(8)Currency Crises and Institutions

download (8)Currency Crises and Institutions

of 21

Transcript of (8)Currency Crises and Institutions

  • 8/12/2019 (8)Currency Crises and Institutions

    1/21

    Currency crises and institutions

    Pattama L. Shimpalee a, Janice Boucher Breuer b,*

    a Faculty of Economics, Chiang Mai University, Chiang Mai 52000, Thailandb Department of Economics, Moore School of Business, University of South Carolina,

    Columbia, SC 29208, USA

    Abstract

    This study furthers recent literature on currency crises and institutions. The main objective is to re-

    evaluate the causes of currency crises by focusing on the role played by a broader array of institutional

    factors and crisis episodes than have previously been considered while at the same time controlling for

    economic factors. Our dataset consists of over 30 countries covering 13 institutional factors for the period

    1984e2002. Two questions are addressed. They are (1) what mix of institutions may contribute to or set

    the stage for a currency crisis? and (2) what mix of institutions may affect the depth of currency crises asmeasured by a decline in output? Our findings reveal that institutional as well as economic factors affect

    the probability of currency crises and that worse institutions are associated with bigger contractions in out-

    put during the crisis. In general, our strongest results regarding institutions show that corruption, a de facto

    fixed exchange rate regime, weak government stability, and weak law and order increase the probability of

    a currency crisis. We find mixed evidence that deposit insurance, the removal of capital controls, a lack of

    central bank independence, financial liberalization, and civil law increase the chance of crisis. We find

    a similar set of factors worsens the contraction in output during a crisis except for deposit insurance, which

    we find moderates the contraction in output.

    2005 Elsevier Ltd. All rights reserved.

    JEL classification:F3; F4

    Keywords:Currency crises; Institutions

    * Corresponding author. Tel.: 1 803 777 7400/7419.E-mail address: [email protected](J.B. Breuer).

    0261-5606/$ - see front matter 2005 Elsevier Ltd. All rights reserved.

    doi:10.1016/j.jimonfin.2005.10.008

    Journal of International Money and Finance 25 (2006) 125e145www.elsevier.com/locate/econbase

    mailto:[email protected]://www.elsevier.com/locate/econbasehttp://www.elsevier.com/locate/econbasemailto:[email protected]
  • 8/12/2019 (8)Currency Crises and Institutions

    2/21

    1. Overview of the issue

    Are currency crises ordinary events? A review of international monetary history seems to

    suggest perhaps so. From 1972e2002, there have been 169 currency crises across 56 countries.1

    Currency crises have occurred in all but three of those years and have shocked developing,emerging, and developed countries alike, with the most recent examples being the Argentine

    crisis of 2001e2002 and the Turkish crisis of 2001e2002. No country seems immune and in-

    deed, the media portrays the United States as poised for its own currency crisis.

    By definition, crises are bad outcomes. In economic terms, crises take a toll on countries in

    terms of reduced income and increased unemployment. Corporate balance sheets are adversely

    affected and bankruptcies may occur. Banks may become illiquid or worse, insolvent. Payments

    of households and businesses become onerous and credit becomes circumscribed. Economic

    growth is interrupted. As well, political and social unrest may be ignited. All in all, citizens

    suffer. The economic, financial, political, and social fallout a currency crisis brings make un-

    derstanding their causes critical.

    A proliferation of theoretical and empirical papers investigating the causes and consequences

    of crises as well as how to guard against them has been sparked by the cluster of currency crises

    since the mid-1990s. Indeed, there has been much work devoted to predicting crises.2 Culling

    extant studies of currency crises, a shortlist of economic factors contributing to currency crises

    emerges. These include real exchange rate overvaluation, a lack of foreign reserve adequacy

    relative to short-term debt or broad money, domestic credit growth, current account deficits,

    poor export growth, and declining foreign reserves. While these studies show poor macroeco-

    nomic fundamentals as causes of currency crises, they leave open the question of the role of

    institutions. Weak institutions may contribute to poor macrofundamentals and hence predisposea country to sudden stops of capital. It is also possible that institutions, in addition to macroeco-

    nomic fundamentals, may contribute to crises.

    In two related papers,Alesina and Wagner (2003)andCalvo and Mishkin (2003) consider

    the quality of institutions and exchange rate arrangements. Alesina and Wagner (2003)found

    that countries with poor institutional quality related to the business environment and the so-

    cio-political environment, have difficulty in maintaining an announced peg and are more likely

    to abandon it. Calvo and Mishkin (2003),in re-assessing the debate over fixed versus floating

    exchange rates, argue that deeper institutional features relating to fiscal stability, financial sta-

    bility, and price stability are more important to macroeconomic stability and the avoidance of

    crises than the exchange rate regime, per se.This study furthers the literature on currency crises and institutions. The main objective is to

    re-evaluate the causes of currency crises by focusing on the role played by a broader array of

    institutional factors and crisis episodes than have previously been considered, while controlling

    for economic factors. Since good macroeconomic performance and good institutions often go

    hand in hand, it is important to control for economic factors in order to correctly estimate the

    1 Calculations based onBordo et al. (2001) and updated by authors.2

    Contributions in this vein includeEichengreen et al. (1995), Sachs et al. (1996), Frankel and Rose (1996), Goldfajnand Valdes (1997), International Monetary Fund (1998), Milesi-Ferretti and Razin (1998), Berg and Pattillo (1999),

    Bussiere and Mulder (1999), Kaminsky and Reinhart (1999), Rossi (1999), Aziz et al. (2000), Edison (2000), Goldstein

    et al. (2000), Hawkins and Klau (2000), Johnson et al. (2000),Osband and Van Rijckeghem (2000), Kamin et al. (2001),

    Block (2002), Ghosh and Ghosh (2002), Kumar et al. (2002), Martinez-Peria (2002), andMulder et al. (2002).

    126 P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

  • 8/12/2019 (8)Currency Crises and Institutions

    3/21

    contribution of institutions to currency crises. Our dataset consists of over 30 countries covering

    13 institutional factors for the period 1984e2002.

    Two questions are addressed. They are (1) what mix of institutions may contribute to or set

    the stage for a currency crisis, controlling for macroeconomic factors? and (2) what mix of in-

    stitutions may affect the depth of currency crises as measured by a decline in output, controllingfor macroeconomic factors? Along the way, we offer hypotheses relating institutions to currency

    crises.

    Our findings reveal that institutional as well as economic factors affect the probability of

    currency crises and that worse institutions are associated with bigger contractions in output dur-

    ing the crisis. In general, our strongest results regarding institutions show that corruption, a de

    facto fixed exchange rate regime, weak government stability, and weak law and order increase

    the probability of a currency crisis. We find ambiguity over deposit insurance e in some cases,

    deposit insurance increases the probability of currency crises and in others, reduces it. We find

    mixed evidence that the removal of capital controls, a lack of central bank independence, finan-

    cial liberalization, and civil law increase the chance of crisis. We find little evidence that bu-

    reaucratic quality, ethnic tensions, external conflict, and internal conflict are important

    factors in currency crises. We find a similar set of factors worsens the contraction in output dur-

    ing a crisis except for deposit insurance, which we find moderates the contraction in output.

    The outline of the paper is as follows: Section2reviews the literature on currency crises,

    paying particular attention to the latest work on institutions. Section 3 presents the data and

    the estimation methodology. Section4 presents the empirical results. Section5offers conclud-

    ing remarks and directions for future work.

    2. Background on models of currency crises

    The question of what causes crises has been addressed with sequential generations of models

    developed largely along historical lines e first, to explain the sovereign debt crisis of Latin

    America, next to explain the European and Mexican crises, and on to explain the Asian crisis.3

    In first generation models, poor macroeconomic fundamentals incite speculative capital out-

    flows, which in turn generate a currency crisis. In second generation models, speculative capital

    outflows are triggered when announced policy and the credibility that it can be maintained are

    called into question. Third generation models are based on a boom/bust (overlending/overbor-

    rowing) cycle and model currency crises as co-terminus with banking crises.

    Fourth generation models are relatively new and introduce institutional factors (elsewheretermed social capital, social infrastructure, deep determinants) as determinants of currency

    crises. In these models, weak institutions worsen problems associated with risk and uncertainty

    and contribute to a misallocation of resources thereby setting the stage for currency crises.Ra-

    jan and Zingales (1998) consider contract enforcement and the opportunity for malfeasance;

    Bussiere and Mulder (1999)examine, e.g. political factors such as divisive and polarized parlia-

    ments;Rossi (1999)considers capital account openness, bank supervision, and depositor safety;

    Johnson et al. (2000)consider a number of variables including rule of law, judicial efficiency,

    and corruption; Li and Inclan (2001) consider central bank independence, coordinated wage

    bargaining, stock-market development, and more;Acemoglu et al. (2002) consider constraints

    on the executive branch; Block (2002) considers the strength of the government, Ghosh and

    3 SeeBreuer (2004)for a review.

    127P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    4/21

    Ghosh (2002) consider governance, rule of law, creditor and shareholder rights; and Mulder

    et al. (2002)consider among other factors, the legal regime, contract enforcement, and account-

    ing standards.

    Generally speaking, institutions affect currency crises through two causal mechanisms (Li

    and Inclan, 2001). First, institutions tend to have an impact and correlate with the health ofthe national economy. Therefore, institutions that lead to bad economic fundamentals may con-

    tribute to currency crises whereas institutions that help produce good economic fundamentals

    remove one reason for currency crises to occur. Second, institutions are informative. Institutions

    signal market agents about future economic fundamentals, and can thereby shape market ex-

    pectations. Consequently, institutions that correlate with bad economic fundamentals destabi-

    lize market expectations, increase market uncertainty about the likelihood of currency crisis,

    and make currency crises motivated by speculative capital outflows more likely. On the other

    hand, institutions that correlate with good economic conditions stabilize market expectations,

    reduce market uncertainty about the probability of currency crisis, and make speculative capital

    outflows less likely.

    3. Data, hypotheses, and estimation methodology

    We next present a set of hypotheses and an empirical model relating institutions to the prob-

    ability of currency crises and the depth of contraction in output suffered during them. We then

    discuss the data.

    3.1. Hypotheses on institutional variables

    Table 1presents the 13 institutional variables considered in our study. The third column of

    the table presents the hypothesized (directional) effect of these institutions on the probability of

    a currency crisis. We briefly summarize the institutions and offer hypotheses about each.

    3.1.1. Bureaucratic quality

    Bureaucratic quality measures the strength and quality of civil service and bureaucrats and

    how able they are to manage political problems without interruption of services.4 A higher level

    of bureaucratic quality means that government services and policies are less likely be interrup-

    ted and/or altered and that agencies are less likely to be influenced by political pressure. Con-

    sequently, there will be less uncertainty with respect to the conduct of government and lessuncertainty with respect to economic outcomes. With less uncertainty, capital outflows may

    not be as subject to panic and herding. Thus, currency crises may be less likely to arise with

    greater bureaucratic quality. Since a higher index value indicates greater bureaucratic quality,

    the expected sign of bureaucratic quality is negative.

    3.1.2. Government stability

    A higher level of government stability means that a government is more likely to be able to

    continue its announced programs and to stay in office. Factors such as type of governance, the

    command of the legislature, and popular approval of policies are considered.4 Government sta-

    bility consequently leads to less uncertainty as to what government policy toward businesses

    4 Source: International Country Risk Guide. PRS Group.

    128 P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

  • 8/12/2019 (8)Currency Crises and Institutions

    5/21

    (banking and non-banking) will be in the future. With less uncertainty, there is less likely to be

    a misallocation of resources and associated inefficiencies. So, government stability strengthens

    the economy. Thus, a higher degree of government stability is less likely to lead to capital flightand thus it is less likely that currency crises will arise. Also, with a higher degree of government

    stability and therefore less uncertainty, panic selling of currency may be less likely to arise.

    Since a higher index value indicates more government stability the expected sign of govern-

    ment stability is negative.5

    3.1.3. Corruption

    Corruption includes the demand for bribes connected with import and export licenses, ex-

    change controls, tax assessments, police protection and loans. It also includes nepotism and cro-

    nyism.4 Higher levels of corruption are more likely to lead to inefficient economic decisions

    Table 1

    Institutional variables: hypotheses and data sources

    Variable Index value Effect on probability

    of currency crisis or

    depth of contraction

    Data source Frequency

    Bureaucratic quality 0e4, Higher values betterbureaucratic quality

    Negative International Country

    Risk Guide

    Monthly

    Government stability 0e12 Higher values moregovernment stability

    Negative International Country

    Risk Guide

    Monthly

    Absence of

    corruption

    0e6, Higher values lesscorruption

    Negative International Country

    Risk Guide

    Monthly

    Law and order 0e6, Higher values morelaw and order

    Negative International Country

    Risk Guide

    Monthly

    Absence of ethnic

    tensions

    0e6, Higher values lessethnic tensions

    Negative International Country

    Risk Guide

    Monthly

    Absence of external

    conflict

    0e12, Higher values lessconflict

    Negative International Country

    Risk Guide

    Monthly

    Absence of internal

    conflict

    0e12, Higher values lessconflict

    Negative International Country

    Risk Guide

    Monthly

    Exchange rate

    regime

    0, 1 Where 1 fixed rateregime

    Ambiguous Reinhart and Rogoffs

    (2002) de facto exchange

    rate regime

    Annual

    Capital controls 0, 1 Where 1 capitalcontrols

    Ambiguous Annual report on

    exchange rate

    arrangements and

    restrictions, IMF

    Annual

    Central bank

    independence

    0e1 Where 1 maximum

    independence

    Negative Cukierman (1992)and

    Cukierman et al. (1992)

    By country

    Deposit insurance 0, 1 Where 1 depositinsurance

    Ambiguous Barth et al. (2001) By country

    Financial

    liberalization

    Proxied by the real interest

    rate on deposits

    Ambiguous Constructed from

    International Financial

    Statistics

    Monthly

    Legal origin 0, 1 Where 1 civil law Ambiguous La Porta et al. (1998) By country

    5 Since a higher index value for government stability may also be associated with autocratic regimes and an absence

    of democracy and voice, it is possible that the expected sign could be positive.

    129P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

  • 8/12/2019 (8)Currency Crises and Institutions

    6/21

    and to a greater misallocation of resources. Inefficiencies and misallocation of resources put the

    economy at greater risk for suffering poor economic outcomes (be it in the banking sector,

    manufacturing sector, government sector, export and import sector, etc.). Since poor economic

    outcomes lead to capital flight, higher levels of corruption may increase the likelihood of cur-

    rency crises. Corruption can also make more uncertain the outcome of contracts and transac-tions which may increase the propensity for panic selling of assets. Since a higher index

    means less corruption, the expected sign of corruption on the probability of currency crises

    is negative.

    3.1.4. Law and order

    Law and order measures the strength and impartiality of the legal system and popular obser-

    vance of the law.4 A higher degree of law and order means that not only is there greater ob-

    servance of the law by the populace but also that the judicial system is fair and impartial. Thus,

    a higher degree of law and order also implies less uncertainty in all types of transactions. Con-

    tractual obligations are more likely to be fulfilled according to the terms of the agreement and

    the judicial system is more likely to settle cases fairly. By reducing uncertainty in transactions,

    there is less likely to be a misallocation of resources and fewer inefficient outcomes. Thus, law

    and order strengthens an economy in ways that may not be directly observable in macroeco-

    nomic performance and makes it less likely to be subject to capital flight and currency crises.

    Since a higher number indicates higher observance of law and order, the expected sign of law

    and order is negative.

    3.1.5. Ethnic tensions

    Ethnic tensions are attributed to racial, nationality, or language division and gauge how in-tolerant groups might be to compromise.4 Ethnic tensions may affect the way in which contrac-

    tual agreements between two parties of different ethnic backgrounds are arranged as well as

    how likely they are to be abided. Ethnic tensions thus raise the level of uncertainty in the econ-

    omy for all types of transactions related to commerce, borrowing and lending, government pol-

    icy toward business, and so on, particularly in principal-agent relationships. A higher level of

    ethnic tensions may thus increase the level of uncertainty in business and financial transactions.

    This may lead to inefficient outcomes and/or outcomes that have a negative impact on the econ-

    omy. Thus, capital flight and consequently currency crises may be more likely to arise in coun-

    tries with greater ethnic tensions. Since a higher number indicates lower ethnic tensions, the

    expected sign of this coefficient is negative.

    3.1.6. External conflict

    External conflict ranges from trade restrictions and embargoes to geopolitical disputes to in-

    cursions, insurgencies, and warfare. A higher level of external conflict means that there is more

    uncertainty in international transactions with the domestic country. A higher level of external

    conflict thus may make a country more subject to capital flight and more likely to experience

    currency crises. Since a higher index means lower external conflict, the expected sign of this

    coefficient is negative.

    3.1.7. Internal conflictInternal conflict is based on the extent of political violence toward the incumbent. A higher

    level of internal conflict reduces the willingness of parties to abide by contracts and heed prop-

    erty rights and thus raises the level of uncertainty in all types of transactions. A higher level of

    130 P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

  • 8/12/2019 (8)Currency Crises and Institutions

    7/21

    uncertainty leads to more inefficient outcomes and a greater misallocation of resources which

    can weaken the economy. This can cause capital flight. The higher level of uncertainty can also

    make capital flight more likely. Thus, currency crises may be more likely the higher the level of

    internal conflict. Since a higher index indicates lower internal conflict, the expected sign of this

    coefficient is negative.

    3.1.8. Exchange rate regime

    In the aftermath of the recent financial crises, a view emerged that the exchange rate regime

    was in part responsible for the likelihood and the depth of these crises. In other words, a number

    of emerging market economies have experienced devastating financial crises and macroeco-

    nomic turbulence because they had kept exchange rates fixed. It is possible that an exchange

    rate peg may encourage borrowers to ignore exchange rate risk. Thus, a fixed exchange rate

    regime may contribute to macroeconomic instability and hence currency crises. A counterargu-

    ment is that a fixed exchange rate regime promotes monetary discipline and thus improves long

    run macroeconomic performance and helps stabilize expectations. We code fixed exchange rate

    regimes with a 1, and 0 otherwise; however, the expected sign on this coefficient is ambiguous.

    3.1.9. Capital controls

    The classic view such as Greenwald et al. (1984), Quirk and Evans (1995), and Cooper

    (1998)argues that free capital mobility enhances a more efficient allocation of resources raising

    welfare in the process. Controls on inflows seem to hamper economic performance. Thus, cap-

    ital controls may predispose a country to currency crisis. On the other hand,McKinnon and Pill

    (1997)argue that it is also possible that the absence of controls encourages overlending and

    overborrowing which can put a country at risk for a currency crisis. Thus, the expected signof capital controls is ambiguous. We code controls on capital account transactions, or restric-

    tions on capital movements, especially inflows with a 1, and 0 otherwise.

    3.1.10. Central bank independence

    Central bank independence may improve real economic performance and the avoidance of

    currency crises for several reasons. For example, an independent central bank that is free from

    political pressure may behave more predictably, promoting economic stability and reducing

    a risk premium in real interest rates. Moreover, to the extent that high inflation has adverse ef-

    fects on economic performance by creating distortions, encouraging rent seeking activity, or

    raising a risk premium, one would expect central bank independence to improve economic per-formance. In other words, central bank independence leads to lower inflation rates and greater

    price stability; it contributes to the long run health of national economy and removes one im-

    portant cause of speculative attacks against ones currency. Thus, the more independent a central

    bank, the lower the probability of currency crises. Therefore, the expected sign of central bank

    independence is negative. We use the index of legal central bank independence taken from

    Cukierman (1992), andCukierman et al. (1992). It is coded on a scale between 0 and 1 where

    0 stands for the minimum level of independence and 1 for the maximum level.

    3.1.11. Deposit insurance

    The effect of deposit insurance on currency crises is ambiguous. On the one hand, depositinsurance offered to banks reduces the downside risk of depositor losses and hence banking cri-

    ses which in turn, decreases the probability of currency crises. On the other hand, deposit in-

    surance offered to banks can set the stage for overborrowing/overlending which can lead to

    131P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

  • 8/12/2019 (8)Currency Crises and Institutions

    8/21

    currency crises. Since we code countries with explicit deposit insurance protection systems with

    a 1, and 0 otherwise, the expected sign of deposit insurance is ambiguous.

    3.1.12. Financial liberalization

    There are different opinions about the role of financial liberalization in currency crises.Some argue that financial liberalization increases the allocation efficiency of the financial sector

    laying the foundation for sounder macroeconomic fundamentals. Consequently, financial liber-

    alization should reduce the likelihood of currency crises. Others argue that financial liberaliza-

    tion, by encouraging competition in the financial sector, can lead to imprudent lending practices

    which lower the stability of the banking and corporate sectors and thus make currency crises

    more likely. Consequently, the effect of financial liberalization on currency crises is ambiguous.

    We use a real interest rate as a proxy for financial liberalization as in Rossi (1999). The real

    interest rate can be interpreted a proxy for financial liberalization since real interest rates are

    usually lower, or negative, in repressed financial systems (Rossi, 1999).

    3.1.13. Legal origin

    La Porta et al. (1998) argue that civil law countries afford less protection to creditors and

    shareholders and thus increase financial sector instability and the propensity for currency crises.

    On the other hand, in civil law countries, judicial decisions are determined according to a cod-

    ified set of laws whereas in common law countries, cases are decided on precedent. Thus, civil

    law countries may reduce risk and uncertainty and thus indirectly reduce the likelihood of cur-

    rency crises. Consequently, the effect of legal origin (civil law) on currency crises is ambigu-

    ous. Countries with a civil law system are coded as 1 and common law countries as 0.

    The hypotheses relating institutions to the probability of currency crisis are the same regard-ing the depth of the currency crisis. That is, a worse set of institutions should correspond to

    larger contractions in output during crisis.

    3.2. Estimation methodology

    The two questions we raised in Section 1 require different estimation strategies be em-

    ployed. To answer the first question relating institutional factors to the probability of a currency

    crisis, we use a multivariate probit. The basic equation is:

    Pr

    Currency Crisisi;t 1FXi;t1b

    ;

    1

    where Pr (Currency Crisisi,t) denotes the probability of a currency crisis in monthtin countryi,

    and takes a value of 1 at the onset of the crisis and 0 otherwise. Xi, t1is a vector of explanatory

    variables (at time t 1) partitioned to include economic factors and institutional factors; F isthe standard cumulative normal distribution. The panel of countries used in estimating Eq.(1)

    includes those that experienced one or more crises over the sample period as well as some that

    did not experience any crisis over the entire period. Consequently, we control for heteroskedas-

    ticity and report heteroskedastic-consistentp-values.

    The second question we address is whether economic and institutional factors systematically

    affect the depth or severity of a currency crisis. To answer it, a measure of the depth or severityof a currency crisis must be constructed. We measure the depth of the crisis (DC) by an index of

    a trend forecast of output (per capita) relative to actual output using annual instead of monthly

    data. Increases in DC thus imply an increase in the depth of the crisis and allow us to interpret

    132 P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

    http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    9/21

    the coefficients on the institutional variables similarly to those hypothesized for the probability

    of crisis. The equation we use is:

    DCi;t b0 b1X1;i;t1 b2X2;i;t1 wi;t; 2

    where DCi,t is the depth of the crisis. The subscripts i and tindicate the country and the crisis num-

    ber for that country. For example, there are four crises in Indonesia in November 1978, April 1983,

    September 1986, and August 1997. Thus, DCi,4 would indicate the fourth crisis for country i.

    Note that since crisis events are not uniformly distributed over time, the tsubscript is not an

    indicator of time (e.g. months or years). Thus, the data no longer have a time-series dimension.

    Moreover, the panel used in estimating Eq. (2) includes only countries that have suffered a currency

    crisis.X1,i,t1is a vector of economic variables and X2,i,t1is a vector of institutional variables.

    The set of explanatory variables is the same as the set used for the probit models.

    In order to test empirically the connection between institutional variables and the depth of

    a currency crisis, controlling for economic factors, we estimate several specifications that differ

    in their treatment of the error term. First, we use ordinary least squares (OLS). This is the sim-

    plest treatment and does not admit the possibility that there are country-specific effects or that

    the error terms across crises may be correlated.

    To account for the possibility that the depth of the crises may differ by country, we use two

    estimation methodologies that introduce country-specific effects. These are the random effects

    estimator (RE) and the between estimator (BE), respectively. We cannot use fixed effect esti-

    mation since several of our institutional variables are country-specific.

    Third, we re-estimate Eq. (2)using what is best described as pseudo seemingly unrelated

    regression (SUR). SUR estimation has the advantage of allowing error terms across observationsto be correlated. In our case, we allow the error terms across crises within a country to be cor-

    related, rather than at a point in time as is usual with standard SUR applied to panel data. The

    estimation strategy is not a straightforward application of SUR for a few reasons. First, the data

    on the depth of the currency crises are neither cross-sectional nor time series. The data are ob-

    servations on the size of the contraction in trend forecast of output per capita relative to actual

    output per capita during a crisis. Since countries historical experience with crises differs by date

    and number, these data do not have a true time-series dimension. Second, in the estimation, we

    restrict the coefficient estimates to be identical across countries and crisis events. Standard SUR

    with panel data permits the coefficient estimates to differ by cross-sectional unit.

    Lastly, because the methods above may suffer from sample selection since only crisis epi-sodes were included in theestimation, we apply Heckmans two-step selection across both cri-

    sis and non-crisis episodes.6 Naturally, this dramatically increases the sample size since there

    are many years in which no crises are experienced by countries. In all estimations, heteroske-

    dasticity has been corrected.

    3.3. Data

    Our data consist of monthly (and annual) observations from January 1984 to December 2002

    for 35 or 44 countries, depending on data availability and the method used to date the crises,

    discussed below. The starting date for the sample period was constrained by data availability on

    6 We thank Roberto Rigobon for suggesting this.

    133P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

  • 8/12/2019 (8)Currency Crises and Institutions

    10/21

    some institutions. Institutional variables are drawn from several sources as listed in Table 1.

    The economic data are from the IMFs International Financial Statistics CD-ROM.

    A few points about the data are worth noting. First, we identify a currency crisis using three

    alternative techniques. Currency crises are identified followingEichengreen et al. (1995), Ka-

    minsky and Reinhart (1999), and Frankel and Rose (1996). Eichengreen et al. (1995) use anindex computed from a weighted average of exchange rate changes, interest rate changes,

    and foreign reserve changes to identify crises whereas Kaminsky and Reinhart (1999) use

    a weighted average of exchange rate changes and foreign reserve changes. Frankel and Rose

    (1996)use an index based on exchange rate changes only. Each of the techniques requires con-

    structing an index of speculative exchange market pressure based on (log) exchange rate

    changes, (log) foreign reserve changes, and/or interest rate changes. When the index breaches

    a threshold value, the start date of a crisis episode (date) is identified. Since the exact identifi-

    cation of crisis start datesis not always the same across techniques, we present results for all

    three dating procedures.7

    Second, the data on institutions from International Country Risk Guide are index values

    where, in all cases, higher index values are associated with better institutions. For example,

    the higher the score on corruption, the less corruption there is. The same applies to the data

    on ethnic tensions, internal conflict, and external conflict. This is why we have modified these

    variable names inTable 1with the term absence of..

    Third, the exchange rate regime classification we use is the de facto classification produced

    byReinhart and Rogoff (2002)instead of the IMFs de jure classification. The IMF classifica-

    tion may misrepresent exchange rate arrangement type because it is based on the arrangement

    type countries self-report to the IMF. These may be different from the exchange rate regim e

    countries maintain in practice. Reinhart and Rogoffs data attempt to adjust for this problem.

    8

    Fourth, eight macroeconomic variables are also included in the estimation to control for

    macroeconomic factors that may contribute to a currency crisis independent of institutional fac-

    tors. The economic control variables included are domestic credit/real GDP, exports/real GDP,

    foreign reserves/real GDP, the inflation rate, M2/foreign reserves, the real exchange rate, the

    trade balance/real GDP, and the U.S. interest rate.

    4. Currency crises and institutions: empirical results

    In this section, the effects of institutions are measured in two dimensions e on the proba-

    bility of currency crisis, and on the depth of contraction in output during the crisis. We turnto each of these next.

    4.1. Institutions and the probability of currency crisis

    Tables 2e4report results of the probit analysis applied to Eq. (1)using the three different

    techniques for identifying currency crises (discussed in Section3.3). We report results across all

    7 UsingEichengreen et al. (1995),we identify 67 crisis episodes across 11 countries in Asia, 15 in Europe, and 9 in

    the Latin America; withKaminsky and Reinhart (1999),we identify 68 episodes across 11 countries in Asia, 16 in Eu-rope, and 17 in the Latin America; and withFrankel and Rose (1996), we identify 55 episodes across 11 countries in

    Asia, 16 in Europe, and 17 in the Latin America.8 Based on Kaminsky and Reinharts crisis identification method, we identified 55 crisis episodes under a fixed ex-

    change rate regime and 13 crisis episodes under a floating rate regime.

    134 P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

    http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    11/21

    Table 2

    Probit estimates of probability of currency crisis, January 1984eDecember 2002 (Eichengreen et al. (1995) method)

    Variable Full model

    a

    Models excluding some variables

    a

    (1) (2) (3) (4) (5)

    Bureaucratic quality 0.115 (0.281) e 0.111 (0.290) 0.094 (0.363) e

    Government stability 0.051* (0.067) 0.046* (0.071) 0.053* (0.066) 0.053* (0.066) 0.048* (0.074Absence of corruption 0.074** (0.023) 0.049** (0.045) 0.077** (0.023) 0.101** (0.015) 0.051** (0.04Law and order 0.067** (0.039) 0.055** (0.046) 0.061** (0.044) 0.065** (0.039) 0.051** (0.05Absence of ethnic

    tensions

    0.023 (0.694) 0.020 (0.743) e 0.031 (0.560) e

    Absence of external

    conflict

    0.036 (0.344) 0.037 (0.321) 0.038 (0.309) 0.032 (0.408) 0.039 (0.296)

    Absence of internal

    conflict

    0.019 (0.331) 0.015 (0.338) 0.012 (0.429) e 0.008 (0.534)

    Exchange rate regime 0.301* (0.058) 0.256 (0.071) 0.312* (0.042) 0.240* (0.074) 0.266* (0.068)

    Capital controls 0.316* (0.098) 0.326* (0.087) 0.309* (0.103) 0.235 (0.206) 0.320* (0.091Central bank

    independence

    0.076 (0.209) 0.053 (0.341) 0.079 (0.188) 0.068 (0.260) 0.056 (0.308)

    Deposit insurance 0.196* (0.090) 0.195* (0.090) 0.210* (0.087) 0.097* (0.098) 0.123* (0.095Financial liberalization 0.004 (0.521) 0.003 (0.521) 0.002 (0.527) 0.001 (0.547) 0.004 (0.521)Legal origin 0.349* (0.094) 0.311 (0.126) 0.325* (0.104) 0.271 (0.187) 0.293* (0.135)

    Number of countries 35 35 35 35 35

    Number of crises 67 67 67 67 67

    Number of observations 7980 7980 7980 7980 7980

    Pseudo R2 0.406 0.393 0.384 0.377 0.382

    Note: currency crises are identified using the Eichengreen et al. (1995)method and takes a value of 1 at the start of crises. The tab

    by 100. Numbers in parentheses are p-values. A dash (e) denotes the variable was dropped from the analysis.

    *Significant at the 10% marginal significance level; **significant at the 5% marginal significance level; ***significant at the 1a Eight macroeconomic control variables (see Section3.3) and year dummies are included in the estimation but not reported

    http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    12/21

    Table 3

    Probit estimates of probability of currency crisis, January 1984eDecember 2002 (Kaminsky and Reinhart (1999)method)

    Variable Full modela Models excluding some variablesa

    (1) (2) (3) (4) (5)

    Bureaucratic quality 0.025 (0.735) e 0.026 (0.731) 0.021 (0.773) e Government stability 0.054** (0.028) 0.056** (0.023) 0.054** (0.029) 0.051** (0.037) 0.056** (0.02Absence of corruption 0.086 (0.162) 0.094* (0.096) 0.082 (0.172) 0.094 (0.123) 0.091* (0.102Law and order 0.037** (0.035) 0.040** (0.030) 0.039** (0.035) 0.029** (0.047) 0.042** (0.03Absence of ethnic

    tensions

    0.013 (0.750) 0.014 (0.745) e 0.026 (0.527) e

    Absence of external

    conflict

    0.001 (0.981) 0.001 (0.979) 0.000 (0.993) 0.006 (0.809) 0.000 (0.992)

    Absence of internal

    conflict

    0.035 (0.270) 0.034 (0.276) 0.037 (0.219) e 0.037 (0.223)

    Exchange rate regime 0.011** (0.052) 0.009* (0.055) 0.011** (0.053) 0.017** (0.050) 0.009* (0.056) Capital controls 0.058 (0.671) 0.047 (0.723) 0.066 (0.623) 0.036 (0.788) 0.056 (0.673)Central bank

    independence

    0.105* (0.077) 0.112* (0.065) 0.105* (0.077) 0.104* (0.077) 0.121** (0.05

    Deposit insurance 0.060** (0.048) 0.068* (0.045) 0.046** (0.052) 0.034* (0.072) 0.053* (0.057Financial liberalization 0.019 (0.879) 0.022 (0.860) 0.010 (0.934) 0.004 (0.957) 0.013 (0.915)Legal origin 0.067* (0.055) 0.054* (0.068) 0.054* (0.068) 0.092** (0.035) 0.040* (0.071)

    Number of countries 44 44 44 44 44

    Number of crises 68 68 68 68 68

    Number of observations 10,032 10,032 10,032 10,032 10,032

    Pseudo R2 0.410 0.409 0.409 0.394 0.407

    Note: currency crises are identified using the Kaminsky and Reinhart (1999)method and takes a value of 1 at the start of crisemultiplied by 100. Numbers in parentheses are p-values. A dash (e) denotes the variable was dropped from the analysis.

    *Significant at the 10% marginal significance level; **significant at the 5% marginal significance level; ***significant at the 1a Eight macroeconomic control variables (see Section3.3) and year dummies are included in the estimation but not reported

    http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    13/21

    Table 4

    Probit estimates of probability of currency crisis, January 1984eDecember 2002 (Frankel and Rose (1996) method)

    Variable Full modela Model excluding some variablesa

    (1) (2) (3) (4) (5)

    Bureaucratic quality 0.077 (0.350) e 0.076 (0.352) 0.073 (0.373) e Government stability 0.068** (0.019) 0.072** (0.012) 0.067** (0.020) 0.062** (0.029) 0.071** (0.012)Absence of corruption 0.089** (0.027) 0.113*** (0.012) 0.088** (0.027) 0.085*** (0.010) 0.118*** (0.010Law and order 0.050** (0.023) 0.063** (0.015) 0.052** (0.023) 0.027** (0.042) 0.064** (0.015)

    Absence of ethnic

    tensions

    0.017 (0.713) 0.016 (0.719) e 0.001 (0.986) e

    Absence of external

    conflict

    0.033* (0.077) 0.032 (0.117) 0.029 (0.180) 0.021 (0.183) 0.032* (0.104)

    Absence of internal

    conflict

    0.059 (0.106) 0.059 (0.111) 0.056 (0.114) e 0.056 (0.119)

    Exchange rate

    regime

    0.031* (0.085) 0.011* (0.097) 0.013* (0.094) 0.046* (0.074) 0.023* (0.097)

    Capital controls 0.189 (0.226) 0.158 (0.301) 0.178 (0.245) 0.154 (0.317) 0.148 (0.324)Central bank

    independence

    0.029 (0.533) 0.035 (0.452) 0.030 (0.515) 0.014 (0.759) 0.036 (0.436)

    Deposit insurance 0.235** (0.044) 0.209* (0.058) 0.224** (0.051) 0.285** (0.033) 0.199* (0.060)

    Financial

    liberalization

    0.006* (0.057) 0.004* (0.057) 0.007** (0.054) 0.009** (0.052) 0.005* (0.057)

    Legal origin 0.243 (0.203) 0.289 (0.120) 0.225 (0.222) 0.206 (0.276) 0.270 (0.129)

    Number of countries 44 44 44 44 44

    Number of crises 55 55 55 55 55

    Number of

    observations

    10,032 10,032 10,032 10,032 10,032

    Pseudo R2 0.407 0.392 0.404 0.384 0.370

    Note: currency crises are identified using the Frankel and Rose (1996)method and takes a value of 1 at the start of crises. The tab

    by 100. Numbers in parentheses are p-values. A dash (e) denotes the variable was dropped from the analysis.

    *Significant at the 10% marginal significance level; **significant at the 5% marginal significance level; ***significant at the 1a Eight macroeconomic control variables (see Section3.3) and year dummies are included in the estimation but not reported

    http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    14/21

    three strategies as a check on robustness. In each, the sample period covers January 1984 eDe-

    cember 2002 and the equation is estimated with year dummies and eight economic control var-

    iables listed in Section3.3. The macroeconomic variables are not included in the tables since

    our focus is on the effect of institutions on the probability of currency crises. 9

    Table 2presents the results usingEichengreen et al. (1995)s crisis definition.Table 3reportsresults following Kaminsky and Reinhart (1999)s crisis definition, and Table 4 reports results us-

    ingFrankel and Rose (1996)s crisis definition. Country coverage varies by crisis definition with

    35 countries in Table 2 and 44 countries in Tables 3 and 4. In each of the tables, results from seven

    different specifications are shown. The seven different specifications are each pared down ver-

    sions of the full (first) specification where several variables are excluded from the estimation if

    they consistently turned out to be statistically insignificant across tables. Model (2) omits bureau-

    cratic quality while model (3) omits ethnic tensions and model (4) omits internal conflict. Models

    (5), (6) and (7) omit bureaucratic quality and ethnic tensions, bureaucratic quality and internal

    conflict, and bureaucratic quality, ethnic tensions, and internal conflict, respectively.

    Overall, the results suggest that controlling for economic factors, several institutions are asso-

    ciated with an increased probability of currency crises. Looking across Tables 2e4,the results

    uniformly show that corruption, a de facto pegged exchange rate regime, a worsening of govern-

    ment stability and a worsening of law and order increase the probability of a currency crisis.10

    These results are robust across the three techniques used to date crises. The results show some

    differences regarding deposit insurance.Tables 2 and 3show that deposit insurance reduces the

    likelihood of a crisis andTable 2 also shows some evidence that capital controls reduce the likeli-

    hood of crisis although this result is not uniform across the seven specifications. Results fromTa-

    ble 3also show that the probability of a currency crisis increases with a decline in central bank

    independence and that civil law countries are more prone to crises than common law countries.In contrast toTables 2 and 3,Table 4shows that deposit insurance now raises the probability

    of a currency crisis. Results fromTable 4also differ from those inTable 2by adding financial

    liberalization as a significant factor in increasing the probability of a currency crisis. However,

    the coefficient estimate has the wrong sign. The result may be due to the method used to iden-

    tify currency crises. TheFrankel and Rose (1996)estimation method relies only on a criteria

    based on exchange rate changes to date currency crisis episodes. Since countries typically em-

    ploy interest rate increases to defend a currency in crisis, and because the financial liberaliza-

    tion measure is the real interest rate, it is not surprising that the coefficient is significant. The

    negative sign suggests that an interest rate defense can work to reduce the probability of crisis.

    In all tables and specifications, there is not much support for the hypotheses that weakbureaucratic quality, ethnic tensions, external conflict, and internal conflict increase the prob-

    ability of currency crises.

    9 InTables 2e4,six of eight economic variables (not reported) are statistically significant at the 10% level or better.

    ForTable 2, they are the ratio of domestic credit to real GDP, the ratio of exports to real GDP, the ratio of foreign re-

    serves to real GDP, inflation rate, the real exchange rate, and the U.S. interest rate. M2/foreign reserves and the trade

    balance/real GDP is not. For Table 3, M2/foreign reserves and the trade balance/real GDP gain significance and ex-

    ports/real GDP and the U.S. interest rate lose significance. ForTable 4,exports/real GDP gain significance and the tradebalance/real GDP loses significance.10 Recall that corruption, government stability, and law and order are measured as indexes where increases imply an

    improvement or better outcome. For example, an increase in the corruption index implies less corruption and thus

    should reduce the probability of a currency crisis. So, the expected coefficient sign is negative.

    138 P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

    http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    15/21

    4.2. Institutions and the depth of currency crises

    Tables 5e7report results from the estimation of Eq.(2), using the three different techniques

    for identifying currency crises (discussed in Section3.3). Because construction of the depth of

    crisis measure is limited by annual data, results reported in Tables 5e

    7are based on annual-ized observations during crisis episodes only. As discussed in Section 3.3, the data are not panel

    data. Again, we report results across all three strategies for identifying crises as a check on

    robustness. In each, the equation is estimated with the same economic control variables (not

    reported).11 We present results for four different estimation methods applied to Eq. (2).

    These are OLS, random effects, between effects, and pseudo-SUR which were discussed in

    Section3.2.

    Table 5presents the results usingEichengreen et al. (1995)s crisis definition.Table 6reports

    results followingKaminsky and Reinhart (1999)s crisis definition, andTable 7reports results

    usingFrankel and Rose (1996)s crisis definition.Overall, the results suggest that controlling for economic factors, many of the same institu-

    tions that increase the probability of a currency crisis also increase the depth of the crisis (as

    measured by a loss in output relative to trend). Looking acrossTables 5e7, several institutions

    are consistently associated with the depth of the crisis. These are corruption, deposit insurance,

    government stability and law and order. More widespread corruption (measured by a decrease

    in the corruption index), lower levels of government stability, and lower levels of law and order

    increase the depth of the contraction in output during the crisis, whereas deposit insurance re-

    duces the size of the contraction. The findings for deposit insurance further the debate over it;

    we find that while deposit insurance increases the probability of a currency crisis, it moderates

    the loss in output during the crisis.Results fromTables 5 and 6also show that civil law countries suffer bigger declines in out-

    put than common law countries.Tables 5e7provide modest evidence that countries with more

    independent central banks suffer smaller contractions in output during crisis. There is also mod-

    est evidence that de facto fixed exchange rate regimes are likely to experience bigger contrac-

    tions in output and that capital controls correspond to a smaller loss in output during a crisis.

    Lastly, there is modest evidence that a lessening of ethnic tensions and internal conflict (a rise

    in the index value) corresponds to a smaller decline in output.

    5. Conclusions and directions for future research

    The last two decades have seen a proliferation of currency crises among countries worldwide

    with a reinforced interest in them among academicians and policymakers. However, only re-

    cently has there been a major shift in focus away from macroeconomic fundamentals toward

    understanding the deeper institutional determinants of crises. While crises are directly initi-

    ated by large scale capital outflows, it begs the question what causes large scale capital out-

    flows? Risk and uncertainty are certainly fundamental but this only further begs the question

    what determines risk and uncertainty? The newest research answers with institutions. In-

    stitutions are broadly conceived but in general can be framed as economic, financial, political,

    legal, or social arrangements, explicit or implicit, that affect expectations, and hence risk and

    11 See Section3.3.

    139P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    16/21

    Table 5

    Depth of contraction in output during crises (Eichengreen et al. (1995) method)

    Variable Full modela

    OLS Random effects Between effects Pse

    Bureaucratic quality 0.007 (0.606) 0.007 (0.543) 0.024 (0.677) 0.0Government stability 0.037* (0.059) 0.057** (0.017) 0.016** (0.052) 0Absence of corruption 0.018* (0.084) 0.017* (0.074) 0.020* (0.077) 0Law and order 0.015* (0.077) 0.011* (0.090) 0.038* (0.059) 0Absence of ethnic tensions 0.001 (0.995) 0.002 (0.802) 0.004 (0.901) 0Absence of external conflict 0.001 (0.774) 0.005 (0.152) 0.013 (0.345) 0.0Absence of internal conflict 0.006 (0.286) 0.010** (0.040) 0.020 (0.499) 0

    Exchange rate regime 0.005 (0.282) 0.019* (0.077) 0.021* (0.069) 0.0Capital controls 0.015** (0.037) 0.020** (0.044) 0.011* (0.080) 0Central bank independence 0.005 (0.954) 0.014* (0.064) 0.017 (0.960) 0Deposit insurance 0.073** (0.016) 0.081*** (0.003) 0.023* (0.079) 0Financial liberalization 0.002 (0.886) 0.003 (0.821) 0.001 (0.610) 0Legal origin 0.064*** (0.003) 0.068*** (0.001) 0.070*** (0.001) 0.0

    Number of countries 28 28 28 28

    Number of observations 67 67 67 67

    R2 0.452 0.477 0.425 0.4

    Note: dependent variable is a trend forecast of output per capita divided by actual output per capita during crisis episodes. See Se

    estimation. Crisis episodes are identified using the Eichengreen et al. (1995) method.

    *Significant at the 10% marginal significance level; **significant at the 5% marginal significance level; ***significant at the 1a Eight macroeconomic control variables (see Section3.3) are included in the estimation but not reported. See also Footnote

    http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    17/21

    Table 6Depth of contraction in output during crises (Kaminsky and Reinhart (1999)method)

    Variable Full modela

    OLS Random effects Between effects Pse

    Bureaucratic quality 0.007 (0.300) 0.004 (0.463) 0.013 (0.464) 0Government stability 0.021** (0.052) 0.031* (0.062) 0.025* (0.065) 0Absence of corruption 0.045** (0.024) 0.028** (0.040) 0.057** (0.045) 0Law and order 0.005* (0.091) 0.004* (0.069) 0.013** (0.053) 0Absence of ethnic tensions 0.003 (0.237) 0.004 (0.239) 0.005 (0.343) 0.0

    Absence of external conflict 0.003 (0.168) 0.000 (0.970) 0.006 (0.145) 0Absence of internal conflict 0.001 (0.491) 0.002 (0.887) 0.004 (0.915) 0

    Exchange rate regime 0.005 (0.282) 0.019* (0.077) 0.021* (0.069) 0.0

    Capital controls 0.006* (0.065) 0.003 (0.660) 0.011 (0.695) 0Central bank independence 0.029* (0.083) 0.037* (0.072) 0.016* (0.092) 0Deposit insurance 0.015* (0.083) 0.022** (0.046) 0.048** (0.038) 0Financial liberalization 0.007 (0.253) 0.005 (0.232) 0.003 (0.274) 0Legal origin 0.028** (0.016) 0.033*** (0.013) 0.034*** (0.011) 0.0

    Number of countries 38 38 38 38

    Number of observations 68 68 68 68

    R2 0.471 0.548 0.490 0.4

    Note: dependent variable is a trend forecast of output per capita divided by actual output per capita during crisis episodes. See Se

    estimation. Crisis episodes are identified using the Kaminsky and Reinhart (1999) method.

    *Significant at the 10% marginal significance level; **significant at the 5% marginal significance level; ***significant at the 1a Eight macroeconomic control variables (see Section3.3) are included in the estimation but not reported. See also Footnote

    http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    18/21

    Table 7

    Depth of contraction in output during crises (Frankel and Rose (1996) method)

    Variable Full modela

    OLS Random effects Between effects Pseu

    Bureaucratic quality 0.011 (0.421) 0.008 (0.463) 0.020 (0.636) 0.Government stability 0.038* (0.089) 0.045** (0.053) 0.054** (0.041) 0.Absence of corruption 0.012*** (0.011) 0.027** (0.045) 0.032** (0.037) 0.Law and order 0.011* (0.078) 0.015* (0.079) 0.039** (0.047) 0.Absence of ethnic tensions 0.005 (0.923) 0.009 (0.897) 0.003 (0.839) 0.Absence of external conflict 0.002 (0.662) 0.007** (0.047) 0.005 (0.548) 0.Absence of internal conflict 0.008 (0.864) 0.004 (0.305) 0.004 (0.722) 0.

    Exchange rate regime 0.018* (0.064) 0.007* (0.073) 0.009* (0.069) 0.01Capital controls 0.002 (0.906) 0.016* (0.061) 0.015* (0.075) 0.00Central bank independence 0.110* (0.096) 0.075 (0.506) 0.169 (0.631) 0.Deposit insurance 0.008* (0.058) 0.018** (0.024) 0.014** (0.029) 0.Financial liberalization 0.006* (0.081) 0.012* (0.085) 0.017* (0.079) 0.00Legal origin 0.011 (0.621) 0.018 (0.446) 0.015 (0.923) 0.03Number of countries 31 31 31 31

    Number of observations 55 55 55 55

    R2 0.422 0.441 0.453 0.41

    Note: dependent variable is a trend forecast of output per capita divided by actual output per capita during crisis episodes. See Se

    estimation. Crisis episodes are identified using the Frankel and Rose (1996) method.

    *Significant at the 10% marginal significance level; **significant at the 5% marginal significance level; ***significant at the 1a Eight macroeconomic control variables (see Section3.3) are included in the estimation but not reported. See also Footnote

    http://-/?-http://-/?-
  • 8/12/2019 (8)Currency Crises and Institutions

    19/21

    uncertainty, in transactions. Their effects may be heightened for international transactions,

    making them particularly important to understanding currency crises.

    We find the notion institutions matter as plausible and so undertake research that seeks to

    determine how so with regard to understanding currency crises. We consider 13 institutional

    factors ranging from law and order to legal origin to ethnic tensions to deposit insurance to cen-tral bank independence and investigate their effects on the probability of currency crises. How-

    ever, we also control for macroeconomic factors which are widely accepted as playing some

    role in the genesis of crises, or may themselves depend on institutions. We also consider

    how institutions affect the depth of the contraction in output during a crisis. The set of institu-

    tions we consider is longer than in previous studies and we offer hypotheses about each. Our

    work also covers a wider set of countries ranging from developing to developed and those

    with fixed and floating exchange rates.

    Our findings offer evidence that institutions, in addition to macroeconomic factors, do indeed

    matter to the onset of currency crises and the contraction in output that ensues. Using a multi-

    variate probit analysis, we find consistent support that the following institutions increase the

    probability of currency crisis: less stable governments; weak law and order; more widespread

    corruption; and a de facto fixed exchange rate regime. There is some ambiguity over the results

    regarding deposit insurance which vary with the type of method used to identify crisis episodes.

    We find more modest support that capital controls and central bank independence reduce the

    likelihood of crises, and that civil law countries are more prone to crises. We find little to no

    evidence that bureaucratic quality, ethnic tensions, external conflict, or internal conflict matter

    to currency crises. Using panel estimation techniques, we find, for the most part, that these

    same factors worsen the contraction in output during the crisis, save for deposit insurance

    which reduces the contraction.Several other questions remain open for future research. First, the out-of-sample perfor-

    mance of the probit models is not tested in this study. This is left for future research. Second,

    we think it is possible that nonlinearities are important. We wonder whether the breaching of

    certain threshold levels for institutions, separately or jointly, increases theprobability of crisis.

    Another issue that could be explored is the duration of fixed rate systems.12 The importance of

    economic and institutional factors to explaining how long a country is able to maintain a fixed

    rate may be worthwhile to investigate.

    Acknowledgements

    Thanks to participants at the JIMF/CRIF/TAFI conference in San Juan, Puerto Rico and

    especially to Roberto Rigobon. Research generously supported by a grant from the Center

    for International Business Education and Research.

    References

    Acemoglu, D., Johnson, S., Robinson, J., Thaicharoen, Y., 2002. Institutional causes, macroeconomic symptoms: vol-

    atility, crises, and growth. NBER Working Paper No. 9124, National Bureau of Economic Research, Cambridge,

    Massachusetts.

    12 SeeKlein and Marion (1997) and Allsopp (2000).

    143P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

  • 8/12/2019 (8)Currency Crises and Institutions

    20/21

    Alesina, A., Wagner, A., 2003. Choosing (and reneging on) exchange rate regimes. NBER Working Paper No. 9809.

    Allsopp, L., 2000. A model to explain the duration of a currency crisis. International Journal of Finance and Economics

    5, 331e337.

    Aziz, J., Caramazza, F., Salgado, R., 2000. Currency crises: in search of common elements. IMF Working Paper WP/00/

    67, Washington D.C.

    Barth, J.R., Caprio Jr., G., Levine, R., 2001. The regulation and supervision of banks around the world: a new database.World Bank Working Paper No. 2588.

    Berg, A., Pattillo, C., 1999. Predicting currency crises: the indicators approach and an alternative. Journal of Interna-

    tional Money and Finance 18, 561e586.

    Block, S.A., 2002. Political conditions and currency crises: empirical regularities in emerging markets. CID Working

    Paper No. 79.

    Bordo, M., Eichengreen, B., Klingebiel, D., Martinez-Peria, S., 2001. Is the crisis problem growing more severe?

    Economic Policy 24, 51e82.

    Breuer, J.B., 2004. An exegesis on currency and banking crises. Journal of Economic Surveys 18, 293e320.

    Bussiere, M., Mulder, C., 1999. Political instability and economic vulnerability. IMF Working Paper WP/99/46,

    Washington D.C.

    Calvo, G.A., Mishkin, F.S., 2003. The mirage of exchange rate regimes for emerging market countries. Journal of

    Economic Perspectives 17, 99e118.

    Cooper, R.N., 1998. Should capital-account convertibility be a world objective? In: Kenen, P.B. (Ed.), Should the IMF

    Purse Capital-Account Convertibility. Essays in International Finance, vol. 207. International Finance Section,

    Department of Economics, Princeton University.

    Cukierman, A., Webb, S.B., Neyapti, B., 1992. Measuring the independence of central banks and its effect on policy

    outcomes. The World Bank Economic Review 6, 353e398.

    Cukierman, A., 1992. Central Bank Strategy, Credibility, and Independence: Theory and Evidence. The MIT Press,

    Cambridge, Massachusetts, London, England.

    Edison, H.J., 2000. Do indicators of financial crises work: an evaluation of an early warning system. International

    Finance Discussion Papers No. 675, Board of Governors of the Federal Reserve System, Washington D.C.

    Eichengreen, B., Rose, A., Wyplosz, C., 1995. Exchange market mayhem: the antecedents and aftermath of speculative

    attacks. Economic Policy 21, 249e

    312.Frankel, J., Rose, A.K., 1996. Currency crashes in emerging markets: an empirical treatment. Journal of International

    Economics 41, 351e366.

    Ghosh, S., Ghosh, A., 2002. Structural vulnerabilities and currency crises. IMF Working Paper WP/02/09, Washington

    D.C.

    Goldfajn, I., Valdes, R.O., 1997. Are currency crises predictable? IMF Working Paper WP/97/159, Washington D.C.

    Goldstein, Morris, Kaminsky, G.K., Reinhart, C.M., 2000. Assessing financial vulnerability: an early warning

    system for emerging markets. Unpublished Working Paper, Institute for International Economics, Washington

    D.C.

    Greenwald, B., Stiglitz, J., Weiss, A., 1984. Information imperfections in the capital markets and macroeconomic fluc-

    tuations. American Economic Review 74, 194e199.

    Hawkins, J., Klau, M., 2000. Measuring potential vulnerabilities in emerging market economies. BIS Working Paper

    No. 91.International Monetary Fund, 1998. Financial crises: characteristics and indicators of vulnerability. World Economic

    Outlook. International Monetary Fund, Washington D.C. (Chapter IV).

    Johnson, S., Boone, P., Breach, A., Friedman, E., 2000. Corporate governance in the Asian financial crisis. Journal of

    Financial Economics 58, 141e186.

    Kamin, S., Schindler, J., Samuel, S., 2001. The contributions of domestic and external factors to emerging market de-

    valuation crises: an early warning systems approach. Board of Governors of the Federal Reserve System, Interna-

    tional Finance Discussion Paper No. 711.

    Kaminsky, G., Reinhart, C., 1999. The twin crises: the causes of banking and balance-of-payments problems. American

    Economic Review 89, 473e500.

    Klein, M.W., Marion, N.P., 1997. Explaining the duration of exchange rate pegs. Journal of Development Economics 54,

    387e404.

    Kumar, M., Moorthy, U., Perraudin, W., 2002. Predicting emerging market currency crashes. IMF Working Paper WP/

    02/7.

    La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 1998. Law and finance. Journal of Political Economy 106,

    1113e1155.

    144 P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145

  • 8/12/2019 (8)Currency Crises and Institutions

    21/21

    Li, Q., Inclan, M., 2001. Fundamentals, expectations, institutions, and currency crisis. Prepared for Presentation at the

    Annual Meeting of the American Political Science Association. 20.

    Martinez-Peria, M.S., 2002. A regime switching approach to studying speculative attacks: a focus on EMS crises. Em-

    pirical Economics 27, 299e334.

    McKinnon, R., Pill, H., 1997. Credible economic liberalization and over borrowing. American Economic Review 87,

    189e

    193.Milesi-Ferretti, G., Razin, A., 1998. Current account reversals and currency crises: empirical regularities. IMF Working

    Paper WP/98/99, Washington D.C.

    Mulder, C., Perrelli, R.A., Rocha, M.D.S., 2002. The role of corporate, legal and macroeconomic balance sheet indi-

    cators in crisis detection and prevention. IMF Working Paper WP/02/59, Washington D.C.

    Osband, K., Van Rijckeghem, C., 2000. Vulnerability to currency crises. IMF Staff Papers 47, 238e258.

    Quirk, P., Evans, O., 1995. Capital account convertibility: review of experience and implications for IMF policies. IMF

    Occasional Paper No. 131, Washington D.C.

    Rajan, R.G., Zingales, L., 1998. Which capitalism? Lessons from the East Asian crisis. Journal of Applied Corporate

    Finance 11, 40e48.

    Reinhart, C., Rogoff, K.S., 2002. The modern history of exchange rate arrangements: a reinterpretation. NBER Working

    Paper No. 8963, Cambridge, Massachusetts.

    Rossi, M., 1999. Financial fragility and economic performance in developing economies: do capital controls, prudential

    regulation and supervision matter? IMF Working Paper WP/99/66, Washington D.C.

    Sachs, J., Tornell, A., Velasco, A., 1996. Financial crises in emerging markets: the lessons from 1995. Brookings Papers

    on Economic Activity 16, 147e215.

    145P.L. Shimpalee, J.B. Breuer / Journal of International Money and Finance 25 (2006) 125e145