ISSN 0268-6902Volume 34 Number 3 2019

ManagerialAuditing Journal

Organizational risk, fraud, forensics, anti moneylaundering laws and controls, and corporate corruption

Guest Editor: Jagdish Pathak

EDITORIAL ADVISORY BOARD

Professor Ali AbdolmohammadiBentley College, USA

Professor Urton AndersonUniversity of Kentucky, USA

Professor Richard A. BernardiRoger Williams University, USA

Professor Niamh BrennanUniversity College Dublin, Ireland

Professor Priscilla A. BurnabyBentley University, USA

Associate Professor Maria Cadiz DyballUniversity of Sydney, Australia

Professor Peter CareyDeakin University, Australia

Professor Garry CarnegieRMIT University, Australia

Dr Moon-Kyung ChoTexas A&M International University, USA

Professor Paul CoramUniversity of Adelaide, Australia

Professor Steven DellaportasRMIT University, Australia

Associate Professor John DumayMacquarie University, Australia

Associate ProfessorWendy GreenThe University of New South Wales, Australia

Associate Professor Ahsan HabibMassey University, New Zealand

Associate Professor Jacqueline S. HammersleyUniversity of Georgia, USA

Assistant Professor Clark HamptonDarla Moore School of Business, USA

Professor Roszaini HaniffaHeriot Watt University, UK

Professor Susan HassSimmons College, USA

Professor David HayThe University of Auckland, New Zealand

Professor Christine HelliarUniversity of South Australia, Australia

Professor Rani HoitashBentley University, USA

Professor Christopher HumphreyUniversity of Manchester, UK

Associate Professor Duane B. KennedyUniversity of Waterloo, Canada

Professor W. Robert KnechelUniversity of Florida, USA

Associate Professor KevinW. KobelskyUniversity of Michigan, Dearborn, USA

Professor Philomena LeungMacquarie University, Australia

Senior Lecturer Youngdeok LimUniversity of New South Wales, Australia

Professor JerryW. LinUniversity of South Florida – St Petersburg(from Spring 2015), USA

Professor Nonna Martinov-BennieMacquarie University, Australia

Professor Reza MonemGriffith University, Griffith Business School, Australia

Professor Gary MonroeThe University of New South Wales, Australia

Professor Robyn MoroneyMonash University, Australia

Professor Karen van PeursemVictoria University of Wellington, New Zealand

Professor Chris PongUniversity of Nottingham Business School, UK

Professor Vaughan RadcliffeUniversity of Western Ontario, Canada

Professor Dr Nicole Ratzinger-SakelUniversität Hamburg, Germany

Professor Alan ReinsteinWayne State University, USA

Professor Glennda ScullyCurtin University of Technology, Australia

Associate ProfessorWilliam ShaferLingnan University, Hong Kong

Professor Divesh SharmaKennesaw State University, USA

Professor Michael ShaubTexas A&M University, USA

Professor Roger SimnettThe University of New South Wales, Australia

Professor Nava SubramaniamRMIT University, Australia

Professor Stuart TurleyUniversity of Manchester, UK

Associate Professor Joost Van BuurenNyenrode Business Universiteit, Netherlands

Professor Karen Van PeursemVictoria University of Wellington, New Zealand

Professor Miklos A. VasarhelyiRutgers Business School, USA

Editorialadvisory board

245

Managerial Auditing JournalVol. 34 No. 3, 2019

p. 245© Emerald Publishing Limited

0268-6902

Trim size: 174mm x 240mm

Guest editorialEditorial note on risk taking, fraudulent reporting, forensic auditing andanti money launderingBusinesses in modern World are fully connected directly or indirectly with the internal andexternal stakeholders. Stockholders wait for increased profits. Financial analysts need moreinformation to disseminate the same to the investors and stock markets. Regulatory and lawenforcement machinery wants compliances. Employees and executives are always in theneed of additional bonus and other incentives. Scenario of modern-day commerce isextremely complex, and the outcome is not dependent on alone financial but non-financialinformation too. Relationships of various functions have turned non-linear with theindependent and the dependent variables. Several independent variables in the objectivefunction’s optimization have become extremely noisy. Decision-making by the CEOs in suchdynamic environment might not be generalizable unlike in the past.

Political environment in which an enterprise grows and sustains too is no longer logical.Comparative cost advantage theory was the benchmark for globalization of economicactivities, but the present politically restrictive trade practices and dosages of nationalismacross the globe have given birth to newer kind of issues and problems in the economicsphere. A new normal is settling in the commerce and industrial undertakings.Maximization of individual advantages has become rampant and where professionalaccountants are facing agency issues. Fraudulent financial reporting by an undertaking hasnot only affected its investors but started reflecting viciously in the society. Moneylaundering, which was assumed before being done only by illicit and illegal activities is nolonger the case. Large-scale money laundering is deeply involving several bankinginstitutions intentionally or unintentionally, despite various efforts made globally.

Illicit and illegally earned money is seeping into the regular banking channels andthrough that in our life and national economies. National economic accounting is at the crossroads and effect of this untaxed money or black money is reflecting in the failure ofcontrolling the inflationary effects on the economy. Corporate governance since SOXimproved to a greater extent in the USA. However, most of the third-World economies havenever really bothered to implement the best practices seriously. Mostly, corporategovernance is an instrument of choice in such economies and remains on paper only.

This special issue has brought some excellent articles on these matters of concern. Thecoverage is global and shows that how various economies in the World are facing suchgrievous issues.

Jagdish PathakDepartment of Accounting, University of Windsor Odette School of Business,

Windsor, Canada

MAJ34,3

246

Managerial Auditing JournalVol. 34 No. 3, 2019p. 246© Emerald PublishingLimited0268-6902DOI 10.1108/MAJ-03-2019-017

The role of shell entities in fraudand other financial crimes

Carl PaciniKate Tiedemann College of Business, University of South Florida-St. Petersburg,

St Petersburg, Florida, USA

William Hopwood and George YoungCollege of Business, Florida Atlantic University, Boca Raton, Florida, USA, and

Joan CrainBNYMellon Wealth Management, Ft. Lauderdale, Florida, USA

AbstractPurpose – The purpose of this paper is to review the use and application of shell entities, as they facilitatecrime and terrorism, impede investigations and harm societies.Design/methodology/approach – The study details the types and characteristics of shell entities,reviews actual cases to exhibit how shells are abused, outlines reasons shells disguise beneficial ownershipand analyzes steps taken by countries and organizations to thwart the abuse of shell entities.Findings – Many types of shell entities are used by white-collar criminals and are often layered in an intricatenetwork which conceals the identity of beneficial owners. Nominees and bearer shares are used in tandem withshell entities to optimize concealment. Accountants, lawyers and trust and company service providers facilitateand promote the use and abuse of shell entities by lawbreakers. The G-8, Financial Action Task Force and G-20have begun steps to improve ownership transparency, but the effort is moving at a modest pace.Research limitations/implications – The analysis makes clear the reasons for and means by whichthe wealthy and powerful, along with criminals, conceal trillions of dollars of income and wealth that remainuntaxed and may be used for nefarious purposes. The paper is limited by the paucity of data on concealedassets and their beneficial owners.Practical implications – The findings clearly show the need for more concerted action by nationalgovernments, organizations, the United Nations and law enforcement and to improve ownership transparencyand information exchange regarding shell entities.Social implications – The findings demonstrate that shell entities used to conceal wealth prevent untoldtrillions in taxes from being collected by governments worldwide. This lack of revenue facilitates incomeinequality and skews national economic and fiscal policies. Also, more white-collar criminals and terroristfinanciers could be brought to justice if ownership transparency is improved.Originality/value – This study adds to the limited literature on shell entities, their characteristics and usesand abuses.

Keywords Tax evasion, Beneficial ownership, Money Laundering, Ownership transparency,Shell entities

Paper type Research paper

“The secret to success is to own nothing, but control everything.” – Nelson Rockefeller

1. Introduction/historical contextThe Financial Action Task Force’s (FATF) 40 Recommendations, published in 2004, addressedthe connection between business secrecy and financial crime [Wolos and Reuters, 2017;

Role of shellentities

247

Received 16 January 2018Revised 24May 2018

Accepted 27 June 2018

Managerial Auditing JournalVol. 34 No. 3, 2019

pp. 247-267© EmeraldPublishingLimited

0268-6902DOI 10.1108/MAJ-01-2018-1768

The current issue and full text archive of this journal is available on Emerald Insight at:www.emeraldinsight.com/0268-6902.htm

Financial Action Task Force (FATF), 2004]. Terrorists, members of criminal organizations, taxevaders, fraudsters, money launderers, drug dealers and others hide illegal assets and activitiesby availing themselves of the secrecy provided by the legal domestic and offshore businessstructures commonly referred to as “shell entities.” In 2011, a World Bank study found thatperpetrators in 70 per cent of 213 largescale corruption cases relied on the secrecy of shellentities to hide their identity (Global Witness, 2013). The legal structures of such entitiestypically include domestic and offshore limited liability companies (LLCs), limited liabilitypartnerships (LLPs), international business companies (IBCs), private foundations, companyfoundations (CFs) and asset protection trusts. Regardless of the entity type, the identities of thebeneficial owners are hidden behind the secrecy of financial “shells” (Baradaran et al., 2014).

The FATF defines a beneficial owner as the:

natural person(s) who ultimately owns or controls a customer and/or the person on whose behalf atransaction is being conducted. This definition is very broad and also incorporates those personswho exercise ultimate effective control over a legal person or arrangement [Financial Action TaskForce (FATF), 2014].

A beneficial owner is always a natural person – a legal person cannot be a beneficial owner.Ultimate control gives the natural person the ability to benefit from the asset in question(Willebois et al., 2011).

The extensive, illegitimate use and abuse of domestic and offshore entities to conceal andtransfer assets (including the proceeds of crime), fund terrorism, evade taxes and facilitateother crimes (such as human slavery, prostitution, racketeering, illegal gambling, securitiesfraud and financial fraud) make such entities an important focus of the work of forensicaccountants, auditors, regulators and law enforcement officials. Hence, the purpose of thisarticle is to review the use and application of shell entities, as they facilitate white-collarcrime and terrorism, impede investigations and ultimately harm society. This paper isintended to educate forensic accountants, auditors, regulators, legislators and others aboutthe structure and workings of various legal entities used as “shells” and provide a policyupdate on what steps have been taken to curb their abuse.

Section 1.1 discusses the importance of secrecy or concealment and then analyzesthe use and abuse of shell entities to achieve such secrecy (often for nefarious or illegalpurposes). Section 2 analyzes the different types of legal or business structures thathave been used as shell entities. The first two sections are intended to provide thereader with the historical context of shell entities. Section 3 discusses suggestedreasons for the vulnerability of various business structures that, when privately owned,are easily manipulated to operate as shell entities. Section 4 highlights policy reactionsand responses, including legislation aimed at dealing with issues relating to beneficialownership problems, curbing the abuse of various shell entities and enhancing theability of forensic accountants and law enforcement officers to combat tax evasion,terror funding, money laundering, hiding stolen assets and defrauding creditors. Thefinal section concludes the paper.

1.1 Concealment and secrecyA correlation exists between corruption and anonymous shell entities. Tracing illicit fundsor assets to a shell entity is not that useful if the individuals who control it (i.e. the beneficialowners) cannot be identified (Anonymous, 2016a). Obscured beneficial ownership via shellentities impedes law enforcement officials and forensic accountants from tracking themovement of money, and investigating and recovering stolen assets (Kalant, 2009). This is

MAJ34,3

248

true for a wide range of investigations, ranging from foreign terrorists, narcotics traffickers,sanctioned regimes, cyber hackers andmoney launderers (Szubin, 2016).

A “shell company” refers to an LLC or other business entity with no significant assets orongoing business activities that are capable of moving assets and large sums of moneyglobally. Shell companies typically have no physical presence other than a mailing address,have no employees and produce little or no independent economic value [Financial CrimesEnforcement Network (FinCEN), 2006]. It is not uncommon to find hundreds, if notthousands, of shell entities registered to the same address because most shell companieshave no operations (Anonymous, 2013; Hubbs, 2014)[1]. In Delaware, 285,000 companies areregistered at just one building. This figure is ten times higher than the total number ofcompanies registered in the Isle of Man, an offshore bank secrecy haven (Anonymous,2016a). Shell entities are sometimes formed with a generic stated purpose, such as “toconduct legitimate transactions, such as domestic and cross-border currency and assettransfers, or to facilitate corporate mergers and reorganizations” (Hubbs, 2014).

Shell companies are not always formed for illegal purposes, and they are even a valuablesource of tax revenues in some countries. For example, shell entities in The Netherlands areinvolved “in about $1 trillion in transactions each year with the taxes paid on thesetransactions serving as an important government revenue source” (Couch, 2004). Shellentities can be publicly traded or privately owned.

Privately owned shell entities tend to be more susceptible to illicit uses because limitedownership limits public exposure and eases the cloaking of beneficial ownership. For thisreason, privately owned shell entities have become the financial and deception vehicle ofchoice for tax evaders, money launderers, corrupt politicians, fraudsters, dishonest businesspeople, celebrities, terrorists, drug dealers, organized crime members and cybercriminals(Hubbs, 2014).

A massive leak of documents from the Panamanian law firm Mossack Fonseca, knownas the “Panama Papers,” has illuminated the extent of the “vast” murky world of shellentities, providing an extraordinary look at how the wealthy and powerful conceal theirmoney (Hall and Taylor, 2016). Data analytics firms such as Linkurious extracted metadatafrom the 2.6 terabytes of leaked data and helped connect the dots through data visualizationtools (Santiso and Roseth, 2017)[2]. Individuals:

[. . .] exposed in the leak include the prime ministers of Iceland and Pakistan [. . .] and companieslinked to the family of Chinese President Xi Jinping. Add to those [. . .] honchos in [. . .] FIFA that[control] international soccer and 29 billionaires featured in Forbes Magazine’s list of the world’s500 richest people. [. . .] The documents within the leak [. . .] expose how secretive offshorecompanies at times subvert U.S. policy and mock U.S. regulators. [. . .] Plenty of criminals are[listed] in the documents, from drug traffickers to [. . .] fraudsters (Hall and Taylor, 2016; Ryle,2013).

As a result of the Panama Papers leak, the Canada Revenue Agency (CRA) executed threesearch warrants on February 14, 2018, during an offshore tax evasion criminalinvestigation. The CRA’s investigation identified a series of transactions involving foreigncorporations and several transfers through offshore bank accounts allegedly used to evadetaxes (Canada Revenue Agency, 2018).

Another, more recent, leak is known as the “Paradise Papers.” The Paradise Paperscontain 13.4 million leaked documents, mostly from the Bermuda-based law firm Applebyand Singapore-based Asiaciti Trust (Murphy, 2017). The documents detail the wayspoliticians, celebrities, and the ultra-rich protect their cash from taxation, hide theirownership of major assets and secretly conduct their business. Most of the exposed practicesare legal, but some may be unethical (Murphy, 2017). For example, Stephen Bronfman, a top

Role of shellentities

249

aide and key fundraiser for Canadian Prime Minister Justin Trudeau, was found to haveavoided millions in taxes through offshore accounts. Although this conduct was not illegal,it was very embarrassing for Trudeau, who had pledged to crack down on tax havens(Murphy, 2017). Another embarrassing disclosure revealed that US Secretary of CommerceWilbur Ross had business ties with Russian President Vladimir Putin’s son-in-law, aRussian oligarch under US sanctions. Ross failed to disclose the ties during his confirmationhearings (Murphy, 2017).

By necessity, criminals and terrorists are attracted to jurisdictions with lax financialsecrecy laws and practices. Such jurisdictions facilitate secrecy and thereby providerelatively weak barriers to the abuse of domestic and offshore shell entities, trusts,foundations, “shelf” corporations, IBCs, LLCs and other business structures. In 2009, theTax Justice Network (TJN) launched an online database that shows how the legal, judicialand regulatory details of different jurisdictions contribute to the environment of financialsecrecy (Christensen, 2012). The Financial Secrecy Index (FSI) is a global ranking that drawsattention to the various aspects of financial secrecy. According to the TJN, an estimated$21tn to $32tn of private financial wealth is located in secrecy jurisdictions around the world[Tax Justice Network (TJN), 2018]. Secrecy jurisdictions use concealment and anti-disclosurelaws to attract illicit and illegitimate financial flows [Tax Justice Network (TJN), 2018]. Illicitcross-border financial flows have been estimated at $1-$1.6tn per year [Tax Justice Network(TJN), 2018]. Offshore companies in the British Virgin Islands (BVI) alone had assets inexcess of $1.5tn in early 2017 (Houlder, 2017).

The FSI reveals that the traditional stereotypes of financial secrecy and tax havens areinaccurate. The world’s most significant providers of financial secrecy, the places thatharbor the most concealed assets, are not small, palm-fringed islands but rather the USA,Switzerland, Hong Kong, Singapore, the UK and its overseas territories (e.g. the CaymanIslands), Germany and Luxembourg [Browning et al., 2018; Tax Justice Network (TJN),2018]. In 2017, the USA ranked second in terms of providing financial secrecy [Browning etal., 2018; Tax Justice Network (TJN), 2018].

Not only does the US system of shell entities and liberal secrecy restrictions promote taxevasion, money laundering and other crimes it also entices the most dangerous criminals inthe world to establish shell corporations in the USA (Szubin, 2016). Viktor Bout, a Russianoligarch, referred to as the “Merchant of Death” by a US Senate committee and considered tobe the world’s largest arms trafficker, used 12 US shell corporations formed in Delaware,TX, and Florida to finance his illegal activities (Browning, 2009). Bout was able to carry onhis business until he was arrested in Bangkok in 2008 (Stempel, 2013). Jose Trevino Morales,the brother of two Los Zetas drug cartel leaders, one of whom dismembered his victimsalive, transferred tens of millions of dollars of illicit drug money under the guise of a horseranch business in Oklahoma. The cartel leaders’ names were nowhere to be found because ofthe shell entities they used, which were completely secretive (Hicken and Ellis, 2015).

2. Shell entities and other legal structures and devices used to achieve secrecyLegitimate business people and lawbreakers alike may form various types of domestic andforeign entities, such as domestic LLCs, foreign LLCs, asset protection trusts, LLPs, IBCs,private interest foundations (PIF) and CFs. Dishonest parties can use these entities to holdlegal title to stolen or hidden assets, cloak beneficial ownership, engage in illegaltransactions such as drug dealing, fund terrorism, evade taxation and commit various othercrimes. After the entities have been formed, the asset hiders open and fund bank accounts inthe names of these entities.

MAJ34,3

250

Criminals can use shell entities to commit a panoply of crimes, such as tax evasion,money laundering, securities fraud, financial fraud, corruption and bribery. One example ofthe abuse of a shell company can be found in U.S. v. Lake, 571 Appx. 303 (5th Cir. 2014).Larry Lake operated VIP Finance, a car title loan business with six locations in the Dallasarea. Lake also owned Cash Auto Sales, which handled “auto club” memberships for VIPFinance, as well as a drugstore called Grapevine Drug Mart. The day before he filed forpersonal bankruptcy, Lake transferred $2,763,000 from an E*Trade account in his name toan E*Trade account held jointly with his wife. Also on that day, he purchased a cashier’scheck in the amount of $348,000 payable to Air I.Q., a shell company formed in his wife’sname. Because Lake did not disclose these transactions during the bankruptcy case, he wascharged with bankruptcy fraud.

Another instance of shell companies facilitating white-collar crime occurred in 2014.Hewlett-Packard A.O. (HP Russia) agreed to pay more than $100m in penalties for felonyviolations of the Foreign Corrupt Practices Act (FCPA) [Foreign Corrupt Practices Act(FCPA), 2017; Athanas, 2010; Deming, 2006][3]. According to the plea agreement, HP Russiaexecutives created a multimillion dollar secret slush fund that was used to bribe RussianGovernment officials [U.S. Department of Justice (DOJ), 2014]. To effect the payment of thesebribes, money was laundered using an intricate web of shell companies and bank accounts.The conspirators kept two sets of books to conceal and track the corrupt payments. Theslush fund proceeds were spent on such items as travel, luxury automobiles, jewelry,clothing and furniture.

These abuses are compounded by the fact that many states, provinces and countriespermit shell entities to own and manage other shell entities [Financial Crimes EnforcementNetwork (FinCEN), 2006]. The result can be multiple layers of cloaked ownership that makeit virtually impossible for forensic accountants or law enforcement officials to determine theidentity of the beneficial owners.

2.1 Limited liability companies as shell entitiesDomestic and offshore LLCs can be used to defeat, or at least delay, the claims of creditorsand other claimants. Because LLCs may be owned and managed anonymously, they can besubject to abuse. They are creatures of statute, and transparency-of-ownership requirementsvary from state to state and country to country [Government Accountability Office (GAO),2006].

LLCs are a hybrid of corporations and partnerships. LLCs provide members (rather thanshareholders) the same limited liability afforded to corporate shareholders while providingthe “pass through” taxation benefits (in most jurisdictions) to members. When used as shellentities, LLCs exist only to own other entities, to hold bank accounts, or as simply a transferpoint for moving funds from one account or business to another.

An LLC shell entity’s ownership can be structured in a variety of ways, including havingshares issued to a natural or legal person or in registered or bearer form (Biedermann, 2015).Bearer shares confer rights of ownership to the physical holder or possessor of the share(s).They are commonly and legitimately used in a number of countries, but they are notpermitted in the USA. However, given that bearer shares are not registered with regard toownership, they enable illegal activities to be hidden and can be controlled by beneficialowners who cannot be identified.

LLC shells are easy to form (in as little as two hours for about $100-$200 in some states)and can be linked or layered across different jurisdictions, creating a confusing path forforensic accountants, auditors and investigators (Martinez, 2017). If they are established in ajurisdiction that has no regard for ownership transparency (e.g. Wyoming, NV, DE),

Role of shellentities

251

identifying the beneficial owners may be virtually impossible (Adkisson and Riser, 2004).The USA is one of the world’s foremost jurisdictions for those seeking to avoid ownershiptransparency. In fact, US LLC shells are used more often in laundering the proceeds of grandcorruption than the shells of any other country (Sharman, 2013).

A fine example of the abuse of LLC shells occurred in U.S. v. Beecroft, 825 F. 3d 991 (9thCir. 2016). From 2003 to 2008, Melissa Beecroft participated in a multimillion dollarresidential mortgage fraud scheme in the Las Vegas area. Beecroft and her co-conspiratorsrecruited and paid straw purchasers (buyers who purchase homes on behalf of undisclosedpersons, with no intent to retain the property themselves) to acquire homes at substantiallyinflated prices. After each sale, straw buyers would transfer ownership in the properties tothe LLC shell entities. Buyers defaulted on the mortgages and lenders were unable to locatethe responsible parties. Altogether the scheme involved more than 400 straw buyertransactions and 227 properties bought for more than $100m.

2.2 Shelf corporationsExisting, but unused, shell companies may be converted to current, possibly illegal, use.Such companies are known as “shelf” or aged corporations. The established age of thesecompanies adds to their credibility. Shelf companies are attractive because any legal filingrequirements have already been satisfied, so they are instantly available, no shares have yetbeen offered, and ownership can be immediate. In general, domestic and offshore shelfcorporations possess all the necessary corporate prerequisites in the appropriate jurisdiction(e.g. Wyoming, St. Kitts and Nevis, Anguilla, Belize) for legal operation and quick transfer ofownership (Weiss, 2011). Shelf corporations may be purchased on the internet for a fewthousand dollars from trust company and service providers (TCSPs) such as www.offshorecompany.com and www.companiesinc.com. In some cases, shelf entities can even bebought with ready-made, established bank accounts.

Once a shelf company is bought, the buyer may acquire the shelf corporation’sestablished credit and tax history, which further enhances its credibility (Willebois et al.,2011). The lack of accurate, recorded information about shelf entities can create almostinsurmountable obstacles for auditors, forensic accountants, regulators and lawenforcement officials in any attempt to identify the beneficial owner(s).

2.3 Nominees or nominee directors in shell entitiesAnother legal device or approach to optimize concealment involves the shell entity’sbeneficial owner or owners electing to hire a nominee as a company director. A nominee isone who holds bare legal title for another, who is designated to act in place of another in alimited way, or who receives and distributes funds for the benefit of others [LiButti v. U.S.,107 F. 3d 110 (2nd Cir. 1997); Martinez, 2017]. A nominee can be a relative, friend, trustedassociate or a person who has no link to the true beneficial owner(s). Nominee incorporationservice providers (e.g. www.offshoresimple.com/nominee.htm) provide local and third-partynominees to serve as the director and/or manager of the shell company. The nomineetypically signs a general power of attorney giving the beneficial owner(s) full power tomanage the shell entity. The nominee also provides a signed and undated letter ofresignation to further protect the anonymity and interest of the beneficial owner(s).

Globally, a small number of nominee directors are used multiple times in many shellentities. Specifically, a mere 28 nominee directors have established, or are in control of, morethan 21,000 companies (Ball, 2012). Many of these individuals have been identified asinvolved with criminal organizations and individuals (Ball, 2012). They market theirservices by selling their names and addresses in obscure global locations. The shell entities

MAJ34,3

252

themselves are often registered anonymously in Switzerland, the USA, Hong Kong,Singapore, Luxembourg, Germany, Taiwan, Guernsey, Panama, St. Kitts-Nevis, Lebanon,The Netherlands, Bahrain, the United Arab Emirates, Jersey, the Bahamas, Malta, Cyprus,the British Virgin Islands, Ireland, China, Japan and Canada [Ball, 2012; Tax JusticeNetwork (TJN), 2018]. In early 2014, the International Consortium of InvestigativeJournalists (ICIJ) (representing more than 190 journalists from 65 countries) published adatabase of the incorporation records of shell companies, directors (some nominees) andaddresses that were leaked to them. The database showed the extent of shell companynetworks and revealed how many companies and nominee directors are linked (See http://tinyurl.com/jwxbg2z and https://tinyurl.com/mu5483k).

2.4 TrustsTrusts are another vehicle subject to abuse by fraudsters, white-collar criminals and assethiders. The salient characteristic of a trust is that it provides for a separation of legal andbeneficial ownership (Danforth, 2002). Legal control is granted to a trustee by a settlor (a/k/acreator or grantor). The trustee manages the trust asset(s) according to the terms of a trustagreement for the benefit of the beneficiaries (Danforth, 2002). A settlor (or creator orgrantor) who establishes the trust can minimize the likelihood of transfer or attachment of abeneficiary’s trust interest by creating a spendthrift trust or by including a spendthriftprovision in a trust (Miller v. Kresser, 34 So. 3d 172 (Fla. Dist. Ct. Appl. 2010) (quotingCroom v. Ocala Plumbing & Elec. Co., 57 So. 243 [1911]). Spendthrift trusts are subject tovarious laws, regulations and exceptions that vary from one jurisdiction to the other. MostUS states follow the principle that it is impermissible to have a spendthrift trust when thesettlor is also a trust beneficiary (In re Brown, 2002). In some offshore jurisdictions, however,spendthrift trusts are permitted even when the settlor is a beneficiary (Ausness, 2007).

Self-settled spendthrift trusts, which have been given the misnomer of “asset protectiontrusts”, are a “booming business for banks, trust companies, and estate planners, both [inthe USA] and abroad. They [are] a multi-billion-dollar-a-year business” (Morse, 2008). ManyUS and offshore promoters attract US, Canadian and other citizens with promises of taxavoidance (not evasion) and asset protection through the use of trusts (Morse, 2008). This isquite similar to the use of LLC shell companies.

The popularity of asset protection trusts is based, in part, on the fact that trusts canprovide beneficiaries with more privacy and autonomy than traditional business entities.Trusts have historically had no registration requirements or central registries where thenames of trustee, settlor and beneficiary must be listed (Fidelity Investor, 2017). Even wherebeneficiaries’ identities must be disclosed, the beneficiary can be a limited partnership (LP),LLC or another trust, thus adding layers of opacity to the trust’s ownership structure(Simser, 2008).

“[E]stimates indicate that between $1 and $5 trillion in assets are located in [offshoreasset protection trusts, or OAPTs]” (Maxwell, 2014). Various jurisdictions that permitOAPTs include Anguilla, the Bahamas, Barbados, Belize, the British Virgin Islands, theCayman Islands, the Cook Islands, Cyprus, Gibraltar, the Isle of Man, Saint Kitts and Nevisand the Turks and Caicos Islands (Maxwell, 2014).

OAPTs possess a number of features that permit the settlor to exercise protective controlover trust assets. Protective features of an OAPT may include a trust protector clause, ananti-duress clause, a flee or flight clause and a non-binding letter of intent or wishes(Ausness, 2007). A trust protector clause provides for a “trust protector” being appointed bythe grantor to act as an advisor and who is responsible for making sure the trusteeimplements the settlor’s wishes. An anti-duress clause prohibits the trustee from complying

Role of shellentities

253

with any order imposed upon the settlor or trustee (Lorenzetti, 1997). A flee or flight clauseauthorizes or requires the trustee to transfer the trust to another jurisdiction upon theoccurrence of certain events, such as an inquiry from a foreign government or Interpol(Taylor, 1998). A letter of intent or wishes is written by the settlor and states his or herwishes as to the disposition of trust assets (Ausness, 2007).

Money launderers and other criminals are able to avail themselves of the layering andmisdirection features of OAPTs. One example of these features can be found in U.S. v.Brennan, 395 F. 3d 59 (2nd Cir. 2005). In this case, the Second Circuit upheld RobertBrennan’s conviction for bankruptcy fraud, based in part on money laundering usingOAPTs. Brennan owned and operated First Jersey Securities, Inc. (FSJ), a brokerage tradingin penny stocks. Brennan and FSJ were found guilty of securities fraud and were ordered topay $75m to 500,000 customers. Subsequently, FSJ and Brennan filed for bankruptcy. Nearthe end of his trial for securities fraud, Brennan created an OAPT known as the CardinalTrust and funded it by use of $4m in bearer bonds. Brennan did not disclose Cardinal’sassets in the bankruptcy action, allowing Cardinal Trust’s assets to grow to $22m by mid-1997. Brennan used $12m in trust assets to buy and refurbish the Palm Beach Princess – agambling boat. The Cardinal Trust’s venue or situs was moved twice (via a flight or fleeclause) to avoid detection. Brennan was convicted of money laundering and other offensesand sentenced to over nine years in prison.

OAPTs can be used by those bent on committing financial crimes in four ways:(1) by hiding legitimate assets for the purpose of evading taxes (Silets and Drew,

2001);(2) by integrating illicitly obtained funds into an economy as “clean assets” (i.e. money

laundering) (Silets and Drew, 2001);(3) by moving legitimately obtained funds into an economy to be used for nefarious

purposes (i.e. reverse money laundering) (Arce, 2009); and(4) by hiding legitimate assets from creditors with bona fide claims or spouses in

divorce proceedings (Silets and Drew, 2001).

The linchpins to illegitimate uses or abuses of OAPTs are layering andmisdirection.The most clever asset-hiding and asset-protection schemes insulate the identity of the

wrongdoer through many layers of trusts and other shell entities. They also incorporatemisdirection by creating the appearance that the wrongdoer has no control of the OAPT,and that their sole administrator is in an offshore jurisdiction. One example of how a taxevader layered OAPTs is found in U.S. v. Scott, 37 F. 3d 1564 (10th Cir. 1994). Anorganization named International Business Associates devised a scheme involvingtransfers to and among four successive trusts. Trust I was a domestic trust established asa shell with an apparently fictitious contribution of $100 by some entity other than theclient. Trust I was required to distribute all taxable income to Trust II (a Belizean trust).Trust II was a conduit trust that passed its income to Trust III, a foreign trust that coulddistribute and accumulate income. Trust IV was a passive foreign trust until thepurchaser settlor of the trust scheme needed funds. Like most OAPT tax evasionschemes, power rested with the purchaser-client while the true beneficial ownersremained unnamed in all documentation.

Just as with other types of shell entities, the goal is to transfer assets through enoughlayers of OAPTs (and other entities) so that a banker, lawyer, forensic accountant,auditor, bankruptcy trustee or law enforcement officer will not suspect or discover the

MAJ34,3

254

sources of assets (Zagaris, 1999). By accomplishing this goal, individuals and entities cancontrol their assets without being named as a beneficiary or trustee.

2.5 Limited partnerships and family limited partnershipsLPs and family limited partnerships (FLPs) are superb places to secrete assets [Tax JusticeNetwork (TJN), 2018]. In a typical scheme involving an LP, the fraudster (general partner)provides assets to trusted associates, friends or family members to invest in an LP. These“investors” become limited partners who have no personal legal liability for the debts ofthe business and cannot take an active role in operating the business. In another type of schemeinvolving an LP, the fraudster or criminal conveys assets to an LP of which he is the sole limitedpartner and then transfers the partnership interest to a trust of which he is the sole trustee andbeneficiary (usually done on an offshore basis).

In an attempt to hide assets or obscure the beneficial owner a married couple mightarrange to contribute all of their assets to the FLP [Association of Certified Fraud Examiners(ACFE), 2009]. In this structure, each spouse retains a 1 per cent general partnership interestand a 49 per cent LP interest. General partners in an FLP have unlimited personal liabilityfor the FLP’s debts and obligations. Using this arrangement, subjects only 2 per cent of thecouple’s assets to unlimited liability. If an FLP is structured correctly, creditors cannotdirectly reach the FLP assets, but are instead restricted to obtaining a charging order, whichgives the creditor the right to receive any distributions made to the partner [Association ofCertified Fraud Examiners (ACFE), 2009]. But creditors cannot force the FLP to makedistributions to the partner. In the USA, however, there have been a few cases in which acourt has ruled that a partner’s interest in an FLP can be foreclosed upon by a creditor[Firmani v. Firmani, 752 A. 2d 854 (N.J. Super. Ct. App. Div. 2000].

2.6 International business companiesAn IBC is an offshore corporation closely related to a traditional corporation because it usesarticles of incorporation or association and requires company directors. An IBC is asubcategory of LLC targeted at non-residents of the jurisdiction in which it is sited. It maynot engage in economic activities within its situs jurisdiction. Historically, in most offshorejurisdictions, an IBC is governed by strict confidentiality regulations, as the names of itsshareholders and directors need not be published in any public register. Also, althoughshareholders of many IBCs are required to elect directors, once elected the board may runthe IBC with little recourse to shareholders. In some jurisdictions, the abolition of sharecapital allows the IBC to ignore capital retention in making distributions to shareholders(Offshore, 2005). The lack of restriction when making distributions to shareholders makesthe IBC a convenient vehicle for money laundering and moving money to many differentlocations, thus obscuring the money trail for auditors, forensic accountants andinvestigators.

IBCs are also used as tools by corporations and individuals throughout the world todirect profits away from high tax countries into offshore jurisdictions that have low or zerotax rates and tax treaties with other nations (double tax treaties). For example, more than140 listed businesses in London, New York and Hong Kong have a unit in the BVI, which isuseful as a tax neutral hub (Houlder, 2017).

Businesses which are suitable to be conducted through an IBC include internationaltrading, international construction and engineering, royalty firms, real estate companies,shipping and ship management and holding companies.

Role of shellentities

255

2.7 Private interest foundations and foundation companiesThe private interest foundation (PIF) is a vehicle provided by civil law countries (and nowbeing adopted by some common law jurisdictions) for asset protection and estate planningpurposes, but which serves as an alternative to trusts (Berbey de Rojas, 2008). The PIF wasfirst introduced by Liechtenstein in 1926 but has attracted followers such as Panama,Aruba, Bonaire and Curacao; the Bahamas; Costa Rica; the Seychelles Islands; St. Kitts andNevis; Anguilla; Antigua; and the Cook Islands (Wiggin, 2008). Like an asset protectiontrust, a PIF may be used to conceal the identity of individuals, entities and assets. Althoughthere is no single definition of a foundation, a number of common features can be identifiedin jurisdictions that offer PIFs.

Panama is a good example because it has permitted PIFs since 1995, has more than400,000 registered offshore corporations and PIFs, does not require foundations to keepfinancial records or submit tax returns and offers much secrecy (Thompson, 2010). A PIFhas four main parts:

(1) Founder – the person or entity that forms the foundation in the public registry.Usually a nominee founder is provided by a professional service firm along with apresigned, undated letter of resignation. At that point, the nominee founder holdsno further control. A founder is analogous to a trustee.

(2) Foundation Council – this serves the same function for a PIF as a board ofdirectors does for a corporation. The council members’ names and passportnumbers are noted in the public registry when the foundation is established. Oftena nominee council is provided by a professional service firm along with presigned,undated letters of resignation from the nominee council.

(3) Protector – this is the ultimate controller of the foundation. Immediately uponestablishment, the council appoints a protector through a notarized privateprotectorate document. Because the document is a private, non-publicly registereddocument, the protector remains anonymous. At that point, the protector has fullcontrol over the foundation (which holds legal title to any assets) and its assets.

(4) Beneficiaries – a PIF does not have owners but beneficiaries. The latter areappointed by the protector either through a private letter of wishes or through aformal set of bylaws. The contents of both remain confidential. Privately appointedbeneficiaries remain anonymous (Boschini, 2006; Berbey de Rojas, 2008; Elements,2017; Aspen Group Limited, 2012).

In sum, no legal requirement exists to disclose the name of a founder, beneficiary orprotector. Annual tax returns or financial statements do not need to be filed. A foundationmay engage in any business or civil transaction in any part of the world and in anycurrency. Moreover, the foundation charter may be signed by an attorney without disclosingthe name of the founder.

In some foundation jurisdictions (e.g. Anguilla), any assets available for distribution to abeneficiary are neither capable of being alienated or passed by bankruptcy, insolvency orliquidation nor liable to be seized, sold, attached or otherwise taken in execution, by processof law (Wiggin, 2008). The secrecy and lack of transparency and flexibility of PIFs have ledto their spread throughout the offshore world (China Offshore, 2009). Such foundationsrepresent another vehicle that can be used by money launderers, tax evaders, terroristfunders, organized crimemembers andwhite-collar criminals.

In the last quarter of 2017, the Cayman Islands implemented a new legal structure(which could be used as a shell entity) known as a CF (Davern and Way, 2017). A CF

MAJ34,3

256

shares some features with a trust, but it may be established so that beneficiaries aregiven no rights to make a claim against the company foundation (CF) (Davern and Way,2017). The CF law itself describes possible objects of a CF as acting as a holdingcompany or an investment company. Another feature of a CF (which trusts do notpossess) is that any kind of power can be given to any person, whether as a personalpower, as a benefit for the CF or for any other lawful purpose (Davern and Way, 2017).These objects and powers make the CF vulnerable to abuse by white-collar criminalssuch as tax evaders, money launderers and terrorist financiers. Only the passage oftime will indicate whether the CF will be added to the list of legal structures thatcommonly serve as shell entities.

3. Three main reasons shell entities provide secrecy and disguise beneficialownershipThree main reasons explain the continued ability of white-collar criminals, terrorists andorganized crimemembers to use shell entities to conceal the identity of their actual beneficialowners and to operate in the shadows. One reason is the lack of transparency in mostjurisdictions (including US states) with regard to actual or beneficial owners, directors,corporate officers, members, partners, trustees, beneficiaries and others. The second reasonis that drug traffickers, money launderers, tax evaders, terrorist funders,and other white-collar criminals are able to engage professionals such as accountants, lawyers, financialadvisors and TCSPs or “gatekeepers” to create shell entities, layer them together intocomplicated webs, hide assets and launder funds. The third reason is that the layering orpyramiding of different shell entities (often different legal types) in various jurisdictionsaround the globe makes an impenetrable trail for law enforcement officers and forensicaccountants to follow.

3.1 Lack of transparency in identifying beneficial ownersOwnership transparency refers to the disclosure of majority and minority shareholders,members, beneficiaries, protectors, trustees, founders and directors, depending upon thetype of legal entity or any other natural person who is in a position to control and/or benefitfrom an asset. Transparency also includes knowledge of the controlling structure of otherlegal entities. Knowledge of beneficial owners and the control structures of entities mustthen be accompanied by effective investigation and enforcement mechanisms regarding thedisclosed information (Lakhani, 2016).

The issue of transparency is captured by FATF Recommendation 24, which states thatcountries should ensure that there is adequate, accurate and timely information available onthe beneficial ownership of all legal persons [Financial Action Task Force (FATF), 2014].The identity of the natural persons who ultimately have a controlling ownership interest in alegal person and/or the identity of the natural persons exercising control of the legal personthrough other means is the goal of transparency requirements [Financial Action Task Force(FATF), 2014].

In practical terms, ownership transparency can be achieved by the use of a centralregistry that collects, stores and verifies the detailed information necessary to determineactual beneficial ownership of any and all types of entities, including trusts andfoundations. Relevant information captured in a central registry would include name;legal entity type; formation documents; related bylaws; address of a registered office orprincipal place of business or address of the entity itself; name and address of a registeredagent; names and addresses of persons in position of legal control within the entity (e.g.

Role of shellentities

257

directors, officers and council members); and the name(s) of the beneficial owner(s)(Willebois et al., 2011).

One huge obstacle to the achievement of practical or actual transparency is thattransparency of ownership requirements vastly differs from jurisdiction to jurisdiction.Transparency practices vary even on an international basis (Anonymous, 2016b). In theUSA alone, where entity formation legal requirements are controlled by the states, vastdifferences make for more favorable entity formation and maintenance in some statesthan others. For example, in 2006, a GAO study found that no state collected beneficialownership information on corporations, only a few collected it on LLCs and othercorporate-like entities, and only four states collected minimal information on LLCs. Lessthan half of the states collected information about management, directors and officers ofcorporations. Although most states collected information on corporate officers and LLCmanagers in periodic reports, the information in these reports was not verified, includinginformation pertaining to beneficial ownership [Government Accountability Office(GAO), 2006]. This is still the situation today (Martinez, 2017). Finally, the states do notscreen information against criminal watch lists. The FBI has open investigations thathave not been resolved because beneficial owners are virtually untraceable (Martinez,2017).

3.2 Gatekeepers and trust company service providersThe services of gatekeepers are often essential for fraud and other illegal schemes tosucceed. Gatekeepers sometimes facilitate the commission of a predicate offense, such asdisguising a person’s involvement in a commercial transaction, commingling property andproceeds or disguising property ownership/control by the ultimate beneficial owners (Bakerand Shorrock, 2009). Gatekeepers’ services help sever the connection between the illegalschemer and the safe enjoyment of the assets. They can also provide criminals with a veneerof respectability (Baker and Shorrock, 2009).

Gatekeepers include TCSPs, lawyers, accountants or other business professionals thatcreate and provide administrative services for various types of entities, such ascorporations, IBCs, LLCs, foundations and trusts. In some jurisdictions, gatekeepers are theonly means of establishing certain kinds of vehicles, such as an IBC. In code or civil lawcountries, certain entities, such as foundations and IBCs, require a notarial deed for creation,meaning that notaries must be employed (i.e. a TCSPmust be hired).

Indispensable administrative procedures performed by TCSPs include checking forthe availability of an entity name, filing appropriate documents with the authorities,opening bank accounts, providing nominees (when necessary), acting as registeredagents, paying fees, handling annual reporting obligations, mail forwarding andproviding virtual office facilities. Many gatekeepers furnish clients with entities from awide range of different jurisdictions. Large TCSPs may form an entity for a client in onejurisdiction (e.g. Belize) but retain client data on file in a different jurisdiction. This makesit more difficult for regulators and forensic accountants to access the information(Willebois et al., 2011).

Money laundering experts at FATF have concluded that gatekeepers should beregulated because they can form a vital link in the chain of performing due diligence (i.e. finding out who they are dealing with and filing suspicious activity reports [SARs]), ifit becomes necessary (Global Witness, 2012). TCSPs typically possess varying degreesof awareness of or involvement in the illicit purposes underlying their clients’ activities[Financial Action Task Force (FATF), 2006]. TCSPs or formation agents have littleincentive to push clients for accurate information. TCSPs already up and running tend

MAJ34,3

258

to be archival; they do not verify incoming data because of the cost (Anonymous,2016b).

TCSPs in many offshore jurisdictions have become subject to formal licensing andregulation, including being audited, meeting anti-money laundering requirements andapplying suitability tests to directors. TCSPs in onshore jurisdictions, particularly USstates, are more loosely regulated (Willebois et al., 2011). This factor contributes to thelarge number of foreign persons creating LLCs and other entities in the USA. In sum,TCSPs from alleged “tax havens” were found in one survey to have higher standards incorporate transparency, at least in the entity formation stage, than those in othercountries (Willebois et al., 2011). By far, the worst performer in the survey was the USA(Willebois et al., 2011). Some TCSPs in Nevada and Wyoming actually offered to use theiremployees’ social security numbers to spare clients the need to obtain an EmployerIdentification Number (Willebois et al., 2011)[4]. Moreover, several shell entitiesoriginating in Cheyenne, WY (where Wyoming Corporate Services is a shell/shelfincorporator) have been used in international frauds and criminal activity (Carr andGrow, 2011). These facts are disturbing given the huge number of legal entities formed inthe USA every year-around ten times more than in all 41 tax havens combined (Willeboiset al., 2011).

In those cases, where an attorney is the TCSP or works for the TCSP, the attorney-clientprivilege may erect another barrier to gleaning information by forensic accountants andinvestigators. The extent to which this barrier exists varies depending on laws of therespective jurisdiction.

3.3 Layering and chaining of shell entitiesPerpetrators of tax evasion, securities fraud and other illicit activities often use a layer orchain of entities established in different jurisdictions to maximize anonymity. This makes italmost impossible for forensic accountants and investigators to determine beneficialownership. In a layered or tiered vehicle structure, various legal entities are insertedbetween the individual beneficial owner(s) and the assets or funds of the shell entity thatholds legal title to those assets or funds. The layering or chaining of various legal entitiesacross numerous jurisdictions (e.g. Jersey, Gibraltar, the USA and the BVI) facilitates accessto the international financial or banking system. Investigators and forensic accountantsmay, for example, obtain ownership information on an entity in Country A and discover thatthe legal owners of that entity are corporations or trusts registered in Countries B and C.Offshore countries and entities by no means possess a monopoly on this type ofarrangement. Legal entities in the USA and UK are also used frequently in layeringarrangements. The ability to layer or chain within and across jurisdictions faces few, if any,restrictions (Willebois et al., 2011).

An example of the use of layered or chained domestic entities for tax purposes occurredin Nevada Partners, LLC v. U.S., 720 F. 3d 594 (5th Cir. 2013). In this case, James KelleyWilliams, a successful Mississippi businessperson, expected to realize an $18m capital gainin tax year 2001 from the cancellation of a loan he had guaranteed. Williams entered into along-term investment program offered by Bricolage Capital, LLC. Bricolage enlisted CreditSuisse First Boston as the bank integral to Williams’ program and used LLCs orpartnerships to execute foreign exchange transactions and other transactions that generatedtax losses.

The three domestic LLCs were Nevada Partners, Carson Partners and Reno Partners.Williams made his required investments in the LLCs using the JKW 1991 RevocableTrust, which held most of his wealth. Numerous purchases of LLCs and other interests

Role of shellentities

259

occurred to transfer the necessary tax losses to Williams, which he claimed on his 2001tax return.

The Fifth Circuit affirmed the federal district court’s decision that:� the transactions lacked economic substance and must be disregarded for tax

purposes; and� the negligence penalty applied, and the three domestic entities were not entitled to

the reasonable cause defense.

This case represents one illustration of the abuse of domestic chained entities and theimportance of TCSPs in structuring shell entities and their illicit transactions.

Criminals and others trying to cloak their identity using shell entities create complexlayered entity networks which result in a labyrinth for forensic accountants andinvestigators. TCSPs involved in providing these professional services are often of little helpin investigations, as they do not deal with the beneficial owners personally (Martinez, 2017;Komisar, 2011). The TCSPs and other gatekeepers are often left untouched by authoritieseven when a fraudster is caught and prosecuted in a shell entity scheme (Hicken and Ellis,2015). Gatekeepers and TCSPs will continue to capitalize on the needs of white-collarcriminals, drug traffickers and others, making for a vicious cycle.

4. Policy reactions and responses to the beneficial ownership issue4.1 US domesticForensic accountants and law enforcement officials have a very difficult time untangling theintricate shell entity networks created by money launderers, tax evaders and othercriminals. Aware of this issue, the US Government has attempted policy initiatives toimprove shell entity ownership transparency.

The US Government is slowly moving forward on the beneficial ownership issue. In June2016, FinCEN finalized its long outstanding beneficial ownership rule, which extendscustomer due diligence requirements to the natural person behind a legal entity (Wolos andReuters, 2017). In June 2017, a bipartisan group of legislators introduced the CorporateTransparency Act, which would require FinCEN to collect information on the beneficialowner(s) of entities created in the USA if it has not been collected at the state level (Wolosand Reuters, 2017).

The Foreign Account Tax Compliance Act, which became law in 2014, has assisted inthe fight against tax evasion (FATCA) [26 USC sections 1471 et seq. (2017)]. FATCAlooks at all types of entities to identify a US taxpayer’s “financial assets”, heldin “financial accounts” outside the USA (Levine et al., 2016). The terms “financial asset”and “financial account” do not refer simply to bankable assets or accounts with regulatedfinancial institutions. They are broadly defined to include such things as equity interestsin partnerships, corporations and beneficial interests in trusts [Foreign Account TaxCompliance Act (FATCA), 2017]. FATCA requires all non-US financial institutions tosearch their records for customers with indicia of “USA person status, such as a USAplace of birth, and to report the assets and identities of such persons to the USA TreasuryDepartment” [Foreign Account Tax Compliance Act (FATCA), 2017]. FATCA intends toferret out US persons who may be hiding as anonymous beneficiaries of corporatevehicles (Biedermann, 2015).

When launched, FATCA threatened to impose a 30 per cent withholding tax on certainUS source payments for non-participating persons. The 30 per cent tax was considerednecessary to get the attention of other governments (Levine et al., 2016). When over 100countries entered into intergovernmental (bilateral) agreements with the USA and

MAJ34,3

260

pledged to incorporate FATCA into their domestic laws, banks and other regulatedfinancial institutions became de facto responsible for implementing FATCA (Levine et al.,2016).

4.2 Other nationsIn 2013, the risks of hidden entity beneficial ownership and various white-collar crimesreached the attention of high-level political leaders at the G-8 summit in Lough Erne,Northern Ireland. The G-8 countries announced the “G-8 Principles” a set of eightprinciples, or a “beneficial ownership action plan,” to combat the abuse of entities vialegal arrangements [Lakhani, 2016; Biederman, 2015; Government, UK (UK Govt), 2013].One significant outcome from this summit was that it led to the G-20 and Organization forEconomic Cooperation and Development’s (OECD) call for the adoption of a multilateralexchange of information on beneficiaries (Lakhani, 2016; Biedermann, 2015).

Inspired by FATCA, in early 2016, the Group of 20 (G-20) countries committed toadvance towards the implementation of an automatic exchange of information targeting taxevasion (Reis, 2016). The G-20 embraced the OECD’s proposal of a global model of automaticfinancial data exchange known as the Common Reporting Standard (CRS) (Reis, 2016). TheCRS resembles FATCA with a global reach, seeking information about individuals andentities residing in CRS-signatory jurisdictions but held outside their own countries ofresidence (Reis, 2016).

As of March 2018, 102 countries have committed to the adoption of CRS. As of December21, 2017, over 2,600 bilateral exchange relationships have been activated with respect to 78jurisdictions (OECD, 2018a). Despite the implementation and cooperation of CRS countries,information from academic studies, media leaks and governmental authorities show thatprofessional advisors and other intermediaries continue to design, market or assist in theimplementation of offshore structures and arrangements that can be used by non-compliantpersons to circumvent government regulation and commit various white-collar crimes(OECD, 2018b).

Similar to FATCA, CRS obligations depend on an entity’s classification and country ofresidence. Entities include corporations, partnerships, trusts and foundations that areclassified as reporting or non-reporting financial institutions, passive or active non-financialentities (NFEs) (Reis, 2016). Financial institutions include banks, brokers, custodians andinvestment funds. Reporting financial institutions are subject to a comprehensive set ofduties and obligations (Reis, 2016). TCSPs are generally financial institutions for CRSpurposes (Reis, 2016).

The way that CRS rules function demonstrates a clear policy focus on control apart fromownership (Levine et al., 2016). CRS regulations blur the lines between controlling personsand beneficial owners (OECD, 2018a). On the other hand, FATCA seeks to identify UStaxpayers’ beneficial ownership in non-passive assets, the notion being that the taxliabilities are attached to income derived from ownership [Foreign Account Tax ComplianceAct (FATCA), 2017; Levine et al., 2016].

In November 2016, the G-20 nations published a set of principles for governments tofacilitate identification of the beneficial owners of shell entities (Smyth and Parker, 2016). Inthe EU, the 4th Anti-money Laundering Directive (AMLD) requires member states tointroduce registries of company beneficial owners. The UK beneficial ownership registryopened in April 2016, but it excluded trusts (Radon and Aehuthan, 2017). The UK has set aprecedent by creating the world’s first fully open register of beneficial ownership, albeit onethat only discloses those beneficial owners that meet a 25 per cent threshold (Radon andAehuthan, 2017). Wherever registries have become available in other European countries,

Role of shellentities

261

the quality of data (often collected but not verified) has been criticized by industry experts(Wolos and Reuters, 2017).

In December 2017, an agreement was reached between the European Parliament and theEU Council on the latest amendments to the 5th AMLD. The amendments attempt toprevent the use of the financial system for funding white-collar crime. The followingmeasures will be introduced in EUmember states:

� Registers of beneficial owners of firms will be made publicly accessible and nationalregistries will be better interconnected.

� Registers of beneficial owners of trusts and similar legal arrangements will only bepublicly accessible where there is a legitimate need.

� Information on national banks and safe deposit boxes, as well as data on real estateownership, will be registered (only to public authorities) (KPMG, 2017).

� Member states must verify beneficial ownership submitted to their registries(Ivanovsky et al., 2017).

� EU bank customers who send funds internationally must provide personal data so itcan be transmitted to all banks in the payment chain (O’Connor, 2017).

The potential implementation of the 5th AMLD requirements remains to be seen since fewmember states took up the 4th AMLD option of implementing publicly accessible centralregistries of corporate beneficial owners (O’Connor, 2017).

Moreover, as of March 1, 2018, an amended Companies Ordinance requires Hong Kong-incorporated firms to obtain and maintain specified information on their beneficial ownersin a Significant Controllers’ Registry in English or Chinese. The registry is open toinspection by Hong Kong law enforcement (Clifford Chance, 2018).

Beneficial ownership disclosure by itself is not the complete answer to the problems offinancial crimes and lost revenues. Such disclosure is most effective when accompanied bywell-drafted criminal and tax laws, sustained enforcement, modern technology andsustained political will (Radon andAehuthan, 2017).

5. ConclusionWhite-collar criminals, terrorist financiers, tax evaders and other individuals form and usevarious types of domestic and offshore legal shell entities to conceal their identities asbeneficial owners of assets or funds. A beneficial owner is a natural person who controls andenjoys an asset and/or its benefits. Shell entities often have no significant assets or ongoingbusiness activity. The vulnerability of shell entities for illicit purposes is amplified whenthey are privately rather than publicly owned. The extensive abuse of domestic and offshoreshell entities to conceal and transfer assets make them an important aspect of the work offorensic accountants and law enforcement officers.

Trillions of dollars are located in secrecy jurisdictions around the globe. Thetraditional stereotypes of financial secrecy and tax havens are inaccurate. The locationsthat provide the most secrecy are the USA, Switzerland, Hong Kong, Singapore, the UKand its overseas territories, Germany, Luxembourg and certain other nations — notsmall, tropical islands.

White-collar criminals use various shell entities to commit a panoply of crimes suchas tax evasion, money laundering, securities and financial fraud, corruption and bribery.Such criminals may choose from various types of legal structures: LLCs, shelfcorporations, LLPs, FLPs, asset protection trusts, IBCs and PIFs. Each entity type hasits own unique structure and legal characteristics. Nominees, nominee directors and

MAJ34,3

262

bearer shares are legal devices used in combination with shell entities to optimizeconcealment.

Three principal reasons explain the ability of white-collar criminals and others tocontinue to hide their identity as beneficial owners and operators. One reason is that manyjurisdictions possess a legal framework that promotes lack of ownership transparency.Another reason is that those who abuse shell entities need the services of gatekeepers, suchas accountants, lawyers and other TCSPs. A third reason is the layering or chaining ofnumerous shell entities in different jurisdictions that make it virtually impossible forforensic accountants and law enforcement officials to discern the real identity of beneficialowners and operators.

Various global organizations, such as the FATF, and international groups, such as theG-8 and G-20, have started to cooperate in dealing with the issue of hidden entitybeneficial ownership. Improved information exchange (e.g. that provided for in FATCA)is one of the means being used to combat concealed beneficial ownership. The creation ofownership registries is another goal that is starting to receive attention and governmentaction. The use of ownership registries is complicated by numerous issues and concerns,such as privacy infringement, the burdens on financial institutions, infringements onnational sovereignty, bank secrecy, violations of contractual relationships and others.Global efforts to improve entity ownership transparency are moving ahead but at amodest pace.

Notes

1. One building in the Grand Caymans, known as Ugland House, is officially the registered home of18,000 companies. Another address of interest is P.O. Box 3444, Road Town, Tortola, BritishVirgin Islands. A Google search of this address yields more than 600,000 hits (Anonymous, 2013;Hubbs, 2014).

2. The Panama Papers scandal shows the power of data analytics to uncover corruption. Open datacan place lots of data into the hands of those who can transform it into valuable information toidentify, trace and predict financial crime (Santiso and Roseth, 2017).

3. Two central themes are captured by the FCPA. The first is that no entity or person may offer orpay anything of value to an official of a foreign government or certain international organizationsthat would cause the official to misuse power or influence to benefit a business interest of anyentity or person. The second is that if any payment is made to an official, whether the purpose isproper or corrupt, the payment must be reported in the payer’s financial statements according toUS GAAP (Athanas, 2010; Deming, 2006).

4. The most thorough study, entitled “Global Shell Games” by Michael Findley, Daniel Nielson, andJason Sharman, was done in 2012. The authors sent 3,773 formation agents a request posing asconsultants trying to establish untraceable shells. In offshore havens, such as the CaymanIslands, few TCSPs took the bait. Dozens of TCSPs in America took the bait and offered suchservices.

ReferencesAdkisson, J. and Riser, C. (2004), Asset Protection: Concepts and Strategies for Protecting Your Wealth,

McGraw-Hill Education, New York, NY.Anonymous (2013), “Themissing $20 trillion”,The Economist, 16 February.Anonymous (2016a), “Tax haven or . . . tax hell?”, International Tax Review, 26 May, pp. 1-4.

Role of shellentities

263

Anonymous (2016b), “Corporate ownership and corruption: how to crack a shell”, The Economist,Vol. 419, p. 56.

Arce, B. (2009), “Taken to the cleaners: panama’s financial secrecy laws facilitate the laundering ofevaded U.S. taxes”, Brooklyn Journal of International Law, Vol. 34, pp. 465-490.

Aspen Group Limited (2012), “A guide to Panamanian private interest foundations”, available at: www.aspenoffshore.com/files/docs/2012/11/a_guide_to_panamanian_private_interest_foundation_new_2.pdf.

Association of Certified Fraud Examiners (ACFE) (2009), Money Laundering: Tracing Illicit Funds,pp. 1-217.

Athanas, W. (2010), “When doing business internationally becomes a crime: assisting clients inunderstanding and complying with the Foreign Corrupt Practices Act”, The Alabama Lawyer,Vol. 71, pp. 382-387.

Ausness, R. (2007), “The offshore asset protection trust: a prudent financial planning device or the lastrefuge of a scoundrel?”,Duquesne Law Review, Vol. 45, pp. 147-193.

Baker, S. and Shorrock, E. (2009), “Gatekeepers, corporate structures, and their role in moneylaundering”,Tracing Stolen Assets: A Practitioner’s Handbook, Basel Institute on Governance.

Ball, J. (2012), “Offshore secrets: how many companies do ‘sham directors’ control?”, The Guardian,November 26, available at: www.theguardian.com/uk/datablog/2012/nov/26/offshore-secrets-companies-sham-directors.

Baradaran, S., Findley, M., Nielson, D. and Sharman, J. (2014), “Funding terror”, University ofPennsylvania Law Review, Vol. 162, pp. 477-536.

Berbey de Rojas, D. (2008), “Panama: the role of the protector in the private interest foundation”,Trusts & Trustees, Vol. 14 No. 5, pp. 350-353.

Biedermann, M. (2015), “G8 principles: identifying the anonymous”, Brigham Young UniversityInternational Law&Management Review, Vol. 11, pp. 72-92.

Boschini, F.D. (2006), “Private foundations and reserved powers trusts”, Trusts & Estates, Vol. 145,pp. 46-50.

Browning, L. (2009), “Delaware laws, helpful to arms trafficker, to be scrutinized”, N.Y. Times,November, Vol. 4.

Browning, L., Davison, L., Basu, K. and Lee, R. (2018), “Hill briefs: US second largest tax haven”, BNADaily Tax Report, January 31.

Canada Revenue Agency (2018), “Canada revenue agency conducts panama papers related searches inmultiple locations”, February 14.

Carr, K. and Grow, B. (2011), “Special report: a little house of secrets on the Great Plains”, Reuters, 28 June,available at: www.reuters.com/article/us-usa-shell-companies-idUSTRE75R20Z20110628

China Offshore (2009), “Secure the wealth: private foundation structures are spreading throughout theoffshore world as jurisdictions look to diversify their client bases.”

Christensen, J. (2012), “The hidden trillions: secrecy, corruption, and the offshore interface”, Crime, Lawand Social Change, Vol. 57 No. 3, pp. 325-343.

Clifford Chance (2018), “Hong Kong eyes international best practices with 2018 amendment to amlordinance”, February.

Couch, G. (2004), “Shaken trust: The Netherlands rethinks an offshore industry”, N.Y. Times,19 February, p. C1.

Danforth, R. (2002), “Rethinking the law of creditors’ rights in trusts”, Hastings Law Journal, Vol. 53,pp. 287-366.

Davern, R. and Way, A. (2017), “Notes from a small island: some observations on the new CaymanIslands foundation company”,Trusts & Trustees, Vol. 23 No. 9, pp. 916-920.

MAJ34,3

264

Deming, S. (2006), “The potent and broad-ranging implications of the accounting and record-keepingprovisions of the Foreign Corrupt Practices Act”, Journal of Criminal Law and Criminology,Vol. 96, pp. 465-502.

Elements of a Panama private interest foundation (Elements) (2017), available at: www.aspenoffshore.com/files/docs/2012/11/a_guide_to_panamanian_private_interest_foundation_new_2.pdf

Fidelity Investor (2017), “Six reasons you should consider a trust”, Jane 30, available at: www.fidelity.com/viewpoints/personal-finance/reasons-to-consider-a-trust

Financial Action Task Force (FATF) (2004), “FATF 40 recommendations”, available at: www.fatf-gafi.org/media/documents/FATF%20Standards%20-%2040%20Recommendations%20rc.pdf

Financial Action Task Force (FATF) (2006), “The misuse of corporate vehicles, including trust andcompany service providers”, Recommendation 12.

Financial Action Task Force (FATF) (2014), Transparency and Beneficial Ownership, availableat: www.fatf-gafi.org/media/fatf/documents/reports/Guidance-transparency-beneficial-ownership.pdf

Financial Crimes Enforcement Network (FinCEN) (2006), The Role of Domestic Shell Companies inFinancial Crime andMoney Laundering, US Department of the Treasury.

Firmani v. Firmani, 752 A. 2d 854 (N.J. Super. Ct. App. Div. 2000).Foreign Account Tax Compliance Act (FATCA) (2017), “(IRC, 2017), sections 1471-1474 of the internal

revenue code of 1986, as amended, pub. L. No. 111-147, 124 stat. 71 (2010)”, available at: www.irs.gov/Businesses/Corporations/Foreign-Account-Tax-Compliance-Act

Foreign Corrupt Practices Act (FCPA) (2017), codified as amended at 15 U.S.C. sections 78m(b), (d)(1),(g)-(h), 78dd-1, 78dd-2, 78dd-3, 78ff.

GlobalWitness (2012), Grave Secrecy.Global Witness (2013), “Anonymous companies”, available at: www.globalwitness.org/library/

anonymous-companies-global-witness-briefingGovernment Accountability Office (GAO) (2006), Company Formations: Minimal Ownership

Information is Collected and Available, available at: www.gao.gov/new.ites/d06376.pdfGovernment, UK (UK Govt) (2013), “2013 lough erne G8 leaders’ communique”, 18 June, available at:

www.gov.uk/government/publications/2013-lough-erne-g8-leaders-communique

Hall, K. and Taylor, M. (2016), “Massive leak exposes how the wealthy and powerful hide their money”,McClatchyDC, available at: www.mcclatchydc.com/news/nationworld/national/article69994502.html

Hicken, M. and Ellis, B. (2015), “These US companies hide drug dealers, mobsters, and terrorists”, CNNMoney, available at: http://money.cnn.com/2015/12/09/news/shell-companies-crime

Houlder, V. (2017), “Offshore assets held in British Virgin Islands double over 7 years to $1.5 tn”,Financial Times, June 21, p. 1.

Hubbs, R. (2014), “Shell games”, Fraud, July/August., pp. 1-6.In re Brown, 303 F. 3d 1261 (11th Cir. 2002).Ivanovsky, A., Dineen, M., Correia, F. and Valente, P. (2017), “CFE’s tax top 5”, 18 December.

Kalant, D. (2009), “Who’s in charge here?: Requiring more transparency in corporate America:advancements in beneficial ownership for privately held companies”, JohnMarshall Law Review,Vol. 42, pp. 1049-1072.

Komisar, L. (2011), “Shells, scams, and corporate scams”, American Interest, 1 January, available at:www.the-american-interest.com/2011/01/01/shells-shams-and-corporate-scams

KPMG (2017), “Euro tax flash from KPMG’s tax Centre”, Dec.r 22, available at: https://home.kpmg.com/xx/en/home/insights/2017/12/etf-351-amld5-and-ubo-agreement.html

Role of shellentities

265

Lakhani, A. (2016), “Imposing company ownership transparency requirements: opportunities foreffective governance of equity capital markets on corporate performance”, Chicago-Kent Journalof International & Comparative Law, Vol. 16, pp. 122-163.

Levine, R., Schumacher, A. and Zhou, S. (2016), “FATCA and the common reporting standard”, Journalof International Taxation, pp. 43-51.

LiButti v.U.S., 107 F. 3d 110 (2nd Cir. 1997).Lorenzetti, J. (1997), “The offshore trust: a contemporary asset protection scheme”, Commercial Law

Journal, Vol. 102, pp. 138-165.Martinez, I. (2017), “The shell game: an easy hide-and-seek-game for criminals around the world”, St.

Thomas Law Review, Vol. 29, pp. 185-210.Maxwell, T. (2014), “Domestic asset protection trusts: a threat to child support?”, Brigham Young

University Law Review, pp. 477-507.Miller v.Kresser, 34 So. 3d 172 (Fla. Dist. Ct. Appl. 2010) (quoting Croom v. Ocala Plumbing & Elec. Co.,

57 So. 243 (1911)).

Morse, J. (2008), “Nevada self-settled spendthrift trusts or offshore trusts?”, Nevada Lawyer, March,pp. 16-20.

Murphy, M. (2017), “Paradise papers: 6 things to know about report exposing tax havens of the mega-rich”,Marketwatch, pp. 1-9.

Nevada Partners, LLC v.U.S., 720 F. 3d 594 (5th Cir. 2013).O’Connor, D. (2017), “EU fifth anti-money laundering directive: can banks handle it?”, November 21,

available at: https://kyc360.com/article/eu-fifth-anti-money-laundering-directive-key-points-banks

OECD (2018a), “Automatic exchange portal”, available at: www.oecd.org/tax/automatic-exchange-international-framework-for-the-crs

OECD (2018b), “Model mandatory disclosure rules for crs avoidance arrangements and opaque offshorestructures”, available at: www.oecd.org/tax/exchange-of-information/model-mandatory-disclosure-rules-for-crs-avoidance-arrangements-and-opaque-offshore-structures.pdf

Offshore (2005), “Offshore: the British Virgin Islands”,The Lawyer, 13 June, pp. 26-29.Radon, J. and Aehuthan, M. (2017), “Beneficial ownership disclosure: the cure for the Panama papers

ills”, Journal of International Affairs, Vol. 70 No. 2, pp. 87-108.Reis, G. (2016), “Common reporting standard explained”,Trusts & Estates, Vol. 155 No. 5, pp. 37-42.Ryle, G. (2013), “Inside the shell: drugs, arms, and tax scams”, International Consortium of Investigative

Journalists (ICIJ), available at http://icij.org/offshore/geoffrey-taylor

Santiso, C. and Roseth, B. (2017), “Data disrupts corruption”, Stanford Social Innovation Review, Spring,pp. 51-55.

Sharman, J. (2013), “Shell companies and asset recovery”, in Zinkernagel, G., Monteith, C. and Pereira,P.G. (Eds),Emerging Trends in Asset Recovery, Basel Institute on Governance, Switzerland.

Silets, H. and Drew, M. (2001), “Offshore asset protection trusts: tax planning or tax fraud?”, Journal ofMoney Laundering Control, Vol. 5 No. 1, pp. 9-15.

Simser, J. (2008), “Money laundering and asset cloaking techniques”, Journal of Money LaunderingControl, Vol. 11 No. 1, pp. 15-24.

Smyth, J. and Parker, G. (2016), “G20 leaders back drive to unmask shell companies”, FT.com,November 16, available at www.ft.com/search?q=g20þleadersþbackþdriveþtoþunmaskþshellþcompanies

Stempel, R. (2013), “Russian arms dealer Viktor Bout’s US conviction upheld”, Reuters, available at:www.reuters.com/article/2013/09/27/us-usa-crime-bout-idUSBRE98Q0PG20130927

Szubin, A. (2016), “A dangerous shell game”, The Hill, available at http://thehill.com/opinion/op-ed/287291-a-dangerous-shell-game

MAJ34,3

266

Tax Justice Network (TJN) (2018), Financial Secrecy Index, available at: http://financialsecrecyindex.com

Taylor, R. (1998), “Domestic asset protection trusts: the ‘estate planning tool of the decade’ or a‘charlatan’?”, BrighamYoung Journal of Public Law, Vol. 13, pp. 163-183.

Thompson, A. (2010), “The cost of privacy: tax and trading in panama”, FT.com, 30 September.

U.S. Department of Justice (DOJ) (2014), “Hewlett-Packard Russia pleads guilty to and sentenced forbribery of Russian government officials”, available at: www.justice.gov/opa/pr/Hewlett-packard-russia-pleads-guilty-and-sentenced-bribery-russian-government-officials

U.S. v. Beecroft, 825 F. 3d 991 (9th Cir. 2016).U.S. v. Brennan, 395 F. 3d 59 (2nd Cir. 2005).U.S. v. Lake, 571 Appx. 303 (5th Cir. 2014).U.S. v.Mazzarella, 784 F. 3d 532 (9th Cir. 2015).U.S. v. Scott, 37 F. 3d 1564 (10th Cir. 1994).Weiss, D. (2011), “Wyoming home is a ‘little Cayman Island’ for shell companies”, American Bar

Association Journal, available at www.aba.journal.com/news/article/wyoming_home_is_a_little_cayman_island_for_shell_companies

Wiggin, H. (2008), “Anguilla: foundations and trusts-a comparison”, Trusts & Trustees, Vol. 14 No. 5,pp. 287-295.

Willebois, E.V.D., Halter, E., Harrison, R.A., Park, J.W. and Sharman, J. (2011), The Puppet Masters,World Bank,WA, D.C.

Wolos, S. and Reuters, T. (2017), “The ultimate beneficial ownership identification requirement: why itmatters to all of us”, Acamstoday, available at: http://acamstoday.org/the-ultimate-beneficial-ownership-requirement-why-it-matters-to-all-of-us

Zagaris, B. (1999), “A brave new world: recent developments in anti-money laundering and relatedlitigation traps for the unwary in international trust matters”, Vanderbilt Journal ofTransnational Law, Vol. 32, pp. 1023-1116.

Corresponding authorCarl Pacini can be contacted at: cpacini@mail.usf.edu

For instructions on how to order reprints of this article, please visit our website:www.emeraldgrouppublishing.com/licensing/reprints.htmOr contact us for further details: permissions@emeraldinsight.com

Role of shellentities

267

Relevance of big data to forensicaccounting practice and education

Zabihollah RezaeeFogelman College of Business and Economics, University of Memphis, Memphis,

Tennessee, USA, and

JimWangSchool of Business, Tung Wah College, Hong Kong, Hong Kong

AbstractPurpose – This paper aims to examine the relevance of Big Data to forensic accounting practice andeducation by gathering opinions from a sample of academics and practitioners in China.Design/methodology/approach – The authors conduct a survey of academics and practitionersregarding the desired demand, importance and content of Big Data educational skills and topics forforensic accounting education to effectively respond to challenges and opportunities in the age of BigData.Findings – Results indicate that the demand for and interest in Big Data/data analytics and forensicaccounting will continue to increase; Big Data/data analytics and forensic accounting should be integratedinto the business curriculum; many of the suggested Big Data topics should be integrated into forensicaccounting education; and some attributes and techniques of Big Data are important in improving forensicaccounting education and practice.Research limitations/implications – Readers should interpret the results with caution because of thesample size (95 academics and 103 practitioners) and responses obtained from academics and practitioners inone country (China) that may not be representative of the global population.Practical implications – The results are useful in integrating Big Data topics into the forensicaccounting curriculum and in redesigning the forensic accounting courses/programs.Social implications – The results have implications for forensic accountants in effectively fulfilling theirresponsibilities to their profession and society by combating fraud.Originality/value – This study provides educational, research and practical implications as Big Data andforensic accounting are advancing.

Keywords Big data, Big data analytics, Business curriculum,Forensic accounting education and practice

Paper type Research paper

JEL classification –M40, M41, M42This manuscript benefits from the invaluable comments and suggestions of participants at the 2016World Business Ethics Forum in Hong Kong, the 2017 Midyear Forensic Accounting Conference ofthe American Accounting Association in Orlando in March 2017, and the 2017 American AccountingAssociation Annual Meeting in San Diego, in August 2017. The authors thank the workshopparticipants at the University of Memphis, the United International College and the Sino-US Collegein Zhuhai, China for their valuable comments on the earlier draft of the manuscript with gratitude,this research is funded by a research grant provided by the University Grant Committee (projectreference: UGC/FDS17/B02/14). The authors appreciate support given by Tung Wah College in HongKong.

MAJ34,3

268

Received 26August 2017Revised 30March 2018Accepted 21 July 2018

Managerial Auditing JournalVol. 34 No. 3, 2019pp. 268-288© EmeraldPublishingLimited0268-6902DOI 10.1108/MAJ-08-2017-1633

The current issue and full text archive of this journal is available on Emerald Insight at:www.emeraldinsight.com/0268-6902.htm

1. IntroductionForensic accounting has emerged as a major area of accounting practices, which includesfraud examination, anti-corruption and anti-bribery, business valuation, litigation support,expert witnessing and cyber security (Crumbley et al., 2015; Rezaee et al., 2004). The demandfor forensic accounting services is growing as businesses, regulators and investors havecontinued to raise concerns about fraud, financial irregularities, corruption and briberycases. For example, business organizations lose about 5 per cent of their revenues to fraudeach year, which can exceed US$3.5tn worldwide (ACFE, 2016). Information technologyadvances (e.g. cloud, social media and analytics) enable organizations to have anunprecedented amount of structured, semi-structured and unstructured data. Theemergence of “high-volume, high-velocity, and high-variety” information that can beprocessed electronically to facilitate decision-making is usually described as Big Data(Gartner, 2014)[1]. The evidence of relevance of Big Data skills and knowledge to forensicaccounting practice and education is rare (Rezaee et al., 2016). To help update and advanceforensic accounting curricula with Big Data content, this research conducts a survey of bothacademics and practitioners in China to gather insight regarding the integration of forensicaccounting and Big Data into the business curricula.

Insight regarding Big Data and its integration to forensic accounting education fromacademics and practitioners from Hong Kong and mainland China are obtained for threereasons. First, forensic accounting in Hong Kong and China has grown significantly in thepast several years (Rezaee et al., 2016). Second, anecdotal evidence suggests that AsianPacific countries (APAC) are the fastest-growing region for the Big Data market (Marketsand Markets, 2017). Third, China has advanced as the world’s second largest economyvulnerable to corporate scandals and fraud (Hung et al., 2015)[2]. The primary purposes ofthis paper are to:

� describe the use of Big Data in forensic accounting practice;� investigate the relevance and importance of Big Data to the practice and education

of forensic accounting;� examine the demand for and interest in integrating Big Data and forensic

accounting into business curricula; and� present Big Data topics relevant to forensic accounting education and their

integration into the business and accounting curricula.

Analysis of the insight gained from a sample of academics and practitioners indicates that:� the demand for and interest in Big Data/data analytics and forensic accounting will

continue to increase;� Big Data/data analytics and forensic accounting can be integrated into the business

curricula at both undergraduate and graduate levels;� many of the suggested 25 Big Data topics should be integrated into forensic

accounting education; and� some attributes and techniques of Big Data such as the availability of Big Data, data

analytics techniques, accuracy, reliability, accessibility, relevancy and predictive,descriptive and prescriptive analytics are important in improving forensicaccounting education and practice.

This study makes several contributions to the Big Data and forensic accounting practice andeducation literature. First, results support recent initiatives taken by the Association to

Relevance ofbig data

269

Advance Collegiate Schools of Business (AACSB) International accounting program tointegrate information technology throughout academic curricula. The AACSB Standard A7(AACSB, 2014, p. 3) underscores the importance of Big Data and data analytics to businesseducation by suggesting “the development of skills and knowledge related to data creation,data sharing, data analytics, data mining, data reporting, and storage within and acrossorganizations.” Thus, this study provides evidence of the importance of Big Dataand forensic accounting education. Second, the findings of this study can be used to advanceforensic accounting education by integrating Big Data (data analytics)-related topics into theforensic accounting curricula as suggested by Cao et al. (2015) and Rezaee et al. (2018).Finally, this study presents evidence of the use of data analytics in forensic accountingpractice and provides some survey results of the relevance of the use Big Data (dataanalytics) in forensic accounting practice. Future research can use the insights to furtherinvestigate and promote the use of Big Data in forensic accounting education and practice[3].

Section 2 describes the use of Big Data in forensic accounting practices. Section 3 reviewsthe literature and presents the research questions. Section 4 illustrates research method,survey questionnaire and sample selection procedures. Section 5 reports the results anddiscussion and Section 6 concludes.

2. Use of big data in forensic accounting practicesForensic accounting services have emerged as an important practice in accounting firms(Crumbley et al., 2015). Forensic accountants can play an important role in discovering andpreventing fraud, corruption and bribery cases. Forensic accounting has recently gainedconsiderable attention as evidenced by a large number of universities offering forensicaccounting courses and programs, many published textbooks in forensic accounting, andseveral established certifications in forensic accounting (Crumbley et al., 2015). Big Data hasalso received common acceptance and practical application in the business community. Forexample, more than 98 per cent of all stored information is now electronic, compared withabout 25 per cent of digital information in 2000 (Cukier and Mayer- Schonberger, 2013;Crumbley et al., 2015). A recent survey conducted by Ernst and Young (EY) shows that “79per cent of respondents use more than 10 million records, which are typically outside thedomain of spreadsheets and require more sophisticated tools for analysis” (EY, 2016, p. 25).

Big Data and business analytics are also being extensively used in recent years (Marketsand Markets, 2017). The Big Data market has grown dramatically from $16.1bn in 2014(Forbes, 2013a) to over $50bn by the end of 2016 (Kelly et al., 2015) and is expected tocontinue to grow. The estimated market for Big Data is about $67bn by 2021, with dataanalytics software holding the highest market share (Markets and Markets, 2017). However,there is an inadequate supply of professionals with Big Data skills. Laney and Kart (2012)predicts a shortage of over 100,000 analytics talents by 2020. McKinsey Global Institute(2011) reports similar shortage of 140,000 to 190,000 professionals with analytical expertiseby 2018 in the USA. Forensic accountants are now able to obtain a huge amount of bothstructured (e.g. general ledger or transaction data) and unstructured data (e.g. email, voice orfree-text fields in a database), together with an increasing amount of nontraditional datasources such as third-party watch lists, news media, free-text payment descriptions, emailcommunications and social media. As a result, forensic accountants use advancedtechnological tools in their investigative practices. For example, forensic accountants usesocial media andWeb monitoring, voice searching and analysis, visualization and reportingtools (EY, 2016).

Two surveys conducted by EY (2014, 2016) report the trends toward the use of IT andBig Data/analytics in forensic accounting practices. First, there is an increasing use of

MAJ34,3

270

forensic data analytics in forensic and investigative accounting services. Second, cyberbreaches (illicit transferring of funds, disrupting critical operations or stealing intellectualproperty/confidential personal data) and insider threats (malicious insiders stealing,manipulating or destroying data, fraud and unauthorized trading information technologysabotage) are emerging as the fastest growing fraud risks faced by forensic accountants.Forensic accountants increasingly use Big Data (analytics) in their practices to deal withdata sets exceeding the typical constrains of a traditional spreadsheet (EY, 2016). Forensicaccountants also use data visualization, predictive analytics, behavior analytics, contentanalytics, social network analysis, geo-spatial analytics and numerous advanced anti-fraudtechniques to overcome the shortcomings of the traditional rules-based relational databasetechniques, such as matching, sorting, filtering and query design (EY, 2016).

There is a shortage of forensic accountants who possess Big Data/analytical skills andare able to use sophisticated analytical tools to effectively and accurately identify potentialrisks and proactively search for irregularities and assess and manage their risk profile. Thegrowing demand and inadequate supply of Big Data professionals raises the questionwhether there is adequate training related to Big Data at the undergraduate/graduate levelin the forensic accounting education. The integration of and Big Data into forensicaccounting and business curricula supports the market demand for digital forensicaccounting services and keeps business curricula aligned with digital forensic accountingpractices. Anecdotal evidence fromWang et al. (2016) show that only 3 out of 19 universitieswith the forensic accounting program in China have a standalone course on Big Data.Rezaee et al. (2018) examine forensic accounting syllabi of many universities worldwide fortheir coverage of Big Data topics and conclude Big Data and data analytics topics are notsufficiently covered in forensic accounting courses and programs.

The following examples illustrate the use of Big Data (analytics) in forensic accountingpractice. First, when forensic accountants investigate fraud, corruption or bribery cases, theytake industry-specific norms or regulations into consideration and use keyword phrases toidentify potential fraud. Second, by using historical activities or transaction data, forensicaccountants can use predictive modeling and other advanced analytics to detect suspicious andanomalous transactions, high-risk events, or potential fraudulent behavior or activities. Third,by mining across multiple databases (such as customer or third-party databases), forensicaccountants can use entity resolution algorithms to identify hidden relationships, addressesand aliases and investigate conflicts of interest, fake identities or sanctioned individuals andentities. Fourth, forensic accountants use social network analytics to detect hiddenrelationships, bogus vendors or fake bank accounts when they analyze both structured andunstructured data in the format of visuals and links from social media. Fifth, a large amount ofunstructured text data is available from the free text field of journal entries, paymentdescription, expense details, e-mails, social media, documents, presentations and hard drives ofindividual employees or organizations. Forensic accountants use text mining or text analyticswith heuristic rules and statistical techniques to discover the sentiments and conceptualmeanings of large amounts of text data, which help to identify potential fraud or non-compliance in the organization. Finally, besides traditional simple spreadsheets or static chartsand graphs, forensic accountants use data visualization techniques and interactive dashboardsto present evidence in an easy to understandmanner.

3. Literature review and research questionsForensic accounting education and practice have gained attention from research scholars.Seda and Peterson-Kramer (2014) examine the availability of forensic accounting educationin the USA and other English-speaking countries and find that in the USA, there are 422

Relevance ofbig data

271

universities and colleges offering forensic accounting courses, with 97 of them providingforensic accounting programs; while in Canada and other English-speaking countries, thereare 25 universities and colleges offering forensic accounting courses, with 23 forensicaccounting programs among 186 Canadian and other English-speaking universities andcolleges. Rezaee et al. (2004) conducted a survey of both academics and practitioners in theUSA and provided evidence of the importance of forensic accounting education and practiceand its integration into business curricula. Rezaee et al. (2016) examine forensic accountingpractice in China and survey both Chinese and international students regarding theimportance, demand, relevance, benefits, coverage and delivery of forensic accountingeducation and suggest that business and accounting universities in China integrate manysuggested forensic accounting topics into their curricula.

There is a growing demand for Big Data knowledge and skills, particularly among businessand accounting professionals. Russom (2011) made predictions of an increasing use of BigData, particularly in the areas of predictive analytics, machine learning, artificial intelligence,visualization techniques (dashboards), data warehouses, dedicated database managementsystems and Big Data technology (e.g. Hadoop, distributed file system). Forbes (2013b)reported that the Securities and Exchange Commission in the USA has used Big Data/analytics, which were labeled as “RobotCop,” to identify securities law violations and financialstatements irregularities and audit failures. Issa and Kogan (2014) point out that there is anincreasing demand for Big Data knowledge and skills for auditors to understand the qualityand relevance of the Big Data to make professional judgments. For example, auditors can usepredictive analytics to overcome the cognitive limitations associated with ambiguity withinfrom voluminous data and to identify meaningful patterns. Cao et al. (2015) argue that becauseclients use Big Data, auditors should adopt Big Data/analytics in risk assessment, substantiveanalytical procedures, collection of audit evidence and obtaining audit confirmations. TheCharted Global Management Accountant (2014) survey suggests that the use of cloud-basedsolutions in accounting information systems encourage professional accountants to knowmoreabout Big Data skills, such as cyber/information security. The recent survey of Fortune 1,000firms by New Vantage Partner (2016) find that they increase investment in Big Data to obtaingreater insights into business and customers, and a majority of these firms (about 54 per cent)have well-defined roles of Chief Data Officers.

There is an inadequate supply of Big Data professionals. Laney and Kart (2012) forecasta shortage of data specialists that “the need, for data scientists, is growing at about threetimes that for statisticians and business intelligence analysts, and there is an anticipated100,000-plus-person analytic talents shortage through 2020.” Similarly, McKinsey GlobalInstitute (2011) predicted:

[. . .] the United States alone faces a shortage of 140,000 to 190,000 people with analytical expertiseand 1.5 million managers and analysts with the skills to understand and make decisions based onthe analysis of Big Data.

There is a growing demand for, and short of supply of, forensic accountants with knowledgeand skills in digital investigation and Big Data analytics as well. Consistent with theanticipation that demand for Big Data professionals will rise, Wixom et al. (2014) documentthat the market demand for students with Business Intelligence (BI) and Business Analytics(BA) skill sets are growing. Global Times (2017) reports an estimated demand of 1.8 millionBig Data professionals in the next three to five years in China, which is 1.5 million more thanthe current supply.

The growing demand, and shortage, of Big Data professionals raises the questionwhether there is adequate training related to Big Data at the undergraduate or graduate

MAJ34,3

272

level in the forensic accounting education. Hong Kong Institute of CPAs and ChineseInstitute of CPAs call on universities to upgrade their accounting curricula to better prepareaccounting students for the challenges provided by Big Data (HKICPA, 2013). This study isa response to the call for more coverage of digital forensic accounting education and thusdiffers from prior research (Rezaee et al., 2004; Seda and Peterson-Kramer, 2014; Rezaeeet al., 2016) and contributes to forensic accounting education literature. Rezaee and Burton(1997) examine coverage of forensic accounting and the role of forensic accounting educationas perceived by academics and practicing certified fraud examiners and conclude that theintegration of forensic accounting education into business curricula either through aseparate course or through modules in business, accounting and auditing courses is vital.Rezaee et al. (2004) survey both academics and practitioners and collect their opinions on theimportance, relevance and delivery of forensic accounting education. Seda and Peterson-Kramer (2014) update the status of forensic accounting education in the USA and otherEnglish-speaking countries and document an increasing provision of forensic accountingeducation in the USA and other English-speaking countries. Wang et al. (2016) analyze theenvironmental factors and find the increasing availability of forensic accounting educationin Hong Kong and Mainland China. These prior studies did not examine the IT issue orrelevance of Big Data to forensic accounting education. This study further investigates therelevance of Big Data to forensic accounting education to prepare students for the demand ofBig Data skills in their future employment.

China has an institutional setting different from many Western countries in terms ofeconomic, legal, social, cultural and political environment (Wang et al., 2016). However, priorresearch of forensic accounting practice and education in a Chinese setting is rare (Rezaee et al.,2016). Wang et al. (2016) examine the environmental factors in China and concludes anincreasing demand for forensic accounting services and forensic accounting education. Anotherresearch conducted by Rezaee et al. (2016) examines the forensic accounting education in Chinaand surveys Chinese students and international students in China on the coverage of forensicaccounting topics in business curricula. This research differs from Rezaee et al. (2016), whichsurvey students on forensic accounting topics, and conducts a survey of both practitioners andacademics on the issues of Big Data in the forensic accounting practice and education. As theinput of both academics and practitioners are important in the curriculum design, this studyhelps to advance forensic accounting education by incorporating Big Data topics into curricula.Thus, the research questions addresses in this study are:

RQ1. What is the status of forensic accounting practice and education in China one ofthe fastest emerging economies andmarkets?

RQ2. What is the status of Big Data and Data analytics in China?

RQ3. How can both Big Data and forensic accounting can be integrated into businesscurricula?

RQ4. What are the topical contents of Big Data and forensic accounting businesscurricula integration?

RQ5. What are the attributes, skills and techniques of Big Data that can be integratedinto forensic accounting courses and programs?

4. Research method and proceduresThe review of the literature performed in Section 3 enables us to identify Big Data topicsrelevant to forensic accounting education. Then, we conduct a survey of academics and

Relevance ofbig data

273

practitioners regarding the desired demand, importance and content of Big Data educationalskills and topics for forensic accounting education[4].

4.1 QuestionnaireWe prepared, pretested and revised the draft of the four-page, six-section questionnaire. Weconducted a pilot and pretesting of the questionnaire by sending it to several academics andpractitioners known to authors and experts in the areas of forensic accounting and Big Data.They were asked to review, correct and suggest for improvements and refinements of theoriginal draft of the questionnaire for its relevance, content and wordings. A revised, refinedand pre-tested four-page, six-section questionnaire was then sent to the participants. The sixmain sections of the survey asked respondents for their perceptions of the future demandfor, and interest in Big Data and forensic accounting, ways that forensic accounting and BigData education can be integrated into business curricula and educational content of BigData and forensic accounting education. The last sections address the relevance andimportant of the practice and education in forensic accounting with a focus on Big Data andsought comments on forensic accounting and Big Data. To improve the response rate, weincluded with each questionnaire a cover letter stating the survey objectives, definingforensic accounting and Big Data, assuring the confidentiality of the responses, agreeing toshare the summary of findings and giving respondents, the appropriate amount of timeneeded to complete the questionnaire. The original draft of the questionnaire was pre-testedby asking several academics and practitioners to review it for content, format, completenessand accuracy. Corrections were made in the final draft submitted to participants throughonline survey. The survey link was generated by an automated survey system in the emailof selected participants, which assured their responses would be completely anonymous[5].A copy of the questionnaire is provided in the Appendix.

4.2 SampleA survey was introduced to participants at three places. First, participants at the 6th WorldBusiness Ethics Forum in December 2016 were approached and invited to participate in thesurvey and respond to the related on-line questionnaire[6]. Second, authors made apresentation to forensic accounting practitioners through the forensic accounting interestgroup of Hong Kong Institute of Certified Public Accountants (HKICPA) in December 2016and invited them to participate in the survey[7]. Third, one of the authors conductedworkshops on forensic accounting and Big Data at two universities in mainland China andencouraged colleagues to participate in the survey.

The survey to both academics and practitioners was conducted using an on-linequestionnaire. We sent the cover letters to 500 academics and 500 forensic accountingpractitioners. Participation was voluntary with no compensation, and all participants wereensured that no identifying information was collected and only summary results reported.To improve the response rate, we distribute the cover letter to participants and remindedthem to complete the on-line survey at the conference venue of the 6th World Business

Table I.Sample andresponses

Chinese academics Chinese practitioners Total

Conducted 500 500 1,000No responses 405 397 802Usable responses 95 103 198Response rate (%) 19 20.6 19.8

MAJ34,3

274

Ethics Forum and at the training workshops of Hong Kong Institute of Certified PublicAccountants forensic accounting interest group and two universities in China. We receivedresponses from 95 academics and 103 practitioners, with a response rate of 19 and 20.6 percent, respectively. The overall response rate is comparable to a prior study (Rezaee et al.,2004, 15.4 per cent for academics and 10.7 per cent for practitioners). The response rate iscommon in the survey study of certain types of individuals (Hodge, 2003).

We used the chi-square test of independence to test the differences in responses ofcategorical variables for different groups of respondents. We used Kruskal–Wallis non-parametric analysis of variance to examine the difference in the responses of ranked data.We also applied the t-test to test the differences between the two groups of Chineseacademics and practitioners with regard to their responses. The results are qualitativelysimilar between chi-square tests and t-tests, which indicate the independence between thesetwo groups. To evaluate and determine the strength of response from each group in a five-point Likert Scale, we follow Rezaee et al. (2004) and Campbell and Mutchler (1988) andcalculate the absolute value of the difference between the mean response of the group andthe neutral response of 3.0 mean responses falling within 0.5 points of the mean response 3.0are considered neutral ratings.

5. Results and discussionsResults are presented in the following four categories:

(1) the horizon for Big Data and forensic accounting practice;(2) method of integrating Big Data and forensic accounting into business curricula;(3) outcome and coverage of Big Data curricula content; and(4) use and relevance and importance of Big Data in forensic accounting practice and

education.

5.1 Horizon for big data and forensic accountingTable II summarizes the responses to a question regarding the demand for and interest inBig Data and forensic accounting. Amajority of both academic and practitioner respondentsreported the increased demand for and interest in Big Data and forensic accounting. Thedifferences in responses between practitioners and academics are statistically significant asa higher percentage of academics, 94 per cent compared to 83 per cent of practitioners,reported that demand for and interest in Big Data will continue to increase. The differencesin responses regarding the demand for and interest in forensic accounting are not

Table II.Big data and forensic

accounting

(%)Big data Forensic Accounting

Chinese academics Chinese practitioners Chinese academics Chinese practitioners

Increase? 94 83 66 66Remain the same? 2 8 20 21Decrease? 0 4 3 9Unsure? 4 5 11 4Total 100% 100% 100% 100%

Note: Demand and interest in Big Data and Forensic Accounting percentage for both groups. Do youexpect future demand and interest in the following two areas?

Relevance ofbig data

275

statistically significant between the two groups. Overall, there is more demand for, andinterest in, Big Data than forensic accounting by both groups of respondents as 83 per centof practitioners believe that the demand for Big Data will increase, while only 66 per cent ofthem think that the demand for forensic accounting will increase. Similar results arerevealed for academics as a high percentage of academics (94 per cent) felt that the demandfor Big Data will increase, but only 66 per cent of them expect that the demand for forensicaccounting will increase. These results suggest that both groups of respondents areexpecting that demand for, and interest, in Big Data will increase more than demand forforensic accounting. One explanation is that Big Data has emerged substantially in recentyears, whereas forensic accounting has been in practice in many years and has reached itsmaturity acceptance by both academics and practitioners. However, academics are moreoptimistic than practitioners regarding the demand for, and interest, in Big Data.

Table III shows that a majority of both academics and practitioners believe that Big Dataand forensic accounting courses should be offered at both graduate and undergraduatelevels. Panel A of Table III shows that the majority of academics (71 per cent) andpractitioners (65 per cent) reported that a Big Data course should be offered at bothundergraduate and graduate levels with some support also for graduate coverage (over 22per cent) and not much in favor of an undergraduate course in Big Data. Panel B of Table IIIindicates that the majority of academics (65 per cent) and about 47 per cent of practitionersfelt that a forensic accounting course should be offered at both undergraduate and graduatelevels. Again, there is some support for offering a graduate forensic accounting course (over27 per cent), whereas there is less preference for an undergraduate coverage of forensicaccounting (less than 20 per cent) for both groups. Differences in responses between twogroups are not statistically significant. These results suggest that both groups ofrespondents believe that forensic accounting and Big Data subjects are more advancedtopics that require prerequisite accounting and business knowledge for students in lowerundergraduate accounting and business courses[8].

5.2 Perception toward integration of big data and forensic accounting educationRespondents were asked to express their opinion on the integration of Big Data into forensicaccounting education. We asked both academics and practitioners to respond to nine relatedquestions by ranking their responses on a five-point Likert scale, with “5” indicating

Table III.

Level Chinese academics (%) Chinese practitioners (%)

Panel A: At what level do you think a Big Data course should be offered?Graduate 22 20Undergraduate 6 13Both graduate and undergraduate 71 65None 1 2Total 100 100

Panel B: At what level do you think a Forensic accounting course should be offered?Graduate 27 29Undergraduate 7 19Both graduate and undergraduate 65 47None 1 5Total 100 100

MAJ34,3

276

“strongly agree” and “1” representing “strongly disagree.” Results presented in Table IVindicate that both groups agree that Big Data will help forensic accounting students to:

� perform data-mining and modeling in forensic accounting investigations (4.22 meanresponse for academics and 3.87 for practitioners);

� have advanced analytical and other data management skills (mean responses of 4.21and 3.94 for academics and practitioners, respectively);

� extract, transform and leverage syndicated data for use in forensic accountingpractices (4.09 and 3.95, respectively);

� have a clear and coherent database digital strategy (3.92 and 3.74, respectively);

Table IV.Opinion on big data

in forensicaccountingpractice

Chinese academics Chinese practitionersSignificantlydifferent

RankMean

responseStandarddeviation Curriculum content 5% 1% Rank

Meanresponse

Standarddeviation

1 4.37 0.36 Forensic data analytics No No 1 4.05 0.552 4.33 0.27 Data base management Yes Yes 13 3.70 0.783 4.31 0.30 Ethical issue in business

intelligenceYes Yes 24 3.49 0.95

4 4.29 0.31 Forensic analytical tools(EnCase)

No No 7 3.83 0.70

5 4.28 0.32 Data structure/data warehouse Yes Yes 17 3.63 0.696 4.26 0.34 Data mining/predictive

modeling analysisYes No 10 3.78 0.74

7 4.21 0.40 Digital investigation Yes Yes 4 3.88 0.678 4.20 0.41 Data visualization No No 2 3.93 0.579 4.19 0.60 Information assurance and

authenticationNo No 5 3.86 0.85

10 4.18 0.48 Big Data technologies(Hadoop, Map Reduce)

Yes No 14 3.68 0.84

11 4.18 0.47 Data integration Yes No 18 3.61 0.8412 4.11 0.51 Expert system/artificial

intelligenceYes Yes 23 3.51 0.79

13 4.08 0.62 Business intelligence usertools (OLAP)

Yes Yes 16 3.64 0.69

14 4.07 0.68 Data governance Yes Yes 12 3.72 0.7615 4.06 0.56 Computer forensics No No 3 3.91 0.6816 4.05 0.58 Mobile digital forensics No No 8 3.82 0.7617 4.03 0.55 Networks, internet and

E-commerceYes No 22 3.55 0.81

18 3.99 0.59 Data movement (in-memorydata)

Yes No 20 3.59 0.78

19 3.98 0.53 Text analytics No No 19 3.61 0.7820 3.96 0.78 Dimensional modeling Yes Yes 21 3.57 0.7221 3.94 0.79 Cybercrime, computers and

auditorsNo No 6 3.83 0.74

22 3.93 0.75 Data streaming management Yes Yes 25 3.48 0.8723 3.87 0.70 Recovery of digital data No No 11 3.73 0.7224 3.87 0.71 Digital evidence seizure No No 9 3.81 0.7125 3.79 0.90 Data encryption Yes Yes 15 3.68 0.59

Relevance ofbig data

277

� use and interpret data sets that may not have standard data formats (3.87 and 3.73,respectively);

� effectively present findings to diverse audiences using strong verbal, written, andvisual communication skills (3.81 and 3.83, respectively); and

� connect the dots in mining data for patterns that lead to evidence (3.78 and 3.68,respectively). The last two questions regarding “sharing Big Data with others” and“align people, processes, and culture” receive relatively low agreement from eithergroup of respondents.

Although the two groups rank the importance of various opinions regarding Big Data verysimilarly, there are some differences in raking among them. For example, practitioners aremore in agreement with statements that Big Data should help forensic accountants toextract, transform and leverage syndicated data for use in forensic accounting practices(Rank 1st); which academics believe less important (Rank 3rd). All responses to statementspresented in Table IV between the two groups are statistically insignificant at the 0.05 and0.01 levels[9], which suggest that academics and practitioners unanimously agree on theoutcomes of Big Data knowledge and skills in the forensic accounting education.

5.3 Curriculum content of big data in forensic accounting educationGiven that the demand for and interest in use of Big Data in forensic accounting willincrease, what should be the curriculum content of Big Data in forensic accountingeducation? We asked both groups of respondents to indicate the importance of 25 suggestedBig Data topics by using a Likert scale of one to five, with five being the “most important”and one being the “least important.” These 25 topics come from a review of existingliterature (Business Intelligence Congress, 2012; Wixom et al., 2014; Gupta et al., 2015; EY,2014; EY, 2016; Rezaee et al., 2018)[10]. The results in Table V reveal that academics rankthe importance of 25 topics for integration into forensic accounting or auditing curriculamuch higher than practitioners, and the differences are statistically significant for mostresponses. Results show that both academics and practitioners ranked the topics “forensicdata analytics,” “forensic analytical tools (Encase),” “data mining/predictive modelinganalysis,” “digital investigation,” “data visualization” and “information assurance andauthentication” on the top of the list.

The main disparity between the two groups involved the topics of “databasemanagement” (Rank Number 2 and 13 for academics and practitioners, respectively),“ethical issue in business intelligence” (3 and 24, respectively), “data structure/datawarehouse” (5 and 17, respectively), “expert system/artificial intelligence”(12 and 23,respectively), “computer forensics” (15 and 3, respectively), “mobile digital forensics” (16 and8, respectively),“cybercrime, computers and auditors” (21 and 6, respectively) and “digitalevidence seizure” (24 and 9, respectively). Compared with practitioners, academics perceivethat these four topics: “database management” “ethical issue in business intelligence” “datastructure/data warehouse” and “expert system/artificial intelligence” are more important.However, practitioners rank “digital investigation” higher than academics (Rank Number 4and 7 by practitioners and academics, respectively).

The differences between academics and practitioners, regarding 10 Big Data topics, arestatistically insignificant, including “forensic data analytics,” “forensic analytical tools(Encase),” “data visualization” and “information assurance and authentication”. Overall, allthe suggested 25 topics are considered by both groups of academics and practitioners asimportant (with mean responses of greater than 3.68) to be integrated into forensicaccounting and business education.

MAJ34,3

278

5.4 Relevance of big data in forensic accounting practiceWe ask two groups of respondents regarding the three attributes of Big Data that mayinfluence forensic accounting practice and education:

(1) the effect of Big Data on the effectiveness of forensic accounting;(2) the use of Big Data in practicing forensic accounting; and(3) importance of several other attributes Big Data in forensic accounting.

Table V.Big data curriculum

content

Chinese academicsChinese

practitionersSignificantlydifferent

RankMean

responseStandarddeviation Curriculum content 5% 1% Rank

Meanresponse

Standarddeviation

1 4.37 0.36 Forensic data analytics No No 1 4.05 0.552 4.33 0.27 Data base management Yes Yes 13 3.70 0.783 4.31 0.30 Ethical issue in business

intelligenceYes Yes 24 3.49 0.95

4 4.29 0.31 Forensic analytical tools(EnCase)

No No 7 3.83 0.70

5 4.28 0.32 Data structure/datawarehouse

Yes Yes 17 3.63 0.69

6 4.26 0.34 Data mining/predictivemodeling analysis

Yes No 10 3.78 0.74

7 4.21 0.40 Digital investigation Yes Yes 4 3.88 0.678 4.20 0.41 Data visualization No No 2 3.93 0.579 4.19 0.60 Information assurance and

authenticationNo No 5 3.86 0.85

10 4.18 0.48 Big Data technologies(Hadoop, Map Reduce)

Yes No 14 3.68 0.84

11 4.18 0.47 Data integration Yes No 18 3.61 0.8412 4.11 0.51 Expert system/artificial

intelligenceYes Yes 23 3.51 0.79

13 4.08 0.62 Business intelligence usertools (OLAP)

Yes Yes 16 3.64 0.69

14 4.07 0.68 Data governance Yes Yes 12 3.72 0.7615 4.06 0.56 Computer forensics No No 3 3.91 0.6816 4.05 0.58 Mobile digital forensics No No 8 3.82 0.7617 4.03 0.55 Networks, internet and

E-commerceYes No 22 3.55 0.81

18 3.99 0.59 Data movement (in-memorydata)

Yes No 20 3.59 0.78

19 3.98 0.53 Text analytics No No 19 3.61 0.7820 3.96 0.78 Dimensional modeling Yes Yes 21 3.57 0.7221 3.94 0.79 Cybercrime, computers and

auditorsNo No 6 3.83 0.74

22 3.93 0.75 Data streaming management Yes Yes 25 3.48 0.8723 3.87 0.70 Recovery of digital data No No 11 3.73 0.7224 3.87 0.71 Digital evidence seizure No No 9 3.81 0.7125 3.79 0.90 Data encryption Yes Yes 15 3.68 0.59

Relevance ofbig data

279

Panel A of Table VI show that academics rank “availability of Big Data” as the mostimportant factor whereas practitioners feel that the “availability of analytical talent”is the most important factor. However, both academics and practitioners agree that“analytical techniques and tools to support forensic accounting” is the second mostimportant factor that can influence the effectiveness of forensic accounting. Theresults of Panel A show that the differences between academics and practitioners aresignificant at the 0.05 level, except for “availability of analytical talent” (insignificantat 0.05 level), which suggests that academics are more interested in “Big Data” itself,whereas practitioners are more concerned with the analytical talent and techniques inusing Big Data.

Panel B of Table VI indicates that both academics and practitioners agree that“descriptive analytics,” “capturing data” and “aggregating/integrating data” are very usefulin forensic accounting practices suggesting that both academics and practitioners believethese Big Data techniques are important in advancing forensic accounting education andpractice. However, academics and practitioners rank “predictive analytics,” “prescriptiveanalytics” and “disseminating data” differently. Academics feel that forensic accountantsuse “predictive analytics” the most, whereas practitioners believe that “disseminating data”and “capturing data” are the most important use of Big Data. However, in most cases,responses from academics are not significantly different from neutral response of 3.0 and sothe strength of the response is weak. In comparison, the responses from practitioners arestrong and significant from neutral response of 3.0. The differences between academics andpractitioners on the use of three Big Data techniques (“capturing data,” “aggregating/integrating data” “and disseminating data”) are not statistically significant at the 0.05 and0.01 level[11].

Panel C of Table VI reveals that both academics and practitioners agree that accuracy,reliability and completeness attributes of Big Data are important in forensic accountingpractices with the mean responses of greater than 3.84. Attributes of Big Data such ascompleteness, accessibility, relevance and consistency (while considered important by bothgroups) received different rankings. For instance, academics rank completeness as thenumber one attribute, whereas practitioners believe the reliability is the most importantattribute. This suggests that academics focus on the volume and size of Big Data, butpractitioners are more interested in the information quality when they use of Big Data inforensic accounting practices. Overall, academics feel that accessibility is significantly moreuseful than practitioners and the two groups are insignificantly different with regard theiropinion of three attributes, namely “accuracy,” “reliability” and “detailed.”

We ask two groups of respondents several questions regarding the role of Big Data andforensic accounting in organizations. The results of Table VII show that majority of bothacademics and practitioners agree that Big Data/Analytics improve forensic accountingpractices (81 and 80 per cent for academics and practitioners, respectively); data securitythreats limit the use of Big Data in forensic accounting (80 and 73 per cent, respectively) andBig Data/Analytics and forensic accounting overlap (67 and 62 per cent, respectively).Compared with academics, practitioners are significantly more optimistic thanorganizations’ Big Data/Analytics roles and forensic accounting roles are well defined(significant at 0.05 and 0.01 level). This suggests that the organizations have used the BigData/Analytics and forensic accounting functions to meet the challenges of the digital age.

6. ConclusionsThis paper examines the relevance of Big Data to forensic accounting practice andeducation. We conduct a survey of both practitioners and academics in China. The

MAJ34,3

280

Chineseacadem

ics

Chinesepractitioners

Sign

ificantly

different

Rank

Mean

response

Standard

deviation

Curriculum

content

5%1%

Rank

Mean

response

Standard

deviation

PanelA

:Towhatextentd

othefollowingaffecttheeffectivenessinforensicaccoun

ting?

13.97

0.59

AvailabilityofBig

Data

Yes

No

33.56

0.72

23.85

0.58

Analyticaltechniqu

esand

toolstosupp

ortforensic

accoun

ting

Yes

No

23.70

0.66

33.96

0.58

Availabilityofanalytical

talent

No

No

13.73

0.72

PanelB

:Towhatextentd

oyouusethefollowinginpracticingforensicaccoun

ting?

13.69

1.32

Predictiv

eanalytics

Yes

No

53.43

0.82

23.31

1.28

Descriptiv

eanalytics

Yes

No

33.55

0.81

33.24

1.34

Capturingdata

No

No

23.56

0.86

43.21

1.18

Prescriptiv

eanalytics

Yes

No

63.38

0.82

53.19

1.40

Agg

regatin

g/integratingdata

No

No

43.53

0.91

63.03

1.24

Disseminatingdata

No

No

13.61

0.77

PanelC

:Towhatextentd

othefollowingattributes

ofBigDataisusefulin

forensicaccoun

tingpractices?

5%1%

14.07

0.77

Completeness

Yes

Yes

43.83

0.50

24.05

0.84

Accuracy

No

No

23.86

0.57

34.03

0.94

Reliability

No

No

13.92

0.67

44.02

0.81

Accessibility

Yes

Yes

63.83

0.50

53.94

0.86

Relevancy

Yes

No

33.84

0.67

63.93

0.81

Consistency

Yes

No

53.83

0.56

73.82

0.73

Detailed

No

No

73.82

0.44

Table VI.

Relevance ofbig data

281

results show that the integration of Big Data into forensic accounting educationenables students to achieve outcomes, including performing data-mining and modelingin forensic accounting investigation, acquiring advanced analytical data managementskills and extracting, transforming and leveraging data for usage in the forensicaccounting practices. Results also indicate that the demand for and interest in bothforensic accounting and Big Data is expected to increase; forensic accounting and BigData should be offered at both the undergraduate and graduate levels; and many of thesuggested 25 Big Data topics are considered important to be integrated into Big Data/forensic accounting courses. Both academics and practitioners viewed a majority of thesetopics, 25 Big Data and data analytics topics (Table V) as important for integration intoforensic accounting education. The convergence of these Big Data related forensicaccounting topics requires the classification of interrelated topics into smaller subsets ortiers. For example, the top ranked topics of 1-10 can be labeled as “fundamentals of BigData for forensic accounting” and be taught in a separate module. The ranked of topics11-20 can be covered in a module as “data analytics and techniques,” whereas topics 21-25 can be considered as “Big Data application in forensic accounting.” Coverage of theseBig Data topics in the forensic accounting curricula should enable students tosuccessfully prepare for their future professional career.

The survey reports that both groups of respondents believe the use of Big Data/analyticsimprove the forensic accounting practices, but data security threats limit the use of BigData/analytics. This is consistent with EY (2016) that the use of Big Data/analyticsincreased dramatically, but cyber-security remains an important issue in the use of BigData/analytics. With regards to forensic accounting practices, the survey finds thatBig Data techniques including descriptive and prescriptive analytics, capturing data andBig Data attributes such as accuracy, reliability, accessibility and consistency are importantto forensic accounting practices.

This research has a few limitations. First, this paper does not address the relevance of BigData topics at undergraduate or graduate (MS andMBA) levels. Gupta et al. (2015) argue that BigData Analytics courses are different at undergraduate, MS andMBA levels where undergraduatecourses emphasize an understanding of business information (BI) tools, while graduate coursesemphasize BI applications. Second, the 25 Big Data topics reported in Table V come from theextensive review of related literature (Gupta et al., 2015; Rezaee et al., 2018). It is possible that

Table VII.Big data and forensicaccounting roles inorganizations

Chinese academicsSignificantlydifferent

Chinesepractitioners

ank Yes (%) No (%) 5% 1% Rank Yes (%) No (%)

1 81 19 Have Big Data/Analytics improvedforensic accounting practices?

No No 1 80 20

2 80 20 Do data security threats limit the useof Big Data in forensic accounting?

No No 2 73 27

3 67 33 Does your organization’s Big Data/Analytics overlap with forensicaccounting?

No No 3 62 38

4 36 64 Are your organization’s forensicaccounting roles well defined?

Yes Yes 4 61 39

5 35 65 Are your organization’s Big Data/Analytics roles well defined?

Yes Yes 5 53 47

MAJ34,3

282

these topics do not represent all of the topics that should be covered regarding Big Data. The listof the suggested 25 topics is by nomeans all-inclusive and overlap can exist among them. Finally,readers should interpret the results with caution because of the sample size (95 academics and 103practitioners) and responses obtained from academics and practitioners in one country (China)that may not be representative of the global population. The surveyed practitioners could haveself-interest bias toward forensic accounting and naturally report higher demand for forensicaccounting. Thus, future research could extend our survey to several countries. Specifically, acomparison of survey results from academics and practitioners in the developed countries (e.g.the USA) with those in emerging markets (e.g. China) should be interesting to both academicsand practitioners. The nature, and speed of the process by which the business community, theaccounting profession and business schools adopt Big Data curricula design, including Big Datacourses, concentrations and programs, is another area for future research.

Notes

1. IBM (2015) defines Big Data with five attributes as volume, velocity, variety, veracity and value.

2. Authors are affiliated with universities and professional organizations in Asia.

3. Indeed, we are not aware of any prior research providing insight into the integration of Big Datatopics into the forensic accounting curriculum.

4. Insight from both academics and practitioners was sought to ensure coverage of forensicaccounting education and practice, respectively.

5. One condition that the participants agree to join the survey is that they remain anonymous. Wedistribute the survey to them through the group e-mail list. So, we will not be able to provide theprofiles of the participants.

6. The World Business Ethics Forum was set up by the Business Schools of Hong Kong BaptistUniversity and Macau University in 2006. The Forum attracted many academics from businessschools in Hong Kong, Macau and Mainland China.

7. The HKICPA has developed a Forensic Accounting Interest Group in 2011 with a membership ofmore than 500 professionals in 2016.

8. Although curriculum design and development for Big Data/forensic accounting courses ingeneral and the level of course offerings in particular are faculty decisions, inputs frompractitioners can be relevant and valuable.

9. Variances within each group are quite big, which result in statistically insignificant differencesbetween mean responses of two groups.

10. We acknowledge the limitation of the 25 Big Data topics in the conclusion.

11. Variances within each group are quite big, which result in statistically insignificant differencesbetween mean responses of two groups.

ReferencesAssociation of Certified Fraud Examiners (ACFE) (2016), “Report to the nations on occupational fraud

and abuse”, available at: www.acfe.com/rttn2016.aspx (accessed 19 August 2016).Business Intelligence Congress (2012), “BIC 3 survey”, available at: www2.commerce.virginia.edu/bic3/

content/2012%20State%20of%20BI%20FINAL.pdf (accessed 1 August 2017).Campbell, J.E. and Mutchler, J.F. (1988), “The ‘expectation gap’ and going-concern uncertainties”,

Accounting Horizons, Vol. 2 No. 1, pp. 42-49 (March).

Relevance ofbig data

283

Cao, M., Chychyla, R. and Stewart, T. (2015), “Big data analytics in financial statement audits”,Accounting Horizon, Vol. 29 No. 2, pp. 423-429.

Charted Global Management Accountant (2014), “Readying business for the big data revolution”, availableat: www.cgma.org/Resources/Reports/DownloadableDocuments/CGMA-briefing-big-data.pdf

Crumbley, L., Heitger, L. and Smith, S. (2015), Forensic and Investigative Accounting, 7th ed., CommerceClearing House, Chicago.

Cukier, N. andMayer- Schonberger, V. (2013), “The rise of big data”, ForeignAffairs (May/June), pp. 28-40.

Ernst and Young (EY) (2014), “Global forensic data analytics survey 2014”, available at: www.ey.com/gl/en/services/assurance/fraud-investigation–dispute -services/ey-global-forensic-data-analytics-survey-2014 (accessed 12 July 2017).

Ernst and Young (EY) (2016), “Global forensic data analytics survey 2016”, available at: www.ey.com/gl/en/services/assurance/fraud-investigation–dispute-services/ey-shifting-into-high-gear-mitigating-risks-and-demonstrating-returns (accessed 12 July 2017).

Forbes (2013a), “$16.1 billion big data market: 2014 prediction from IDC and IIA”, available at: www.forbes.com/sites/gilpress/2013/12/12/16-1-billion-big-data-market-2014-predictions-from-idc-and-iia/#3f285f304482 (accessed 1 August 2017).

Forbes (2013b), “How SEC’s new RoboCop profiles companies for accounting fraud”, available at:www.forbes.com/sites/janetnovack/2013/08/09/how-secs-new-robocop-profiles-companies-for-accounting-fraud/#2342b7f43226 (accessed 4 January 2017).

Gartner (2014), “IT dictionary”, available at: www.gartner.com/it-glossary/?s=big6data (accessed 31July 2017).

Global Times (2017), “Alibaba to help build china’s first big data college to offer postgraduate courses”,available at: www.globaltimes.cn/content/1048287.shtml (accessed 31 July 2017).

Gupta, B., Goul, M. and Dinter, B. (2015), “Business intelligence and big data in higher education: statusof a Multi-Year model curriculum development effort for business school undergraduates, MSgraduates and MBAs”, Communications of the Association for Information Systems, Vol. 36No. 23, pp. 449-476.

Hodge, F.D. (2003), “Investors’ perceptions of earnings quality, auditor independence, and the usefulness ofaudited financial information”,Accounting Horizons, Vol. 17 No. Supplement, pp. 37-48.

Hong Kong Institute of Certified Public Accountants (HKICPA) (2013), “Sixth long rang plan”, availableat: www.hkicpa.org.hk/file/media/section1_about_us/organization/pdf-file/6-plan.pdf (accessed22March 2016).

Hung, M., Wong, T.J. and Zhang, F. (2015), “The value of political ties versus market credibility:evidence from corporate scandals in China”, Contemporary Accounting Research, Vol. 32 No. 4,(winter) pp. 1641-1675.

IBM (2015), “Why only one of the 5Vs of big data really matters”, available at: www.ibmbigdatahub.com/blog/why-only-one-5-vs-big-data-really-matters (accessed 19 August 2016).

Issa, H. and Kogan, A. (2014), “A predictive ordered logistic regression model as a tool forquality review of control risk assessments”, Journal of Information Systems, Vol. 28 No. 2,pp. 209-229.

Kelly, J. Vellante, D. and Floyer, D. (2015), “Big data market size and vendor revenues”, available at:http://wikibon.org/wiki/v/Big_Data_Market_Size_and_Vendor_Revenues (accessed 22 March2016).

Laney, D. and Kart, L. (2012), “Defining and differentiating the role of data scientist”, Gartner, Inc,available at: http://blogs.gartner.com/doug-laney/defining-and-differentiating-the-role-of-the-data-scientist/ (accessed 25March 2012).

Markets and Markets (2017), “Big data market worth 66.79 billion USD by 2021”, available at: www.marketsandmarkets.com/PressReleases/big-data.asp (accessed 5 July 2017).

MAJ34,3

284

McKinsey Global Institute (2011), “Big data: the next frontier for innovation, competition, andproductivity”, available at: www.mckinsey.com/business-functions/digital-mckinsey/our-insights/big-data-the-next-frontier-for-innovation (accessed 5 July 2017).

New Vantage Partner (2016), “Big data executive survey 2016”, available at: http://newvantage.com/wp-content/uploads/2016/01/Big-Data-Executive-Survey-2016-Findings-FINAL.pdf (accessed 1August 2017).

Rezaee, Z. and Burton, J. (1997), “Forensic accounting education: insights from academicians andcertified fraud examiner practitioners”,Managerial Auditing Journal, Vol. 12 No. 9, pp. 479-489.

Rezaee, Z., Larry Crumbley, D. and Elmore, R.C. (2004), “Forensic accounting education: a survey ofacademicians and practitioners”,Advances in Accounting Education, Vol. 6, pp. 193-231.

Rezaee, Z., Lo, D., Ha, M. and Suen, A. (2016), “Forensic accounting education and practice: insightsfrom China”, Journal of Investigative and Forensic Accounting, Vol. 8 No. 1, pp. 106-119.

Rezaee, Z., Wang, J. and Lam, B. (2018), “Toward the integration of big data into forensic accountingeducation”, Journal of Forensic and Investigative Accounting, Vol. 10 No. 1, (January-June 2018).

Russom, P. (2011), Big Data analytics. The Data Warehousing Institute (TDW1) Best PracticesReport (Fourth Quarter) available at: https://tdwi.org/research/2011/09/�/media/TDWI/TDWI/Research/BPR/2011/TDWI_BPReport_Q411_Big_Data_Analytics_Web/TDWI_BPReport_Q411_Big%20Data_ExecSummary.ashx (accessed 5 July, 2017).

Seda, M. and Peterson-Kramer, B.K. (2014), “An examination of the availability and composition offorensic accounting education in the United States and other countries”, Journal of Forensic andInvestigative Accounting, Vol. 6 No. 1, pp. 1-46.

The Association to Advance Collegiate Schools of Business (AACSB) International (2014), “Accountingaccreditation standard A7: information technology skills and knowledge for accountinggraduates: An interpretation”, available at: www.aacsb.edu/�/media/AACSB/Publications/white-papers/accounting-accreditation-standard-7.ashx (accessed 19 August 2016).

Wang, J., Lee, G. and Crumbley, L. (2016), “Current availability of forensic accounting education andstate of forensic accounting services in Hong Kong and mainland China”, Journal of Forensicand Investigative Accounting, Vol. 8 No. 3, pp. 515-534.

Wixom, B., Ariyachandra, T., Douglas, D.E., Goul, M., Gupta, B., Iyer, L., Kulkarni, U.R., Mooney, J.G.,Phillips-Wren, G.E. and Taretken, O. (2014), “The current state of business intelligence inacademia: the arrival of big data”, Communications of the Association for Information Systems,Vol. 34, pp. 1-34.

Corresponding authorZabihollah Rezaee can be contacted at: zrezaee@memphis.edu

Relevance ofbig data

285

AppendixMAJ34,3

286

Relevance ofbig data

287

MAJ34,3

288

Executive compensation andcompensation risk: evidence

from technology firmsPaul Dunn, Zhongzhi He and Samir Trabelsi

Goodman School of Business, Brock University, Saint Catharine’s, Canada, and

Zhimin (Jimmy) YuDepartment of Accounting, University of Houston-Downtown,

Houston, Texas, USA

AbstractPurpose – The purpose of this research is to investigate factors that contribute to technology firms payinghigher compensation than non-technology firms, and why the mix of compensation at technology firms isdifferent than the compensation packages at non-technology firms.Design/methodology/approach – This research used a sample of 1,009 firm-year observations for thefive-year period from 2001 to 2005 and random-effects regression models.Findings – It was found that the total compensation paid to the CEOs of technology firms is higher than thetotal compensation paid to the CEOs of non-technology firms, and that the value of the stock options grantedto the former is greater than the value of the stock options granted to the latter.Research limitations/implications – The results are largely consistent with the labour marketefficiency perspective. The higher compensation paid to CEOs in technology firms seems to be commensuratewith the higher compensation risk that CEOs in technology firms bear.Practical implications – Compensation designers should consider both the benefits and costs of grantingstock and stock options to executives. An increased portion of stock options definitely aligns the interests ofshareholders and CEOs together, and could maximize the retentive effect if CEOs have a significant amount oftheir wealth in unvested in-the-money options.Social implications – Consistent with the literature, a CEO could earn much higher pay if he or she alsoserves as the chair of the board of directors. Practically, firms do not require all governance mechanisms.They just require one set of suitable governance mechanisms.Originality/value – This paper is the first to investigate factors that contribute to technology firms payinghigher compensation than non-technology firms, and that do explain why the mix of compensation attechnology firms is different than the compensation packages at non-technology firms.

Keywords Executive compensation, Compensation risk, Technology firms

Paper type Research paper

IntroductionPrevious research (Conyon and Murphy, 2000; Conyon et al., 2011; Fernandes et al., 2012)reveals that American firms tend to pay their managers more than non-American firms,both with respect to direct compensation and incentive pay. Furthermore, within the USA,

JEL classification – J33, J44, M12, M52The authors thank Amin Mawani for his insightful feedback and guidance. They also acknowledgevaluable comments from Steve Balsam, Jeffrey Callen, Fabrizio Ferri and workshop participants atBrock University and Manouba University.

Evidence fromtechnology

firms

289

Received 27 October 2017Revised 10 July 2018

Accepted 28 July 2018

Managerial Auditing JournalVol. 34 No. 3, 2019

pp. 289-304© EmeraldPublishingLimited

0268-6902DOI 10.1108/MAJ-10-2017-1687

The current issue and full text archive of this journal is available on Emerald Insight at:www.emeraldinsight.com/0268-6902.htm

technology firms tend to pay their CEOs more than the CEOs of non-technology firms, bothwith respect to direct compensation and incentive pay (Ittner et al., 2003).

The purpose of this research is to investigate factors that contribute to technology firmspaying higher compensation than non-technology firms, and why the mix of compensationat technology firms is different than the compensation packages at non-technology firms.

Fernandes et al. (2012) compare the pay of 2,543 CEOs at firms in 27 countries. They findthat the compensation paid to the CEOs of US firms is significantly higher than thecompensation paid to their foreign counterparts. Furthermore, a greater portion of theAmericans’ pay is in the form of stock and stock options. Similarly, Conyon and Murphy(2000) find that US CEOs receive significantly higher pay as well as higher incentives ofstock and stock options than the CEOs of UK firms, even controlling for firm and industrycharacteristics. Within the USA, Ittner et al. (2003) document that the CEOs of technologyfirms are paid significantly more than the CEOs of all other firms, and that the technologyfirms remunerate their CEOs with significantly greater levels of stock and stock optionsthan non-technology firms. They find that technology firms use these incentive pay devicesto attract and retain employees.

Part of the explanation for compensation differences has to do with risk premium(Pratt, 1964). Risk premium occurs when part of an executive’s pay package includesreceiving employee stock options. Because a manager cannot sell or short employee stockoptions, the risk associated with these options is non-diversifiable. Consequently, from themanager’s point of view, the utility of holding $1m of stock options is less than the utility ofreceiving a $1m in cash.

American CEOs receive a large portion of their pay in the form of employee stock options(Conyon et al., 2011; Fernandes et al., 2012; Ittner et al., 2003). Conyon and Murphy (2000)find that the difference between the pay of American and British CEOs is attributable to thehigher risk premium associated with the former’s compensation mix (the value of theemployee stock options as a percentage of total compensation). After controlling forindustry and managerial characteristics, they find that the difference in the pay levelsbetween these two countries is because of the risk premium of being paid in the form of non-diversifiable employee stock options.

So, if risk premium helps to explains whyAmericanmanagers are paid more than Britishmanagers, can risk premium also help to explain why the managers of technology firms arepaid more than the managers of non-technology firms? Using a sample of 1,009 firm-yearobservations for the period 2001-2005, we find that the total compensation paid to the CEOsof technology firms is higher than the total compensation paid to the CEOs of non-technology firms, and that the value of the stock options granted to the former CEOs isgreater than the value of the stock options granted to the latter CEOs. We argue that thisdifference can be explained by the risk premium that technology CEOs have in theircompensation packages.

The remainder of the paper is laid out as follows. The next section reviews thecompensation literature and develops our testable hypothesis. The third section describesthe methodology and the fourth section the results of the study. The final section concludesthe paper.

Executive compensation and compensation riskAgency theory predicts that incentive plans are used to align the interests of the firm’semployees with the interest of the firm’s shareholders (Eisenhardt, 1989). The theoryassumes that CEOs are risk-averse and that shareholders are risk-neutral. As such,shareholders want management to accept all of the risky projects that have a positive net

MAJ34,3

290

present value, whereas management, on the other hand, would not naturally take on anyrisky projects. To align the interest of management with those of the shareholders,shareholders write compensation contracts with incentives that encourage management toaccept a certain level of risk. The inventive is in the form of stock options. Options increasein value as the firm’s stock increases in value, and the firm’s stock price increases as a resultof the firm successfully investing in risky projects. However, given that managers are risk-averse, they will demand higher compensation to offset the risk associated with having aportion of their pay in the form of employee stock options.

Why do CEOs demand higher pay if a portion of their pay is in the form of stock options?The answer is that CEOs cannot diversify a firm-specific portfolio of employee stockoptions.

Stock market theory predicts that a diversified portfolio reduces risk (Markowitz, 1952).However, employee stock options are not diversifiable. There is no market for employeestock options. As a result, there is a much higher risk to a CEO in holding a non-diversifiableportfolio of employee stock options than there is in holding a portfolio of publicly tradedstock options. Because of the greater risk, there is, therefore, a lower utility associated withbeing paid $1m in stock options than $1m in either cash or diversifiable stocks (Conyonet al., 2011; Hall and Murphy, 2002). Consequently, the less their wealth is diversified, themore risk CEOs bear, and the greater the pay that they demand to be compensated for therisks embedded in holding employee stock options.

Pratt (1964) develops a utility function of consumption to explain risk-aversion and riskpremium. He predicts that a decision-maker with relative risk-aversion will demand greaterpay to compensate the risks that he or she takes. A risk-averse decision-maker discounts thevalue of the consumable assets that are at risk. This means that at every level of risk, theestimated cash equivalent (the amount for which the risky asset could be exchanged) is lessthan the expected value of the risky assets. Pratt (1964) defines “risk premium” as theexpected monetary value of the asset minus its cash equivalent. He concludes that a risk-averse decision-maker always has a risk premium greater than zero.

The market value of stock options is calculated on the assumption that investors candiverse their investment portfolios to reduce risk. But top executives cannot sell nor shorttheir employee stock options. Because they cannot diverse their risks if they hold employeestock options, managers evaluate the value of their compensation packages at prices lessthan the market value of publicly traded stock options. Lambert et al. (1991) develop a modelto calculate the value of employee stock options. They conclude that the higher the portionof the manager’s wealth that is tied to the company’s stock price, the higher the relative risk-aversion the manager has, and consequently the less perceived value of the stock option tothe manager.

Conyon et al. (2011) suggest that compensation has two aspects. The first is “risk-adjusted pay”. This is the amount of compensation that the CEO receives based on thatindividual’s effort and ability. The other part of executive compensation is Pratt’s (1964)notion of “risk premium”. This is the amount of compensation that is tied to the fact that it isgiven in the form of non-diversifiable employee stock options. Risk premium occurswhenever a compensation package requires the CEO to hold the firm’s stock and/or stockoptions. As the CEO cannot sell these equities and diversify his or her portfolio to reducerisk, he or she demands additional compensation:

Another way to think of this risk premium is: How much less pay would the CEO accept if hewere released from the restriction that he holds a substantial fraction of his wealth in firm stock?(Conyon et al., 2011, p. 417).

Evidence fromtechnology

firms

291

Conyon et al. (2011) assume that a CEO is indifferent between receiving either risk-adjustedpay or risk premium pay, as long as the expected values of the cash compensation and theincentive-based compensation are the same. Given that the expected value of risk premiumpay is lower than the value of risk-adjusted pay, the total compensation of a CEO willincrease if a greater portion of the CEO’s pay is in the form of risk premium incentive pay.

The empirical evidence (Ittner et al., 2003) indicates the compensation mix that the CEOsof technology firms receive is different than the compensation mix of non-technology CEOs.The CEOs of the former are paid more than the other CEOs, and they are granted moreincentive pay, in the form of stock and stock options, than the CEOs of non-technologyfirms. So, it could be that this difference in compensation between technology and otherfirms is because of the compensation mix. In other words, the higher compensation paid tothe CEOs at technology firms is associated with their inability to diversify the risk ofholding employee stock options. As a result, our hypothesis is the following:

H1. Ceteris paribus, differences in CEO compensation between technology and non-technology firms are associated with the mix of their compensation packages.

MethodologyRegression modelThe following panel regression is used to study the composition of executive compensation:

Compensation ¼ a þ b 1 CompensationMixð Þ þ b 2 Technologyð Þþ b 3 CompensationMix *Technologyð Þ þ b 4 controlsð Þ

The definitions of the variables are contained in the Appendix.Dependent variable. The dependent variable, Compensation, is the logarithm of total

CEO composition. Total compensation is the sum of salary, bonus, other annualcompensation, restricted stock grants, value of options (Black–Scholes value), long termincentive plan and all other compensation.

Independent variables. The continuous variable, Compensation Mix, is the proportion ofstock-based compensation to total compensation. Stock-based compensation is the totalvalue of employee stock options and restricted stock awarded to the CEO. Totalcompensation is as defined above.

The dummy variable, Technology, is coded 1 if the firm is in the technology sector and 0otherwise. Consistent with Murphy (2003), the technology sector includes firms withprimary SIC codes 3570, 3571, 3572, 3576, 3577, 3661, 3674, 4812, 4813, 5045, 5961, 7370,7371, 7372 and 7373.

The dummy variable, Compensation Mix � Technology, is the interaction of theCompensation mix variable and the Technology variable.

Control variables. Consistent with previous research (Gabaix and Landier, 2008; Smith andWatts, 1992), the model controls for both size and firm performance. Net sales, net income andmarket value are the proxies for size. Return on assets (ROA) and stock market returns, bothone-year and five-year, control for performance. Stock return is calculated using the dividendreinvestmentmethod.

Following Balsam et al. (2011), Bebchuk et al. (2010) and Core et al. (1999), there are sixcorporate governance variables to control for CEO power over the board as a whole as wellas the CEO’s influence over the compensation, audit and governance committees. CEO/chairduality is an indicator variable, equal to 1 if the CEO serves as chair of the board of directors,

MAJ34,3

292

and 0 otherwise. As the compensation committee sets the CEOs compensation package, thetwo control variables, composition chair and composition member, are coded 1 if the CEO is,respectively, the chair or a member of the composition committee, and 0 otherwise.Similarly, because of the importance of the audit committee for good governance, the twocontrol variables, audit chair and audit member, are coded 1 if the CEO is, respectively, thechair or a member of the audit committee, and 0 otherwise. Finally, the control variable,governance member, is coded 1 if the CEO is a member of the corporate governancecommittee and 0 otherwise.

The model also controls for institutional ownership, which is measured as the percent ofshares held by institutional investors.

SampleTable I summarizes our sample selection procedure. We collected compensation data for USfirms from the Compustat ExecuComp database, which includes the S&P 500, S&P MidCap400 and S&P Small Cap 600. There are 8,513 observations in the database for the period2001-2005. Financial information, including net sales, net income, ROA, market value, one-year return on equity and five-year return on equity, is drawn from Compustat. Corporategovernance data are retrieved from Risk Metrics. After eliminating observations withmissing data, the final sample consists of 1,009 firm-year observations.

Descriptive statisticsTable II Panel A shows that sales and net incomes in technology firms are significantlylower than those in non-technology firms. The market values of equity in technology firmsare higher than those in non-technology firms. The difference is statistically significant atthe 10 per cent level. The one-year stock returns to shareholders are not statisticallysignificantly different between technology firms and non-technology firms. Finally, the five-year stock returns to shareholders in technology firms are significantly lower than those innon-technology firms.

Table II Panel B documents that salary, bonus, restricted stock grants, option grants andtotal compensation in technology firms differ significantly from those in non-technologyfirms. The mean of salaries in technology firms is $543,000, which is significantly lowerthan that in non-technology firms at $712,000. The mean bonus in technology firms is$597,000, which is significantly lower than the mean bonus in non-technology firms at$945,000. The mean value of the stocks in technology firms is $568,000, which issignificantly lower than that in non-technology firms at $752,000. Most importantly,however, the mean of the options in technology firms is $4,941,000, which is significantlyhigher than that in non-technology firms at $2,027,000. The higher stock options in

Table I.Sample selection

Compensation years 2001-2005

Total CEO observations in ExecuComp 8,513No missing value for net sales and net income in Compustat 8,471No missing value for return on assets 8,470No missing value for market value 8,231No missing value for five-year return 7,814No missing value for one-year return 8,394No missing value for governance attributes in Risk Metrics 1,009Final sample 1,009

Evidence fromtechnology

firms

293

technology firms cause the higher total compensations in technology firms at $6,956,000,which is significantly higher than that in non-technology firms at $4,964,000. The results areconsistent with the findings of Ittner et al. (2003). The CEOs in technology firms earn higherpay than those in non-technology firms.

Table II Panel C shows that, for each year, the percent of stock-based pay in technologyfirms is significantly higher than that in non-technology firms. The results support thenotion that CEOs in technology firms earn higher pay and that they also take much highercompensation risks embedded in stock-based compensation.

Table III provides Pearson correlations for the variables. These correlationsappear to be reasonable. For example, the correlation between Compensation Mix andTechnology (which is an indicator that equals 1 for companies in the technologysector) is 0.162, and this correlation is highly significant. This suggests that the CEOsof companies in the technology sector are more likely to incur higher compensationrisk. The logarithm of market value of equity is significantly associated to thelogarithm of sales, and net income. Hence, we only use logarithm of market value as aproxy for size.

Table II.Descriptive statistics

Technology firms Non-technology firmsMean Median SD Mean Median SD

Panel A: Economic determinantsNet sales (thousands) 3,519*** 494*** 10,437 5,382 1,383 15,519Net income (thousands) 122*** 15 1,925 278 66 1,939Market value (thousands) 8,447* 1,137 29,169 6,889 1,548 21,631Return on assets (%) 2.2*** 2 11 1.7 3 3One-year return to shareholders (%) 291 0 6,828 131 12 6,410Five-year return to shareholders (%) �1*** �1*** 24 9 8 20

Panel B: Compensation (in thousands of dollars)Salary 543*** 459*** 346 712 656 379Bonus 597*** 228*** 1,553 945 499 1,730Restricted stock 568*** 0 3,082 752 0 2,582Options granted 4,941*** 1,436 16,186 2,027 697 4,412Total compensation 6,956*** 2,809 17,095 4,964 2,770 6,998

Panel C: Compensation mix2001 0.60*** 0.76*** 0.37 0.45 0.49 0.312002 0.57*** 0.68*** 0.33 0.42 0.46 0.292003 0.51*** 0.60*** 0.34 0.39 0.41 0.282004 0.54*** 0.62*** 0.31 0.40 0.42 0.282005 0.51*** 0.59*** 0.30 0.40 0.43 0.27

Panel D: Corporate governanceCEO/Chair duality 0.61 1 0.495 0.68*** 1 0.467Compensation committee chair 0.05 0 0.225 0.05 0 0.226Compensation committee member 0.23 0 0.422 0.22 0 0.415Audit committee chair 0.03 0 0.162 0.04 0 0.196Audit committee member 0.18 0 0.384 0.20 0 0.399Governance committee member 0.15 0 0.359 0.17 0 0.373Institutional holdings 0.003 0 0.015 0.028 0 0.55

Notes: ***, **, and * are statistically significant at the 0.01, 0.05 and 0.10 levels, respectively, under thetwo-tailed test; The variables are defined in the Appendix

MAJ34,3

294

12

34

56

78

1Co

mpensation

1.000

2Co

mpensationmix

0.576***

1.000

3Techn

ology

�0.001

0.162***

1.000

4CE

O/Chairdu

ality

0.144***

0.010

�0.076***

1.000

5Co

mpensationcommittee

chair

�0.002

�0.010

�0.001

�0.031***

1.000

6Co

mpensationcommittee

mem

ber

0.030**

0.047***

0.008

�0.076***

0.272***

1.000

7Aud

itcommittee

chair

�0.005

�0.004

�0.023

�0.061***

�0.074***

0.008**

8Aud

itcommittee

mem

ber

0.007

0.022

�0.015

�0.115***

�0.014***

0.041***

0.266***

1.000

9Governancecommittee

mem

ber

0.079***

0.025*

�0.013

�0.077***

0.095***

0.158***

0.017***

0.031***

10Institu

tionalholding

�0.009

0.018

�0.013

�0.008

0.002

0.002

0.002

0.001

11Marketv

alue

0.640***

0.252***

�0.060***

0.149***

0.010

0.044***

0.002

0.028**

12Netsales

0.575***

0.109***

�0.189***

0.168***

0.009

0.049***

0.023*

0.045***

13Netincome

0.601***

0.173***

�0.086***

0.151***

�0.004

0.038***

0.004

0.028*

14Five-years

tock

return

0.152***

0.060***

�0.160***

0.000

�0.005

�0.041***

�0.020

�0.001

15One-years

tock

return

0.021*

0.018*

0.009

�0.032**

0.029**

�0.011

0.001

0.011

16ROA

0.080***

�0.009

�0.079***

0.005

0.031**

0.013

0.011

0.022*

Notes

:***

,**and*arestatisticallysign

ificant

atthe0.01,0.05and0.10

levels,respectively,un

derthe

two-tailedtest;T

hevaria

bles

aredefinedintheAppendix

(contin

ued)

Table III.Pearson correlation

matrix

Evidence fromtechnology

firms

295

910

1112

1314

1516

1Co

mpensation

2Co

mpensationmix

3Techn

ology

4CE

O/Chairdu

ality

5Co

mpensationcommittee

chair

6Co

mpensationcommittee

mem

ber

7Aud

itcommittee

chair

8Aud

itcommittee

mem

ber

9Governancecommittee

mem

ber

1.000

10Institu

tionalholding

�0.010

1.000

11Marketv

alue

0.125***

�0.055*

1.000

12Netsales

0.125***

�0.130***

0.784***

1.000

13Netincome

0.105***

0.068**

0.887***

0.794***

14Five-yearstock

return

�0.046***

�0.078**

0.265***

0.046***

0.040***

1.00

15One-yearstock

return

0.035***

�0.051*

�0.006

0.008

0.037**

�0.011

1.000

16ROA

0.034**

�0.111***

0.191***

0.208***

0.100***

0.217***

0.174***

1.000

Table III.

MAJ34,3

296

Empirical resultsPrimary findingsTable IV presents the results for the panel regressions that test for the differences in CEOcompensation between technology and non-technology firms. The dependent variable is thelog of total compensation. The basic model is contained in Column I. Of the two controlvariables, size (proxied by the logarithm of market value) is statistically significant, whereasperformance (proxied by the five-year stock return to shareholders) is not. This is consistentwith Gabaix and Landier (2008), who find that the CEOs of large firms receive higher levelsof compensation than the CEOs of small firms. The variable of interest, Technology, ispositive and statistically significant. This indicates that the CEOs of technology firms arepaid more than the CEOs of non-technology firms.

Next, compensation mix is added to the model. The results are contained in Column II.The coefficients of all the variables, except for the year dummy variables, are statisticallysignificant. These results indicate that total compensation is associated withcompensation mix after controlling for firm size and performance. The regression modelexplains a reasonable amount of the cross-sectional variation, as evidenced by theadjusted R2 of 60.2 per cent. More importantly, the R2 increases substantially (from 43.2per cent in the first model) when compensation mix is added as an independent variable.This indicates that compensation risk explains a large percent of the variation inexecutive compensation.

Column III contains the results of adding the interaction term, Compensation Mix �Technology, to the model. Once again the coefficients of all the variables, with the exceptionof the year dummies, are statistically significant. The results show that compensation ispositively associated with both Compensation Mix and Compensation Mix � Technology.This shows that CEOs in technology firms are paid more than the CEOs in non-technologyfirms after controlling for firm size and profitability. As compensation mix is the proxy forcompensation risk, the results suggest that the CEOs of technology firms are given morecompensation because they have higher compensation risk.

We conclude that compensation risk explains the differences in pay between technologyand non-technology firms. The association between total compensation and compensationmix is positive and statistically significant. The magnitude of the coefficient suggests that a10 per cent increase in the percentage of compensation mix (i.e. stock-based compensation to

Table IV.Results from

regressing totalcompensation on

compensation risk

I II III

Market value 0.66 0*** (73.680) 0.44 0*** (68.680) 0.540*** (69.570)Five-year return �0.014 (�1.487) �0.020*** (�2.66) �0.020 *** (�3.000)Technology 0.034 *** (3.916) �0.043 *** (�5.727) �0.18 *** (�13.32)Compensation mix 0.440 *** (57.090) 0.40 *** (47.450)Compensation mix� Technology 0.180 *** (12.110)Years No effect No effect No effectObservations 1,009 1,009 1,009Adj-R2 0.432 0.602 0.610

Notes: ***, ** and * are statistically significant at the 0.01, 0.05 and 0.10 levels, respectively, under two-tailed test. The coefficient is reported on the first line and the t-statistic in the parentheses. The coefficientestimates are based on firm clustered standard errors (Petersen, 2009); the variables are defined in theAppendix

Evidence fromtechnology

firms

297

total compensation) translates into a 4.4 per cent increase in total compensation. This resultis consistent with Fernandes et al. (2012), who find that total compensation is significantlyassociated with compensation mix. Using a sample of 2,543 firms from 27 countries, theirresults show that a 10 per cent increase in the percentage of compensation mix translatesinto a 20 per cent increase in total compensation.

Board characteristics and ownership structureTo check the robustness of our results, we examined whether total compensation isassociated with board structure and ownership structure. Board structure includesindicators for CEO/chair duality, CEO compensation committee chair, CEO compensationcommittee member, CEO audit committee chair, CEO audit committee member and CEOgovernance committee member. Ownership structure was measured as the percentage ofshares held by institutional investors.

Table V reports the results. With the exception of CEO/chair duality, none of the boardstructure variables is statistically significant. Also, none of the interaction terms of boardstructure variables and technology firms is statistically significant. This indicates that, forthis sample of firms, CEO membership on the powerful board committees of compensation,audit and governance does not affect CEO remuneration. This applies both to the technologyand the non-technology firms.

However, compensation does increase when the CEO is the chair of the board ofdirectors. Core et al. (1999) argue that agency costs and managerial power are higher whenthe CEO is also the chair of the board of directors. Our results support this position. CEO/chairs, who have high levels of power, appear to be unaffected by committee-levelgovernance structures when it comes to determining the amount of their compensation.

The coefficient on institutional holdings is not statistically significant. We do not findany evidence that total compensation is associated with the shareholding of institutionalinvestors. This result is different from Fernandes et al. (2012), who found that totalcompensation was significantly associated with institutional investors. This differencecould be due to two factors. Fernandes et al. (2012) had a sample of international firms,whereas ours is restricted to US firms only. Second, they selected their sample from theLionShare database and ours was drawn from ExecuComp. So, it is possible thatinstitutional investors play a more important role in determining CEO compensation indifferent countries and/or in different market sectors.

Sensitivity testsTotal compensationIn our models, compensation has been measured using the TDC1 number in the ExecuCompdatabase. TDC1 is the total compensation paid to the executive, comprising salary, bonuses,restricted stock options, exercised stock options and all other forms of compensation. TDC2,in the ExecuComp database, is the total compensation realized during the year, includingany backdated stock options. Kaplan and Rauh (2010) state that TDC1 represents thecompensation authorized by the board of directors, and TDC2 approximates the executive’sgross income.

To test the sensitivity of the compensation number, we substituted TDC2 for TDC1.Table VI Column I shows that the logarithm of TDC2 is significantly associated withcompensation mix. This is consistent with the results of Table IV Column I. Both show thatthe level of compensation received by a CEO is dependent on the percentage of stock optionscontained in the executive’s compensation mix.

MAJ34,3

298

III

III

IV

Marketv

alue

0.627***

(23.679)

0.619***

(22.729)

0.614***

(22.335)

0.452(20.296)

Five-yearreturn

0.014(0.545)

0.015(0.562)

0.015(0.551)

�0.032

(�1.529)

Techn

ology

0.114***

(4.636)

0.114***

(4.631)

0.089*

(1.941)

�0.173***(�

3.564)

CEO/Chairdu

ality

0.056**(2.216)

0.051*

(1.868)

0.052**(2.504)

Compensationcommittee

chair

�0.008

(�0.311)

�0.014

(�0.518)

�0.014

(�0.636)

Compensationcommittee

mem

ber

�0.005

(�0.199)

�0.006

(�0.195)

�0.029

(�1.321)

Aud

itcommittee

chair

�0.024

(�0.880)

�0.013

(�0.468)

�0.009

(�0.418)

Aud

itcommittee

mem

ber

0.033(1.199)

0.020(0.680)

0.027(1.157)

Governancecommittee

mem

ber

0.009(0.344)

0.016(0.608)

0.011(0.053)

Institu

tionalholding

s0.028(1.132)

0.027(1.110)

0.003(0.162)

Techn

ology�CE

O/Chairdu

ality

0.017(0.399)

0.041(1.258)

Techn

ology�Co

mpensationcommittee

chair

0.022(0.793)

0.026(1.193)

Techn

ology�Co

mpensationcommittee

mem

ber

�0.003

(�0.092)

0.011(0.434)

Techn

ology�Aud

itcommittee

chair

�0.025

(�0.882)

0.012(0.534)

Techn

ology�Aud

itcommittee

mem

ber

0.033(0.963)

�0.014

(�0.515)

Techn

ology�Governancecommittee

mem

ber

�0.021

(�0.788)

�0.003

(�0.162)

Techn

ology�Institu

tionalholding

s0.022(0.841)

0.009(0.457)

Compensationmix

0.489***

(21.372)

Compensationmix

�Techn

ology

0.181***

(4.335)

Years

Noeffect

Noeffect

Noeffect

Noeffect

Observatio

ns1,009

1,009

1,009

1,009

Adj-R

20.412

0.412

0.410

0.651

Notes

:***

,**and* a

restatistically

sign

ificant

atthe0.01,0.05and0.10

levels,respectively,un

derthetw

o-tailedtest.T

hecoefficientisreported

onthefirstline

andthet-statistic

intheparentheses.The

coefficientestim

ates

arebasedon

firm

clusteredstandard

errors

(Petersen,

2009);thevariablesaredefinedin

the

App

endix

Table V.Results from

regressing totalcompensation on

compensation riskand board structures

and ownershipstructures

Evidence fromtechnology

firms

299

III

III

IVV

Compensation

TDC2

TDC1

TDC1

TDC1

TDC1

Techn

ology

�0.067***(�

3.865)

�0.090***(�

6.680)

�0.134***(�

8.475)

�0.193***(�

14.588)

�0.190***(�

14.364)

Compensationmix

0.071***

(6.730)

0.467***

(58.940)

0.444***

(49.670)

0.405***

(49.940)

0.407***

(50.280)

Compensationmix

�Techn

ology

�0.001

(�0.047)

0.153***

(10.828)

0.162***

(9.852)

0.189***

(13.396)

0.184***

(13.11)

Marketv

alue

0.532***

(54.560)

0.525***

(72.100)

0.528***

(71.400)

Logsales

0.525***

(73.050)

Lognetincom

e0.522***

(63.4)

Five-yearreturn

0.096***(10.030)

0.098***

(13.580)

0.022***

(2.700)

One-yearreturn

0.018***

(2.570)

ROA

�0.019***(�

2.628)

Years

Yes

Yes

Yes

Yes

Yes

Observatio

ns1,009

1,009

1,009

1,009

1,009

Adj-R

20.380

0.619

0.601

0.605

0.605

Notes

:***

,**and*arestatistically

sign

ificant

atthe0.01,0.05and0.10

levels,respectively,un

derthetw

o-tailedtest.T

hecoefficientisreported

onthefirstline

andthet-statistic

intheparentheses.The

coefficientestim

ates

arebasedon

firm

clusteredstandard

errors

(Petersen,

2009);thevariablesaredefinedin

the

App

endix

Table VI.Results fromregressing totalcompensation(TDC2) oncompensation mixand technologyresults fromregressing totalcompensation(TDC1) oncompensation mixand technology byusing alternativecontrolled variables

MAJ34,3

300

Firm sizeConsistent with Balsam et al. (2011), we use the logarithm of sales instead of the logarithm ofmarket value as a proxy for firm size. Table VI Column II shows the results of thisalternative specification. The regression model is well specified and explains 61.9 per cent ofcross-sectional variation in executive compensation. The coefficient of the CompensationMix variable retains the same sign and significance. So, total compensation is stillsignificantly associated with the compensation mix.

Firm performanceFollowing Gabaix and Landier (2008), we use the logarithm of net income as the proxy forfirm performance rather than the market value of equity. The results of this specification arepresented in Table VI Column III. With an adjusted R2 of 60.1 per cent, the explanatorypower of this model is almost the same as the previous model. Once again, the coefficient ofthe Compensation Mix variable remains positive and statistically significant. Totalcompensation is still significantly associated with the composition of the executive’scompensation package.

We also used a one-year return to shareholders instead of a five-year return as a proxyfor firm performance, as suggested by Bebchuk et al. (2010). Table VI Column IV showsresults similar with the previous models. Finally, we used ROA, an accounting measure ofperformance, as the proxy for firm performance, as per Balsam et al. (2011). Column V ofTable VI shows similar results again. Overall, in all of these models, total compensation ispositively associated with the compensation mix.

Change analysisThe previous results, that use a levels-based model, show that total compensation ispositively associated with compensation mix. To exclude the possibility ofmisspecification in the levels-based model because of omitted variables andendogeneity problem (Lu et al., 2011; Klassen and Mawani, 2000), we regress thechange of the logarithm of total compensation on the change of compensation risk.The control variables include the change of five-year return, the change of marketvalue, a technology dummy and year dummies. Table VII shows the results. Theregression model explains slightly over half of the cross-sectional variation in totalcompensation. In this alternative specification, the coefficient on the change of thecompensation mix variable is 0.68. The magnitude of this suggests that a 10 per cent

Table VII.Results from

regressing change oftotal compensation

on change ofcompensation mix

Market value 0.167*** (13.908)Five-year return 0.008 (0.632)Technology �0.210*** (-2.246)Compensation mix 0.683*** (73.020)Compensation mix� Technology 0.481*** (4.135)Years No effectObservations 1,009Adj-R2 0.504

Notes ***, ** and * are statistically significant at the 0.01, 0.05 and 0.10 levels, respectively, under thetwo-tailed test. The coefficient is reported on the first line and the t-statistic in the parentheses. Thecoefficient estimates are based on firm clustered standard errors (Petersen, 2009); the variables are definedin the Appendix

Evidence fromtechnology

firms

301

increase in the change of the compensation mix translates into a 6.8 per cent increaseof the change of the executive’s total compensation. Overall, a comparison of theresults of the levels-based regression and the change-based regression suggests thatthe results of the levels-based regression reported in Tables IV-VI are robust topotential problems associated with omitted correlated variables and/or endogeneityconcerns.

Summary and concluding remarksWe present evidence of a direct link between CEO compensation and compensation risk.The CEO total compensation in technology firms is higher than that in non-technology firmsbecause technology firms impose greater risk on their CEOs by giving them a greaterproportion of their total pay in the form of variable pay through stock options. Thisadditional compensation risk explains the pay differences between technology firms andnon-technology firms.

We use the logarithm of total compensation as the dependent variable, compensationmix asa proxy for compensation risk, compensation mix and a technology dummy as the independentvariables and the logarithm of market value of equity and five-year return to shareholder as thecontrol variables. Compensation mix is defined as stock-based compensation to totalcompensation. The results of the regression analysis, as well as the additional robustness tests,reveal that the total compensation is significantly associated with compensation risk. Theresults are largely consistent with the labour market efficiency perspective. The highercompensation paid to CEOs in technology firms seems to be commensurate with the highercompensation risk that CEOs in technology firms bear. This result provides further support forthemarket equilibriummodel of CEO compensation.

The findings of this paper have practical implications for corporations and compensationconsultants. Compensation designers should consider both the benefits and costs ofgranting stock and stock options to executives. An increased portion of stock optionsdefinitely aligns the interests of shareholders and CEOs together and could maximize theretentive effect if CEOs have a significant amount of their wealth in unvested in-the-moneyoptions. However, the compensation risk embedded in options could escalate the totalcompensation. The optimal compensation contract should balance the benefits and the costsof the options.

These results also have practical implication on corporate governance. Consistentwith the literature, a CEO could earn much higher pay if he or she also serves as the chairof the board of directors. Though some activists argue for the separation of these tworoles so as to improve corporate governance, it is also possible that the CEO dualityreflects the firm’s demand for a high-quality CEO. In addition, many of the guidelines forimproving corporate governance state that the members of the compensation, audit andcorporate governance committees should be independent. If a CEO takes any of thesepositions, he or she is more likely to extract rents from the firm. We do not find anyevidence that the executive compensation is significantly associated with these boardstructures. It is possible that one kind of governance mechanism could be effective even ifother kinds of governance mechanisms are not effective. Chhaochharia and Grinstein(2009) argue that independent committee requirements to enhance board oversight couldreduce excessive compensation only if the firm lacks any other governance mechanisms.Practically, firms do not require all governance mechanisms. They just require one set ofsuitable governance mechanisms.

MAJ34,3

302

ReferencesBalsam, S., Fan, H. and Mawani, A. (2011), “How Canadian CEO compensation is approaching US

levels”, Working paper.Bebchuk, L.A., Grinstein, Y. and Peyer, U. (2010), “Lucky CEOs and lucky directors”, The Journal of

Finance, Vol. 65 No. 6, pp. 2363-2401.Chhaochharia, V. and Grinstein, Y. (2009), “CEO compensation and board structure”, The Journal of

Finance, Vol. 64 No. 1, pp. 231-261.Conyon, M.J. and Murphy, K. (2000), “The prince and the pauper? CEO pay in the United States and

United Kingdom”,The Economic Journal, Vol. 110 No. 467, p. F640-71.Conyon, M.J., Core, J.E. and Guay, W.R. (2011), “Are US CEOs paid more than UK CEOs? Inferences

from risk-adjusted pay”, Review of Financial Studies, Vol. 24 No. 2, pp. 402-438.Core, J.E., Holthausen, R.W. and Larcker, D.F. (1999), “Corporate governance, chief executive

officer compensation, and firm performance”, Journal of Financial Economics, Vol. 51 No. 3,pp. 371-406.

Eisenhardt, K.M. (1989), “Building theories from case study research”, Academy of ManagementReview, Vol. 14 No. 4, pp. 532-550.

Fernandes, N., Ferreira, M.A., Matos, P. and Murphy, K.J. (2012), “Are US CEOs paid more? Newinternational evidence”, SSRNWorking paper.

Gabaix, X. and Landier, A. (2008), “Why has CEO pay increased so much?”, Quarterly Journal ofEconomics, Vol. 123 No. 1, pp. 49-100.

Hall, B. and Murphy, K.J. (2002), “Stock options for undiversified executives”, Journal of Accountingand Economics, Vol. 33 No. 1, pp. 3-42.

Ittner, C.D., Lambert, R.A. and Larcker, D.F. (2003), “The structure and performance consequences ofequity grants to employees of new economy firms”, Journal of Accounting and Economics,Vol. 34 Nos 1/3, pp. 89-127.

Kaplan, S.N. and Rauh, J. (2010), “Wall Street and main street: What contributes to the rise in thehighest incomes?”, Review of Financial Studies, Vol. 23 No. 3, pp. 1004-1050.

Klassen, K. and Mawani, A. (2000), “The impact of financial and tax reporting incentives onoption grants to Canadian CEOs”, Contemporary Accounting Research, Vol. 17 No. 2,pp. 227-262.

Lambert, R.A., Larcker, D.F. and Verrecchia, R.E. (1991), “Portfolio considerations in valuing executivecompensation”, Journal of Accounting Research, Vol. 29 No. 1, pp. 129-149.

Lu, H., Richardson, G. and Salterio, S. (2011), “Direct and indirect effects of internal control weaknesseson accrual quality: evidence from a unique Canadian regulatory setting”, ContemporaryAccounting Research, Vol. 28 No. 2, pp. 675-707.

Lustig, H., Syverson, C. and Van Nieuwerburgh, S. (2011), “Technological change and the growinginequality in managerial compensation”, Journal of Financial Economics, Vol. 99 No. 3,pp. 601-627.

Markowitz, H. (1952), “Portfolio selection”,The Journal of Finance, Vol. 7 No. 1, pp. 77-91.Murphy, K.J. (2003), “Stock-based pay in new economy firms”, Journal of Accounting and Economics,

Vol. 34 Nos 1/3, pp. 129-147.Petersen, M. (2009), “Estimating standard errors in finance panel data sets: comparing approaches”,

Review of Financial Studies, Vol. 22 No. 1, pp. 435-480.Pratt, J.W. (1964), “Risk aversion in the small and in the large”, Econometrica: Journal of the

Econometric Society, Vol. 32 Nos 1/2, pp. 122-136.Smith, C., Watts, R., (1992), “The investment opportunity set and corporate financing, dividend, and

compensation policies”, Journal of Financial Economics, Vol. 32 No. 3, pp. 263-292.

Evidence fromtechnology

firms

303

Appendix

Corresponding authorSamir Trabelsi can be contacted at: strabelsi@brocku.ca

For instructions on how to order reprints of this article, please visit our website:www.emeraldgrouppublishing.com/licensing/reprints.htmOr contact us for further details: permissions@emeraldinsight.com

Table AI.Definition ofvariables

Variable Measurement

Dependent variableCompensation Sum of salary, bonus, other annual compensation, restricted stock grants, value

of options (Black–Scholes value), long-term incentive plan and othercompensation

Independent variablesCompensation mix Percent of restricted stock grant and value of options to total compensationTechnology An indicator variable is equal to 1 if a firm is a technology firm, and 0 otherwise.

Technology firms defined as companies with primary SIC codes 3570, 3571, 3572,3576, 3577, 3661, 3674, 4812, 4813, 5045, 5961, 7370, 7371, 7372 and 7373

Compensation mix�Technology

Interaction of the Compensation mix variable and the Technology variable

Control variablesMarket value Logarithm of stock market capitalization in thousands of dollars at the end of

yearNet sales Logarithm of sales in thousands of dollars at the end of the yearNet income Logarithm of after tax accounting earnings for the year in thousands of dollarsFive-year stock return The percentage of stock market return to shareholders in five yearsOne-year stock return The percentage of stock market return to shareholders in one yearReturn on assets The percentage of earnings before interest and taxes to total assetsCEO duality An indicator variable equal to 1 if the CEO serves as the chairman of the board,

and 0 otherwiseCompensation chairman An indicator variable equal to 1 if the CEO serves as the chairman of the

compensation committee, and 0 otherwiseCompensation member An indicator variable equal to 1 if the CEO serves as a member of the

compensation committee, and 0 otherwiseAudit chair An indicator variable equal to 1 if the CEO serves as the chair of the audit

committee, and 0 otherwiseAudit member An indicator variable equal to 1 if the CEO serves as a member of the audit

committee, and 0 otherwiseCorporate governance An indicator variable equal to 1 if the CEO serves as a member of the corporate

governance committee, and 0 otherwiseInstitutional holding Percent of shares held by institutional investors

MAJ34,3

304

The impact of corruption onanalyst coverage

Omaima HassanAccounting and Finance Department, Aberdeen Business School,

Robert Gordon University, Aberdeen, UK, and

Gianluigi GiorgioniManagement School, University of Liverpool, Liverpool, UK

AbstractPurpose – This study aims to investigate the impact of country-level corruption and firms’ anti-briberypolicies on analyst coverage. Analyst coverage has been identified as a powerful tool to detect fraud andshould equally act as a possible tool to reduce corruption.Design/methodology/approach – This study used a negative binomial count regression method on alongitudinal data set of a sample of S&P Global 1200 companies for the years 2010-2015. To control forpotential endogeneity bias and improve the reliability of the estimation, both country-level corruption andfirms’ anti-bribery policies variables were instrumented.Findings – After controlling potential endogeneity bias, the results show that the adoption of anti-briberypolicies at firm level attracts more analysts to follow a firm. The results for corruption at country level showthat analyst coverage increases in less corrupted countries indicating that the costs of corruption exceed itspotential benefits. When the variables corruption at country level and anti-bribery policies are interacted, therelationship is positive and highly significant.Practical implications – Given the potential important role played by anti-corruption measures, firmsare encouraged to adopt them to reduce the incidence of corruption and to increase analyst coverage, whichwill reinforce the benign effect of monitoring.Originality/value – Although the literature on corruption at the country level is rich, it is geared towardsthe determinants of corruption in contrast to its consequences, and fewer studies have focused on the impactof corruption at firm level because of data limitations. This paper addresses this gap and contributes to theliterature on the consequences of corruption at firm level.

Keywords Corruption, Sell-side analysts, Analyst coverage, Analysts following,Anti-bribery policies, Multi-country

Paper type Research paper

1. IntroductionThis study contributes to the growing literature on the impact of corruption. Judge et al.(2011) observed that the literature on corruption has been skewed towards examining thecauses of corruption in contrast to its consequences, as they observed that the number ofpapers published on the causes is twice as large as the number of papers published on theconsequences of corruption. This paper reduces this gap in the literature by empirically

JEL classification – G12, G14, M4, G230, G240, M490, D73The authors thank two anonymous referees for their constructive feedback, as well as delegates whoprovided feedback at the research seminar series at the Robert Gordon University, AberdeenBusiness School October 2017 and the Annual Congress of the European Accounting Association2018, Milan, Italy. Remaining errors are the responsibility of the authors.

Analystcoverage

305

Received 31 January 2018Revised 8 July 2018

Accepted 26 July 2018

Managerial Auditing JournalVol. 34 No. 3, 2019

pp. 305-323© EmeraldPublishingLimited

0268-6902DOI 10.1108/MAJ-01-2018-1783

The current issue and full text archive of this journal is available on Emerald Insight at:www.emeraldinsight.com/0268-6902.htm

examining the impact of corruption on analyst coverage. It investigates the impact ofcountry-level corruption and firms’ anti-bribery policies on analyst coverage. Corruption atfirm level can affect analyst coverage due to increased costs of information acquisition andcosts associated with the revelation of corruption (Dyck et al., 2010). If a firm adopts anti-bribery policies, this may provide some assurance to analysts that the management isproactive in dealing with corruption at firm level. Corruption at country level can affectanalyst coverage through increased levels of information costs, thereby reducing thenumber of analysts following. Alternatively, high levels of corruption at country level mayinduce the demand for analyst service for monitoring purposes, increasing thereby analystcoverage. Thus, the association between country-level corruption and analyst coveragedepends on a trade-off between the costs and benefits associated with analyst coverage. Theimportance of analyst coverage was identified by Dyck et al. (2010) in Table II page 2,225, inthe context of corporate fraud, as the second most effective fraud detectors within externalgovernance, well ahead of other groups such as auditors and industry regulators. Therefore,it is important to examine the determinants of analyst coverage, related to the actions takenby companies to tackle corruption. To do so, this paper uses a novel panel data set of asample of S&P Global 1200 companies for the years 2010-2015. This index represents 70 percent of global market capitalization, which provides good coverage and international reach,as well as enhances the generalizability of the results.

There are many definitions of corruption (Cuervo-Cazurra, 2016; Jain, 2001; Svensson,2003). Cuervo-Cazurra (2016, p. 36) defined corruption as the “abuse of entrusted power forprivate gains”. Entrusted power could originate from the government, shareholders,employees and trustees. Therefore, the abuse of entrusted power goes beyond thegovernment and includes firms, international organisations, not-for-profit organisations andnon-governmental organisations. Also, this definition indicates that people abuse thatentrusted power for their own benefits at the cost of their organizations (Cuervo-Cazurra,2016). Whether it is explicit in the form of bribes or implicit in the form of commission,corruption is perceived as a social evil and unethical practice that increases the hidden andreal costs of doing business and can result in misallocation of resources (Alam, 1995; Kehoe,1998; Dal B�o and Rossi, 2007; Chen et al., 2010). Corruption creates barriers to foreign directinvestment, dislocates international trade and misleads public policy (Cuervo-Cazurra, 2006;Zhao et al., 2003). However, the consequences of corruption on the economic performance offirms is less clear (Jong and Ees, 2014; Cuervo-Cazurra, 2016) with a few studies (Wu, 2005;Chen et al., 2010) indicating that corruption increases the information asymmetry betweenthe management and outside providers of funds, increasing thereby information costs.

On the other hand, following Bhushan’s (1989) seminal paper, prior studies haveinvestigated different determinants of analyst coverage at both firm and country levelsincluding firm characteristics, corporate governance issues, levels of investor protection,cross-listing, dynamics of coverage, the quality of accounting standards and listing location(Bushman et al., 2004; Boubakri and Bouslimi, 2010; Yu, 2010; Chen et al., 2007; Giraldo,2011; Abed et al., 2012; Kim and Shi, 2012; Hassan and Skinner, 2016). However, to date,there is a lack of direct empirical evidence on the association between analyst coverage andcorruption, although corruption is perceived as a deterioration factor to the informationenvironment and the economic performance of companies (Wu, 2005; Dal B�o and Rossi,2007; Doh et al., 2003; Rodriguez et al., 2006; Uhlenbruck et al., 2006). In addition, althoughthe literature on corruption is growing, it is mostly skewed towards examining the causes ofcorruption rather than its consequences (Judge et al., 2011). In this context, this studycontributes to the literature in several aspects. It contributes to the literature on corruptionby investigating the consequences of corruption at firm level. It also contributes to the

MAJ34,3

306

literature on analyst coverage by extending the determinants of analysts following beyondfirm characteristics to include firm policy and institutional factors. It postulates that theassociation between country-level corruption and analyst coverage depends on a trade-offbetween the benefits and costs associated with analyst coverage. In addition, while self-reported anti-bribery policies might not be enough to draw on their effectiveness incombating bribery and corruption at the firm level, it postulates that such policies maysignal the serious commitment of the management to prevent, monitor and addresscorruption. It extends the work of Hassan and Giorgioni (2015) by using a more recent,longitudinal data set in contrast to their cross-sectional analysis and controlling for potentialendogeneity bias. In addition, the results of Hassan and Giorgioni (2015) are likely to sufferfrom selection bias because they limited their sample to companies with analyst coverageonly, whereas companies that do not attract analyst coverage are excluded from theirsample. In contrast, the current study performs a two-stage regression analysis using apanel data set of a sample of S&P Global 1200 companies for the years 2010-2015 to controlfor potential endogeneity bias in the estimation of the research model. In this data set, thenumber of analysts following a company ranges from 0 to 63. The use of panel data analysisprovides many advantages over both the traditional cross-sectional and time-seriesanalyses. It provides the researcher with a larger number of observations, increasing thedegrees of freedom for any statistical testing and lessening the problem of multicollinearityamong the explanatory variables (Hsiao, 2002), improving thereby the efficiency ofestimates. In addition, prior studies on analyst coverage tend to apply variations of theclassical linear regression model to data sets where the dependent variable (the number ofanalysts following) can only take non-negative integer values. In contrast, this study uses aneconometric procedure that adjusts for the count nature of the dependent variable, as usedby Rock et al. (2001), Boubaker and Labégorre (2008) and Hassan and Skinner (2016), exceptthat we apply the count data regression technique on a larger and more recent multi-countrysample. Specifically, we use a negative binomial count (NBC) regression method to a sampleof 5,888 firm-year observations from 30 different countries. The current study also uses theTransparency International’s Corruption Perception Index (CPI) as a proxy for corruptionscore at country level, which is typically used in prior studies (Yu, 2010; Chen et al., 2010;Judge et al., 2011). This index captures the perceptions of business people, academics andrisk analysts of the level of corruption at the country level. The CPI ranks countries based onhow corrupt their public sector is perceived to be on a scale of 0 to 100, where 0 means that acountry is perceived as highly corrupt and 100 means that it is perceived as very clean. TheCPI has been validated as a proxy for corruption at country level in prior studies (Wilhelm,2002; Davis and Ruhe, 2003). After controlling for potential endogeneity bias in the variablesof interest, the results show that the adoption of anti-bribery policies at firm level attractsmore analysts to follow a firm, in line with our hypothesis. The results for corruption atcountry level show that analyst coverage increases in less corrupted countries indicatingthat the costs of corruption exceed its benefits. When the variables corruption at countrylevel and anti-bribery policies are interacted, the relationship is positive and highlysignificant.

The importance of the current study arises from the importance of analyst coverage tofirm visibility. According to Arbel et al. (1983) and Merton (1987), neglected or less-visiblefirms are ones that are not followed by large numbers of financial analysts (investors) on aregular basis. Financial analysts increase the visibility of the firms they follow by signallinginformation about their performance, increasing thereby the demand for their commonshares even when they do not actively add new information about these firms (Mola et al.,2013; Li and You, 2015). Also, higher analyst coverage may lead to the company being

Analystcoverage

307

included either in industry reports or as an industry comparison in a report on a largercompany, creating both visibility and credibility (Bushee andMiller, 2012). In addition, priorempirical studies suggest that higher[1] analyst coverage is associated with lowertransaction costs, higher stock liquidity, lower cost of capital and lower audit fees (Doukaset al., 2000; Jung et al., 2012; Lang et al., 2012; Easley and O’hara, 2004; Gotti et al., 2012).Therefore, it is not surprising that analyst coverage is actively pursued by firms. Forexample, Anantharaman and Zhang (2011) provided evidence that managers value analystcoverage and are willing to expend resources to maintain a particular level of that coverage.Also, Cliff and Denis (2004) found that firms are willing to compensate for analyst coveragethrough initial public offerings premiums. Bushee and Miller (2012) noted that some firmsmay resort to hiring investor relations professionals to pitch their business to analysts, andKirk (2011) stated that firms are prepared to buy paid-for research. Sibilkov et al. (2013)found that firms value analyst coverage and are prepared to strategically use the choice ofappointments for mergers and acquisitions advisors to secure analyst coverage.

The remainder of this paper is organized as follows. Section 2 provides a brief discussionof the literature and develops the research hypotheses. Section 3 explains the researchmodel. Section 4 describes the research samples and discusses the results. Section 5 providesconcluding remarks.

2. Literature review and hypotheses developmentFinancial analysts collect a wide variety of information about the firms they follow and theirindustries andmarkets, analyze it and produce their reports. These reports may include buy,sell or hold recommendations, the competitive position of the firm relative to its rivals andanalysts’ forecasts of earnings and cash flows. According to the seminal paper of Bhushan(1989), the equilibrium total expenditure by investors on analyst services for a particularfirm in a given period is a function of the aggregate demand for and supply of analystservices.

Demand for analyst services arises in situations characterised by informationasymmetry where agency problems can arise between managers and outside providers offunds. Analyst coverage helps mitigate these problems by providing information about thequality of investment opportunities to outside providers of funds and to the financialintermediaries they represent (Palepu et al., 2013). This information intermediary role ofanalysts goes beyond interpreting corporate disclosure to discovering new information andclarifying and confirming corporate disclosures (Huang et al., 2017). Chen et al. (2015)showed how analysts help reducing the agency problem between shareholders andmanagers through direct and indirect monitoring. Analysts can provide a direct monitoringrole through regular examination of the financial statements of the firms they follow andfrequent interactions with the management via conference calls for example. They can alsoprovide an indirect monitoring role through the dissemination of information to institutionalinvestors and ordinary investors through research reports and media outlets, such asnewspapers and TV programs. Therefore, we expect that in a highly corrupt country, themonitoring role of analysts will be even more important, driving thereby the demand foranalyst services and increasing analyst coverage. Alternatively, analysts themselves mightbe bribed to provide coverage (Cuervo-Cazurra, 2016).

However, higher level of corruption can also reduce analyst coverage due to increasedlevel of information costs. According to Bhushan’s (1989) model, the aggregate supply ofoutside analyst service is a function of the information costs they incur. The higher theinformation costs associated with following a specific company, the less likely that analystswill follow that company.

MAJ34,3

308

If country-level corruption is high, it is more likely that bribes and corruption will beperceived as normal ways of doing business even if this is illegal (Hauser and Hogenacker,2014). According to transaction cost economics theory (Cuervo-Cazurra, 2016), operating in acorrupt country increases transaction costs because firms incur the costs of bribes to dobusiness, in addition to the time and attention devoted by its employees to manage corruptactivates. However, bribes can also facilitate business activities and increase firms’revenues. In addition, operating in a corrupt country increases uncertainty becausemanagers are not sure whether they are bribing the right people, and whether they willreceive the promised benefits of these bribes. Thus, the net impact of corruption on firms’profitability depends on a trade-off between the costs and benefits associated with bribes.This, in turn, increases the level of uncertainty that analysts face in projecting futureearnings and cash flows for these firms (Chen et al., 2010). In addition, managers who areengaged in corrupt practices will try to hide the bribe payments (Wu, 2005), increasingthereby the level of information asymmetry between the management and outside providersof funds and market participants in general. This, in turn, makes it even harder for analyststo collect and process relevant information about the firms they follow to produce theirreports. Moreover, high level of corruption may induce access to private informationthrough bribes, reducing thereby the demand for outside analyst services.

Furthermore, Dal B�o and Rossi (2007) found that more corruption at the country level isstrongly associated with more inefficient firms, in the sense that they use more inputs toproduce a given level of output. This, in turn, means that analysts will have to restate thefinancial statements to eliminate the effect of corruption, increasing thereby the costs ofcollecting and processing information. Following companies operating in highly corruptcountries can also increase the likelihood of financial loss for outside analysts if, for instance,they fail to provide an effective monitoring tool for stakeholders (Dyck et al., 2010). Karpoffet al. (2008) noted that the largest monetary penalties inflicted to companies caught incorruption cases are not imposed by regulators or courts, but are imposed by the market,where presumably analyst coverage does play an important role. Karpoff et al. (2008)estimated that a quarter of the loss in market value is because of a re-assessment of the truefinancial situation of a firm (i.e. without the advantages obtained through corruption), buttwo-thirds are because of the loss of reputation as stakeholders of the company (e.g.investors, customers and suppliers) modify the way they conduct business with thecompany. Legal costs only play a minor role (around 10 per cent) of the overall loss of stockmarket value.

Thus, the association between country-level corruption and analyst coverage depends ona trade-off between the costs and benefits associated with analyst coverage. If the benefits ofthe monitoring role of analysts in highly corrupt countries are higher (lower) than theirassociated costs, higher level of corruption will be associated with higher (lower) analystcoverage. Therefore, we develop our first hypothesis without estimating a direction of suchassociation as a priori:

H1. There is an association between the level of corruption and the number of analystsfollowing

Recent years have witnessed an international shift towards strengthening anti-corruption laws in many countries worldwide such as the US Foreign Corrupt PracticesAct, the UK Bribery Act, the Council of Europe Criminal Law Convention on Corruptionand the Canadian Corruption of Foreign Public Officials Act, among others. One aspectof these strengths is to criminally prosecute corruption practice by individuals andcompanies in their home countries even when the corrupt practice occurs abroad

Analystcoverage

309

(Hauser and Hogenacker, 2014; Carr-Howard, 2014; Cuervo-Cazurra, 2016). Failure toadhere to such laws can result in serious reputational damage, significant fines,imprisonment of individuals and debarment of organizations from conducting businesswith national and local governments (Carr-Howard, 2014). According to the neo-institutional theory, companies are motivated to address the norms, rules andregulations that prevail in their respective external institutional environments[2] toearn the legitimacy needed to operate in their societies (Baum and Oliver, 1991; Cuervo-Cazurra, 2016). In this context, firms can take a wide range of internal measures totackle corruption and preserve its legitimacy. These measures are called firm-levelcontrols of corruption and include, but not limited to, financial record keeping,statements by executive officers, internal monitoring, whistle-blowing facilities, use ofcompliance manuals, appointment of compliance officers or committees, threats ofdisciplinary actions and training for compliance, among others (Gordon and Miyake,2001). Cuervo-Cazurra (2016) suggested future research to study the effectiveness offirm-level controls of corruption in reducing bribery by their managers. Adam andRachman-Moore (2004) found that the social norms of the organization are perceived byemployees to have the most influence on their conduct. Hauser and Hogenacker (2014)found that firms tend to follow a reactive approach toward the management ofcorruption risks in foreign countries and only implement anti-corruption measures ifthey have been confronted with the issue. Dyck et al. (2010) demonstrated that theincentives for the existing network of whistleblowers (auditors, analysts andemployees) are weak, thus it may be easier for the analyst just to drop coverage ofcompanies involved in corrupt activities rather than revising their recommendationsand developing a reputation of exposing corporate scandals (Young and Peng, 2013).While self-reported anti-bribery policies might not be enough to draw on theireffectiveness in combating bribery and corruption at the firm level, the fact that a firmadopts and reports such policies may signal the serious commitment of themanagement to prevent, monitor and address corruption (Transparency International,2009). In fact, recent empirical evidence shows that self-reported anti-corruptionpolicies do reflect companies’ real efforts to combat bribery and corruption and notmerely cheap talk (Healy and Serafeim, 2016). Therefore, it is expected that firms thathave signed such measures will have larger analyst coverage, hence the secondresearch hypothesis is:

H2. The adoption of anti-bribery policies at firm level is expected to induce the numberof analysts following a company

Another issue worth investigating is whether the adoption of anti-bribery policies in firmsthat operate in highly corrupt countries would mitigate the impact of country-levelcorruption and induce analyst coverage. In other words, we examine the impact of theinteraction term between corruption at country level and firms’ self-reported anti-corruptionpolicies on analyst coverage. Therefore, the third research hypothesis is:

H3. The presence of anti-bribery policies at firm level increases the number of analystsfollowing a company that operates in a highly corrupt country.

3. Research modelBhushan (1989) considered various company characteristics that affect the number ofanalysts following a firm, such as ownership structure, firm size and return variability.Following Bhushan (1989), the current study examines the impact of country-level

MAJ34,3

310

corruption and firms’ anti-bribery policies on analyst coverage, controlling for several firm-level and country-level factors. Therefore, the research model takes the following form(cross-section and time identifiers are suppressed for simplicity):

NOA ¼ b 0 þ b 1ABP þ b 2 CPI þ b 3 ABP *CPIð Þ þ b 4 NINST þ b 5 INST

þ b 6 INSIDþ b 7 MCAP þ b 8VOLþ b 9 SEGþ b 10 BETAþ b 11ROA

þ b 12GAAP þ b 13INDUSTRY þ b 14CGOV þ b 15GDP þ «

However, it could be argued that analyst coverage could itself induce firms to change theirbehaviour. For instance, firms may be more inclined to adopt anti-bribery policies as aconsequence of analyst coverage. Therefore, the variables ABP and NOA may beendogenously determined. Endogeneity may lead to biased results. Therefore, this paper

Table I.A summary of

variablemeasurement

Variable Definition

NOA The total number of analysts making recommendations for the security at the financialyear-end

CPI The corruption score at the country level provided by Transparency International. Ittakes a value from 0 to 100, where 0 means that a country is perceived as highly corruptand 100 means it is perceived as very clean

ABP It represents firms’ self-reported anti-corruption policies. It is a dummy variable thattakes the value of one if the company has policies in place to prevent corruption and zerootherwise

HRP A dummy variable that takes the value of 1 if the company has implemented anyinitiatives to ensure the protection of the rights of all people it works with, and 0 if thecompany has not explicitly disclosed any such efforts in its Annual or CompanyResponsibility reports

BEP A dummy variable which takes the value of 1 if the company has established ethicalguidelines and/or a compliance policy for its non-management/executive employees inthe conduct of company business, and 0 if the company has not explicitly disclosed thispolicy in its Annual or Company Responsibility Reports

FDI The percentage of foreign direct investment net inflow to GDPOpenness Imports of goods and services (% of GDP)NINST The number of institutions holding shares in a companyINST (INSID) The percentage of institutional (insiders’) holding in a companyMCAP The market value of equity of the firm at the fiscal year-endSEG The number of recorded business segmentsVOL The standard deviation of the relative price change for the 360 calendar days closing

price, expressed as a percentageBETA The market model beta for each stock measured via the market model using weekly

dataROA The return on assets measured as net income divided by total assetsGAAP A dummy variable that takes the value of 1 if the country applies US accounting

standards or international financial reporting standards (IFRS) and 0 otherwiseINDUSTRY A dummy variable that takes the value of 1 for the relevant industry and 0 otherwiseCGOV A governance index developed using the principal component analysis. It is the first

principal component of all corporate governance indicators at country level except thecontrol of corruption indicator

GDP The gross domestic product per capita

Note: This table reports the definition of the different variables

Analystcoverage

311

uses a two-stage regression analysis to remove the problem of endogeneity between theadoption of anti-bribery measures (ABP) and analyst coverage. The instrumental variablesare HRP and BEP (please see Table I for more details). This study uses a popular measurefor corruption at the country level, i.e. the Transparency International’s CPI, which istypically used in prior studies (Yu, 2010; Chen et al., 2010) and has been validated againstother alternative proxies of corruption in prior studies (Wilhelm, 2002; Davis and Ruhe,2003). This index captures the perceptions of business people, academics and risk analystsof the level of corruption at country level. However, some issues have been raised in theliterature concerning the use of perception indexes in general (Olken, 2009; Razafindrakotoand Roubaud, 2010) and the CPI in particular (Bertrand and Mullainathan, 2001; Knack,2007; Donchev and Ujhelyi, 2014). For example, the surveys document “expert opinions” of asmall number of respondents (or perceptions of what “other firms” are doing) rather thantheir actual experience of corruption. Respondents may also be influenced by previousratings or by recent events positive (recent economic performance) or negative (recentpublicized corruption scandals). In addition, Donchev and Ujhelyi (2014) claimed thateconomic development, political system and cultural variables tend to bias perceptions ofcorruption away from experience. Furthermore, Fan et al. (2009) reported that countries withsimilar levels of corruption frequency (gleaned from surveys) may have very different levelsof corruption perceptions and vice versa. Moreover, the methodology for measuring theindex has been changed in 2012, forcing a reduction in the number of observations thatcould be used. Therefore, we have re-estimated CPI with instrumental variables (FDI is thepercentage of foreign direct investment net inflow to gross domestic product [GDP] andopenness is the imports of goods and services [per cent of GDP]). This procedure could alsoallay the fear that the variable CPI (perception of corruption) on aggregate may itself beinfluenced by the level of corruption of firms, which in itself could be influenced by analystcoverage. Variable definition andmeasurement are explained in Table I.

3.1 Dependent variables3.1.1 NOA. NOA is the total number of analysts making recommendations for the securityat the financial year-end.

3.2 Explanatory variables3.2.1 ABP. ABP is a dummy variable to represent firms’ self-reported anti-corruptionpolicies. It takes the value of one if the company has policies in place to prevent corruptionand zero otherwise.

3.2.2 CPI. It is the corruption score at the country level provided by TransparencyInternational, which is typically used in prior studies; see Judge et al. (2011) for a review ofpapers that have used this measure. It reflects the perceived level of corruption atthe country level, which aggregates information frommultiple surveys into one indicator forthe country (Cuervo-Cazurra, 2016, p. 38). It takes a value from 0 to 100, where 0 means thata country is perceived as highly corrupt and 100 means it is perceived as very clean.

3.3 Firm-level control variablesSeveral firm-level controls are considered in the research model, namely, ownershipstructure (NINST; INST; INSID), firm size (MCAP), business complexity (SEG), returnvolatility (VOL; BETA), firm profitability (ROA), the quality of the accounting standards(GAAP) and industry-type (INDUSTRY). Where, NINST is the number of institutionsholding shares in a company, and INST (INSID) is the percentage of institutional (insiders’)holding in a company. According to Bhushan (1989), more concentrated institutional

MAJ34,3

312

ownership (NINST; INST) may increase (decrease) the demand on in-house analyst servicesrather than outside analyst services if this is cost effective (ineffective) for individualinstitutional investors. In addition, a negative association between INSID and NOA isexpected because insiders would have access to private information, reducing thereby thedemand on outside analyst services (Bhushan, 1989).

Firm size (MCAP) is the market value of equity. SEG is the number of recorded businesssegments. Larger firms (MCAP) are expected to attract higher analyst coverage becausethey are widely held and have wider investors’ bases with more potential transactions ofbusiness to outside analysts. However, as the number of business segments (SEG) increases,the number of analysts following will decrease as the firm will become more complex andexpensive to follow (Bhushan, 1989).

VOL is the standard deviation of the relative price change for the 360 calendar days’closing price, expressed as a percentage, whereas BETA is the market model beta for eachstock measured via the market model using weekly data. Higher return volatility (VOL;BETA) implies higher uncertainty in predicting future returns. Thus, if outside analystinformation is perceived to add valuable information over and above public information inpredicting future returns, then higher return volatility is expected to increase the demand foroutside analyst services (Bhushan, 1989; Hussain, 2000; Boubaker and Labégorre, 2008).

ROA is the return on assets to proxy for firm profitability. Prior empirical studiessuggest that analysts will be reluctant to follow less profitable companies (Boubaker andLabégorre, 2008), implying a positive association between ROA and NOA. Furthermore, weinclude a dummy variable (GAAP), take the value of 1 if the country applies US accountingstandards or international financial reporting standards (IFRS) and 0 otherwise, to proxy forthe quality of the accounting standards applied by the sample firms, where high-qualityaccounting standards are expected to attract more analyst coverage (Kim and Shi, 2012). Inaddition, the quality of the accounting standards might also be associated with the extent ofthe perceived corruption at country level since Khalil et al. (2015) found that firms are lesslikely to grant gift to secure a government contract in countries having more extensivefinancial reporting requirements. Also, we include industry dummies (INDUSTRY) thattakes the value of 1 for the relevant industry and 0 otherwise to control for the type ofindustry.

3.4 Country-level control variablesCGOV is the governance index at country level, which is developed using the principalcomponent analysis. It is the first principal component of all corporate governanceindicators at country level except the control of corruption indicator. It explains 75 per centof the variation in the corporate governance indicators. It is expected that better corporategovernance environment will attract more analyst coverage. Alternatively, in a badcorporate governance environment, there is more demand for the monitoring role of outsideanalysts. GDP is the gross domestic product per capita to measure country income.

4. Sample and results4.1 Research sampleThis study examines the impact of corruption on analyst coverage using a sample of S&PGlobal 1200 companies from 2010 to 2015. The S&P Global 1200 index represents 70 percent of global market capitalization, which is likely to offer good variation in the variables ofinterest and has international reach. The S&PGlobal 1200 consists of seven indices, many ofwhich are accepted leaders in their regions. These include the S&P 500 (the USA), S&PEurope 350, S&P TOPIX 150 (Japan), S&P/TSX 60 (Canada), S&P/ASX All Australian 50,

Analystcoverage

313

S&P Asia 50 and S&P Latin America 40. According to the OECD (2014), most internationalbribes are paid by large companies to win contracts in advance economies rather than in thedeveloping world, and that most bribe payers and takers are from wealthy countries, whichjustifies the choice of our sample and setting.

We collected data for the constituents of S&P Global 1200 for the years 2010-2015deliberately avoiding the recent financial crisis of 2008/2009. Data at the firm level arecollected from Bloomberg database. Data on the CPI scores are obtained from TransparencyInternational database. Data on GDP per capita, foreign direct investments, corporategovernance indicators and imports of goods and services to GDP are obtained from theWorld Bank. This has yielded an initial sample of 1,187 companies for the years 2010-2015with 7,122 firm-year observations. However, the final sample size depends on dataavailability for each variable included in the model. It is worth noting that TransparentInternational changed its methodology in calculating corruption score in 2012. The finalcommon sample consists of 1,050 firms covering 30 different countries for the years2010-2015 with 5,888 firm-year observations. Table II shows a breakdown of the sample bycountry. Interestingly, countries like South Africa and China that are at the lower end of theGDP and CPI spectrums attract higher than average analyst coverage. This table also tellsan interesting story about the average of analyst coverage, corruption and GDP, which goes

Table II.A breakdown of thesample by country

Country Obs. NOA CPI GDP

Australia 230 15 81 53,363Austria 24 23 80 47,571Belgium 54 22 73 44,509Bermuda 4 19 71 83,752Brazil 12 12 38 11,483Britain 583 22 77 3,9742Canada 273 20 82 48,987Chile 25 12 68 13,969China 42 29 24 5,533Denmark 30 28 85 58,894Finland 42 26 87 45,972France 270 27 70 41,265Germany 168 34 78 44,138Greece 30 14 47 23,593Hong Kong 65 20 75 34,574Ireland 35 20 77 50,183Italy 105 26 54 34,713Japan 821 15 74 45,701Luxembourg 12 26 83 104,425Mexico 28 13 35 9,235The Netherlands 78 29 83 50,550Norway 30 28 87 88,442Portugal 24 16 63 21,830Singapore 35 24 77 49,716South Africa 6 23 45 7,534South Korea 44 42 61 23,624Spain 89 30 61 29,898Sweden 134 24 86 53,156Switzerland 153 27 86 75,056USA 2,442 21 74 49,791Grand Total 5,888 22 74 46,851

MAJ34,3

314

against intuition or common sense. For example, a country like China has an average GDPper capita of 5,533 USD with corruption score (CPI) of 24 but attracts higher than averageanalyst coverage (29), whereas a country like Australia has an average GDP per capita of US$53,363 with a corruption score of 81 but attracts less than average analyst coverage of 15.Another example, countries like South Africa, Austria, Brazil and Chili have the sameaverage number of analyst following but they have significantly different levels ofcorruption. A third example is that it is not the USA but South Korea that attracts thehighest average number of analyst coverage. Thus, the perception that highly cleancountries would attract more analyst coverage is not observed from these figures, whichhighlights that the association between analyst coverage and the extent of corruption is anempirical issue.

4.2 Results and discussionPanel A of Table III shows the descriptive statistics of all the variables. It shows that theaverage company in the research sample comes from a moderately clean country with a CPIscore of 74 and an average GDP of US$46,851. It also shows that 62 per cent of the samplefirms have policies in place to tackle corruption at the firm level. The ownership structure ofa typical firm in the sample is 66 per cent institutional holding (INST) and less than 2 percent insiders holding (INSID). The average number of analysts following a firm (NOA) isabout 22 with an average number of 633 institutions holdings (NINST). Marketcapitalization (MCAP) for a typical firm is on average US$28,361m. Average returnvariability (VOL) is 31 and market beta (BETA) is 1 on average. A typical firm has foursegments (SEG) on average and an average return on assets (ROA) of 5 per cent. About 89per cent of the sample firms apply either US accounting standards or IFRS.

Panel B of Table III shows the pair-wise correlation matrix. The correlations do notidentify any need for concerns about multicollinearity, as most correlations are quite weak.The results show no correlation between NOA and CPI in contrast to H1. However, theyshow that the higher the probability that a firm adopts an anti-bribery policy (ABP), thehigher the number of analysts following it, consistent withH2.

Table III also shows that the number of analysts following (NOA) is positively associatedwith the number of institutions holding shares in a company (NINST), firm size (MCAP),firm profitability (ROA) and the quality of accounting standards (GAAP) consistent withresults from prior studies (Bhushan, 1989; Hussain, 2000; Boubaker and Labégorre, 2008;Kim and Shi, 2012). The association between the number of analysts following and thenumber of recorded segments (SEG) is positive and significant, contrary to expectations,indicating higher analyst coverage for more complex companies. Table III also shows anegative and significant correlation between NOA and the percentage of institutionalholdings (INST) indicating that higher percentage of institutional holdings motivates morereliance on the service of in-house analysts rather than outside analysts (Bhushan, 1989).Furthermore, Table III also shows a negative and significant association between analystcoverage and return volatility (VOL) inconsistent with prior expectations. However, thesefindings reflect pair-wise correlations only; a multi-regression analysis could yield differentresults.

Table IV reports the results for the research model explaining analyst coverage. Theestimation is based on a count regression model that better suits an integer dependentvariable, after controlling for heteroscedasticity.

Ordinary least squares (OLS) assume that the dependent variable is continuous whereclearly our dependent variable, i.e. NOA, is a count integer taking on values from 0 to 63.Using OLS in this context can lead to biased and inconsistent estimates (Long and Freese,

Analystcoverage

315

NOA

ABP

CPI

NINST

INST

INSID

MCA

PVOL

SEG

BETA

ROA

GAAP

GDP

PanelA

:Mean

21.59

0.62

74.30

633.35

66.43

1.34

28,361.17

31.47

3.86

1.05

4.70

0.89

46,851.33

Median

21.00

0.70

74.37

498.50

70.43

0.23

13,577.58

28.62

3.00

1.02

3.93

1.00

48,455.21

SD8.85

0.24

8.00

492.07

25.99

4.13

45,079.35

13.69

2.66

0.30

6.52

0.31

9,907.94

PanelB

:NOA

1.000

ABP

0.399a

1.000

CPI

�0.005

0.048a

1.000

NINST

0.389a

0.253a

0.038a

1.000

INST

�0.029

b0.097a

0.092a

0.357a

1.000

INSID

0.017

�0.063

a�0

.004

0.013

�0.012

1.000

MCA

P0.409a

0.291a

�0.056

a0.614a

�0.080

a�0

.014

1.000

VOL

�0.115

a�0

.275

a�0

.178

a�0

.328

a�0

.089

a�0

.011

�0.231

a1.000

SEG

0.097a

0.024c

0.022c

0.061a

�0.200

a�0

.106

a0.166a

�0.056

a1.000

BETA

0.010

�0.061

a�0

.008

�0.059

a0.105a

0.018

�0.120

a0.534a

0.038a

1.000

ROA

0.100a

0.082a

0.060a

0.234a

0.148a

0.104a

0.193a

�0.309

a�0

.154

a�0

.227

a1.000

GAAP

0.292a

0.550a

0.023c

0.263a

0.311a

0.069a

0.126a

�0.075

a�0

.131

a0.042a

0.112a

1.000

GDP

�0.001

0.002

0.724a

0.191a

0.125a

0.033a

0.004

�0.130

a0.010

0.088a

0.042a

0.042a

1.000

Notes

:N=5,888firm

-yearo

bservatio

ns;a,bandc :correlations

aresign

ificant

atthe0.01,0.05and0.10

levels(tw

o-tailed),respectively

Table III.Descriptive analysisfor variables

MAJ34,3

316

2006); therefore, we rely on a count regression method to predict analyst coverage. Poissonmodel is a count regression model which corrects for the discrete, count data nature of thedependent variable and is especially suitable when the (conditional) mean and variance ofthe dependent variable are equal. If the equality of the (conditional) mean and variance doesnot hold, then the standard errors generated by a maximum likelihood of the Poisson will beunderestimated thereby conveying incorrectly high level of significance, in such case a NBCregression method is more suitable (Rock et al., 2001). In this paper, an NBC regressionmethod is used. As endogeneity is an issue for our research model, we use two-stageregression analysis to control for this potential bias. We run the analysis with one and twoestimated variables and compare the results.

Table IV shows the results of the second-stage regression model after estimating each ofABP and CPI using instrumental variables in results not tabulated. It shows that the resultsseem to strongly indicate that firms operating in a highly clean country (measured by highlevels of the variable CPI) attract fewer analysts, once the effect of presence of the anti-bribery policy (and the control variables) is taken into consideration. This appears to bequite a consistent finding. This finding contradicts the results from Chen et al. (2010) thatoperating in corrupt countries makes it harder for analysts to predict firms’ future earningsand cash flows. However, once the variable CPI is replaced by estimated instrumental

Table IV.Method: ML –

negative binomialcount (quadratic hillclimbing) with QML

(Huber/White)standard errors and

covariance

One estimated variable* Two estimated variables**

B0 1.460a 1.362a �1.432a 1.917a

ABP 0.126a 0.276 0.164a �6.984a

CPI �0.007a �0.006c 0.034a �0.011ABP� CPI �0.002 0.097a

NINST 5.33E-05a 5.29E-05a 7.53E-05a 8.20E-05a

INST �3.52E-04 �3.41E-04 �5.27E-04b �5.48E-04b

INSID �1.74E-03 �1.76E-03 �1.44E-04 �4.16E-04LOG(MCAP) 0.177a 0.177a 0.161a 0.159a

VOL 0.001 0.001 0.001c 0.001c

SEG �0.004c �0.004c �0.002 �0.002BETA 0.131a 0.131a 0.081a 0.078a

ROA �0.004a �0.004a �0.003a �0.003a

GAAP 0.322a 0.322a 0.327a 0.317a

COMMUNICATIONS 0.152a 0.152a 0.153a 0.151a

INDUSTRIAL 0.013 0.013 �0.004 �0.004DIVERSIFIED �0.364a �0.359a �0.546a �0.528a

ENERGY 0.150a 0.149a 0.133a 0.133a

TECHNOLOGY 0.202a 0.202a 0.195a 0.197a

UTILITIES �0.097a �0.097a �0.109a �0.112a

BASIC MATERIALS 0.114a 0.114a 0.067a 0.066a

CONSUMER CYCLICAL 0.101a 0.101a 0.079a 0.082a

CONSUMER NONCYCLICAL �0.029 �0.029 �0.012 �0.012CGOV 0.044a 0.044a 0.007b 0.007b

GDP �3.09E-06a �3.08E-06a �2.78E-06a �2.93E-06a

Adjusted R2 0.40 0.40 0.39 0.39Log likelihood �13,021.92 �13,021.79 �19,742.91 �19,728.39Obs. 3,897 3,897 5,888 5,888

Notes : Dependent variable: number of analysts following (NOA); a, band c: coefficients are significant atthe 0.01, 0.05 and 0.10 levels (two-tailed), respectively; *: ABP is estimated using instrumental variables; **:both ABP and CPI are estimated using instrumental variables

Analystcoverage

317

variable, the relationship becomes positive. The results also show that adopting anti-briberypolicies at the firm level (ABP) has a positive and significant association with analystcoverage consistent with H2. This finding supports the recent empirical evidence by Healyand Serafeim (2016) that self-reported anti-corruption policies reflect a genuine commitmentby companies to combat bribery and corruption at firm level and not merely cheap talks,thus inducing high analyst coverage. Finally, the results of the interaction term capturingthe effect of operating in a less corrupt country and the adoption of anti-bribery dependupon the choice of the variable approximating corruption. With the CPI index, the adoptionof anti-corruption policies does not induce analyst coverage, rejecting thereby H3. However,if the estimated CPI variable is used, anti-corruption policies appear to induce analystcoverage, in line withH3.

In terms of the control variables, the results are remarkably consistent regardless of thechoice of variable CPI. Analyst coverage (NOA) is significantly and positively associatedwith the number of institutional investors (NINST), firm size (MCAP), market beta (BETA)and the quality of accounting standards (GAAP) consistent with results prior studies(Bhushan, 1989; Hussain, 2000; Rock et al., 2001; Boubaker and Labégorre, 2008; Kim andShi, 2012). NOA is also significantly and positively (negatively) associated with CGOV(GDP). Consistent with results from prior studies, NOA is negatively associated withbusiness complexity (SEG) (Bhushan, 1989; Rock et al., 2001). Surprisingly, firm profitability(ROA) is negatively associated with NOA. This result indicates more analyst coverage forless profitable firms unlike results from prior studies (Boubaker and Labégorre, 2008). Thisis potentially because of less profitable firms having more volatile stock returns, as can beseen from the negative and significant association between ROA and each of VOL andBETA in Table III, which induces more demand on outside analyst service.

Finally, Khalil et al. (2015) found that firms are less likely to bribe bureaucrats in countrieswhich havemore extensive financial reporting requirements and those where audit firms face ahigher litigation and sanction risk. At the firm level, Khalil et al. (2015) found that firms are lesslikely to bribe bureaucrats when their financial statements are reviewed by an external auditfirm. Taken together, the results of Khalil et al. (2015) might indicate that the current analysis isprone to omitted variable bias[3]. Thus, as a robustness check, we re-run the analysis aftercontrolling for audit quality; a dummy variable[4] (BIG4) that takes the value of 1 if a firm’saccounts are audited by one of the big four audit firms and 0 otherwise, and the business extentof disclosure (DISC), a disclosure index of the business extent of disclosure at the country-levelprovided by the World Bank, in results not tabulated to test the consistency of our results. It isalso worth noting that the quality of accounting standards at the firm level is considered usingGAAP and the quality of the governance and monitoring roles at the country level areconsidered using CGOV. Our main results continue to hold after the inclusion of both BIG4 andDISC in the analysis. Also, the BIG4 variable was interacted with the variable capturinggovernance rules at country level (CGOV). The results were in line with the previous findings.Furthermore, we test whether the three variables of interest, i.e. ABP, CPI and ABP*CPI, areredundant and might thus be deleted from the research model, in results not tabulated, afteraccounting for the several firm-level and country-level controls including BIG4. The redundantvariable hypotheses are rejected for any pair of these variables and for the three variablesjointly, which indicates that these variables are significant estimators of the number of analystsfollowing a company.

5. Concluding remarksThis study investigates the impact of corruption at country level and firm’s action to tacklecorruption on analyst coverage using an NBC regression method for a sample of S&P Global

MAJ34,3

318

1200 companies for the years 2010-2015. To remove the issue of endogeneity and improvethe reliability of the variable corruption, both variables corruption perception (CPI) and anti-bribery policies (ABP) were instrumented. The results are affected by the choice ofcorruption index. However, they show that the adoption of anti-bribery policies at firm levelattracts more analysts to follow a firm, in line with our hypothesis. However, firmsoperating in countries with higher levels of corruption appear to attract higher analystcoverage, consistent with the importance of the monitoring role of analyst coverage inhighly corrupt countries, only if the CPI is used. Once the CPI is estimated by instrumentalvariables, the relationship is in line with the transaction cost hypothesis that analystcoverage increases in less corrupted countries. The same conclusion applies to the casewhen the variables corruption at country level and anti-bribery policies are interacted. Therelationship is strongly positive and significant in the case of the instrumented CPI.

The findings of this paper have several important implications. First, the results of the currentstudy provide robust evidence that proactive actions by firms, such as the adoption of anti-bribery policies, can initiate or strengthen a positive process of increased analyst coverageleading to higher degrees of monitoring. Given the significant role played by analysts inuncovering fraud, as documented by Dyck et al. (2010), this is an important finding. Second, thispaper also contributes to the extant literature on corruption by providing empirical evidence onthe consequences of corruption at firm level, rather than the causes of corruption. This is animportant contribution because there is a serious imbalance between the number of papersexamining the causes of corruption than its consequences as observed by Judge et al. (2011).Third, this study also contributes to the literature on the determinants of analyst coverage byextending the determinants of analysts following beyond firm characteristics to include firmpolicy and institutional factors. This contribution is particularly crucial at a time wheninvestment firms are now required to clearly state the costs of any investment advice they mayprovide because of the recent introduction of the new regulatory framework contained in theEuropean Union Market in Financial Instruments Directive and Regulation (MiFID II).Investment firms are required to inform clients whether the investment advice is provided on anindependent basis to ensure that clients are not sold inappropriate products, and thatinducements are only allowed if they benefit the client (Lannoo, 2017). The paper also sheds animportant light on the consequences of the adoption of anti-corruption measures at firm levelsuch as anti-bribery policies. One practical implication is to encourage firms to adopt anti-corruption measures to reduce the incidence of corruption and increase analyst coverage. Thisresult is particularly timely and practically relevant since the new European Directive 2014/95/EU requires large companies to include non-financial information about anti-corruption andbribery in their annual reports from 2018 onwards. A political implication is to expand efforts toproduce reliable and consistent controls of corruption across companies and countries, perhapsthrough the introduction of compulsory anti-briberymeasures. Finally, the paper establishes boththeoretical and empirical links between two strands of research, i.e. analyst coverage andcorruption, which may provide new material to incorporate in the syllabus of some courses infinance and motivate future research on the determinants and consequences of the adoption ofanti-corruption and briberymeasures at firm level.

Notes

1. However, analyst coverage may also have some negative effects on firm performance. Forinstance, He and Tian (2013) showed that firms covered by a larger number of analysts generatefewer patents or patents with lower impact. The reason is that innovation involves a lengthyprocess that is full of uncertainty and has a high probability of failure, making the firm morevulnerable to, for example, hostile takeovers. As a reaction, managers tend to invest less in

Analystcoverage

319

innovation and prioritise routine tasks that offer quicker and more certain returns, possiblysacrificing the long-term success of the firm.

2. This gives rise to a potential ethical dilemma if the perceived norms, rules and regulations in dealingwith corruption are different between the home and host countries. An example is a multinationalbusiness that has its headquarter in a highly clean country, but its subsidiaries operate in corruptcountries where the social norms about corruption could be very different. Although this is aninteresting area for research, the focus in the current study is the anti-corruption policies thatcompanies adopt in their respective home countries and their impact on analyst coverage.

3. We would like to thank an anonymous referee for this suggestion. The results of this extraregression are not included in the paper for reasons of space, but are available, upon request,from the authors.

4. Auditor size is often used in prior studies as a surrogate of audit quality (Tendeloo and Vanstraelen,2008; Chen et al., 2011). This line of research might be interpreted as either larger auditors providehigher-quality audits, as they have more wealth and reputation at risk (DeAngelo, 1981), or smallerauditors supplying unacceptably low levels of audit quality (DeFond and Francis, 2005). Smaller auditfirms report less conservatively (issue fewer non-clean audit reports), and their clients are more likely tohave abnormal accruals, which is suggestive of more aggressive earnings management.

ReferencesAbed, A., Abdallah, W. and Ismail, A. (2012), “Do accounting standards matter to financial analysts?

An empirical analysis of the effect of cross-listing from different accounting standards regimeson analyst following and forecast error”, The International Journal of Accounting, Vol. 47 No. 2,pp. 168-197.

Adam, A.M. and Rachman-Moore, D. (2004), “The methods used to implement an ethical code ofconduct and employee attitudes”, Journal of Business Ethics, Vol. 54 No. 3, pp. 225-244.

Alam,M.S. (1995), “A theory of limits on corruption and some applications”,Kyklos, Vol. 48No. 3, pp. 419-435.Anantharaman, D. and Zhang, Y. (2011), “Cover me: managers’ responses to changes in analyst coverage

in the Post-Regulation FD period”,TheAccounting Review, Vol. 86 No. 6, pp. 1851-1885.Arbel, A., Carvell, S. and Strebel, P. (1983), “Giraffes, institutions and neglected firms”, Financial

Analysts Journal, pp. 2-8.Baum, J.A.C. and Oliver, C. (1991), “Institutional linkages and organizational mortality”,Administrative

Science Quarterly, Vol. 36 No. 2, pp. 187-218.Bertrand, M. and Mullainathan, S. (2001), “Do people mean what they say? Implications for subjective

survey data”,The American Economic Review, Vol. 91 No. 2, pp. 67-72.Bhushan, R. (1989), “Firm characteristics and analyst following”, Journal of Accounting and Economics,

Vol. 11 Nos 2/3, pp. 255-274.Boubaker, S. and Labégorre, F. (2008), “Ownership structure, corporate governance, and analyst following:

a study of French listed firms”, Journal of Banking and Finance, Vol. 32 No. 6, pp. 961-976.Boubakri, N. and Bouslimi, L. (2010), “Analyst following of privatized firms around the world: the role

of institutions and ownership structure”, The International Journal of Accounting, Vol. 45 No. 4,pp. 413-442.

Bushee, B.J. and Miller, G.S. (2012), “Investor relations, firm visibility, and investor following”, TheAccounting Review, Vol. 87 No. 3, pp. 867-897.

Bushman, R.M., Piotroski, J.D. and Smith, A.J. (2004), “What determines corporate transparency?”,Journal of Accounting Research, Vol. 42 No. 2, pp. 207-252.

Carr-Howard, M. (2014), “Global anti-corruption laws: an executive summary, dentons”, available at: file:///C:/Users/Omaima/Downloads/Global%20AntiCorruption%20Laws%20Dentons%20(3).pdf (accessed 5January 2018).

MAJ34,3

320

Chen, C.J., Ding, Y. and Kim, C.F. (2010), “High-level politically connected firms, corruption, and analystforecast accuracy around the world”, Journal of International Business Studies, Vol. 41 No. 9,pp. 1505-1524.

Chen, F., Weiss, I.S. and Zheng, L. (2007), “The predictive role of analyst coverage intensity: evidencefrom cross-listing in the US”, Journal of Contemporary Accounting and Economics, Vol. 3 No. 1,pp. 34-57.

Chen, H., Chen, J.Z., Lobo, G.J. and Wang, Y. (2011), “Effects of audit quality on earnings managementand cost of equity capital: evidence from China”, Contemporary Accounting Research, Vol. 28 No.3, pp. 892-925.

Chen, T., Harford, J. and Lin, C. (2015), “Do analysts matter for governance? Evidence from naturalexperiments”, Journal of Financial Economics, Vol. 115 No. 2, pp. 383-410.

Cliff, M.T. and Denis, D.J. (2004), “Do initial public offering firms purchase analyst coverage withunderpricing?”,The Journal of Finance, Vol. 59 No. 6, pp. 2871-2901.

Cuervo-Cazurra, A. (2006), “Who cares about corruption?”, Journal of International Business Studies,Vol. 37 No. 6, pp. 807-822.

Cuervo-Cazurra, A. (2016), “Corruption in international business”, Journal of World Business, Vol. 51No. 1, pp. 35-49.

Dal B�o, E. and Rossi, M.A. (2007), “Corruption and inefficiency: theory and evidence from electricutilities”, Journal of Public Economics, Vol. 91 Nos 5/6, pp. 939-962.

Davis, J.H. and Ruhe, J.A. (2003), “Perceptions of country corruption: antecedents and outcomes”,Journal of Business Ethics, Vol. 43 No. 4, pp. 275-288.

DeAngelo, L. (1981), “Auditor size and quality”, Journal of Accounting and Economics, Vol. 3 No. 3,pp. 183-199.

DeFond, M.L. and Francis, J.R. (2005), “Audit research after Sarbanes–Oxley”, Auditing: A Journal ofPractice and Theory, Vol. 24, pp. 5-30.

Doh, J.P., Rodriguez, P., Uhlenbruck, K., Collins, J. and Eden, L. (2003), “Coping with corruption inforeignmarkets”,Academy ofManagement Perspectives, Vol. 17 No. 3, pp. 114-127.

Donchev, D. and Ujhelyi, G. (2014), “What do corruption indices measure?”, Economics and Politics,Vol. 26 No. 2, pp. 309-331.

Doukas, J.A., Kim, C. and Pantzalis, C. (2000), “Security analysis, agency costs, and companycharacteristics”, Financial Analysts Journal, Vol. 56 No. 6, pp. 54-63.

Dyck, A., Morse, A. and Zingales, L. (2010), “Who blows the whistle on corporate fraud?”, The Journalof Finance, Vol. 65 No. 6, pp. 2213-2253.

Easley, D. and O’hara, M. (2004), “Information and the cost of capital”, The Journal of Finance, Vol. 59No. 4, pp. 1553-1583.

Fan, C.S., Lin, C. and Treisman, D. (2009), “Political decentralization and corruption: evidence fromaround the world”, Journal of Public Economics, Vol. 93 Nos 1/2, pp. 14-34.

Giraldo, M. (2011), “Dynamics of analysts’ coverage and the firms’ information environment”,International Review of Financial Analysis, Vol. 20 No. 5, pp. 345-354.

Gordon, K. and Miyake, M. (2001), “Business approaches to combating bribery: a study of codes ofconduct”, Journal of Business Ethics, Vol. 34 Nos 3/4, pp. 161-174.

Gotti, G., Han, S., Higgs, J.L. and Kang, T. (2012), “Managerial stock ownership, analyst coverage, andaudit fee”, Journal of Accounting, Auditing and Finance, Vol. 27 No. 3, pp. 412-437.

Hassan, O.A.G. and Giorgioni, G. (2015), “Analyst coverage, corruption and financial secrecy: a multi-country study”, Working Paper, available at: https://ssrn.com/abstract=2566848; http://dx.doi.org/10.2139/ssrn.2566848 (accessed 10 January 2018).

Hassan, O.A.G. and Skinner, F. (2016), “Analyst coverage: does the listing location really matter?”,International Review of Financial Analysis, Vol. 46, pp. 227-236.

Analystcoverage

321

Hauser, C. and Hogenacker, J. (2014), “Do firms proactively take measures to prevent corruption in theirinternational operations?”, EuropeanManagement Review, Vol. 11 Nos 3/4, pp. 223-237.

Huang, A.H., Lehavy, R., Zang, A.Y. and Zheng, R. (2017), “Analyst information discovery and interpretationroles: a topicmodeling approach”,Management Science, doi: 10.1287/mnsc.2017.2751.

He, J.J. and Tian, X. (2013), “The dark side of analyst coverage: the case of innovation”, Journal ofFinancial Economics, Vol. 109 No. 3, pp. 856-878.

Healy, P.M. and Serafeim, G. (2016), “An analysis of firms’ self-reported anti-corruption efforts”, TheAccounting Review, Vol. 91 No. 2, pp. 489-511.

Hsiao, C. (2002),Analysis of Panel Data, 2nd ed., Cambridge University Press, London.Hussain, S. (2000), “Simultaneous determination of UK analyst following and institutional ownership”,

Accounting and Business Research, Vol. 30 No. 2, pp. 111-124.Jain, A.K. (2001), “Corruption: a review”, Journal of Economic Surveys, Vol. 15 No. 1, pp. 71-121.Jong, G. and Ees, H. (2014), “Firms and corruption”, European Management Review, Vol. 11 Nos 3/4,

pp. 187-190.Judge, W.Q., Mcnatt, D.B. and Xu, W. (2011), “The antecedents and effects of national corruption: a

meta-analysis”, Journal ofWorld Business, Vol. 46 No. 1, pp. 93-103.Jung, B., Sun, K.J. and Yang, Y.S. (2012), “Do financial analysts add value by facilitating more

effective monitoring of firms’ activities?”, Journal of Accounting, Auditing and Finance,Vol. 27 No. 1, pp. 61-99.

Khalil, S., Saffar, W. and Trabelsi, S. (2015), “Disclosure standards, auditing infrastructure, and briberymitigation”, Journal of Business Ethics, Vol. 132 No. 2, pp. 379-399.

Karpoff, J.M., Lee, D.S. and Martin, G.S. (2008), “The cost to firms of cooking the books”, Journal ofFinancial and Quantitative Analysis, Vol. 43 No. 3, pp. 581-611.

Kehoe, W.J. (1998), “The environment of ethics in global business”, Journal of Business and BehavioralSciences, Vol. 2, pp. 47-56.

Kim, J.B. and Shi, H. (2012), “Voluntary IFRS adoption, analyst coverage, and information quality:international evidence”, Journal of International Accounting Research, Vol. 11 No. 1, pp. 45-76.

Kirk, M. (2011), “Research for sale: determinants and consequences of paid-for analyst research”,Journal of Financial Economics, Vol. 100 No. 1, pp. 182-200.

Knack, S. (2007), “Measuring corruption: a critique of indicators in Eastern Europe and Central Asia”,Journal of Public Policy, Vol. 27 No. 3, pp. 255-291.

Lang, M., Lins, K.V. and Maffett, M. (2012), “Transparency, liquidity, and valuation: international evidenceonwhen transparencymattersmost”, Journal of Accounting Research, Vol. 50 No. 3, pp. 729-774.

Lannoo, K. (2017), “MiFID II and the new market conduct rules for financial intermediaries: will complexitybring transparency?”, European Capital Markets Institute, available at: www.ceps.eu/publications/new-market-conduct-rules-financial-intermediaries-will-complexity-bring-transparency (accessed 27July 2018).

Li, K.K. and You, H. (2015), “What is the value of sell-side analysts? Evidence from coverage initiationsand terminations”, Journal of Accounting and Economics, Vol. 60 Nos 2/3, pp. 141-160.

Long, J.S. and Freese, J. (2006), Regression Models for Categorical Dependent Variables Using Stata,Stata Press, TX.

Merton, R. (1987), “A simple model of capital market equilibrium with incomplete information”, Journalof Finance, Vol. 42 No. 3, pp. 483-510.

Mola, S., Rau, P.R. and Khorana, A. (2013), “Is there life after the complete loss of analyst coverage?”,The Accounting Review, Vol. 88 No. 2, pp. 667-705.

OECD (2014), Foreign Bribery Report: An Analysis of the Crime of Bribery of Foreign Public Officials, OECDPublishing, available at: http://dx.doi.org/10.1787/9789264226616-en (accessed 30 February 2018).

MAJ34,3

322

Olken, B.A. (2009), “Corruption perceptions vs corruption reality”, Journal of Public Economics, Vol. 93Nos 7/8, pp. 950-964.

Razafindrakoto, M. and Roubaud, F. (2010), “Are international databases on corruption reliable? acomparison of expert opinion surveys and household surveys in Sub-Saharan Africa”, WorldDevelopment, Vol. 38 No. 8, pp. 1057-1069.

Rock, S., Sedo, S. and Willenborg, M. (2001), “Analyst following and count-data econometrics”, Journalof Accounting and Economics, Vol. 30 No. 3, pp. 351-373.

Rodriguez, P., Siegel, D.S., Hillman, A. and Eden, L. (2006), “Three lenses on the multinationalenterprise: politics, corruption, and corporate social responsibility”, Journal of InternationalBusiness Studies, Vol. 37 No. 6, pp. 733-746.

Sibilkov, V., Straska, M. and Waller, H.G. (2013), “Do firms use M&A business to pay for analystcoverage?”, Financial Review, Vol. 48, pp. 725-751.

Svensson, J. (2003), “Whomust pay bribes and howmuch? Evidence from a cross section of firms”,TheQuarterly Journal of Economics, Vol. 118 No. 1, pp. 207-230.

Tendeloo, B.V. and Vanstraelen, A. (2008), “Earnings management and audit quality in Europe: evidencefrom the private client segmentmarket”,EuropeanAccounting Review, Vol. 17 No. 3, pp. 447-469.

Transparency International (2009), Transparency in Reporting on Anti-Corruption: A Report onCorporate Practices, Transparency International, Berlin.

Uhlenbruck, K., Rodriguez, P., Doh, J. and Eden, L. (2006), “The impact of corruption on entry strategy:evidence from telecommunication projects in emerging economies”, Organization Science,Vol. 17 No. 3, pp. 402-414.

Wilhelm, P.G. (2002), “International validation of the corruption perceptions index: implications for businessethics and entrepreneurship education”, Journal of Business Ethics, Vol. 35 No. 3, pp. 177-189.

Wu, X. (2005), “Corporate governance and corruption: a cross-country analysis”, Governance: AnInternational Journal of Policy, Administration, and Institutions, Vol. 18 No. 2, pp. 151-170.

Young, S.M. and Peng, E.Y. (2013), “An analysis of accounting frauds and the timing of analystcoverage decisions and recommendation revisions: evidence from the US”, Journal of BusinessFinance and Accounting, Vol. 40 Nos 3/4, pp. 399-437.

Yu, M. (2010), “Analyst forecast properties, analyst following and governance disclosures: a globalperspective”, Journal of International Accounting, Auditing, andTaxation, Vol. 19 No. 1, pp. 1-15.

Zhao, H., Kim, S. and Du, J. (2003), “Corruption, transparency and foreign direct investment: a cross-country analysis”,Management International Review, Vol. 43 No. 1, pp. 41-63.

Further readingLang, M.H. and Lundholm, R.J. (1996), “Corporate disclosure policy and analyst behavior”, The

Accounting Review, pp. 467-492.Serafeim, G. (2014), “Firm competitiveness and detection of bribery”, Working Paper 14-012, Harvard

Business School, available at: www.hbs.edu/faculty/Publication%20Files/14-012_42a7455b-4a8a-4393-a16a-18b0de5278ba.pdf (accessed 5 July 2017).

Corresponding authorOmaima Hassan can be contacted at: o.hassan@rgu.ac.uk

For instructions on how to order reprints of this article, please visit our website:www.emeraldgrouppublishing.com/licensing/reprints.htmOr contact us for further details: permissions@emeraldinsight.com

Analystcoverage

323

Financial fraud detection andbig data analytics – implications

on auditors’ use of fraudbrainstorming session

Jiali TangDepartment of Accounting, University of Hartford,

West Hartford, Connecticut, USA, and

Khondkar E. KarimDepartment of Accounting, Robert J. Manning School of Business,University of Massachusetts Lowell, Lowell, Massachusetts, USA

AbstractPurpose – This paper aims to discuss the application of Big Data analytics to the brainstorming session inthe current auditing standards.

Design/methodology/approach – The authors review the literature related to fraud, brainstormingsessions and Big Data, and propose a model that auditors can follow during the brainstorming sessions byapplying Big Data analytics at different steps.

Findings – The existing audit practice aimed at identifying the fraud risk factors needs enhancement, dueto the inefficient use of unstructured data. The brainstorming session provides a useful setting for suchconcern as it draws on collective wisdom and encourages idea generation. The integration of Big Dataanalytics into brainstorming can broaden the information size, strengthen the results from analyticalprocedures and facilitate auditors’ communication. In the model proposed, an audit team can use Big Datatools at every step of the brainstorming process, including initial data collection, data integration, fraudindicator identification, groupmeetings, conclusions and documentation.

Originality/value – The proposedmodel can both address the current issues contained in brainstorming (e.g.low-quality discussions and production blocking) and improve the overall effectiveness of fraud detection.

Keywords Big data analytics, Brainstorming session, Fraud detection

Paper type Research paper

1. IntroductionFinancial fraud remains one of the most discussed topics in accounting literature. Accordingto Cotton (2002), the financial scandals of Enron, WorldCom, Qwest, Global Crossing andTyco resulted in approximately $460bn loss. The detection of financial fraud, therefore, hasbecome a critical task for accounting practitioners.

In the fraud triangle put forward by Cressey (1973), three factors determine the likelihoodof fraud occurrence, including pressure, opportunity and rationalization. The core of thesefactors lies in people’s belief and behavior. Due to the unpredictability and uncertainty infraudsters’ incentives and techniques, fraud detection requires the skillset that encompassesboth diligence and judgment.

Although the current auditing standards intend to provide a comprehensive guidelinethat governs the process in fraud examination, the actual implementation is conducted on a

MAJ34,3

324

Received 16 January 2018Revised 14May 201813 July 2018Accepted 6 August 2018

Managerial Auditing JournalVol. 34 No. 3, 2019pp. 324-337© EmeraldPublishingLimited0268-6902DOI 10.1108/MAJ-01-2018-1767

The current issue and full text archive of this journal is available on Emerald Insight at:www.emeraldinsight.com/0268-6902.htm

case-by-case basis that heavily relies on auditor judgment thus leads to inconsistent successrate. For example, AS 2401 contains explanations on understanding fraud, exercisingprofessional skepticism, responding to fraud risk, communicating with the managementand documenting auditors’ comments[1]. However, these standards might be interpretedand executed differently by auditors, and they cannot effectively address the fraud riskembedded in nonfinancial information (e.g. meeting content, management conversations,tone at the top, language in annual reports, etc.) without competent personnel andsupplementary tools.

A recent trend is to apply data analytics to fraud detection. An example is the use of datamining techniques aimed at finding patterns from journal entries to identify fraud(Debreceny and Gray, 2010). While the study contributes greatly to the research on frauddetection method by integrating data and technology, it still ignores the non-numericalsignals from parties who prepare the financial statements.

To cope with the flaws mentioned above and provide an addition to the fraud detectiontoolset, in this paper, we discuss the potential use of Big Data analytics in identifying fraudrisk, particularly in the brainstorming sessions required by current auditing standards. Inresponse to the accounting scandals in the early 2000s, an auditing statement aimed atdetecting fraud (SAS No.99) was established by AICPA. The standard specifically requires abrainstorming session to be held by auditors to identify fraud-related risks. More recently,the idea of brainstorming has been added to multiple sections of the auditing standards.AICPA mentions in AU-C 240 that the engagement team is required to brainstorm anddiscuss areas potentially subject to material misstatement, management’s fraudulentreporting and asset misappropriation. AU-C 315 of AICPA considers such discussions asauditors’ risk assessment activities. Similar requirement is also included in PCAOBstandards (AS 2110).

As the purpose of brainstorming is to gather intellectual input from different individualsand inspire new ideas, such format seems to be an effective way of processing unstructureddata and capturing anomaliesWhile we believe that brainstorming sessions can enhance theoverall fraud assessment, the execution can be quite challenging given the complexity inconducing quality discussions and the limitations contained in the format of brainstorming(e.g. production blocking)[2].

The integration of Big Data analytics provides one solution to improving theperformance of the brainstorming sessions. First, Big Data can enlarge the information baseused in brainstorming. By combining or aggregating different types of information throughBig Data tools, auditors can have access to a database that contains both the financial (e.g.accounting record) and nonfinancial information (e.g. news on management, board meetingminutes, contract details, etc.) of the client firm. Second, Big Data can enhance theinformation content. When conducting analytical procedures, auditors can efficientlycompare data across time and industries to quickly identify anomalies. A larger sample data(or the full population) will also increase the accuracy of the prediction models. Thus, BigData can generate reliable results that more precisely point to the fraud risks. Finally, BigData can facilitate the communications among the engagement team members, or evenbetween the predecessor and successor auditors. For example, during the brainstormingsessions, auditors can use electronic devices to record their thoughts while reading othermembers’ comments simultaneously. In addition, Big Data can also incorporate the industryexpertise of individual auditors by selectively displaying relevant information (e.g. news,industry index, competitors) on the monitor to inspire new ideas. Overall, the application ofBig Data analytics in brainstorming sessions allows auditors to use unstructured data andanalyze fraud factors closely related to the fraud triangle. The larger information set and

Financialfraud detection

325

more reliable evidence help ensure quality discussions, and the computer-based setting canreduce production blocking and redundant process.

To strengthen the practical contribution of our arguments, we propose a model that canbe adopted by the audit team engaged in the brainstorming sessions. We suggest a six-stepsystem that includes initial data collection, data integration, fraud indicator identification,group meetings and discussions, drawing conclusions and documentation. We discuss thepossible application of Big Data analytics to each step of the process. The model offers aninnovative and potentially effective approach of conducting brainstorming and identifyingfraud risks.

The remainder of the paper is organized as follows. Section 2 presents the background offraud detection, Section 3 explains the brainstorming sessions in current auditing standards,Section 4 discusses the application of Big Data analytics, and Section 5 concludes.

2. Financial fraud and issues in its detectionBased on different behavior and consequences, Reurink (2016) classifies financial fraud intothree types: financial statement fraud, financial scams and fraudulent financial mis-selling.Of the three categories, financial statement fraud is closely related to bookkeeping and day-to-day transactions as it involves intentional falsification of financial records, inappropriateapplication of accounting standards, earnings management and other activities that createinformation asymmetry (Rezaee, 2005). Errors and misrepresentations in accounting recordoften cause financial restatements. GAO (2002) reports that approximately 10 per cent of thepublic firms underwent financial restatements during 1997 to 2002. For the other twocategories, financial scams refer to the deceptive schemes designed to unlawfully obtainother parties’ assets, and financial mis-selling includes deliberately marketing financialproducts to the unsuitable clients (Reurink, 2016).

Financial fraud has led to tremendous loss over the years. Karpoff et al. (2008) show thatfinancial misconduct is penalized by both the legal system and the market, and the latterusually imposes a significantly stronger penalty. Of the 585 firms investigated in theirstudy, the loss incurred in the market due to reputation damage is 7.5 times higher than thelegal penalty. GAO (2002) finds that the announcement of restatements triggers a negativemarket reaction of�10 per cent. Stockholders also respond unfavorably to corporate crimes,especially for companies with similar record previously (Davidson et al., 1994). Theconsequences of fraudulent activities are not limited to financial loss. Tian et al. (2016) showthat venture capitalists will face difficulty in obtaining future IPO business if they overlookthe fraud in prior cases.

The factors that determine fraud likelihood has been discussed extensively in theliterature. Theoretically, Cressey (1973) argues that an individual or organization needs topossess the pressure, opportunity and rationalization (the fraud triangle) to conductfraudulent activities. Later research proposed a fourth factor, capability, and presented thefraud diamond structure (Wolfe and Hermanson, 2004; Boyle et al., 2015). Power (2013)states that fraud has been considered as a risk rather than an event and discusses thecurrent shift to a universal acknowledgement of fraud risk within organizations. Otherfactors that have contributed to fraudulent activities include morality (Morales et al., 2014),competition (Bolton et al., 2007), conflicts of interest (Mehran and Stulz, 2007), option-basedcompensation (Burns and Kedia, 2006; Efendi et al., 2007), clawback adoptions and insidersales (Fung et al., 2015), the knowledge and trust embedded in accounting positions(Dellaportas, 2013) and lack of scrutiny on internal control systems (Baker et al., 2017).

Due to the uncertainty and variability in these factors, fraud detection has become achallenging task that requires both talent and technology. While current auditing standards

MAJ34,3

326

(AU 316) put a strong emphasize on exercising professional skepticism, they are perhapsinsufficient in addressing the amount of ambiguity and variations in different individualsand organizations. For example, Davis and Pesch (2013) find that the auditors cannot treatfraud detection uniformly without adjustment, and that the detection mechanisms areinfluenced by organization types and individual social awareness. The incentive toinvestigate fraud also depends how the reports are framed (Huerta et al., 2012). Trotman andWright (2012) show that auditors might evaluate external evidence related to fraud based onwhether the evidence clashes with business goals, resulting in potential subjectivity.Further, internal auditors can experience stress from reporting to audit committees and alterthe level of fraud risk when reporting (Norman et al., 2010). Knapp and Knapp (2001) showthat auditors’ experience matters in adopting analytical procedures during fraudexamination.

To deal with the complexity in fraud detection and enhance effectiveness, different typesof technologies are developed and implemented. Some of the highly discussed technologyapplications include using data mining techniques to find patterns from financial records(Debreceny and Gray, 2010; Grabski, 2010; Gray and Debreceny, 2014), using descriptivedata mining tools to identify internal fraud risks (Jans et al., 2010), using outlier techniquesto flag fraudulent insurance claims (Capelleveen et al., 2016), using computer algorithms todetect abnormal stock price movement (Williams, 2013) and applying natural languageprocessing (NLP), queen genetic algorithm (QGA) and support vector machine (SVM) toanalyze annual reports (Chen et al., 2017).

3. Brainstorming sessions in current auditing standardsStatement on Auditing Standards No. 99 (SAS 99) is a set of standards established in 2002after the occurrence of the accounting scandals. The statement aims at more effectivelypreventing and detecting fraudulent activities. One of the highlights of SAS 99 is therequirement of a brainstorming session held by the audit team members. Since then, thestandards have been updated and the idea of implementing a brainstorming session hasbeen shown in different parts of AICPA (AU-C 240 and AU-C 315) and PCAOB (AS 2110)standards. According to AU-C 240, the content of the brainstorming sessions should includethe occurrence of potential material misstatement, the methods that management might useto commit fraud and the possibility of asset misappropriation. In particular, the sessionshould address management behavior in the context of the fraud triangle (incentive,opportunity and rationalization) and a high degree of professional skepticism must bemaintained[3]. Trompeter et al. (2013) place brainstorming in the framework of auditors’ rolein detecting fraud and argue that auditors should consider fraudulent act, concealment andconversion methods during the brainstorming sessions. Based on a survey conducted on 22auditors by Bellovary and Johnstone (2007), the brainstorming sessions can be summarizedas a four-step process, including client information review, fraud triangle analysis, fraudlikelihood assessment and audit response indication.

Prior studies have documented several benefits of conducting brainstorming. Osborn(1957) argues that group interactions create stimulation and synergy. Carpenter (2007)concludes that brainstorming auditors can generate new and quality ideas on frauddetection and higher risk assessment relative to individual auditors. Brazel et al. (2010) statethat fraud risk factors (incentive, opportunity and rationalization) can more effectivelyindicate fraud likelihood when the brainstorming is high-quality. Hoffman and Zimbelman(2009) find that both strategic reasoning and group brainstorming facilitate the auditprocedure modifications.

Financialfraud detection

327

However, there is also evidence that group interactions cause process loss (Hill, 1982;Straus et al., 2011). Nunamaker et al. (1991) list several possible sources of process loss frombrainstorming, and three sources are most discussed in the literature, including productionblocking, evaluation apprehension and free riding (Dennis and Valacich, 1993). Based on theexperiments conducted in Diehl and Stroebe (1987), production blocking appears to be themain cause. Paulus et al. (1995) find that group brainstorming can diminish productivityeven when members are well-trained in group work.

A further line of research investigates face-to-face and electronic brainstorming. Overall,the studies are in favor of brainstorming conducted electronically. Lynch et al. (2009) showthat computer-mediated brainstorming sessions enhance the effectiveness of fraud riskassessment. Specifically, electronic brainstorming can mitigate concerns such as productionblocking, evaluation apprehension and free riding, and strengthen synergy duringinteractions (Dennis and Valacich, 1993). Additionally, Smith et al. (2012) state that thesuperior performance of electronic brainstorming is mainly due to stronger task focus andlonger comments made by auditors. In particular, computed-based groups and groupsupport systems (GSS) have demonstrated strong performance in tasks related to idea-generation (Valacich et al., 1994; Fjermestad and Hiltz, 1998).

Literature has also shown some benefits of face-to-face brainstorming. For example,Cockrell and Stone (2011) find that face-to-face brainstorming encourages more in-depthdiscussion that leads to better performance than electronic format. Brazel et al. (2004) showthat face-to-face audit workpaper review is associated with more accountability and higherquality judgment from the preparers relative to the review in electronic mode. Overall, face-to-face brainstorming seems to suffer more from process loss, particularly productionblocking, while electronic brainstorming might result in low-quality dialogue and judgment.

4. Application of big data analyticsBased on the above, financial fraud is the product of internal and external factors thatinvolve incentives, opportunities and rationalization. The detection mechanism iscomplicated and tends to overlook clues from non-financial information. The auditingstandards introduce a critical mechanism in fraud assessment, the brainstorming sessions,which allow auditors to draw on collective wisdom and generate ideas. In this section, wediscuss the integration of Big Data analytics into the brainstorming sessions.

The meaning of such integration is twofold. One is to propose detailed procedures thatcan make use of the non-financial information in detecting fraud. The fraud triangleemphasizes incentives, opportunity and rationalization, all of which involve a high degree ofuncertainty and cannot be identified simply through the numbers on the financialstatements. Big Data analytics can be useful in this dimension by incorporatingunstructured data and providing more reliable results. The other purpose is to address theconcerns in both face-to-fact and electronic brainstorming. As mentioned above, theeffectiveness of brainstorming can be compromised due to production blocking and lack ofin-depth discussions. The use of Big Data can, at least to some extent, mitigate theseproblems and eventually enhance the effectiveness of fraud detection.

Big Data is usually defined as possessing the characteristics of four Vs, namely, Volume,Velocity, Variety, Veracity (Zhang et al., 2015), and requires the support of large, complexinformation systems (Vasarhelyi et al., 2015). Yoon et al. (2015) argue that Big Data providesufficient, reliable and relevant information that should be considered as complementaryaudit evidence. Big data also increase audit quality due to its ability to enable population-based audit compared to sample-based (Ramlukan, 2015). While the application of Big Datain accounting and auditing practice has been introduced before (Tang and Karim, 2017), the

MAJ34,3

328

focus on brainstorming sessions remains undiscussed and could generate usefulimplications.

We propose three channels through which Big Data can be used in brainstorming. First,Big Data can combine structured and unstructured data. A key part of the brainstormingsession is to review and analyze the client’s prior information in the framework of the fraudtriangle. Thus, the information should be presented in various formats that can effectivelyindicate fraud risk factors. Prior studies support the use of nonfinancial measures (Brazelet al., 2009) and text-mining method (Humphreys et al., 2011) in fraud assessment.

For example, the news related to management, board members, employees and company-wide reputation can be selectively combined with the firms’ financial record to provide morecomplete evidence. The observation that a CEO increases personal spending on luxuriousitems when the company is incurring a loss and layoff should serve as a potential fraud risk.Board meeting minutes can also be extracted and matched with the actual outcomes.Auditors can check whether the agenda is indeed carried out by gathering evidence fromdifferent sources. In addition, Big Data can help auditors examine alumni relationships,especially those between managers and audit committee members. Information derivedfrom social media can assist auditors in identifying appointments or new positions obtainedthrough personal network, which often leads to fraudulent activities. The messages fromanonymous whistleblowers can also be analyzed as an additional source of information.Kaplan et al. (2012) show that non-anonymous whistleblowers are discouraged fromreporting when they witness no repercussions to the fraudsters. With the help of Big Datatechnology, the brainstorming sessions can combine the whistleblowers’ information withdata in other formats, which retains the confidentiality and increase evidence completeness.

In short, Big Data provides a larger and more resourceful information base that enablesauditors to efficiently translate audit evidence into fraud risk factors. The integration ofunstructured data also enhances the veracity of audit evidence. The reason is that datacollected in such large quantity and from very different sources offer more completeevidence compared to the traditional data and are subject to less management manipulation.

Second, Big Data can enhance the performance of analytical procedures, thus yieldingmore insightful and reliable results for decision-making. In trend/pattern analysis, allaccounting data can be pooled and compared across years and industries. Perols and Lougee(2011) find that earnings management in prior years is an indicator for committing fraud.Accordingly, auditors can conduct regression analysis to find associations on the populationlevel to increase the model’s explanatory power. Big Data can also be useful in identifyinganomalies. Kedia and Philippon (2009) show that firms involve in earnings managementtend to invest aggressively to match the performance of the better firms in the industry.Auditors can use computer algorithms to detect any sudden change in profits or revenuethat falls out of the normal range of the industry average. PCAOB (2007) states that theperformance of analytical procedures can be strengthened if the data analyzed are subject toless manipulation. By gathering timely raw data from real-life sources, auditors arepresented with more reliable information thus improving the relevance of the ideasgenerated during the brainstorming process. That is, Big Data analytics tools cansignificantly contribute to the brainstorming sessions if they are designed to provide timelyand accurate feedback upon auditors’ request.

Finally, Big Data can facilitate the communications during brainstorming sessions.Building on the idea of electronic brainstorming, Big Data can further incorporatecomputerized programs to record fraud discussions, trace the chain of thoughts andestablish possible scenarios. Auditors can write down longer comments that include photos,flowcharts and videos to express their ideas. Devices can be used to simultaneously record

Financialfraud detection

329

all auditors’ comments and present them on the same screen. The devices can beprogrammed to combine the unstructured data and group auditors’ comments intocategories. Analysis, such as descriptive data mining, can then be performed within eachcategory to generate conclusions. Therefore, Big Data analytics not only eliminates the waittime for speaking during discussions, a common issue in face-to-face brainstorming, but alsoallows auditors to quickly use dialogue in idea generation.

In addition, Big Data can collect the comments from both predecessor and currentauditors and draw on prior fraud cases occurred in real life. This allows auditors to trace thethought pattern of potential fraudsters and construct hypothetical fraud scenarios. Further,Big Data tools can research on each auditor’s industry background and work experience.Displaying information such as news reports and industry index can provide auditors witha sense of familiarity and inspires logical and creative thinking.

To summarize, the above procedures require careful design and an efficient combinationof Big Data, technology and accounting professionals. Such integration can help mitigateconcerns found in brainstorming (e.g. production blocking and low-quality discussions) anduse non-financial and unstructured data to better identify fraud factors within theframework of the fraud triangle.

Following the advantages mentioned above, we propose a model that can effectivelyincorporate Big Data analytics in the brainstorming sessions. The model contains six steps,and the performance at each step can be enhanced by the suggested application of Big Datatools, as shown in Figure 1.

Step 1 and Application 1: The brainstorming process should begin with initial datacollection that helps uncover potential fraud indicators, which later serve as the outline forthe group meetings. The system can automatically generate pre-saved data queries thatcontain the company’s basic information, such as business operations, financial statements,prior audit results, board composition and recent media coverage. Further data requests forspecific items can be submitted by members of the audit team based on their experience, andthe system will populate the results along with the original request. For example, an auditteammember might suspect the existence of a close relationship with the supplier could leadto bribery or kickbacks. Therefore, the detailed purchase history with the supplier should be

Figure 1.Proposedmodel forthe application of bigdata analytics to thebrainstormingsession

MAJ34,3

330

added to the query. If a firm’s revenue performance consistently beats external expectation, itcould be a signal of earnings overstatement and source documents that support the salestransactions should be collected. In short, Big Data analytics can facilitate the data collection taskby searching and storing a large amount of data in different formats.

Step 2 and Application 2: data in various formats collected in Step 1 are processed. Themain objective of this step is to combine the structured and unstructured data so a unifiedversion of suitable evidence can be presented. For example, surveillance video can becombined with inventory record to detect risk in inventory theft. The software can alsomerge data of system logs with high-risk accounts such as cash to discover unauthorizedaccess and fraudulent activities.

Step 3 and Application 3: based on the data collected and organized in the previous steps,some primary analytical tasks can be performed to identify potential risk factors. Theapplication of Big Data analytics in such analysis varies across companies and industries.For banks and financial service companies, a close examination of the high default rateloans (e.g. type, loan amount, interest rate, location, account managers) can reveal thebusiness line with high fraud risk. In an environment where intellectual properties aresusceptible to theft, tests can be conducted to distinguish patterns, such as the number andtype of similar products released around the same time by competitors. This helps uncoverthe department and product line that require stronger security checks and protectionsystems. In the insurance industry, analysis can be carried out to identify irregularities inbilling activities. If a department consistently processes insurance claims right before thepolicy lapses, it might be a signal of potential fraud committed by the insurers, which shouldtrigger further auditing work.

Step 4 and Application 4: the audit team can then coordinate group meetings, in which thepreviously identified risk factors are used for idea stimulation and inspiration. The purposeof incorporating the fraud indicators determined by Big Data tools is to provide helpfulguidelines and build relevant context so that new ideas are more likely to be generated. Thesystem-identified risk factors should not serve as a checklist that limits the scope of the auditteam’s discussion. Rather, they should provide additional evidence that could not be capturedby auditors’ intuition.

Even when the technology can establish an extremely secure environment, such asblockchain, it is impossible to eliminate all risk factors, which makes manual monitoringnecessary. For example, while companies that adopt blockchain technology seem to containalmost zero risk due to the difficulty in altering transaction records, auditors should stillconsider the potential flaws in the system design. A dominating miner node could controlthe blockchain to commit fraud, and identity theft can comprise the entire network (Xu,2016). Hence, the existence of Big Data tools should serve to complement auditors’professional judgment.

The effect of the group meetings can be further enhanced by the continued use of BigData tools throughout the discussion process. For example, similar fraud cases in the pastcan be requested and presented during meetings for comparison. A Big Data software canrecord each person’s thoughts, questions and notes simultaneously and display them on thesame screen for the team to share. Team members can follow-up with an analytical testwhen new comments emerge to solidify the proposed ideas. Auditors can also use themassive database to explore hidden clues and reconcile different opinions as they proceedwith the brainstorming. In short, the incorporation of Big Data encourages group membersto delve into new areas and offers instant data support.

Step 5 and Application 5: in this step, new ideas proposed during the group meetings areevaluated and finalized. With the assistance of recording devices, each team member can

Financialfraud detection

331

view the ideas of others and leave comments. The system can categorize the auditors’comments into different sessions according to the content and tones. Within each session,the software can collect additional data to perform analytical procedures. These procedurescan address questions include: are the concerns raised by auditors supported by real-lifecases? Is the fraud indicator particularly relevant to a specific industry? Can we observe apattern from similar fraud cases? Further, an anonymous voting on the mostly mentionedideas can be conducted. Based on the feedback provided by team members and dataanalysis, the audit partner/leader can then proceed to draw the conclusions.

Step 6 and Application 6: the final step completes the entire process by documenting thecontent and results of the brainstorming sessions. While the conclusions reached by theaudit team remain the most important product, it is also necessary to document the detailedthought process of each team member. As Big Data tools can store and combine data invarious formats, even voice and video records of the group meetings can be properly saved.These “work-in-progress” notes from the discussion will help form a useful source ofdatabase for future brainstorming sessions.

The implementation of the proposed model will require some necessary training, both onauditors’ knowledge of current Big Data trends and their skills in using technology that iscompatible with handling audit tasks. A separate department can be established to designthe system, and workshops can be organized to teach members helpful techniques. Thetraining should be systematically conducted and offered to all auditors, particularly for theaudit partners who might lead the brainstorming discussions. Given the scale of resourceand revenue, Big 4 might be more suitable candidates to adopt the model at the currentstage. However, there is essentially no constraint on making the choice as long as the budgetallows, or the client’s business operations demand a Big-Data-supported audit.

We further discuss the potential benefits and extension of the model above. According toPCAOB (2007), the effectiveness of brainstorming can be impaired when the session was notproperly held, or key members were absent. By incorporating Big Data tools, thebrainstorming session can be conducted remotely, which encourages more participation. Asthe comments are recorded and processed in the system, the content of the meetings can beshared via a network when a member fails to attend. This also allows continuous discussionand communication among the engagement team, especially when new ideas or facts arediscovered after the planning stage.

However, the integration of Big Data in brainstorming sessions should be conductedwith caution to avoid interacting with factors that negatively affect auditors’ judgment.Simon et al. (2018) find that the separation of likelihood and magnitude assessment of risk,as recommended by the auditing standards, actually results in worse auditor’s judgment.Rose et al. (2017) find that Big Data visualizations should be introduced after the traditionalaudit evidence to yield positive results. These findings suggest that the design of a Big-Data-supported brainstorming session should consider auditors’ cognition and behavior.

Finally, the proposed integration of Big Data tools can be extended to different steps ofthe audit process. For example, initial data collection (Step 1), data integration (Step 2) andfraud indicator identification (Step 3) can be used in the planning stage to establish the scopeof the audit and risk assessment procedures. Hammersley et al. (2010) find thatdocumentation specificity during planning stage affects the subsequent fraud riskassessment; thus, documenting (Step 6) of the model can be helpful in this task. Forsubstantive testing, analytical procedures using Big Data technology described in Step 3 canbe carried out to detect material misstatement. The review stage can benefit from Step 3 andStep 6 so auditors can have easy access to the previously discussed memos and performanceeffective analytical procedures when assessing the conclusions.

MAJ34,3

332

5. ConclusionsThis paper discusses the application of Big Data analytics to the brainstorming sessionsrequired in the current auditing standards. We review literatures on the causes andconsequences of financial fraud. We further comment on the existing problems in frauddetection mechanisms, particularly the difficulty in analyzing unstructured and non-financial data. The brainstorming sessions offer a potential solution by encouragingdiscussions and new ideas. Given the complexity in organizing individual thoughts anddrawbacks in the format of brainstorming, whether it is face-to-face or electronic, wepropose the integration of Big Data analytics into the brainstorming sessions. Theobvious advantages of such integration include comprehensive information base, usefulresults from analytical procedures and efficient communications. Our proposed modeloutlines a six-step brainstorming process that can benefit from the application of BigData analytics.

One limitation of the model proposed is its cost-effectiveness. The implementation of BigData tools during brainstorming sessions and other parts of the audit process can be costly,particularly given the need to continuously update the system and collect various types ofdata. Moreover, the choice of integrating Big Data tools partially depends on the audit feesand whether the clients themselves use Big Data technology in the accounting practice. Thisimplies that larger audit firms with stronger revenue performance are perhaps more suitablecandidates to adopt the model. Another caveat is that the quantity of Big Data might createunwanted burden for the engagement team. As stated in Appelbaum et al. (2017), if an auditby exception (ABE) approach is adopted, Big Data can generate too many exceptions whichhinder the audit process. Brown-Liburd et al. (2015) identify information overload andrelevance as potential drawbacks of Big Data. This implies that adjustments to the overallaudit strategy need to be made if auditors decide to involve Big Data tools in the auditprocess. Finally, the security of Big Data also deserves attention. Appelbaum (2016) arguesthat the provenance of audit-related data should be safely managed so that the auditors canretrace the events for verification purpose. The same might apply to using Big Data tools inthe brainstorming sessions, which is to emphasize the importance of verifying the source ofall data.

Future research can be conducted in areas that extend the arguments in this paper. Forexample, theoretical framework with respect to the application of Big Data analytics shouldbe developed. Alles and Gray (2016) argue that Big Data can be interpreted incorrectly andgenerate false positives without formal theories for guidance. Other potential questionsinclude: what are possible sources of the unstructured data considered in the brainstormingsessions? What analytical procedures can be developed using Big Data tools? What are theadvantages and disadvantages of the existing brainstorming methods and future ones thatincorporate Big Data? We suggest future studies use survey results or case analysis toaddress these questions.

Notes

1. Please refer to the official website of PCAOB, available at: https://pcaobus.org/Standards/Auditing/Pages/AS2401.aspx

2. Production blocking refers to the situation within a group where only one member can speak at atime. This consequently leads to wait time for other members (Diehl and Stroebe, 1987).

3. Please refer to the official website of AICPA, available at: www.aicpa.org/Research/Standards/AuditAttest/DownloadableDocuments/AU-C-00240.pdf

Financialfraud detection

333

ReferencesAlles, M. and Gray, G. (2016), “Incorporating big data in audits: identifying inhibitors and a research

agenda to address those inhibitors”, International Journal of Accounting Information Systems,Vol. 22, pp. 44-59.

Appelbaum, D. (2016), “Securing big data provenance for auditors: the big data provenance black boxas reliable evidence”, Journal of Emerging Technologies in Accounting, Vol. 13 No. 1, pp. 17-36.

Appelbaum, D., Kogan, A. and Vasarhelyi, M. (2017), “Big data and analytics in the modern auditengagement: research needs”,Auditing: A Journal of Practice and Theory, Vol. 36 No. 4, pp. 1-27.

Baker, C.R., Cohanier, B. and Leo, N.J. (2017), “Breakdowns in internal controls in bank tradinginformation systems: the case of the fraud at société générale”, International Journal ofAccounting Information Systems, Vol. 26, pp. 20-31.

Bellovary, J.L. and Johnstone, K.M. (2007), “Descriptive evidence from audit practice on SAS no. 99brainstorming activities”, Current Issues in Auditing, Vol. 1 No. 1, pp. A1-A11.

Bolton, P., Freixas, X. and Shapiro, J. (2007), “Conflicts of interest, information provision, andcompetition in the financial services industry”, Journal of Financial Economics, Vol. 85 No. 2,pp. 297-330.

Boyle, D.M., DeZoort, F.T. and Hermanson, D.R. (2015), “The effect of alternative fraud model use onauditors’ fraud risk judgments”, Journal of Accounting and Public Policy, Vol. 34 No. 6, pp. 578-596.

Brazel, J.F., Agoglia, C.P. and Hatfield, R.C. (2004), “Electronic versus face-to-face review: the effects ofalternative forms of review on auditors’ performance”, The Accounting Review, Vol. 79 No. 4,pp. 949-966.

Brazel, J.F., Carpenter, T.D. and Jenkins, J.G. (2010), “Auditors’ use of brainstorming in the consideration offraud: reports from the field”,TheAccounting Review, Vol. 85 No. 4, pp. 1273-1301.

Brazel, J., Jones, K. and Zimbelman, M. (2009), “Using nonfinancial measures to assess fraud”, Journal ofAccounting Research, Vol. 47 No. 5, pp. 1135-1166.

Brown-Liburd, H., Issa, H. and Lombardi, D. (2015), “Behavioral implications of big data’s impact onaudit judgment and decision making and future research directions”, Accounting Horizons,Vol. 29 No. 2, pp. 451-468.

Burns, N. and Kedia, S. (2006), “The impact of performance-based compensation on misreporting”,Journal of Financial Economics, Vol. 79 No. 1, pp. 35-67.

Capelleveen, G.V., Poel, M., Mueller, R.M., Thornton, D. and Hillegersberg, J.V. (2016), “Outlier detectionin healthcare fraud: a case study in the medicaid dental domain”, International Journal ofAccounting Information Systems, Vol. 21, pp. 18-31.

Carpenter, T.D. (2007), “Audit team brainstorming, fraud risk identification, and fraud risk assessment:implications of SAS no. 99”,The Accounting Review, Vol. 82 No. 5, pp. 1119-1140.

Chen, Y., Wu, C., Chen, Y., Li, H. and Chen, H. (2017), “Enhancement of fraud detection for narratives inannual reports”, International Journal of Accounting Information Systems, Vol. 26, pp. 32-45.

Cockrell, C. and Stone, D.N. (2011), “Team discourse explains media richness and anonymity effects inaudit fraud cue brainstorming”, International Journal of Accounting Information Systems,Vol. 12 No. 3, pp. 225-242.

Cotton, D.L. (2002), “Fixing CPA ethics can be an inside job”, The Washington Post, 20 October,available at: www.washingtonpost.com/archive/opinions/2002/10/20/fixing-cpa-ethics-can-be-an-inside-job/b7441564-e0a6-431b-9280-8c27c6267ebc/?utm_term=.d914cee83738 (accessed 15 January2018).

Cressey, D. (1973), Other People’s Money: A Study in the Social Psychology of Embezzlement, PattersonSmith, Montclair, NJ.

Davidson, W.N., Worrell, D.L. and Lee, C.I. (1994), “Stock market reactions to announced corporateillegalities”, Journal of Business Ethics, Vol. 13 No. 12, pp. 979-987.

MAJ34,3

334

Davis, J.S. and Pesch, H.L. (2013), “Fraud dynamics and controls in organizations”, Accounting,Organizations and Society, Vol. 38 Nos 6/7, pp. 469-483.

Debreceny, R.S. and Gray, G.L. (2010), “Data mining journal entries for fraud detection: anexploratory study”, International Journal of Accounting Information Systems, Vol. 11 No. 3,pp. 157-181.

Dellaportas, S. (2013), “Conversations with inmate accountants: motivation, opportunity and the fraudtriangle”,Accounting Forum, Vol. 37 No. 1, pp. 29-39.

Dennis, A.R. and Valacich, J.S. (1993), “Computer brainstorms: more heads are better than one”, TheJournal of Applied Psychology, Vol. 78 No. 4, pp. 531-537.

Diehl, M. and Stroebe, W. (1987), “Productivity loss in brainstorming groups: toward the solution of ariddle”, Journal of Personality and Social Psychology, Vol. 53 No. 3, pp. 497-509.

Efendi, J., Srivastava, A. and Swanson, E.P. (2007), “Why do corporate managers misstate financialstatements? The role of option compensation and other factors”, Journal of Financial Economics,Vol. 85 No. 3, pp. 667-708.

Fjermestad, J. and Hiltz, S.R. (1998), “An assessment of group support systems experimentresearch: methodology and results”, Journal of Management Information Systems, Vol. 15No. 3, pp. 7-149.

Fung, S.Y.K., Raman, K.K., Sun, L. and Xu, L. (2015), “Insider sales and the effectiveness of clawbackadoptions in mitigating fraud risk”, Journal of Accounting and Public Policy, Vol. 34 No. 4,pp. 417-436.

Grabski, S. (2010), “Discussion of ‘data mining journal entries for fraud detection: an exploratorystudy”, International Journal of Accounting Information Systems, Vol. 11 No. 3, pp. 182-185.

Gray, G.L. and Debreceny, R.S. (2014), “A taxonomy to guide research on the application of data miningto fraud detection in financial statement audits”, International Journal of AccountingInformation Systems, Vol. 15 No. 4, pp. 357-380.

Hammersley, J., Bamber, E.M. and Carpenter, T. (2010), “The influence of documentation specificity andpriming on auditors’ fraud risk assessments and evidence evaluation decisions”,The AccountingReview, Vol. 85 No. 2, pp. 547-571.

Hill, G. (1982), “Group versus individual performance: are nþ1 heads better than one?”, PsychologicalBulletin, Vol. 91 No. 3, pp. 517-539.

Hoffman, V.B. and Zimbelman, M.F. (2009), “Do strategic reasoning and brainstorming help auditorschange their standard audit procedures in response to fraud risk?”, The Accounting Review,Vol. 84 No. 3, pp. 811-837.

Huerta, E., Glandon, T. and Petrides, Y. (2012), “Framing, decision-aid systems, and culture: exploringinfluences on fraud investigations”, International Journal of Accounting Information Systems,Vol. 13 No. 4, pp. 316-333.

Humphreys, S., Moffitt, K., Burns, M., Burgoon, J. and Felix, W. (2011), “Identification of fraudulentfinancial statements using linguistic credibility analysis”, Decision Support Systems, Vol. 50No. 3, pp. 585-594.

Jans, M., Lybaert, N. and Vanhoof, K. (2010), “Internal fraud risk reduction: results of a data miningcase study”, International Journal of Accounting Information Systems, Vol. 11 No. 1, pp. 17-41.

Kaplan, S.E., Pany, K., Samuels, J. and Zhang, J. (2012), “An examination of anonymous and non-anonymous fraud reporting channels”, Advances in Accounting, Incorporating Advances inInternational Accounting, Vol. 28 No. 1, pp. 88-95.

Karpoff, J.M., Lee, D.S. and Martin, G.S. (2008), “The cost to firms of cooking the books”, The Journal ofFinancial and Quantitative Analysis, Vol. 43 No. 3, pp. 581-612.

Kedia, S. and Philippon, T. (2009), “The economics of fraudulent accounting”, Review of FinancialStudies, Vol. 22 No. 6, pp. 2169-2199.

Financialfraud detection

335

Knapp, C.A. and Knapp, M.C. (2001), “The effects of experience and explicit fraud risk assessment indetecting fraud with analytical procedures”, Accounting, Organizations and Society, Vol. 26No. 1, pp. 25-37.

Lynch, A.L., Murthy, U.S. and Engle, T.J. (2009), “Fraud brainstorming using computer-mediatedcommunication: the effects of brainstorming technique and facilitation”,The Accounting Review,Vol. 84 No. 4, pp. 1209-1232.

Mehran, H. and Stulz, R.M. (2007), “The economics of conflicts of interest in financial institutions”,Journal of Financial Economics, Vol. 85 No. 2, pp. 267-296.

Morales, J., Gendron, Y. and Guénin-Paracini, H. (2014), “The construction of the risky individual andvigilant organization: a genealogy of the fraud triangle”, Accounting, Organizations and Society,Vol. 39 No. 3, pp. 170-194.

Norman, C.S., Rose, A.M. and Rose, J.M. (2010), “Internal audit reporting lines, fraud riskdecomposition, and assessments of fraud risk”, Accounting, Organizations and Society, Vol. 35No. 5, pp. 546-557.

Nunamaker, J.F., Dennis, A.R., Valacich, J.S., Vogel, D.R. and George, J.F. (1991), “Electronic meetingsystems to support group work”, Communications of the Acm, Vol. 34 No. 7, pp. 40-61.

Osborn, A. (1957),Applied Imagination, Scribner, New York, NY.Paulus, P.B., Larey, T.S. and Ortega, A.H. (1995), “Performance and perceptions of brainstormers in an

organizational setting”, Basic and Applied Social Psychology, Vol. 17 Nos 1/2, pp. 249-265.Perols, J.L. and Lougee, B.A. (2011), “The relation between earnings management and financial

statement fraud”, Advances in Accounting, Incorporating Advances in International Accounting,Vol. 27 No. 1, pp. 39-53.

Power, M. (2013), “The apparatus of fraud risk”,Accounting, Organizations and Society, Vol. 38 Nos 6/7,pp. 525-543.

Public Company Accounting Oversight Board (PCAOB) (2007), “Observations on auditors’implementation of PCAOB standards relating to auditors’ responsibilities with respect to fraud”,available at: https://pcaobus.org/Inspections/Documents/2007_01-22_Release_2007-001.pdf

Ramlukan, R. (2015), “How big data and analytics are transforming the audit”, Financial ExecutivesInternational Daily, 16 December, available at: http://daily.financialexecutivesorg/how-big-data-and-analytics-are-transforming-the-audit/ (accessed 15 January 2018).

Reurink, A. (2016), Financial Fraud: A Literature Review, Max Planck Institute for the Study ofSocieties, Cologne, Germany.

Rezaee, Z. (2005), “Causes, consequences, and deterrence of financial statement fraud”, CriticalPerspectives on Accounting, Vol. 16 No. 3, pp. 277-298.

Rose, A., Rose, J., Sanderson, K. and Thibodeau, J. (2017), “When should audit firms introduce analysesof big data into the audit process?”, Journal of Information Systems, Vol. 31 No. 3, pp. 81-99.

Simon, C., Smith, J. and Zimbelman, M. (2018), “The influence of judgment decomposition on auditors’fraud risk assessments: some tradeoffs”,The Accounting Review, In-Press.

Smith, A.L., Murthy, U.S. and Engle, T.J. (2012), “Why computer-mediated communication improvesthe effectiveness of fraud brainstorming”, International Journal of Accounting InformationSystems, Vol. 13 No. 4, pp. 334-356.

Straus, S., Parker, A. and Bruce, J. (2011), “The group matters: a review of process and outcomes inintelligence analysis”,Group Dynamics: Theory, Research, and Practice, Vol. 15 No. 2, pp. 128-146.

Tang, J.J. and Karim, K. (2017), “Big data in business analytics: implications for the audit profession”,The CPA Journal, Vol. 87 No. 6, pp. 34-39.

Tian, X., Udell, G.F. and Yu, X. (2016), “Disciplining delegated monitors: when venture capitalists fail toprevent fraud by their IPO firms”, Journal of Accounting and Economics, Vol. 61 Nos 2/3,pp. 526-544.

MAJ34,3

336

Trompeter, G., Carpenter, T., Desai, N., Jones, K. and Riley, R. (2013), “A synthesis of fraud-relatedresearch”,Auditing: A Journal of Practice and Theory, Vol. 32 No. Supplement 1, pp. 287-321.

Trotman, K.T. and Wright, W.F. (2012), “Triangulation of audit evidence in fraud risk assessments”,Accounting, Organizations and Society, Vol. 37 No. 1, pp. 41-53.

US General Accounting Office (GAO) (2002), Financial Statement Restatements: Trends, MarketImpacts, Regulatory Responses, and Remaining Challenges, GAO-03-138, GAO, Washington, DC,available at: www.gao.gov/new.items/d03138.pdf (accessed 15 January 2018).

Valacich, J.S., Dennis, A.R. and Connolly, T. (1994), “Idea generation in computer-based groups: a newending to an old story”, Organizational Behavior and Human Decision Processes, Vol. 57 No. 3,pp. 448-467.

Vasarhelyi, M.A., Kogan, A. and Tuttle, B.M. (2015), “Big data in accounting: an overview”,AccountingHorizons, Vol. 29 No. 2, pp. 381-396.

Williams, J.W. (2013), “Regulatory technologies, risky subjects, and financial boundaries:governing ‘fraud’ in the financial markets”, Accounting, Organizations and Society, Vol. 38Nos 6/7, pp. 544-558.

Wolfe, D.T. and Hermanson, D.R. (2004), “The fraud diamond: considering four elements of fraud”, TheCPA Journal, pp. 38-42.

Xu, J. (2016), “Are blockchains immune to all malicious attacks?”, Financial Innovation, Vol. 2 No. 1,pp. 1-9.

Yoon, K., Hoogduin, L. and Zhang, L. (2015), “Big data as complementary audit evidence”, AccountingHorizons, Vol. 29 No. 2, pp. 431-438.

Zhang, J., Yang, X. and Appelbaum, D. (2015), “Toward effective big data analysis in continuousauditing”,Accounting Horizons, Vol. 29 No. 2, pp. 469-476.

About the authorsJiali Tang is an Assistant Professor of Accounting at the University of Hartford. Her research interestincludes earnings quality, corporate governance, audit quality and corporate social responsibility.Jiali Tang is the corresponding author and can be contacted at: jtang@hartford.edu

Khondkar E. Karim has published over 80 articles in various refereed journals includingAccounting Organizations and Society (AOS), Behavioral Research in Accounting (BRIA), Journal ofCorporate Finance, Accounting Horizons, Journal of Accounting Auditing and Finance, ManagerialAuditing Journal, Review of Quantitative Finance and Accounting, Applied Financial Economics,Advances in Accounting Behavioral Research, International Journal of Finance, Advances inAccounting, Advances in International Accounting, Advances in Quantitative Analysis of Finance andAccounting, Research in Finance, International Journal of Auditing and The Mid-Atlantic Journal ofBusiness. He is also the recipient of a 1999 ANBAR Citation of Excellence for AOS paper. He hasreceived the outstanding teaching award and outstanding research award in 1997 and 1999respectively. He was recognized as one of the most productive accounting doctoral graduates in astudy published in Advances in Accounting. It identified the top ten doctoral graduates by the yearof their graduation and the number of publications within the top 40 selected accounting journals. Hehas co-edited a monograph on environmental disclosure practices and financial performance and twospecial issues of managerial finance on performance measurement and evaluation. He served on theeditorial board of Issues in Accounting Education. Currently, he serves as the Editor of Advances inAccounting Behavioral Research.

For instructions on how to order reprints of this article, please visit our website:www.emeraldgrouppublishing.com/licensing/reprints.htmOr contact us for further details: permissions@emeraldinsight.com

Financialfraud detection

337

Corporate governance, Europeanbank performance and the

financial crisisMohamed A. Ayadi

Goodman School of Business, Brock University, Saint Catharines, Canada

Nesrine AyadiUniversity of Sfax, Sfax, Tunisia, and

Samir TrabelsiGoodman School of Business, Brock University, Saint Catharines, Canada

AbstractPurpose – This paper aims to analyze the effects of internal and external governance mechanisms on theperformance and risk taking of banks from the Euro zone before and after the 2008 financial crisis.Design/methodology/approach – To avoid macroeconomic problems and shocks and because of dataavailability, the authors select some countries of the Euro zone, namely, France, Belgium, Germany andFinland, during the 2004-2009 period. These countries share similar macroeconomic environments(unemployment, inflation and economic growth rates). All the data relating to the banks are manually drawnfrom the supervising reports submitted to banks and are available on the banks’ websites and/or on that ofthe AMF website. The banks included in our sample are drawn from the list of European central banks onwww.ecb.intFindings – The empirical results show that banks undertake tradeoffs between different governancemechanisms to alleviate the intensity of the agency conflicts between the shareholders and managers. Thefindings also confirm that internal mechanisms and capital regulations are complementary and significantlyimpact bank performance.Research limitations/implications – This analysis can be extended through studying the interactionbetween bondholders’ governance and shareholders’ governance and their impact on the 2008 financial crisis.Practical implications – The changes in banking governance help banks find a useful and necessaryway to avoid ill-considered risks that can cause a systemic risk. Therefore, some conditions should be met sothat banking governance can contribute to the economic development.Social implications – Culture and mentality of good banking governance must grow as much as possiblethrough awareness-raising, training, promotion, recognition of performance, enhancing proceduretransparency and stability of good banking governance and regulations, strengthening the national capacityto fight against corruption, and preventive mechanisms.Originality/value – This paper complements previous studies, mainly those of Andres and Vallelado(2008) who examine the impact of the components of the board on banking performance and of Laeven andLevine (2009) who estimate the combined effect of regulatory and ownership structure on the risk-taking ofeach bank.

Keywords Financial crisis, Financial performance, Capital regulation, Insolvency risk,Internal mechanisms

Paper type Research paper

JEL classification – G340, G380, G320, G330, G21, G01The authors would like to thank Skander Lazrak and Robert Welch for their helpful comments.

MAJ34,3

338

Received 13 November 2017Revised 18 July 2018Accepted 15August 2018

Managerial Auditing JournalVol. 34 No. 3, 2019pp. 338-371© EmeraldPublishingLimited0268-6902DOI 10.1108/MAJ-11-2017-1704

The current issue and full text archive of this journal is available on Emerald Insight at:www.emeraldinsight.com/0268-6902.htm

1. IntroductionThe 2008 financial crisis has affected international asset and credit markets. It has revealedsome weaknesses in control and risk management mostly in the financial sector. Actually,financial stability requires the establishment of effective control mechanisms in banks. It isobvious that banking governance is among the mechanisms responsible for the difficultiesfacing the global economies. Although the financial scandals that have shaken the worldover the recent years stimulated studies on corporate governance, there has been limitedresearch on banking governance.

Several studies, such as Shleifer and Vishny (1997) and Baker and Gompers (2003), findthat governance mechanisms are costly to implement. Jensen and Meckling (1976) state thatmonitoring can mitigate the agency problems when insider ownership is low. Companiesadopt governance mechanisms to align the director’s interests with the shareholders’(Shleifer and Vishny, 1997). In the regulated industries, governance mechanisms may be lessimportant. Much of the literature has so far literally interpreted this relationship by statingthat the regulation should substitute governance. However, Hadlock et al. (2002) andHouston and James (1995) show that this empirical evidence does not fully support thisconcept when they wonder why regulated firms apply highly monitoring governancestructures due to cost.

The Basel Committee on Banking Supervision (2006) raises the need for banks tostrengthen their governance practices. The role of managerial discipline must be sharedbetween the board of directors and supervisory authorities. Internal control exercised by theboard of directors and external control exercised by the supervisory authorities arecomplementary (Basel Committee on Banking Supervision, 2006). Young et al. (2009) andVeltri and Silvestri (2011) argue that the financial sector needs a higher level of knowledge,mainly in terms of skills and abilities, a higher degree of technological innovation and ahigher degree of interaction between the staff and customers to generate strategies forcompetitive differentiation based on the level of the service and assistance provided to thecustomers. Moreover, they show that it is necessary for companies of the financial sector toinvest in the development of the human capital to improve the financial stability of thebanking sector. Furthermore, according to the OECD (1996), the knowledge economy is theone in which the production, distribution and use of knowledge are the main drivers ofgrowth, the creation of wealth and employment in all industries, not only the ones classifiedas high-tech industries or knowledge-based industries. More specifically, Joshi et al. (2013)state that the strategic emphasis on the training and management of human capital throughtraining and education, and understanding of the components of intellectual capital, isessential in the knowledge-based economy.

This paper focuses on one of the elements that contribute to the stability of the financialsystem, namely, the banking governance which aims at an enhanced understanding of themechanisms of internal governance, the board and the CEO’s compensation in banks andtheir impact on both risk-taking and performance. It also contributes to a betterunderstanding of the regulatory capital mechanism, as an external governance mechanism,and its impact on risk-taking and banking performance. Moreover, it sheds bright lights onthe relationship between internal and external governance mechanisms (complementaryrelationship or substitutability). Overall, this study provides some answers to the issue ofthe efficiency and adequacy of both internal and external governance mechanisms that mayinfluence risk-taking and banking performance. In particular, we aim to examine the effectof both combined mechanisms (internal governance mechanism and regulatory capital) onrisk-taking and banking performance by taking into account the effect of the CEO’sremuneration, size, board’s independence, proportion of financial expert directors of the

Corporategovernance

339

bank’s board of directors and activities, leadership structure, regulatory capital andpressure on risk-taking and banking performance.

Our paper complements previous studies, mainly those of Andres and Vallelado (2008)who examine the impact of the components of the board on banking performance. Laevenand Levine (2009) estimate the combined effect of regulatory and ownership structure on therisk-taking of each bank. Becher and Frye (2011) assess the interaction between regulationand governance on a sample of bank holding companies and other financial institutions.However, to our knowledge, few previous studies have evaluated the effect of the internalmechanisms of corporate governance on the banking performance; moreover, they have notspecified the relationship between the attributes of the board of directors (BD), CEO’scompensation, capital regulation and effect of these combined mechanisms on both the risk-taking and banking performance. This prompted us to study the effect of the internalmechanisms of corporate governance and the capital regulation on the bankingperformance. More specifically, we sought to answer the following research question:

RQ1. Do internal mechanisms constitute a substitute or complement to the externalmechanisms?

To test this link, we study the effect of the internal mechanisms and capital regulation on thebanking performance, which is our research focus.

Hence, the objective of the paper is to examine the combined effect of internal governancemechanisms and capital regulation on bank risk-taking and performance using a sample of30 banks in four Eurozone countries (France, Belgium, Germany and Finland). We rely onthe CEO’s compensation and activities of the BD as internal governance mechanisms. Weempirically test whether and how internal and external governance mechanisms influencebanking performance before and during the financial crisis period.

The remainder of this paper is organized as follows. Section 2 presents the theoreticalframework and the research hypotheses. Section 3 deals with the methodological aspectsand econometric models. The analysis and discussion of the results are presented in Section4. The conclusion summarizes themain results of this research.

2. Literature review and research hypothesesIn this section, we review the theoretical and empirical literature dealing with therelationship between the governance internal mechanisms, capital regulation andperformance and banking risk-taking. In light of this review, our research hypotheses aredeveloped.

2.1 Impact of the governance internal mechanisms and capital regulation on bankingperformance and risk-takingIn the context of the agency theory, governance attempts to influence and monitor theleader’s behavior through internal mechanisms. Moreover, governance uses externalmechanisms, such as regulation, which is extremely important in the banking sector.However, the BD within the banking sector plays a crucial role. Given the opacity of thebanking activities and existence of deposit insurance, the control exercised by the BD ismuch more important than that of the other stakeholders (Levine, 2004). Macey and O’Hara(2003) explain that the role played by the BD in banks is more important than that in otherfirms. This is due to the fact that directors have a fiduciary responsibility to shareholders,depositors and regulatory authorities.

MAJ34,3

340

Laeven and Levine (2009) study risk-taking by banks, their ownership structures and thenational banking regulations. They deal with the possible conflicts of interest between thebank executives and riskiest owners. Moreover, they estimate the risk to the bank facecomparative shareholders in the corporate governance structure of each bank. They alsoobserve the link between national regulation, bank risk and ownership structure.

The empirical analysis of Laeven and Levine (2009) rests on three theoreticalfoundations. First, the diversified owners tend to argue for more risk-taking decisions thanthe bank borrowers and non-proprietary managers. It is the same with any limited liabilitycompany; the diversified owners have incentives to increase bank risk after fundraising andbondholders (Galai and Masulis, 1976; Esty, 1998). Similarly, managers in private bankstend to advocate for less risk regarding the decisions of the shareholders (Jensen andMeckling, 1976; Demsetz and Lehn, 1985; Kane, 1985). Second, the regulation affects theincentive risk-taking of individual owners as opposed to borrowers and non-proprietarymanagers. Deposit insurance intensifies the ability and shareholders’ incentives to increasethe risk (Merton, 1977; Keeley, 1990).

The impetus for a greater risk-taking generated by the deposit insurance does notnecessarily work with the owners or managers non-owners. Capital regulation is alsoconsidered. Indeed, one of the objectives of the capital regulation is to reduce the risk-takingincentives for owners. The latter will be forced to put in the bank most of their personalwealth at risk (Kim and Santomero, 1988). However, capital requirements do not need toreduce the incentive risk-taking of the majority owners. Although it may push the bank toraise capital, regulation may not force the owners to invest in the bank most of their wealth.Capital regulation may increase the risk-taking. Owners could compensate for the loss ofmore stringent capital requirements in selecting a riskier investment portfolio (Koehn andSantomero, 1980; Buser et al., 1981). The intensification of conflicts between the owners andmanagers leads to an excessive risk-taking. Many countries are trying to reduce the bank’srisk by restricting activities such as securities and insurance (Boyd et al., 1998). In additionto the capital requirements, these activity restrictions may reduce the usefulness of owning abank and intensifying the owners’ incentive risk-taking in relation to the managers. Theimpact of banking regulation depends on the owners’ comparative influence within thegovernance structure of each bank.

In the third foundation, even if the banking regulation affects the owners’ incentive risk-taking in a different way from the managers’, the governance theory justifies the ownershipstructure that affects the ability of the owners to influence risk (Jensen and Meckling, 1976).Shareholders with greater voting rights and cash flow have a stronger incentive power(Shleifer and Vishny, 1986). In this perspective, the ownership structure affects the owners’ability to modify the banking risk in response to both the incentive of risk transfer and onecreated by the government regulations (Boyd and Hakenes, 2014).

The theoretical pioneering studies can combine to make two testable hypotheses. Firstly,the diversified owners of powerful banks have more incentives to increase the non-ownermanagers’ risk. They will be subject to more risk than banks which enjoy other constantfactors. Secondly, banking regulations such as capital requirements, activity restrictionsand deposit insurance affect the incentive risk-taking of the owners rather than of themanagers. The purpose of this is to make the real impact of regulations on risk-takingdependent on the power of shareholders against the managers of corporate governance ofeach bank. However, this framework does not consider optimal risk-taking. Inversely, theobjective of Laeven and Levine (2009) is to produce the first assessments of the theoreticaldata of the empirical predictions related to the way the banking ownership structureinteracts with the national regulations in developing the banking risk-taking. Laeven and

Corporategovernance

341

Levine (2009) compile new data about the bank savings with different regulations with morethan 250 private banks in 48 countries. The latter particularizes these majority-owned banksby calculating the voting and cash flow rights of the majority owner. They collect dataabout banks’ owners and managers based on the theory which indicates the potentialtensions between the stockholders and manager. Even if the managers accumulate bank-specific human capital or enjoy control private benefits, looking therefore for a lower risk-taking compared to shareholders who do not have these skills and benefits (Demsetz andLehn, 1985; Kane, 1985). As tensions between the owners and managers could be alleviatedonce the senior managers have major shareholdings (Houston and James, 1995), the theoryalso suggests that incentives to the owners’ risk-taking be mitigated when the owners investa large part of their personal wealth in the bank.

Laeven and Levine (2009) evaluate family-oriented banks and test the non-linearrelationship between the cash flow and control of the voting rights and the shareholders’risks. Their main measure of risk is the Z-score, which is inversely proportional to theprobability of insolvency. They also use the volatility of the asset returns of each bank,volatility of the stock returns, volatility of earnings and ratio of capital to assets. Laeven andLevine conclude, on the one hand, that the credit risk is generally higher in banks withmajority owners having the rights of important cash flow. This promotes greatercompliance with the theory which states that the cash flow rights by a majority owner areassociated with greater risk. The fact of ignoring the ownership structure is an incompleteanalysis of the bank risk-taking. On the other hand, the authors find that the relationshipbetween risk and regulation depends on the ownership structure of each bank. Therelationship between the regulation and bank risk can actually change sign according tothe ownership structure. Thus, deposit insurance is associated with an increased risk whenthe bank holds a substantial part of the shareholders’ participation with a high enoughpower to react to the incentive of an additional risk-taking created by deposit insurance. Theowners seek to compensate for the utility loss of capital regulation with activity restrictionsand for the increase of bank risk. Stricter capital regulations and reduced restrictions onactivities are associated with a higher risk when the bank has a powerful enough owner(majority), but more stringent capital regulation that has an opposite effect widely spread inbanks. Laeven and Levine (2009) state that the regulation and ownership structure affect thebanking risks, mainly by changing the assessment of the bank. Therefore, they allow thesimultaneous determination of the risk assessment by extending prior research (Keeley,1990; John et al., 2008).

Other researchers argue that regulations and shareholder control are substitutablegovernance mechanisms. Black (1998) shows that the regulations prevent the shareholdersfrom effective potential monitoring. When a company operates in a regulated industry, thebenefits of monitoring the shareholders may be limited. The implication of this statement isan internal governance mechanism, namely, monitoring by shareholders, which can be asubstitute for an external governance mechanism, which is regulation.

Shleifer and Vishny (1986, 1997) argue that the shareholders should be encouraged togather information and monitor the management to maximize the profit. However, it isevident that, when the shareholders are inactive, their inactivity cost is significantly theirresponsibility (Demsetz and Lehn, 1985). The ability of a shareholder to monitor and controlrepresents the main reason of a large owner (Demsetz, 1983, 1986). The observed ownershipconcentration is the weight in which the majority-shareholders have control. However, thebenefits of control are not likely to be equal between firms (Grossman and Hart, 1986).

Demsetz and Lehn (1985) show that the large owner will be present in risky businessesbecause the potential profitability of active monitoring of risky businesses is high. The mere

MAJ34,3

342

presence of majority-owners in risky companies shows clearly that they are tracked.Monitoring is weak, particularly in risky businesses.

In their empirical tests, Demsetz and Lehn (1985) report that regulated companies areindustries with a low large owner. Besides, in the unregulated ones, they find a positiverelationship between ownership concentration and risk. However, large owner andregulation are negatively correlated, which confirms the substitution hypothesis. Largeowner and firms’ risks are positively correlated. This relationship supports the hypothesisof the shareholders’ monitoring. Moreover, Holderness et al. (1999) and Himmelberg et al.(1998) find a positive relationship between large owners and corporate risk. This monitoringevidence is indicative for the majority shareholders.

Other researchers also find a support for the substitution hypothesis. Controlling theshareholders in the US airlines has implemented many governance mechanisms afterthe industry deregulation (Kole and Lehn, 1999). Similarly, Anderson and Fraser (2000)examine US banks during more regulated and less regulated periods and find thatparticipation and risk management are positively related only during the less regulatedperiod. Subsequently, Konishi and Yasuda (2004) assess risk-taking of the Japanese banksonce the capital requirement is increased. They find that risk-taking by stable banks’shareholders decreased when capital adequacy requirement increased. All the findings ofthese previous studies can be considered compatible with the substitution hypothesis. Infact, the owners are more active when there is less regulation.

La Porta et al. (1998) conduct a cross-country test of the substitution hypothesis. Theyfind that firms in countries with weak shareholders’ protection laws have a high large ownersupporting than substitution hypothesis of the shareholders. Caprio et al. (2007) also run thesame test focusing on the relationship between bank governance and its value and find thatbanks in countries with strong legal environment are often widespread. This relationship isconsistent with that of La Porta et al. (1998) and then with the substitution hypothesis.

In his work on takeovers in the banking sector, Prowse (1995) examines the disciplinarymechanisms applicable to bank managers. He compares the banking sector with othersectors of the industry. He emphasizes that the banks’ BD are low-effective mechanisms ofthe managers’ discipline compared to the threat of a hostile takeover as an externalmechanism. Having found that the rotation of the manager in banks is twice lower than inother companies, he explains this by reference to the structure of the board. The banks’ BDare mainly composed of inside directors who enhance the validation observed by themanager. The low proportion of outside directors is explained by the low ownership ofshares in the bank by providing few incentives to the manager’s effective discipline. Prowse(1995) also shows that the intervention of regulatory authorities increases with the size ofthe bank andwith the degree of performance degradation.

Furthermore, Bouwman (2009) studies the interaction between three banking governancemechanisms (capital requirements, regulatory supervision and control by majority owner)and the impact of these mechanisms on risk-taking and on performance after an acquisition.The capital requirements and regulatory control are substitutable monitoring mechanismsof the bank acquisition behavior, the thing which impacted risk-taking and performance ofthe banks. Monitoring by the shareholders replaces then the supervision by the regulatoryauthorities when it is ineffective, and vice versa.

Regulatory discipline also acts as a complement to the discipline exerted by theshareholders. Indeed, the control by the board, in the presence of independent directors andmanagers who are also leaders and control the managers of other companies, improvesperformance through an effective management of the risk-return. In addition, experienceand the number of positions held by the manager, which reflect his/her management ability,

Corporategovernance

343

shows that he/she manages the bank for the benefit of the shareholders by investing in riskyactivities while respecting the governance external constraints imposed by the BaselCommittee (Alexandre and Bouaiss, 2009).

2.2 Hypotheses and research modelThe complementary hypotheses tackled in this study are as follows.

The research hypotheses for financial performancemeasures are:

H1A. The internal mechanisms have a positive impact on financial performance.

H1B. The minimum capital requirement has a positive impact on financial performance.

H1C. Regulatory pressure has a negative impact on financial performance.

The research hypotheses for the insolvency risk measures are:

H2A. The internal mechanisms have a negative impact on the insolvency risk.

H2B. Theminimum capital requirement has a negative impact on the insolvency risk.

H2C. Regulatory pressure has a positive impact on the insolvency risk.

The research model is fully described in Figure 1.

3. Methodology3.1 DataStudying the effects of the global financial crisis in the banking section is of paramountimportance across all markets. Our analysis focuses on the role of internal and externalgovernance mechanisms in bank risk-taking and performance for a sample of Europeanbanks.

To avoid macroeconomic problems and shocks and because of data availability, weselect some countries of the Eurozone, namely, France, Belgium, Germany and Finland,during the 2004-2009 period. These countries share similar macroeconomic environments(unemployment, inflation and economic growth rates). All the data relating to the banks are

Figure 1.Internal governancemechanisms, capitalregulation and banks’performance

Banking Performance

Internal Governance Mechanisms Capital Regulation

Board of Directors

Compensation of the CEO

Power of the CEO

Minimum Capital

Requirements

Regulatory Pressure

MAJ34,3

344

manually drawn from the supervising reports submitted to banks and are available on thebanks’ websites and/or on that of the AMF website. The banks included in our sample aredrawn from the list of European central banks on www.ecb.int

The only banks retained in this study are those for which the data about the mechanismsof banking governance and financial information are available (See Appendix for the list ofbanks). Several banks are dropped out because they lack data on CEO’s compensation. Ourfinal sample consists of 30 banks from four Eurozone countries over a six-year periodbetween 2004 and 2009 (See Appendix for the list of banks). The choice of the panel data ismade to take advantage of the temporal and cross-sectional dimensions of the availableinformation.

3.2 MethodsThe panel data estimates of the relationship between the governance internal mechanisms,capital regulation and banking performance have a number of advantages. They providenot only analysis elements of the relationship between the governance internal and externalmechanisms and banking performance but also the evolution of this relationship over time.In fact, the panel structure and design reduce the bias associated with the estimation.

To better understand the impact of banking governance mechanisms on performanceand to avoid the criticisms of panel studies, the use of more efficient econometric techniquesin the study of the link between the governance internal mechanisms, the capital regulationand banking performance, is then required. Among these techniques, we can mention themethod of static panel data which enable us to bring solutions to the raised problems.

The panel data, or longitudinal data, which have two dimensions (individual andtemporal), give the values of the variables considered for a set or for a panel of individualsover a given period. According to Sevestre (2002), the quality of the information contained inthe panel data is, therefore, very high, and the ability to discriminate between the differentalternative hypotheses is, therefore, significantly greater when working with this type ofdata. Dormont (1989), for example, indicates that this double dimension simultaneouslyaccounts for the behavioral dynamics and their possible heterogeneity. Unlike the timeseries or cross-sections, the double dimension also makes it possible to increase the numberof observations and degree of freedom that implies several variances. On the other hand,Pirotte (2011) indicates that the lack of taking into account the heterogeneity of theindividuals’ behavior during the assessment, the estimated coefficients can be biased. Thisrequires making a hypothesis about the homogeneity of behaviors, which implies thepresence of other constraints.

The panel data have cross-sectional and time dimensions. Therefore, it is ofteninteresting to identify the associated effect for each individual, an effect that does not varyover time, but varies from one individual to another; besides, it can be fixed or random. Thequestion of individual effects in the context of panel data is addressed. To determinethe appropriate method of estimating equations, we will perform some tests, namely, theindividual effects presence test, effect specification test (Hausman test), fixed or randomeffects.

The Hausman (1978) specification test is a general test that can be applied to manyspecification problems in econometrics dealing with the problem of endogeneity. Its mostwidespread application is that of random individual panel specification tests. If the test issignificant (p-value of 5 per cent), we retain the estimators of the fixed effects model that areunbiased. In the opposite case (which is unlikely), we retain those of the composite errormodel because they are efficient.

Corporategovernance

345

3.3 Econometric modelsTo examine the governance internal mechanisms and capital regulation, which affect theEurozone banks’ performance, we propose the following panel models:

Zi;t ¼a0 þaTXi;t þ « i;t (1)

where i refers to an individual bank, i= 1, 2, . . ., 30 and n= 30 (individuals), t refers to the year,t = 2004, . . ., 2009 and t = 6 (time), Z is the model-dependent variable which includes return onassets ROA, return on equity ROE, or insolvency risk IR, X is a vector of explanatory variablesof bank i in year t. It includes BOARDSIZE, which is a measure of the board’s size calculated asthe number of directors in the annual board meeting; INDDIR is a measure of the proportion ofindependent directors calculated as the ratio of the number of independent directors to the totalnumber of directors; EXPERT is a measure of the proportion of financial expert directors in theBD calculated as the ratio between the number of inside directors and the total number ofadministrators; COMITE is a dichotomous measure of the Compensation Committee which isequal to 1 if the bank has a compensation committee, and 0 otherwise; MEYEAR is a measureof the board’s meeting frequency calculated as the number of directors’ meetings held everyyear; DUAL is a measure of duality, which is dichotomous and takes value 1 when bothchairman and CEO’s positions are held by the same person, and 0 otherwise; LRISK is a creditrisk measure calculated as the ratio of the loan loss provision to total loans; CRISK is a liquidityrisk measure calculated as the ratio of the liquid assets to the customer and short-terminvestment; CAP is a measure of the capital ratio calculated as the ratio of the shareholders’equity to total assets; REGP is a measure of the regulatory pressure, which is equal to 1 ifsolvency ratio is below 8 per cent, and 0 if it is above 8 per cent; REMU is the CEO’s cashcompensation,a is a vector of coefficients and « is a random disturbance.

3.4 Measures of variablesThe dependent variables include the ROA, ROE and IR. The explanatory variables arerelated to internal and external governance mechanisms, as well as few control variables.The internal governance mechanisms are described by the size of the BD, proportion ofindependent directors on the board, proportion of financial expert directors of the bank’s BD,board activity, CEO’s remuneration, CEO’s age, CEO’s tenure, CEO’s rotating and duality.The external governance mechanisms are measured using the minimum required capitaland regulatory pressure variables. The proportion of financial expert directors of the bank’sBD is determined by the ratio of the number of internal administrators in the bank and thetotal number of administrators. It is a proxy measure of the proportion of expert directors ofthe bank’s BD. As the internal directors in banks are specialists in the field of finance andbanking, and often attend training in banking sector, they will be increasingly qualifiedthanks to their working careers in the banking sector. The internal administrators is thenumber of directors with experience (present or past) as an executive officer or directors in abank or insurance company (Belkhir, 2009; Aebi et al., 2012).

To operationalize the hypotheses to be tested in the literature review, we define, inTable I, all the variables used in this study for the statistical analyses, as well as thepredicted relationships between the variables.

3.5 Descriptive statisticsTable II presents the descriptive statistics of the dependent variables (performance metrics)and the independent variables of the internal and external governance mechanisms duringthe period between 2004 and 2009.

MAJ34,3

346

Expected Sign

Variables Acronym DefinitionROA/ROE IR Sources

Dependent variablesReturn onassets (%)

ROA ROA is defined as the ratio of netincome to total assets (Barro andBarro, 1990; Gedajlovic andShapiro, 2002)

Authors’ calculationsbased on the annualreports

Return onequity (%)

ROE ROE is defined as the ratio of netincome to total equity (Ang et al.,2002)

Authors’ calculationsbased on the annualreports

Insolvencyrisk (%)

IR The numerator in the formulaabove represents the standarddeviation of the value of ROA ofthe banking assets. Thedenominator is the sum of theaverages of the ratio of ROA andROE/total assets (Ayadi andBoujelbène, 2012; Sinkey, 1999;McAllister and McManus, 1993)

Authors’ calculationsbased on the annualreports

Net interestmargin (%)

NIM NIM is expressed as the percentageof total assets. The net interestincome is the difference betweenthe interest income and the interestcosts

Authors’ calculationsbased on the annualreports

Independent variablesInternal governance mechanismsBoard size BOARDSIZE The number of directors sitting in

the board of directors at the annualmeeting (Ayadi and Boujelbène,2012; Belkhir, 2009)

(þ) (�) Annual reports

Boardindependence (%)

INDDIR The proportion of independentdirectors is determined by thenumber of independent directorsdivided by the total number ofdirectors (Ayadi and Boujelbène,2012; Belkhir, 2009)

(þ) (�) Annual reports

Financial expertdirectors on theBD (%)

EXPERT The proportion of financial expertdirectors in the board isdetermined by the ratio of insidedirectors to the total number ofdirectors in the bank (Belkhir,2009; Aebi et al., 2012)

(þ) (�) Annual reports

Remunerationcommittee

COMITE This is a dummy variable thattakes the value 1 if the bank has aremuneration committee, and 0otherwise (Ayadi and Boujelbène,2012; Broye and Moulin, 2010)

(þ) (�) Annual reports

BD meeting MEYEAR (þ) (�) Annual reports

(continued )

Table I.Description of allretained variables

Corporategovernance

347

Expected Sign

Variables Acronym DefinitionROA/ROE IR Sources

The number of directors’meetingsheld each year (Ayadi andBoujelbène, 2012; Andres andVallelado, 2008)

CEO–chairmanduality

DUAL This is a dummy variable thattakes the value 1 when bothfunctions are occupied by the sameperson (CEO = Chairman of theboard), and 0 otherwise (Ayadi andBoujelbène, 2012; Belkhir, 2009)

(þ) (�) Annual reports

Total cashcompensation ofthe CEO

REMU The total cash compensation of theCEO is defined as annual salaryplus bonus (Ayadi and Boujelbène,2012; Barro and Barro, 1990)

(þ) (�) Annual reports

The CEO’s age CEO AGE The CEO’s age is expressed inyears

Annual reports

CEO’s tenure CEO TENURE The CEO’s mandate is expressedin terms of years of the bank’s CEOpower

Annual reports

The CEO’srotating

CEOROTATION

This is a dummy variable thattakes value 1 when the CEO ischanged, and 0 otherwise.

Annual reports

Capital regulation variablesMinimum capitalrequirement (%)

CAP The CAP variable is equal equityto total assets as a proxy formeasuring the solvency ratio(Shrieves and Dahl, 1992)

(þ) (�) Authors’ calculationsbased on the annualreports

Regulatorypressure

REGP The regulatory pressure variable(Reg-Press) is equal to unity if thebanks solvency ratio (CAP) is less8%, and 0 otherwise (Shrieves andDahl, 1992; Godlewski, 2005)

(�) (þ) Authors’ calculationsbased on the annualreports

Control variablesCredit risk (%) CRISK Credit risk variable is measured by

the ratio of the provisions for loanlosses to the overall loans(Gonzalez, 2005)

Authors’ calculationsbased on the annualreports

Liquidity risk (%) LRISK We use the liquidity risk ratios as aproxy to the ratio of “liquid assetsto the customer and to short-terminvestments” (Valverde andFernandez, 2007)

Authors’ calculationsbased on the annualreports

Bank size LNCSIZE Log (book value of total assets)(Ayadi and Boujelbène, 2012;Belkhir, 2009)

Authors’ calculationsbased on the annualreports

Leverage (%) END The ratio of total debt to totalequity (Ayadi and Boujelbène,2012; Hoshi et al., 1990)

Authors’ calculationsbased on the annualreportsTable I.

MAJ34,3

348

Variable

Beforeanddu

ring

thecrisis[2004-2009]

Beforethecrisis[2004-2006]

Duringthecrisis[2007-2009]

Observatio

nMean

SDMinim

umMaxim

umObservatio

nMean

SDMinim

umMaxim

umObservatio

nMean

SDMinim

umMaxim

um

ROA

180

1.46%

4.83%

�23%

31%

901.99%

4.81%

�0.44%

26.15%

900.92%

4.82%

�23.25%

31.46%

ROE

180

10.08%

12.07%

�33%

77%

9013%

8.66%

�13.80%

43.49%

907.15%

14.17%

�33.42%

77.46%

IR180

79.59%

40.20%

5%2.02

9074.94%

3.54%

4.92%

1.649

904.23%

44.14%

4.76

%201.52%

NIM

162

1.97%

5.42%

�35.62%

31.43%

802.76%

4.31%

�0.45%

24.9%

821.20%

6.25%

�35.61%

31.43%

BOARDSIZE

171

15.333

3.977

622

8115.679

3.885

622

9015.022

4.055

622

INDDIR

157

45.03%

31.93%

01

7243.33%

33.51%

01

8546.46%

30.65%

01

EXPE

RT

157

55.01%

31.94%

01

7256.67%

33.52%

01

8553.60%

30.66%

01

COMITE

171

71.35%

45.34%

01

8166.66%

47.43%

01

9075.55%

43.21%

01

MEYEAR

161

8.658

4.599

326

748.081

3.950

323

879.149

5.056

326

DUAL

172

21.51%

41.21%

01

8225.6%

43.91%

01

9017.77%

38.44%

01

REMU

180

62.92%

36.87%

01.78

9061.82%

36.71%

0.004%

1.770

9064.00%

37.20%

0.008%

1.779

CEOAGE

131

56.92

5.61

4570

6656.54

5.18

4667

6557.30

6.02

4570

CEOTENURE

149

3.66

1.39

16

753.67

1.39

16

743.66

1.38

16

CEOROTATION

155

35.48%

48%

01

7832.05%

46.96%

01

7738.96%

49.08%

01

CAP

180

10.35%

18.70%

01

9010.89%

19.00%

1.46%

96.51%

909.81%

18.48%

0.93%

0.997

REGP

180

78.33%

41.31%

01

9077.77%

41.80%

0.00

1.00

9078.88%

41.03%

01

CRISK

152

450.733

3109.386

035144.60

75669.615

4255.824

0.00

35144.60

77237.535

1216.802

07986.72

LRISK

172

25.39%

1.633

�11.90

13.75

8513.25%

2.114

�11.90

13.75

8737.25%

95.47%

06.71

LNCS

IZE

180

11.268

2.171

4.74

14.63

9011.070

2.146

4.738

14.190

9011.466

2.190

5.20584

14.626

END

180

23.433

17.412

0106.47

9020.893

13.207

3.61%

67.412

9025.973

20.550

0.21%

106.471

Notes

:Thistablepresents

descriptivestatisticsof

thegovernance

variableson

bank

ingperformance

over

the2004-2009period,d

uringthepre-crisisperiod

from

2004

to2006

and

during

thecrisisperiod

2007-2009.The

includ

edvariablesare:ROAisthereturn

onassets

definedas

theratio

ofnetincom

eto

totalassets;ROEisthereturn

onequity

definedas

the

ratio

ofnetincom

eto

totalequ

ity;IRistherisk

ofinsolvency

where

thenu

merator

ofthisratio

representstheSD

ofthevalueof

theROAof

thebank

where

thedenominator

isthesum

oftheaverages

oftheratio

oftheROAandtheROEto

totalassets;NIM

:the

netinterestm

arginisexpressedas

thepercentage

oftotalassets.NIM

isthedifference

betw

eentheinterest

incomeandtheinterestcosts;BOARDSIZE

isameasure

ofthesize

oftheboardof

directors.Itisequaltothenu

mberof

directorssittingon

theBDat

theannu

almeetin

g;INDDIR

isthe

proportio

nof

independ

entdirectorsis

determ

ined

bythenu

mberof

independ

entdirectorsdividedby

thetotalnu

mberof

directors;EXPE

RT

istheproportio

nof

financialexpert

directorsin

theboardisdeterm

ined

bytheratio

ofinside

directorsto

thetotaln

umberof

directorsin

thebank

;COMITEis

adu

mmyvariablethat

takesvalue1ifthebank

hasa

remun

erationcommittee,and

0otherw

ise;MEYEARisameasure

ofthefrequencyof

meetin

gsof

theboardcalculated

asthenu

mberof

directors’meetin

gsheld

each

year;D

UALisa

measure

ofdu

ality

basedon

adu

mmyvariablethat

takesvalue1whenboth

functio

nsareoccupied

bythesameperson

(CEOisthechairm

anof

theboard),and

0otherw

ise;REMUis

thetotalcash

compensationof

theCE

Oandis

definedas

theannu

alsalary

plus

bonu

s;CE

OAGE:T

heCE

O’sageis

expressedin

years;CE

OTENUREis

theCE

O’smandate

isexpressedin

term

sof

yearsof

thebank

’sCE

Opower;C

EOROTATION:T

hisisadu

mmyvariablethat

takesvalue1whentheCE

Oischanged,and0otherw

ise;CA

Pistheminim

umcapitalrequirementv

ariableistheequalequ

ityto

totalassetsas

aproxyformeasuring

thesolvency

ratio

;REGPisaregu

latory

pressure

variablethat

isequaltotheun

ityifthebank

’ssolvency

ratio

(CAP)

isless

than

8percent,and

0otherw

ise;CR

ISKisacreditrisk

variablemeasuredby

theratio

oftheprovisions

forloan

losses

totheoverallloans;L

RISKisthe

liquidity

risk

ratio

definedas

theratio

of“liquidassetstothecustom

erandto

short-terminvestments”;LN

CSIZEisameasureof

thesize

ofthebank

andiscalculated

asthelogarithm

ofthebook

valueof

totalassets;ENDisameasureof

indebtedness

andiscalculated

astheratio

oftotaldebttototalequ

ity

Table II.Summary statistics

of all variables

Corporategovernance

349

We observe that the average number of board directors is 15, which, in no case, exceeds 22for the banks of our sample. Regarding the board’s activity, we find that the average numberof annual meeting it holds is eight for each bank, but increased during the crisis. However,since the 2008 crisis, the activities of banks’ BD have intensified. Moreover, it is observedthat the rate of the banks where the function of the board chairman and that of the CEO areheld by the same person was in the order of 17.8 per cent between 2007 and 2009. This ratedeclines slightly during this period compared to the period before the crisis, which was ofthe order of 25 per cent. The limited effect of the crisis on the structures in our study showsheterogeneity in the subject. Some banks have dual structures (CEO/board chairman), whileothers have monistic structures (board chairman). Some banks had decided before the crisisto a separate the chairman and CEO’s roles. However, due to the crisis, few banks changetheir governance to pass from one system to another. Most banks choose the regulatorypressure which exhibits a marginal increase from 78 per cent before the crisis to 79 per centduring it. Moreover, in most of these banks, there is a compensation committee. Theproportion of banks having this committee rises from 67 per cent before the crisis to 76 percent during it. On the other hand, the post-crisis research shows that the CEO’sremuneration method and that of the employees in charge of the market risk andcounterparty management is a major problem inmodern finance. Hence, for political reasonsand sound management, it is necessary to impose restrictions on this issue. Theserestrictions consist primarily in strengthening the governance framework for executivecompensation, through the creation of a compensation committee. For this reason, thenumber of compensation committees increases after the crisis with the purpose ofstrengthening governance over remuneration and slowing down banking CEO’scompensation. We also notice that the average bank managers’ pay rose from e618,266during the crisis to e640,055 before it. However, good remuneration management must passby the control governance bodies of the board, including the largest employees’compensation. The bank remuneration committee should have in this context means whichenable it to more reliably compare the information released by the human resourcesmanagement with the compensation proposals of this or that leader.

An examination of the variables related to the risk management shows that, on average,credit risk equals to 669 before the 2008 crisis and then falls to the order of 237 during theperiod between 2007 and 2009. Similar pattern exists for liquidity risk, as banks have, onaverage, a liquidity risk equal to 0.1325 before the 2008 crisis, and then rises to 0.3725 duringthe crisis period. As risk management is one of the key elements of the financial institutions’governance, several major financial institutions disappeared because they neglected thebasic rules of control and risk management. The European Commission (2010) shows theabsence of a healthy culture for risk management. In this context, the leaders of the financialinstitutions bear a special responsibility because it is essential to instill a healthy culture forrisk management at all levels. However, the use of hedging instruments entails the ability tocover all or part of those risks that led to the sophistication of the banking business. It is thisuse of the most sophisticated financial techniques that led risk and audit committees to bemore expert, although the risk committee is still largely an internal banking body and not agovernance mechanism of the board. The audit committees of large banking institutionsintegrated in their working program a review of the major risks and purely financial aspectsrelated to accounts.

4. Results and discussionWe first calculate the Pearson correlation coefficients between the dependent andindependent variables in Table III to uncover the univariate association between governance

MAJ34,3

350

Variable

ROA

ROE

IRBOARDSIZE

INDDIR

EXPE

RT

COMITEMEYEAR

DUAL

REMU

CAP

REGP

CRISK

LRISK

END

LNCS

IZE

ROA

1ROE

0.454***

1IR

�0.574***�0

.780

1BOARDSIZE

�0.381***�0

.143***

0.375*

1INDDIR

0.230**

�0.012***�0

.151

�0.390***

1EXPE

RT

�0.226***

0.012

0.149**

0.389***

�0.999

1CO

MITE

0.006

0.113

�0.063*

�0.015

�0.196**

0.197**

1MEYEAR

�0.134***�0

.173*

�0.216**

�0.155**

0.335***

�0.336***

0.180**

1DUAL

�0.041

0.160

0.052**

�0.134*

�0.196**

0.195**

0.186**�0

.048

1REMU

�0.214***

0.043***

0.345

0.196**

0.157**

�0.158**

�0.071

0.064

0.016

1CA

P0.872***

0.150***

�0.689**

�0.391***

0.239***

�0.233***

0.047

�0.064

�0.124*

�0.309***

1REGP

�0.695***�0

.203***

0.640***

0.351***

�0.403***

0.400***

0.054

�0.083

0.033

0.205***

�0.742***1

CRISK

�0.044

�0.038

�0.070

�0.088

0.271***

�0.271***�0

.108

0.361***

�0.081

�0.119*

�0.021

0.070

1LR

ISK

�0.092***�0

.119

0.239*

0.112*

�0.143*

0.143*

0.027

�0.077

�0.081

�0.012

�0.090

0.103

�0.061

1END

�0.432***

0.001***

0.957

0.324***

�0.091

0.088

�0.069

�0.204***

0.018

0.299***

�0.539***0.487***

�0.075

0.277***

1LN

CSIZE

�0.613***�0

.243***

0.496***

0.466***

�0.349***

0.345***

�0.017

0.086

0.169**

0.358***

�0.653***0.489***

�0.162**

0.151**

0.422***

1

Notes

:Thistablepresents

thePearsoncorrelationmatrixof

thevariablesof

governance

mechanism

andbank

ingperformance.T

hesampleinclud

es30

bank

sfrom

four

European

coun

tries(France,Belgium

,GermanyandFinland).T

heinclud

edvariablesare:ROAisthereturn

onassetsdefinedas

theratio

ofnetincom

eto

totalassets;ROEisthereturn

onequity

definedas

theratio

ofnetincometo

totalequity;IRis

therisk

ofinsolvency

where

thenu

merator

ofthis

ratio

represents

theSD

ofthevalueof

theROA

ofthebank

where

the

denominator

isthesum

oftheaverages

oftheratio

oftheROAandtheROEto

totalassets;BOARDSIZE

isameasure

ofthesize

oftheBD.Itequ

alsto

thenu

mberof

directorssittingon

theboardof

directorsat

theannu

almeetin

g;INDDIR

istheproportio

nof

independ

entdirectorsisdeterm

ined

bythenu

mberof

independ

entdirectorsdividedby

thetotaln

umberof

directors;EXPE

RTisthee

proportio

nof

financialexp

ertd

irectors

intheboardisdeterm

ined

bytheratio

ofinside

directorsto

thetotaln

umberof

directorsin

thebank

;COMITEisa

dummyvariable

that

takesvalue1ifthebank

hasaremun

erationcommittee,a

nd0otherw

ise;MEYEARis

ameasure

ofthefrequencyof

meetin

gsof

theboardcalculated

asthe

numberof

directors’meetin

gsheld

each

year;D

UALisameasure

ofdu

ality

basedon

adu

mmyvariablethat

equalsto

1whenboth

functio

nsareoccupied

bythesameperson

(CEOis

thechairm

anof

theboard),and

0otherw

ise;REMUisthetotalcashcompensationof

theCE

Oisdefinedas

theannu

alsalary

plus

bonu

s;CA

Pistheminim

umcapitalrequirement

variable,equ

alsto

theratio

ofequity

tototala

ssets,andisused

asaproxyformeasuring

thesolvency

ratio

;REGPisaregu

latory

pressure

variablethat

isequalto1ifthebank

’ssolvency

ratio

(CAP)

isless

than

8percent,and

0otherw

ise;CR

ISKisacreditrisk

variableismeasuredby

theratio

oftheprovisions

forloan

losses

totheoverallloans;L

RISKisa

liquidity

risk

ratio

definedas

theratio

of“liquidassetsto

thecustom

erandto

short-term

investments”;ENDisameasure

ofindebtedness

andiscalculated

astheratio

oftotald

ebtto

totalequ

ity;L

NCS

IZEisameasure

ofthesize

ofthebank

andiscalculated

asthelogarithm

ofthebook

valueof

totalassets;***,**

and*indicatestatisticalsign

ificance,respectiv

ely,

atthelevels0.01,0.05and0.10

Table III.The Pearson

correlation matrix

Corporategovernance

351

and performance metrics and to assess the multicollinearity problem between theindependent variables. Most of the correlation coefficients are smaller than 0.8, which is thelimit line drawn by Kennedy (1985) for potential problems of multicollinearity.

The examination of the correlation matrix shows that there is a strong correlationbetween ROA and capital regulation CAP which is equal to 0.872 and between IR andindebtedness END which is equal to 0.952. As a consequence, we have to eliminate variableCAP from regression ROA and variable END from regression IR. Moreover, there is a strongcorrelation between the independent directors INDDIR and financial expert directors at theboard EXPERT, which is equal to –0.999. In this case, both the INDDIR and EXPERTindependent variables should not be included in the same regression. We further provideexplanations on the banking performance determinants regarding governance before andduring the 2008 crisis.

Tables IV, V and VI present the parameter estimates relating to the ROA, ROE and IRregressions. The empirical results reported that each regression includes both internal andexternal mechanisms combined together, as well as the control variables during the 2004-2009 period (Table IV), before the 2008 crisis (Table V) and during the 2008 crisis (Table VI).

Each regression includes the BD’ attributes, CEO’s remuneration, capital regulationmechanism and control variables, such as the credit and liquidity risk, indebtedness andbank’s size. Themodel equations are as follows:

ROAi;t ¼ a0 þ a1 Board of Directorsi;t þ a2 Compensation of the CEOi;t

þa3 Power of the CEOi;t þ a4 MinimumCapital Requirementi;t

þa5 Regulatory Pressurei;t þ a6 Control Variablesi;t þ « it (2)

ROEi;t ¼ a0 þ a1 Board of Directorsi;t þ a2 Compensation of the CEOi;t

þa3 Power of the CEOi;t þ a4 MinimumCapital Requirementi;t

þa5 Regulatory Pressurei;t þ a6 Control Variablesi;t þ « it (3)

IRi;t ¼ a0 þ a1 Board of Directorsi;t þ a2 Compensation of the CEOi;t

þa3 Power of the CEOi;t þ a4 MinimumCapital Requirementi;t

þa5 Regulatory Pressurei;t þ a6 Control Variablesi;t þ « it (4)

Tables IV, V, and VI show good adjustment quality where the coefficient of determinationR2 ranges between 15 and 95 per cent. This would indicate that our variables explain thefinancial performance of banks.

The size of the board of directors BOARDSIZE has a negative and statisticallysignificant link with the ROA regression during the 2004-2009 period (Table IV). However,an increase of the BD’ size of banks in the Eurozone has led to a decline of financialperformance. One can deduce that a larger board can decrease the motivation to collect and/or interpret information when information acquisition is costly. Furthermore, thecoordination losses are also more likely to happen. This link rejects our H1A and supportsthe results of Andres and Vallelado (2008) and Persico (2004). Moreover, it confirms theresults of Jensen (1993), suggesting that large boards are less efficient for the management of

MAJ34,3

352

Variable

ROA

ROE

IRGLS

GLS

GLS

GLS

GLS

GLS

BOARDSIZE

�0.00049*[0.0003]

�0.00080***[0.0003]

�0.00256

[0.0034]

�0.00315

[0.0033]

0.01044[0.0067]

0.00667[0.0068]

INDDIR

�0.00642*[0.0038]

�0.04527

[0.0431]

0.36416***

[0.0886]

EXPE

RT

0.00177[0.0041]

0.08694*

[0.0447]

�0.19941**

[0.0847]

COMITE

0.00175[0.0022]

0.00364[0.0022]

0.02944[0.0251]

0.03616[0.0245]

0.031681

[0.0488]

0.06122[0.0504]

MEYEAR

�0.00054**

[0.0002]

�0.00098***[0.0002]

�0.00337

[0.0028]

�0.00381

[0.0027]

�0.02271***[0.0049]

�0.02646***[0.0053]

DUAL

�0.00083

[0.0024]

�0.00295

[0.0024]

0.04696*

[0.0269]

0.03682[0.0264]

0.02711[0.0540]

�0.01120

[0.0542]

REMU

0.00353[0.0027]

0.00127[0.0028]

0.05830*

[0.0313]

0.06798**[0.0306]

0.05537[0.0629]

0.11172*

[0.0626]

CAP

0.09951[0.1686]

�2.59334***[0.3110]

REGP

�0.02528***[0.0030]

�0.02706***[0.0033]

�0.09978***[0.0361]

0.57803***

[0.0647]

CRISK

1.87e-07

[0.0000]

4.60e-07

[0.0000]

1.14e-06

[0.0000]

3.74e-06

[0.0000]

4.78e-06

[0.0000]

LRISK

�0.00001

[0.0011]

�0.00005

[0.0012]

�0.01733

[0.0134]

�0.02043

[0.0130]

0.04809***

[0.0152]

0.07494***

[0.0259]

LNCS

IZE

�0.00245***[0.0006]

�0.01866**

[0.0074]

�0.01591**

[0.0068]

0.04830***

[0.0132]

0.01663[0.0147]

END

�0.00011*[0.0001]

0.00110[0.0008]

0.00174**[0.0008]

Constant

0.06899***

[0.0690]

0.05091***

[0.0048]

0.29634***

[0.0973]

0.27600***

[0.0607]

�0.40999***[0.1572]

0.90517***

[0.1748]

R2

0.688

0.676

0.248

0.255

0.517

0.516

Num

bero

fobservatio

ns132

132

132

132

146

146

F-Statistic

3.66***

3.63***

3.01***

2.72***

20.86***

20.86***

Hausm

antest(p-value)

9.52

(0.48)

8.12

(0.52)

19.10(0.04)

16.26(0.09)

19.93(0.02)

19.65(0.02)

Durbin–

Watson

1.366

1.380

1.459

1.489

1.003

1.125

Notes

:Thistablereportstheim

pactof

internalgovernance

mechanism

sandcapitalregulationon

bank

ingperformance

over

the2004-2009period.T

hesampleinclud

es30

bank

sfrom

four

Europeancoun

tries(France,Belgium

,GermanyandFinland).T

heinclud

edvariablesare:ROAisthereturn

onassetsdefinedas

theratio

ofnetincom

eto

totalassets;ROEisthe

return

onequity

definedas

theratio

ofnetincometo

totalequ

ity;IRistherisk

ofinsolvency

where

thenu

merator

ofthisratio

represents

theSD

ofthevalueof

theROA

ofthebank

where

thedenominator

isthesum

oftheaverages

oftheratio

oftheROAandtheROEto

totalassets;BOARDSIZE

isameasure

ofthesize

oftheBD.Itequ

alsto

thenu

mbero

fdirectors

sittingon

theboardof

directorsat

theannu

almeetin

g;INDDIR

istheproportio

nof

independ

entd

irectors

andisdeterm

ined

bythenu

mberof

independ

entd

irectors

dividedby

thetotal

numberof

directors;EXPE

RTistheproportio

nof

financialexp

ertd

irectors

intheboardandisdeterm

ined

bytheratio

ofinside

directorsto

thetotaln

umberof

directorsin

thebank

;CO

MITEisadu

mmyvariablethat

equalsto

1ifthebank

hasaremun

erationcommittee,and

0otherw

ise;MEYEARisameasure

ofthefrequencyof

meetin

gsof

theboardcalculated

asthenu

mberof

directors’meetin

gsheld

each

year;D

UALisameasure

ofdu

ality

.Thisisadu

mmyvariablethat

takesvalue1whenboth

functio

nsareoccupied

bythesameperson

(CEO

isthechairm

anof

theboard),and

0otherw

ise;REMUisthetotalcashcompensationof

theCE

Oandisdefinedas

theannu

alsalary

plus

bonu

s;CA

Pistheminim

umcapitalrequirement

variablethat

equalsto

equity

tototalassetsandused

asaproxyformeasuring

thesolvency

ratio

;REGPisaregu

latory

pressure

variablethat

isequalto1ifthebank

’ssolvency

ratio

(CAP)

isless

than

8percent,and

0otherw

ise;CR

ISK:creditriskvariableismeasuredby

theratio

oftheprovisions

forloan

losses

totheoverallloans;L

RISKistheliq

uidity

risk

ratio

definedas

theratio

of“liquidassetsto

thecustom

erandto

short-term

investments”;LN

CSIZEisameasure

ofthesize

ofthebank

andiscalculated

asthelogarithm

ofthebook

valueof

totalassets;ENDisameasure

ofindebtedness

andiscalculated

astheratio

oftotald

ebttototalequ

ity;R

2istheR-squ

ared

oftheregression;F

-statistics:in

apaneld

atasample,thetest

ofthepresence

ofindividu

aleffectsmakes

itpossibleto

checkwhether

thestructureof

thepanelishomogeneous

orheterogeneous;theHausm

antestdeterm

ines

whether

theindividu

aleffectsarefixedor

rand

om;the

Durbin–

Watsontestisastatisticaltestwhich

enablesto

testtheauto-correlatio

nof

residu

als;***,**

and*indicatestatisticalsign

ificance,respectiv

ely,

atthelevels0.01,0.05,and0.10.Stand

arderrors

oftheestim

ated

coefficientsarereported

inbrackets

Table IV.The effect of the

internal governancemechanisms and

capital regulation onthe performance ofbanks during theperiod 2004-2009

Corporategovernance

353

ROA

ROE

IRGLS

GLS

GLS

GLS

GLS

GLS

BOARDSIZE

*B_C

RISIS

�0.00083*[0.0005]

�0.00090*[0.0005]

0.00002[0.0030]

0.00048[0.0030]

0.00060[0.0073]

�0.00650

[0.0041]

INDDIR*B

_CRISIS

�0.00456

[0.0064]

�0.03119

[0.0374]

0.18009**[0.0916]

EXPE

RT*B

_CRISIS

0.01271[0.0096]

0.07158[0.0567]

0.03857[0.0791]

COMITE*B

_CRISIS

0.00382[0.0055]

0.00406[0.0055]

0.00922[0.0332]

0.01140[0.0323]

0.00008[0.0801]

�0.02629

[0.0474]

MEYEAR*B

_CRISIS

�0.00111*[0.0006]

�0.00100

[0.0006]

0.00449[0.0036]

0.00525[0.0037]

�0.01441

[0.0090]

0.00482[0.0054]

DUAL*

B_C

RISIS

0.00323[0.0054]

0.00189[0.0055]

0.01850[0.0324]

0.01314[0.0325]

0.02368[0.0783]

�0.07739

[0.0499]

REMU*B

_CRISIS

0.05475***

[0.0130]

0.04004***

[0.0122]

0.03140[0.0694]

�0.02632

[0.0719]

�0.44081**

[0.1879]

0.12211[0.0951]

CAP*

B_C

RISIS

REGP*

B_C

RISIS

�0.03585***[0.0066]

�0.03610***[0.0065]

�0.04857

[0.0385]

0.40333***

[0.0940]

CRISK

�8.33e-08[5.76e-07]

2.24e-07

[6.21e-07]

�4.68e-06[3.40e-06]

�2.64e-06[3.68e-06]

3.44e-06

[8.37e-06]

�5.52e-08[5.12e-06]

LRISK

0.00097[0.0022]

0.00101[0.0022]

�0.00802

[0.0128]

�0.01500

[0.0129]

0.07610[0.0309]

0.01272[0.0202]

LNCS

IZE

�0.00450***[0.0009]

�0.00463***[0.0009]

�0.01538***[0.0047]

�0.01956***[0.0052]

0.07330[0.0119]

0.34972***

[0.0548]

END

�0.00010

[0.0001]

�0.00011

[0.0001]

0.00161***

[0.0007]

Constant

0.05999***

[0.0095]

0.06148***

[0.0095]

0.24602***

[0.0543]

0.25439***

[0.0561]

�0.02062

[0.1370]

�3.10636***[0.6145]

R2

0.6285

0.6335

0.1644

0.1577

0.3919

0.3777

Num

bero

fobservatio

ns147

147

147

147

147

147

F-statistic

3.80***

3.77***

2.99**

3.20**

24.2***

27.34***

Hausm

antest(p-value)

5.99

(0.74)

2.97

(0.96)

5.35

(0.72)

6.76

(0.66)

11.65(0.23)

14.71(0.06)

Notes

:Thistablereportstheim

pact

ofinternal

governance

mechanism

sandcapitalregulationon

bank

ingperformance

before

the2008

financialcrisis(2004-2006)u

sing

thebefore

crisisdu

mmyvariable(B_C

RISIS).The

sampleinclud

es30

bank

sfrom

four

Europeancoun

tries(France,Belgium

,GermanyandFinland).T

heinclud

edvariablesare:ROAisthereturn

onassetsdefinedas

theratio

ofnetincom

eto

totalassets;ROEisthereturn

onequity

definedas

theratio

ofnetincom

eto

totalequ

ity;IRistherisk

ofinsolvency

where

thenu

merator

ofthisratio

representstheSD

ofthevalueof

theROAof

thebank

where

thedenominator

isthesum

oftheaverages

oftheratio

oftheROAandtheROEto

totalassets;BOARDSIZE

isameasure

ofthesize

oftheBD.Itequalsto

thenu

mberof

directorssittingon

theboardof

directorsat

theannu

almeetin

g;INDDIR

istheproportio

nof

independ

entdirectorsandis

determ

ined

bythenu

mberof

independ

entd

irectors

dividedby

thetotaln

umberof

directors;EXPE

RTistheproportio

nof

financialexp

ertd

irectors

intheboardandisdeterm

ined

bytheratio

ofinside

directorsto

thetotaln

umberof

directorsin

thebank

;COMITEisadu

mmyvariablethat

equalsto

1ifthebank

hasaremun

erationcommittee,and

0otherw

ise;

MEYEARis

ameasure

ofthefrequencyof

meetin

gsof

theboardcalculated

asthenu

mberof

directors’meetin

gsheld

each

year;D

UALis

ameasure

ofdu

ality

.Thisis

adu

mmy

variablethat

takesvalue1whenboth

functio

nsareoccupied

bythesameperson

(CEOisthechairm

anof

theboard),and

0otherw

ise;REMUisthetotalcashcompensationof

theCE

Oandisdefinedas

theannu

alsalary

plus

bonu

s;CA

Pistheminim

umcapitalrequirementv

ariablethat

equalsto

equity

tototala

ssetsandused

asaproxyformeasuring

thesolvency

ratio

;REGPisaregu

latory

pressure

variablethat

isequalto1ifthebank

’ssolvency

ratio

(CAP)

isless

than

8percent,and

0otherw

ise;CR

ISK:creditriskvariableismeasuredby

the

ratio

oftheprovisions

forloan

losses

totheoverallloans;L

RISKistheliq

uidity

risk

ratio

definedas

theratio

of“liquidassetsto

thecustom

erandto

short-term

investments”;LN

CSIZE

isameasure

ofthesize

ofthebank

andiscalculated

asthelogarithm

ofthebook

valueof

totalassets;ENDisameasure

ofindebtedness

andiscalculated

astheratio

oftotald

ebtto

totalequ

ity;R

2istheR-squ

ared

oftheregression;F

-statistics:in

apaneld

atasample,thetestof

thepresence

ofindividu

aleffectsmakes

itpossibleto

checkwhether

thestructureof

the

panelishomogeneous

orheterogeneous;theHausm

antest

determ

ines

whether

theindividu

aleffectsarefixedor

rand

om;the

Durbin–

Watsontest

isastatistical

testwhich

enablesto

testtheauto-correlatio

nof

residu

als;***,**

and*indicatestatisticalsign

ificance,respectiv

ely,at

thelevels0.01,0.05and0.10.Stand

arderrors

oftheestim

ated

coefficientsarereported

inbrackets

Table V.The effect of theinternal governancemechanisms andcapital regulation onthe performance ofbanks before the2008 crisis

MAJ34,3

354

ROA

ROE

IRGLS

GLS

GLS

GLS

GLS

GLS

BOARDSIZE

*P_C

RISIS

�0.00012

[0.0004]

0.00039[0.0004]

�0.00444

[0.0040]

�0.00347

[0.0040]

0.01012[0.0095]

0.00892[0.0102]

INDDIR*P

_CRISIS

0.00514[0.0045]

0.00760[0.0426]

0.20910**[0.1032]

EXPE

RT*P

_CRISIS

�0.01616***[0.0058]

�0.11574**

[0.0553]

�0.00855

[0.1422]

COMITE*P

_CRISIS

�0.00204

[0.0032]

�0.00168

[0.0034]

0.03942[0.0304]

0.05164*

[0.0304]

�0.01224

[0.0728]

�0.03667

[0.0780]

MEYEAR*P

_CRISIS

�0.00046

[0.0003]

�0.00029

[0.0003]

�0.00609**

[0.0029]

�0.00810***[0.0030]

�0.01710***[0.0064]

�0.01704**

[0.0076]

DUAL*

P_CR

ISIS

0.00388[0.0042]

0.00382[0.0035]

0.04033[0.0395]

0.05085[0.0368]

0.08726[0.0953]

0.02932[0.0945]

REMU*P

_CRISIS

�0.00017

[0.0095]

�0.00958

[0.0075]

0.07801[0.0944]

0.18022**[0.0890]

�0.54925**

[0.2175]

�0.24192

[0.2285]

CAP*

P_CR

ISIS

REGP*

P_CR

ISIS

0.02074***

[0.0046]

�0.07800*[0.0430]

�0.07496*[0.0414]

0.52921***

[0.0957]

0.43963***

[0.1032]

CRISK

�6.47e-07**[3.23e-07]

1.12e-07

[3.07e-07]

�2.03e-06[3.04e-06]

�2.66e-06[3.00e-06]

2.48e-06

[7.72e-06]

LRISK

0.00088[0.0014]

0.00013[0.0012]

�0.01587

[0.0128]

�0.01541

[0.0126]

0.05590***

[0.0171]

0.05257*

[0.0317]

LNCS

IZE

�0.00392***[0.0006]

�0.00723*[0.0038]

�0.01093*[0.0058]

�0.00836

[0.0058]

0.05115[0.0133]

0.05311***

[0.0146]

END

�0.00023***[0.0001]

0.00003[0.0001]

0.00168**[0.0007]

0.00169**[0.0007]

Constant

0.059830***[0.0064]

0.09159**[0.0406]

0.22290***

[0.0650]

0.19469***

[0.0649]

0.18338[0.1524]

0.15118[0.1668]

R2

0.5002

0.3612

0.2599

0.2402

0.4736

0.4279

Num

bero

fobservatio

ns132

132

132

132

146

132

F-statistic

7.45***

9.33***

2.45***

2.75***

22.99***

21.7***

Hausm

antest(p-value)

5.01

(0.83)

16.68(0.96)

6.51

(0.77)

9.92

(0.45)

0,00

(1.00)

11.29(0.26)

Notes

:Thistablereportstheim

pactof

internalgovernance

mechanism

sandcapitalregulationon

bank

ingperformance

postthe2008

financialcrisis(2004-2006)u

sing

thepost-crisis

dummyvariable(P_C

RISIS).The

sampleinclud

es30

bank

sfrom

four

Europeancoun

tries(France,Belgium

,GermanyandFinland).T

heinclud

edvariablesare:ROA

isthereturn

onassetsdefinedas

theratio

ofnetincom

eto

totalassets;ROEisthereturn

onequity

definedas

theratio

ofnetincom

eto

totalequ

ity;IRistherisk

ofinsolvency

where

thenu

merator

ofthisratio

representstheSD

ofthevalueof

theROAof

thebank

where

thedenominator

isthesum

oftheaverages

oftheratio

oftheROAandtheROEto

totalassets;BOARDSIZE

isa

measure

ofthesize

oftheBD.Itequalsto

thenu

mberof

directorssittingon

theboardof

directorsat

theannu

almeetin

g;INDDIR

istheproportio

nof

independ

entdirectorsandis

determ

ined

bythenu

mberof

independ

entd

irectors

dividedby

thetotaln

umberof

directors;EXPE

RTistheproportio

nof

financialexp

ertd

irectors

intheboardandisdeterm

ined

bytheratio

ofinside

directorsto

thetotaln

umberof

directorsin

thebank

;COMITEisadu

mmyvariablethat

equalsto

1ifthebank

hasaremun

erationcommittee,and

0otherw

ise;

MEYEARis

ameasure

ofthefrequencyof

meetin

gsof

theboardcalculated

asthenu

mberof

directors’meetin

gsheld

each

year;D

UALis

ameasure

ofdu

ality

.Thisis

adu

mmy

variablethat

takesvalue1whenboth

functio

nsareoccupied

bythesameperson

(CEOisthechairm

anof

theboard),and

0otherw

ise;REMUisthetotalcashcompensationof

theCE

Oandisdefinedas

theannu

alsalary

plus

bonu

s;CA

Pistheminim

umcapitalrequirementv

ariablethat

equalsto

equity

tototala

ssetsandused

asaproxyformeasuring

thesolvency

ratio

;REGPisaregu

latory

pressure

variablethat

isequalto1ifthebank

’ssolvency

ratio

(CAP)

isless

than

8percent,and

0otherw

ise;CR

ISK:creditriskvariableismeasuredby

the

ratio

oftheprovisions

forloan

losses

totheoverallloans;L

RISKistheliq

uidity

risk

ratio

definedas

theratio

of“liquidassetsto

thecustom

erandto

short-term

investments”;LN

CSIZE

isameasure

ofthesize

ofthebank

andiscalculated

asthelogarithm

ofthebook

valueof

totala

ssets;ENDisameasure

ofindebtedness

andiscalculated

astheratio

oftotald

ebtto

totalequ

ity;R

2istheR-squ

ared

oftheregression;F

-statistics:in

apaneld

atasample,thetestof

thepresence

ofindividu

aleffectsmakes

itpossibleto

checkwhether

thestructureof

the

panelishomogeneous

orheterogeneous;theHausm

antest

determ

ines

whether

theindividu

aleffectsarefixedor

rand

om;the

Durbin–

Watsontest

isastatistical

test

which

enablesto

testtheauto-correlatio

nof

residu

als;***,**

and*indicatestatisticalsign

ificance,respectiv

ely,at

thelevels0.01,0.05and0.10.Stand

arderrors

oftheestim

ated

coefficientsarereported

inbrackets

Table VI.The effect of the

internal governancemechanisms and

capital regulation onthe performance ofbanks during the

2008 crisis

Corporategovernance

355

monitoring due to problems of conflict of interest between the directors and the rise of thedecision-making time. On the other hand, this variable has neither a positive nor a negativeeffect on banking performance before the crisis (Table V) and during the crisis (Table VI).However, the board is one of the best indicators for the improvement of the quality ofgovernance in the European banks, which implies that the size of the BD affects the qualityof the banking governance although it does not improve financial performance. Contrary toour result, several papers consider that the board’s size has an impact either positive ornegative on the financial performance. García-Meca et al. (2015) specify that the board’s sizehas a significant and positive relationship.

The presence of the proportion of outside independent directors in the BD INDDIR has anegative and statistically significant link in the ROA regression but a positive andstatistically significant impact on the IR during the 2004-2009 period. This result does notsupport our H1A and H2A. As a consequence, a rise of the proportion of independentoutside directors in the boards of the Eurozone banks led to a decline of their financialperformance. This link is well confirmed during the period before the 2008 crisis (Table V)and during the 2008 crisis (Table VI). This negative relationship may be caused by severalfactors that need to be adjusted. The European Commission (2010) then raised the strictobservance of the governance principles, which, in this case, reduces the financial crisis.However, the representation of the independent directors on the board is, therefore, one ofthe most important issues of good governance. Therefore, it is not important for banks tohave a significant proportion of independent directors in their boards but to have a majorityproportion. However, a study of the European banks shows diversity. In this context, theEuropean Commission (2010) states that the composition of the management bodies ofinstitutions should be sufficiently diversified regarding age, gender, geographical origin andeducational and professional background so as to represent the various opinions andexperiences. The extreme situation diversity, either the total number of directors or the largenumber of independent directors, as well as their position in the board, especially in theAudit Committee, is a feature related to independent directors that can improve the bankingperformance. Pathan and Faff (2013) report a negative relationship between directors’independence and the banking performance, suggesting that the independent directorswithin banks are chosen rather to comply with the regulatory requirements. On the otherhand, Cornett et al. (2009) show that there is a positive impact of the independent memberson the banking performance because of an increasing monitoring on themanagers.

Regarding the expertise of the banking directors EXPERT, which is measured by theproportion of inside directors at the bank, it can be noticed that there is a positive andstatistically significant relationship between the directors’ expertise of banks EXPERT andfinancial performance beside a negative and significant relationship between the proportionof expert directors of the bank’s BD and the IR during the period 2004-2009 (Table IV).However, we found no effect until the crisis period of 2008 (Table V), and an effect isreversed during the crisis (Table VI) and during the period 2004-2009 (Table IV). However,we found no effect until the crisis period of 2008 (Table V), and an effect is reversed duringthe crisis (Table VI). However, the presence of the board’s directors who are specialists andexperts in the bank field is very beneficial for the bank’s BD, especially from the perspectiveof their role in risk management. In fact, without having sufficient knowledge of the bank,the members of the supervisory board cannot effectively control the executive council. Thisis verified by Fernandes and Fitch (2009) and Haan and Vlahu (2016) who define thatexpertise is the average number of years of professional experience in the financial sector.On the other hand, several other studies report that outside directors of financial institutionsdo not often have significant recent experience in the banking sector (Minton et al., 2014).

MAJ34,3

356

Furthermore, Minton et al. (2014) rank an independent director as a financial expert if he/sheworks within a banking institution, non-banking financial institution or if he/she has a rolerelated to funding within a non-financial company or an academic institution, or he/she is aprofessional investor. In addition, our results confirm the conclusions of Hau and Thum(2009), suggesting that the lack of experience of the BD’ members in the German banks hasbeen positively linked to losses incurred in 2007-2008.

The examination of the coefficients of the estimated variables related to the activity ofthe board shows that the number of directors’ meetings in the board MEYEAR has anegative impact on the asset returns during the 2004-2009 period (Table IV). In fact, weobtain the same effect before the 2008 crisis (Table V) and during the crisis period(Table VI). This result contradicts both our H1A and the results of Andres and Vallelado(2008). Nevertheless, this result shows that the banks tend to reduce the number of the boardmeetings MEYEAR so as to raise their financial performance. It was also noted that there isa negative and statistically significant relationship between the number of the board’smeetings MEYEAR and IR, which link confirms our H2A. It is reported that the increase ofthe number of board’s meetings during the 2008 financial crisis period is insufficientbecause the number of the directors’ meetings in the board is revealed to have a negativerelationship with the asset returns. However, there are other criteria related to theimprovement of the number of meetings and the increase of banking performance, such asthe extension of the duration of the meeting with a large attendance more than 90 per cent asa general rule. These criteria may be considered as methods to be followed so as to foreseewhether there is a risk of a financial crisis. More particularly, the European Commission(2010) revealed that the board evaluation is required to make sure that important issues areproperly discussed, and that each director is sufficiently involved. However, it declared thatthe board should have more freedom to carry out this assessment. For his part, Vafeas (1999)shows that the frequency of the board’s meetings is negatively related to performance. Onthe other hand, Liang et al. (2013) state that the number of board’s meetings and proportionof independent directors have positive effects on banking performance. We can concludethat the increase in the size of the BD and that of meetings adversely affects performance,which indicates that a BD that meets regularly increases the divergence of decisions and theasymmetry of information. Then, we argue that performance may be weakened.

Regarding the Remuneration Committee COMITE, it was found to have a positive effecton the financial performance. This effect invalidates our research hypothesis (H1A), whichconfirms the results of Laing and Weir (1999) and Dalton et al. (1999) who find thata compensation committee will lead to better performance. However, according to the BaselCommittee on Banking Supervision (2015), systemically important financial institutionsmust have a compensation committee on the BD, which is an integral part of theirgovernance structure, and monitor the development and the functioning of the remunerationsystem. The 2015 Basel Committee also requires that the BD, if necessary in collaborationwith its compensation committee, approve the compensation of the senior executives(mainly, the chief executive officer, chief risk officer and chief internal auditor) andsupervise the development and functioning of the compensation policies and systems, aswell as the related control processes. The regulatory standards also state that financialinstitutions should have a compensation committee to supervise the design and functioningof the compensation system on behalf of the directors of the board. Therefore, thecompensation committee should be made of in a way that enables it to proceed with anindependent and specific judgment on the compensation policies and practices and theincentives created for the management of risk, capital and liquidity. Moreover, thecommittee should carefully evaluate the practices by which compensation is paid for

Corporategovernance

357

potential future income, the timing and likelihood of which remain uncertain. The committeeshould also work closely with the corporate risk committee in assessing the incentivescreated by the compensation system and ensure that the company’s compensation policycomplies with the relevant principles and standards. On the other hand, the regulatorystandards emphasize the role of the compensation committee in establishing andsupervising the executives’ compensation and require its members to work with theenterprise risk committee to ensure the compliance with the relevant requirements. Overall,the focus of the principles and standards on the “efficient governance of compensation”deserves approval and reflects a consolidated trend in the banking regulation byrecognizing the role of corporate governance for financial stability.

Regarding the position of the chief executive officer sitting as the chairman of the boardof directors DUAL, there is a positive and statistically significant impact on financialperformance during the 2004-2009 period (Table IV), which confirms ourH1A. However, wefind no effect before the crisis (Table V) and during the crisis (Table VI). Inspired by thegreen book of the European Commission (2010), some banks have changed their method ofgovernance passing from one system to another after the 2008 financial crisis. However, thegrowing need for control over the decisions justifies a more general use of the dual structureor of the dissociation between the function of the CEO and that of the chairman of the board.Nevertheless, there are institutions that have even chosen the opposite way during the 2008financial crisis, moving from a dissociated governance structure to a unified structurebetween the CEO and the chairman. According to Moody, the function dissociation isinteresting only if the chairman of the board is actually independent of the generalmanagement. For their part, Hagendorff et al. (2010) find no link between the duality and thebanking performance. In contrast, El-Chaarani (2014) concludes that duality negativelyaffects banking performance.

Regarding the CEO’s remuneration REMU, the results show that it has a positive andstatistically significant impact on the return on equity ROE during the 2004-2009 period(Table IV). In fact, we find the same effect before and during the 2008 crisis (Tables V andVI). This result confirms ourH1A and contrasts the conclusions of Alshimmiri (2004).

On the other hand, a positive and statistically significant influence (at 10 per cent level) isobserved on the IR during the 2004-2009 period, which does not support our H2A. Thiseffect is reversed both before and during the 2008 crisis (Tables V and VI). In this context, noserious study helps economically justify the compensation levels of banks’ executives.Therefore, due to political reasons and sound management, it is necessary to imposerestrictions on this subject. However, the European Commission’s (2010) objective was tostrengthen the governance mechanism in terms of remuneration by creating a compensationcommittee to support the monitoring of the CEO’s remuneration, reduce the IR and increasebanking performance; besides, it aims at limiting the variable remuneration. Themanagement and staff remuneration must be aligned with the bank’s long-term interestsand should not encourage excessive risk. However, this device, which depends on theremuneration outlined by the European Commission (2010), is largely inadequate, in that itdoes not provide a comprehensive and restrictive framework which, alone, cannot bring thelevels of the banking compensation closer to equal qualifications and to the rest of theeconomy.

According to our results, we observe a negative and statistically significant relationshipat the level of 1 per cent between the regulatory pressure REGP, which is considered as anexternal control mechanism, and financial performance (ROA) during the 2004-2009 period(Table IV) and after the 2008 crisis (Table V). These results confirm ourH1C. The impact of

MAJ34,3

358

the regulatory pressure becomes positive and significant with the ROA and negative andsignificant with the ROE during the 2008 crisis (Table VI).

Moreover, a positive and statistically significant relationship at the threshold of 1 percent is obtained between the regulatory pressure and IR during the 2004-2009 period(Table IV), and before and during the 2008 crisis (Tables V and VI), which does not confirmour H2C. This means that the presence of regulatory pressure implies the presence ofinsufficient capital not meeting the regulatory standards, since after the 2008 financial crisis,the Basel Committee was committed to improving the resilience of the banking sector bystrengthening the regulatory framework of equity. However, the use of regulatory pressuremakes the capital increase to reach the regulatory standards for the purpose of beingprotected against the bankruptcy risk. Therefore, the higher the pressure is, the betterfinancial performance will be. On the basis of this result, we would conclude that the internalcontrol strengthens the external one in this regression.

The banking capital regulation is chosen as an external governance mechanism on theperformance of 30 Eurozone banks. According to the general model, there is a negativecorrelation between the bank’s capital regulation CAP and level of IR during the 2004-2009period (Table IV). This result confirms our H2B. However, not link is revealed before andduring the 2008 crisis (Tables V and VI). In sum, the estimation results confirm ourhypotheses, as well as the findings of Laeven and Levine (2009). The latter underlines theconflicts of interest between the banking CEOs and shareholders and shows that bank risk-taking positively varies with the shareholders’ power within the governance structure ofeach bank. The results of Laeven and Levine (2009) and Becher and Frye (2011) haveimportant political implications, as they assume that the same regulations have differenteffects on banks’ risk-taking depending on the governance structure of banks.

The relationship between credit risk (CRISK) and the ROA is negative and statisticallysignificant during the 2008 crisis (Table VI). Therefore, any increase of the risk leads to adecline of performance. This result appears to be in line with the Basel Committee’sregulation which introduced the solvency ratio to limit the effects of this negative influenceof risk-taking on the banking performance. Hence, banks with a riskier loan portfolio areexpected to have a higher loan loss provision, and vice versa. The statistical significance isthen low for a high loan portfolio quality and high for a low loan portfolio quality. As aresult, the overall banking risk decreases as the loan portfolio quality increases. In addition,the credit risk coefficient is significant and negative in the regression. This proves that theeffect of credit risk on banking performance is negative, which confirms our hypothesis andthe results of Miller and Noulas (1997) and Athanasoglou et al. (2008).

Liquidity risk LRISK is positive and statistically significant (at the 1 per cent level)during the 2004-2009 period (Table IV) and the 2008 crisis (Table VI), suggesting that thegreater liquidity risk is, the higher the insolvency risk will be, which confirms the results ofAngbazo (1997) and Valverde and Fernandez (2007). Overall, the obtained results suggestthat there is a link of compromise between the internal and external governancemechanisms during the period between 2004 and 2009. Similar investigations are conductedbefore and during the crisis, to assess the robustness of the total period based result.

The bank’s size (LNCSIZE) is significant and has a negative sign with the financialperformance and a positive sign with the insolvency risk during the 2004-2009 period(Table IV), before and during the 2008 crisis (Table V) the 2008 crisis (Table VI). This resultsuggests that the larger the size, the lower the performance of the bank.

However, performance may be reduced with the increase of debts and possible losses.Moreover, a financial structure marked by a high indebtedness has a negative andsignificant impact on banking performance measured by the ROA during the period

Corporategovernance

359

between 2004 and 2009 (Table IV) and during the 2008 crisis period (Table VI). This can beexplained by the large financial burden caused by indebtedness, which reduces net incomeand subsequently the book yield. We note that there is a positive and statistically significantrelationship between debt and ROE over the 2004-2009 period (Table IV), before the 2008crisis period (Table V) and during the 2008 crisis period. (Table VI). Therefore, the higherthe debt is, the lower private equity will be and, as a result, the ROE increases. On the otherhand, indebtedness and the ROE, which grow in the same direction, means that theshareholders are satisfied with their investments. However, the higher indebtedness is, thelower the financial costs and the higher the profitability will be.

On the whole, the results reported in Tables IV, V and VI suggest that there is acomplementary relationship between internal and external governance mechanisms duringthe 2004-2009 period, before the crisis and during the 2008 crisis, mainly between 2007 and2009. The crisis period results corroborate our tested hypotheses and those in Table IVwhich are obtained during the period between 2004 and 2009. In fact, the externalmechanisms seem to be reinforced by the internal governance mechanisms so as to ensurebetter soundness and efficiency of the banking industry by improving and ensuring itsperformance during the 2008 crisis. This result corroborates the opinion of the BaselCommittee on Banking Supervision, which in October 2010, after a public consultation,published a set of principles to strengthen good governance practices in banks to deal withthe 2008 crisis.

In addition, and in accordance with the study of De Coussergues (1996), prudentialregulation intervened for the purpose of harmonizing of the banking securitystrengthening conditions and modernizing the banks’ operation. Moreover, the crucialobjective behind prudential regulation is to avoid any crisis that may jeopardize thevulnerability of the banking system. Therefore, regulators are required to permanentlyestablish standards for the control of the stability of the banking system so that thesystem will be trustful and the systemic crises caused by “bank runs” will be avoided.In the end, the wave of financial innovations in recent years has affected the banks’functioning through new markets, businesses and banking practices. The regulationauthority must, therefore, adapt both to these changes, particularly regarding thedevelopment, and to accounting operations.

5. Robustness tests5.1 Alternative measure of banking performance and extended model to incorporate of theinternal governance variablesTo check the robustness of the relation between internal governance, capitalregulation and performance, equations (2, 3 and 5) are extended to include the age ofCEO, mandate of the CEO and rotation of the CEO as additional explanatory variablesand NIM as an additional dependent variable. The resulting extended equation (5) isbelow:

Performi;t 5a0 þ a1 Board of Directorsi;t þ a2 Compensation of the CEOi;t

þa3 Power of the CEOi;t þ a4 CEOAgei;t þ a5 CEOTenurei;t

þa6 Rotation of the CEOi;t þ a7 MinimumCapital Requirementi;t

þa8 Regulatory Pressurei;t þ a9 Control Variablesi;t þ « it (5)

MAJ34,3

360

Table VII presents the results of the static panel estimates of the regression equation (5) interms of ROA, ROE, IR and NIM.

The interpretation of the coefficients related to the measures of the BD’ structure(BOARDSIZE, INDDIR and EXPERT), activity of the board (COMITE, MEYEAR), CEO’spower (DUAL), remuneration of the CEO (REMU), age, mandate and rotation of the CEO, thecapital regulation (CAP and REGP) in Table VII is qualitatively similar to that of Table IV.

Table VII.The effect of the

internal governancemechanisms and

capital regulation onthe alternativeperformance of

banks during theperiod 2004-2009:extended model

Independent variablesROA ROE IR NIMGLS GLS GLS GLS

BOARDSIZE �0.0004 [0.0002] �0.00120 [0.0043] 0.01631 [0.0127] �0.00090** [0.0004]INDDIR �0.0010 [0.0037] �0.14299** [0.0586] �0.04305 [0.1652] �0.02640*** [0.0053]EXPERTCOMITE 0.0005 [0.0022] 0.02682 [0.0343] �0.00792 [0.0741] �0.00627** [0.0029]MEYEAR �0.0007*** [0.0002] �0.00603 [0.0045] �0.01160* [0.0068] �0.00205*** [0.0003]DUAL 0.00321 [0.0026] 0.04805 [0.0395] �0.40419*** [0.1134] 0.00593** [0.0027]REMU 0.00105 [0.0028] 0.08497* [0.0436] �0.00488 [0.0915] 0.00561 [0.0034]CEO AGE 0.00029 [0.0002] �0.00193 [0.0030] 0.00398 [0.0046] �0.00082*** [0.0002]CEO TENURE �0.00111 [0.0008] 0.00219 [0.0120] �0.00195** [0.0008]CEO ROTATION �0.00159 [0.0017] �0.02007 [0.0272] �0.03006 [0.0304] �0.00145 [0.0019]CAP 0.78793* [0.4630] 0.19729*** [0.0317]REGP �0.01787*** [0.0035] �0.12796* [0.0677] 0.36857*** [0.0946] �0.00248 [0.0047]CRISK 0.06101* [0.0318] �1.31847** [0.5402] 1.83962* [0.9654] 0.02988 [0.0368]LRISK 0.00447* [0.0024] �0.08322** [0.0381] 0.12038 [0.0766] �0.00055 [0.0026]LNCSIZE �0.00157*** [0.0006] �0.01463 [0.0094] 0.18016 [0.0508] 0.00085 [0.0006]END �0.00018*** [0.0001] 0.00350*** [0.0012] �0.00005 [0.0001]Constant 0.04378*** [0.0137] 0.43345* [0.2400] �1.90467*** [0.6128] 0.09519*** [0.0176]R2 0.8591 0.1874 0.5943 0.3798Number of observations 97 97 97 90F-statistic 2.06** 3.54*** 15.49*** 32.54***Hausman test(p-value)

14.13 (0.36) 14.24 (0.43) 94.74 (0.00) 18.35(0.19)

Notes: The included variables are: ROA is the return on assets defined as the ratio of net income to total assets; ROE isthe return on equity defined as the ratio of net income to total equity; IR is the risk of insolvency where the numerator ofthis ratio represents the SD of the value of the ROA of the bank where the denominator is the sum of the averages of theratio of the ROA and the ROE to total assets; NIM: the net interest margin is expressed as the percentage of total assets.NIM is the difference between the interest income and the interest costs; BOARDSIZE is a measure of the size of the boardof directors. It is equal to the number of directors sitting on the BD at the annual meeting; INDDIR is the proportion ofindependent directors is determined by the number of independent directors divided by the total number of directors;EXPERT is the proportion of financial expert directors in the board is determined by the ratio of inside directors to thetotal number of directors in the bank; COMITE is a dummy variable that takes value 1 if the bank has a remunerationcommittee, and 0 otherwise; MEYEAR is a measure of the frequency of meetings of the board calculated as the number ofdirectors’ meetings held each year; DUAL is a measure of duality based on a dummy variable that takes value 1 when bothfunctions are occupied by the same person (CEO is the chairman of the board), and 0 otherwise; REMU is the total cashcompensation of the CEO and is defined as the annual salary plus bonus; CEO AGE: the CEO’s age is expressed in years;CEO TENURE is the CEO’s mandate is expressed in terms of years of the bank’s CEO power; CEO ROTATION: this is adummy variable that takes value 1 when the CEO is changed, and 0 otherwise; CAP is the minimum capital requirementvariable is the equal equity to total assets as a proxy for measuring the solvency ratio; REGP is a regulatory pressurevariable that is equal to the unity if the bank’s solvency ratio (CAP) is less than 8 per cent, and 0 otherwise; CRISK is acredit risk variable measured by the ratio of the provisions for loan losses to the overall loans; LRISK is the liquidity riskratio defined as the ratio of “liquid assets to the customer and to short-term investments”; LNCSIZE is a measure of the sizeof the bank and is calculated as the logarithm of the book value of total assets; END is a measure of indebtedness and iscalculated as the ratio of total debt to total equityequity; ***, ** and *indicate statistical significance, respectively, at thelevels 0.01, 0.05 and 0.10. Standard errors are reported in brackets

Corporategovernance

361

The coefficient of the BD’ size is negative and statistically significant at 5 per cent level forthe NIM. This suggests that banking performance is negatively related to the board’s size.Similarly, the coefficients of the proportion of independent directors (INDDIR) are allnegative for all the measures of performance but statistically significant for (ROE) and(NIM). Therefore, these provide evidence of a negative association between the board’sindependence and banking performance.

Moreover, the CEO’s age-related coefficients and mandate are negative and statisticallysignificant for the NIM. In this case, it can be concluded that there is a positive associationbetween the CEO” age andmandate and banking performance.

Finally, the regulatory pressure coefficients (REGP) are negative and statisticallysignificant for the ROA, ROE and NIM. They, therefore, indicate that there is a negativerelationship between capital regulation and performance measured by the ROA and ROE.Moreover, there is a positive and statistically significant association between the regulatorypressure (REGP) and banking performance measured by the IR.

Overall, the static panel estimates of equation (5) using an alternative performancemeasure (NIM) and other performance measures, such as ROA, ROE and IR, and additionalinternal governance measures in Table VII also support a complementary link betweeninternal governance mechanisms and capital regulation.

5.2 Impact on the periods, before and during the crisis of 2008The results of the impact of internal governance mechanisms and capital regulation onbanking performance may differ over time. We re-estimate equation (5) using the staticpanel data technique for different periods: from 2004 to 2009, before the 2008 crisis (from2004 to 2006) and during the crisis (from 2007 to 2009). We present the results of the NIM inTable VIII.

The results for the periods before and during the 2008 crisis and during the 2004-2009period show that the ratio of the proportion of independent directors (INDDIR) isstatistically negative for all the periods. However, the negative coefficient for the number ofdirectors’meetings on the BD is statistically significant for all the periods.

The CEO’s age coefficient is negative and statistically significant at 1 per cent levelduring the crisis and during the 2004-2009 period. Moreover, the CEO’s term has a negativeand statistically significant impact at 5 per cent before the crisis and during the 2004-2009period.

Similarly, the regulatory pressure negatively affects banking performance. In fact, it wasstatistically significant at 5 per cent threshold before the 2008 crisis. On the other hand, theminimum CAP positively affects banking performance. In fact, it is statistically significantat 1 per cent level during the 2004-2009 period. It seems that there are differences about theimpact of the internal mechanisms and CAP on banking performance between the periods.

Actually, the adjusted R2 of the regression before the crisis (0.3268) is more than thedouble and almost the triple of that of the regression during the 2008 crisis, indicating thatbanking governance is more important as an internal governance for the period before thecrisis, while during the crisis, despite the fact that there is a compromise between internaland external mechanisms; however, this control is insufficient; therefore, the control needs tobe further strengthened.

6. ConclusionThis paper analyzes the effect of the governance internal mechanisms and capital regulationon risk-taking and banking performance before and during the 2008 crisis. Our empiricaltests are conducted using a sample of 30 Eurozone banks during the 2004-2009 period. We

MAJ34,3

362

Table VIII.The effect of the

internal governancemechanisms and

capital regulation onthe performance ofbanks before the

2008 crisis, duringthe 2008 crisis andduring the period

2004-2009

Independent variables

NIM NIM NIM2004-2006 2007-2009 2004-2009

GLS GLS GLS

BOARDSIZE �0.00090** [0.0004]BOARDSIZE*B_CRISIS 0.00092* [0.0005]BOARDSIZE*D_CRISIS 0.00105* [0.0006]INDDIR �0.02640*** [0.0053]INDDIR*B_CRISIS �0.01493*** [0.0041]INDDIR*D_CRISIS �0.01540*** [0.0052]EXPERTEXPERT*B_CRISIS �0.03181*** [0.0103]EXPERT*D_CRISIS �0.00408 [0.0144]COMITE �0.00627** [0.0029]COMITE*B_CRISIS �0.00125 [0.0056]COMITE*D_CRISIS 0.00822 [0.0052]MEYEAR �0.00205*** [0.0003]MEYEAR*B_CRISIS �0.00131* [0.0007]MEYEAR*D_CRISIS �0.00292*** [0.0007]DUAL 0.00593** [0.0027]DUAL*B_CRISIS 0.01359*** [0.0042]DUAL*D_CRISIS �0.01018* [0.0060]REMU 0.00561 [0.0034]REMU*B_CRISIS 0.05037*** [0.0117]REMU*D_CRISIS 0.02756** [0.0125]CEO AGE �0.00034 [0.0002] �0.00121*** [0.0003] �0.00082*** [0.0002]CEO TENURE �0.00191** [0.0009] �0.00102 [0.0010] �0.00195** [0.0008]CEO ROTATION �0.00284 [0.0026] �0.00059 [0.0028] �0.00145 [0.0019]CAP 0.19729*** [0.0317]CAP*B_CRISISCAP*D_CRISISREGP �0.00248 [0.0047]REGP*B_CRISIS �0.01316** [0.0064]REGP*D_CRISIS �0.01188 [0.0094]CRISK 0.12715*** [0.0472] 0.24617*** [0.0519] 0.02988 [0.0368]LRISK �0.00174 [0.0030] 0.00356 [0.0034] �0.00055 [0.0026]LNCSIZE �0.00091 [0.0008] �0.00003 [0.0008] 0.00085 [0.0006]END �0.00012 [0.0001] �0.00031*** [0.0001] �0.00005 [0.0001]Constant 0.05231*** [0.0154] 0.09349*** [0.0167] 0.09519*** [0.0176]R2 0.3268 0.1324 0.3798Number of observations 93 90 90F-statistic 29.36*** 33.64*** 32.54***Hausman test (p-value) 5.37 (0.97) 12.71(0.54) 18.35(0.19)

Notes: NIM: The net interest margin is expressed as the percentage of total assets. The net interest income is thedifference between the interest income and the interest costs; B_CRISIS (D_Crisis) is a dummy variable for the before(during) the 2008 financial crisis period; BOARDSIZE is a measure of the size of the board of directors. It is equal to thenumber of directors sitting on the BD at the annual meeting; INDDIR is the proportion of independent directors isdetermined by the number of independent directors divided by the total number of directors; EXPERT is the proportion offinancial expert directors in the board is determined by the ratio of inside directors to the total number of directors in thebank; COMITE is a dummy variable that takes value 1 if the bank has a remuneration committee, and 0 otherwise;MEYEAR is a measure of the frequency of meetings of the board calculated as the number of directors’ meetings held eachyear; DUAL is a measure of duality based on a dummy variable that takes value 1 when both functions are occupied by thesame person (CEO is the chairman of the board), and 0 otherwise; REMU is the total cash compensation of the CEO and isdefined as the annual salary plus bonus; CEO AGE: the CEO’s age is expressed in years; CEO TENURE is the CEO’smandate is expressed in terms of years of the bank’s CEO power; CEO ROTATION: this is a dummy variable that takesvalue 1 when the CEO is changed, and 0 otherwise; CAP is the minimum capital requirement variable is the equal equity tototal assets as a proxy for measuring the solvency ratio; REGP is a regulatory pressure variable that is equal to the unity ifthe bank’s solvency ratio (CAP) is less than 8 per cent, and 0 otherwise; CRISK is a credit risk variable measured by theratio of the provisions for loan losses to the overall loans; LRISK is the liquidity risk ratio defined as the ratio of “liquidassets to the customer and to short-term investments”; LNCSIZE is a measure of the size of the bank and is calculated asthe logarithm of the book value of total assets; END is a measure of indebtedness and is calculated as the ratio of total debtto total equity; ***, ** and *indicate statistical significance, respectively, at the levels 0.01, 0.05 and 0.10. Standard errorsare reported in brackets

Corporategovernance

363

adopt panel regression models that link the banking internal governance mechanisms (BDand CEO’s compensation), external mechanisms (minimum required capital and regulatorypressure) and banking performance (ROA, ROE and IR).

This research contributes to the literature dealing with the CEO’s compensation, board’sattributes, banking regulation and governance. It also shows that CEO’s compensation is aninternal mechanism for his/her control. On the other hand, this assessment indicates that theattributes of the BD represent one of the mechanisms of the CEO control, and on the otherhand, banking regulation is an external supervising mechanism to control the bank’scapital.

The obtained results show that banks conduct arbitrages between the differentgovernance mechanisms to alleviate the intensity of agency conflict between theshareholders and managers. Actually, these results confirm the complementary hypothesisbetween the internal and external governance mechanisms during the 2008 financial crisis.In fact, all of our results show that the banks’ supervision via banking regulations should bestrengthened using internal governance mechanisms, which implies a good bankingperformance during the 2008 financial crisis. Our results reflect the complementaritybetween internal governance mechanisms and banking regulations in the Eurozone banksduring the 2008 financial crisis.

In light of our results, we believe that, during the 2008 financial crisis, mainly between2007 and 2009, there were two mutually strengthened complementary pressures which wereexerted on the management decisions regarding risk-taking, financial performance.However, before the 2008 financial crisis, between 2004 and 2006, an alternative link isobserved between the governance internal mechanisms and capital regulation. Overall,during the 2004-2009 period, there was a compromise link. Actually, the pressure exerted byregulation seems to be reinforced using the internal governance mechanisms for the purposeof ensuring the soundness and efficiency of the banking activity while improving andensuring its performance during the period of the 2008 crisis. Moreover, it can be confirmedthat the internal governance mechanisms and capital regulation are determinants ofbanking performance.

Indeed, the conclusion drawn from the results of this analysis is that the extent to whichthe governance mechanisms are used depends not only on factors related to the bank and itsenvironment but also on the extent of using other governance mechanisms available for thebank, which can perform the same function. However, the 2008 financial crisis that severelyaffected the financial markets and the world economy underlined the limitations of theregulatory framework defined by the Basel II agreements. Even if Basel II enabled a majorbreakthrough in improving the risk measurement methods in banking institutions, thiscrisis revealed the incapacity of this device to properly cover all the risks (liquidity risk, pro-cyclicality risk, and market risk). Under the leadership of the G20, the Basel Committeedeveloped a set of new measures called Basel III to strengthen Basel II capital adequacyratios and improve more broadly the supervision system of the financial sector on theinternational basis to prevent future systemic crises.

Furthermore, although the banking activity has become more than ever a key toeconomic development in the world, good governance is the very expression of developmentas a perfect regime. Modernizing the European banking system is, therefore, anunprecedented challenge. The changes in banking governance help banks find a useful andnecessary way to avoid ill-considered risks that can cause a systemic risk. Hence, someconditions should be met so that banking governance can contribute to economicdevelopment. Consequently, culture and mentality of good banking governance must growas much as possible through awareness-raising, training, promotion, recognition of

MAJ34,3

364

performance, enhancing procedure transparency and the stability of good bankinggovernance and regulations, strengthening national capacity to fight against corruption andpreventive mechanisms.

Finally, our analysis can be extended through studying the interaction between internaland external governance mechanisms and their impact on the 2008 financial crisis.

ReferencesAebi, V., Sabato, G. and Schmid, M. (2012), “Risk management, corporate governance, and bank

performance in the financial crisis”, Journal of Banking and Finance, Vol. 36 No. 12,pp. 3213-3226.

Alexandre, H. and Bouaiss, K. (2009), “The complementarity of regulatory and internal governancemechanisms in banks”, Bankers, Market and Investors, Vol. 98, pp. 6-15.

Alshimmiri, T. (2004), “Board composition, executive remuneration, and corporate performance: thecase of REITS”, Corporate Ownership and Control, Vol. 2, pp. 104-118.

Anderson, R.C. and Fraser, D.R. (2000), “Corporate control, bank risk taking, and the health of thebanking industry”, Journal of Banking and Finance, Vol. 24 No. 8, pp. 1383-1398.

Andres, P. and Vallelado, E. (2008), “Corporate governance in banking: the role of the board ofdirectors”, Journal of Banking and Finance, Vol. 32 No. 12, pp. 2570-2580.

Ang, J., Lauterbach, B. and Schreiber, B.Z. (2002), “Pay at the executive suite: how do US bankscompensate their top management teams?”, Journal of Banking and Finance, Vol. 26 No. 6,pp. 1143-1163.

Angbazo, L. (1997), “Commercial bank net interest margin, default risk, interest rate risk, and off-balance sheet banking”, Journal of Banking and Finance, Vol. 21 No. 1, pp. 55-87.

Athanasoglou, P.P., Brissimis, S.N. and Delis, M.D. (2008), “Bank-specific, industry-specific andmacroeconomic determinants of bank profitability”, Journal of International Financial Markets,Institutions andMoney, Vol. 18, pp. 121-136.

Ayadi, N. and Boujelbène, Y. (2012), “Compensation of the CEO, board of directors and bank risktaking”, Journal of Advances Research inManagement, Vol. 3, pp. 4-16.

Baker, M. and Gompers, P.A. (2003), “The determinants of board structure at the initial public offering”,Journal of Law and Economics, Vol. 46 No. 2, pp. 569-598.

Barro, J.R. and Barro, R.J. (1990), “Pay, performance, and turnover of bank CEOs”, Journal of LaborEconomics, Vol. 8 No. 4, pp. 448-481.

Basel Committee on Banking Supervision (2006), Enhancing Corporate Governance for BankingOrganizations, Bank for International Settlement, Basel.

Basel Committee on Banking Supervision (2015), Corporate Governance Principles for Banks, Bank forInternational Settlement, Basel.

Becher, D.A. and Frye, M.B. (2011), “Does regulation substitute or complement governance?”, Journal ofBanking and Finance, Vol. 35 No. 3, pp. 736-751.

Belkhir, M. (2009), “Board structure, ownership structure and firm performance: evidence frombanking”,Applied Financial Economics, Vol. 19 No. 19, pp. 1581-1593.

Black, B.S. (1998), “Institutional investors and the case for institutional voice”, in Stern, J.M. andChew, D.H. (Eds), The Revolution in Corporate Finance, Blackwell Publishers, Hoboken,NJ.

Bouwman, C.H.S. (2009), “The governance of banks: how do bank capital and monitoring impact post-acquisition risk taking and performance?”, Working Paper, CaseWestern Reserve.

Boyd, J. and Hakenes, H. (2014), “Looting and risk shifting in banking crises”, Journal of EconomicTheory, Vol. 149, pp. 43-64.

Corporategovernance

365

Boyd, J., Chang, C. and Smith, B.D. (1998), “Moral hazard under commercial and universal banking”,Journal of Money, Credit, and Banking, Vol. 30 No. 3, pp. 426-468.

Broye, G. and Moulin, Y. (2010), “Rémunération des dirigeants et gouvernance des entreprises: Le casdes entreprises françaises cotées”, Revue Finance Contrôle Stratégie, Vol. 13, pp. 67-98.

Buser, S., Chen, A. and Kane, E. (1981), “Federal deposit insurance, regulatory policy, and optimal bankcapital”, Journal of Finance, Vol. 36, pp. 51-60.

Caprio, G., Laeven, L. and Levine, R. (2007), “Governance and bank valuation”, Journal of FinancialIntermediation, Vol. 16 No. 4, pp. 584-617.

Cornett, M.M., McNutt, J.J. and Tehranian, H. (2009), “Corporate governance and earningsmanagement at large US bank holding companies”, Journal of Corporate Finance, Vol. 15No. 4, pp. 412-430.

Dalton, D.R., Daily, C.M., Johnson, J.L. and Ellstrand, A.E. (1999), “Number of directors and financialperformance: a meta-analysis”,Academy ofManagement Journal, Vol. 42, pp. 674-686.

De Coussergues, S. (1996), La Banque (Structure- Marché- Gestion), 2e éd., Dalloz, Paris.Demsetz, H. (1983), “The structure of ownership and the theory of the firm”, Journal of Law and

Economics, Vol. 26 No. 2, pp. 375-390.Demsetz, H. (1986), “Corporate control, insider trading, and rates of return”, American Economic

Review, Vol. 76, pp. 313-316.Demsetz, H. and Lehn, K. (1985), “The structure of corporate ownership: causes and consequences”,

Journal of Political Economy, Vol. 93 No. 6, pp. 1155-1177.Dormont, B. (1989), “Introduction à l’économétrie des données de panel: théory et applications à des

échantillons d’entreprises”, Editions du Centre National de la Recherche Scientifique, p. 125.

El-Chaarani, H. (2014), “The impact of corporate governance on the performance of Lebanese banks”,International Journal of Business and Finance Research, Vol. 8, pp. 35-46.

Esty, B. (1998), “The impact of contingent liability on commercial bank risk taking”, Journal ofFinancial Economics, Vol. 47 No. 2, pp. 189-218.

European Commission (2010), “Communication from the commission to the European parliament, thecouncil, the European economic and social committee and the committee of the regions”,Brussels, available at: http://ec.europa.eu/research/innovation-union/pdf/innovation-union-communication_en.pdf

Fernandes, N. and Fich, E.M. (2009), “Does financial experience help banks during credit crises?”,Working Paper, Drexel University.

Galai, D. and Masulis, R. (1976), “The option pricing model and the risk factor of stock”, Journal ofFinancial Economics, Vol. 3 Nos 1/2, pp. 53-81.

García-Meca, E., Garcia-Sanchez, I.M. and Martínez-Ferrero, J. (2015), “Board diversity and its effects onbank performance: an international analysis”, Journal of Banking and Finance, Vol. 53,pp. 202-214.

Gedajlovic, E. and Shapiro, D.M. (2002), “Ownership structure and firm profitability in Japan”, TheAcademy ofManagement Journal, Vol. 45, pp. 565-575.

Godlewski, C. (2005), “Bank capital and credit risk taking in emerging market economies”, Journal ofBanking Regulation, Vol. 6 No. 2, pp. 128-145.

Grossman, S. and Hart, O. (1986), “The costs and benefits of ownership: a theory of vertical and lateralintegration”, Journal of Political Economy, Vol. 94 No. 4, pp. 691-719.

Haan, J. and Vlahu, R. (2016), “Corporate governance of banks: a survey”, Journal of Economic Surveys,Vol. 30, pp. 228-277.

Hadlock, C.J., Lee, D.S. and Parrino, R. (2002), “Chief executive officer careers in regulatedenvironments: evidence from electric and gas utilities”, Journal of Law and Economics, Vol. 45No. 2, pp. 535-563.

MAJ34,3

366

Hagendorff, J., Collins, M. and Keasey, K. (2010), “Board monitoring, regulation, and performance in thebanking industry: evidence from the market for corporate control”, Corporate Governance: AnInternational Review, Vol. 18 No. 5, pp. 381-395.

Hau, H. and Thum, M. (2009), “Subprime crisis and board (in-)competence: private vs. public banks inGermany”, Economic Policy, Vol. 24 No. 60, pp. 701-751.

Hausman, J.A. (1978), “Specification tests in econometrics”, Econometrica, Vol. 68, pp. 1251-1271.Himmelberg, C.P., Hubbard, R.G. and Palia, D. (1998), “Understanding the determinants of managerial

ownership and the link between ownership and performance”, Journal of Financial Economics,Vol. 53 No. 3, pp. 353-384.

Holderness, C.G., Kroszner, R.S. and Sheehan, D.P. (1999), “Were the good old days that good? Changesin managerial stock ownership since the great depression”, Journal of Finance, Vol. 54 No. 2,pp. 435-569.

Hoshi, T., Kashyap, A. and Scharfstein, D. (1990), “The role of banks in reducing the costs of financialdistress in Japan”, Journal of Financial Economics, Vol. 27 No. 1, pp. 67-88.

Houston, J. and James, C. (1995), “CEO compensation and bank risk: is compensation in bankingstructured to promote risk taking? ”, Journal of Monetary Economics, Vol. 36 No. 2, pp. 405-431.

Jensen, M.C. (1993), “Themodern industrial revolution, exit, and the failure of internal control systems”,Journal of Finance, Vol. 48 No. 3, pp. 831-880.

Jensen, M. and Meckling, W. (1976), “Theory of the firm: managerial behavior, agency costs, andownership structure”, Journal of Financial Economics, Vol. 3 No. 4, pp. 305-360.

John, K., Litov, L. and Yeung, B. (2008), “Corporate governance and risk taking”, The Journal ofFinance, Vol. 63 No. 4, pp. 1679-1728.

Joshi, M., Cahill, D., Sidhu, J. and Kansal, M. (2013), “Intellectual Capital and financial performance: anevaluation of the Australian financial sector”, Journal of Intellectual Capital, Vol. 14 No. 2,pp. 264-285.

Kane, E. (1985),The Gathering Crisis in Federal Deposit Insurance, MIT Press, Cambridge, MA.Keeley, M. (1990), “Deposit insurance, risk, and market power in banking”,American Economic Review,

Vol. 80, pp. 1183-1200.

Kennedy, P. (1985),AGuide to Econometrics, 2nd ed., MIT Press, Cambridge, MA.Kim, D. and Santomero, A.M. (1988), “Risk in banking and capital regulation”, Journal of Finance,

Vol. 43 No. 5, pp. 1219-1233.Koehn, M. and Santomero, A.M. (1980), “Regulation of bank Capital and portfolio risk”, Journal of

Finance, Vol. 35 No. 5, pp. 1235-1250.Kole, S. and Lehn, K. (1999), “Deregulation and the adaptation of governance structure: the case of U.S.

airline industry”, Journal of Financial Economics, Vol. 52 No. 1, pp. 79-117.Konishi, M. and Yasuda, Y. (2004), “Factors affecting bank risk taking: evidence from Japan”, Journal of

Banking and Finance, Vol. 28 No. 1, pp. 215-232.La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R.W. (1998), “Law and finance”, Journal of

Political Economy, Vol. 106 No. 6, pp. 1113-1155.Laeven, L. and Levine, R. (2009), “Bank governance, regulation and risk taking”, Journal of Financial

Economics, Vol. 93 No. 2, pp. 259-275.Laing, D. and Weir, C.M. (1999), “Governance structures, size and corporate performance in UK firms”,

Management Decision, Vol. 37 No. 5, pp. 457-464.Liang, Q., Xu, P. and Jiraporn, P. (2013), “Board characteristics and Chinese bank performance”, Journal

of Banking and Finance, Vol. 37 No. 8, pp. 2953-2968.Levine, R. (2004), “The corporate governance of banks: a concise discussion of concepts and evidence”,

World Bank Policy ResearchWorking Paper.

Corporategovernance

367

Macey, J. and O’Hara, M. (2003), “The corporate governance of banks”, Federal Reserve Bank ofNew York Economic Policy Review, Vol. 9, pp. 91-107.

McAllister, P.H. and McManus, D. (1993), “Resolving the scale efficiency puzzle in banking”, Journal ofBanking and Finance, Vol. 17 Nos 2/3, pp. 389-405.

Merton, R. (1977), “An analytical derivation of the cost of deposit insurance and loan guarantees:an application of modern option pricing theory”, Journal of Banking and Finance, Vol. 1No. 1, pp. 3-11.

Miller, S.M. and Noulas, A.G. (1997), “Portfolio mix and large-bank profitability in the USA”, AppliedEconomics, Vol. 29 No. 4, pp. 505-512.

Minton, B.A., Taillard, J.P.A. and Williamson, R. (2014), “Financial expertise of the board, risk taking,and performance: evidence from bank holding companies”, Journal of Financial and QuantitativeAnalysis, Vol. 49 No. 2, pp. 351-380.

OECD (1996),The Knowledge Based Economy, OECD, Paris.Pathan, S. and Faff, R. (2013), “Does board structure in banks really affect their performance?”, Journal

of Banking and Finance, Vol. 37 No. 5, pp. 1573-1589.Persico, N. (2004), “Committee design with endogenous information”, The Review of Economic Studies,

Vol. 71 No. 1, pp. 165-194.

Pirotte, A. (2011), “Econométrie des données de panel: théorie et applications”, Economica, CollectionCorpus, p. 278.

Prowse, S.D. (1995), “Alternative methods of corporate control in commercial banks”, Economic Review– Federal Reserve Bank of Dallas, Vol. 24, pp. 24-36.

Sevestre, P. (2002), Econométrie Des Données de Panel, Dunod, Paris, pp. 109-152.Shleifer, A. and Vishny, R.W. (1986), “Large shareholders and corporate control”, Journal of Political

Economy, Vol. 94 No. 3, Part 1, pp. 461-488.Shleifer, A. and Vishny, R.W. (1997), “A survey of corporate governance”, Journal of Finance, Vol. 52,

pp. 737-783.Shrieves, R.E. and Dahl, D. (1992), “The relationship between risk and Capital in commercial banks”,

Journal of Banking and Finance, Vol. 16 No. 2, pp. 439-457.Sinkey, J.F. (1999), Commercial Bank Financial Management, 6th ed., Lavoisier, Paris, p. 696.Vafeas, N. (1999), “Board meeting frequency and firm performance”, Journal of Financial Economics,

Vol. 53 No. 1, pp. 113-142.

Valverde, S.C. and Fernandez, F.R. (2007), “The determinants of bank margins in European banking”,Journal of Banking and Finance, Vol. 31, pp. 2043-2063.

Veltri, S. and Silvestri, A. (2011), “Direct and indirect effects of human capital on firm value: evidencefrom Italian companies”, Journal of Human Resource Costing and Accounting, Vol. 15 No. 3,pp. 232-254.

Young, C.S., Su, H.Y., Fang, S.C. and Fang, S.R. (2009), “Cross-country comparison of IC performance ofcommercial banks in Asian economies”, The Service Industries Journal, Vol. 29 No. 11,pp. 1565-1579.

Further readingBasel Committee on Banking Supervision (2009), Principles for Sound Stress Testing Practices and

Supervision, Bank for International Settlement, Basel.

Basel Committee on Banking Supervision (2010), Range of Methodologies for Risk and PerformanceAlignment of Remuneration, Bank for International Settlement, Basel.

Brunnermeier, M. (2009), “Deciphering the liquidity and credit crunch 2007-08”, Journal of EconomicPerspectives, Vol. 23, pp. 7-100.

MAJ34,3

368

Erkens, D.H., Mingyi, H. and Pedro, M. (2012), “Corporate governance in the 2007-2008 financial crisis:evidence from financial institutions worldwide”, Journal of Corporate Finance, Vol. 18 No. 2,pp. 389-411.

Gonzales, F. (2005), “Bank regulation and risk-taking incentives: an international comparison of bankrisk”, Journal of Banking and Finance, Vol. 29, pp. 1153-1184.

Moody’s Manuals (2018), Moody’s Manuals: Industrial, public utility, government and municipalities,and banking and financial, New York, various volumes, 1921-1941.

Pathan, S. (2009), “Strong boards, CEO power and bank risk-taking”, Journal of Banking and Finance,Vol. 33 No. 7, pp. 1340-1350.

Stigler, G.J. and Friedland, C. (1962), “What can regulators regulate – the case of electricity”, Journal ofLaw and Economics, Vol. 5, pp. 1-16.

Taylor, J.B. (2009), “Economic policy and the financial crisis: empirical analysis of what went wrong”,Critical Review, Vol. 21 Nos 2/3, pp. 341-364.

Weir, C., Laing, D. and McKnight, P.J. (2002), “Internal and external governance mechanisms: theirimpact on the performance of large UK public firms”, Journal of Business Finance andAccounting, Vol. 29 Nos 5/6, pp. 579-611.

Corporategovernance

369

Appendix

About the authorsMohamed A. Ayadi is a Professor of Finance, Department of Finance, Operations, and InformationSystems, Goodman School of Business, Brock University, St Catharines, ON, Canada, L2S 3A1, (905)688-5550, x3917. Dr Ayadi’s research interests are in the areas of investment management, mutualfunds, corporate governance, financial markets regulation and derivative securities. He is particularlyinterested in Canadian and international mutual funds. He has made numerous presentations atseveral domestic and international finance conferences. His papers have been published in journalssuch as the Journal of Banking and Finance, Journal of Empirical Finance, Journal of Futures Markets,International Review of Finance, Journal of Financial Services Research, European Journal ofOperations Research, Computers and Operations Research, Quarterly Journal of Business andEconomics, Quarterly Review of Economics and Finance and Research in International Business andFinance. He is an active reviewer for a number of finance and OR journals, books, conferences,FQRSC, SSHRC and the Romanian National Council for Scientific Research.

Nesrine Ayadi is an Assistant Professor, Department of Finance, Faculty of Economics andManagement, University of Sfax, Tunisia. Professor Ayadi’s research interests are in the areas incorporate governance, assets pricing and corporate finance.

Table AI.List of Europeanbanks

No. Banks Country

1 Group Banque Populaire France2 Natexis France3 Crédit Coopératif France4 Crédit Lyonnais France5 Credit Foncier France France6 Postal Bank France7 CIC France8 BNP Paribas France9 Société Général France10 Banque Populaire Caisse d’Epargne France11 Crédit du Nord France12 HSBC France France13 Crédit Agricol France14 CFCAL Banque France15 Group BCGE France16 BFCM France17 ABN Amro NV France18 RBS France19 National Bank of Pakistan Belgium20 Banque Privée Belgium Belgium21 JP Morgan Chase Bank Belgium Belgium22 NIBC Belgium Belgium23 ING Belgium SA Germany24 Barclays Bank Germany25 Deutch Bank Germany26 Commerzbank Germany27 Deuttsh Postbank Germany28 Aareal Bank AG Germany29 Nordea Bank Finland30 Danske Bank Finland

Note: The sample consists of 30 banks from four Eurozone countries

MAJ34,3

370

Samir Trabelsi is a Professor of Accounting, Department of Accounting, Goodman School ofBusiness, Brock University, St Catharines, ON, Canada, L2S 3A1, (905) 688-5550, x4463. Dr Trabelsi’sresearch interests are in the area of voluntary disclosure, corporate governance, mutual fundgovernance, earning conservatism and eXtensible Business Reporting Language (XBRL). Dr Trabelsiwon the Best Paper Award at the 2008 Journal of Contemporary Accounting and Economics (JCAE)and Auditing: A Journal of Practice and Theory (AJPT) Joint Symposium. He received the 2008 BestPaper Award in the Accounting Division at the Administrative Sciences Association of Canada. Hiswork is currently supported by the Social Sciences and Humanities Research Council (SSHRC) ofCanada through its International Opportunities Fund: Project Grants. Samir Trabelsi is thecorresponding author and can be contacted at: strabelsi@brocku.ca

For instructions on how to order reprints of this article, please visit our website:www.emeraldgrouppublishing.com/licensing/reprints.htmOr contact us for further details: permissions@emeraldinsight.com

Corporategovernance

371

ErratumIt has come to the attention of the publishers that the articles “A framework for enterpriserisk identification and management: the resource-based view”, by Birendra K. Mishra, ErikRolland, Asish Satpathy and Michael Moore, and “Mitigating financial leakages througheffective money laundering investigation”, by Sharifah Nazatul Faiza Syed Mustapha Nazri,Salwa Zolkaflil and Normah Omar, were incorrectly published in Managerial AuditingJournal Volume 34, Issue 2, rather than in the full Organizational Risk, Fraud, Forensics,Anti Money Laundering Laws and Controls and Corporate Corruption special issue. Thiserror was introduced in the editorial process, and the publisher sincerely apologises for thiserror and for any inconvenience caused.

MAJ34,3

372

Managerial Auditing JournalVol. 34 No. 3, 2019p. 372© Emerald PublishingLimited0268-6902DOI 10.1108/MAJ-03-2019-018