do markets care about central bank governor changes? evidence from emerging markets
TRANSCRIPT
CHRISTOPH MOSER
AXEL DREHER
Do Markets Care about Central Bank Governor
Changes? Evidence from Emerging Markets
Based on a new daily data set for 20 emerging markets over the period1992–2006, we examine the reactions of foreign exchange markets, do-mestic stock markets, and sovereign bond spreads to central bank governorchanges. We find that the replacement of a central bank governor negativelyaffects financial markets on the announcement day, which is in line withthe hypothesis that newly appointed central bank governors suffer from asystematic credibility problem at the beginning of their tenure. We also findsome evidence that changes in perceived central bank independence affectmarkets.
JEL codes: F30, F34, G14, H63Keywords: central bank governor turnover, monetary policy, emerging
markets, risk premium.
“After arguing behind the scenes with his central bank governor over the directionof interest rates, Prime Minister Thaksin Shinawatra of Thailand . . . dismissed thebanker . . . brought a sharp reaction of financial markets, where it cast doubts over thepolitical independence of the Thai central bank.” (The New York Times [“Chief of ThaiBank Is Dismissed Over Rates,” May 30, 2001, Source: Proquest])
We thank two anonymous referees and the editors, Robert DeYoung and Deborah Lucas, for helpfulcomments. Furthermore, we are grateful to Christian Bjørnskov, Ricardo Caballero, Christian Conrad,Etienne Farvaque, Martin Gassebner, Ashok Kaul, Silke Rath, Jan-Egbert Sturm, Dieter Urban, participantsat the First BBQ Conference, the Verein fur Socialpolitik, the KOF Research Seminar, the Verein furSocialpolitik: Research Committee Development Economics, the Central Banking Conference “DoesCentral Bank Independence Still Matter?” at the Bocconi University 2007, the Brown Bag Seminarat the University of Mainz, the Annual Meeting of the Austrian Economics Association (NOeG), theMacroeconomics Research Meeting 2007, and in particular Helge Berger for suggestions and discussion.We thank Simon Holzhammer, Stefan Keitel, Michaela Lischer, Elisabeth Munch, Nadja Panzer, andChristoph Woodli for excellent research assistance; Jan Schopen for help in compiling the stock marketand foreign exchange market data; and Hendrik van Broekhuizen for excellent proofreading.
CHRISTOPH MOSER is at the ETH Zurich, Swiss Federal Institute of Technology, KOF SwissEconomic Institute (E-mail: [email protected]). AXEL DREHER is a Faculty of EconomicSciences, Georg-August University Goettingen, KOF Swiss Economic Institute, Switzerland,and IZA and CESifo, Germany (E-mail: [email protected]).
Received October 26, 2007; and accepted in revised form July 12, 2010.
Journal of Money, Credit and Banking, Vol. 42, No. 8 (December 2010)C© 2010 The Ohio State University
1590 : MONEY, CREDIT AND BANKING
THE DYNAMIC INCONSISTENCY of low-inflation monetary policycan be overcome by delegating monetary policy to independent and conservativecentral bankers (Rogoff 1985). While many countries have recently granted theircentral banks legal independence, the experience of some countries suggests limitedactual independence, with the head of the central bank frequently being replaced atshort notice and outside the legal schedule. How do financial markets react to changesof central bank governors?1 We expect them to react to such changes if changes conveynew information about future monetary policy. Economic theory suggests that theinflationary bias is determined by the degree of central bank independence and thedegree of the bank’s conservatism.2 Hence, if market participants’ perceptions changewith respect to one of these two dimensions, asset prices should change to the extentof their sensitivity to inflation. There are two transmission channels. First, when thegovernment interferes in the replacement procedure, irregular turnovers are likelyto affect markets’ perceptions about the central bank’s independence. Second, if theperceived inflation aversion of the new central bank governor differs from that of thepredecessor’s, this will alter expected financial market returns.
The literature has shown that the identity of central bank council members hasimportant bearing on economic outcomes. Drawing on a sample of 15 industrializedcountries, Kuttner and Posen (2010) conclude that markets do care about who chairsthe central bank. Central bank governor changes apparently convey new informationabout the future monetary policy, thereby affecting exchange rates and domesticbond yields. Kuttner and Posen (2010) do not, however, find evidence of a genericcredibility problem, that is, a systematic (at least) transitory increase in inflationexpectations at the beginning of a central bank governor’s tenure. This is hardlysurprising, since central bank governor turnovers are mostly predictable and highlydeveloped institutions are likely to reduce the individual governor’s influence inadvanced countries. By contrast, replacing the central bank governor is among themost sensitive decisions for emerging market governments, as these policymakersplay a crucial role in communicating with international markets (Santiso 2003).Credibility is more likely to be an issue here. Surprisingly, whether and to whatextent central bank governor changes in transition countries affect financial marketshas so far not been investigated.
This paper examines the impact of central bank governor changes on domestic andinternational financial markets in emerging economies. Based on a new daily dataset on 20 emerging markets over the period 1992–2006, we study whether foreignexchange rates, domestic stock market indices, and sovereign bond spreads react to theannouncement of a change at the helm of a central bank and whether new governorshave systematic credibility problems. Moreover, we investigate whether changes in
1. We call the heads of the central bank “governors” independent of whether their actual job title is“governor,” “director,” or “president.”
2. By “conservatism” we mean that the central bank has a stronger bias against inflation than thegovernment.
CHRISTOPH MOSER AND AXEL DREHER : 1591
perceived central bank independence or changes in the perceived conservatism of thecentral banker affect markets.
To foreshadow our main results, we find that the replacement of a central bankgovernor negatively affects financial markets on the announcement day. Overall, ourresults are in line with the hypothesis that incoming governors in emerging marketsface a credibility problem. Furthermore, we find evidence that the negative effecton financial markets is primarily driven by central bank governor changes that occurbefore the officially scheduled end of tenure, indicating reactions to perceived changesin central bank independence.
The remainder of the paper is structured as follows. Section 1 derives our hy-potheses, while Section 2 describes the data set. Section 3 discusses the methodologyused, derives the empirical specifications, and describes the results. The final sectionconcludes.
1. HYPOTHESES
Evidence on the impact of who is in charge of economic policies on economicoutcomes is scarce. It is only very recently that studies have started analyzing thisissue. Among them, Jones and Olken (2005) demonstrate that national leaders matterfor economic growth. Dreher et al. (2008) show that the educational and professionalbackground of a head of government matters for the implementation of reforms. Fora sample of Latin American countries, Moser (2006) finds that changing the financeminister had the effect of increasing sovereign bond spreads due to a perceivedworsening of the fiscal policy stance.
Turning to central banks, Gohlmann and Vaubel (2007) provide recent empiricalevidence that the educational qualifications and professions of the central bank’sgoverning council members matter for its effectiveness when attempting to controlinflation. Kuttner and Posen (2010) find that changing a central bank governor conveyssignals about the future course of monetary policy, thereby affecting exchange ratesand financial market returns. However, the direction of these effects subsequent tosuch announcement is not obvious a priori. What are the channels by which turnoversaffect markets and are market reactions positive or, instead, negative? Policymakerscannot credibly commit themselves to low-inflation policy (Kydland and Prescott1977). One approach to overcome this time-inconsistency problem is to establishreputation (Backus and Driffill 1985). While the public may not know central bankers’preferences, policymakers’ behavior conveys information about their characteristics,and the public will adapt their expectations about inflation accordingly. Cukiermanand Meltzer (1986) emphasize the importance of uncertainty about the underlyingpreferences of governors. A less conservative governor (dovish) has an incentive tomimic the behavior of the more conservative one (hawkish) for a while, but sooner orlater, it becomes optimal to behave opportunistically. Uncertainty can be expected tobe highest at the beginning of a governor’s tenure. In a similar vein, Schaumburg and
1592 : MONEY, CREDIT AND BANKING
Tambalotti (2007) and Kara (2007) state that newly appointed central bank governorssuffer from a systematic credibility problem: while incumbent governors can committo policies during their own administration, their successors might deviate and pursuediscretionary policies. We therefore hypothesize the following:
Hypothesis 1. Investors react negatively to central bank governor changes due to asystematic credibility problem at the start of new governors’ tenure.
One way to remedy the inconsistency problem is proposed by Rogoff (1985) andentails the delegation of monetary policy to a conservative central bank. Eijffinger andHoeberichts (1998) and Berger, de Haan, and Eijffinger (2001) develop an argumentin the spirit of Rogoff. The government seeks to minimize the following loss function,representing the preferences of society:
LGov = 1
2π2
t + χ
2(yt − y∗
t )2,
where π t is the rate of inflation at day t, yt is output, y∗t denotes desired output, and
χ is the government’s weight on output stabilization (χ > 0). The loss function ofthe central banker is expected to differ from that of the government in one importantaspect:
LC B = 1 + ε
2π2
t + χ
2(yt − y∗
t )2,
where ε denotes the additional inflation aversion of the central bank governor, that is,his conservatism. Furthermore, Eijffinger and Hoeberichts (1998) argue that centralbankers’ preferences only matter to the extent to which they can pursue monetarypolicy without (much) government interference. This can be captured in the followingway:
Mt = γ LC B + (1 − γ ) LGov,
where γ and Mt denote the degree of central bank independence and monetarypolicy, respectively. Assuming that output is determined by a simplified Lucas supplyfunction and assuming rational expectations, inflation turns out to be:
πt = χy∗t − χ
χ + 1μt and πt = χ
1 + γ εy∗
t − χ
1 + γ ε + χμt .
The equation on the left (right) represents the inflation outcome without (with)delegation of monetary policy.3 Comparing these two outcomes, it becomes clearthat it is the product of central bank independence and conservatism that mattersfor monetary policy. We label this product “effective conservatism” of monetary
3. μt denotes a random shock with zero mean and variance σμ and is part of the simplified Lucassupply function.
CHRISTOPH MOSER AND AXEL DREHER : 1593
policy. For positive values of γ and ε, any increase of either the degree of centralbank independence or the central bank governor’s conservatism will decrease theinflationary bias, ceteris paribus. The same level of monetary policy can be achievedthrough various combinations of the two dimensions.
From the above, changing the central bank governor might affect financial marketexpectations along these two dimensions. First, changes at the head of the central bankmay carry signals about the future stance of the incumbent government on the centralbank’s independence. If central bank governors’ resignations are politically motivatedand/or incoming governors lack political independence, financial markets are likelyto react negatively. Second, a newly appointed central bank governor’s attitude towardinflation might deviate from that of her predecessor. If the new governor is perceivedto be more conservative than the old one, market reactions should be positive. Fromthis, we derive the following two hypotheses:
Hypothesis 2. Investors react to central bank governor changes due to changes inperceived conservatism.
Hypothesis 3. Investors react to central bank governor changes due to changes inperceived central bank independence.
We employ various characteristics of outgoing and newly appointed governors andtheir nominating governments to distinguish between these two channels.4,5
Clearly, governor changes might affect exchange markets, stock markets, and bondmarkets to different degrees and even in different directions. We address each ofthese markets in turn. We implicitly assume that the semistrong form of the efficientmarket hypothesis holds. Under this hypothesis, security prices are assumed to reflectall public information and to adjust swiftly to the arrival of such information. In thisvein, if markets fully anticipate a change of governor, prices will not react on theofficial announcement day.
1.1 Foreign Exchange Markets
Following Kuttner and Posen (2010), our starting point for analyzing the impact ofcentral bank governor changes on the exchange rate is uncovered interest rate parity:
Et�et+1 = i∗t − it ,
4. In the working paper version of this paper, we also investigated whether markets react to personalcharacteristics of the incoming central bank governor. We test whether incoming governors educated inthe United States or United Kingdom are perceived to be more conservative and credible than thosewithout such education, since investors can better anticipate their preferences. We also analyzed incominggovernors with a history in their central bank (insiders) and those without (outsiders). Moreover, weinvestigated whether market reactions depend on the central bank’s degree of independence.
5. To the best of our knowledge, the only other study that seeks to disentangle the two Rogoffdimensions is Berger and Woitek (2005). They find that conservatism does matter for Germany, where theBundesbank enjoyed a virtually unchanged high degree of independence in the postwar period.
1594 : MONEY, CREDIT AND BANKING
with e being the log of the foreign exchange rate (foreign currency/domestic cur-rency), i being the domestic interest rate, and i∗ being the foreign interest rate.Solving forward, we obtain
et = Et
[T −1∑s=0
(it+s − i∗t+s) + eT
],
where eT , the nominal exchange rate at some future date T , can be thought of asthe expected equilibrium exchange rate, determined by—for large enough values ofT—purchasing power parity. Hence, expected changes in monetary policy can affectthe foreign exchange rate either through expected changes in the nominal interest ratedifferentials and/or changes in the expected long-run exchange rate. For an increasein effective conservatism, that is, either an increase in central bank independence orconservatism, we expect the foreign exchange rate to appreciate. Alternatively, thedomestic currency might depreciate during a period of financial crisis (Caballero andKrishnamurthy 2001, 2002).
1.2 Domestic Stock Markets
The effect of a change in the expected monetary policy on stock markets is lessobvious than for exchange markets. Following Campbell et al. (1997), the stock pricecan be expressed as the expected value of future dividends (D) out to the infinitefuture, discounted at a constant rate (R).
Pt = Et
[ ∞∑i=1
(1
1 + R
)i
Dt+i
].
For this classic “Gordon growth model,” changes in monetary policy expectationscan affect stock prices through two different channels. Policy expectations mightsimply affect the discount rate or more subtly affect the expected future dividendstream. In summary, while the direct effect of higher expected inflation on stockprices is ambiguous, to the extent that it is associated with greater economic turmoil,investors may demand a higher risk premium.
1.3 Foreign-Currency-Denominated Bond Markets
The classical approach to model sovereign bond yields dates back to Edwards(1984), where the spread is denoted as a function of the probability of default (pd)and the risk-free interest rate (i*)
s = pd
1 − pd(1 + i∗).
Since our bond data are restricted to foreign-currency-denominated public or pub-licly guaranteed debt, bonds are not sensitive to changing inflation expectations andany possible transmission channel is of an indirect nature via the changing perception
CHRISTOPH MOSER AND AXEL DREHER : 1595
of the probability of default. News that increases the sovereign’s perceived defaultprobability is expected to lower bond prices and, hence, increase sovereign bondspreads.6
2. DATA
Our analysis is based on several types of data. Our main selection criterion isthe availability of reliable daily data on the foreign exchange, and stock and bondmarkets for emerging countries. The resulting sample spans the period 1992–2006for the following 20 economies: Argentina, Brazil, Chile, China, Colombia, Egypt,Hungary, Malaysia, Mexico, Morocco, Pakistan, Peru, Philippines, Poland, Russia,South Africa, South Korea, Thailand, Turkey, and Venezuela. The number of obser-vations varies, however, across countries. The exact data availability and summarystatistics are provided in Appendices A and B.
A central component of our daily data set is the unique sample of exact announce-ment dates of changes of central bank governors, which we have identified through afull-text analysis of the Economist, the Wall Street Journal, and the Financial Times,using the database LexisNexis.7 Overall, this procedure yields 65 events, compris-ing 44 resignations and 21 appointments at the head of a central bank. Appendix Cpresents the number of events per country in our sample, with Argentina and Brazilshowing the greatest number of turnovers.8
We have cautiously sought to infer from newspaper articles whether and to whatextent the respective changes had been anticipated. If the change was largely an-ticipated by the markets, we would expect a softer reaction in market movements.While we can be confident about the hard facts, that is, the name and position of thegovernor and his date of departure, the soft facts about the surprise content should beinterpreted more cautiously. Appendix D gives a precise listing of the timing, nature,and reason of central bank governor changes. We complete our sample with data onthe partisanship of the nominating government, drawn from the Database of PoliticalInstitutions (see Beck et al. 2001).
Our sovereign bond spread data are denominated in U.S. dollars and drawn fromJ.P. Morgan. The Emerging Market Bond Index (EMBI) sovereign bond spreads are
6. We can think of two different channels along which central bank governor changes affect theperceived probability of default. First, if the change at the head of the central bank is interpreted as a signof political interference, investors will be expected to demand a higher yield. A less independent centralbanker (for any given level of conservatism) makes it more likely that fiscal policy ultimately dominatesmonetary policy, driving up public debt and the probability of default. Second, a more conservativeresponse to excess inflation raises (for given level of independence) expected real interest rates, reducingexpected investment and growth rates. This in turn will worsen the debt sustainability, increasing theprobability of default and, hence, the sovereign bond spread. Alternatively, one can argue once more inline with Caballero and Krishnamurthy (2001, 2002).
7. These three financial newspapers are backed by other press sources available through LexisNexis,where necessary. We also compared the dates of resignation with official sources, most of which are listedin the database of central bank governor turnovers provided in Dreher, Sturm, and de Haan (2010); seehttp://www.kof.ethz.ch/centralbankgovernors.
8. Neither the exclusion of Brazil nor Argentina would qualitatively change the results below.
1596 : MONEY, CREDIT AND BANKING
calculated as the yield difference between the (basket of a country’s) emerging marketbond(s) and a comparable U.S. bond.9 These country indices are closely monitoredindicators for perceived country risk in emerging markets and can be interpretedas a default premium charged by investors above the risk-free interest rate. Onlysovereign bonds that comply with well-defined liquidity requirements are eligible forJ.P. Morgan’s bond indices.
Turning to stock markets, we use Morgan Stanley Capital International (MSCI)local market indices, obtained from Datastream. The MSCI data used here are dailyreturns of indices, excluding dividends, and measured in local currency. The in-dices measure market performance for selected securities, capturing the market-capitalization-weighted return of all constituents included.10
The exchange rate data is obtained from Bloomberg. Daily foreign exchange ratesvis-a-vis the U.S. dollar are employed, whereby an increasing foreign exchange ratemeans a depreciation of the domestic currency vis-a-vis the U.S. dollar.
We also employ a number of control variables. We control for U.S. financial marketindicators using the yield of 10-year U.S. Treasury bonds and 3-month U.S. T-bills.Both variables are widely used to control for international liquidity. Finally, we addthe Volatility Index (VIX) of the Chicago Board Options Exchange (CBOE). TheVIX measures the implied volatility from option contracts on the Standard & Poor’s100 (S&P 100) index and, hence, can be interpreted as a forward-looking indicatorof global risk aversion.
3. METHOD AND RESULTS
To test for the effects of central bank governor changes on financial markets, weemploy three different dependent variables: sovereign bond spreads, stock marketindices, and foreign exchange rates vis-a-vis the U.S. dollar.
We start by considering the average market reaction of our variables of interest (�y)to news announcements, following Kuttner and Posen (2010). As the volatility of ourdependent variables varies over time and country, we normalize �y by dividing it byits estimated standard deviation σ over the 90 days preceding the announcement. Thestatistic we use is thus zi ≡ (�yt )/σ . Under the null hypothesis that news regardingthe change of the central bank governor contains no relevant information, �y followsthe preannouncement distribution with zero mean and unit variance (Kuttner andPosen 2010). We test whether the average change in our normalized dependentvariables significantly differs from zero on days where the replacement of a governoris announced. We thus assume that the effect of a governor change materializes on that
9. Henceforth, the notion EMBI is used synonymously for EMBI, Emerging Markets Bond Index PlusEMBI+, and Emerging Markets Bond Global EMBIG. We mainly rely on the EMBI+ due to its largecoverage, its liquidity requirements, and its up-to-date record. Bond spread data from the early 1990s areobtained from EMBI. For Chile and Uruguay, only EMBIG data are available.
10. Note that the returns on MSCI and the respective country index are highly correlated (Pantzalis,Stangeland, and Turtle 2000).
CHRISTOPH MOSER AND AXEL DREHER : 1597
TABLE 1
AVERAGE MARKET REACTION TO EVENTS
Sample Exchange rate Bond yield Stock price
CB resignation (all) Avg. change 0.493 0.297 −0.225N 41 44 44p-value 0.001 0.049 0.136
CB resignation (irregular) Avg. change 0.770 0.248 −0.272N 32 34 34p-value 0.000 0.148 0.113
CB resignation (regular) Avg. change −0.493 0.462 −0.063N 9 10 10p-value 0.139 0.144 0.842
Partisanship Avg. change 0.797 0.097 −0.294N 8 9 9p-value 0.024 0.771 0.378
CB appointment Avg. change 3.217 −0.263 0.766N 20 21 21p-value 0.000 0.228 0.000
NOTES: The table evaluates whether the mean change of our variables of interest (�y) over the sample period, normalized by dividingit by its estimated standard deviation σ over the 90 days preceding the announcement, equals their mean change on the event days. Wetest whether the average change in our normalized dependent variables significantly differs from zero on days where the replacement of agovernor is announced. CB resignation refers to the announcement day of the resignation of a central bank governor, whereby a resignationis coded as irregular if the central bank governor change occurred before the expiration of the central bank’s tenure and as regular otherwise.Partisanship indicates that the current nominating government is less conservative than the government that nominated the previous governor.CB appointment denotes the announcement day of the appointment of a central bank governor.
same day. The average dependent variables are approximately normally distributedwith variance 1/N, where N is the number of events in our sample.11
Table 1 shows the value of the z-statistics, indicating the respective levels ofsignificance. As can be seen, the results show some interesting patterns. Resignationsincrease bond spreads at the 5% level of significance and lead to an exchange ratedepreciation at the 1% level.12 Relating these results to our hypotheses, investors doindeed react to central bank governor changes. In line with Hypothesis 1, the averagemarket reaction to the announcement of the resignation of the central bank governor isnegative in two out of the three financial markets. On average, emerging markets thusseem to react differently than markets in advanced countries. While Kuttner and Posen(2010) report that markets do react significantly to governor changes for their sampleof 15 industrialized countries, their results do not follow a unidirectional pattern:domestic-currency-denominated bond yields do not, on average, rise, nor does theexchange rate depreciate, in response to the announcement of a new governor. Kuttnerand Posen thus reject the widely held belief that markets might react due to a lack ofcredibility at the start of a governor’s tenure. This is contrary to our own results.
The appointments of new governors do not affect bond spreads significantly butlead, on average, to a depreciation of the exchange rate and an increase in stock prices
11. In the working paper version of this paper, we provide two further simple methods, which producequalitatively similar results.
12. In a test for robustness, we increased the event period to include 1 day prior and 1 day after theannouncement day. Our results are sensitive to this change to some extent. While our findings on theexchange rate effects remain very similar, bond spreads are no longer significantly affected.
1598 : MONEY, CREDIT AND BANKING
(at the 1% level of significance).13 The stock market reaction can be interpreted in linewith Brown, Harlow, and Tinic (1988) who show that the resolution of uncertaintyleads to positive market reactions, on average. Once the successor is announced,pending uncertainty about the new head of the central bank is resolved.
What about the impact of effective conservatism on financial markets? Turning tothe first dimension of effective conservatism, we draw on the Database of PoliticalInstitutions (Beck et al. 2001) to construct a proxy for perceived conservatism. Thedatabase classifies the partisanship of an incumbent government as left, center, orright. Our variable takes a value of 1 when partisanship of the government nominat-ing the current governor is to the left of the government who nominated the previousgovernor, –1 when the nominating government is more conservative than the previousone, and 0 if the government’s degree of conservatism remains unchanged. In con-structing this proxy for the central bank governor’s conservatism, we follow Bergerand Woitek (2005), who assume that right-wing governments tend to nominate moreinflation-averse central bank governors, and vice versa. Arguably, the assumptionthat the nominating governments’ political color corresponds to those of the gover-nor is debatable. We therefore checked our coding by employing full-text analysisin LexisNexis. While exclusive reliance on LexisNexis would reduce the number ofevents and would not allow meaningful econometric analysis, those cases where wecould identify the relative conservatism of an incoming governor largely confirm thevalidity of our coding (see Appendix E).
Table 1 reports results for a sample of resignations where the government nomi-nating the incoming governor has a different political partisanship than the one thatappointed the previous governor. More specifically, the table shows market reactionswhen central bank governors are appointed by less conservative governors, as de-fined above. As argued in Hypothesis 2, markets react to changes in the governors’perceived conservatism. When a governor resigns and the incumbent government onthe announcement day is less conservative than the one that nominated the outgo-ing governor, markets should thus react negatively. As can be seen, we find such anegative reaction in favor of Hypothesis 2 only for the inflation sensitive exchangerate. Note, however, that the very small number of (a maximum of nine) observationsmakes this finding rather tentative.
We also gain insight into Hypothesis 3 by separating regular from irregular res-ignations. While irregular events apparently drive the negative announcement effecton the exchange rate, the effects are marginally insignificant for the bond and stockmarket. We do not find evidence for significant market reactions to regular resig-nations. If markets were primarily concerned about the incoming governor lackingcredibility, we would expect regular and irregular events to be equally likely to affectmarkets. If, however, central bank independence is perceived to be at risk, irregularchanges of central bank governors should drive the results. Hence, irregular events
13. Note, however, that the very small sample size might drive the insignificant effect of appointmentson bond spreads.
CHRISTOPH MOSER AND AXEL DREHER : 1599
can be expected to alter perceived central bank independence, which is one of thetwo key determinants of inflationary bias.14
Arguably, stock markets and U.S. dollar-denominated sovereign bonds are ratherinsensitive to changes in expected inflation. This might explain why we cannot findevidence in favor of Hypothesis 3 for these financial markets. However, note onceagain that these conclusions rest on a very small number of observations, and are thusonly tentative.
As a next step, we conduct panel data analysis. Our baseline regression is estimatedby pooled ordinary least squares (OLS) with robust standard errors clustered at thecountry level15 and takes the following form:
�Yi,t = α + λ�Yi,t−1 + βRESIGNi,t + γ APPOINTi,t + η�Xi,t
+ νw Dw + εi,t , (1)
where the subscripts i and t indicate country and time, respectively. Yi,t representsthe respective dependent variables denoted in log-differences. The one-period laggedvalue of the dependent variable also enters the equation as a covariate. Our coefficientsof interest are β and γ , accounting for the impact of resignations, RESIGNi,t, and newappointments, APPOINTi,t, at the head of the central bank.16 The two announcementvariables are 1 on the day of the change and 0 otherwise. We denote the error termεi,t and employ dummy variables Dw, running from Monday to Thursday, in order tocontrol for day-of-the-week effects.
The matrix Xi,t includes the three U.S. financial market indicators introduced above.We employ (log) changes of the volatility index, and (log) changes in 10-year U.S.Treasury bond yields and 3-month U.S. T-bill yields, defined as 100 × log (1 + it,US).
By definition, fixed exchange rate regimes do not allow for daily market reactionsto the announcement of the turnover. For this reason, we interact the resignation orappointment variable with a dummy variable that takes the value of 1 when there is aflexible exchange rate (or at least not a fully pegged exchange rate) and 0 otherwise.17
Finally, as one potential caveat to the analysis, central bank governors might bedismissed as a consequence of economic crises, giving rise to endogeneity. Whengovernors are dismissed due to economic shocks, market reactions might reflect theseshocks rather than the exogenous change in who governs the central bank. While thisargument appears reasonable for quarterly or yearly data, endogeneity is unlikelyto be an issue when data frequency is daily, as is the case in our study. Even if
14. In coding the dummy for irregular changes, we follow Kuttner and Posen (2010) and also classifycases in which the incumbent governor was eligible for reappointment, but did not receive it, as irregular.Note that these kinds of irregular changes constitute a very small number of all irregular resignations.
15. We do not include fixed country effects as they are not jointly significant at conventional levels.Our key results are not changed by their exclusion.
16. On average, the new central bank governor is announced 10 days after the resignation of hispredecessor. We only measure appointments when they constitute an independent event.
17. We code an exchange rate regime as flexible as long as at least some market reactions are observablein the weeks around the day of the event.
1600 : MONEY, CREDIT AND BANKING
TABLE 2
FINANCIAL MARKETS AND GOVERNOR CHANGES, 20 COUNTRIES, 1992–2006
(1) (2) (3) (4) (5)Stock price Stock price IV Exchange rate Bond yield Bond yield IV
� log y, lagged 0.1062∗∗∗ 0.1854∗∗∗ 0.0249 −0.1925∗∗∗ 0.5765∗∗∗(6.55) (3.12) (0.53) (3.38) (14.23)
CB resignation −0.0042∗ −0.0045∗∗ 0.0097∗∗∗ 0.0047 0.0152∗∗∗(1.96) (2.23) (2.89) (1.43) (4.23)
CB appointment 0.0140 0.0129 −0.0073 0.0002 −0.0086(1.40) (1.36) (0.72) (0.02) (0.83)
� log volatility index (VIX) −0.0534∗∗∗ −0.0538∗∗∗ 0.0053∗∗ 0.0825∗∗∗ 0.0801∗∗∗(4.66) (4.70) (2.62) (5.45) (5.44)
� log U.S. T-bond 10 years 0.0682∗∗∗ 0.0658∗∗ 0.0176∗ −0.9960∗∗∗ −0.9299∗∗∗(3.94) (3.73) (1.97) (4.09) (3.91)
� log U.S. T-bill 3 months 0.0158 0.0167 −0.0043 −0.1355 −0.2050∗∗(1.01) (1.01) (0.77) (1.71) (2.22)
Observations 51,423 51,422 51,423 51,399 51,375R2 0.04 0.03 0.003 0.07 —
NOTES: The dependent variable is the (log) change in y. Results are based on clustered standard errors. (absolute) values of t-statistics inparentheses. Week-day effects and a constant are estimated but not reported. The instrumental variable (IV) estimation in columns (2) and(5) uses the second lag of the dependent variable as instrument for the lagged dependent variable. Testing for first-order autocorrelation inthe error term indicates no first-order auto correlation.∗∗∗ , ∗∗ , ∗denote 1%, 5%, 10% level of significance.
the governor is fired as a consequence of macroeconomic crisis, such crises usuallyunfold over a longer period of time, so that daily data can still be used to identify thecausal impact of the turnover itself on market reactions. Endogeneity is thus unlikelyto be an issue here. Still, the effect of central bank governor turnovers might bedifferent during times of crisis, a matter to which we return below.
Table 2 reports the results of the panel data analysis. Column (1) reports thebase specification while the potential bias introduced by the correlation between thelagged dependent variable and the error term is addressed in column (2). We thereforeinstrument the (highly significant) lagged dependent variable with its second lag(Anderson and Hsiao 1982). Our results are hardly changed by this. Column (3)shows the results for the exchange rate. Finally, results for bond spreads are reportedin columns (4) and (5). The lagged dependent variable is completely insignificantin column (3) and we therefore do not instrument it. While the lagged dependentvariable is highly significant according to the OLS specification of column (4), witha negative coefficient, it becomes significantly positive once instrumented (column(5)).
Turning to our variables of interest, Table 2 again shows some support for our hy-potheses and, specifically, that financial markets do react to the resignation of centralbankers. By contrast, the announcement of the appointment of a governor has nosignificant impact. Overall, the results regarding appointments are thus weaker thanthose reported earlier in Table 1. We base our conclusions on the more stringent re-gression analysis. As can be seen in the table, domestic stock markets react negativelyto central bank governor resignations. The estimated coefficients in columns (1) and
CHRISTOPH MOSER AND AXEL DREHER : 1601
(2) imply a small decline in returns of about 0.5 percentage points. While being farfrom a dramatic crash of markets, this effect is, arguably, economically relevant. Thestock market announcement effect is about five times higher than the one triggeredby changes in sovereign ratings and sovereign rating outlooks in emerging markets(Kaminsky and Schmukler 2002).18
Column (3) shows that the resignation of a central banker leads to a depreciation ofthe exchange rate, at the 1% level of significance.19 The estimated coefficient impliesa depreciation of almost 1 percentage point following the resignation of a centralbanker. Finally, columns (4) and (5) show that bond spreads do not increase followingthe resignation of the head of the central bank according to the OLS regressions butdo increase (at the 1% level of significance) once the lagged dependent variable isinstrumented. According to the coefficient, the resignation of a central bank governorincreases bond spreads by more than 1.5 percentage points.
Based on our results, we cannot reject Hypothesis 1, stressing the credibility prob-lem at the start of a governor’s tenure. As potential objection to this interpretation, thenegative market reactions might, of course, simply reflect the presence of a financialcrisis, indicated by the resignation of the governor. We therefore include interactionterms between our proxies for crises and changes of central bank governors. Thefollowing specification accounts for financial crises:
�Yi,t = α + λ�Yi,t−1 + β1RESIGNi,t + β2RESIGNi,t ∗ CRISISi,t
+ γ1APPOINTi,t + γ2APPOINTi,t ∗ CRISISi,t + η�Xi,t + νw Dw
+φCRISISi,t + εi,t , (2)
where our variable CRISISi,t takes the value of 1 in case of an ongoing currencycrisis in country i in month t and 0 otherwise. We follow Kaminsky and Reinhart(1999) and compute a foreign exchange market pressure index as an average ofthe rate of change of the exchange rate and of international reserves, weighted bytheir standard deviations.20 Following Dreher, Herz, and Karb (2006), a country isdefined to be in a currency crisis when its index deviates from the index mean for thatcountry by at least one standard deviation. According to this definition, 11 centralbankers from nine countries have been dismissed during a currency crisis in oursample.
Table 3 shows the results. As can be seen, currency crises never enter significantlyin any of our regressions. Similarly, the interaction of crises and dismissals is notsignificant at conventional levels in any specification, and with the exception of
18. This comparison can only be made for the U.S. dollar-denominated MSCI index (results notshown).
19. Note that this result remains when the insignificant lagged dependent variable is omitted from theregression.
20. Ideally, we would also like to include interest rates that monetary authorities can use to defendparities. However, this leaves us with a drastically reduced sample, and so, following Kaminsky andReinhart (1999), we do not include interest rates.
1602 : MONEY, CREDIT AND BANKING
TABLE 3
FINANCIAL MARKETS, GOVERNOR CHANGES, AND FINANCIAL CRISES, 20 COUNTRIES, 1992–2006
(1) (2) (3) (4) (5)Stock price Stock price IV Exchange rate Bond yield Bond yield IV
� log y, lagged 0.1060∗∗∗ 0.1851∗∗∗ 0.0221 −0.1926∗∗∗ 0.5766∗∗∗(6.52) (3.13) (0.52) (3.38) (14.2)
CB resignation [1] −0.0051∗ −0.0054∗∗ 0.0090∗∗ 0.0051 0.0136∗∗∗(1.97) (2.23) (2.54) (1.44) (4.19)
CB resignation (crisis) [2] 0.0046 0.0046 0.0044 −0.0027 0.0089(0.82) (0.85) (0.37) (0.44) (0.59)
CB appointment [3] 0.0080 0.0075 0.0048 −0.0119 −0.0172(1.68) (1.50) (1.51) (1.12) (1.51)
CB appointment (crisis) [4] 0.0314 0.0284 −0.0607 0.0635∗∗ 0.0451∗(0.68) (0.65) (1.46) (2.75) (1.88)
� log volatility index (VIX) −0.0534∗∗∗ −0.0538∗∗∗ 0.0053∗∗ 0.0824∗∗∗ 0.0801∗∗∗(4.67) (4.70) (2.64) (5.44) (5.44)
� log U.S. T-bond 10 years 0.0682∗∗∗ 0.0658∗∗∗ 0.0176∗ −0.9961∗∗∗ −0.9299∗∗∗(3.93) (3.73) (1.97) (4.09) (3.91)
� log U.S. T-bill 3 months 0.0156 0.0165 −0.0042 −0.1357 −0.2051∗∗(1.01) (1.00) (0.76) (1.71) (2.22)
Currency crisis period −0.0001 −0.0001 −0.0003 −0.0002 −0.0003(1.01) (0.40) (1.49) (0.33) (1.14)
Joint significance [1] + [2] 0.16 0.10 0.03 0.36 0.00Joint significance [3] + [4] 0.19 0.25 0.16 0.04 0.16Observations 51,423 51,422 51,423 51,399 51,375R2 0.04 0.03 0.006 0.07 —
NOTES: The dependent variable is the (log) change in y. Results are based on clustered robust standard errors. The (absolute) values oft-statistics are in parentheses. Week-day effects and a constant are estimated but not reported. The instrumental variable (IV) estimation incolumns (2) and (5) uses the second lag of the dependent variable as instrument for the lagged dependent variable. The coefficients [2] and [4]represent the respective interaction with our crisis dummy. Testing for first-order autocorrelation. The in the error term indicates no first-ordercorrelation. p-values for test of joint significance are reported.∗∗∗ , ∗∗ , ∗denote 1%, 5%, 10% level of significance.
the bond market, the same is true regarding the interaction of appointments andcrises. We actually do find some statistically weak evidence that appointments areperceived more negatively by sovereign bond holders during currency crises than intranquil times. Note, however, that while the point estimates for the stock market,foreign exchange, and bond markets marginally change when we control for financialcrises, overall, there is neither evidence for an increased sensitivity to central bankturnovers in crisis times nor for an additional risk premium following turnovers duringcrises.21
Coming back to Hypothesis 1, we conclude that our results do not reflect a riskpremium. However, the negative market reactions are also in line with a decreasein perceived effective conservatism. Given the results in Table 2, we cannot knowwhether Hypothesis 1 or, alternatively, Hypotheses 2 or 3 hold true. In order to
21. Note that the coefficient on central bank appointments is marginally insignificant in column (1).This is even though the t-statistic of 1.68 with more than 50,000 observations might suggest otherwise.However, the standard errors are clustered at the country level, so the number of clusters (M) becomesthe relevant dimension to determine the degrees of freedom. The confidence values Stata 11 reports whenclustering standard errors are based on (more conservative) critical values from a Student’s t-distributionwith M–1 degrees of freedom.
CHRISTOPH MOSER AND AXEL DREHER : 1603
disentangle the two dimensions of “effective conservatism” we estimate the followingequation:
�Yi,t = α + λ�Yi,t−1 + β1RESIGNi,t + β2RESIGNi,t ∗ PARTISANi,t
+β3RESIGNi,t ∗ IRREGi,t + γ APPOINTi,t + η�Xi,t
+ νw Dw + εi,t , (3)
where PARTISANi,t is our proxy for the governor’s perceived conservatism describedabove and IRREGi,t is a dummy variable that takes the value of 1 when the centralbank governor change occurred before the expiration of the central bank governor’stenure and 0 otherwise.22 According to the results (not shown in a table), almostno coefficient of interest is significant at conventional levels. There is one importantexception, however. Irregular turnovers lead to a depreciation of the exchange rate,with a coefficient significant at the 5% level. This provides some evidence thatinvestors are indeed worried about central bank independence when central bankgovernors resign before the end of their tenure.
4. CONCLUSION
Central bank governor changes in emerging markets may convey important signalsabout future monetary policy. Based on a new daily data set on 20 countries over the1992–2006 period, this paper has examined the reactions of foreign exchange markets,domestic stock market indices, and sovereign bond spreads to the announcement ofa central bank governor change.
Our results show, first, that the resignation of a central bank governor negativelyaffects financial markets on the announcement day, with average market reactionsbetween 0.5 and 1.5 percentage points. By contrast, we find little evidence thatappointments of new governors convey relevant news to investors.
Second, comparing our results with those found in the existing literature, we findthat our results for emerging market economies are distinct from industrialized coun-tries in an interesting and important way. Newly appointed central bank governorsapparently suffer from a systematic credibility problem at the beginning of theirtenure. In contrast to their counterparts in industrialized countries, emerging marketgovernors initially have to face (at least) a transitory rise in inflation expectations,because investors are uncertain about the true type of the central bank governor(hawkish versus dovish).
Third, the negative announcement effect for resignations is mainly driven by irreg-ular changes, that is, changes occurring before the scheduled end of tenure. Foreign
22. Note that we cannot include PARTISAN and IRREG individually, as they, by definition, can onlybe observed when the governor changes. We do not interact the appointment dummy with changes inpartisanship, since the incumbent government’s partisanship, and hence potential changes in ideologyvis-a-vis the government nominating the previous central bank governor, are already priced in on theresignation day.
1604 : MONEY, CREDIT AND BANKING
exchange market participants are apparently sensitive to signals about perceivedcentral bank independence, expecting higher inflationary bias.
Finally, we find little evidence that the governor’s degree of conservatism mattersfor market reactions. Still, this lack of significance might well be due to the verylimited number of observations for changes in conservatism.
Overall, our study complements Santiso (2003), pointing out that key policymakersin emerging markets are crucial for building credibility in international financialmarkets in one important aspect. Investors are apparently sensitive to the way inwhich an incumbent government handles the replacement of key policymakers. Withrespect to central bank governor changes, investors seem to care most about perceivedcentral bank independence. Future research to further disentangle the two dimensionsof effective conservatism is clearly desirable.
CHRISTOPH MOSER AND AXEL DREHER : 1605
APPENDIX A: DATA AVAILABILITY, EMBI (G), FX TO U.S. DOLLAR, ANDMSCI AVAILABLE
Country Start End
Argentina April 30, 1993 July 31, 2006Brazil January 15, 1992 July 31, 2006Chile May 28, 1999 July 31, 2006China December 31, 1997 July 31, 2006Colombia December 31, 1997 July 31, 2006Egypt July 31, 2001 July 31, 2006Hungary January 1, 1999 July 31, 2006Malaysia December 31, 1997 July 31, 2006Mexico December 31, 1991 July 31, 2006Morocco December 31, 1997 July 31, 2006Pakistan June 29, 2001 July 31, 2006Peru May 30, 1997 July 31, 2006Philippines December 31, 1991 July 31, 2006Poland December 31, 1997 July 31, 2006Russia December 31, 1997 July 31, 2006South Africa December 31, 1997 July 31, 2006South Korea December 31, 1997 July 31, 2006Thailand December 31, 1997 March 31, 2006Turkey December 31, 1997 July 31, 2006Venezuela December 31, 1992 July 31, 2006
APPENDIX B: DESCRIPTIVE STATISTICS
Variable Mean Std. dev. Min Max
(log) Spread 5.68 1.07 0.00 8.88(log) MSCI 5.53 1.06 2.55 8.61(log) Exchange rate 3.09 −2.44 7.82 8.00Central banker change 0.00 0.03 0.00 1.00(log) VIX 2.91 0.32 2.23 3.90(log) U.S. T-bond 10 years 1.94 0.22 1.41 2.35(log) U.S. T-bill 3 months 1.62 0.42 0.59 2.31
APPENDIX C: NUMBER OF CENTRAL BANK GOVERNOR CHANGES BYCOUNTRY
Country Number events Country Number events
Argentina 6 (3) Pakistan 1 (1)Brazil 9 (2) Peru 4 (3)Chile 1 (1) Philippines 2 (1)China 1 (0) Poland 1 (2)Colombia 1 (0) Russia 2 (1)Egypt 2 (1) South Africa 0 (1)Hungary 1 (1) South Korea 2 (0)Malaysia 2 (1) Thailand 2 (0)Mexico 1 (0) Turkey 2 (2)Morocco 1 (0) Venezuela 3 (1)
NOTE: This table reports the number of central bank governor resignations (appointments). Our data are for the period of 1992–2006.
1606 : MONEY, CREDIT AND BANKINGA
PPE
ND
IXD
:CE
NT
RA
LB
AN
KG
OV
ER
NO
RC
HA
NG
ES:
LO
CA
TIO
N,D
AT
E,N
AM
E,A
ND
DIF
FER
EN
TN
EW
SC
HA
RA
CT
ER
IST
ICS
(199
2–20
06).
Cou
ntry
Dat
eaN
ame
Irre
gula
r(y
es/n
o)R
easo
nN
ews
char
acte
rN
ews
type
Arg
entin
a19
96R
oque
Fern
ande
zY
esFe
rnan
dez
step
sdo
wn
asgo
vern
orof
the
cent
ralb
ank
tobe
com
eec
onom
ym
inis
ter.
Not
antic
ipat
edR
esig
natio
n
Arg
entin
a19
96Pe
dro
Pou
–Po
uis
anno
unce
das
new
cent
ralb
ank
gove
rnor
.N
otan
ticip
ated
App
oint
men
t
Arg
entin
a20
01Pe
dro
Pou
Yes
Pres
iden
tDe
La
Rua
dism
isse
sPo
uon
alle
ged
bad
adm
inis
trat
ion.
Part
lyan
ticip
ated
Res
igna
tion
Arg
entin
a20
02R
oque
Mac
caro
neY
esM
acca
rone
sudd
enly
resi
gns
and
Mar
ioB
leje
rbe
com
eshi
ssu
cces
sor.
Not
antic
ipat
edR
esig
natio
n
Arg
entin
a20
02M
ario
Ble
jer
Yes
Ble
jer
step
sdo
wn
afte
rre
peat
edly
butti
nghe
ads
with
econ
omy
min
iste
rL
avag
na.
Part
lyan
ticip
ated
Res
igna
tion
Arg
entin
a20
02A
ldo
Pign
anel
li–
Pign
anel
liis
appo
inte
dby
the
pres
iden
t.Pa
rtly
antic
ipat
edA
ppoi
ntm
ent
Arg
entin
a20
02A
ldo
Pign
anel
liY
esPi
gnan
elli
isth
ird
top
bank
erto
resi
gnw
ithin
aye
ar.
Part
lyan
ticip
ated
Res
igna
tion
Arg
entin
a20
02A
lfon
soPr
atG
ay–
Form
erW
allS
tree
tban
ker
Prat
Gay
appo
inte
d.Pa
rtly
antic
ipat
edA
ppoi
ntm
ent
Arg
entin
a20
04A
lfon
soPr
atG
ayY
esPr
esid
entK
irch
ner
ina
surp
rise
mov
ere
plac
esPr
atG
ayby
Mar
tinR
edra
do.
Not
antic
ipat
edR
esig
natio
n
Bra
zil
1992
Fran
cisc
oG
ros
Yes
Gro
sis
succ
eede
dby
Gus
tavo
Loy
ola.
Part
lyan
ticip
ated
Res
igna
tion
Bra
zil
1993
Gus
tavo
Loy
ola
Yes
Loy
ala
resi
gns
alon
gw
ithth
eba
nk’s
who
ledi
rect
orat
e.N
otan
ticip
ated
Res
igna
tion
Bra
zil
1993
Paul
oC
esar
Xim
enes
–Pa
ulo
Ces
arX
imen
esis
tore
plac
eL
oyol
a.N
otan
ticip
ated
App
oint
men
t
Bra
zil
1993
Paul
oC
esar
Xim
enes
Yes
Pedr
oM
alan
repl
aces
Ces
arX
imen
esas
new
Cen
tral
Ban
kgo
vern
oraf
ter
disp
ute
with
Pres
iden
tFra
nco.
Part
lyan
ticip
ated
Res
igna
tion
Bra
zil
1994
Pedr
oM
alan
Yes
Mal
anis
nam
edto
beco
me
finan
cem
inis
ter.
Part
lyan
ticip
ated
Res
igna
tion
Bra
zil
1995
Pers
ioA
rida
Yes
Ari
daab
rupt
lyre
sign
sfo
rpe
rson
alre
ason
s.N
otan
ticip
ated
Res
igna
tion
Bra
zil
1997
Gus
tavo
Loy
ola
Yes
Loy
ola
step
sdo
wn
for
pers
onal
reas
ons
and
isim
med
iate
lyre
plac
edby
Gus
tavo
Fran
co.
Part
lyan
ticip
ated
Res
igna
tion
Bra
zil
1999
Gus
tavo
Fran
coY
esM
arke
tspl
umm
etw
hen
Cen
tral
Ban
kch
ief
Gus
tavo
Fran
coab
rupt
lyre
sign
s.Pa
rtly
antic
ipat
edR
esig
natio
n
Bra
zil
1999
Fran
cisc
oL
opes
Yes
Arm
inio
Frag
are
plac
esL
opes
.N
otan
ticip
ated
Res
igna
tion
(Con
tinu
ed)
CHRISTOPH MOSER AND AXEL DREHER : 1607A
PPE
ND
IXD
:CO
NT
INU
ED
Cou
ntry
Dat
eaN
ame
Irre
gula
r(y
es/n
o)R
easo
nN
ews
char
acte
rN
ews
type
Bra
zil
2002
Arm
inio
Frag
aY
esPr
esid
ent-
elec
tspo
kesm
anm
akes
clea
rth
atFr
aga
isno
tto
stay
atth
eC
entr
alB
ank.
Ant
icip
ated
Res
igna
tion
Bra
zil
2002
Hen
riqu
eM
eire
lles
–M
eire
lles
isan
noun
ced
tofo
llow
Frag
a.Pa
rtly
antic
ipat
edA
ppoi
ntm
ent
Chi
le20
03C
arlo
sM
assa
dY
esM
assa
dfin
ally
resi
gns
due
toa
finan
cial
scan
dal.
Part
lyan
ticip
ated
Res
igna
tion
Chi
le20
03V
ittor
ioC
orbo
–C
orbo
isof
ficia
llyno
min
ated
byPr
esid
ent
Lag
os.
Not
antic
ipat
edA
ppoi
ntm
ent
Chi
na20
02D
aiX
iang
long
Yes
Zho
uX
iaoc
huan
isna
med
gove
rnor
ofth
ePe
ople
’sB
ank
ofC
hina
,tak
ing
over
from
Xia
nglo
ng.
Part
lyan
ticip
ated
Res
igna
tion
Col
ombi
a20
04M
igue
lUrr
utia
No
Urr
utia
step
sdo
wn
afte
rhi
sth
ird
4-ye
arte
rm.J
ose
Dar
ioU
ribe
ishi
ssu
cces
sor.
Ant
icip
ated
Res
igna
tion
Egy
pt20
01Is
mai
lHas
san
No
Has
san’
ste
rmen
dsan
dhe
aske
dno
tto
bere
appo
inte
dfo
rpe
rson
alre
ason
s.A
ntic
ipat
edR
esig
natio
n
Egy
pt20
01M
ahm
oud
Abu
el-A
youn
–E
l-A
youn
,afo
rmer
econ
omic
spr
ofes
sor,
isap
poin
ted
tobe
com
ene
wC
entr
alB
ank
gove
rnor
.
Part
lyan
ticip
ated
App
oint
men
t
Egy
pt20
03M
ahm
oud
Abu
el-A
youn
Yes
Pres
iden
tMub
arak
appo
ints
Faro
ukel
-Oqd
aas
gove
rnor
ofth
eC
entr
alB
ank
repl
acin
gE
l-A
youn
.
Part
lyan
ticip
ated
Res
igna
tion
Hun
gary
2000
Gyo
rgy
Sura
nyi
Yes
Sura
nyi’s
rene
wab
leco
ntra
ctw
illno
tbe
exte
nded
.A
ntic
ipat
edR
esig
natio
n
Hun
gary
2001
Zsi
gmon
dJa
rai
–Ja
rair
ecei
ves
his
offic
iala
ppoi
ntm
entf
rom
Pres
iden
tMad
l.A
ntic
ipat
edA
ppoi
ntm
ent
Mal
aysi
a19
98A
hmad
Moh
amed
Don
Yes
Ahm
adD
onre
sign
sal
ong
with
his
depu
ty.
Part
lyan
ticip
ated
Res
igna
tion
Mal
aysi
a19
98A
liA
bdul
Has
san
–T
hehe
adof
the
Eco
nom
icPl
anni
ngU
nit,
Ali
Abd
ulH
assa
n,is
pick
edas
succ
esso
r.N
otan
ticip
ated
App
oint
men
t
Mal
aysi
a20
00A
liA
bdul
Has
san
No
Zei
tAkh
tar
Azi
zfo
llow
sup
onA
liA
bdul
Has
san’
sco
ntra
ctex
piry
.Pa
rtly
antic
ipat
edR
esig
natio
n
Mex
ico
1997
Mig
uelM
ance
raN
oM
ance
rare
tires
asex
pect
ed,b
utno
min
atio
nof
Fina
nce
Min
iste
rG
uille
rmo
Ort
izco
mes
asa
surp
rise
.
Ant
icip
ated
Res
igna
tion
(Con
tinu
ed)
1608 : MONEY, CREDIT AND BANKING
APP
EN
DIX
D:C
ON
TIN
UE
D
Cou
ntry
Dat
eaN
ame
Irre
gula
r(y
es/n
o)R
easo
nN
ews
char
acte
rN
ews
type
Mor
occo
2003
Moh
amed
Sekk
atN
oSe
kkat
isre
plac
edby
Abd
ella
tifJo
uahr
iWal
i.Pa
rtly
antic
ipat
edR
esig
natio
nPa
kist
an20
05Is
hrat
Hus
sain
No
Ishr
atou
tline
spl
ans
afte
rup
com
ing
retir
emen
t.A
ntic
ipat
edR
esig
natio
n
Paki
stan
2005
Sham
shad
Akh
tar
–Sh
amsh
adA
khta
ris
nam
edC
entr
alB
ank
gove
rnor
.Pa
rtly
antic
ipat
edA
ppoi
ntm
ent
Peru
2000
Ger
man
Suar
ez–
Suar
ezis
reap
poin
ted
byPr
esid
entF
ujim
ori.
Part
lyan
ticip
ated
App
oint
men
tPe
ru20
01G
erm
anSu
arez
Yes
Ric
hard
Web
bis
chos
enby
new
Pres
iden
tTo
ledo
.N
otan
ticip
ated
Res
igna
tion
Peru
2003
Ric
hard
Web
bY
esW
ebb
tend
ers
his
resi
gnat
ion
afte
rru
mor
sab
outi
nfigh
ting
ingo
vern
or’s
boar
d.Pa
rtly
antic
ipat
edR
esig
natio
n
Peru
2003
Javi
erSi
lva
Rue
te–
Fina
nce
Min
iste
rSi
lva
Rue
teis
nam
edne
wC
entr
alB
ank
gove
rnor
.Pa
rtly
antic
ipat
edA
ppoi
ntm
ent
Peru
2004
Osc
arD
anco
urt
–D
anco
urtb
ecom
esac
ting
Cen
tral
Ban
kgo
vern
ordu
eto
Silv
a’s
IMF
dire
ctor
ate
appo
intm
ent
Not
antic
ipat
edA
ppoi
ntm
ent
Peru
2004
Javi
erSi
lva
Rue
teY
esSi
lva
step
sdo
wn
afte
rcr
itici
smre
gard
ing
the
inco
mpa
tibili
tyof
his
new
posi
tion
asIM
Fdi
rect
orw
hile
pres
iden
tof
the
Cen
tral
Ban
k.
Part
lyan
ticip
ated
Res
igna
tion
Peru
2006
Osc
arD
anco
urt
Yes
New
pres
iden
tGar
cia
repl
aces
actin
gpr
esid
ent
ofth
eC
entr
alB
ank
Dan
cour
tby
Julio
Vel
arde
.
Not
antic
ipat
edR
esig
natio
n
Phili
ppin
es19
93Jo
seC
uisi
aN
oG
over
nor
Cui
sia
resi
gns
for
pers
onal
reas
ons.
Not
antic
ipat
edR
esig
natio
nPh
ilipp
ines
1993
Gab
riel
Sing
son
–Pr
esid
entR
amos
appo
ints
Sing
son
asne
wC
entr
alB
ank
gove
rnor
.Pa
rtly
antic
ipat
edA
ppoi
ntm
ent
Phili
ppin
es20
05R
afae
lBue
nave
ntur
aN
oPr
esid
entA
rroy
oan
noun
ces
that
Am
ando
Teta
ngco
beco
mes
Cen
tral
Ban
kpr
esid
ent
whe
nB
uena
vent
ura
retir
esin
July
.
Not
antic
ipat
edR
esig
natio
n
Pola
nd19
98H
anna
Gro
nkie
wic
z-W
altz
–Pr
esid
entK
was
niew
skiw
ants
Gro
nkie
wic
z-W
altz
tobe
reap
poin
ted
for
anot
her
6-ye
arte
rm.
Part
lyan
ticip
ated
App
oint
men
t
Pola
nd20
00H
anna
Gro
nkie
wic
z-W
altz
Yes
Gro
nkie
wic
z-W
altz
resi
gns
inor
der
tota
keE
BR
Dvi
ce-p
resi
denc
y.no
tant
icip
ated
resi
gnat
ion
Pola
nd20
00L
esze
kB
alce
row
icz
–Pr
esid
entK
was
niew
skip
ropo
ses
Bal
cero
wic
zto
beco
me
new
Cen
tral
Ban
kgo
vern
or.
Ant
icip
ated
App
oint
men
t
(Con
tinu
ed)
CHRISTOPH MOSER AND AXEL DREHER : 1609A
PPE
ND
IXD
:CO
NT
INU
ED
Cou
ntry
Dat
eaN
ame
Irre
gula
r(y
es/n
o)R
easo
nN
ews
char
acte
rN
ews
type
Rus
sia
1998
Serg
eiD
ubin
inY
esD
ubin
inco
nfirm
sth
athe
inte
nds
tore
sign
.Pa
rtly
antic
ipat
edR
esig
natio
nR
ussi
a19
98V
ikto
rG
eras
chen
ko–
Pres
iden
tYel
tsin
prop
oses
Ger
asch
enko
asD
ubin
in’s
succ
esso
r.Pa
rtly
antic
ipat
edA
ppoi
ntm
ent
Rus
sia
2002
Vik
tor
Ger
asch
enko
Yes
Pres
iden
tPut
inac
cept
sG
eras
chen
ko’s
resi
gnat
ion
and
nom
inat
esSe
rgei
Igna
tyev
.N
otan
ticip
ated
Res
igna
tion
Sout
hA
fric
a19
98T
itoM
bow
eni
No
Mbo
wen
iis
anno
unce
dto
follo
wC
hris
Stal
sup
onhi
sre
tirem
enti
n19
99.
Not
antic
ipat
edA
ppoi
ntm
ent
Sout
hK
orea
1998
Lee
Kyu
ng-s
ikY
esC
hon
Cho
l-H
wan
isap
poin
ted
asgo
vern
orof
the
cent
ralB
ank
ofK
orea
byne
wpr
esid
ent
Dae
-Jun
g.
Not
antic
ipat
edR
esig
natio
n
Sout
hK
orea
2002
Cho
nC
hol-
Hw
anN
oPr
esid
entD
ae-j
ung
nam
esPa
rkSe
ung
asgo
vern
orof
the
Ban
kof
Kor
easu
ccee
ding
Cho
l-H
wan
.
Part
lyan
ticip
ated
Res
igna
tion
Tha
iland
1998
Cha
iyaw
atW
ibul
swas
diY
esW
ibul
swas
dian
dse
vera
ltop
depu
ties
resi
gnun
der
polit
ical
pres
sure
.Cha
tum
ongk
olSo
naku
lis
imm
edia
tely
nom
inat
edas
succ
esso
r.
Ant
icip
ated
Res
igna
tion
Tha
iland
2001
Cha
tum
ongk
olSo
naku
lY
esSo
naku
lis
forc
edto
resi
gnby
Prim
eM
inis
ter
Shin
awat
radu
eto
polic
yco
ntro
vers
y.Pr
idiy
atho
rnD
evak
ula
beco
mes
succ
esso
r.
Part
lyan
ticip
ated
Res
igna
tion
Tur
key
2001
Gaz
iErc
elY
esE
rcel
offe
red
his
resi
gnat
ion,
afte
rT
urke
yha
dto
over
haul
anam
bitio
usre
form
prog
ram
.Pa
rtly
antic
ipat
edR
esig
natio
n
Tur
key
2001
Sure
yya
Serd
enge
cti
–Se
rden
gect
itak
esov
eras
gove
rnor
.N
otan
ticip
ated
App
oint
men
tT
urke
y20
06Su
reyy
aSe
rden
gect
iN
oSe
rden
gect
i’s5-
year
term
expi
res.
Ant
icip
ated
Res
igna
tion
Tur
key
2006
Dur
mus
Yilm
az–
Yilm
az’s
nom
inat
ion
ends
am
onth
ofun
cert
aint
yov
erth
edi
rect
ion
ofm
onet
ary
polic
y.Pa
rtly
antic
ipat
edA
ppoi
ntm
ent
Ven
ezue
la19
94R
uth
deK
rivo
yY
esD
eK
rivo
yan
dot
her
boar
dm
embe
rsre
sign
over
Cen
tral
Ban
kin
depe
nden
ceco
ncer
ns.
Not
antic
ipat
edR
esig
natio
n
Ven
ezue
la19
94A
nton
ioC
asas
Gon
zale
z–
Cas
asis
appo
inte
dby
Pres
iden
tCal
dera
.N
otan
ticip
ated
App
oint
men
t
Ven
ezue
la20
00A
nton
ioC
asas
Gon
zale
zY
esD
iego
Lui
sC
aste
llano
sis
appr
oved
tofo
llow
Cas
asas
Cen
tral
Ban
kpr
esid
ent.
Part
lyan
ticip
ated
Res
igna
tion
Ven
ezue
la20
05D
iego
Lui
sC
aste
llano
sN
oG
asto
nPa
rra
Luz
uard
ois
prop
osed
toas
sum
eth
epo
stof
the
outg
oing
Cen
tral
Ban
kgo
vern
or.
Not
antic
ipat
edR
esig
natio
n
a The
exac
tdat
esar
eav
aila
ble
onre
ques
t.
1610 : MONEY, CREDIT AND BANKING
APPENDIX E: CENTRAL BANK GOVERNORS AND CONSERVATISM—SOMEEXAMPLES FROM THE PRESS
Country Date Name
Argentina 2001 Roque Maccarone President De La Rua (of the Alliance forWork, Justice and Education, which isconsidered to be a center party) dismissedPedro Pou on alleged bad administration in2001. The incoming Roque Maccarone wasindeed considered to be less conservativethan Pou (nominated by the conservativePartido Justicialista), as illustrated in TheMiami Herald (April 27, 2001, p. 2):“[Minister of the Economy] Cavallopublicly blamed the current 33-monthrecession on Pou’s failure to loosen liquidityrequirements for banks, saying the policyprevented businesses from obtainingmuch-needed credit. . . . Maccarone said hewill consider changing the level of reservescommercial banks are required to keep ondeposit with the Central Bank.”
Argentina 2002 Mario Blejer Maccarone suddenly resigned in 2002 andMario Blejer became his successor,appointed by the more conservative PartidoJusticialista. Again, the change in ideologyis reflected in the news. According to TheNew York Times (January 18, 2002, p. 1)“Mr. Maccarone’s replacement, MarioBlejer, is a former International MonetaryFund official with a good reputation amonginternational lenders.”
South Africa 1998 Tito Mboweni Another example that fits our coding is theappointment of Tito Mboweni in SouthAfrica in 1998. Being appointed by theAfrican National Congress—defining itselfas force of the left—he is coded to be lessconservative than his predecessor ChristianStals, who was nominated by theconservative National Party in 1989.According to the Guardian (July 7, 1998, p.12) Mboweni was perceived to be “generousto a fault to the unions. He has been heavilydependent on left-wing advisers at the laborministry.”
Turkey 2006 Durmus Yilmaz In Turkey, Durmus Yilmaz followed SureyyaSerdengecti in 2006. Serdengecti had beenappointed by the Democratic Left Party,while Yilmaz was appointed by the moreconservative Justice and Development Party.Consequently, Yilmaz is coded as moreconservative than Serdengecti. According tothe Washington Post (April 19, 2006, p. 5),Yilmaz was indeed perceived to beconservative at the time of his appointment:“Yilmaz’s appointment was seen as anassurance that the bank would not deviatefrom its tight monetary policy.”
CHRISTOPH MOSER AND AXEL DREHER : 1611
LITERATURE CITED
Anderson, Theodore, and Cheng Hsiao. (1982) “Formulation and Estimation of DynamicModels Using Panel Data.” Journal of Econometrics, 18, 47–82.
Backus, David, and John Driffill. (1985) “Inflation and Reputation.” American EconomicReview, 75, 530–38.
Beck, Thorsten, George Clarke, Alberto Groff, Philipp Keefer, and Patrick Walsh. (2001)“New Tools in Comparative Political Economy: The Database of Political Institutions.”World Bank Economic Review, 17, 165–76.
Berger, Helge, and Ulrich Woitek. (2005) “Does Conservatism Matter? A Time-Series Ap-proach to Central Bank Behaviour.” Economic Journal, 115, 745–66.
Berger, Helge, Jakob de Haan, and Silvester Eijffinger. (2001) “Central Bank Independence:An Update of Theory and Evidence.” Journal of Economic Surveys, 15, 3–40.
Brown, Keith, William Van Harlow, and Seha Tinic. (1988) “Risk Aversion, Uncertain Infor-mation, and Market Efficiency.” Journal of Financial Economics, 22, 355–85.
Caballero, Ricardo, and Arvind Krishnamurthy. (2001) “International and Domestic CollateralConstraints in a Model of Emerging Market Crises.” Journal of Monetary Economics, 8,513–48.
Caballero, Ricardo, and Arvind Krishnamurthy. (2002) “A Dual Liquidity Model of EmergingMarkets.” American Economic Review, 92, 33–37.
Campbell, John, Andrew Lo, and Craig MacKinlay. (1997) The Econometrics of FinancialMarkets. Princeton, NJ: Princeton University Press.
Cukierman, Alex, and Alan Meltzer. (1986) “A Theory of Ambiguity, Credibility, and Inflationunder Discretion and Asymmetric Information.” Econometrica, 54, 1099–1128.
Dreher, Axel, Bernhard Herz, and Volker Karb. (2006) “Is There a Causal Link betweenCurrency and Debt Crises?” International Journal of Finance and Economics, 11, 305–25.
Dreher, Axel, Michael Lamla, Sarah Lein, and Frank Somogyi. (2008) “The Impact of PoliticalLeaders’ Profession and Education on Reforms.” Journal of Comparative Economics, 37,169–93.
Dreher, Axel, Jan-Egbert Sturm, and Jakob de Haan. (2010) “When Is a Central Bank GovernorFired? Evidence Based on a New Data Set.” Journal of Macroeconomics, 32, 766–81.
Edwards, Sebastian. (1984) “LDC Foreign Borrowing and Default Risk: An Empirical Inves-tigation, 1976–80.” American Economic Review, 74, 726–34.
Eijffinger, Sylvester, and Marco Hoeberichts. (1998) “The Trade Off between Central BankIndependence and Conservativeness.” Oxford Economic Papers, 50, 397–411.
Gohlmann, Silja, and Roland Vaubel. (2007) “The Educational and Professional Backgroundof Central Bankers and its Effect on Inflation: An Empirical Analysis.” European EconomicReview, 51, 925–41.
Jones, Benjamin, and Benjamin Olken. (2005) “Do Leaders Matter? National Leadership andGrowth since World War II.” Quarterly Journal of Economics, 120, 835–64.
Kara, Hakan. (2007) “Monetary Policy under Imperfect Commitment: Reconciling Theorywith Evidence.” International Journal of Central Banking, 3, 149–77.
Kaminsky, Graciela L., and Carmen M. Reinhart. (1999) “The Twin Crises: The Causes ofBanking and Balance-of-Payments Problems.” American Economic Review, 89, 473–500.
1612 : MONEY, CREDIT AND BANKING
Kaminsky, Graciela L., and Sergio Schmukler. (2002) “Emerging Markets Instability: DoSovereign Ratings Affect Country Risk and Stock Returns.” World Bank Economic Review,16, 171–95.
Kuttner, Kenneth N., and Adam S. Posen. (2010) “Do Markets Care Who Chairs the CentralBank?” Journal of Money, Credit and Banking, 42, 347–71.
Kydland, Finn, and Edward Prescott. (1977) “Rules Rather Than Discretion: The Inconsistencyof Optimal Plans.” Journal of Political Economy, 85, 473–90.
Moser, Christoph. (2006) “The Impact of Political Risk on Sovereign Bond Spreads—Evidencefrom Latin America.” Mimeo, University of Mainz.
Pantzalis, Christos, David Stangeland, and Harry Turtle. (2000) “Political Elections and theResolution of Uncertainty: The International Evidence.” Journal of Banking and Finance,24, 1575–1604.
Rogoff, Kenneth. (1985) “The Optimal Degree of Commitment to an Intermediate MonetaryTarget.” Quarterly Journal of Economics, 100, 1169–89.
Santiso, Javier. (2003) The Political Economy of Emerging Markets. Actors, Institutions andFinancial Crises in Latin America. CERI Series in International Relations and PoliticalEconomy. New York: Palgrave Macmillan.
Schaumburg, Ernst, and Andrea Tambalotti. (2007) “An Investigation of the Gains fromCommitment in Monetary Policy.” Journal of Monetary Economics, 54, 302–24.