HOW DO MNCS VOTE IN DEVELOPINGCOUNTRY ELECTIONS?

PAUL M. VAALERUniversity of Minnesota

Research on multinational corporations (MNCs) and host government political risk indeveloping countries has largely ignored local electoral politics, economic policies,and the MNC investment incentives they may generate. In response, I develop and testa framework for understanding MNC risk and investment behavior based on politicalbusiness cycle considerations. Analyses of 408 MNC investments worth $199 billion in18 developing countries holding 35 presidential elections from 1987 through 2000 areconsistent with these considerations: MNCs perceive higher (lower) risk and announcefewer (more) investment projects as right-wing (left-wing) incumbents appear morelikely to be replaced by left-wing (right-wing) challengers.

Management research over five decades has in-vestigated political risk and investment behaviorrelated to the divergent interests of foreign-domi-ciled multinational corporations (MNCs) and hostgovernments in developing countries. Robinson(1963) identified political risk to internationalfirms operating in newly independent and “nation-alistic” countries with occasional interest inbreaching contracts or outright expropriation offirm assets. Vernon (1971) called attention to polit-ical risk associated with “obsolescing bargains” be-tween investing MNCs and developing countryhost governments over time. Kobrin (1979, 1987)articulated different components of political riskassociated with a “bargaining hypothesis” and dif-ferent MNC responses intended to mitigate thatrisk. A general decline in expropriations during the1980s and 1990s (Minor, 1994) coincided with newresearch directions regarding strategic actions de-veloping country host governments might take toattract larger shares of foreign investment (Murtha& Lenway, 1994) and what legal policies (LaPorta,

Lopez-de-Silanes, Shleifer, & Vishny, 1998), eco-nomic policies (Murtha, 1993), and political insti-tutional arrangements (Henisz, 2000) might con-strain government actions.

In the 2000s, interest remains strong in under-standing these topics: how governments matter forbusiness investment incentives (Ring, Bigley,D’Aunno, & Khanna, 2005); how MNC investmentdecisions in emerging economies differ from thosein industrialized democracies generally (Hoskis-son, Eden, Lau, & Wright, 2000) and, in particular,how MNCs’ investment willingness (Henisz & De-lios, 2001) and modes of investment (Delios & Hen-isz, 2000) evolve with their experience in managinglocal policy environments in emerging economies;and how MNCs identify and mitigate risks associ-ated with investing in and transforming formerlystate-owned enterprises (Zahra, Ireland, Gutierrez,& Hitt, 2000).

With such richly developed research streams, itis surprising that there are, to date, no theoreticalmodels or quantitative empirical evidence to guideunderstanding of MNC risk and investment behav-ior when host government economic policies, pol-itics, and political institutions are arguably mostvulnerable to change—that is, during elections.Past and present management research on obsolesc-ing bargains between MNCs and developing coun-try host governments, on reversals of economic pol-icies inducing MNC investment, and on MNCinvestment modes and strategies for privatizing en-terprises has not necessarily been tied to local elec-toral dynamics. Until the 1980s, this oversight mayhave been understandable. Developing countriesoften occupying researcher attention had one-partysystems, as in Mexico or Poland, or military-ledgovernments, as in Brazil or South Korea. With no

I thank Ruth Aguilera, Bill Bernhard, Steve Block,Hadi Esfahani, Glenn Hoetker, Matt Kraatz, David Leb-lang, Joe Mahoney, Gerry McNamara, Steve Michael,Greg Northcraft, Mike Peng, Mike Preis, Dennis Quinn,Burkhard Schrage, Mike Sher, Sushil Vachani, LouWells, Bernie Yeung, three AMJ anonymous reviewers,and AMJ Associate Editor Amy Hillman for helpful com-ments, criticisms, and suggestions. I thank Ben Esty,Barclay James, Shawn Riley, Burkhard Schrage, NatalyaShipachova, Farzad Taheripour, and Guzel Tulegenovafor help with data collection. I thank Randy Westgrenand the University of Illinois at Urbana-Champaign Cen-ter for International Business Education and Research forgenerous financial support. All remaining errors aremine.

� Academy of Management Journal2008, Vol. 51, No. 1, 21–43.

21

Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download or email articles for individual use only.

competitive electoral system, there is little likeli-hood of policy changes linked to voter preferences.But the last two decades saw substantial democra-tization in developing countries, often carried outwith the expectation that political modernizationwould enhance country attractiveness for foreigninvestment and economic growth (Goldsmith,1994; Haggard, 1990). In many developing coun-tries, parties from across the political spectrumhave competitive opportunities to hold office andshape policies affecting MNC risk and investmentbehavior. Management research should respond tothese developments with theoretical models andempirical evidence designed to observe and ex-plain risk and investment behavior during increas-ingly frequent election periods.

In this study, I develop and test hypotheses de-rived from a framework of election-period MNCrisk and investment behavior based on politicalbusiness cycle theory, which is more familiar topolitical economy than management researchers.Since the seminal work of Nordhaus (1975), polit-ical business cycle models and empirical evidencehave been debated largely in the context of indus-trialized democracies and interactions betweenelected officials and voters. These original modelsand their descendants (Drazen, 2000; Rogoff, 1990)posit that opportunistic incumbent politicians useexpansionary fiscal, monetary, and related policiesduring election periods to garner voter support,even though such policies often have detrimentalpostelection economic consequences. Models de-veloped by Hibbs (1977) and refined by others(Alesina, 1987; Alesina, Roubini, & Cohen, 1997)also suggest that politicians implement economicpolicies for electoral purposes. But “partisan mod-els” in this stream differ from the earlier “opportu-nistic models.” In partisan models, right-wing andleft-wing politicians implement different types ofexpansionary policies. Right-wing politicians putin place policies promoting lower inflation and theinterests of investors, but left-wing politicians im-plement policies promoting lower unemploymentand the interests of workers. Thirty years of empir-ical work summarized recently by Drazen (2000)and Block and Vaaler (2004) have shown mixedsupport for both the opportunistic and partisan po-litical business cycle models in industrialized de-mocracies but consistent support for both types ofmodels during the last decade in recently democ-ratizing countries from the developing world.

My study builds on these foundations. It extendsthe political business cycle domain beyond inter-actions among elected officials, voters, and localeconomies. It promises at least two contributions tomanagement research on MNC risk and investment

behavior related to host government politics in de-veloping countries. The first contribution is theo-retical. My study provides management researcherswith the first theoretical framework for understand-ing MNC risk and investment behavior during elec-tion periods in developing countries where politi-cal business cycle theories suggest that localpoliticians have incentives to vary economic poli-cies to suit their electoral aspirations. My theoreti-cal framework is motivated by the proposition thatforeign-domiciled MNCs watch local politicians,their policies, and likely electoral outcomes and“vote” during election periods on the basis of op-portunistic and partisan considerations. In keepingwith opportunistic considerations, MNCs may per-ceive more (less) risk to the extent that incumbentpoliticians are unpopular (popular), thus prompt-ing (avoiding) election-period spending sprees thatmay be detrimental to postelection investment en-vironments. In keeping with partisan consider-ations, MNCs may also perceive more (less) risk tothe extent that right-wing (left-wing) incumbentswith investor-friendly (worker-friendly) policiesare likely to go down in defeat to the possibledetriment (benefit) of postelection investment en-vironments. Unlike previous political business cy-cle research, which has followed either an oppor-tunistic or a partisan logic, I combine both logicsinto an integrated theoretical framework to derivehypotheses about election-period MNC risk and in-vestment behavior—or voting—in developingcountries.

My integrated theoretical framework finds prece-dents in previous studies by Vaaler, Schrage, andBlock (2005, 2006), who developed similar frame-works to explain election-period risk assessmentsby investors in “sovereign” bonds issued by devel-oping country governments (Vaaler et al., 2005) andelection period sovereign risk assessments of na-tional government creditworthiness by major cred-it-rating agencies (Vaaler et al., 2006). Althoughnoteworthy, these previous frameworks and relatedevidence about foreign financial actors and elec-tion-period risk may not easily extend to MNCmanagers and the investment projects they sponsorin developing countries. Skeptics might argue thatforeign-domiciled financial actors, like bondhold-ers and rating agencies, work in a world quite dif-ferent from that of strategic managers working inMNCs. Bondholders determining sovereign bondyields and credit agencies rating government cred-itworthiness operate in institutional settings thatpermit fast, low-cost responses to changes in localpolitics. But strategic managers in MNCs assessrisks and take decisions about whether to constructand operate hydroelectric dams, automobile man-

22 FebruaryAcademy of Management Journal

ufacturing plants, and hotel resorts whose con-struction and operating costs run in the millions orbillions of U.S. dollars and whose expected life-spans are measured in years or decades. These in-vestment decisions are exemplars of the kind ofdifficult-to-reverse commitment that Ghemewat(1991) labeled “the dynamic of strategy” and high-lighted as a key source of firm performance differ-ences. Short-term electoral politics and economicpolicies in developing countries may be largelyirrelevant to decisions about investment projectsspanning several national governments, cam-paigns, and votes. Thus, my research propositionabout significant opportunistic and partisan politi-cal business cycle effects on MNC risk and invest-ment behavior competes with a plausible alterna-tive expectation of no political business cycleeffects.

In this research context, a second empirical con-tribution is promised. I test two hypotheses derivedfrom the integrated political business cycle frame-work using a novel empirical context. I analyzeelection-period trends in announcements of projectinvestments, a form of foreign direct investment(FDI) MNCs frequently choose in developing coun-tries. Project-finance-based FDI typically involvescreating a company in which “sponsoring” MNCsbecome equity investors with limited liability if theproject company fails. Most of the capital forproject companies comes from debt provided bylenders, who agree in advance to limit their re-course to project company assets (but not otherassets of sponsoring MNCs) in the event of failure.MNCS often use this approach to fund large-scale,long-term infrastructure, manufacturing, and ser-vice investment opportunities. I analyze annualcounts of 408 project investment announcementsworth $199 billion announced by foreign-domi-ciled MNCs in 18 developing countries holding 35presidential elections from 1987 through 2000. Noprevious empirical research in management has ex-amined this FDI form, particularly with the breadthof industry coverage and length of time comprisedby my sample. As will be seen below, this empiri-cal context also proves advantageous for assessingthe robustness of empirical model assumptions, in-cluding whether and how election-period politicalbusiness cycle considerations affect or are affectedby MNC project announcements.

My analyses yield results consistent with bothhypotheses and the broader theoretical frameworklinking election-period MNC risk and investmentbehavior to political business cycle considerations.The annual count of new investment projects an-nounced by MNCs decreases significantly and sub-stantially as the likelihood of right-wing incumbent

government defeat on election day increases. Thecount and implied dollar amount of announcedinvestment projects drop to zero in years whenleft-wing challengers with less investor-friendlypolicies are likely to replace right-wing incum-bents. By contrast, the count and implied dollaramount of announced investment projects increasesignificantly and substantially as more investor-friendly right-wing challengers appear more likelyto defeat left-wing incumbents, an indication thatpartisan considerations dominate contrary oppor-tunistic considerations in MNC risk and invest-ment behavior. These election-year effects on thecount of MNC project announcements translateinto swings worth hundreds of millions or evenbillions of dollars in FDI. Developing country po-litical business cycles have statistically significantand economically substantial effects on long-terminfrastructure, manufacturing, and service projectssponsored by MNCs that are similar to politicalbusiness cycle effects documented previously(Vaaler et al., 2005, 2006) for developing countrybondholders and major credit agencies. Morebroadly, these results suggest that political busi-ness cycle theoretical models and developing coun-try empirical settings provide management re-searchers with new lenses and evidentiary sourcesfor broadening and deepening understanding of of-ten divergent but perhaps at times convergent in-terests of investing MNCs and host governments.

EMPIRICAL CONTEXT

Additional explanation of institutional practicesassociated with project investment in developingcountries provides helpful context for building atheoretical framework to predict changes in MNCrisk and investment behavior linked to opportunis-tic and partisan political business cycle consider-ations. For this description, I rely primarily on Esty(2003, 2004). Project investment, also described asproject finance investment, is defined as direct in-vestment using a legally independent project com-pany financed by equity from a sponsoring firm orsyndicate of sponsoring firms, and by nonrecoursedebt. Typically, a project company has a lead spon-sor with the largest single equity stake, oversightresponsibility for project operations, and controlover strategic decisions. A lead sponsoring MNCoften engages junior sponsors in an investing syn-dicate as well as lenders to provide additionalfunds. It may also engage specialist suppliers toprovide equipment and services for project con-struction and operation. In contrast to the lendersin other MNC investment structures, project fi-nance lenders and other suppliers typically agree to

2008 23Vaaler

rely exclusively for repayment on receipts gener-ated by and guarantees given to the investmentproject. The project company is bankruptcy remote,thus effectively separating its risk profile fromthose of the MNC parents of the lead sponsor andany junior sponsors in the syndicate. Under theseconditions, sponsors can undertake riskier invest-ments with less concern that an individual projectfailure will threaten MNC assets elsewhere. Aproject company and its various stakeholders aretightly focused on a single line of business—aproject—which typically has a construction andoperation life span of 5–15 years in manufacturingand upwards of 30 years if it is an infrastructureproject such as a hydroelectric generator or sewagetreatment plant.

This investment structure lends itself well to thehigher-risk-and-return environment of developingcountries. Since the mid 1960s, the number of an-nounced project investments in countries outsidethe Organisation for Economic Cooperation and De-velopment has topped 2,200; these projects areworth more than US$1.6 trillion. In some develop-ing countries, such as the Philippines and Indone-sia, more than 75 percent of inward FDI in the1990s came through project investment companies.For other developing countries, project invest-ments have become a substantial percentage ofoverall inward FDI.

Project investments in developing countries fo-cus primarily on infrastructure industries such asconstruction, transportation, energy generation andtransmission, telecommunications, and water andsewage. In the 1990s, project investments in devel-oping countries were frequently established as partof host government privatization policies. For ex-ample, the Philippines’ Maynilad Water Serviceswater treatment project announced in 1997 in-volved a syndicate led by a France-based MNC,Lyonnaise des Eaux, S.A. Initial construction andfacilities upgrade costs were valued at announce-ment at approximately $150 million. The Mayniladwater project was expected to generate over $7billion worth of infrastructure investments over its25-year life (Manila Times, 2003). Project invest-ment structures have also proved popular histori-cally and currently for mining and power genera-tion. One of the earliest examples of projectinvestment in mining, Bougainville Copper Ltd. inPapua New Guinea (Hammond & Allan, 1974), isalso one of the most popular cases used in manage-ment education to analyze issues of MNC invest-ment and risk related to host government politics.United Kingdom– and Australia-based Rio-TintoZinc sponsored a project company in the mid 1960sto construct a multimillion dollar copper mine,

preliminary refining facility, deep-water port, andrelated housing. The start of operations in the early1970s coincided with Papua New Guinea’s inde-pendence from Australia and founding elections.Competing factions and policies for dealing withthe now foreign-domiciled MNC led to substantialrenegotiation of the original mining concessionterms. Case study interest extends into the 1990swith Enron Development Corporation’s DahbolPower Project misadventure in Maharashtra State,India (Wells, 1997). State elections and a change ingovernment led to renegotiation of Enron’s earlierconcession agreement and the project’s eventualabandonment. My study complements case re-search interest with more formal theoretical mod-eling and broad-sample quantitative study of elec-tion period risk and MNC project investmentbehavior in developing countries guided by politi-cal business cycle considerations.

THEORETICAL FRAMEWORKAND HYPOTHESES

With this institutional context, I develop a theo-retical framework of MNC risk and investment be-havior integrating both partisan and opportunisticpolitical business cycle considerations. From thisframework, I derive two hypotheses. The frame-work follows similar ones explaining electoral pe-riod risk assessments by sovereign bondholders(Vaaler et al., 2005) and major credit rating (Vaaleret al., 2006). The framework builds on two impor-tant assumptions drawn from political business cy-cle theory. The first assumption relates to opportu-nistic incentives and MNC project investment. Ifollow Nordhaus’s (1975) model and other oppor-tunistic political business cycle models showingthat elected politicians have incentives to engage inexpansionary economic policies in the run-ups toelections and contractionary policies in postelec-tion environments, the net effect of which can bedetrimental to sustained economic development.1

1 The assumption underlying the Nordhaus (1975) op-portunistic model, for example, is that all incumbents,both left- and right-wing, behave the same. They tend toengage in fiscal spending sprees that increase output anddecrease unemployment just before an election. Inflationaccelerates in the run-up to election day, but it peaks andis observed by voters after the election. At that time,incumbents (or successful challengers) typically reduceinflation with fiscal austerity policies that also loweroutput and increase unemployment. Alternatively, poli-ticians tolerate permanently higher inflation and the ero-sion of gains in nominal wages, salaries, and fixed assetvalues. Supporting his opportunistic political business

24 FebruaryAcademy of Management Journal

But I assume in my framework that the likelihoodof incumbent electoral victory modifies opportu-nistic political business cycle incentives. Incum-bents certain of victory have fewer incentives toresort to opportunistic policies than those facingeither close elections or likely defeat. This assump-tion follows Schultz (1995), who showed that ex-pectations of incumbent party victory in Britishparliamentary elections were negatively correlatedwith the likelihood of pre-election expansionaryeconomic policies, and Block, Singh, and Feree(2003), who observed similar trends in sub-SaharanAfrica.

Opportunistic political business cycle incentivesmoderated in intensity by incumbent popularitymay have substantial impact on MNC willingnessto invest, even when the projects involved have lifespans of years or decades. Higher inflation in theaftermath of an election can erode the real value ofnominal returns from MNC project operations inearly years. Decreasing near-term returns can alsodepress longer-term project valuation and attrac-tiveness. Similarly, fiscal contraction in the after-math of an election may decrease the pool of gov-ernment funds available to subsidize MNC projectconstruction. Higher construction costs requirehigher future operating returns. Together, these ar-guments suggest that postelection investment envi-ronments become less attractive for MNCs to theextent that incumbents resort to expansionary eco-nomic policies during election years.

My second assumption relates to partisan politi-cal business cycle incentives and MNC project in-vestment. I assume that right-wing policies favorMNC project investment more than left-wing poli-cies. Partisan political business cycle researchsince Hibbs (1977) has articulated differences inright- versus left-wing economic policies in termsof a Phillips curve, whereby right-wing policiesfavor less inflation at the expense of more unem-ployment and left-wing policies favor the oppositetrade-off. More recent partisan political businesscycle research (Alesina et al., 1997) has expandedthis simple distinction to contrast the broader in-

vestor friendliness of right-wing policies loweringinflation and taxes, and preserving fixed asset val-ues to the broader worker friendliness of left-wingpolicies permitting more inflation, taxes, and assetdevaluation if such policies also lower unemploy-ment. Leblang and Mukherjee (2005) documentedmovements in U.K. and U.S. stock market pricesconsistent with right- versus left-wing policy pref-erences in U.S. and U.K. governments from 1930through 2000. In the run-up to developing countryelections in which right-wing (left-wing) incum-bents were likely to lose to left-wing (right-wing)challengers, Vaaler, Schrage, and Block (2005)showed that the credit risk premium demanded byinvestors in developing country sovereign bondsincreased (decreased) in line with partisan politicalbusiness cycle considerations of higher (lower)credit risks when a left-wing (right-wing) party vic-tory was likely.

Partisan political business cycle incentives mayalso have substantial impact on MNC willingnessto invest in projects with life spans of years ordecades. Job creation policies stoking inflation inthe aftermath of a right- to left-wing switch can alsoerode the real value of nominal returns from theearly years of a project’s operation, thus depressingproject valuation and attractiveness. On the otherhand, investor-friendly policies, such as targetedtax cuts for new project construction coupled withfiscal discipline and balanced budgets, may bemore likely after a left- to right-wing switch. Such apartisan policy switch can reduce constructioncosts and protect the real value of nominal returnsfrom project operations, both of which increaseproject valuation and attractiveness. This logic sug-gests that postelection investment environmentsbecome less (more) attractive to MNCs to the extentthat less (more) investor-friendly left-wing (right-wing) parties are likely to prevail in election years.

Guided by these two assumptions, I define inFigure 1 an integrated political business cycle the-oretical framework for explaining MNC project in-vestment willingness during election years. Thetwo columns of this framework define the partisanorientation of an incumbent party seeking to retainoffice in a general election. The three rows of theframework define different MNC expectations (ex-pressed as “�”) regarding the likelihood that a right-wing party candidate will prevail. Values rangefrom zero to one (0 � � � 1); “� � 1” indicates MNCexpectations of a right-wing victory, “� � 0” indi-cates MNC expectations of a right-wing defeat; and“� � 0.5” indicates balanced MNC expectations (anelection that is a close call). The six political busi-ness cycle scenarios in this two by three matrix(I–VI) summarize predicted effects that incumbent

cycle model, Nordhaus documented economic expan-sion cum contraction associated with U.S. presidentialelections in the 20th century, and Block, Singh, andFerree (2003) documented increased government expen-diture, output, and money supply foretelling acceleratinginflation across sub-Saharan African countries with com-petitive electoral systems since the 1980s. Block andVaaler (2004) reviewed other empirical research docu-menting pre- and postelection trends consistent with op-portunistic and partisan political business cycle modelsin other country contexts.

2008 25Vaaler

partisan orientation and incumbent reelection like-lihood have on MNC willingness to sponsor projectinvestments, as indicated by increasing (�) or de-creasing (–) willingness. I depict these two effectsin pairs in which the first sign summarizes partisanpolitical business cycle effects and the second signsummarizes opportunistic political business cycleeffects on MNC willingness to invest. For example,a “0, 0” pair indicates an election-year scenariowith no political business cycle effects, but a “�, –”indicates an election-year scenario in which parti-san effects increase but opportunistic effects de-crease MNC willingness to invest.

For right-wing incumbents, a shift from likelyreelection (I, � � 1) to a close call (III, � � 0.5) andthen to a partisan switch through a left-wing vic-tory (V, � � 0) decreases MNC willingness to investduring election years. Right-wing incumbents areincreasingly likely to be replaced by less investor-friendly left-wing challengers, and those embattledright-wing incumbents are more likely to engage inopportunistic spending sprees to avoid losing. Bothtypes of political business cycle considerations de-crease MNC willingness to invest, moderately (–, –)in close call scenarios and strongly in left-wingvictory scenarios (––, ––):

Hypothesis 1. Given a right-wing incumbent,MNC investment decreases as the likelihood of

reelection decreases (as the scenario shiftsfrom likely reelection [I] to a close call [III] to aswitch [V]).

For left-wing incumbents, partisan and opportu-nistic political business cycle considerations op-pose rather than reinforce each other as an electoralscenario shifts from likely left-wing reelection (VI,� � 0) to a close call (IV, � � 0.5) and then to likelypartisan switch (II, � � 1). Increasing prospects ofinvestor-friendly right-wing victory increase MNCwillingness to invest, but they also increase incen-tives to stave off right-wing challenges with oppor-tunistic spending sprees, which decrease MNCwillingness to invest. These opposing consider-ations are moderate (�, –) in close call scenarios(IV) and stronger (��, ––) in right-wing victoryscenarios (II). Therefore, no a priori basis exists fordetermining whether partisan or opportunistic po-litical business cycle effects will dominate. Accord-ingly, I formulate alternative hypotheses. If parti-san political business cycle effects dominate, then Iexpect election-year MNC investment to increaserelative to the base-case left-wing reelection sce-nario (VI):

Hypothesis 2a. Given a left-wing incumbent,MNC investment increases as the likelihood ofreelection decreases (as the scenario shifts

FIGURE 1Integrated Political Theoretical Model of MNC Investment during Election Yearsa

aThe predicted direction of MNC investment is based on either partisan or opportunistic considerations.

26 FebruaryAcademy of Management Journal

from likely reelection [VI] to a close call [IV] toa switch [II]).

Given previous research indicating the domi-nance of partisan political business cycle effects onrisk and investment behavior among developingcountry sovereign bondholders (Vaaler et al., 2005)and major credit-rating agencies (Vaaler et al.,2006), I pay particular attention to Hypothesis 2a.Yet I do not dismiss the other theory-driven predic-tion, that opportunistic political business cycle ef-fects will dominate, in which case, election-periodMNC investment is expected to decrease (not in-crease) relative to the base-case left-wing reelectionscenario:

Hypothesis 2b. Given a left-wing incumbent,MNC investment will decrease as the likeli-hood of reelection decreases (as the scenarioshifts from likely reelection [VI] to a close call[IV] to a switch [II]).

This section closes by noting a third assumptionin my framework: It is foreign-domiciled MNCsrather than domestic firms that vary their risk andinvestment behavior during election periods con-sistently with reinforcing or counteracting politicalbusiness cycle considerations. This assumption fol-lows from a rich line of research pursued for over40 years documenting MNC vulnerability to obso-lescing bargains (Vernon, 1971) with host countrygovernments and related local individuals. Thisvulnerability amounts to a “liability of foreignness”(Zaheer, 1995) for MNCs to manage in developingcountry political environments. My frameworksuggests that electoral dynamics and the politicalbusiness cycle incentives they generate can in-crease or decrease such liability substantially andvary MNC project investment activity during elec-tion periods.

METHODOLOGY

MNC Project Investment Empirical Model andImplied Hypothesis Tests

To test the two hypotheses, I define the followingempirical model for estimation:

Project countit � �0intercept � �1–17countryi

� �1–13yeart � �1–15macroeconomic factorsit

� �1election yearit � �2right-wing incumbentit

� �3right-wing incumbent election yearit

� �4expectations election yearit

� �5expectations right-wing incumbent

election yearit � �6election yearit � 1

� �7right-wing incumbent election yearit � 1

� �8election yearit – 1

� �9right-wing incumbent election yearit – 1

� �10project countit � 1 � errorit. (1)

The dependent variable, project count, is definedas the count of project investments announced byforeign-domiciled MNCs for developing country iin year t. To explain project count, I first includecontrols for unobserved effects related to individ-ual countries (country) and years (year). The firstcountry in the sample, Argentina, is omitted, and17 0–1 dummies for the other countries in thesample are included. I omit the last year of obser-vation, 2000, and include 13 0–1 year dummies forthe other years. Next, I include 15 macroeconomicand related country control variables (macroeco-nomic factors) that previous researchers have usedto explain the broader attractiveness of countriesfor lending, investment, and economic develop-ment (Cantor & Packer, 1996; Humphreys & Bates,2005; La Porta et al., 1998; Vaaler & McNamara,2004; Vaaler et al., 2006). The data on these vari-ables are updated on an approximately annual ba-sis; thus, final data on each of the 15 terms may notbe available in each year t in which MNCs makeinvestment project decisions. To reflect that possi-bility, I measure these controls as rolling two-yearaverages using observations from years t and t – 1.The 15 macroeconomic factors include thefollowing:

External balance is measured as the average cur-rent account balance (exports less imports) dividedby gross domestic product (GDP) and expected tobe positively related to project count.

External debt is measured as the sum of public,publicly guaranteed, and private nonguaranteedlong-term debt, use of International Monetary Fund(IMF) credit, and short-term debt divided by GDP,stated as a percentage (multiplied by 100) and ex-pected to be negatively related to project count.

Per capita income is measured as average GDP inconstant (1995) thousands of U.S. dollars dividedby the average midyear country population andexpected to be positively related to project count.

Economic size is measured as the natural loga-rithm of average GDP and expected to be positivelyrelated to project count.

Economic growth is measured as the average an-nual real GDP percentage growth rate and expectedto be positively related to project count.

Inflation is measured as the average annual per-

2008 27Vaaler

centage of consumer price inflation, divided by100, and expected to be negatively related toproject count.

Fiscal balance is measured as the average annualoverall budget balance (receipts less expenditures)divided by GDP and expected to be positively re-lated to project count.

Fuel exports is measured as the value of all en-ergy exports (e.g., coal, oil, natural gas) in currentU.S. dollars divided by GDP and expected to benegatively related to project count.2

Government size is measured as government fi-nal consumption expenditure, including all gov-ernment current expenditures for purchases ofgoods and services except the military, divided byGDP and expected to be negatively related toproject count.

Openness is measured as the sum of exports andimports of goods and services divided by GDP andexpected to be positively related to project count.

Currency crisis is a dummy (1 if in crisis, 0 oth-erwise) indicating whether a country’s local cur-rency has depreciated at least 20 percent againstthe U.S. dollar in the current year and expected tobe negatively related to project count.3

Recent default is a dummy variable (1 if in de-fault, 0 otherwise) indicating whether a nationalgovernment has defaulted on its foreign-currency-denominated debt (excluding bank debt) in the lastfive years, and expected to be negatively related toproject count.

Investment grade rating is a dummy (1 if invest-ment grade, 0 otherwise) indicating whether a na-tional government has an investment-grade credit

rating according to the Standard & Poor’s credit-rating agency (where an investment-grade rating is“BBB–”or higher according to the following ordinalranking: AAA, AA�, AA, AA–, A�, A, A–, BBB�,BBB, BBB–, BB�, BB, BB–, B�, B, B–, and C � 0)and expected to be positively related to projectcount.

Lack of political and civil rights is measured asthe average of political rights (1–7 integral measure;1 � “strong political rights,” and 7 � “weak polit-ical rights”) and civil rights (1–7 integral measure;1 � “strong civil rights,” and 7 � “weak civilrights”) and expected to be negatively related toproject count.

Political checks is measured as the extent ofchecks on political authority, derived from an as-sessment of the number of relevant veto holders ina national polity (1–18 integral measure; 1 � “no/minimal checks on political authority” and 18 �“substantial checks on political authority”) and ex-pected to be positively related to project count.

These 15 controls generally follow intuition.Countries will attract more MNC investmentprojects with net exports, lower external debt,higher per capita incomes, greater economic size,faster economic growth, lower inflation, govern-ment budget surpluses, lower government profilein the overall economy, more trade, no recent his-tory of large domestic currency depreciation, norecent history of defaulting on foreign financialobligations, an investment grade rating by a majorcredit-rating agency, and stronger political andcivil rights.

In keeping with my integrated political businesscycle theoretical framework, the variables of cen-tral interest relate to the occurrence of elections,the partisan orientation of incumbents during elec-tions, and MNC electoral expectations. Nine suchvariables are defined. I first include election year(�1), a dummy coded 1 if the year of an election, 0otherwise, to probe for current election year t ef-fects on project count. I also include one-year lead-ing and lagged election-year dummies, electionyearit � 1 and election yearit – 1 (�6 and �8). Theypermit investigation of the duration of election-year effects on project count. Next, I include right-wing incumbent (�2), 1 if incumbent is right-wing,0 if left-wing, to control for the partisan orientationof elected incumbents. When right-wing incumbentis interacted with election year as right-wing in-cumbent � election yearit (�3) and when right-wingincumbent is interacted with leading and laggedelection year as right-wing incumbent � electionyearit � 1 and right-wing incumbent � electionyearit – 1 (�7 and �9), I can partition current, leading,

2 At first glance, fuel exports has an ambiguous impacton country attractiveness for MNC project investment.On the one hand, oil, gas, and related energy exports mayhave a positive effect on investment attractiveness. Theseexports attract energy project investors as well as gener-ate foreign currency reserves thus stabilizing developingcountry finances. On the other hand, country reliance onenergy industry development and export can starve otherindustry sectors of capital for new projects. Because itgenerates foreign currency, it may also lessen the needfor foreign projects and capital (Humphrey & Bates,2005). Given my focus on foreign projects across a rangeof industry settings, I resolved this ambiguity by predict-ing a net negative impact on project count for fuelexports.

3 I continue to categorize a country as experiencing acurrency crisis beyond the first year if the rate of depre-ciation in subsequent years grows by at least 10 percent.Thus, in the first year, if there is a 20 percent deprecia-tion, then the country is in currency crisis. In the nextyear, depreciation must increase to at least 22 percent tocontinue in currency crisis (Vaaler & McNamara, 2004).

28 FebruaryAcademy of Management Journal

and lagged election year effects on project count bythe partisan orientation of the incumbent.

Two additional variables, expectations � elec-tion yearit and expectations � right-wing incum-bent � election yearit (�4 and �5), deal specificallywith MNC electoral expectations. Expectationstakes on three possible values related to three ex-pected electoral outcomes MNCs might consider. Avalue of 1 (expectations � 1) indicates MNC expec-tations that right-wing parties and policies will pre-vail. A value of –1 (expectations � –1) indicatesMNC expectations that left-wing parties and poli-cies will prevail. A value of 0 (expectations � 0)indicates there is no clear MNC expectation of ei-ther right- or left-wing parties and policies comingto power—a close call. I interact expectations withelection year and with the election year and right-wing incumbent terms to examine MNC expecta-tions under different conditions of incumbent par-tisanship (right-wing and left-wing).

A final term in the empirical model is a one-yearlagged dependent variable, project counti – 1. Thisterm acts as a catch-all control for other unspecifiedpast factors influencing current-year project count.Inclusion of this lagged dependent variable term inthe presence of country fixed effects leads to esti-mation challenges discussed below.

Hypothesis 1 predicts decreasing annual MNCinvestment project announcement counts as right-wing base-case scenarios of likely reelection (�1 ��3 � �4 � �5)4 shift to close call scenarios (�1 � �3),and then to switch scenarios (�1 � �3 – �4 – �5). Atest of differences in this hierarchy reduces to Hy-pothesis 1: �4 � �5 � 0.5 Hypothesis 2a predictsthat partisan political business cycle consider-ations will dominate MNC risk and investment be-havior during election years. Accordingly, Hypoth-esis 2a predicts increasing annual MNC investmentproject announcement counts as left-wing base-case scenarios of likely reelection (�1 – �4) giveway to a close call scenarios (�1) and then to switch

scenarios (�1 � �4). Hypothesis 2b predicts thatopportunistic political business cycle effects willdominate. Accordingly, Hypothesis 2b predicts de-creasing annual MNC investment project an-nouncement counts for left-wing incumbent elec-tions as one moves from base case to close call toswitch. A test of differences in these alternativehierarchies reduces to Hypothesis 2a: �4 � 0 orHypothesis 2b: �4 � 0.6

Ideally, MNC expectations would be measuredwith data from pre-election polls of MNC managersconsidering investment projects. Alternatively,pre-election polls of likely voters would be used.Both measures are problematic. First, pre-electionpolling data for likely voters are not widely avail-able in all developing countries. Polling data forMNC managers are nonexistent. Aside from Shultz(1995), a study that used regular, comparable, andreliable U.K. pre-election polling data, only a hand-ful of published studies have examined the moder-ating effect of electoral expectations and exploitedsuch empirical luxury. These studies are exclusiveto industrialized country contexts such as theUnited States (Alesina et al., 1997) and the UnitedStates and United Kingdom (Leblang & Mukherjee,2005). Second, even if pre-election polling datawere available, MNC investment project announce-ments occur throughout election years. Researcherswould be challenged to decide appropriate weightsfor various polling results, including the most im-portant polling results on election day.

I deal with these measurement issues by assum-ing that MNC expectations during election year tare not systematically different from results ob-served on election days. This approach followsVaaler, Schrage, and Block (2005, 2006), who usedelection day results as retrospective proxies for in-vestor expectations shaping 60- and 90-day pre-election trends in credit premia on national govern-ment bonds and as proxies for major credit-ratingagency expectations affecting the likelihood ofchanges in country creditworthiness throughoutelection years. I then review available pre-electionpolling and related prognostications 30–90 daysprior to each election in this sample to confirm thatthere are no “surprise” differences between pre-election trends and election day results. Using thisapproach, I measure MNC expectations by noting

4 In this scenario, election year is 1, right-wing incum-bent is 1, and expectations is 1. Thus, in my fully parti-tioned empirical model, the set of coefficients corre-sponding to this scenario becomes election year (�1) �right-wing incumbent � election year (�3) � electionyear � expectations (�4) � right-wing incumbent � elec-tion year � expectations (�5). Appropriate variable mea-sures, equation terms, and corresponding coefficients arederived for the other five political business cycle elec-toral scenarios similarly.

5 Hypothesis 1 is derived from reduction of the follow-ing inequality: �1 � �3 � �4 � �5 � �1 � �3 � �1 ��3 – �4 – �5. This inequality reduces to: �4 � �5 � 0 �–�4 – �5 � �4 � �5 � 0.

6 Hypothesis 2a is derived from reduction of the fol-lowing inequality: �1 – �4 � �1 � �1 � �4. This inequal-ity reduces to –�4 � 0 � �4 � �4 � 0. Hypothesis 2b isderived from reduction of the following inequality: �1

–�4 � �1 � �1 � �4. This inequality reduces to: –�4 � 0 ��4 � �4 � 0.

2008 29Vaaler

election day victors, their partisan orientations,and their victory margins, which are defined asdifferences in percentage points between winningand second-place (runner-up) candidates in a finalround of voting, typically the general election dayvote. Right-wing victories by substantial marginsare coded as 1. Left-wing victories by substantialmargins are coded as –1. Victory margins of lessthan 3 percent are close call elections no matter thewinner and coded as 0.7

Data Sources and Sampling

I collected several types of data to estimate myempirical model. Data from the World Bank’s Da-tabase of Political Institutions (DPI), Version 4(Beck, Clarke, Groff, Keefer, & Walsh, 2001) and theInternational Foundation for Election Systems(2006) were my primary sources of information onpresidential elections held in developing countrieswith competitive electoral systems from 1987through 2000. I sampled only from presidentialelectoral systems with fixed election dates to avoidissues of endogeneity in election timing that arepossible with parliamentary systems. Countrieshad to have competitive presidential systems to besampled, meaning that they had to score a 6 or a 7on a DPI scale of 1–7 for competitiveness. The DPIsets criteria for incumbent and challenger partisanorientation with left-wing, centrist, right-wing, andother classifications based primarily on contentanalysis of party titles and secondarily on contentanalysis of party platforms and historical commit-ments to investor (right-wing and centrist) versusworker (left-wing) interests. Following these crite-ria, I aggregated electoral incumbents and challeng-ers from right-wing and centrist party orientationsinto a single right-wing bloc.

For the 15 macroeconomic factors in my empiri-cal model, I collected annual data from the follow-ing sources: the World Bank’s World DevelopmentIndicators (World Bank, 2006) provided data onexternal balance, external debt, per capita income,economic size, economic growth, inflation, fiscalbalance, fuel exports, government size, opennessand currency crisis; Standard & Poor’s (S&P’s;1999, 2000) yielded data on recent default;

Bloomberg International’s (2006) online servicesyielded data on investment grade rating; FreedomHouse (2006) online sources provided informationon lack of political and civil rights; and the DPIgave information on political checks.

Data on MNC project announcements were col-lected from the Thomson Securities Data Corpora-tion (2006) online project investment database,which Thomson Financial compiles from regula-tory filings and media reports covering dates ofproject announcement, financing and construction,estimated dollar costs, and lead and junior sponsorsyndicate information. I identified 18 countrieswith competitive presidential electoral systems,fixed election dates, parties with discernible parti-san distinctions, reliable final voting results, andconditions sufficient for investment projects spon-sored by foreign-domiciled MNCs in the period1987–2000: Argentina, Bolivia, Brazil, Bulgaria,Chile, Colombia, Ecuador, Indonesia, South Korea,Mexico, Paraguay, Peru, Philippines, Poland, Rus-sia, South Africa, Uruguay, and Venezuela. Projectinvestments sponsored by MNCs typically requiremajor credit-rating agency ratings of project com-panies and of the sovereign governments in thecountries where the projects are to be located.Country sampling thus begins in the first year thatcountries had sovereign ratings published by one ofthe six major credit-rating agencies active in thesovereign credit-rating business from 1987 through2000: Moody’s, Standard & Poor’s, Fitch, DuffCredit Rating, Thomson Bank Watch, and Invest-ment Bank Credit Analysis.

My sampling approach results in 154 annual ob-servations of MNC investment project announce-ment counts in 18 countries, an average of approx-imately 9 annual observations per country. Thesample includes 408 project announcements worthapproximately $199 billion sponsored by MNCsdomiciled outside project host countries. On thebasis of the partisan orientations of incumbent par-ties and electoral expectations and outcomes, Icategorize the countries (election years) into theintegrated political business cycle theoreticalframework.8 The sample distributes itself into all

7 I use percentage of votes cast or valid votes depend-ing on availability. Legislative electors chose presidentsin South Africa (1994, 1999), Bolivia (1997), and Indone-sia (1999) after general elections, so I use that legislativevote information. When close call election victory mar-gins are redefined as less than 5 percent, I obtain consis-tent results (These results are available on request).

8 The sampled countries (election years) are placed inthe following election scenarios: Argentina (89, 95, 99),Bolivia (97), Brazil (89, 94, 98), Chile (93), Colombia (98),Paraguay (98), Peru (95, 00), Russia (96, 00), Uruguay (99)in the right-wing base-case scenario (I); Chile (99), Co-lombia (94), Indonesia (99), Korea (92, 97), Uruguay (94)in the right-wing close call scenario (III); Ecuador (96),Philippines (98), South Africa (94), Venezuela (98) in theright-wing switch scenario (V); Mexico (94), Poland (95,00), South Africa (99), Venezuela (88, 00) in the left-wing

30 FebruaryAcademy of Management Journal

six scenarios of the framework. Not surprisingly,most countries and elections fall into the two in-cumbent base-case scenarios (I, VI) with the right-wing base case (likely reelection; I) including 16 ofthe 35 elections sampled. Yet there is representa-tion in the five other cells, permitting testing fordifferences in project count hierarchies predictedby Hypotheses 1, 2a, and 2b.

Estimation Strategy

To analyze this study’s cross-sectional (country i)time series (year t) panel data, I use three differentestimators, all of which are available in Stata Ver-sion 9.2 (StataCorp, 2005). Analysis begins withpanel generalized least squares (GLS) regressionestimation of project count with only the 15 mac-roeconomic control variables and the country andyear dummies. Since the dependent variable is anannual count and the data are overdispersed, I alsouse negative binomial regression (NBR) estimationto regress project count on the same set of basiccontrols and then on an increasing number of po-litical business cycle–related terms.

Hypothesis tests based on interaction terms aloneor in linear combination can yield inconsistent es-timates when derived from nonlinear models likeNBR (Ai & Norton, 2003; Hoetker, 2007). Thus, Icomplement the NBR-based hypothesis tests withinvestigation of predicted trends based on simula-tion methods developed by King, Tomz, and Wit-tenberg (2000). Their Clarify Version 2.1 software,an add-on to Stata, permits Monte Carlo simulationof project count results under different electionscenarios. I run 1,000 NBR-based simulations foreach given equation specification and then set vari-ables at their mean values, except for political busi-ness cycle–related terms, which I set at values cor-responding to different election scenarios. I thenassess the probability of a project announcementunder these scenarios and compare it with NBR-based hypothesis test results.

In addition to GLS and NBR estimators, I use athird generalized method of moments (GMM) dy-namic panel regression estimator developed byArellano and Bond (1991) to evaluate the robust-ness of key results to reasonable changes in equa-tion specification and sampling. Although notbased on negative binomial distributional assump-tions and thus not ideal for count data, this GMMestimator is ideal for inclusion of a lagged depen-

dent variable, project countit – 1, to address year-to-year serial correlation in the panel error structureand to act as a catch-all control variable for currentyear project count. The Arellano-Bond GMM esti-mator yields consistent estimates in the presence offixed country and year effects and permits hypoth-esis tests using interaction terms alone or in linearcombination. Lagged dependent variable and otherGMM estimation requirements reduce the samplesize from 154 to 118. GLS, NBR, and GMM estima-tors include robust (to panel heteroskedasticity)Huber-White sandwich standard errors (Huber,1967; White, 1980).

RESULTS

Descriptive Statistics and Pairwise Correlations

The mean value of the dependent variable,project count, is 2.65, with a standard deviation of3.40, a minimum value of 0, and maximum value of19. On average, MNCs announce two to three in-vestment projects in a country annually. The meanvalue of project count in an election year is 2.60,with a standard deviation of 4.16. At first glance,elections and the political business cycle incen-tives they may engender appear to have little im-pact on the count of investment projects an-nounced by MNCs. The average U.S. dollar value ofan announced investment project is approximately$500 million, which translates to annual dollarvalue of $1.3 billion in new project announcementsper sampled country. If this trend does not varysignificantly and substantially in election years,then political business cycle assumptions andframeworks may provide little explanation of MNCrisk and investment behavior.

Table 1 provides descriptive information andpairwise correlations for terms in my empiricalequation. Approximately 76 percent of the country-year observations involve countries with right-wing incumbents, a statistic consistent with thegeneral dominance of right-wing parties in devel-oping country governments during the 1990s.Other macroeconomic and related institutionalterms present a profile of developing countries inthe 1980s and 1990s consistent with most expecta-tions. They have midrange per capita incomes($3,895) and higher (as compared to industrializedcountries) inflation rates (133%) and external debtand fiscal deficits (respectively, 41.63 and 1.80percent of GDP). About one-fifth of the observa-tions come from countries that had recently beenin default of their financial obligations to inves-tors holding foreign-currency-denominated govern-ment bonds (21%) or experiencing currency crises

base-case scenario (IV); Ecuador (98) in the left-wingclose call scenario (IV); and Bulgaria (96), Mexico (00),and Venezuela (93) in the left-wing switch scenario (II).

2008 31Vaaler

TA

BL

E1

Des

crip

tive

Sta

tist

ics

and

Pai

rwis

eC

orre

lati

onsa

Var

iabl

esM

ean

s.d

.1

23

45

67

89

1011

1213

1415

1617

1819

2021

2223

2425

1.P

roje

ctco

un

t2.

653.

402.

Ext

ern

alba

lan

ce�

0.49

5.75

�.0

33.

Ext

ern

ald

ebt

41.6

319

.30

.03

.03

4.P

erca

pit

ain

com

e3.

892.

56�

.08

.19

�.6

15.

Eco

nom

icsi

ze25

.38

1.29

.29

.37

�.5

2.4

76.

Eco

nom

icgr

owth

1.64

3.67

.12

�.1

1�

.27

.24

.15

7.In

flat

ion

1.33

4.16

�.1

8.1

8�

.18

.07

.24

�.2

48.

Fis

cal

bala

nce

�1.

803.

25.0

4�

.08

.25

�.0

0�

.29

.33

�.3

09.

Fu

elex

por

ts0.

040.

06�

.07

.53

.34

�.2

4�

.08

�.2

1�

.13

.08

10.

Gov

ern

men

tsi

ze12

.50

4.12

.05

�.0

4�

.20

�.0

3.1

8�

.08

�.0

1�

.46

.18

11.

Op

enn

ess

49.9

122

.44

�.0

2�

.07

.43

�.2

9�

.37

�.0

4�

.40

�.3

4.1

3�

.11

12.

Cu

rren

cycr

isis

0.18

0.39

�.1

2.1

6�

.03

�.0

4.1

6�

.36

.33

�.3

9.1

3�

.02

�.0

013

.R

ecen

td

efau

lt0.

210.

41�

.18

.37

.04

�.0

5.1

6�

.23

.49

�.2

2.2

3�

.12

�.2

8.2

914

.In

vest

men

tgr

ade

rati

ng

0.25

0.43

.14

�.0

3�

.31

.36

.12

.46

�.1

7.2

6�

.22

�.0

1.0

7.2

0.3

015

.L

ack

ofp

olit

ical

-civ

ilri

ghts

2.91

0.90

.28

�.1

6.2

0�

.48

.04

�.1

0�

.10

.08

.08

�.2

5.0

6�

.01

�.0

1�

.13

16.

Pol

itic

alch

ecks

3.50

1.33

�.1

0.1

5�

.18

.30

.09

�.0

1.1

8�

.20

.11

.12

�.1

9.0

5.2

4�

.05

�.4

517

.E

lect

ion

year

(�1)

0.23

0.42

�.0

1.0

0.0

1.0

0�

.01

�.0

7.0

3�

.12

.03

.08

�.0

5.0

3.0

2�

.07

�.0

2.0

118

.R

igh

t-w

ing

incu

mbe

nt

( �2)

0.76

0.43

.00

�.0

9�

.06

.14

�.0

7.0

2.1

5.0

0�

.26

�.1

2�

.14

�.0

1.2

2.1

5.1

5.0

4�

.06

19.

Rig

ht-

win

gin

cum

ben

t�

elec

tion

year

( �3)

0.16

0.36

�.0

4.0

1�

.06

.05

.01

�.0

5.0

8�

.09

�.0

7.0

5�

.11

.02

.07

�.0

1.0

4�

.01

.81

.25

20.

Ele

ctio

nye

ar�

�D

(�4)

0.05

0.42

.10

�.0

2�

.01

.04

.01

�.0

2.1

5�

.13

�.1

1�

.06

�.1

1.1

4.0

5�

.04

.06

.14

.23

.18

.40

21.

Rig

ht-

win

gin

cum

ben

t�

elec

tion

year

��D

(�5)

0.07

0.34

.04

�.0

2�

.07

.05

.05

�.0

1.1

6�

.11

�.1

2.0

2�

.21

.05

.08

�.0

3.0

4.1

1.3

8.1

2.4

7.8

2

22.(

Ele

ctio

nye

art

�1)

(�6)

0.19

0.39

�.0

6�

.06

�.0

3.0

3�

.00

.11

�.0

5.0

1.0

0�

.01

�.0

1.0

7.0

7�

.01

�.0

8.0

2�

.26

.04

�.2

1�

.06

�.1

023

.R

igh

t-w

ing

incu

mbe

nt

�el

ecti

onye

art

�1

(�7)

0.15

0.36

�.0

8�

.05

�.0

8.0

2�

.06

.05

�.0

2�

.02

�.0

1�

.07

�.0

3.0

4.1

4.0

1.0

1.0

2�

.23

.19

�.1

8�

.05

�.0

9.8

2

24.(

Ele

ctio

nye

art

�1)

(�8)

0.21

0.41

�.1

1�

.00

.01

.01

�.0

5�

.27

.10

�.0

8�

.04

.06

.02

.08

�.0

0�

.05

�.0

7.0

7�

.28

�.0

0�

.23

�.0

6�

.11

�.2

5�

.22

25.

Rig

ht-

win

gin

cum

ben

t�

elec

tion

year

t�

1(�

9)

0.15

0.36

�.0

3�

.01

�.0

5.0

9.0

2�

.17

.14

�.0

7�

.09

.04

�.1

2�

.01

.05

.01

�.0

4.0

9�

.23

.15

�.1

8�

.05

�.0

9�

.20

�.1

8.8

0

26.P

roje

ctco

un

t t�

1(�

10)

2.31

3.10

.59

�.0

4.0

6�

.08

.27

.07

�.2

0.0

4�

.04

.09

.05

�.0

3�

.12

.17

.27

�.1

8�

.07

.05

�.0

6�

.06

�.0

5.0

3�

.03

.06

�.0

0

an

is15

4ex

cep

tfo

rth

ela

gged

dep

end

ent

vari

able

,p

roje

ctco

un

t t�

1,

for

wh

ich

nis

118.

Cor

rela

tion

sgr

eate

rth

anap

pro

xim

atel

y.1

3or

less

than

�.1

3ar

esi

gnif

ican

tat

p�

.10;

Cor

rela

tion

sgr

eate

rth

anap

pro

xim

atel

y.1

6or

less

than

�.1

6ar

esi

gnif

ican

tat

p�

.05;

Cor

rela

tion

sgr

eate

rth

anap

pro

xim

atel

y.2

0or

less

than

�.2

0ar

esi

gnif

ican

tat

p�

.01

(all

one-

tail

edte

sts)

.

(18%). Another quarter (25%) come from countrieswith national governments deemed investmentgrade in creditworthiness by major credit-ratingagencies. Political and civil rights are middling,scoring about 3 on a 1–7 scale. Checks on politicalauthority range from 1 (minimal constraints given acompetitive system for electing the national exec-utive) to 6, with a mean of 3.50.

Regression and Hypothesis Test Results

Estimating project count with country and yearfixed effects as well as the macroeconomic controlsleads to intuitive results, presented in columns 1and 2 of Table 2. GLS estimation in column 1 yieldsmacroeconomic controls with the predicted sign in9 of 15 instances and statistical significance at the10 percent level or better in three instances. Thecountry, year, and macroeconomic controls aloneexplain substantial variation in the dependent vari-able (overall model R2 � 0.60). The NBR estima-tion, presented in column 2, yields expected signsin 11 of 15 instances, with statistical significance at5 percent or a higher level in six instances. Incolumns 2–5 of Table 2, the signs on fiscal balance,external debt, inflation, fuel exports, governmentsize, and recent default are consistent with predic-tions, statistically significant at commonly ac-cepted levels, and practically substantial.9

The GLS and GMM results in presented in col-umns 1 and 6–8 are interpreted in terms of annualproject investment announcement levels, but theNBR results in columns 2–5 should be read interms of annual project investment announcementrates. Thus, in column 1, recent default decreases

the number of project investment announcementsby 3.21, effectively reducing project count from itssample mean of 2.65 to 0. In column 6, recentdefault decreases the number of project investmentannouncements by 1.58. In column 2’s NBR results,recent default has a –0.77 coefficient estimate. Itransform the NBR estimate into a percentage usingthe following formula: 100% � (exp[–0.77] – 1) �–53.6%. After transformation, I infer that recentdefault decreases the rate of new project announce-ments by 53.6 percent, holding other factors at theirmean levels. Similarly, transforming the coefficientestimate of –0.08 for external debt (100% � (exp([–0.08] –1) � –7.5%) implies that a one-unit in-crease in external debt decreases project count byapproximately 7.5 percent, again with other factorsheld at their mean levels. These and other controlsexhibit statistically significant and practically sub-stantial effects on developing country attractive-ness for MNC investment projects worth, on aver-age, more than $1 billion annually.

Column 3 of Table 2 shows the results of reesti-mating project count after addition of the currentelection year dummy. The coefficient sign is nega-tive but statistically insignificant at commonly ac-cepted levels. Additions of the variables for right-wing incumbent and the interaction of right-wingincumbent and election year in column 4 permitpartitioning current year election effects on projectcount into those related to elections with left-wingincumbents (where election year � 1, right-wingincumbent � 0, and right-wing incumbent � elec-tion year � 0) and elections with right-wing incum-bents, where the same three terms take the value of1. After NBR estimation, coefficients for electionyear (�1), for right-wing incumbent � election year(�3), and for their linear combination (�1 � �3) arenot significantly different from zero at commonlyaccepted levels. Complementary investigationbased on Monte Carlo simulation also indicates nosubstantial change in the probability of project an-nouncements during election years. After 1,000simulations, I set all column 4 variables at theirmean values and estimate a 10.8 percent probabil-ity of a project being announced in a given year andcountry sampled. This probability decreases onlyslightly, to 10.6 percent, if I reset election year,right-wing incumbent, and right-wing incumbent �election year to values of 1, representing a countryholding an election with a right-wing incumbent.The probability of project announcement decreasesagain only slightly, to 10.4 percent, if election yearis kept at 1 but right-wing incumbent and the in-teraction term are set equal to 0, to represent acountry holding an election with a left-wing in-cumbent. Were one to stop here, the conclusion

9 On the other hand, I note two anomalous controlswith consistently contrary signs at commonly acceptedlevels of significance: country economic size (natural logof GDP) and lack of political and civil rights. I predict apositive sign for economic size and observe this in pair-wise correlations with project count in Table 1. Yet Iobserve significant negative coefficients after severalmultivariate estimations, reported in Table 2. These es-timates include country dummies, thus rendering themwithin-subjects (within-countries) estimates of the im-pact of economic size on annual project counts. In thiscontext, increasing economic size does not correlate withhigher annual investment project counts. I also predict anegative sign on lack of political and civil rights butobserve significant positive coefficients in both pairwiseand multivariate analyses. This anomaly also surprisesme, as it signals at least an indifference to political open-ness and individual liberties among project-investingMNCs, and further evidence contrary to a view held bysome scholars that political modernization improves thebusiness investment climate (Goldsmith, 1994).

2008 33Vaaler

TA

BL

E2

Res

ult

sof

Reg

ress

ion

An

alys

esfo

rA

nn

ual

MN

CIn

vest

men

tP

roje

cts

An

nou

nce

d,

1987

–200

0a

Var

iabl

esC

olu

mn

1:C

ontr

ols;

GL

SC

olu

mn

2:C

ontr

ols;

NB

R

Col

um

n3:

Con

trol

s,E

lect

ion

;N

BR

Col

um

n4:

Con

trol

s,E

lect

ion

,P

arti

san

;N

BR

Col

um

n5:

Con

trol

s,P

olit

ical

Bu

sin

ess

Cyc

leE

ffec

ts;

NB

R

Col

um

n6:

Con

trol

s,P

olit

ical

Bu

sin

ess

Cyc

leE

ffec

ts;

GM

M

Col

um

n7:

Dom

esti

cP

roje

cts;

GM

M

Col

um

n8:

For

eign

Pro

ject

Syn

dic

ates

;G

MM

Con

stan

t26

9.11

(241

.16)

227.

88**

(75.

59)

225.

46**

(77.

35)

211.

47**

(75.

49)

173.

44**

(72.

92)

0.20

(0.3

0)�

0.01

(0.2

4)0.

04(0

.21)

Ext

ern

alba

lan

ce0.

01(0

.11)

0.01

(0.0

4)0.

01(0

.04)

0.02

(0.0

4)0.

03(0

.04)

0.13

(0.1

2)0.

03(0

.11)

�0.

02(0

.10)

Ext

ern

ald

ebt

�0.

15*

(0.0

7)�

0.08

**(0

.02)

�0.

08**

(0.0

2)�

0.07

**(0

.02)

�0.

06*

(0.0

2)�

0.13

*(0

.07)

0.04

(0.0

4)�

0.11

†(0

.06)

Per

cap

ita

inco

me

0.90

(0.9

6)1.

34**

(0.4

0)1.

33**

(0.4

0)1.

23**

(0.3

7)1.

00**

(0.3

4)0.

73(1

.55)

�0.

31(0

.93)

0.56

(1.0

1)E

con

omic

size

�9.

29(8

.96)

�8.

29**

(2.8

0)�

8.20

**(2

.86)

�7.

70**

(2.7

9)�

6.29

*(2

.68)

�0.

23(1

0.69

)5.

87(7

.93)

1.89

(7.2

5)E

con

omic

grow

th0.

03(0

.10)

�0.

03(0

.03)

�0.

03(0

.02)

�0.

03(0

.03)

�0.

05**

(0.0

2)�

0.05

(0.1

1)0.

00(0

.04)

�0.

00(0

.08)

Infl

atio

n�

0.12

(0.0

9)�

0.06

(0.0

4)�

0.06

(0.0

4)�

0.06

(0.0

4)�

0.09

†(0

.05)

�0.

11†

(0.0

7)�

0.06

(0.0

5)0.

05(0

.04)

Fis

cal

bala

nce

0.25

†(0

.14)

0.16

**(0

.05)

0.16

**(0

.05)

0.15

**(0

.06)

0.15

**(0

.05)

0.31

**(0

.09)

0.09

(0.0

7)0.

02(0

.06)

Fu

elex

por

ts�

41.0

7†(2

3.75

)�

15.8

4**

(5.8

0)�

15.9

4**

(5.8

6)�

17.6

9**

(6.1

1)�

19.1

4**

(5.9

5)�

62.7

2†(3

4.36

)�

35.9

0†(1

9.83

)8.

99(1

5.56

)G

over

nm

ent

size

�0.

42**

(0.1

5)�

0.16

**(0

.04)

�0.

16**

(0.0

4)0.

15**

(0.0

5)�

0.15

**(0

.05)

�0.

30†

(0.1

8)�

0.07

(0.1

1)1.

89(7

.25)

Op

enn

ess

0.04

(0.0

4)0.

01(0

.01)

0.01

(0.0

1)0.

01(0

.01)

0.02

*(0

.01)

0.06

†(0

.04)

0.07

(0.2

1)0.

02(0

.04)

Cu

rren

cycr

isis

�0.

33(0

.83)

�0.

18(0

.22)

�0.

19(0

.23)

�0.

22(0

.23)

�0.

42*

(0.1

9)�

1.06

†(0

.58)

�1.

06**

(0.4

0)�

0.52

(0.2

2)R

ecen

td

efau

lt�

3.21

**(0

.89)

�0.

77**

(0.2

8)�

0.78

**(0

.29)

�0.

77**

(0.3

0)�

0.57

†(0

.32)

�1.

58*

(0.7

5)�

0.97

†(0

.51)

�0.

59(0

.66)

Inve

stgr

ade

rati

ng

0.23

(1.0

8)�

0.14

(0.3

2)�

0.16

(0.3

2)�

0.01

(0.3

5)0.

20(0

.33)

1.55

(1.6

1)0.

61(1

.16)

0.72

(0.8

1)L

ack

ofp

olit

ical

-civ

ilri

ghts

1.23

*(0

.89)

0.49

**(0

.17)

0.50

**(0

.18)

0.44

**(0

.17)

0.47

**(0

.17)

1.32

*(0

.65)

2.10

**(0

.37)

0.96

(0.6

9)P

olit

ical

chec

ks0.

11(0

.29)

0.06

(0.0

8)0.

07(0

.08)

0.05

(0.0

8)�

0.02

(0.0

8)�

0.16

(0.2

4)�

0.20

(0.1

5)0.

38**

(0.1

3)E

lect

ion

year

( �1)

�0.

05(0

.19)

0.22

(0.3

6)0.

14(0

.26)

2.50

(1.9

3)�

0.04

(0.8

9)0.

73(1

.13)

Rig

ht-

win

gin

cum

ben

t( �

2)

0.43

(0.4

0)0.

59(0

.45)

2.56

†(1

.41)

0.93

(0.6

9)1.

49**

(0.5

5)R

igh

t-w

ing

incu

mbe

nt

�el

ecti

onye

ar( �

3)

�0.

38(0

.40)

�0.

74†

(0.4

5)�

3.44

(2.2

8)�

0.16

(0.9

8)�

1.95

(1.2

2)

Ele

ctio

nye

ar�

�D

(�4)

1.05

**(0

.28)

4.44

*(2

.19)

0.23

(0.5

9)2.

45**

(0.8

6)R

igh

t-w

ing

incu

mbe

nt

�el

ecti

onye

ar�

�D

(�5)

�0.

46(0

.40)

�3.

19†

(1.9

8)�

0.13

(0.5

0)�

1.61

*(0

.80)

Ele

ctio

nye

art

�1

(�6)

�0.

36(0

.34)

�0.

89(1

.51)

0.00

(0.4

3)�

1.13

†(0

.62)

Rig

ht-

win

gin

cum

ben

t�

elec

tion

year

t�

1(�

7)

�0.

08(0

.40)

0.54

(1.6

6)0.

61(0

.52)

0.88

(0.6

7)

Ele

ctio

nye

art

�1

(�8)

�1.

00**

(0.3

7)�

1.96

**(0

.64)

0.77

†(0

.43)

�0.

66(0

.49)

Rig

ht-

win

gin

cum

ben

t�

elec

tion

year

t�

1(�

9)

0.99

**(0

.40)

2.19

**(0

.88)

�1.

18*

(0.5

7)0.

24(0

.42)

Pro

ject

cou

nt t

�1

(�10)

0.02

(0.0

6)�

0.01

(0.0

7)�

0.51

**(0

.09)

Cou

ntr

yye

ard

um

mie

sY

esY

esY

esY

esY

esY

esY

esY

esY

ear

du

mm

ies

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

n15

415

415

415

415

411

811

811

8W

ald

2

(R2)

(0.6

0)6,

494*

*7,

254*

*6,

338*

*6,

710*

*11

5**

570*

*94

0**

aC

olu

mn

s1–

10re

por

treg

ress

ion

coef

fici

ents

and

Hu

ber-

Wh

ite

robu

stst

and

ard

erro

rs(i

np

aren

thes

es).

“GL

S”

refe

rsto

gen

eral

ized

leas

tsqu

ares

mod

els;

“NB

R,”

ton

egat

ive

bin

omia

lre

gres

sion

mod

els;

and

“GM

M,”

toge

ner

aliz

edm

eth

odof

mom

ents

mod

els.

Reg

ress

ion

resu

lts

for

cou

ntr

yan

dye

ard

um

mie

sar

eav

aila

ble

onre

ques

t.†

p�

.10

*p

�.0

5**

p�

.01

might be that short-term electoral politics and eco-nomic policies, no matter their impact on variousforeign financial actors like bond investors andcredit-rating agencies, have little relevance forMNC managers with a long-term or strategic per-spective on risk and investment.

But the political business cycle theoretical frame-work of interest here suggests finer-grained parti-tioning of election-year effects on MNC risk andinvestment behavior, and my empirical equationpermits such analysis. Columns 5–8 of Table 2provide additional terms sufficient for such parti-tioning as well as construction and evaluation ofdifferences in project count predicted by Hypothe-ses 1 and 2. The base case scenario of likely right-wing incumbent reelection (I) implies that electionyear (�1), right-wing incumbent (�2), right-wing in-cumbent � election year (�3), expectations � elec-tion year (�4), and expectations � right-wing in-cumbent � election year (�5) all equal 1. The right-wing base-case trend is given by the linearcombination of �1 plus �3 plus �4 plus �5. Based onthe NBR results in column 5, this linear combina-tion is –0.01, but it is not significant at commonlyaccepted levels. I then run 1,000 simulations andset all column 5 terms at their mean value exceptthe five terms above, all of which take the value of1 consistent with a likely right-wing reelection sce-nario. I observe a 10.7 percent probability of aproject announcement under this base case.

When expectations shift from likely reelection toa close call (III), the project count trend is given bythe linear combination of �1 plus �3. This linearcombination decreases to –0.60, which is signifi-cant at the 10 percent level. The political businesscycle variables also change for NBR-based simula-tion of project count in a close call scenario (elec-tion year � 1, right-wing incumbent � 1, right-wingincumbent � election year � 1, expectations �election year � 0, and expectations � right-wingincumbent � election year � 0), and the probabilityof a project announcement drops to 7.3 percent.This negative trend continues as one moves from aclose call to a switch (V), and the project counttrend is given by the linear combination �1 plus �3

minus �4 minus �5. This linear combination de-creases to –1.20, which is significant at the 5 per-cent level. The political business cycle variableterms again change for NBR-based simulation ofproject count in a switch scenario (election year �1, right-wing incumbent � 1, right-wing incum-bent � election year � –1, expectations � electionyear � –1, and expectations � right-wing incum-bent � election year � –1), and the probability of aproject announcement drops to 4.6 percent.

In keeping with my integrated political business

cycle theoretical framework, I observe decreasingNBR-based simulated probabilities of project an-nouncement in moving from likely right-wing in-cumbent reelection to close call to switch scenar-ios. Similarly, I observe estimated trends indicatingdecreasing rates of project announcement as reelec-tion prospects dim for more investor friendly right-wing incumbents. This difference in trends is pos-itive and statistically significant at the 10 percentlevel, supporting Hypothesis 1 (H1: �4 � �5 �0.59 � 0, p � .10).

What about MNC project investment trends forcountries with left-wing incumbents facing reelec-tion? The base-case scenario of likely left-wing in-cumbent reelection (VI) implies that election yearequals 1, that right-wing incumbent equals 0, right-wing incumbent � election year and expecta-tions � right-wing incumbent � election yearequals 0, and that expectations � election yearequals –1. The left-wing base-case trend is given bythe linear combination of �1 minus �4. NBR resultsin column 5 indicate that this base-case linear com-bination is –0.91, which is significant at the 1 per-cent level. When I set the five terms at the values asnoted immediately above and leave all other vari-ables set at their mean values, a 3.3 percent proba-bility of a project announcement is observed.

When expectations shift from likely reelection toclose call (IV), the appropriate project count trendis given by the coefficient �1, which equals 0.14 butis not significant at commonly accepted levels. Thepolitical business cycle variable terms change forNBR-based simulation of project count in a closecall scenario (election year � 1, right-wing incum-bent � 0, right-wing incumbent � election year �0, expectations �election year � 0, and expecta-tions � right-wing incumbent � election year � 0),and the probability of a project announcementjumps to 8.4 percent. This increasing trend contin-ues as one moves from a close call to a switchscenario (II). The appropriate trend is now given bythe linear combination �1 plus �4, which is 1.42and significant at the 1 percent level. The politicalbusiness cycle variable terms change again forNBR-based simulation of project count in a switchscenario (election year � 1, right-wing incum-bent � 0, right-wing incumbent � election year �0, expectations � election year � 1, and expecta-tions � right-wing incumbent � election year � 0),and the probability of a project announcementjumps again, to 20.1 percent.

In keeping with partisan political business cycleconsiderations in the current study’s theoreticalframework, I observe increasing NBR-based simu-lated probabilities of project announcement in go-ing from likely left-wing incumbent reelection to

2008 35Vaaler

close call to switch. Similarly, I observe estimatedtrends indicating increasing rates of project an-nouncement as prospects brighten for more inves-tor-friendly right-wing challengers. This differencein trends is positive and statistically significant atthe 1 percent level, supporting Hypothesis 2a (H2a:�4 � 1.05 � 0, p � .01).

The GMM results in column 6 of Table 2 provideadditional support for Hypotheses 1 and 2a.10 Theright-wing base-case linear combination is 0.32, theright-wing close call linear combination is –0.94,and the right-wing switch scenario is –2.19. Al-though none of these estimates are significant atcommonly accepted levels individually, the teststatistic for Hypothesis 1 and the difference overthese estimates is significant at the 5 percent level(H1: �4 � �5 � 1.25 � 0, p � .05). For left-wingincumbent elections, annual project counts drop byabout two (�1 – �4 � –1.94, p � .10) when reelec-tion is likely. A close call increases the number ofprojects to 2.50 (�1), but this estimate is not signif-icant at commonly accepted levels. When left-wingincumbents are likely to be ousted by more investorfriendly right-wing challengers, however, projectannouncements increase by approximately seven(�1 � �4 � 6.94, p � .10). These differences are

significant at the 5 percent level, consistently withHypothesis 2a and the dominance of partisan overopportunistic political business cycle consider-ations (H2a: �4 � 4.44 � 0, p � .01). These differ-ences are also practically substantial. Given thatthe average estimated project cost is approximately$500 million, a swing in left-wing electoral scenar-ios results in project investment increases or de-creases worth as much as $4.5 billion annually.

Figure 2 summarizes these trends graphically, bypresenting two nonparametric, locally weightedscatterplot smoother (Lowess) analyses (StataCorp,2005). Lowess analyses compute linear regressionsaround each observation, xit, with neighborhoodobservations chosen within some sampling band-width and weighted by a tricubic function. Valuesfor yit are based on the estimated regression param-eters. Connecting these xit, yit combinations thenyields a Lowess curve. A higher bandwidth resultsin a smoother Lowess curve.

Figure 2 presents two Lowess analyses using a 90percent sampling bandwidth. The x-axis representsthe winning (�) or losing (–) percentage margin ofvictory on election day. The y-axis represents thechange in the number of investment projects an-nounced in country i in year t compared to theprior year (t � 1). Supporting my integrative po-litical business cycle theoretical framework andHypothesis 1, increasingly positive year-to-yearchanges in the number of project announcementscan be observed as one moves from losing to win-ning margins for more investor friendly right-wing

10 Dynamic panel GMM model assumptions are met,including group exogeneity of all instruments generatedin estimation, and absence of higher (than first-order)serial correlation in the error structure. These results areavailable on request.

FIGURE 2Incumbent Election Results and Change in Announced MNC Project Counts, 1987–2000

36 FebruaryAcademy of Management Journal

incumbents. Consistently with my political busi-ness cycle framework and Hypothesis 2a, I observeincreasingly negative year-to-year changes in thenumber of project announcements in moving fromlosing to winning margins for less investor friendlyleft-wing incumbents. These nonparametric resultsprovide additional support for my overall researchproposition as well as more empirical confirmationthat actual election day results are good proxies forMNC expectations during election years.

Related Results

Results in Columns 5 and 6 of Table 2 also permitexamination of lagged and leading election effectsfor right- and left-wing incumbents. Although notrelated directly to my two hypotheses, they none-theless shed light on the duration of election-yeareffects. NBR results for election yearit – 1 (�8) andright-wing incumbent � election yearit – 1 (�9) incolumn 5 suggest that partisan differences matterfor MNC project risk and investment behavior ayear after elections take place. The –1.00 coeffi-cient on election yearit – 1 (�8) is significant at the 1percent level. After transformation, this result im-plies a 63 percent decrease in the rate of projectannouncement in the year after elections resultingin left-wing presidents. There is no such decreasein the case of lagged effects for right-wing presi-dents, where the linear combination of �8 plus �9

yields an estimate of –0.01 but is not significant atcommonly accepted levels. A similar pictureemerges from NBR-based simulation under a laggedleft-wing president scenario (election yearit – 1 � 1,right-wing incumbent � 0, right-wing incumbent �election yearit – 1 � 0) and a lagged right-wing pres-ident scenario (election yearit – 1 � 1, right-wingincumbent � 1, right-wing incumbent � electionyearit – 1 � 1). The probability of a project an-nouncement the year after elections resulting inright-wing presidents is 10.2 percent but is only 3.1percent after elections resulting in left-wing presi-dents. Column 6’s GMM results are similar. An-nounced investment project counts drop by ap-proximately two (–1.96, p � .01) a year afterelections resulting in left-wing presidents.11 Ratherthan irrelevance, political business cycle consider-ations appear to be significant, practically substan-

tial, and for elections resulting in left-wing incum-bents, persistent over years for MNCs.

Robustness and Heterogeneity

These results prove robust to reasonable varia-tion in equation specification and sampling. I ob-tain consistent results regarding both hypotheses ifI reestimate after: (1) adding additional controls forelections involving a “right-wing” party that theDPI categorizes as centrist; (2) substituting mea-sures of inflation, lack of political and civil rights,and political checks with their natural log values;(3) partitioning external debt into bank and non-bank components; or (4) standardizing external bal-ance, external debt, fiscal balance, fuel exports, andgovernment size with country population ratherthan GDP.12

Reestimation after reasonable alternative sam-pling strategies complements the core results pre-sented in columns 5 and 6 of Table 2. Recall thatone assumption of my theoretical framework is thatMNC willingness to sponsor investment projects isrelated to partisan and opportunistic political busi-ness cycle considerations. Investment projects ledby domestic sponsors may not carry with them aliability of foreignness (Zaheer, 1995) sensitizingMNCs to political business cycle–related risks dur-ing election periods. Column 7 of Table 2 reportsresults from brief empirical investigation of thatframework assumption. I resample from the Thom-

11 Lagged right-wing incumbent effects in column 6are significantly different from left-wing effects (right-wing incumbent � election yearit – 1 (�9) � 2.19, p � .01)but not significantly different from zero at commonlyaccepted levels when assessed as a linear combination.

12 I again obtain results supporting my political busi-ness cycle theoretical framework with an alternative em-pirical equation specification eschewing reliance ondummy-term interaction effects. With this alternativespecification, I drop the election year, right-wing incum-bent, and expectations terms (�1–5) and replace themwith seven 0/1 dummy terms corresponding to seven ofeight political business cycle–related electoral scenarios:(1) right-wing incumbent in a nonelection year, (2) left-wing incumbent in a nonelection year, (3) right-wingincumbent in an election year and likely to be reelected,(4) right-wing incumbent in an election year and facing aclose call, (5) right-wing incumbent in an election yearand likely to be ousted by left-wing challenger, (6) left-wing incumbent in an election year likely to be reelected,(7) left-wing incumbent in an election year and facing aclose call, and (8) left-wing incumbent in an election yearand likely to be ousted by a right-wing challenger. I omitscenario 1 and then compare coefficient estimates for theseven dummies to this omitted category. I find decreas-ing (increasing) coefficients indicating fewer (more)project investment announcements as the right-wing(left-wing) incumbent election scenario shifts from likelyreelection to likely ouster. These results are available onrequest.

2008 37Vaaler

son SDC project finance database annual countswith lead sponsoring organizations from the samecountry as the announced project. One hundredfifty-four country-year project count observationsare now based on 230 domestically sponsoredproject announcements for the 18 developing coun-tries from 1987 through 2000. GMM results revealmany controls and political business cycle–relatedterms neither signed nor statistically significant asbefore. Trends corresponding to the six differentelection scenarios no longer support Hypotheses 1and 2a. A year after elections resulting in left-wingpresidents, domestic project announcements in-crease by about one (election yearit – 1 � 0.77, p �.10) rather than decrease as with MNC-sponsoredproject announcements. Together, these resultsvindicate a focus on MNCs in my political businesscycle theoretical framework and suggest an impor-tant source of organizational heterogeneity in re-sponse to electoral dynamics and political businesscycle considerations.

Column 8 of Table 2 reports another set of com-plementary results. This time, I explore the possi-bility that results might differ if I redefine “foreign-ness” on the basis of the weighted nationality of allsponsoring firms in a project syndicate rather thanthe nationality of the lead sponsoring MNC alone.13

The Thomson SDC project finance database doesnot have complete information on many of the pre-viously sampled projects. Even so, I am able toassemble 154 country-year project count observa-tions based on 230 syndicate-based foreign projectannouncements for the 18 developing countriesfrom 1987 through 2000. The distribution of these230 project announcements over countries andyears is similar to the distribution of the 408 projectannouncements comprising my core sample. Pair-wise correlation between the two counts is 0.84(p � .01). GMM results in column 8 are consistentwith the core results. Test statistics for hierarchicaldifferences in the right-wing incumbent (H1: �4 ��5 � 0.84 � 0, p � .01) and left-wing incumbent(H2a: �4 � 2.45 � 0, p � .01) election scenarios areagain positive, significant at commonly acceptedlevels, and consistent with Hypotheses 1 and 2a.Whether defined by the nationality of a lead-spon-soring MNC or the weighted nationality of lead andjunior sponsors in a syndicate, foreign investmentproject counts vary based on political business cy-cle considerations.

DISCUSSION AND CONCLUSION

Key Findings

I set out to understand whether and how localelectoral factors might shape MNC risk and invest-ment behavior in developing countries, positingthat local factors linked to opportunistic and parti-san political business cycles and the (dis)incen-tives they create for a foreign-based constituencythat votes economically might moderate globalMNC investment trends. I found substantial sup-port for this proposition. MNCs sponsoring invest-ment projects act consistently with political busi-ness cycle considerations, with special emphasison partisan considerations. In developing countryelection scenarios involving right-wing incumbentsunder substantial threat from left-wing challengers,the rate and number of new investment projectannouncements decrease significantly and substan-tially. With elections involving left-wing incum-bents, it is just the opposite. The growing prospectof a partisan shift to the right apparently emboldensMNCs and increases project announcement ratesand numbers to levels potentially worth billions ofdollars in investment and economic development.Partisan political business cycle considerationsseem to be especially important for understandinghow MNCs view local elections and their invest-ment implications. Such considerations persistover time. In the year after elections resulting inleft-wing presidents, I find evidence of continuedincrease in MNC risk and lower investment projectannouncement rates and levels. Short-term electionperiods and the political business cycle incentivesthey create apparently have significant, substantial,and sometimes persistent effects on MNC invest-ments with life spans measured over many years ordecades and (lost) value potentially worth billionsof dollars. MNCs and their strategic managers ap-parently think about risk and investment duringelection periods similarly to bond investors, credit-rating agencies, and perhaps other foreign actorsimportant to supplying capital and capabilities todeveloping countries.

Implications

I draw several research, practice, and public pol-icy implications from these findings. For manage-ment research, these findings underscore the valueof reexamining with novel (to management) politi-cal economy theories and project investment em-pirics a venerable but still critical research issuerelated to the divergent interests of MNCs and de-veloping country host governments. In the process,one begins to understand how dynamics related to

13 Here, I define a project as “foreign” when more than50 percent of the project equity is held by syndicatemembers located outside the project’s country.

38 FebruaryAcademy of Management Journal

democratization and elections shape MNC risk andinvestment behavior.

Almost 30 years ago, Kobrin (1979: 77) chal-lenged management researchers to identify “whichevents matter,” to analyze how “environmentalprocesses affect investor perceptions,” to develop“a conceptual structure relating politics to thefirm.” In response, I identified elections as criticalevents for shaping host government politics andeconomic policies in democratizing countries ofthe developing world. I identified opportunisticand partisan political business cycle consider-ations potentially affecting MNC risk and invest-ment behaviors during these increasingly frequentevents. After developing a theoretical frameworkfor integrating these political business cycle con-siderations, I documented support for my theoreti-cal framework in a broad sample of MNC invest-ment projects, elections, countries, and years.These findings contribute to a broader reexamina-tion of the bargaining hypothesis for MNCs activein developing countries currently experiencing thetwin stresses of political and economic moderniza-tion. As Henisz and Zelner (2003, 2005) alreadynoted, management research in this context willbenefit substantially from crossing disciplinaryboundaries to draw on political economy conceptsand theories, as was done in this political businesscycle–motivated study.

My results have implications for MNC executivesand public policy makers. MNC managers appar-ently take political business cycle–related factorsinto account, particularly partisan factors, whenmulling over investment projects during and some-times after election years. Election-period pre-sumptions about investor friendliness based onright- and left-wing distinctions explained signifi-cant and substantial variance in MNC risk and in-vestment behavior, even after I controlled for thechecks and balances on political authority in agiven country. Henisz (2000) and others (e.g., Becket al., 2001; Humphreys & Bates, 2005) have alreadydemonstrated the importance of political checksand policy uncertainty on foreign investment andeconomic development. Perhaps elections tempo-rarily redirect MNC attention away from con-straints on developing country political authoritiesand toward partisan (and opportunistic) policychanges these same authorities may implementwith the appropriate electoral mandate.

I speculate that MNC partisan presumptions maybe rebutted if developing country governments andpolitical parties communicate credibly about theirelection-period and prospective postelection poli-cies. Managers charged with evaluating countriesfor major investment projects may speculate simi-

larly. Perhaps I have observed here merely pre-sumptive MNC risk perceptions, which may evolvesubstantially through additional study and sus-tained engagement with politicians of differing par-tisan perspectives. If that is true, then more studi-ous MNC managers are also more strategic. Closestudy of local parties and policies beyond simple“left-wing” and “right-wing” labels could lead aselect few managers and MNCs to differ from com-petitors and identify undervalued investment andovervalued divestment opportunities during elec-tion periods.

Limitations and Future Research Directions

This study makes important contributions tomanagement theory and empirical work on politi-cal risk, investment, and democratization in devel-oping countries. It also has limitations. I have al-ready noted thorny issues related to collecting andcategorizing information about partisan orientationand MNC expectations. Simple left-right partisanclassifications follow from partisan political busi-ness cycle theory but are often coarse-grained inempirical applications. Mexico’s Vicente Fox, Rus-sia’s Vladimir Putin, and the Philippines’ FidelRamos fell into the right-wing category in mystudy, but their policy priorities almost certainlycreate substantial distance between them from thestandpoint of MNCs mulling over investmentprojects in their countries. Future research exploit-ing an ever-increasing sample of developing coun-try elections will permit finer-grained partisan cat-egorization schemes to capture not merely left- andright-wing but also more extreme and more moder-ate left-wing, centrist, and right-wing partisan po-sitioning. I also see room for improvement in mod-eling MNC expectations about who will win andlose during election years. Use of actual results isan admittedly second-best option, though empiri-cal confirmation with pre-election polls and mediareports as well as with Lowess analyses suggeststhat this second-best option works well. Going for-ward, I see opportunities for researchers to surveyMNC managers on upcoming elections in the de-veloping world and to develop the regular, compa-rable, and reliable pre-election polling data admit-tedly missing here.

Another limitation relates to my use of invest-ment project counts rather than dollar amounts.Information on the announced dollar value of in-vestment projects is available and provides addi-tional insight on the investment implications of

2008 39Vaaler

“democracy in action” in developing countries.14

On the other hand, cost estimates provided at thetime of a project’s initial announcement oftenprove less reliable than the basic intent to moveforward with the project itself. Even so, with ap-propriate controls, substantial opportunity exists toextend these analyses to understand how the countand dollar amount of MNC project investment varywith political business cycle considerations.

A final limitation concerns model identification.I assumed that elections and the political businesscycle incentives they create shape MNC risk andinvestment behavior, but the opposite relationshipis also possible. MNC investment project an-nouncements during election years may buttressright-wing incumbents’ claims of good economicstewardship and increase their reelection pros-pects. Conspicuous silence by MNCs during left-wing incumbents’ campaigns may underminereelection prospects. My sample of project invest-ments permits brief investigation of this possibility.The Thomson SDC project finance database reportsnot only the dates that investment projects are ini-tially announced but also the dates that final termsfor financing projects are concluded and the datesthat project construction begins. Assume that

MNCs are trying to influence election outcomes bymaking investment project announcements de-signed to buttress claims of good economic stew-ardship by investor-friendly right-wing incum-bents. MNCs would then have incentives for“cheap talk,” whereby they announce many newprojects that they have little interest in actuallycarrying out. Cheap talk incentives are high in anelection years with right-wing incumbents and lowin election years with less investor-friendly left-wing incumbents. I gain insight into the possibilityof cheap talk by MNCs to promote reelection ofright-wing incumbents by examining the percent-age of announcements in which either project fi-nancing is concluded or construction is begun inthe three years after an election year. Given MNCefforts to buttress right-wing candidates, under mycheap talk assumption, the percentage of financedor constructed projects announced during electionyears with right-wing incumbents would be lowerthan the same percentage during elections withleft-wing incumbents. Figure 3 graphs average per-centages under these two scenarios and a thirdnonelection year scenario.

Percentages of projects either financed or underconstruction are higher (not lower) one and twoyears afterward for right-wing incumbent electionscompared to left-wing incumbent elections: 33 and39 percent of projects are financed or under con-struction one and two years after a right-wing in-cumbent election; 27 and 31 percent are financedor under construction one and two years after a

14 Reestimation of my full empirical model using esti-mated (at time of initial announcement) project U.S. dol-lar cost and a panel feasible generalized least squaresestimator yields results consistent with my core projectcount results. These results are available on request.

FIGURE 3Announced MNC Projects Financed or under Construction, 1987–2003

40 FebruaryAcademy of Management Journal

left-wing incumbent election. Three years afterleft-wing incumbent elections, the percentage ofprojects either financed or under construction(55%) noses slightly ahead of the same percentagefor right-wing incumbents (50%). This descriptiveevidence reveals no clear pattern consistent withcheap talk by MNCs designed to influence electionoutcomes. MNCs seem as willing to move aheadwith projects announced during elections withright-wing incumbents as they are with projectsannounced during elections with left-wing incum-bents. These results provide further support for mypolitical business cycle framework assumptions.They also suggest another future research avenueinvolving closer examination of electoral dynamicsaffecting MNC investment project financing andconstruction.

This final point invites further expansion of thepolitical business cycle empirical domain to em-brace other players important to the pricing andavailability of capital and capabilities critical todeveloping country growth and modernization. Isee value in building on research about developingcountry politics and bank lending by Dinc (2005)and applying political business cycle lenses to de-cisions by foreign bankers regarding loan limits andmaturities during elections. I also see value in ex-amining not only broad election-period risk acrossfirms, but also firm-specific differences in responseto such risk. If a key question in strategic manage-ment is why firms differ (Nelson, 1991), then a keyquestion for future research along these linesshould be why firms might differ in their responsesto political business cycle considerations. I demon-strated that firm nationality is a crucial source ofheterogeneity in response to electoral dynamicsand incentives related to political business cycles.Delios and Henisz (2000, 2003) suggest that previ-ous geographic experience moderates firm risk andinvestment in countries with uncertain policy en-vironments. Previous geographic experience mayalso moderate the impact of political business cycleconsiderations. Such future research can direct po-litical business cycle study further into the man-agement research domain and guide managementresearchers crossing disciplinary boundaries intopolitical economy and related domains.

REFERENCES

Ai, C., & Norton, E. 2003. Interaction terms in logit andprobit models. Economics Letters, 80: 123–129.

Alesina, A. 1987. Macroeconomic policy in a two-partysystem as a repeated game. Quarterly Journal ofEconomics, 102: 651–678.

Alesina, A., Roubini, N., & Cohen, G. 1997. Political

cycles and the macroeconomy. Cambridge, MA:MIT Press.

Arellano, M., & Bond, S. 1991. Some tests of specifica-tions for panel data: Monte Carlo evidence and anapplication to employment equations. Review ofEconomic Studies, 58: 277–297.

Beck, T., Clarke, G., Groff, A., Keefer, P., & Walsh, P.2001. New tools and new tests in comparative polit-ical economy: The database of political institutions.World Bank Economic Review, 15: 165–176.

Block, S., Singh, S., & Ferree, K. 2003. Multiparty com-petition, founding elections and political businesscycles in Africa. Journal of African Economies, 12:444–468.

Block, S., & Vaaler, P. 2004. The price of democracy:Sovereign bond spreads and political business cy-cles in developing countries. Journal of Interna-tional Money and Finance, 23: 917–946.

Bloomberg International. 2006. Credit ratings(Bloomberg online data services). New York:Bloomberg International.

Cantor, R., & Packer, F. 1996. Determinants and impactsof sovereign credit ratings. Federal Reserve Bank ofNew York Economic Policy Review, October: 37–53.

Delios, A., & Henisz, W. 2000. Japanese firms’ investmentstrategies in emerging economies. Academy of Man-agement Journal, 43: 305–323.

Delios, A., & Henisz, W. 2003. Political hazards, experi-ence and sequential entry strategies: The interna-tional expansion of Japanese firms, 1980–1998.Strategic Management Journal, 24: 1153–1164.

Dinc, S. 2005. Politicians and banks: Political influenceson government-owned banks in emerging countries.Journal of Financial Economics, 77: 453–479.

Drazen, A. 2000. The political business cycle after 25years. In B. Bernanke & K. Rogoff (Eds.), NBER mac-roeconomic annual: 124–135. Cambridge, MA: MITPress.

Esty, B. 2003. Modern project finance: A casebook.New York: Wiley.

Esty, B. 2004. Why study large projects? An introductionto research on project finance. European FinancialManagement, 10: 213–224.

Freedom House. 2006. Freedom House country rank-ings of political rights and civil liberties, 1972–2006. www.freedomhouse.org.

Ghemewat, P. 1991. Commitment: The dynamic of strat-egy. New York: Free Press.

Goldsmith, A. 1994. Political freedom and the businessclimate: Outlook for development in newly democ-ratizing states. Social Science Quarterly, 75: 115–124.

Haggard, S. 1990. Pathways from the periphery: The

2008 41Vaaler

politics of growth in the newly industrializingcountries. Ithaca, NY: Cornell University Press.

Hammond, J., & Allan, G. 1974. Bougainville CopperLtd. (B-E) (case no. 174104, 175071–175072,175304). Boston: Harvard Business School Publish-ing.

Henisz, W. 2000. The institutional environment for mul-tinational investment. Journal of Law, Economicsand Organization, 16: 334–364.

Henisz, W., & Delios, A. 2001. Uncertainty, imitation,and plant location: Japanese multinational corpora-tions, 1990–1996. Administrative Science Quar-terly, 46: 443–475.

Henisz, W., & Zelner, B. 2003. The strategic organizationof political risks and opportunities. Strategic Orga-nization, 4: 451–460.

Henisz, W., & Zelner, B. 2005. Legitimacy, interest grouppressures, and change in emergent institutions: Thecase of foreign investors and host country govern-ments. Academy of Management Review, 30: 361–382.

Hibbs, D. 1977. Political parties and macroeconomic pol-icy. American Political Science Review, 71: 467–87.

Hoetker, G. 2007. The use of probit and logit models instrategic management research: Critical issues. Stra-tegic Management Journal, 28: 331–343.

Hoskisson, R., Eden, L., Lau, C., & Wright, M. 2000.Strategy in emerging economies. Academy of Man-agement Journal, 43: 249–267.

Huber, P. 1967. The behavior of maximum likelihoodestimates under nonstandard conditions. In Pro-ceedings of the Fifth Berkeley Symposium onMathematical Statistics and Probability, vol. 1:221–223. Berkeley: University of California Press.

Humphreys, M., & Bates, R. 2005. Political institutionsand economic policies: Lessons from Africa. BritishJournal of Political Science, 35: 403–428.

International Foundation for Election Systems. 2006.Election guide. http://www.ifes.org.

King, G., Tomz, M., & Wittenberg, J. 2000. Making themost of statistical analyses: Improving interpretationand presentation. American Journal of PoliticalScience, 44: 347–361.

Kobrin, S. 1979. Political risk: A review and reconsider-ation. Journal of International Business Studies,10: 67–80.

Kobrin, S. 1987. Testing the bargaining hypothesis in themanufacturing sector in developing countries. Inter-national Organization, 41: 609–638.

LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., &Vishny, R. 1998. Law and finance. Journal of Polit-ical Economy, 106: 1113–1155.

Leblang, D., & Mukherjee, B. 2005. Government partisan-ship, elections and the stock market: Examining

American and British stock returns, 1930–2000.American Journal of Political Science, 49: 780–802.

Manila Times. 2003. Maynilad: A model in water pri-vatization springs a leak. http://www.manilatimes.net/others/special/2003/mar/26/20030326spe1.html.Published March 26.

Minor, M. 1994. The demise of expropriation as an in-strument of LDC policy, 1980–1992. Journal of In-ternational Business Studies, 25: 177–188.

Murtha, T. 1993. Credible enticements: Can host govern-ments tailor multinational firms’ organizations tosuit national objectives? Journal of Economic Be-havior and Organization, 20: 171–143.

Murtha, T., & Lenway, S. 1994. Country capabilities andthe strategic state: How national political institu-tions affect multinational corporations’ strategies.Strategic Management Journal, 15: 113–130.

Nelson, R.1991. Why do firms differ, and how does itmatter? Strategic Management Journal, 12: 61–74.

Nordhaus, W. 1975. The political business cycle. Reviewof Economic Studies, 42: 169–90.

Ring, P., Bigley, G., D’Aunno, T., & Khanna, T. 2005.Introduction to special topic forum: Perspectives onhow governments matter. Academy of ManagementReview, 30: 308–320.

Robinson, R. 1963. International business policy. NewYork: Holt, Rinehart & Winston.

Rogoff, K. 1990. Equilibrium political budget cycles.American Economic Review, 80: 21–36.

Schultz, K. 1995. The politics of the political businesscycle. British Journal of Political Science, 25: 79–99.

Standard & Poor’s. 1999. Sovereign defaults: Hiatus in2000. Standard & Poor’s Credit Week, December22.

Standard & Poor’s. 2000. Sovereign defaults declinethrough third quarter. Standard & Poor’s QuarterlySurvey Highlights: Outlook 2001, September 22.

StataCorp. 2005. Stata statistical software: Release 9.2.College Station, TX: StataCorp.

Thomson Securities Data Corporation. 2006. Project fi-nance investment database. New York: ThomsonFinancial.

Vaaler, P., & McNamara, G. 2004. Competition and crisisin organizational decision making: Credit ratingagencies and their response to turbulence in emerg-ing economies. Organization Science, 15: 687–703.

Vaaler, P., Schrage, B., & Block, S. 2005. Counting theinvestor vote: Political business cycle effects on sov-ereign bond spreads in developing countries. Jour-nal of International Business Studies, 36: 62–88.

Vaaler, P., Schrage, B., & Block, S. 2006. Elections, op-portunism, partisanship and sovereign ratings in de-

42 FebruaryAcademy of Management Journal

veloping countries. Review of Development Eco-nomics, 10: 154–170.

Vernon, R. 1971. Sovereignty at bay: The multinationalspread of U.S. enterprise. New York: Basic Books.

Wells, L. 1997. Enron Development Corp.: The DahbolPower Project in Maharashtra State (A-B)(abridged) (case no. 9–797-085–9–797-086). Bos-ton: Harvard Business School Publishing.

White, H. 1980. A heteroskedasticity-consistent covari-ance matrix estimator and a direct test for heteroske-dasticity. Econometrica, 48: 817–830.

World Bank. 2006. World development indicators, 2006.Washington, DC: World Bank.

Zaheer, S. 1995. Overcoming the liability of foreignness.Academy of Management Journal, 38: 341–363.

Zahra, S., Ireland, D., Gutierrez, I., & Hitt, M. 2000. Pri-

vatization and entrepreneurial transformation:Emerging issues and a future research agenda. Acad-emy of Management Review, 25: 509–524.

Paul M. Vaaler (vaal0001@umn.edu) received his Ph.D.from and is currently an associate professor of interna-tional business at the University of Minnesota’s CarlsonSchool of Management. He studies firm strategy and per-formance stability in turbulent industries. His interna-tional business research focuses on risk and investmentbehavior by firms and individuals active in emergingmarket countries experiencing economic and politicalmodernization.

2008 43Vaaler