Paper Autonomy, Private Ambition:

Theory and Evidence Linking Central Bankers’

Careers and Economic Performance∗

Christopher Adolph†

Harvard University

September 13, 2003

Abstract.

Central bankers’ careers are shown to influence inflation outcomes. I present

two theories in which careers explain central bank behavior, develop them

in a game theoretic model, and test them using a comprehensive new data

set of central bankers’ career backgrounds which spans twenty rich democ-

racies and half a century. Career experiences vary considerably over this

sample, and not only mould beliefs about appropriate policy (the social-

ization hypothesis), but also shape career concerns for central bankers who

seek career advancement in either the financial sector or government (the

career incentives hypothesis). Accordingly, time series cross-section anal-

ysis of inflation shows central bankers with financial sector backgrounds

preside over lower inflation, while central bankers with bureaucratic experi-

ence produce higher inflation. The magnitude of career effects on inflation

is on par with standard measures of central bank independence, and inter-

active models suggest both socialization and incentives contribute to these

career effects. The study of central banks in particular and bureaucracy

in general should pay greater attention to individual preferences and their

interaction with organizations; institutions alone are not enough.

∗Prepared for the Annual Meeting of the American Political Science Association, August 28–31, 2003, held inPhiladelphia. An earlier version of this paper was presented at the Comparative Political Economy Workshop,Cornell University, October 25-26, 2002. I thank Torben Iversen, Peter Hall, Jim Alt, Michael Hiscox, RobFannion, Gary King, Amanda Friedenberg, Katerina Linos, Joe Foudy, Dan Gingerich, and Victor Shih for helpfulsuggestions and conversations over the long gestation of this project. I am grateful for the translation assistanceof Christian Brunelli (Japanese) and Dean Hunt (Swedish). I am especially endebted to the Center for BasicResearch in the Social Sciences and the Multidisciplinary Program in Inequality and Social Policy, both of HarvardUniversity, and the National Science Foundation for supporting this research.

†Ph.D. Candidate, Department of Goverment. (Littauer Center, North Yard, Harvard University, CambridgeMA 02138; http://chris.adolph.name, cadolph@fas.harvard.edu).

Will it be sufficient to mark, with precision, the boundaries

of these departments in the constitution of the government,

and trust to these parchment barriers

against the encroaching spirit of power?

James Madison, Federalist 48

1 Introduction

In difficult economic times, observers wonder whether central banks will stimulate a slowing

economy or maintain a policy to keep inflation in check. In today’s gloomy environment, many

worry that central banks will be “too conservative”, paying too much heed to inflation while

global recession and perhaps even deflation loom. This dilemma raises questions not only about

institutions that try to balance independence and accountability, but also about the people

working within central banks themselves. Unfortunately, the political economy literature remains

ill-positioned to address both questions because it usually conflates central bank conservatism

with central bank independence. This confusion of preferences with institutions arises from the

unsupported assumption that independent central bankers are naturally conservative; in other

words, that heavy-handed government meddling is the only source of loose monetary policy.

Rather than ground a large and influential literature in an untested assumption, we should

disentangle our understanding of monetary preferences and institutions. We need a theory and

measure of central bank conservatism to complement existing work on central bank independence.

To understand central bankers’ monetary policy preferences, we should begin with central

bankers’ career paths and career concerns. A central banker’s career background may influence

his personal beliefs about the ideal trade-off between inflation and output stability, while at the

same time providing the basis for exchanges of future career favors for policy influence. Hence a

financial sector veteran may make a more conservative monetary agent than a former bureaucrat

for two reasons: first, because ex-bankers are conditioned to care more about inflation vis-a-

vis output; and second, because ex-bankers are more amenable to and suitable for high-level

financial jobs offered by financial sector firms pleased with conservative monetary policy. In

turn, bureacrats may be less enamored with inflation-fighting for its own sake, and more enticed

by tacit promises of high political office to provide the government accommodating monetary

policy. In sum, the career trajectories of central bankers before and after the central bank may

hold clues to their monetary policy conservatism while still in it.

The argument unfolds in two halves. I first place the career argument in the context of recent

research on central banks, developing both the socialization and career incentives mechanisms. I

also illustrate that career concerns can be embedded in the standard game threoretic model of

monetary policy (Rogoff 1986), thereby allowing outside principals leverage over monetary policy

even when the central bank is legally independent. In the second half of the paper, I test the

career background hypothesis using a comprehensive new dataset of central bankers’ career paths

from twenty countries over fifty years, showing that central bankers’ career experience explains

significant differences in inflation performance across countries and time.

2

2 Career paths and conservatism in the central bank

It is intuitive to most political scientists that we should examine policymakers’ preferences in order

to understand policy outcomes. Yet precisely this question has been overlooked in the study of

central banking. After reviewing the context and consequences of this oversight, I present an

argument for understanding the preferences of officials through their career backgrounds, which

may affect both the pre-existing preferences of central bankers and the incentives which face

them.

2.1 Do we really need to study central bank conservatism?

For twenty years, the problem of time inconsistency has dominated the study of monetary pol-

icy, suggesting that elected governments will always be vulnerable to the (arguably ultimately

futile) temptations of expansionary monetary policy, leading to permanently excessive inflation

(Kydland and Prescott 1977, Barro and Gordon 1983). An influential series of models suggests a

credibly conservative and independent monetary agent can ameliorate this inflationary bias (Ro-

goff 1986, Lohmann 1992). Scholars have measured and policymakers have implemenented central

bank independence, but the other half of the formulation—central bank conservatism—has been

ignored in comparative research.

Autonomy is the ability to act on one’s preferences. It tells us nothing about the content of

those preferences. Yet early studies set the precedent of treating CBI as a sufficient measure of

both autonomy and conservatism (Grilli, Masciandaro and Tabellini 1991; Alesina and Summers

1993; and Cukierman, Webb, and Neyapti 1992), and dozens of published works that rely in some

way on CBI have followed this example. This modeling choice is an oversimplification. It cannot

be justified by the explanatory power of CBI taken alone, since even the best CBI measures fail

to explain inflation performance in models with plausible controls (Campillo and Miron 1997)

or in developing countries generally (Cukierman, Webb, and Neyapti 1992). Moreover, no study

finds the expected positive relation between CBI and the variance of unemployment, suggesting

that CBI is an incomplete measure of nonaccommodation. Finally, and unsurprisingly, efforts

to disaggregate CBI into separate measures of independence and conservatism fail to find any

added effect of statutory injuctions to pursue price stability (Berger, de Haan, and Eijffinger 2001

review the evidence). Policy preferences run deeper than unenforceable commands, and only an

approach focused on the agents themselves will uncover the roots of central bankers’ behavior.

Excepting studies of partisan appointment, however, the central bankers themselves have

been ignored.1 Perhaps stereotypes of conservative, financial-sector-trained central bankers lead

many to assume that these agents, and hence central bank conservatism itself, are invariant across

time and space. But the stereotype is misleading: central bankers hail from a variety of careers,

of which private finance is not even the most common (see Section 3.2). As I will argue, differing

backgrounds may even form the basis for a measure of central bank conservatism.

A tendency to accord central bankers with remarkable self-restraint may also contribute to

1Several authors have examined the effect of partisanship in the appointment of central bankers, particularlyin the US (Chappell, Havrilesky, and MacGregor 1993), but also in Germany (Berger and Woitek 1999). Thesestudies generally find that more conservative parties appoint central bankers who pursue more hawkish monetarypolicy.

3

the neglect of their preferences. This faith—implicit in studies of central bank independence

that presume legal declarations of policy objectives will be followed as a matter of course—is

paradoxical for a literature centered on the inability of the government to faithfully execute the

long-term interests of its own principals. As McCallum (1995) notes, constitutional directives to

the central bank to pursue low inflation may only “relocate” the time inconsistency problem, since

the government has no more incentive to enforce such commands than it has to resist inflationary

policies in the first place. Moreover, given an independent central bank with an official policy

goal, it is unclear what the government could do to police or clarify the mission of a wayward

central bank, short of the costly step of changing the law itself.2 It is the very nature of agent

independence to give the principal as little power to enforce as to override.

Contrast this murky delegation problem with the hopeful view of Alan Blinder, an economist

and former Vice Chairman of the Fed, that central bankers (and political agents generally) check

their preferences at the door:

It is not necessary to find a “truly conservative” central banker whose personal value of the

parameter α [the amount of output the central banker is willing to sacrifice to lower inflation]

is excessive; you can simply direct the central bank to behave as if α were higher. In either

case, central bankers set aside their own personal beliefs about what is best for society (α or

k [the ideal inflation rate]) and adopt instead parameter values that lead them to “do their

duty.”

Blinder concedes that “Homo economicus may not behave this way. But responsible people, put

in positions of authority, do.” (Blinder, 1997, p. 14)

Blinder served on the FOMC, so his views cannot be dismissed out of hand. But from a

principal-agent perspective, Blinder’s claims ring hollow. First, compared with their political

principals, central bankers enjoy faster access to economic data and specialized staff studying

monetary policy questions. Information asymmetry is central to arguments for bureaucratic

dominance of policy, and it is easy to imagine that political principals tasked with dozens of

policy problems may not even be aware that agents are implementing policies the principal would

oppose if he knew more (Peters 1981, Weir and Beetham 1999). Second, Blinder’s “responsible

central banker” must not be subject to the unconscious biases that may follow each agent’s

unique experience, knowledge, and interests. Instead, he can tweak the “preference function in

his head” at the drop of a hat. Finally, the legal strictures on central bankers tend to leave

substantial wiggle-room; Blinder himself laments the lack of discussion and consensus on targets

and weights by the FOMC (Blinder 1997, 5). When the law does not say precisely what α and

k should be, there is no reason to expect all will interpret—or want to interpret—the law in the

same way. If agents can use the law to rationalize their pre-existing policy preferences, they are

2The European Parliament has learned how hard it is to hold an independent central bank “accountable”.Unfortunately, despite the spate of articles on central bank accountability, basic questions—what does it mean tobe accountable, who shall enforce accountability, and how—remain foggy (ironically, the meaning of the subsidiaryconcept of “transparency” is particularly contentious; see, e.g., the debate between central bankers Buiter [1999]and Issing [1999]). Moreover, central bank accountability, proposed in an era of low inflation and prosperity,remains wholly untested—there is not yet even an anecdote regarding a central bank which was held to account,or changed its monetary policy because of accountability institutions (like legally defined targets).

4

not constrained, but shielded from accountability.3

Because central bankers are neither homogenous nor straightjacketed, their own preferences

are likely to show up in policy. And monetary policy matters: most economists agree it has

real effects in the short run, so central bankers will always have to consider the trade-off between

inflation control and maintaining stable economic output.4 In this context, any presumption that

there is a single “right” level of inflation-aversion begs the political question, “Right for whose

interests?” Because the trade-off between inflation and economic stability has distributional

consequences, governments, political parties, and private actors differ in their preferred inflation

hawkishness. To this list of political actors, we must add the central bankers themselves. For

all the attention paid to grants of discretion to central bankers, what central bankers do with

discretionary power is woefully under-theorized. It is time to ask how central bankers’ preferences

vary, whether their preferences are influenced by other actors, and what effect those preferences

have on economic outcomes.

2.2 The career and policy choices of bureaucrats

This paper approaches the preferences of monetary policymakers by way of their career paths.

The effects of career incentives on public officials are often cited in the mass media, but have

received only sporadic attention in political science. It is worth reviewing several examples from

the literature to gain a sense of how career paths influence policy, what options exist for civil

servants to advance their private or political careers, and how opportunity structures vary across

countries.

Political actors’ career ambitions vary; some simply want to stay in place, others to rise to

higher office within a given sector or organization, and still others to rotate between two sectors,

ratcheting higher with each revolution. American politics is replete with examples of all three

career trajectories. Since Mayhew (1974), political scientists’ understanding of Congressional

behavior has centered on legislators’ overriding need to preserve their careers through re-election,

and since Schlesinger (1966), scholars have recognized that many legislators seek to rise above

3This critique of purely legal arrangements that purport to “solve” the monetary delegation problem alsoapplies to the concept of inflation targeting (Svensson 1997), adopted by several countries—the UK, Canada,Spain, Australia, Finland, New Zealand, and the ECB—throughout the 1990s. Many economists and centralbankers consider legally enshrining an inflation goal a crucial step towards central bank accountability. Essentiallyall inflation targeting countries follow a “flexible” inflation target in which the central bank accepts deviationsfrom the target to some degree (or for some period) to minimize output variability (Siklos 2002, Cukierman 2002).However, this degree of flexibility (i.e., the weight of the tradeoff between inflation and output stabilization) is neverdictated, policed, or even publicized, nor is the output target ever explicitly set (Cukierman 2002). Finally, themechanism by which a government would enforce the inflation goal is usually either unclear, untested, or difficultto impose. Indeed, per McCallum’s critique of CBI, governments may not even wish to punish central banks whofail to quickly return to the target inflation level. In this context, a central banker who officially enjoys onlyinstrument independence (in the Debelle and Fischer [1994] sense) could exploit weak oversight and informationadvantages to establish de facto goal independence, at least in the short run. Even central bankers with (partially)assigned goals retain substantial monetary policy flexibility, and the preferences of individual central bankers stillmatter under inflation targeting regimes.

4Walsh (1998, Ch. 1) is a good place to start on the short run trade-off between output volatility and pricestability. The popular Taylor Rule approximation of monetary policy decisions is one embodiment of this dilemma,and empirical work accepting this framework suggests that central bankers’ preference differ and matter. Forexample, Judd and Rudebusch (1998) argue that estimating separate Taylor-rule-like reaction functions for theFed under Chairmen Greenspan, Volcker, and Burns reveals systemically different behavior.

5

their office; from the state house to the US House, from the US House to the Senate, and

so on. In both cases, legislators’ policy preferences result from the career incentives created

by the electorate—and sometimes even the prospective electorate of a new office (Carey 1994,

1996; Rothenberg and Sanders 2000). In contrast, the executive branch is often described as

a revolving door for “in-and-outers”, officials who are first employees of regulated sectors, then

regulators themselves, and finally lobbyists or leaders for the regulated once more (see, e.g., Heclo,

1988). In recent years prominent US officials, including Vice President Dick Cheney, Securities

and Exchange Commission chair Harvey Pitt, and Treasury Secretary Robert E. Rubin, have

circulated through the private and public sectors, and in each case appeared to some observers to

make policy choices based on their career experiences.5 Of course, government officials generally

deny that career-incentives shape their policy choices, which suggests that we will have to look

for indirect evidence of career effects on policy.

Career paths matter in other countries besides the US. Reviewing the comparative evidence,

Schneider (1993) argues that where a bureaucrat stands depends not only on where he sits, but

on where he has sat and will sit. In many countries, senior bureaucrats often rise to the top of the

civil service only to jump to a lucrative private sector job. Known as pantouflage in France, this

pattern is also found in Denmark, Japan, the Netherlands, and Spain (Rouban 1999, Schneider

1993, Jensen and Knudson 1999, van der Meer and Raadschelders 1999, Alvarez de Cienfuegos

1999). On the other hand, pantouflage is limited or forbidden in Belgium, Britain (except the

Treasury), and Sweden—intensifying comptetition for the next rung on the civil service ladder

(Brans and Hondeghen 1999, Dargie and Locke 1999, Pierre and Ehn 1999).

This competition provides partisan governments leverage over agencies. Hence promotion

to the top of the civil service has grown more politicized in Britain and Germany, and has long

been a feature of bureaucratic appointments in most other European democracies (Peters 1997,

Mayntz and Derlien 1989, Dowding 1995, Page and Wright 1999). And though many consider

it the paradigmatic case of bureaucratic dominance, Japan may provide the best example of

political principals manipulating bureaucrats’ careers to ensure policy meets the government’s

needs. Loyal Japanese bureaucrats are often rewarded with private or political posts: thanks

to the support of the ruling party, from 1949–1980 the lower house of the Diet contained, on

average, 51 former bureaucrats, including 10 from the Ministry of Finance; bureaucrats rarely

ran for the opposition (Naka 1980, reproduced in Ramseyer and Rosenbluth 1993; Kim, 1988).

5Critics link Halliburton’s no-bid contract to rebuild Iraqi oil fields to Cheney’s years at Halliburton andlarge retirement bonus (e.g., David Lazarus, “Conflict of interest for vice president?” San Fransisco Chron-

icle, Nov. 3, 2002, G1). To others, the whole of Cheney’s career—through two presidential administra-tions and his interregnum as CEO—revolves around defense department favors to Halliburton; in Cheney’s fiveyears there, the company received $2.3 billion in federal contracts, up from $1.2 billion in the previous five(Robert Bryce, “Cheney’s Multi-Million Dollar Revolving Door”, Mother Jones News Wire, August 2, 2000,http://www.motherjones.com/news wire/cheney.html; Knut Royce and Nathaniel Heller, “Cheney Led Hallibur-ton To Feast at Federal Trough”, Center for Public Integrity, http://www.public-i.org/story 01 080200.htm,accessed July 24, 2003). Daniel Gross argues Harvey Pitt’s tepid response to the corporate accounting scandals of2001–2 reflects his need to “remain viable in the system” after leaving office, on the presumption the lawyer wantsto grow more wealthy representing accounting firms (“The Pitt: Why Chairman Harvey Pitt has failed,” Slate,Nov. 1, 2002, http://slate.msn.com/?id=2073450). And Joseph Stigliz (2002) claims Robert Rubin’s approachto international economic policy can be explained by his career path, which led from Goldman Sachs to Citibankby way of the Treasury.

6

Disloyal bureaucrats face the blacklist.6

The use of career rewards and punishments to affect policy implementation has implications

for the study of delegation that are not fully appreciated. Most of the vast delegation literature

focuses on formal, legal means of monitoring and control (see Bendor, Glazer, and Hammond

[2001] for a review). Initial pessimism that political agents cannot be controlled—since formal

avenues are often cumbersome and seldom used—has given way to tentative optimism that some

principals manage agents effectively without frequent recourse to formal sanctions (Huber and

Shipan 2002). It is often argued that an effective principal need not be an active overseer, but

instead can rely on a combination of credible sanctions and sporadic, perhaps even third-party,

monitoring (Weingast and Moran 1983; McCubbins and Schwartz 1984). But principals may

also control agents through extra-legal mechanisms: rewards and punishments that require no

hearings or legislation. Where available and effective, informal methods may even be preferred,

since they allow bureaucrats to maintain an aura of independence. In particular, contingent

career rewards as a subtle means of mastering agents bears investigation, and in this paper, I

demonstrate the utility of career incentives to uncover new insights in the well-plowed field of

monetary delegation.

Career mechanisms for bureaucratic control also open the playing field to groups outside

the government. Studying career incentives thus gives us theoretical and empirical leverage over

the case of an agent serving several masters. Recent work on delegation by multiple principals

tackles situations where the roles and powers of the principals are clearly defined (Epstein and

O’Halloran 1996, 1999; Morris 2000). But many political principals—such as interest groups

and firms—lack a formal role in policymaking and implementation, yet still exert influence by

informal means. Career rewards may be among the most potent tools these “shadow principals”

possess for manipulating bureaucrats.

2.3 Career incentives for monetary policy making

The application of career incentives to monetary policy delegation is not entirely new. The same

paper which popularized the “independence plus conservatism” solution alludes to a career-based

central bank conservatism (Rogoff, 1985), a notion seconded by Lohmann (1992) and Stigliz (2002,

p. 19). Central bankers are often veterans of the financial sector, the argument goes, and the

promise of future career rewards renders the financial sector a shadow principal of the central

bank, encouraging conservative monetary policy.7 The idea that financial sector experience makes

6Ramseyer and Rosenbluth (1993) document many career mechanisms by which the governing Liberal Demo-cratic Party manipulates Japanese bureaucrats, including threats of dismissal and control over promotion withinthe bureaucracy. Most important, elite bureaucrats’ wages are kept below market to ensure obedience to thegovernment, which has the legal right to grant or deny lucrative private and public jobs. Top civil servants re-tire early to avail themselves of these career rewards, a process known as amakudari, or “descent from heaven”(Koh, 1989). By showing that political principals can control seemingly independent expert agents using careerincentives, Ramseyer and Rosenbluth turned the notion of Japanese bureaucratic dominance on its head.

7Posen (1995) takes the argument further, claiming CBI is irrelevant—only financial sector influence matters.However, his proxies of financial sector strength (a lack of universal banking, banking regulation by the central bank,and the presence of federalism and party fractionalization) are somewhat distant from the concept of financial sectorinfluence and may pick up spurious correlation from alternative sources of central bank independence (federalism,for instance). The argument itself remains provocative, but without stronger evidence, we should not dismisscentral bank institutions and officials as epiphenominal.

7

central bankers more inflation-averse also appears in the literature on the Federal Reserve. For

example, Havrilesky and Gildea (1991) find that years spent in the financial sector predicts

Federal Open Market Committee (FOMC) members’ dissents in favor of tightness, while Wooley

(1984) and Belden (1989) link regional bank presidents’ greater conservatism to their their careers

in private banking. In the US, several former FOMC members attributed quitting the Fed to

the gap between public and private sector salaries.8 The evidence for career-concerned central

bankers grows when one notes from 1950 to 2000, the median Fed Governor chose to serve only

5.2 years of a guaranteed fourteen year term; many Governors re-enter the private sector on

leaving the Fed. To date, however, the career incentives argument has not been extended across

countries or career types.

Recognizing that central bankers are as likely to be career bureaucrats as financial sector

types opens a new front for the career-incentives approach. Just as a former private banker,

hoping to secure a better private banking job later, may bear private banks’ preferences in

mind when setting monetary policy, so may a bureaucrat or politician with ministerial ambitions

accommodate the government’s election-driven desire for economic stability. Career concerns

may loom large for a bureaucrat serving a stint at the central bank since governing parties in

industrialized democracies exercise significant control over the appointment of senior civil servants

and subcabinet ministers, even where the civil service is nominally neutral.

For both bankers and bureaucrats, the human capital, social networks, and preferences ac-

quired over a career point to the likely shadow principals at work behind the scenes. To the extent

that monetary policy makers are either ‘government’ or ‘financial’ types who feel pressured by

these sectors to be doves or hawks, the careers of central bankers constitute an observable mea-

sure of conservatism in central banks. Since governments have an electoral incentive to keep

the economy stable, while banks tend to be particularly concerned with inflation, my basic pre-

diction is simple: central bankers with career backgrounds in the financial sector should be more

anti-inflation than central bankers who are career bureaucrats.

Former finance ministry officials and former central bank staff (i.e., below the monetary

policy making level) comprise a possible exception to this hypothesis. Given their expertise and

opportunities to make connections with banking sector officials, these bureaucrats are more likely

to have developed their own views on monetary policy and are arguably better equipped to plunge

into the financial sector than other bureaucrats. As we shall see in Section 3.3, these officials

also have distinct career patterns from other bureaucrats who find themselves on central bank

boards. Therefore, I distinguish finance ministry and central bank experience from other types

of government experience, holding out the possibility that their career incentives may be quite

different.

2.4 Career socialization and monetary policy

There is an another interpretation of correlation between financial sector experience and cen-

tral bank conservatism: rather than career incentives, perhaps long sojourns in private banking

8This view has been made explicit by former Governors Robert C. Holland, who protested he could not pay hischildren’s tuition bills on a Fed Governor’s salary, and Jeffrey M. Bucher, who lamented the “financial penalty”he paid to leave the private sector for the Fed. Both men served only three years before returning to privateemployment (Katz 1992).

8

engender conservative ideas about inflation and monetary policy. At least three different mech-

anisms could produce pre-existing preferences for hawkish monetary policy: 1.) bankers may be

socialized to believe that inflation-fighting is the primary purpose of monetary policy, 2.) wealthy

private bankers’ material interests may induce anti-inflation feeling,9 and 3.) conservatives with

anti-inflation views may self-select into financial careers.

Of the three possibilities, socialization seems most promising. It is reasonable to suspect cen-

tral bankers’ approaches to policy problems are influenced by prior work experience, and that the

monetary policy convictions of career-based peer groups influence central bankers.10 The premise

that work experience influences attitudes is widely held by organization theorists (Hambrick and

Mason 1984, Gunz and Jalland 1996, van Maanen and Schein 1979). Experiments show that

industry-background and past functional roles within organizations influence executive decision-

making (Dearborn and Simon 1958, Beyer et al 1997, Melone 1994, and Hitt and Tyler 1991)

and in particular what information decisionmakers perceive as relevant (Rosman, Lubatkin, and

O’Neill 1994). Though there is scant recent work, students of political elites noticed in the 1970s

that career socialization had pervasive, lingering effects on the behavior of policymakers (Putnam

1976). One scholar of comparative elites went to far as to assert that “[v]alue-socialization is not

parental, or even based on early political experience, but apparently takes place from working in

a given field or institutional setting” (Barton 1973, p. 242; quoted in Putnam 1976).

For a political example of career socialization, consider American Supreme Court justices.

Long viewed as wise arbiters of legal precedent, justices have turned out to be political beings

whose policy preferences systematically influence their decisions (Segal and Spaeth [1993] built

the coffin for the legal theory, and Bush v. Gore hammered the nails). In part, justices owe

their policy preferences to their career tracks; for example, justices with prosecutorial experience

are more conservative on civil liberties decisions, controlling for the judge’s partisanship and

appointing president’s ideology (Tate and Handberg 1991). Since Supreme court justices generally

lack career concerns, this seems to follow socialization (and perhaps some self-selection), with

the experience, training, and environment of prosecutors imparting conservative policy beliefs.

For financial-sector-trained central bankers, the socialization argument is similar: work

within a sector that fears inflation and considers the struggle against it the only acceptable

monetary stance should shape the central banker’s own attitudes on monetary policy.11 Hence,

9See Scheve (2002a,b) for cross-national evidence that asset wealth leads to anti-inflation attitudes, and Burden(2000) for evidence that policymakers consider their material interests in passing laws

10Some speculate that central bankers might someday form (Kapstein 1992) or perhaps already constitute acohesive community collectively puzzling through policy choices; this perspective would relocate the socializationprocess within the central banking community itself. This paper does not suppose the existence of a policycommunity of central bankers, though the argument made here would jibe with the notion of several distinctcommunities (financiers, bureaucrats, and economists) intersecting the world of central banking. In particular,the growing cooperation among central bankers and academic economists on monetary policy research has struckseveral observers (e.g., McCallum 1999), but as the data collected here show, economists remain a minority ofcentral bank leaders, and comprehensive explanations of central banker behavior will thus need to reach beyondthe economic community.

11Private bankers have a long-standing reputation for conservatism on inflation, though I am unaware of anysystematic surveys of their attitudes. In the US, members of Congress and the Federal Reserve believe that “[i]fone’s goal is to minimize inflation ... a sure way to achieve that goal is to have private bankers—who are among theworld’s fiercest inflation hawks—appoint the regional bank presidents” (S. Greenhouse, “Showdown: The populistversus the Fed.” New York Times October 12, 1993, D1). And in many cases, banks’ fear of inflation surely hasa rational basis. For example, Santoni (1986) provides evidence that the stock prices of banks react negatively to

9

according to the socialization hypothesis, central bankers with career backgrounds in the financial

sector should be more anti-inflation than other central bankers.

The socialization hypothesis holds that past career experience determines preferences over

policy itself, while the career-incentives hypothesis sees in policy choices a means to a private end.

Both perspectives are theories about the effects of career paths, and to the extent that either

or both mechanisms act on central bankers, career variables are an important antecedant of

monetary policy. In the Model Appendix, I show how either career-incentive-induced preferences

or socialized policy preferences could affect the policy choices made by Rogoff (1986) central

bankers; readers interested in a formalization of the argument should turn there now. In the next

section, I test the career effects approach by looking at inflation performance across time and

countries, then present three tests to distinguish career incentives from pre-existing preferences.

3 Testing the career effects approach to monetary policy:

Evidence from advanced industrialized democracies

The theory elaborated thus far suggests a number of testable hypothesis regarding monetary

policy making. For now, I focus on the simplest empirical implication: monetary policy should

be more anti-inflationary in the hands of financial sector types than government bureaucrats.

This argument contains several steps—between careers and preferences, between preferences and

policy choices (votes), and between policy and outcomes (inflation and variation in real variables).

Here I attend only to the overall linkage between central bankers’ careers and inflation because

these are the two concepts most easily measured and compared over a broad array of countries

and periods. I then attempt to distinguish the role of career incentives and socialized preferences

by considered the context in which career effects on inflation operate.

3.1 Measurement

In developing measures of central banker’s careers, the first choice is whether to focus on what

they did before joining the central bank’s board, or what they did after. There are theoretical,

empirical, and practical reasons to concentrate on measures based on past career experience. It is

obvious that prior experience provides the context in which career socialization takes place, and

should therefore be a good measure to test the socialization hypothesis. But prior careers are

important for the incentives argument as well, since experience provides the specialized knowledge

and social networks on which (necessarily subtle and informal) job-for-policy exchanges may be

based. Choosing to work in a sector also reveals preferred career rewards, and I show in Section

3.3 that earlier work in a sector also strongly predicts post-central bank career patterns. Future

careers are central to the career incentives story, but from the perspective of central bankers

they are uncertain. A central banker may aim towards a future career that fails to materialize

for any number of reasons (e.g., poor health, poor performance, or the arrival of an unexpected

alternative). To the extent the central banker’s past reveals their expectations and preferences

at the time they served on the monetary policy authority, prior career measures should not be

unanticipated inflation, as might be expected of net holders of nominal assets.

10

neglected.

Given that past careers arguably proxy both types of career effects, there are three practical

reasons to develop measures using prior experience. First, the alternative—using the future to

explain the past—is not only unsettling, but also rules out prediction using observable variables.

Second, the quality of future jobs is hard to assess from extant records, and it is often difficult

to find any information at all on post-bank activities (especially in the private sector), while

background data are almost always available. Finally, we are interested in developing good mea-

sures of overall central bank conservatism, which may result both from career concerns and from

policy preferences which are induced or at least reflected by career backgrounds. Nevertheless,

the available future jobs data does help distinguish incentive and socialization effects, as I show

below in Section 3.5.

To measure the career background of a particular central banker at a particular time, I

classify his past jobs into six mutually exclusive and exhaustive categories: Financial (private

banking jobs), Government (bureaucrats outside the central bank and finance ministry), Finance

Ministry (bureaucrats in the finance ministry), Central Bank (staffers at the central bank), Eco-

nomics (academic economists), Business, and Other (international organization officials and staff,

other academics, labor union organizers, journalists, etc.).12 Most studies of political actors’ back-

ground use binary variables to capture such experience, but this practice has two key deficiencies:

it groups together specialists who have devoted their careers to one area with those who have

spent perhaps no more than a year in one place, and it overlooks changes in careers over time.

In contrast, I focus on the composition of each person’s career over time. For each job category,

I calculate an experience score, which is the fraction of the central banker’s career spent in that

job type as of his most recent appointment to a monetary policy making post.13 To state this

formally, let j index central banks, let i ∈ {1, ..., Ij} index central bankers, let t index time

periods (e.g., months, quarters, or years), and let d count days from a universal reference date.

Also let Careerij mark the start of i’s career, in days, let Appointijt be the day of i’s most recent

appointment to j, and let Jobsijd indicate the number of jobs i held on day d. Then define the

12To improve the international comparability of the categories, I made substantial efforts to include only privatelyowned and operated financial firms in the Financial category. State-run banks face very different incentives, andvirtually all individuals in the dataset who took a turn at such banks were career bureaucrats, not bankers. HenceI include management of government controlled banks in the Government category.

13The experience score methodology employed here should probably see wider use in the study of bureaucracy,especially to test whether agencies have been captured by private interests. The notion that bueaucratic agentsoften move through revolving doors to regulated sectors of the private economy is widespread (Stiglitz, Krugman,other cites), but is seldom rigorously explored. And studies which do link agents past careers with their actions ingovernment tend not to take full advantage of the available data, using binary or categorical indicators of experiencetypes rather than finer grained measures of experience that distinguish dabblers from careerists. Perhaps this isdue to the perception that coding more detailed data would be more expensive, but this is not necessarily so. Allthat is required is a record of each official’s past jobs, with starting and ending dates, and software to tabulatethese data into experience scores. Given underlying histories, it is easy to produce “simple” quantities, likethe experience scores used here, as well as more complicated variables that weight experience over time, take intoaccount personal context (e.g., interactions based on personal characteristics), and capture contextual relationships.Escore, the software I have written to accomplish these tasks on central banker data, can be applied to any otherbureaucracy. Hopefully this tool will encourage more scholars to investigate oft-invoked career effects in otherareas of policymaking. Escore runs in Gauss, and is available from http://chris.adolph.name.

11

financial experience of the ith central banker in the jth bank in period t as:

FinExpijt =

appointijt∑

d=careerij

FinJobijd

Jobsijd

/

(appointijt − careerij

). (1)

GovExpijt, FMExpijt, CBExpijt, EcoExpijt, BusExpijt, and OthExpijt are defined analogously;

together with FinExpijt, they sum to one.

To produce a set of experience scores for an entire central bank over period t we need an

aggregation mechanism. Work on the Federal Reserve (Chappell, Havrilesky, and MacGregor

1993) suggests that not all central bankers’ have an equal voice in policymaking (in particular,

the Fed Chair is estimated to have the effective voting power of any two Governors), but it

will be hard to estimate voter weights without recorded votes, which are often unavailable or

uninformative. Likewise, modeling appointment and oversight of monetary agents with attention

to institutional interaction (as is state of the art in studies of the Fed; see Morris [2000] and

Chang [1998]) requires substantial country-specific modelling. Here, I aim for a broad sweep

across countries to establish the importance of career variables, and hope that this will serve as a

foundation for more nuanced country-studies. Hence, I simply average the career experiences of all

central bankers who, by virtue of their positions, appear to have significant influence on monetary

policy.14 (I weight the averages by the proportion of the period each member served; I call this

procedure “tenure-weighting”.) Where it is possible to combine central banker characteristics

in a single index, I summarize the insitution’s characteristics by its (tenure-weighted) median

member, in the hope that the median voter theorem will offer some leverage on the preference

aggregation problem.

To define institution-wide experience scores formally, first let Durationt indicate the length

of t in days, and let Officeijt count the number of days i was in a monetary policy post at j during

t. Then define the financial experience of central bank j in period t as the weighted average of

individuals’ scores, where the weights are the fraction of the period each banker served:

FinExpjt =

Ij∑

i=1

FinExpijt

Officeijt

Durationt

/ Ij∑

i=1

Officeijt

Durationt

. (2)

GovExpjt, FMExpjt, CBExpjt, EcoExpjt, and BusExpjt are defined analogously. As for an

individual’s experience scores, the experience scores of a given bank in a given period always sum

to one.

14To determine which officials have voting authority on monetary policy questions, I turned to legal documentsfrom the various central banks, along data collected in Siklos (2002), Eijffinger and Geraats (2002), and Goodman(1992). Two special cases are worth mentioning. The first is the FOMC of the US Fed, which has four rotatingmembers representing the regional Federal Reserve Banks. I determined which regional bank presidents werevoting members at any given time from the Federal Reserve Bulletin. The second case is Canada, which reservesde jure monetary authority for the central bank governor only, but informally grants some power to a “GoverningCouncil” within the bank. Given the ambiguity of the Canadian case, I ran the analysis either including the entireGoverning Council, or excluding all members but the Governor. The result reported in the text fir the de jure

definition, and include the Canadian Governor only.

12

3.2 Data

The data consist of complete or near-complete career histories of about six hundred monetary

policy decision makers (e.g., governors, deputy governors, directors, and policy boards, or their

equivalents, where relevant) from twenty developed countries over the period 1950–2000.15 For

each policymaker, the database includes all jobs worked, by type; starting and ending dates for

each job; all positions at central bank, with dates of service; educational history; birth, gradua-

tion, retirement, and death dates; and gender (95% are male). Data were collected by the author

from central banks’ archives, biographical dictionaries, web resources, and business periodicals

(see Data Appendix for sources). Career histories were then tabulated into the individual- and

central-bank-level experience scores defined in Section 3.1. Tabulated experience scores are avail-

able for various period lengths (monthly, quarterly, annually) and collections of central banking

offficials (e.g., all officials, just governors, etc.). In this paper, I focus on quarterly scores averaged

over all of a given central banks’ officials with de jure authority over monetary policy.

A large majority (about 83%) of all work done by central bankers was in government (in-

cluding the finance ministry and the central bank itself), private finance, or economics (Figure 1,

top). This is a remarkable degree of convergence, given data on careers from university onwards

for hundreds of individuals scattered across twenty countries. Former bureaucrats are the most

common type of central banker, but less than half of past bureaucratic experience is in finance

ministries; the rest is spread over a variety of ministries often lacking any substantive connection

to monetary or economic policy. Private banking backgrounds are rarer than perceived, com-

prising only twelve percent of central bankers’ backgrounds, making it third-most common, after

bureaucrats and CB staffers.

The mixture of career types varies across nations. Some rely heavily on bureaucrats and

politicians to staff their banks; prime examples include Sweden, where parliamentary backgrounds

are common, as well as Belgium and Finland. Others depend more on financiers (New Zealand and

Denmark). The monetary policy authorities of France and Ireland are overwhelmingly veterans

of the Finance Ministry, while those of the United Kingdom, Canada, and Italy tend to be career

central bank staffers. The remaining countries tend to have more balanced boards, epitomized

by the US, Japan, Austria, and the Netherlands.

Looking at individual characteristics, rather than at central bank board averages, supports

the typology as well. At first appointment to a monetary policy position, the average central

banker was 47.8 years old, could expect to stay at the central bank for 6.1 years, and had spent 80

percent of his past career in just one of the sectors listed in Figure 1. Nine of ten central bankers

spent at least half their pre-appointment careers in one sector, while a third spent all of their

careers in just one type of job. Only 29 percent of freshman central bankers had ever worked

in a private bank, while 23 percent had worked in the finance ministry, 47 percent elsewhere in

government, and 23 percent in a private business. About two in five had worked previously as

central bank staff; one in five as an academic economist. Almost half (47 percent) had never

15Countries include Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy,Japan, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, andthe United States. For these countries, I identified a total of 721 monetary policy officials, and was able toassemble reasonably complete career histories for 598 of these, at least through their time at the central bank. Acomplementary database for developing countries is also being compiled.

13

Average Experience Score0.0 0.1 0.2 0.3 0.4

Other

Economics

Business

Financial

CB Staff

Government FinMin Non−FinMin

Financial

Government (Not Fin Min)

Finance Ministry

CB Staff

Economics

Business

Other

0

0.2

0.4

0.6

0.8

1

1950 1960 1970 1980 1990 2000

Car

eer

Exp

erie

nce

Figure 1: Composition and evolution of central bankers’ career experience: Define career experi-ence as the fraction of a central banker’s prior career spent in a job type, averaged across a bank’smonetary policymakers. The first plot shows that over the last half century, the past experiencesof central bankers in twenty industrial democracies can be described almost entirely as fallingwithin the state, central bank, private banking sector, and economics. The second plot showshow the average background across these countries’ central banks evolved, month by month, over1950–2000.

14

had a job outside the financial sector or government (including the finance ministry and central

bank)—which suggests most new central bankers would expect their next job, obtained perhaps

at age 54, to be in one of these places as well.

Returning to central bank level data, the bottom panel of Figure 1 traces the evolving mixture

of backgrounds within central banks over time. The most outstanding feature is the waning and

waxing of financial sector experience. Starting at an average of 17 percent of the cumulative

experience of the central bankers of 1950, financial experience steadily dropped until in made up

only seven percent of the backgrounds of 1970s central bankers. Then, in the first half of the

1980s, financial backgrounds shot up to their former highs, and have stayed around seventeen

percent ever since. The second notable trend is the steady growth in economics backgrounds,

totally absent in 1950, but fifteen percent of the past experience of central bankers in 2000. The

gains of financiers and economists have come primarily at the expense of bureaucrats, especially

those without finance ministry or central banking backgrounds. Over all, the public sector

experience of central bankers formed a majority bloc in 1950 (65 percent of all past careers),

grew to overwhelming dominance by 1979 (73 percent), then rapidly receded after 1980. In

2000, private sector experience—in finance, economics, and business—made up 41 percent of

backgrounds, or twice the 1979 nadir.

3.3 The revolving door at the central bank

A central question for is whether a career bureaucrat can easily pass through the top of the central

bank on his way to a private banking job, or whether central bankers tend to return to the sector

from whence they came. In short, is the central bank an “airlock” or a “revolving door”? A set

of probit models shows what past career experience implies for post-central bank appointments.

First I construct two binary variables, FinJobi and GovJobi, that indicate whether a central

banker obtained a job of corresponding type after leaving the central bank. (Note that GovJob i

includes all government jobs, including posts in the finance ministry and central bank, unlike our

usual experience score categories. The rationale for contrasting these types in the first place rests

on the hypotheses that they differentially predict future careers in banking or government as a

whole.) I regress each of these indicators on the same six explanatory variables: pre-central bank

levels of FinExpi, GovExpi, FMExpi, CBExpi, and EcoExpi, along with the age of the central

banker at the end of his service to the central bank. (For the estimated coefficients from these

models, see the Data Appendix).

Rather than numb the reader with probit parameters, I present expected values that show

how the probability of a post-bank career shifts as the corresponding pre-bank experience score

ranges from 0 to 1. This is always good practice, but especially useful here because of a subtle

issue in the interpretation of composition variables in regression models. By definition, a central

bank(er)’s experience scores must sum to one. Hence any hypothetical which alters one compo-

nent score is incomplete if it does not specify which other scores are adjusted to maintain this

accounting identity. When multiple components enter a model separately, then, it is inappropri-

ate to treat a single coefficient as a complete summary of the effect of a component score, since

there could be countervailing or reinforcing effects working through other parameters. To avoid

over- or understating relationships, I assume that a hypothetical increase in one career score is

15

0.0

0.2

0.4

0.6

0.0 0.5 1.0

Previous FinExp

−>Gov

−>Fin

0.0

0.2

0.4

0.6

0.0 0.5 1.0

Previous GovExp

−>Gov

−>Fin

0.0

0.2

0.4

0.6

0.0 0.5 1.0

Previous FMExp

−>Gov

−>Fin

0.0

0.2

0.4

0.6

0.0 0.5 1.0

Previous CBExp

−>Gov

−>Fin

Probabi-lity of

FutureCareerin. . .

Figure 2: The Revolving Door at the Central Bank. Expected values from probit regressions ofpost-CB careers on pre-CB experience scores. Each plot varies the portion of the central banker’spast career in a particular sector. Solid lines show the expected probability of leaving the centralbank board for a government job (including the finance ministry or central bank staff) and dashedlines show the probability of leaving for the finanical sector. The gray areas mark 95% confidenceintervals.

balanced by proportional reductions in all other scores, thus preserving the logical requirement

that experience scores sum to one.

Before proceeding, the reader is warned that the data suffer from one notable problem:

missing data. While the database has fairly complete data on prior careers, future jobs are often

hard to document, and we cannot usually distinguish missing data on the dependent variable

from an absence of post-bank jobs of that type. There is no easy fix for this problem, and it

surely introduces bias into the results; at a minimum, fitted probabilities of future jobs are likely

too low on average (since many of the zeros are probably ones, but not vice versa).16 Moreover,

this problem afflicts the different dependent variables to varying degrees. Government jobs taken

after central bank service are more likely to be recorded in the public record, and are more likely

to be noted in central bank archives. It is harder to find out whether central bankers took on

private banking roles after leaving the central bank; these jobs are seldom tracked by the central

bank, and are often overlooked by biographical dictionaries that emphasize public service. Web

searches turned up numerous cases where former central bankers joined private banks, raising

the suspicion that post-central bank financial jobs in the pre-internet era are undercounted.

Now to the results, starting with the clearest cases, former private bankers and (non-finance

ministry, non-central bank) bureaucrats. As expected, heavy experience in one makes a future

career in the other much less likely, shown by the downward sloping lines in the first two panels

of Figure 2. Conversely, one-time bureaucrats are likely to return to government after the central

bank, just as former financiers often pick up banking again (the upward sloping lines). These

“revolving doors” support the feasibility of career incentives.

Now consider our two ambiguous cases. In Section 2.3, I argued that central bank and finance

ministry staff may more easily join private banks than other bureaucrats. The third and fourth

panels of Figure 2 show this is the case. But note there is also a difference between the two types:

16Constructing an appropriate imputation model is complicated by the difficulty of establishing any of the zeroson the dependent variable with confidence. An exception is when central bankers die in office, but these cases aresurely not representative of other zeros, and are consequently of limited use in reducing bias through imputation.

16

central bank experience makes government jobs more likely, while finance ministry experience

makes them less so. While our findings on financial and government experience make intuitive

sense and seem likely to hold out of this sample, the career patterns of central bank and finance

ministry staff are likely to vary on a case by case basis. But for the data collected here, we

can conclude that the order of the career types based on their likelihood to lead to government,

rather than financial, future appointments is Fin → FM → CB → Gov. We should expect the

same ordering of the effects of these career types on economic outcomes.

3.4 Career effects and the level of inflation

The first step in exploring the connection between central banker characteristics and monetary

policy outcomes is to look at the linkage between central bankers and inflation in the post-

Bretton Woods era.17 To do this, I employ least squares time-series cross section regression with

standard errors corrected for panel heteroskedasticity (Beck and Katz, 1995). Linear regression is

appropriate since the dependent variable, logged quarterly inflation, is approximately normally

distributed (a few cases of deflation are omitted). I include country fixed effects to mitigate

omitted variable bias, and I include lags of the dependent variable. The estimating equation is

of the form

yit =k∑

j=1

φjyi,t−j + αi + X itβ + uit (3)

where X it is a vector of covariates, β is a vector of associated coefficients, αi is a country fixed

effect, and uit is a normally distributed disturbance.

Model 1 regresses logged inflation on several career components—financial experience, finance

ministry experience, economics experience, and government experience—while controlling for

CBI using an average of three well-known indices (the Cukierman, Webb, and Neyapti [1992]

index, with updated data from Maxfield (1997), the Grilli, Masciandaro and Tabellini [1991]

index, and the Bade and Parkin [1982] rankings). I also control for imports as a share of GDP,

which, according to several theories, should reduce the attraction of loosening the money supply

(Campillo and Miron 1997).18

Results are presented both in terms of estimated model parameters (Table 1) and counterfac-

tuals calculated from the estimated model (Figure 3).19 I focus on the counterfactuals because

they more transparently handle the problem of compositional explanatory variables described

in Section 3.3. We cannot read a single component’s coefficient as a “first difference” without

implicitly assuming that the increase in that category is made up by reductions in categories

with null effects (which may not even exist). The solution is to reduce all other categories in the

same proportion, and calculate expected values.

17That is, the data analyzed in this section and the next cover 1973–2000, with the exceptions of ECB membersafter 1997 and Spain and Portugal before democratic government (before 1979 and 1977 respectively).

18For example, Romer (1993) argues that openness lowers the benefits of raising output through monetary policywhile raising the inflationary cost. Lane (1997) supposes the benefits of surprise inflation act through raising theoutput of rigidly priced non-tradables; hence the more open the economy, the less the benefit of money surprises.

19Substantial serial correlation can endanger the consistency of time-series cross section regressions that in-clude lags of the dependent variable. Lagrange muliplter tests for serial correlation reject the null hypothesis ofautocorrelation when two lags of the dependent variable are included in the model, except in Model 4.

17

Expected DV: ln(Inflation)Variable Sign 1 2 3 4

FinExpj,t−2 − −0.14

(0.08)GovExpj,t−2 + 0.23

(0.08)FMExpj,t−2 −/+ −0.08

(0.06)CBExpj,t−2 +/− 0.12

(0.05)CBIj,t−2 − −0.91 −0.92 −0.90 −0.73

(0.30) (0.29) (0.29) (0.29)

CBCCmedj,t−2 − −0.09 −0.03 −0.06

(0.03) (0.07) (0.03)

CBIj,t−2 × CBCCmedj,t−2 − −0.12

(0.15)(Imports/GDP)j,t−2 − −0.02 0.02 0.05 −0.03

(0.26) (0.25) (0.26) (0.24)ln πworld

it 0.12(0.02)

ln πj,t−1 0.97 0.97 0.97 0.91(0.04) (0.04) (0.04) (0.04)

ln πj,t−2 −0.03 −0.03 −0.03 −0.04(0.04) (0.04) (0.04) (0.04)

Fixed effects x x x xN 1696 1696 1696 1696s.e.r. 0.304 0.305 0.305 0.299R2 0.887 0.887 0.888 0.891LM test (5% level ≥ 3.84) 2.60 2.64 2.57 4.31

Table 1: Natural log of inflation regressed on central banker characteristics, 20 countries, 1973–2000, quarterly. Least squares estimates with panel-corrected standard errors in parentheses.ECB members excluded after 1997. LM test refers to a Lagrange Multiplier test for serialcorrelation.

18

−4

−2

0

2

4

0.0 2.5 5.0

Years after ...

+1 sd FinExp−4

−2

0

2

4

0.0 2.5 5.0

Years after ...

+1 sd GovExp−4

−2

0

2

4

0.0 2.5 5.0

Years after ...

+1 sd FMExp−4

−2

0

2

4

0.0 2.5 5.0

Years after ...

+1 sd CBExp

−4

−2

0

2

4

0.0 2.5 5.0

Years after ...

+1 sd CBCC−4

−2

0

2

4

0.0 2.5 5.0

Years after ...

+1 sd CBI

cbi−c

cbi−3

Printer: this text is invisible filler should not be seen︸ ︷︷ ︸

Change in

Inflation

Change in

Inflation

Figure 3: Effects of individuals and institutions on inflation. Change in inflation following a per-manent 1 standard deviation increase in a career type, the career characteristics index (CBCC),or central bank independence (CBI). Each solid line is a separate counterfactual; those in thetop row summarize model 1, while those in the bottom row show the implications of model 3.When one experience score is increased, all other scores are reduced proportionately from theirmeans to maintain a sum of 1. Initial lags are set at mean observed inflation. All plots showexpected values as solid lines and mark 90% confidence intervals in gray. These intervals reflectthe cumulative estimation uncertainty produced by iterating the model through twenty periods.In the final plot, the solid line marked cbi-3 refers to the average of three CBI indices, whiledashed line marked cbi-c refers to the CWN index alone.

Figure 3 shows counterfactuals calculated from Model 1 using this method. According to the

model, increasing financial experience by one standard deviation corresponds with a 1.1 point

reduction in inflation over a five year period. That is to say, if the government changed the

membership of the central bank board from one with 17 percent of its collective experience in

finance to one with 30 percent, inflation would drop by a little over a point after five years

with the new board in office. In contrast, increasing government experience by one standard

deviation presages a 1.7 point increase in inflation. Both findings are significant, substantial,

and match our expectations under either the incentive or socialization hypotheses. Turning our

two more ambiguous cases, we find finance ministry experience is associated with significantly

lower inflation (1.1 points), while central bank experience is associated with significantly higher

inflation (1.4 points). The effects of the career types are ordered as hypothesized based on career

transitions (from least to most inflationary, they run Fin → FM → CB → Gov), supporting the

career incentive view.

The results from Model 1 suggest we can produce a single number summary of central bank

conservatism (at least as proxied by career effects) by simply summing experience in “conserva-

19

tive” job types and subtracting experience in “liberal” types. Define the Central Banker Career

Characteristics (CBCC) index as:

CBCCijt = FinExpijt − GovExpijt + FMExpjt − CBExpijt (4)

Logically, the index ranges from -1 (all “liberal” career experience) to 1 (all “conservative”). As

discussed in Section 3.1, since this is a single dimension proxying the policy positions of central

bank boards, taking the (tenure-weighted) median for each country-period makes more sense

than the mean. Over the 1973–2000 period, CBCCmed averaged -0.17 (s.d. = 0.53), and varied

substantially over time within countries and even more so across countries.20 Overall, the median

member of the average central bank first shot upwards on the CBCC index, from -0.35 in 1973

to a high of -0.01 in 1988, then drifted back to -0.16 in 1993, before rising again to -0.05 in 2000.

Model 2 regresses logged inflation on CBCC, controlling for CBI. I obtain similar results

using either mean or median CBCC, and either the combined CBI index or just CWN’s time-

varying version. The plot in the center of the second row shows that a one standard deviation

increase in the CBCC of the median central banker preceeds a 1.4 point decline in inflation over

five years.21 This result is highly significant, virtually identical to the effect of a one standard

deviation increase in CWN’s CBI index (1.5 points), and somewhat smaller than the effect of the

combined index (2.2 points).

Since we would expect central banker preferences to matter more when central bankers have

greater autonomy, Model 3 adds an interaction between these two variables. Though correctly

signed, this interaction is far from significant. I suspect this puzzling non-finding can be re-

solved through further work on the interaction between central bank preferences and the relevant

range of institutions (including not just CBI but also labor market organization and government

partisanship; see note 27). In the meantime, CBCC could serve as a provisional measure of

conservatism, especially since it explains inflation about as well as CBI. (CBCC also has the

distinct advantage of varying over time, and is easy to recalculate as the leadership of a central

bank changes.) And since CBI and CBCC are uncorrelated (in this sample, r = −0.02 for the

combined index, and 0.01 for CWN), the strong showing of CBCC casts doubt on the assump-

tion that CBI alone is an adequate proxy for monetary policy non-accommodation. Preferences

matter, and we ought not neglect them in favor of models relying only on institutional variables.

Empirical results are most convincing when they meet our prior expectations, are robust

to plausible respecifications, and resistant to outliers. The key findings presented here perform

well on all three criteria. First, they accord with the theory articulated in Section 2. Second,

the relation between CBCC and inflation persists when we alter the right-hand side variables

in various ways. For example, we could include a control for G7 average level of inflation to

allay concerns that the relations among CBCC, CBI, and inflation are a spurious result of trends

20Examining monthly CBCCmed data over 1973–2000, in the average country, the standard deviation over timewas 0.35, while for the average time period, the standard deviation across countries was 0.53. Including furthertime periods (e.g., going back to 1950) not surprisingly raises the proportion of variance explained within countries.

21The choice of equal weights for each component in CBCC is made for simplicity, but some plausible alterna-tives produce substantively similar results. For example, if we give only half weight to the two more ambiguouscategories—FMExp and CBExp—the effect of one standard deviation higher CBCC is -1.1 points of inflation, witha 90 percent confidence interval of -0.3 to -1.8.

20

in these variables. As Model 4 shows, our explanatory variables remain potent even when this

source of variation is removed. Likewise, we could change our measure of CBI to include just

the CWN index, include measures of the proportion of central bankers appointed by left or right

governments, or omit the imports variable, and obtain substantively similar results. Finally, the

key relations described here do not depend on a few influential observations. Since least squares

is notoriously vulnerable to outliers, I re-estimate equation 3 using robust regression techniques,

and again find similar results.22 For a detailed summary of these and other robustness checks,

see the Data Appendix.

3.5 Career effects on monetary policy: Native and induced preferences

Central bankers’ past careers correlate with inflation outcomes, but is this the result of socialized

policy preferences, career concerns, or both? To get leverage on the question, I test whether

factors that should increase the strength of career incentives augment the effect of past experience

on inflation:

• Future careers. As the putative reward for granting shadow principals their preferred

policies, post-central bank jobs in finance and government may reflect successful career-for-

policy-bargains leading to lower or higher inflation, respectively. If we suppose (consistent

with the model of monetary policy under career effects offered in the Model Appendix)

that career incentives augment socialized preferences, then we should see stronger effects

of career variables on central bankers for whom future jobs materialized.

• Public votes. The model of career incentives discussed in the Model Appendix assumes

shadow principals can observe the policy decisions of particular central bankers, at least

after the fact. This is the case where central banks eventually publish their voting records,

or where a single official makes policy, but secret voting procedures should hamper the

career rewards mechanism.

• Central bankers’ ages. Age may have two contervailing effects on career experience.

Since younger central bankers have more “career” left to worry about, they should face

stronger career incentives than central bankers nearing retirement; central bankers in the

“last period” of their careers face no career incentives at all. However, a 65 year old

central banker with, say, the same FinExp score as a 40 year old has spent more years

in the financial sector, which augurs for stronger socialization effects among older central

bankers.

The simplest of these test to implement is for the effect of public votes. Let PVjt be 1 for

countries that (eventually) publish monetary policy voting records or have a single monetary

decisionmaker, and 0 otherwise. A multiplicative interactive term, CBCCmed × PVj,t−2, should

augment the effect of CBCCmedj,t−2 if public votes help shadow principals apply career incentives,

22To test the resistance of the model to outliers, I re-estimated Model 2 using an M -estimator (specifically,one based on Huber’s influence function). Using this less efficient but more robust technique, I obtain an effectparameter of -0.043 (s.e. = 0.013) for CBCC and -0.177 (s.e. = 0.062) for CBI. Both results are significant andin accord with the LS results. Robust estimation of Model 1 also supports the LS findings. (For an accessibleintroduction to robust regression relevant to comparative political economy, see Western (1995).)

21

but have no effect if career effects work only through socialized preferences. The other two tests

require somewhat more complicated measures, since they involve individual level “microinterac-

tions” rather than institution level interactions.23 To capture the effect of youth on career effects,

we need to multiply a measure of youth with individual level characteristics, then aggregrate up

to the central bank level:

CBCCYY65medjt = median(CBCCijt × YY65ijt, wijt) (5)

In this case, our measure of youth is “years younger than 65”, YY65ijt = max(0, 65 − Ageijt).

If incentives overshadow native preferences, this variable should carry a negative coefficient,

augmenting the effect of CBCCmedj,t−2. But if socialization weighs more heavily, the coefficient

should be positive, since an older central banker with the same CBCC score as a younger official

has had more time to be socialized. Finally, to see whether career effects are stronger when

central bankers are known to have returned to financial or government jobs later, we multiply

the binary future job variables described in Section 3.3 by the relevant past experience. Because

of the ambiguity of finance ministry and central bank experience, I include it in both categories:

CBCCFJmedjt = median[FinJobij,t+ × (FinExpijt + FMExpijt + CBExpijt)

+GovJobij,t+ × (GovExpijt + FMExpijt + CBExpijt), wijt] (6)

As with the other variables, a negative coefficient augmenting the effect of CBCCmedj,t−2 provides

evidence that career incentives are at work.

I included each of these terms seperately in the inflation model, and report the results in

Table 2. The signs of the interactive terms are appropriate and the precision of the estimates

reasonably good for future jobs and public votes, but since interactive coefficients are easy to

misread, I focus on two kinds of counterfactual calculated from the estimated model. One is a

first difference that shows the effect of a standard deviation increase in CBCC under conditions

that either discourage or encourage incentives. The second is the “extra effect” of incentives,

that is, the difference in differences.

Given the same one standard deviation increase in median CBCC, inflation falls further when

the median central banker took a job later in finance than when he did not (1.9 points versus 1.4;

left panel of Figure 4).24 In both cases, the reduction in inflation is significant, suggesting that

socialization plays a role regardless. A look at the difference in differences (right panel) shows

23A common way to investigate the contingent nature of relationships among variables is to employ interationterms. If we suspect that the relation between X and Y depends also on the level of Z, we will usually specifyY = f(X, X · Z). The same technique can be used here to investigate the contigencies that arise from having aparticular level of, say, FinExp, given a certain degree of CBI, or from any other interaction taking place at aninstitutional level. However, where the contingency arises at the level of individual central bankers, we must take theinteraction into account before aggregating across the central bank. Hence, to consider the effect of each banker’s ageon the contribution of their financial experience to policy, we construct FinExpAgejt = a(FinExpijt×Ageijt, wijt),where a is an aggregating function (either a weighted mean or a weighted median) and w are the weights. (Ingeneral, of course, FinExpAgejt 6= FinExpjt × Agejt.)

24To calculate the appropriate hypothetical level of CBCCFJmedj,t−2, I assume the median central banker spent

his career in only two sectors, government and private finance. If CBCC = 0.35 (one standard deviation abovethe mean), this implies a hypothetical FinExp = (CBCC + 1)/2 = 0.675, which in this case is also the value ofCBCCFJmed

j,t−2.

22

Expected DV: ln(Inflation)Variable Sign 5 6 7 8

CBCCmedj,t−2 − −0.087 −0.068 −0.082 −0.073

(0.028) (0.031) (0.034) (0.032)

CBCCFJmedj,t−2 − −0.039 −0.030

(0.022) (0.022)CBCCmed × PVj,t−2 − −0.096 −0.080

(0.062) (0.064)

CBCCYY65medj,t−2 ? −0.0004

(0.0024)CBIj,t−2 − −0.980 −0.932 −0.913 −0.981

(0.294) (0.292) (0.294) (0.293)(Imports/GDP)j,t−2 − 0.047 0.023 −0.002 0.052

(0.257) (0.253) (0.254) (0.257)lnπj,t−1 0.967 0.970 0.970 0.967

(0.038) (0.038) (0.038) (0.038)lnπj,t−2 −0.027 −0.030 −0.029 −0.028

(0.038) (0.038) (0.038) (0.038)

Fixed effects x x x xN 1683 1696 1688 1683s.e.r. 0.306 0.305 0.305 0.306R2 0.886 0.889 0.885 0.886LM test (critical = 3.84) 2.388 2.476 2.840 2.369

Table 2: Natural log of inflation regressed on central banker characteristics, 20 countries, 1973–2000, quarterly: Interactive models. Least squares estimates with panel-corrected standard errorsin parentheses. ECB members excluded after 1997. LM test refers to a Lagrange Multiplier testfor serial correlation.

23

−3 −2 −1 0

Change in inflation five years after +1 s.d. in CBCC and ...

No future jobs

Future jobs

Secret votes

Public votes

Old (65+) CBer

Young (45) CBer

No jobs or votes

Both jobs and votes

−2 −1 0 1

Extra inflation reduction, five years after +1 s.d. in CBCC

Difference, having

a future job & not

Difference, public

& secret votes

Difference, young

& old CBers

Difference, jobs

& votes combined

Figure 4: First differences (left) and difference in differences (right) from interactive career ef-fects models of inflation. Circles and triangles mark point estimates, and horizontal lines 90%confidence intervals, for counterfactuals calculated from the models estimates shown in Table2. Results on the left are five-year first differences assuming a one standard deviation increasein CBCC, but making varying assumptions about the median central banker’s future jobs, votepublication, and age. Circles label scenarios that should increase career incentives; triangles onesthat suppress them. Results on the right are differences of the corresponding categories on theleft.

that the added impact of future jobs is also significantly different from zero (at the 0.1 level),

supporting the presence of incentive effects. Public votes had a similar effect: CBCC lowered

inflation rates whether votes were secret or not, but lowered them substantially more when votes

were eventually revealed (2.0 points versus 1.4). However, the extra impact of public votes is

not quite significant. Youth, on the other hand, appears to have a null effect: central bankers

young and old yielded the same reduction in inflation for an increase in CBCC.25 As the right

panel shows, this null result is imprecisely estimated, so we cannot say with much certainty it

is actually zero. But we can say that it is consistent with socialization and incentives sharing

roughly equal responsibility for career effects.

Finally, I estimate a model with both vote and future job interactions included. (Adding the

age interaction produces similar parameter estimates but larger standard errors, and is discarded

on efficiency grounds.) Using this model, I find the effect of CBCC is 56 percent larger when

votes are public and the median banker takes a financial job than otherwise. The extra effect of

the combined interactions is also significantly different from zero. While I do not consider later

jobs or revealed votes perfect proxies of career incentives, these finding suggests the two main

mechanisms offered by this paper operate on the same order of magnitude. Central bankers’

career backgrounds appear to affect both native and induced preferences over monetary policy.

25I experimented with various other transformations of age, but always obtained substantively similar results.

24

4 Concluding remarks

With a few exceptions, political economists have given short shrift to the role of central banker’s

preferences in explaining differences in monetary policy. To remedy this oversight, I have pre-

sented a theory, data, and empirical tests which suggest that central banker’s career concerns,

observable in their career histories, guide their behavior in office. The financial sector, govern-

ment, and economic experience of central bankers appear to affect inflation as much as the most

commonly used measure of monetary regimes, central bank independence. Some of the effect

of central bankers’ careers is likely a result of socialization and selection within the financial

sector and government, but several pieces of evidence suggest that prospective career hopes play

a significant role.

To the extent career effects are a reaction to career incentives set by the government and

private banks, legally-mandated central bank independence builds only “parchment barriers”

between government and central bank. True, laws may help central bankers act without the

constant threat of easy government vetos. But as long as monetary agents aspire to further

wealth or office, paper autonomy alone cannot guarantee the insulation of monetary policy from

outside interests. Whether one advocates or opposes such insulation, it is clear that the monetary

policy interests and ideas of central bankers and governments should no longer be shoved under

the CBI rug.

The immediate purpose of this research project is to bring preferences back in to the study

of central banks, but models relying on career concerns may be useful in the study of delegation

broadly. Career effects and incentives may permeate regulatory agencies, courts, and political

parties throughout the world. Using career incentives to explore preferences offers an alternative

to models which blithely “deduc[e] officials’ preferences from the attributes of their agencies,

without considering how preferences develop informally and over time.” (Schneider, 1993) At

the same time, attention to career paths can help researchers sort through relationships among

competing principals and organizations using the characteristics of agents themselves. Research

along these lines can enrich our understanding of what happens within and across institutions

by bringing the political actors that inhabit them back to center stage.

25

Appendix A: Model Appendix

Monetary policy delegation with career concerns

To illustrate the career effects argument, I develop a simple model of monetary policymaking

which builds on the standard model by treating the monetary agent as one of a series of ap-

pointees, each with his own career track and ambitions. This constitutes a “career concerns”

approach to monetary policy delegation, akin to models used to study managerial labor mar-

kets.26 Career concerns models have much to offer political science, particularly in the area of

delegation. Ambitious policymakers care not only about the policy discretion or rents they can

extract today, but also about their ability to advance to more prestigious, powerful, or lucrative

posts tomorrow. Principals can exploit career anxiety to reward or punish their agents: compli-

ance wins a plum appointment, shirking a cold shoulder. But shadow principals waiting in the

wings can play the game, too. An organization (such as a party, interest group, firm, or rival

bureaucracy) possessing no present contractual relationship with the agent could still informally

promise a future appointment which beats the original principal’s offer; in exchange, the shadow

principal receives a better policy today. In this way, varied prospective principals can exert pres-

sure on the agent, whose career path links institutions lacking any formal connection. If today’s

central banker is tomorrow’s private banker, today’s monetary policy may belong to the banks

as well.

Economic assumptions

The initial set-up of the game mirrors Barro and Gordon (1983), Rogoff (1986), and Lohmann

(1992). Assume the economy follows a Lucas supply function,

y = π − w + z, (7)

where y is economic output, π is inflation, w is the wage level, and z is a normally distributed

shock with mean zero and standard deviation σz. The labor market is characterized by price-

takers who accept w = E(π).27 If the monetary authority has quadratic utility over inflation π

and output y, with ideal output y and ideal inflation of zero,

U = −(1 − χ)(y − y)2 − χπ2, (8)

26Holmstrom (1982) introduced this approach, which ordinarily focuses on encouraging effort or skills acquisition,rather than issues of policy discretion. Career concerns models assume a principal-agent problem (i.e., monitoringof effort is imperfect) which is at least partly resolved by the agent’s concern for future employability—in hiscurrent firm or elsewhere—which in turn depends on observable outputs from his labors (for an example involvingCEOs, see Gibbons and Murphy, 1992). Tirole (1994) suggests that since public sector actors accrue less monetarycompenstion than private managers (especially in the sense of receiving only a small share of their marginalproduct), career concerns loom at least as large in government agencies as in the private sector.

27Work on labor-market/central-bank interaction casts doubt on the applicability of this assumption, especiallyoutside the US (see, e.g., Iversen [1998a,b and 1999], Hall and Franzese [1998], Franzese [1999], and Cukiermanand Lippi [1999]); I use it here as a starting point, and plan to build more plausible assumptions regarding thestructure of the labor market and dynamics of wage setting into future versions of the model.

26

then monetary policy will be subject to an inflationary bias inversely related to the conservatism

of the policymaker, χ. In turn, equilibrium output is unaffected by money on average, but the

variance in output in response to shocks grows with the conservatism of the monetary agent:

π∗ = (1 − χ)

(y

χ− z

)

, y∗ = χz. (9)

This sets up a trade-off between the level (and variance) of inflation and the variance of the

real economy, over which different policymakers may have different preferences, governed by χ,

y, and σz. Note that if governments could credibly commit to a conservative monetary policy

(one based on a high χ), they would enjoy lower inflation and the same output on average. But

governments will be hard pressed to keep their commitments when a deviation from the rule may

keep them in power, a contingency other actors anticipate. This is the time inconsistency problem

that led Rogoff to suggest delegation to a conservative, independent central banker. (In practice,

this would mean choosing an agent who has conservative beliefs, or it could mean relying upon

other factors to induce conservative behavior.) Of course, since conservatism comes at the price

of greater economic instability, governments will not want an ultraconservative central banker,

either. All else equal, different governments will prefer agents with different preferences, though

more conservative ones than the government has.

Based on these well-known results, we can already state the implications of career-socialized

preferences. To the extent a career in finance produces greater inflation sensitivity, a central

banker will have higher χi, and will thus produce lower inflation and higher real economic vari-

ability. It is crucial to underscore that this is not an implication of independence alone, but of

independence plus conservative preferences; by explaining where those preferences originate and

how they may vary, the socialization hypothesis is a crucial adjunct to the Rogoff model.

Establishing the effects of career incentives requires a further generalization of the model, to

which I now turn.

Players

I generalize the game to include monetary agents of different “types”. Some agents care most

about policy, others wealth, and still others political office. Agents may also have different

personal beliefs about the appropriate course of monetary policy, and different forms of human

and social capital which make them more or less suitable for other kinds of future employment.

From this universe of potential agents, the government delegates monetary authority to a

single central banker who enjoys legal independence to set monetary policy as he wishes, includ-

ing making (unwritten and legally unenforceable) arrangements to exchange monetary policy

influence today for career favors tomorrow.

Central banker i has a three-period career. Period 0i is always spent outside the central

bank, for example in the financial sector or government. The government appoints the central

banker to set monetary policy in period 1i. In the last period of his career, period 2i, a central

banker may either take a job outside the bank or continue as monetary policy agent.

Central bankers derive utility from policy itself (π and y), from wealth-enhancing private

sector jobs (m), and powerful government positions (r). The relevant portions of the ith central

27

Liberal Conservative

Office-seeking 1χ, τ, τ

θχ, τ, τ

θ

Wealth-seeking 1χ, θ, θ

τχ, θ, θ

τ

Policy-seeking 1χ, 1

θ, 1

τχ, χ

θ, χ

τ

Figure 5: A typology of central bankers: A central banker approaches an ideal type when thequantities in the corresponding box are high. τ is the weight on office rents in the centralbanker’s utility function, θ the weight on wealth, and χ the weight on inflation relative to thereal economy.

banker’s utility function are:

Ui = − (1 − χi)(y1 − yi)2 − χiπ

21

︸ ︷︷ ︸

current policy

+δi

[

− (1 − χi)(y2 − yi)2 − χiπ

22

︸ ︷︷ ︸

future policy

+ θim + τir︸ ︷︷ ︸

future jobs

]

(10)

Depending on the value of these parameters, a central banker may be principally concerned with

policy, wealth, or rents from political office, and with respect to policy may be either conservative

or liberal. This suggests six central banker ideal types, summarized in Table 1.

To keep things simple, I assume that the central banker makes his final choice of π by calcu-

lating the value of χ∗ which, when plugged into Equation 9, will maximize his utility according

to Equation 10, taking into consideration both career side-payments (m or r) and the central

bankers true policy preference (χi). In other words, the prize in this game is the policy parameter

χ∗, ultimately chosen by the central banker, which in turn yields π∗ and y∗ according to equation

9.

There are two other players in the game, the financial sector (F) and the government (G),

both of which are treated as indefinitely-lived unitary actors (i.e., they do not change over the

course of the game). Each receives utility from policy, and loses utility by doling out positions

(since this incurs an opportunity cost). F and G have no legal role in setting π, but may make

promises of m or r to the agent in exchange for the chance to choose the equilibrium level of χ

and thus π. Their utility functions are similar:

UF =∑

∀t

δt−1F

[−(1 − χF)(yt − yF)

2 − χFπ2t − θFmt

], (11)

UG =∑

∀t

δt−1G

[−(1 − χG)(yt − yG)2 − χGπ2

t − τGrt

]. (12)

I assume χF > χi > χG to focus on the case in which a.) tension exists between the govern-

ment’s monetary preferences and those of more conservative financial sector, b.) the government

28

Player/Time 0 1 2 3 4 · · ·

CBa Fin (0a) CB (1a) Fin (2a)CBb Fin (0b) CB (1b) Fin (2b)CBc Gov (0c) CB (1c) Gov (2c)

......

.... . .

Figure 6: Example career tracks of successive central bankers. Cell entries show the location ofeach player in each period (either the financial sector, Fin; the government, Gov, or the centralbank, CB. In parentheses are shown the career-period number and agent letter; hence at time 1,player b is in period 1 of his career and serving as central banker while player a is finishing hiscareer up in the financial sector.

attempts to stave off the temptations of monetary policy through delegation to a more conser-

vative, legally independent central banker, but c.) this banker is still not as conservative as the

financial sector desires.

Play of the game

The game takes place over an indefinite period, with each central banker serving at most two

periods. To understand the policies and career tracks supportable in equilibrium, it is necessary

to consider the play of the game over the latter two periods of a central banker’s career:

Period 1i

• F offers CBi a job in Period 2i worth m in exchange for CBi’s promise to set policy according

to χF in Period 1i.

• Simultaneously, G offers a job in Period 2i worth r in exchange for CBi’s promise to set

policy according to χG in Period 1i.

• Subsequently, CBi chooses a policy trade-off χ∗ ∈ {χi, χF, χG}. Policy choices result in

same-period economic outcomes, π∗1i and y∗1i.

Period 2i

• F and G decide whether to make good on their offers, choosing m∗ ∈ {0, m} and r∗ ∈ {0, r}.

• CBi chooses among his available career options, and either stays at the central bank or

heads to the financial sector or government. If CBi stays at the central bank, he sets period

2i monetary policy according to χ∗ = χi, and π∗2i and y∗2i result. Otherwise, the government

appoints a new central banker, and the game begins again.

Equilibrium

Observers of politics often assume that organizations can consummate policy-for-career-rewards

bargains. A game theorist, however, would wonder how these deals can stick. Why don’t organi-

zations leave agents hanging after receiving the policy they want? Repeated play offers one way

29

out of this conundrum, as we will see below. Whether this is the right explanation—and whether

career deals actually stick—remain empirical questions.

In the one-shot version of the game described above, would-be shadow principals face a time

inconsistency problem. Even when Pareto superior outcomes are possible through job-for-policy

bargains, once the second period is reached, the offeror has no incentive to pay, so central bankers

would refuse to make deals in the first place. However, a form of the folk theorem applies to

repeated games played by long- and short-run players (Fudenberg, Kreps, and Maskin, 1990).

Provided the long-run run players (F and G) play last in the stage game, as they do here,

reputational concerns can enforce cooperation (assuming short-run players are aware of the past

behavior of long-run players). Since shadow principals who want to deal with today’s central

banker will also want to deal with tomorrow’s, worthwhile bargains are in equilibrium, so long

as there are gains to trade.

To characterize equilibria allowed by the folk theorem, we must identify Pareto improving

job-for-policy trades—jobs offers which are both feasible (the offeror would be willing to trade the

job for policy) and acceptable (in the sense of being the best option before the central banker). For

example, the financial sector will only offer jobs that cost less than the policy which would have

been implemented otherwise; a central banker will only accept jobs that provide more utility than

either independent policymaking or any counter-offer from the government. Given the actors’

preferences, there may be no such offers (in which case the banker implements πi), only one offer

(which the banker accepts), or two equally good offers (a knife-edge case of little interest).

For a formal characterization of equilibrium in this game, see the final part of this appendix,

‘Characterizing equilibrium’. The next section provides an intuitive view of the implications of

this game for monetary policy choices.

Comparative statics and empirical implications

The easiest way to understand the empirical implications of the model is to compare equilibria

that result when different types of central bankers act as the agent. To illustrate comparative

statics, I use graphics that divide up the parameter space according to the policy the central

banker will adopt at each point in that space: his personally prefered policy, the financial sector’s

policy, and the government’s. These diagrams demonstrate that agents with preferences typical

of a financial career will implement the financial sector’s preferred policy, and vice versa for

government types.

First, we explore the effects of different central banker preferences on the monetary policy

outcome, holding fixed the preferences of the financial sector and government. Figure 7 shows

which sector will win the monetary policy auction given central bankers with varied tastes for

policy, political office, and private banking positions, along the lines sketched in Figure 5. Fo-

cusing on the middle panel, which depicts the case of a central banker with moderate monetary

policy preference, note that when the central banker cares more about winning a government post

(has higher τi) the government tends to prevail in monetary policy, but when the monetary agent

cares more about financial sector rewards (has higher θi), private banks win. But if the banker

cares little for either kind of reward, then he follows his own policy preferences and remains in

the central bank.

30

Liberal CBχi = 0.4

Moderate CBχi = 0.6

Conservative CBχi = 0.8

Desire for Wealth (θi)

Des

ire fo

r Offi

ce (τ

i)

0.0

0.5

1.0

0.0 0.5 1.0

Gov

CB

Fin

Wealth−seeking (θi)O

ffice

−see

king

(τi)

0.0

0.5

1.0

0.0 0.5 1.0

Gov

CB

Fin

Wealth−seeking (θi)

Offi

ce−s

eeki

ng (τ

i)

0.0

0.5

1.0

0.0 0.5 1.0

Gov

FinCB

× Typical government type:

dovish to slightly hawkish on inflation,

prefers gov to fin jobs

→ adopts Gov monetary prefs

N Typical financial type:

strong inflation hawk,

prefers fin to gov jobs

→ adopts Fin monetary prefs

× ×

�

Figure 7: Who sets monetary policy? Monetary policy according to central banker preferencesShaded regions indicate whose monetary preferences the central banker will adopt, as the in-dicated parameters of the model are varied; this is also the sector in which the central bankerwill work in the next period. Fin indicates the financial sector, Gov the government, and CB thecentral banker’s own preferences. The darker the region, the less conservative the policy imple-mented. In the final plot, a thin, barely visible stripe extends down along the vertical axis tothe origin from the CB region (i.e., a central banker with no career concerns always implementshis own policy preferences). Typical outcomes for government types (×) and financial types (N)are shown. For all plots, I assume that χF = 0.9, χG = 0.3, θF = τG = 0.25, yi = yF = yG = 0.2,δi = 0.95, and σz = 1.

Looking from left to right across the panels shows what happens when we consider the range

of central bank inflation preferences. Four patterns emerge:28

1. Regardless of the central banker’s preference over policy, any outcome is possible provided

the right career incentives. In particular, the corner cases are always the same: an agent

who wants office but not wealth always sides with government, an agent who wants wealth

but not office with the financial sector, and an agent who cares for neither implements his

own preferred policy.

2. Career incentives and central banker inflation preferences interact, so bargains are easier

the closer the agent’s preferences are to the principals. Hence a conservative central banker

must strongly prefer government advancement to financial sector jobs if he is to make a

career deal with the government, while a wider range of liberal or moderate central bankers

would accept the government’s offer.

28All conclusion drawn from this section depend, of course, on the specific values of parameters we do not vary.For the most part, this is a fairly minor issue, since reasonable ranges of the parameters have been chosen andchanging them holds no surprises.

31

3. Over the range of central bank preferences shown, opportunistic central bankers, who are

equally and substanstially attracted to private and government posts, always side with the

financial sector, which offers a bigger reward ceteris paribus. This follows from assuming

monetary policy affects the level and variance of inflation, but only the variance of output.

Thus, unless the economy fluctuates wildly (σz is large), the financial sector stands to win

or lose more by intervening in monetary policy that the government, and is willing to pay

more for policy as a result.

4. Central bankers are truly independent only when their career concerns are minimal. This

holds in cases where the costs of job offers (θF and τG) dispersed across whole organizations

are (reasonably) assumed to be smaller than their concentrated benefits to agents (θi and

τi). I relax this assumption presently.

Together these patterns suggest that “financial type” central bankers—who prefer financial sector

jobs and espouse hawkish inflation views—will consummate career-for-policy bargains with the

financial sector, while “government types”—who prefer government posts and have inflation

preferences ranging from dovish to slightly hawkish—will accede to government demands to

advance their careers.

Figure 8 shows how the hiring costs of government and private banks affect their ability

to influence monetary decisions across the range of central banker types. The columns of plots

correspond to the policy preferences of the central banker, as before. But now the rows of plots

reflect whether the agent is office-, policy-, or wealth-seeking, while the axes show the cost of

hiring for the financial sector (vertically) and government (horizontally). Hiring costs encompass

a range of concerns, including the central banker’s potential productivity in a sector (based,

perhaps, on his human capital and accumulated social networks) and the size of the organization

across which the costs and benefits of the bargain are spread. Lower hiring costs make job-for-

policy bargains easier to consummate.

All plots demonstrate that ceteris paribus agents who can be hired cheaply by one sector but

dearly by another will side with the cost-effective sector. Where costs are closer to equality, the

agent’s preferences usually tip the scales. The exception is when strong career preferences are

reinforced by congruent policy biases (liberalism combined with office-orientation or conservatism

with wealth-orientation), in which case agent preferences tend to overwhelm all but the largest

disparities in hiring costs (as in the top left and bottom right plots).

To the extent that financial types are more easily re-hired by the financial sector, and govern-

ment types by the government, these comparative statics reinforce the impression that monetary

policy will be true to central bankers’ types. But consider for moment an appointee who does

not exactly conform to type: suppose the government appoints a career bureaucratic with po-

litical aspirations but conservative inflation preferences. This government is following Rogoff’s

recommendation to appoint a “more conservative” central banker, but the agent’s career concerns

undermine his independence from the government. When in economic trouble, the government

can employ career leverage to get around the central bank’s nominal independence. As the top-

right plot shows, this kind of agent will side with government in the (more likely) case that

government job offers are less costly than financial ones; this central banker is marked “a”. In

contrast, a conservative financial type will strike a career bargain with the financial sector, as the

32

Liberalχi = 0.4

Moderateχi = 0.6

Conservativeχi = 0.8

Office-oriented

θi = 0.2τi = 0.8

Policy-oriented

θi = 0.1τi = 0.1

Wealth-oriented

θi = 0.8τi = 0.2

Government hiring cost (τG)

Priv

ate

bank

hiri

ng c

ost (

θ F)

0.00

0.25

0.50

0.00 0.25 0.50

Gov

Fin

.

.

0.00

0.25

0.50

0.00 0.25 0.50

Gov

Fin

.

.

0.00

0.25

0.50

0.00 0.25 0.50

Gov

CB

Fin

.

.

0.00

0.25

0.50

0.00 0.25 0.50

Gov CB

Fin

.

.

0.00

0.25

0.50

0.00 0.25 0.50

GovCB

Fin

.

.

0.00

0.25

0.50

0.00 0.25 0.50

Gov

CB

Fin

.

.

0.00

0.25

0.50

0.00 0.25 0.50

Gov

Fin

.

.

0.00

0.25

0.50

0.00 0.25 0.50

Gov

Fin

.

.

0.00

0.25

0.50

0.00 0.25 0.50

Gov

Fin

× ×

�

× Typical government type:

dovish to slightly hawkish on inflation,

prefers gov to fin jobs,

more easily employed in Gov

→ adopts Gov monetary prefs

N Typical financial type:

strong inflation hawk,

prefers fin to gov jobs,

more easily employed in Fin

→ adopts Fin monetary prefs

a

b

Figure 8: Who sets monetary policy? Monetary policy according to career-bargain costs. Shadedregions show whose monetary preferences the central banker will use, according to the model, asthe indicated parameters are varied; this is also the sector in which the central banker will workin the next period. Fin indicates the financial sector, Gov the government, and CB the centralbanker’s own preferences. The darker the region, the less conservative the policy implemented.Typical outcomes for government types (×) and financial types (N) are shown. Also demonstratedis that for otherwise identical conservative agents, different career goals can produce differentpolicy outcomes (contrast “a” and “b”). For all plots, χF = 0.9, χG = 0.3, yi = yF = yG = 0.2,δi = 0.95, and σz = 1.

33

banker labelled “b” in the bottom right plot shows. Thus even with a legally independent central

bank, central bankers who prefer financial sector jobs will produce more conservative policy than

those who prefer government jobs.

Finally, two simple implications of the model should be noted. First, if the central banker

has zero concern for political office or private jobs, then he will always implement his preferred

policy. To the extent that central bankers’ policy preference reflect career socialization, then,

financial types will still adopt conservative monetary policy. (Note, however, from the middle row

of Figure 8 that even a modicum of career concerns might be enough to shift policy decisions.)

Second, even if central bankers exhibit career concerns, all careers come to an end. Central

bankers approaching their retirement have little concern for job-policy bargains, and can again

be expected to implement their own preferred policies.

Caveats and extensions

A key assumption in the model is that both the financial sector and the government are

indefinitely-lived unitary actors. While financial sector firms might foresee indefinite lives, they

are not generally unified in behavior or interests. Any given firm could hope to free-ride on

another firm’s efforts to buy monetary policy influence through job offers. There are several

ways this problem could be solved, at least in part: 1.) Since banks are playing a repeated game

with each other, cooperation around a number of equilibria in the job-for-policy game may arise

through the threat of future defection. To consider just one of the many possibilities, banks may

focus on the equilibrium in which each bank offers a job-for-policy bargain to all of its former

employeers on their appointment to the central bank. 2.) Absent cooperation, the largest bank

may be big enough to offer some level of job-for-policy bargains, though perhaps with subopti-

mal m (i.e., the largest bank may be an Olsonian “privileged group”). 3.) Central bankers who

successfully carry out financial sector requests may be sending a signal to banks which makes

them more desirable as senior members.29 These are merely speculative explanations; for the

remainder of the paper, I focus on whether policy appears to be set in a way consistent with

financial sector influence through career paths, and leave the mechanism by which financial sector

coordination might occur to future work.

Democratic governments, on the other hand, may be unitary or not (in the case of coalition),

and do not play uninterrupted for all time. Instead, we might identify as long-run players each of

the major parties or enduring coalitions which alternate in office. For example, we might have GL

and GR which alternate in office with probability p at set election times. We might also suppose

the state of the economy influences election outcomes, so that p = f(π, y). Perhaps central

bankers bargain with the party that appointed them, and expect no payment of government

jobs when that party is out of office—unless, perhaps, the central banker switches to policy

preferred by the opposition party. Hence pre-electoral uncertainty reduces the expected return

on bargains with the government (e.g., E(r) = pr). As transition to a new government grows more

29For example, central bankers may reveal some privately desirable characteristic through pushing price stability(such as loyalty to the bank, belief in conservative monetary principals, or concern for investors). Alternatively,placing an esteemed central banker on a bank’s board may signal clients of the bank’s clout or resources, but theremay be less benefit if this former central banker is associated with failed or inflationary policies, or is not widelyregarded as having financial expertise.

34

likely, bargains with the government must include an ever larger risk premium, lest the banker

decide to side with the incoming government (as in Alt, 1991) or the financial sector. Since

accommodating monetary policy may be key to the government’s re-election election hopes, a

larger payment seems quite likely, but whether it will be enough to prevent the central banker

from hedging his bets depends on just how (un)likely re-election is.30

This model could be usefully extended in several ways. As the forgoing discussion suggests,

adding elections to the model may add important dynamic nuances to central bankers’ motiva-

tions. Second, one could include a richer model of the labor market, in which wage bargain-

ing coordination produces strategic interaction between wage-setters and the central bank, as in

Iversen (1999) or Franzese (1999,2001). Finally, one could add an explicit model of central banker

appointment, re-appointment, and dismissal. To do this, it would also be useful to treat the true

preferences of the central banker (χi) as uncertain from the government’s view. Governments

appointing central bankers will aggregate all known information about a candidate—including

career characteristics—which may convey information on the candidate’s preferences over policy

and over career inducements. As governments update their knowledge of χi, they may decide to

retain or dismiss CBi (at some cost), while the agent may set policy strategically to influence

this learning process. I leave these extensions for future work.

Characterizing equilibrium

Define the difference in player k’s utility across policies χ1 and χ2 as ∆k(χ1, χ2) = E(Wk(χ1) −

Wk(χ2)). Wk denotes the policy terms of k’s utility function subject to job-for-policy trades; for

central bankers, this amounts to Wi(χ1) = (1 + δi)[−(1 − χi)(y − yi)

2 − χiπ2], since the price

CBi pays for an outside offer in period 2i is not one but two periods of policy discretion. Yet for F

(and analogously G), WF(χF) = −(1−χF)(y−yF)2−χFπ

2, since the bargain only buys the present

period’s policy; another bargain with CBi+1 is needed to secure next period’s policy. Define the

reversion policy from F’s view, which obtains when F makes no offer, as χRF = E(χ∗|w = 0).

Using these definitions, we will characterize the equilibrium behavior of F; the equilibrium offer

of r by G is defined analogously.

The folk theorem for games with short- and long-run players suggests F will offer m ∈ [m,m].

The upper bound is the most F can credibly offer. The lower bound, m, reflects that to win the

auction, F must offer more (in CBi’s view) than either G or independent action. In words,

m = F’s added utilityfrom policy control, m = max

(CBi’s added utility from

independent policy , CBi’s added utility frombest alternative bargain

)

Specifically, CBi knows that regardless of how much F promises (in Period 1i) to offer in

period 2i, the most F will be willing to pay once 2i is reached is the one-period value of the

30Career incentives thus bear on the unanswered question of whether political-economic cycles can work throughmonetary policy when central banks are legally independent. Alesina and Roubini with Cohen (1997) show post-electoral partisan cycles in time-series cross section studies of growth, inflation, and unemployment in industrialdemocracies, but despite their expectations, these cycles seem unaffected by the presence of a legally independentcentral bank (i.e., interactions between CBI and Alesina and Roubini’s partisan cycle find no suppressing effect inmodels of inflation or unemployment). Drazen (2000) attributes this to a misapprehension of partisan cycles asinduced by monetary rather than fiscal policy, but it is also possible that partisan cycles in monetary policy workthrough career incentives, and thus circumvent legal guarantees of central bank autonomy.

35

difference in policies, in units of m, or

m =∆F(χF, χ

RF)

θF

(13)

To see this, it suffices to note that if F always reached and fulfilled bargains to pay m, it would

break-even versus the no-bargain solution in every period, and a bargain of ε > 0 less in each

period would ensure F gains from trade in every period. (Given δF < 1, F would be willing to

promise m/δF, but not willing to pay it, hence m is best deal the central banker will accept.)

The lower bound is more complicated, since winning bargains must beat all alternatives

facing CBi:

m = max

[(1 + δi)∆i(χi, χF)

δiθi

,max

(

0,τi∆G(χG, χF)

τGθi

)

+∆i(χG, χF)

δiθi

]

(14)

The first term is simply the utility CBi loses from setting F’s preferred policy; clearly any

acceptable bargain must fully compensate this loss. Winning bargains must also be better than

G’s best offer. The net change in CBi’s utility given G’s best offer is captured in the second term

in Equation 14, which consists of the value to CBi of r and the government’s preferred policy,

respectively.

Successful bargains between F and CBi depend on the existence of gains to trade; i.e., m > m.

But unlike most games between short- and long-run players, bargains can succeed even when the

principal cares very little for the future. This is because payment of m for today’s policy is

deferred to the next period. If it is not paid, the next period’s central banker (whether the

current agent or a replacement) will not accept an offer from F. Since the costs and benefits of

defection are deferred to the next period, any δF ∈ (0, 1] suffices, so long as there are gains to

trade in the next period as well.

Since CBi always maximizes his return over one play of the stage game, he always employs

a pure strategy of accepting the best offer made, setting policy accordingly. F and G, however,

may choose to play either pure or mixed strategies. When both principals play pure strategies,

they offer some m ∈ [m,m] and r ∈ [r, r]. CBi always accepts the best offer and implements

policy accordingly. In period 2, the winning bidder (if any) will make good on its promise, and

CBi will accept. If there was no winning bidder then CBi will remain as central banker, and

implement his own ideal χi.

The arrangement of the game allows F or G to play mixed strategies, but they will not do

so in any way that affects policy or career transitions. Suppose, given m > m, F offers to pay m

in Period 2 such that m ∈ [m,m] with probability q, and m = 0 with probability 1− q. This will

be the winning offer so long as F respects the constraint that E(m) = qm ≥ m. If the central

banker expects any less from F, then he will (rationally) punish F by refusing F’s bargain, but so

long as CBi expects at least m, he will implement F’s desired policy. Hence any mixed strategy

with qm < m is strictly dominated by the pure strategy of paying m = m every round, since the

latter assures F of unbroken policy influence, which is worth at least m per period. But failing

to pay m ≥ m in any particular round of the game leaves CBi in place for a second period, in

which he implements χi regardless. Since we have shown that F would rather pay at least m than

36

reach this outcome, we can conclude that under a mixed strategy, no realization of m will fall

below this threshold. (If we relax the assumption that central bankers receive job offers before

choosing to leave the central bank in period 2, mixed strategies that occassionally pay nothing

become viable, but the policy implications of the model remain the same.) In sum, mixed and

pure strategy equilibria with the same E(m) and E(r) may differ with respect to job offer quality

and perhaps the likelihood of CBi receiving a job at all, but not with respect to any policies

actually implemented.

37

Appendix B: Data Appendix

Sources of data

Central banker’s experience data were coded by the author and tabulated into experience scores

and other quantities using Escore. This software is/will be available from the author’s website.

I benefited greatly from the assistance of central banks themselves in gathering information

on past and present central bankers. I gratefully acknowledge the assistance of the Banca d’Italia,

Banco de Espana, Banco de Portugal, Bank of Canada, the Bank of England, Banque de France,

Banque Nationale de Belgique, the Central Bank of Ireland, Danmarks Nationalbank, Deutsche

Bundesbank, De Nederlandsche Bank, Norges Bank, Oesterreichische Nationalbank, the Reserve

Bank of Australia, the Reserve Bank of New Zealand, Suomen Pankki, Sveriges Riksbank, and the

Swiss National Bank. (Naturally, none of these institutions is responsible for any of the content

of this paper.) In many cases, the information possessed by the central banks was incomplete,

though a simple list of names was often an invaluable starting point. I then turned to a variety of

print, electronic, and web resources to fill in the details of central banker’s careers. The printed

sources are listed in the references.

Central bank independence data were taken from Cukierman, Webb, and Neyapti (1992),

whose dataset ranges over 1950–1989. These data were supplemented using Maxfield’s (1997)

coding of the same countries through 1994. Additional CBI indices taken from Grilli, Masciandaro

and Tabellini (1991) and Bade and Parkin (1982).

Inflation, GDP, imports, and exchange rate data were obtained from the IMF International

Financial Statistics on CD-ROM (Spring 2003 edition).

The left and right leanings of governments were coded based on the scheme of Alesina and

Roubini with Cohen (1997), extended and supplemented with exact election dates by the author,

using the Europe World Yearbook (various years).

Descriptive statistics

Mean Std. Dev. Min Max

FinExp 0.131 0.172 0.000 1.000GovExp 0.166 0.161 0.000 0.729FMExp 0.133 0.195 0.000 1.000CBExp 0.279 0.254 0.000 1.000CBCCmed −0.200 0.493 −1.000 1.000CBI-c 0.436 0.186 0.141 0.852CBI-3 0.383 0.157 0.090 0.690ln(Inflation) 1.417 0.940 −3.664 3.466Inflation 5.976 5.130 0.026 32.012Imports/GDP 0.290 0.133 0.064 0.756

Table 3: Summary statistics for data used in Table 1 regressions Data cover all twenty countriesover 1973–1997, and non-ECB members from 1973–2000. Cases of deflation are omitted.

38

finexp govexp fmexp cbexp ecoexp busexpgovexp −0.14fmexp −0.33 −0.08cbexp −0.18 −0.34 −0.39ecoexp 0.04 −0.14 −0.20 −0.23busexp −0.10 −0.10 −0.08 −0.31 −0.02cbi-3 0.17 −0.08 −0.15 0.10 −0.01 −0.13

Table 4: Correlations across career types and institutions

Section 3.3 Probit Regressions

FinJobi,t+ GovJobi,t+

FinExpi,t− 0.98 −0.27

(0.25) (0.21)GovExpi,t− −0.39 0.43

(0.31) (0.20)FMExpi,t− 0.41 −0.07

(0.29) (0.23)CBExpi,t− 0.67 0.56

(0.21) (0.16)EcoExpi,t− −0.12 0.04

(0.32) (0.21)Agei,t −0.02 −0.03

(0.01) (0.00)Constant −0.19 0.82

(0.34) (0.29)

log-likelihood −0.42 −0.64N 928 928

Table 5: Post central bank appointments as a function of pre-central bank careers. Data drawnfrom twenty countries over the period 1950–2000. The unit of analysis is each appointment tothe monetary policy authority, indexed i; t refers to the time the appointment ends. Entries areprobit parameters and standard errors. See text and Figure 2 for interpretation.

Robustness of Section 3.4 Inflation Regressions

To test the robustness of the inflation findings, I re-estimated the model using a variety of

alternative specifications, a robust estimator, and varied rules for aggregating across members of

a central bank. Though they vary in precision, the results almost uniformly support the substance

of Section 3.4.

Figure 9 provides a concise summary of alternative specifications. The first row reiterates

the findings presented in the text: after five years, inflation falls considerably if FinExp, FMExp,

or CBCC are raised, and it rises following appointment of central bankers with higher GovExp

39

−3 −1 1 −1 1 3 −3 −1 1 −1 1 3 −3 −1 1

FinExp GovExp FMExp CBExp CBCC

Baseline

Add πworld

Add %Left-Apptees

Use Cukierman CBI

Omit Imports/GDP

Robust Estimation

Change in inflation, five years after +1 s.d. in . . .

Figure 9: The career-inflation link under alternative specifications. Each plot shows the five-yearfirst difference in inflation resulting from a one standard deviation in a career variable (FinExp,GovExp, etc.), given the specification noted at left. See text for a description of the baselinemodel and alternatives. Circles and squares indicate point estimates of the first difference, andhorizontal lines 90 percent confidence intervals. The results marked with a square are the sameas those used in Figure 3. The shaded areas highlight the range of point estimates across allalternatives.

or CBExp. These findings reflect a baseline least squares model that controls for CBI (using an

average of three indices), imports as a fraction of GDP, and two lags of the dependent variable.

The succeeding five rows show that these results hold up to changes in this specification. It makes

little difference if we add controls for contemporaneous inflation levels in the (other) G7 countries

or for the fraction of central bankers appointed by a left-wing government. Likewise, we could

substitute Cukierman’s CBI index for the three index average, drop imports from the model,

or re-estimate the baseline model using robust estimation (see note 22), and in every case, the

effects of careers would remain correctly signed, and in the large majority of cases, statistically

significant. And no matter the specification, the combined career index, CBCC, remains strongly

and significantly associated with lower inflation. (The only added control that has a non-zero

effect is world inflation, which is positively related to domestic inflation).

A similar pattern emerges if we keep the baseline model and vary the construction of the

key explanatory variables. Most central banks have multiple, legally established monetary policy

makers, but only one policy outcome. Aggregating across these policymakers is a key challenge.

Taking the (tenure-weighted) means of all de jure policymakers seems like a good first approx-

imation, though the median member is a better bet when a single dimension (like CBCC) is

established. But as Figure 10 shows, it makes little difference if we take means or medians of

the career components or CBCC. In constructing the CBCC index we have still more options.

Rather than weighting FinExp, GovExp, FMExp, and CBExp equally, as in the main text, we

could give only half-weight to the more ambiguous categories (FMExp and CBExp), or we could

use the point estimates of the coefficients on each category as its weight. The middle rows of

Figure 10 show that these alternative make little difference in the aggregate.

40

−3 −1 1 −1 1 3 −3 −1 1 −1 1 3 −3 −1 1

FinExp GovExp FMExp CBExp CBCC

Means

Medians

FM & CB 12-weight

Estimated weights

Governors only

Gov only subsample

Change in inflation, five years after +1 s.d. in . . .

Figure 10: The career-inflation link under alternative aggregation rules. These plots can beinterpreted as Figure 9, save that the rows vary the rule for aggregating across central bankmembers.

A final approach follows the common practice of ignoring all central bankers but the governor

of the bank. Focusing on governors is an ironic practice, likely adopted for convenience only, since

its proponents often insist on the importance of legal institutions while ignoring the group of offi-

cials to whom those laws grant discretion. The procedure eliminates the problem of aggregation

by pretending it does not exist, and is, at best, a half step towards recognizing the importance of

central bank officials in setting monetary policy. Re-running the model using data on only gov-

ernors produces correct signs in four of five cases, but none of the results are significant, and the

fit of the model is inferior to any other considered here. A final refinement restricts the sample

of countries to those where the governor has sole legal authority. This discards three-quarters of

the data, but produces correct signs in all five cases, and an “almost significant” effect of CBCC.

The intuitive conclusion to be drawn is that those officials who are legally deemed to matter are

the right ones to study, and restricting attention to governors, though expedient, is unjustified.

41

References

Sources of data

Bank of Canada. Annual Report of the Governor to the Minister of Finance. Ottawa. Variousyears. (In recent years, Annual Report).

Banco de Portugal. Relatorio do Conselho de Administracao. Various years. Lisbon: Tipographiado Banco de Portugal.

Canadian Who’s Who. Toronto: University of Toronto Press. Various years.

Encyclopaedia of Austria. Archived online at aeiou.at.

Federal Reserve Bulletin. Various issues.

Hennessy, Elizabeth. 1999. Flemings Who’s Who in Central Banking. London: Central BankingPublications, Ltd.

Hennessy, Elizabeth and Yvonne Messenger. 2001. Who’s Who in Central Banking, 2002. Lon-don: Central Banking Publications, Ltd.

Janssens, Valery. 1997. De beheerders van ons geld: Negentien gouverneurs van de NationaleBank van Belgie. Lannoo.

The Japan Biographical Encyclopedia & Who’s Who, 1964–65. 3rd ed. Rengo Press: Tokyo.

The Japan Who’s Who. 1950-51 Edition. Tokyo News Service: Tokyo.

Katz, Bernard S. 1992. Biographical dictionary of the Board of Governors of the Federal Reserve.New York : Greenwood Press.

Morris, Irwin. 2000. Congress, the President, and the Federal Reserve. Ann Arbor, MI: TheUniversity of Michigan Press. [Contains a detailed list of all Governors of the Federal Reserve,with dates in office.]

O Grande Livro dos Portugueses. 1991. Lisbon: Circulo de Leitores.

O Seculo XX Portugues. 2000. Lisbon: Texto Editora.

Pluym, Walter. 1995. L’Hotel du Gouverneur de la Banque Nationale de Belgique. Pandora.[Contains biographies of most Belgian central bank governors.]

Pringle, Robert. 1999. The Morgan Stanley Dean Witter Central Bank Directory. London:Central Banking Publications.

Roselli, Alessandro. 1987. ‘Notes on the Contributors.’ Money and the Economy: CentralBankers’ Views. Ed. Pierluigi Ciocca. New York: St. Martin’s Press. 313–324.

Qui Est Qui en France. Paris: Editions Jacques Lafitte. Various years.

Quien es Quien en Espana. Various years. Madrid:

Vem Ar Det: Svensk biografisk handbok. Various years. P.A. Norstedt & Soner: Stockholm.

von Furstenberg, George M. and Michael K. Ulan. 1998. Learning from the World’s Best Cen-tral Bankers: Principles and Policies for Subduing Inflation. Boston: Kluwer AcademicPublishers. [Includes selected biographical sketches.]

42

Wer Ist Wer in Osterreich. 1953. Vienna: Verlag “Wer Ist Wer in Osterreich”.

Who’s Who. London : A. C. Black. Various years. [United Kingdom]

Who’s Who in Canada. Toronto: International Press Limited. Various years.

Who Was Who. London : A. C. Black. Various years. [United Kingdom]

Who’s Who in America. Chicago: Marquis Who’s Who. Various years.

Who’s Who in Austria. Various years and publishers; most recently Who’s Who AG Zurich.

Who’s Who in Japan, 1984-85. 1984. International Culture Institute: Hong Kong.

Who’s Who in New Zealand. Various years. Reed: Wellington, New Zealand.

Who’s Who in Spain. Various years and publishers; most recently Who’s Who in Spain AG.

Who’s Who in Switzerland. Various years and publishers; most recently Orell Fussli Verlag,Zurich.

Who Was Who in America. Chicago: Marquis Who’s Who. Various years.

Works cited

Aberbach, Joel D., Robert D. Putnam, and Bert A. Rockman. 1981. Bureaucrats and Politiciansin Western Democracies. Cambridge, MA: Harvard University Press.

Alesina, Alberto and Nouriel Roubini with Gerald Cohen. 1997. Political Cycles and the Macroe-conomy. Cambridge, MA: MIT Press.

Alesina, Alberto and Lawrence H. Summers. 1993. ‘Central bank independence and macroeco-nomic performance: Some comparative evidence.’ Journal of Money, Credit, and Banking25.2 (May) 151–162.

Alt, James E. 1991. ‘Leaning into the wind or ducking out of the storm? U.S. monetary policyin the 1980s.’ Politics and Economics in the 1980s. Ed. Alberto Alesina and G. Carliner.Chicago: University of Chicago Press.

Alvarez de Cienfuegos, Ignacio Molina. 1999. ‘Spain: Still the primacy of corporatism?’ Bureau-cratic Elites in Western European States. Ed. Edward C. Page and Vincent Wright. Oxford:Oxford University Press. 32–54.

Bade, R., and M. Parkin. 1982. ’Central bank laws and monetary policy.’ Department ofEconomics, University of Western Ontario.

Barro, Robert J., and David B. Gordon. 1983. ‘Rules, discretion and reputation in a model ofmonetary policy.’ Journal of Monetary Economics 12: 101–121.

Beck, Nathaniel and Jonathan Katz. 1995. “What to do (and not to do) with time-series cross-section data.” American Political Science Review 89.3: 634–647.

Belden, Susan. 1989. “Policy preferences of FOMC members as revealed by dissenting votes.”Journal of Money, Credit, and Banking. (Nov.) 21. 432–41.

Bendor, J, A. Glazer, and T. Hammond. 2001. “Theories of delegation.” Annual Reviews ofPolitical Science 4:235–269.

43

Berger, Helge, Jacob de Haan, and Sylvester C.W. Eijffinger. 2001. “Central bank independence:An update of theory and evidence.” Review of Economic Sturveys 15.1: 3–40.

Beyer, Janice M. et al. 1997. ‘The selective perception of managers revisited.’ The Academy ofManagement Journal 40.3 (Jun.) 716–737.

Blinder, Alan. 1997. “What central bankers could learn from academics—and vice versa.”Journal of Economic Perspectives. 3–19.

Blondel, Jean, and Jean-Louis Thiebault, ed. 1991. The Profession of Government Minister inWestern Europe. New York: St. Martin’s Press.

Brans, Marleen, and Annie Hondeghem. 1999. ‘The senior civil service in Belgium.’ BureaucraticElites in Western European States. Ed. Edward C. Page and Vincent Wright. Oxford:Oxford University Press. 121–146.

Buiter, Willem. 1999. “Alice in Euroland.” Journal of Common Market Studies 37.2: 181–209.

Burden, Barry C. 2000. ‘Representation versus self-interest in U.S. politics: The cases of tobaccoregulation and school choice.’ Working paper, Harvard University.

Campillo, Marta, and Jeffrey A. Miron. 1997. ‘Why does inflation differ across countries?’Reducing Inflation. Ed. Christina D. Romer and David H. Romer. Chicago: The Universityof Chicago Press: 335–357.

Carey, John. 1994. “Political shirking and the last term problem: Evidence for a party-administered pension system.” Public Choice 81: 1-22.

Carey, John. 1996. Term limits and legislative representation. Cambridge: Cambridge UniversityPress.

processes in panel data: Theory, simulations, and examples.” April. Working paper. De-partment of Economics, University of Oklahoma.

Chang, Kelly H. 1998. “Their people equal their policy: The President, Congress, and appoint-ments to the Federal Reserve.” Ph.D. Diss. Stanford University.

Chappell, Henry, Thomas Havrilesky, and Rob Roy McGregor. 1993. ‘Partisan monetary policies:Presidential influence through the power of appointment.’ Quarterly Journal of Economics108:1 (Feb.) 185–218.

Cukierman, Alex. 2002. “Are contemporary central banks transparent about economic modelsand objectives and what difference does it make?” Federal Reserve Bank of St. Louis ReviewJuly/August.

Cukierman, Alex. 1992. Central Bank Strategy, Credibility, and Independence: Theory andEvidence. Cambridge, MA: MIT Press.

Cukierman, Alex, and Francesco Lippi. 1999. “Central bank independence, centralization of wagebargaining, inflation and unemployment: theory and some evidence.” European EconomicReview 43.7 (June) 1395-1434.

Cukierman, Alex, Steven B. Webb, and Bilin Neyapti. 1992. ‘Measuring the independence ofcentral banks and its effect on policy outcomes.’ World Bank Economic Review 6: 353–398.

Dargie, Charlotte, and Rachel Locke. 1999. ‘The British civil service.’ Bureaucratic Elitesin Western European States. Ed. Edward C. Page and Vincent Wright. Oxford: OxfordUniversity Press. 178–204.

44

Dearborn, DeWitt C., and Herbert A. Simon. 1958. ‘Selective perception: A note on thedepartmental identifications of executives.’ Sociometry 21.2 (Jun.) 140–144.

Debelle, G. and S. Fischer. 1994. ‘How independent should a central bank be?’ Goals, Guidelines,and Constraints Facing Monetary Policymakers. Ed. J. Fuhrer. Federal Reserve Bank ofBoston Conference Series, no. 38. Boston: Federal Reserve Bank of Boston.

de Winter, Lieven. 1991. “Parliamentary and party pathways to the cabinet.” The Professionof Government Minister in Western Europe. Ed. Blondel, Jean, and Jean-Louis Thiebault.New York: St. Martin’s Press. 44–69.

Dowding, Keith. 1995. The Civil Service. London: Routledge.

Drazen, Allan. 2000. “The political business cycle after 25 years.” NBER MacroeconomicsAnnual Cambridge, MA: MIT Press.

Eijffinger, Sylvester C.W., and Petra M. Geraats. 2002. ‘How transparent are central banks?’May. Unpublished paper. University of Cambridge.

Franzese, Robert J. 2000. Macroeconomic Policies of Developed Democracies. Cambridge: Cam-bridge University Press.

—. 1999. ‘Partially independent central banks, politically responsive governments, and inflation.’American Journal of Political Science 43.3

Fudenberg, Drew, David M. Kreps, and Eric S. Maskin. 1990. “Repeated games with long-runand short-run players.” Review of Economic Studies. 57.4: 555–573.

Gibbons, Robert and Kevin J. Murphy. 1992. “Optimal incentive contracts in the presence ofcareer concerns: Theory and evidence.” Journal of Political Economy 100.3: 468–505.

Goodman, John B. 1992. Monetary Sovereignty: The Politics of Central Banking in WesternEurope. Ithaca, NY: Cornell University Press.

Grilli, Vittorio, Donato Masciandaro, and Guido Tabellini. 1991. ‘Political and monetary insti-tutions and public financial policies in the industrial countries.’ Economic Policy October341–392.

Gunz, Hugh P., and Michael Jalland. 1996. “Managerial careers and business strategies.”Academy of Management Review 21.3: 718–756.

Hall, Peter, and Robert Franzese. 1998. “Mixed signals: Central bank independence, co-ordinated wage bargaining, and European Monetary Union.” International Organization52.3 (Summer) 505–536.

Hambrick, Donald C., and Phyllis A. Mason. 1984. “Upper echelons: The organization as areflection of its top managers.” Academy of Management Review 9.2: 193–206.

Havrilesky, Thomas, and John A. Gildea. 1991. ‘The policy preferences of FOMC members asrevealed by dissenting votes: Comment.’ Journal of Money, Credit and Banking 23:1 (Feb.)130–138.

Heclo, Hugh. 1988. “The in-and-outer system: A critical assessment.” Political Science Quarterly103.1: 37–56.

Hitt, Michael A., and Beverly B. Tyler. 1991. “Strategic decision models: Integrating differentperspectives.” Strategic Management Journal 12.5: 327–351.

Holmstrom, Bengt. 1982. “Managerial incentive problems: A dynamic perspective.” in Essaysin Economics and Management in Honor of Lars Wahlbeck. Helsinki: Swedish School ofEconomics.

45

Huber, John D. and Charles R. Shipan. 2002. Deliberate Discretion? The Institutional Founda-tions of Bureaucratic Autonomy. Cambridge: Cambridge University Press.

Issing, Otmar. 1999. “The Eurosystem: Transparent and accountable, or ’Willem in Euroland’?”Journal of Common Market Studies 37.3: 503–519.

Iversen, Torben. 1999. Contested Economic Institutions: The Politics of Macroeconomics andWage Bargaining in Advanced Democracies. Cambridge: Cambridge University Press.

—. 1998a. “Wage bargaining, hard money, and economic performance: Theory and evidence fororganized market economies.” British Journal of Political Science January. 31-62.

—. 1998b. ’Wage bargaining and central bank independence.’ International Organization 52:3Summer. 469-504.

Jensen, Hanne Nexø, and Tim Knudson. 1999. “Senior officials in the Danish central admin-istration: From bureaucrats to policy professionals and managers.” Bureaucratic Elites inWestern European States. Ed. Edward C. Page and Vincent Wright. Oxford: OxfordUniversity Press. 229–248.

Judd, John P. and Glenn D. Rudebusch. 1998. “Taylor’s Rule and the Fed: 1970–1997.” FederalReserve Bank of San Francisco Economic Review 3: 1–16.

Kapstein, Ethan Barnaby. 1992. “Between power and purpose: Central bankers and the politicsof regulatory convergence.” International Organization 46.1: 265–287.

Kim, Paul S. 1988. Japan’s Civil Service System. New York: Greenwood Press.

King, Gary, Michael Tomz, and Jason Wittenberg. 2000. ‘Making the Most of Statistical Analy-ses: Interpretation and Presentation.’ American Journal of Political Science 44 (2): 341-355.

Koh, B.C. 1989. Japan’s Administrative Elite. Berkeley: University of California Press.

Kydland, Finn E. and Edward C. Prescott. 1977. ‘Rules rather than discretion: The inconsis-tency of optimal plans.’ Journal of Political Economy 85 (June) 473–491.

Lane, Philip R. 1997. “Inflation in open economies.” Journal of International Economics 327–347.

Lohmann, Susanne. 1992. ‘Optimal commitment in monetary policy: Credibility versus flexibil-ity.’ American Economic Review 82.1 (March) 273–286.

Maxfield, Sylvia. 1997. Gatekeepers of Growth: The International Political Economy of CentralBanking in Developing Countries. Princeton: Princeton University Press.

Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven: Yale UniversityPress.

Mayntz, R. and H. U. Derlien. 1989. ‘Party patronage and politicization of the West Germanadministrative elite 1970–87—Toward Hybridization?’ Governance 2.4: 384–404.

McCallum, Bennett T. 1999. “Recent developments in the analysis of monetary policy rules.”Homer Jones Memorial Lecture, University of Missouri, St. Louis. March 11.

McCallum, Bennett T. 1995. ‘Two fallacies concerning central bank independence.’ AmericanEconomic Review 85.2: 207–211.

McCubbins, Mathew D., and Thomas Schwartz. 1984. “Congressional Oversight Overlooked:Police Patrols versus Fire Alarms.” American Journal of Political Science 28.1: 165–179.

46

Melone, Nancy Paule. 1994. ‘Reasoning in the executive suite: The influence of role/experience-based expertise on decision processes of corporate executives.’ Organization Science 5.3(Aug.) 438–455

Morris, Irwin. 2000. Congress, the President, and the Federal Reserve. Ann Arbor, MI: TheUniversity of Michigan Press.

Naka, Kura. 1980. Kokkai giin no kosei to henka [The Change and Structure of Diet Membership].Tokyo: K. K. Seiji joho Sentaa.

Peters, B. Guy. 1997. ‘Bureaucrats and political appointees in European democracies: Who’swho and does it make any difference?’ Modern Systems of Government: Exploring the Rolesof Bureaucrats and Politicians Ed. Ali Farazmand. Thousand Oaks, CA: Sage. 232–254.

Pierre, Jon, and Peter Ehn. 1999. “The welfare state managers: Senior civil servants in Sweden.”Bureaucratic Elites in Western European States. Ed. Edward C. Page and Vincent Wright.Oxford: Oxford University Press. 249–265.

Posen, Adam. 1995. ‘Declarations are not enough: Financial sector sources of central bankindependence.’ NBER Macroeconomics Annual Cambridge, MA: MIT Press.

Putnam, Robert D. 1976. The Comparative Study of Political Elites. Englewood Cliffs, NJ:Prentice-Hall, Inc.

Ramseyer, J. Mark, and Frances McCall Rosenbluth. 1993. Japan’s Political Marketplace. Cam-bridge, MA: Harvard University Press.

Rogoff, Kenneth. 1985. ‘The optimal degree of commitment to an intermediate monetary target.’Quarterly Journal of Economics 11: 1169–1188.

Romer, David. 1993. “Openness and inflation: Theory and evidence.” Quarterly Journal ofEconomics 869–903.

Rosman, Andrew, Michael Lubatkin, and Hugh O’Neill. 1994. “Rigidity in decision behaviors:A within-subject test of information acquisition using strategic and financial informationalcue.” Academy of Management Journal 37.4: 1017–1033.

Rothenberg, Lawrence S. and Mitchell S. Sanders. 2000. “Severing the electoral connection:Shirking in the contemporary Congress.” American Journal of Political Science 44(2): 316-325.

Rouban, Luc. 1999. ‘The senior civil service in France.’ Bureaucratic Elites in Western EuropeanStates. Ed. Edward C. Page and Vincent Wright. Oxford University Press. 65–89.

Santoni, G. J. 1986. “The effects of inflation on commercial banks.” Federal Reserve Bank of St.Louis Review March.

Scheve, Kenneth. 2002a. ‘Nominal Asset Ownership and Individual Macroeconomic Priorities.’Yale University. Working Paper.

—. 2002b. ‘Public Inflation Aversion and the Political Economy of Macroeconomic Policymak-ing.’ Presented at the Annual Meetings of the Midwest Political Science Association.

Schlesinger, Joseph. 1966. Ambition in Politics: Political Careers in the United States. Chicago:Rand McNally.

Schneider, Ben Ross. 1993. “The career connection: A comparative analysis of bureaucraticpreferences and insulation.” Comparative Politics 25.3: 331–350.

47

Segal, Jeffrey A. and Harold J. Spaeth. 1993. The Supreme Court and the Attitudinal Model.Cambridge: Cambridge University Press.

Siklos, Pierre L. 2002. The Changing Face of Central Banking: Evolutionary Trends Since WorldWar II Cambridge: Cambridge University Press.

Stiglitz, Joseph. 2002. Globalization and its Discontents. New York: Norton.

Svensson, L.E.O. 1997. “Inflation forecast targeting: Implementing and monitoring inflationtargets.” European Economic Review 41:1111-1146.

Tate, C. Neal and Roger Handberg. 1991. “Time binding and theory building in personalattribute models of Supreme Court voting behavior, 1916–88.” American Journal of PoliticalScience 35.2:460–480.

Tirole, Jean. 1994. “The internal organization of government.” Oxford Economic Papers 46:1–29.

Thiebault, Jean-Louis. 1991. “The social background of Western European ministers.” TheProfession of Government Minister in Western Europe. Ed. Blondel, Jean, and Jean-LouisThiebault. New York: St. Martin’s Press. 19–30.

van der Meer, Fritz M. and Jos C.S. Raadschelders. 1999. “The senior civil service in theNetherlands: A quest for unity.” Bureaucratic Elites in Western European States. Ed.Edward C. Page and Vincent Wright. Oxford: Oxford University Press. 205–228.

van Maanen, John, and Edgar H. Schein. 1979. ‘Towards a theory of organizational socialization.’Research in Organizational Behavior. 1: 209–264.

Walsh, Carl E. 1998. Monetary Theory and Policy. Cambridge, M.A.: MIT Press.

Walsh, James P. 1988. ‘Selectivity and selective perception: An investigation of managers’ beliefstructures and information processing.’ The Academy of Management Journal 31.4 (Dec.)873–896.

Weingast, Barry, and Mark Moran. 1983. “Bureaucratic discretion or Congressional control?Regulatory policymaking by the Federal Trade Commission.” Journal of Political Economy91.5: 765–800.

Western, Bruce. 1995. “Concepts and suggestions for robust regression analysis.” AmericanJournal of Political Science 39.3: 786–817.

Wooley, John T. 1984. Monetary Politics: The Federal Reserve and the Politics of MonetaryPolicy. Cambridge: Cambridge University Press. 1984.

48