exchange rate volatility and democratization in emerging market countries

Exchange Rate Volatility and Democratization in Emerging Market CountriesAuthor(s): Jude C. Hays, John R. Freeman and Hans NessethSource: International Studies Quarterly, Vol. 47, No. 2 (Jun., 2003), pp. 203-228Published by: Wiley on behalf of The International Studies AssociationStable URL: http://www.jstor.org/stable/3693542 .

Accessed: 14/06/2014 02:35

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

Wiley and The International Studies Association are collaborating with JSTOR to digitize, preserve and extendaccess to International Studies Quarterly.

http://www.jstor.org

This content downloaded from 185.2.32.109 on Sat, 14 Jun 2014 02:35:54 AMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=black

http://www.jstor.org/action/showPublisher?publisherCode=isa

http://www.jstor.org/stable/3693542?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


International Studies Quarterly (2003) 47, 203-228

Exchange Rate Volatility and Democratization in Emerging Market

Countries

JUDE C. HAYS

University of Michigan

JOHN R. FREEMAN

University of Minnesota

HANS NESSETH

University of Minnesota

We examine some of the consequences of financial globalization for democratization in emerging market economies by focusing on the currency markets of four Asian countries at different stages of democratic development. Using political data of various kinds-including a new events data series-and the Markov regime switching model from empirical macroeconomics, we show that in young and incipient democracies politics continuously causes changes in the probability of experiencing two different currency market equilibria: a high volatility "contagion" regime and a low volatility "fundamentals" regime. The kind of political events that affect currency market equilibration varies cross-nationally depending on the degree to which the polity of a country is democratic and its policymaking transparent. The results help us better gauge how and the extent to which democratization is compatible with financial globalization.

Much has been written in recent years about the economic benefits of globalization. However, scholars have begun to question whether globalization, particularly the internationalization of financial markets, is compatible with democracy. They have

expressed concern that global markets impose costs and constraints on governments in ways that limit the possibilities for democratic politics (Rodrik, 2000).1

Authors' note: Earlier versions of this article were presented on panels at the 2001 meetings of the American Political Science Association and the International Studies Association, and at research seminars at the University of

Colorado, New York University, the University of North Texas, and Yale University. We thank participants on these

panels and in these seminars for comments and criticisms. We especially appreciate the input of William Bernhard, David Leblang, and Tom Willet. For the data used in gauging the impact of extragovernmental political behavior on financial markets we thank Doug and Joe Bond.

1Arguing by analogy to the Mundell-Flemming conditions for open economies Rodrik (2000) contends that,

given the pace and scope of international market integration, countries have to choose between a "golden straight jacket" in which there is little room for government to maneuver and global federalism. Only the latter, in his mind, allows for any meaningful form of mass politics or popular sovereignty over the economy. He calls this the "political trilemma."

( 2003 International Studies Association. Published by Blackwell Publishing, 350 Main Street, Malden, MA 02148, USA, and 9600 Garsington Road, Oxford OX4 2DQ, UK.



204 Exchange Rate Volatility and Democratization in Emerging Market Countries

COST OF CHAOS Political uncertainty has undermined the rupiah

Rupiah/$ 6,500 7,000 Religious omb blast

confrontation atstock 7,500 - in Ambon exchange

8,500' _East Timor unrest

and presidential Tv 9,000 u: `--cc' lI ..." election. ?l 9,500 -Uncertainty before

10,000 the june elections Wahid's impeachment rumour

10,5001 2 0 J F M A M ASO N DJ F•M IM J J A S ON 1999 2000

Source: Far Eastern Economic Review, Nov. 16, 2000.

FIG. 1.

One of the most significant developments in the globalization of finance has been the unprecedented increase in capital flows to developing countries (IMF, 1997). Theoretically, these flows have been linked to contagion and volatility in the financial markets of emerging market economies (Calvo and Mendoza, 2000). And there is considerable anecdotal evidence that periods of financial instability are triggered by political events in young and incipient democracies. Take Southeast Asia, for example. Analysts are convinced that revelations about corruption in the Estrada administration and uncertainty about Estrada's ability to survive an impeachment trial caused volatility in the Philippine peso, and conversely that the smooth presidential transition from Estrada to Arroyo produced positive reactions in the Philippine currency and stock markets.2 In the case of Indonesia, uncertainty about the secession of provinces, the survival of the president, and other political factors presumably were responsible for the "chaos" in the rupiah-dollar exchange rate (Figure 1). This issue is important because "chaos" in foreign exchange markets is costly-currency volatility reduces trade, which diminishes national income, and contributes to the outbreak of financial crises like those that swept through Asia in 1997-and if domestic politics is a contributing cause of this volatility, then there is a potentially serious tension between democratization in emerging market countries and financial globalization.3

But is domestic politics responsible for currency market volatility in the emerging market economies? The answer is not clear. Significant progress has been made in understanding how uncertainty stemming from elections, legislative politics, and other political variables affects the behaviors of currency, equity, and bond markets. This research has shown that certain political institutions--for example, particular kinds of electoral systems--mitigate the effects of political uncertainty on global markets in countries like Germany.4 But, to date, most of this work has been

2 See, e.g., Estrada Impasse, Far Eastern Economic Review, November 23, 2000, p. 80; and After the B Movie, A New Main Attraction to Filipinos, The Economist, January 27, 2001, p. 37.

3 On the link between currency volatility and (reduced) trade see Rose (2000); see also Leblang (2001). The

relationship between trade and income (growth) is studied in such works as Frankel and Roemer (1999). The Asian financial crisis had a devastating impact on several of the region's economies. Annual growth rates in Thailand, South Korea, and Indonesia went from above 5% before the crisis to below -5% in 1998 (Mishkin, 1999).

4 Representative of the burgeoning literature on politics and financial markets in countries in North America and Europe are Roberts (1990), Alesina, Roubini, and Cohen (1997), Bernhard and Leblang (2001, 2002), Freeman, Hays, and Stix (2000), Herron et al. (1999), Herron (2000a, 2000b), Leblang (2001), Leblang and Bernhard (2000b) and Lin and Roberts (2001).



JUDE C. HAYS ET AL. 205

focused on long-standing democracies in North America and Europe rather than on young and incipient democracies like those in the Philippines and Indonesia.

The research that has been done on young and incipient democracies-for instance, the important book by Haggard (2000)-argues that because of the uncertainty they create about the government's ability to generate mass support and manage (cooperate with) its opposition, less democratic and opaque political institutions produce more market uncertainty and hence their respective societies experience comparatively greater social costs in financial crises. For example, the costs of the crisis of 1997 were greater in Indonesia than in South Korea and Thailand because of the greater uncertainties created for markets by Indonesia's nondemocratic institutions. Unfortunately, Haggard and others (e.g., MacIntyre, 2001) focus mostly on this crisis period and do not analyze the continuous effect domestic politics may have on currency markets and hence on the economy. Nor do they characterize or establish empirically the causal mechanism that links political uncertainty and currency market behavior in young and incipient democracies.5

These are the purposes of this article. Using political data of various kinds-including a new events data series-and the Markov regime switching model from empirical macroeconomics, we show that, in young and incipient democracies, politics causes continuous changes in the probability of observing different market equilibria (or regimes), one of which connotes high currency volatility and contagion and hence, by implication, reduced trade, investment, and national income. The kind of politics that affects currency market equilibration varies cross-nationally depending on the degree to which a country is democratic and its policymaking transparent. In young democracies like Thailand and South Korea, which have relatively established democratic institutions and processes, the probability of shifts to and from "contagion" regimes depends on familiar variables like the degree of legislative support for the executive. On the other hand, when democratic institutions are less developed and policymaking is more opaque, as in Indonesia, the probability of shifts to harmful, volatile currency market equilibria depends on extragovernmental politics such as riots and other forms of social violence. Countries like the Philippines represent intermediate cases; both intra- and extragovernmental politics affect the probability of experiencing harmful currency market volatility in these countries.

The discussion is divided into three parts. The first part briefly reviews the relevant political science and economic literatures. It finds in the former some testable propositions about which political variables should affect market behavior and identifies in the latter the Markov switching model as a useful framework to study the impact of domestic politics on international currency markets. Using data from four Asian countries at different stages of democratic development and the model from economics, the propositions are evaluated in the second part of the paper. The debate about the compatibility of globalization and democratization is revisited in light of the findings in the conclusion.

Politics and Currency Markets

A handful of political scientists have studied the connections between domestic politics and international financial markets, including currency markets. They have used a variety of research designs for this purpose and their work provides valuable insights into how currency traders respond to politics in young and incipient democracies. For instance, Bernhard and Leblang (2002) show that politics influences the relationship between spot and forward exchange rates. During electoral campaigns, cabinet negotiations, and cabinet dissolutions, forward

5 The work of Haggard, MacIntyre, and others on the political economy of Asian and other developing democracies is discussed in greater detail below.




exchange rates are a biased predictor of future exchange rates. Similarly, Frieden (2002) found that electoral periods and indicators of the relative size of certain sectors of the economy--indicative of the power of particular interest groups -affect average annual rates of depreciation and coefficients of variation in exchange rates in fourteen OECD countries.6

Research on politics and security markets yields complementary findings. Illustrative is Lin and Roberts's (2001) demonstration that returns for certain stocks were affected by expectations about the outcome of the March 2000 presidential election in Taiwan. Also, Pantzalis, Strangeland, and Tuttle (2000) discovered that stock price movements--patterns of strongest abnormal returns -occur in the weeks immediately preceding but not following elections. This is because in the final weeks of the campaign the resolution of political uncertainty -especially in closed, or "less free" countries-causes stock prices to rise.7

The main implication of these findings is that currency traders continuously monitor and react to information about the workings of young and incipient democracies, particularly information about electoral politics and government formation (maintenance). Electoral outcomes and the popularity of elected officials affect the course of economic policy. Therefore, where political polls are conducted regularly-for example, in the Philippines--they will have a continuous impact on currency trading and on the possibility of currency volatility. Who controls key ministries also affects economic policy and so the information traders receive about cabinet negotiations (survival) will also affect their behavior.

As regards the possibly mitigating effects of institutions, Freeman, Hays, and Stix (2000) have highlighted the importance of electoral systems. They studied the effects of political uncertainty on the currencies of four developed economies -Australia, Germany, United Kingdom, and the United States-and found that their measures of political uncertainty helped to identify low and high volatility exchange rate regimes in every case but Germany. They concluded that this was due to the political constraints arising from German coalition politics, which, in turn, were traced to Germany's electoral system. Generalizing, they argue that the proportional representation (PR) election systems associated with consensual forms of democracy tend to buffer currency markets from the effects of political uncertainty; where PR is used, information about elections and cabinet stability does not affect currency market behavior.

By contrast, in majoritarian, winner-take-all democracies with first-past-the-post electoral systems, elections are more likely to produce significant changes in economic policy, which makes the effect of political uncertainty on their currency markets more pronounced. To the extent that this familiar distinction-consensual versus majoritarian (Lijphart, 1999)-applies to young democracies, our expectations are clear. Thailand and other young democracies that have PR electoral systems should, like their older democratic counterparts (e.g., Germany), not experience any impact of electoral politics and cabinet stability on currency volatility.

6 Bernhard and Leblang (2002) use dummy variables to represent the effect of campaign periods and cabinet

politics. In a companion study, the same authors (Leblang and Bernhard, 2000) employ an EGARCH set-up using dummy variables in the conditional variance expression. Among twenty-five elections, eight have effects on the conditional variance in the log difference of the exchange rates. Lobo and Tufte (1998) report similar results for a GARCH model of currency movements. Frieden (2002) uses the annual average change and the coefficient of variation in the nominal exchange rate as his dependent variable and employs a simple, pooled cross-section, time series linear regression model in his analysis.

7 Lin and Roberts (2001) regress individual stock returns on the daily Iowa Electronic Markets contract for Chen in the weeks leading up to the March 2000 election and find that a number of stock returns are affected by this

political variable. Pantzalis et. al. (2000) argue that the asymmetry of information in closed ("less free") political systems means that elections resolve more uncertainty than in open ("free") political systems, especially when the incumbent is defeated.




But what of countries like Indonesia where democracy is just taking root? In incipient democracies--that is, countries where democratic institutions are emerging but the polity retains a significant degree of authoritarianism-there is a greater possibility of coups and democratic reversals. Hence political uncertainty is often times of a different character and perhaps greater magnitude than in young democracies where the political institutions are new but fully consolidated. Research on these transition cases is more qualitative and suggestive than quantitative and empirical. It suggests that the key determinants of market uncertainty are the lack of (1) political transparency or "information symmetry" between currency traders and policymakers, and (2) government capacity to generate mass support and manage its opposition.

Transparency means traders can be relatively certain about the content and aims of economic policy, and therefore better able to anticipate macroeconomic performance, than where policymaking is more opaque. Of course political transparency is a key correlate of democracy so there should be a direct relationship between the degree of democratic development and the amount of currency market volatility we observe in countries' exchange rates.8

The second factor has to do with governments' abilities to implement their policies. Opposition from parties not in the government and from the general population can undermine the effectiveness of economic and other kinds of policies. Uncertainty about the depth of such opposition makes it much harder for traders to anticipate macroeconomic performance. Traders therefore are more susceptible to speculation and contagion creating market volatility and financial crises (Haggard, 2000:51).

The implication is that, in more opaque, incipient democracies, social unrest -especially riots and other forms of violent opposition to government policy-is indicative of uncertainty about the content and effectiveness of government policy. Where democracy is less developed as in Indonesia, electoral data and assessments of cabinet stability may not have an impact on currency volatility; extragovernmental factors like riots are more likely to be expressions of opposition to government policy and therefore social unrest should be more likely to spawn currency volatility. In contrast, electoral polls and intragovernmental variables such as those that reflect the degree of legislative support for the executive will affect currency volatility in young democracies like South Korea and Thailand; in these cases, social unrest may not be indicative of opposition to government policies and hence it may not have any impact on currency markets.9

These then are our general expectations about the way politics is linked to currency market behavior within and across young and incipient democracies. But what causal mechanism is at the heart of this linkage? Political science has few answers to this question. The economics literature is more useful in this regard.

8 For instance, Haggard (2000:16ff) contends, "nontransparent political relationships between politicians and particular firms contributed to the economic vulnerabilities [and risk associated with the South Asian financial crisis of the late 1990s]." One of the main arguments of his book is that this crisis "showed the democracies [of that region] to be resilient and has enhanced the cause of economic reform in the region. But ... reforms will also require a parallel process of deepening democracy by enhancing accountability and transparency of government" (3). Political risk analysts also consider transparency and democracy to be less prone to market volatility; see, e.g., Howell (2001:24, 230).

9 In a number of places in his book, Haggard (2000:48, 91, 168) stresses that nonelectoral or unconventional forms of political participation like social violence deeply constrain governments' abilities to cope with financial crises. Again, he argues that Indonesia's less democratically developed government was comparatively the least able to cope with such participation. In contrast, in countries like South Korea, political parties openly and constitutionally forged an agreement that reassured financial traders (51, 100ff).

Of course, it need not be the case that extragovernmental politics affects currency volatility in incipient democracies only and intragovernmental politics affects currency volatility in young democracies only. The differences across young and incipient democracies are likely to be differences of degree rather than kind. It is entirely plausible that extra- and intragovernmental politics affects currency volatility in both young and incipient democracies.




a. Economic Understandings of Currency Markets

Empirical research in international economics has produced a number of findings about currency price dynamics. For example, currency returns are nonstationary (unit root property), display periods of marked turbulence and quiescence (clustering property), and exhibit extreme values more often than one would expect if the series were normally distributed (fat tail property). (See DeVries, 1994.) A formalism that accounts for these properties is the Markov regime- switching model (Hamilton, 1989; Engel and Hamilton, 1990). This model holds that currency markets can be in two or more states. The states correspond to distinct dynamic market equilibria, or "regimes" which are represented by different statistical models, for instance, two random walks with different drifts and (or) variances. The exchange rate series we observe is a mixture of these regimes, the proportions of which are determined by probabilistic transitions between states. With a first-order Markov switching process, the probability that a currency market is in either regime at time t is a function of the regime it was in at time t - 1. One body of research argues that changes in economic fundamentals render the transition probabilities between two market regimes time varying (Weinbach, 1993;Diebold, Lee, and Weinbach, 1994).

The Markov switching model with time varying transition probabilities is potentially very useful. It is superior econometrically to the linear regression models used by Bernhard and Leblang (2002) , Frieden (2002), and other political scientists because, with the regimes expressed as random walks, the Markov switching model can account for both the observed nonstationarity in currency returns and the fat-tailed distributions in these returns (Meese, 1990:129-130; see also Kim and Nelson, 1999). Linear regression models do not account for these properties. As a result, hypothesis tests based on the estimates from linear regression models are not as accurate as those based on Markov regime switching models.10

This is not to say that economists' applications of the model are adequate for our purposes. With a few notable exceptions discussed below, the economic research concentrates on market fundamentals and eschews political analysis. Applications of the Markov switching model in economics usually make no provision for the possibility that regime change is precipitated by political factors of any kind. Those who apply the model in economics do not recognize that the economic fundamentals on which the transition probabilities depend may be mediated by, if not the result of, political variables (e.g., Weinbach, 1993). The idea that, in young and incipient democracies, social violence and other kinds of challenges to government authority might be a source of currency regime switching is nowhere to be found in the economic literature.

The Markov switching model with time varying probabilities thus offers an econometrically powerful framework with which to represent the causal mechanism whereby currency markets respond to politics in young and incipient democracies. But to realize its potential we need to give it a "politically relevant" interpretation and translate our theoretical expectations into testable propositions about regime switching.

10 Space does not permit a methodological critique of existing work on democracy and financial markets. Like

most empirical studies in political science, Frieden's (2002) measures, model, and diagnostics do not address let alone capture the three key properties of exchange rates. Hence the hypothesis tests reported in his study are

problematic (e.g., the residuals on which they are based are non-normal). Bernhard and Leblang's (2002)

investigation addresses the unit root property but not the clustering and fat-tail properties. Suffice it to say that GARCH models like those used by Leblang and Bernhard (2000a) and Lobo and Tufte (1998) address the clustering property but do not adequately address the fat tail property of exchange rates. For this reason the distributional

assumptions on which the authors' hypothesis tests are based also are problematic. The residuals from a Markov

switching model are not plagued by the kurtosis in the same way as those from a linear regression model.




b. A Synthesis

The Markov switching model can be interpreted in the following way. The log difference in the spot exchange rate follows a random walk because of the way traders form expectations about the content and consequences of government policy. The random walk specification implies that the change in a currency's value from one period to the next is a random draw from a normal distribution. The parameters of the random walk-its mean (pi) and variance (c2)-reflect both the prevailing politico-economic informational environment and the optimal decision rule of the traders; the random walk represents an equilibrium in this sense. But the informational environment sometimes changes and this alters the optimal trading rule. Together, these changes produce a regime switch: a change in the parameters of the random walk (a new equilibrium). A random walk with a relatively high variance connotes a more frenetic informational environment and a trading rule that is more sensitive to news. The questions are (1) what kinds of changes to the informational environment cause currency traders to revise their decision rules so as to produce random walks with greater variance (volatility)? and (2) what institutions, if any, make switches to higher variance (more volatile) random walks more and less likely?

The metaphor of a rumor mill is frequently used in the popular press to describe the type of informational changes that we have in mind. For example, one commentator from The Economist recently wrote, in a story on Philippine president Arroyo, that the mill is working at "double speed" grinding out rumors about a possible coup and the degree of support for former president Estrada among the masses, military officials, and police corps." During times of potentially significant political change, like the Philippines experienced in early 2002, political speculation abounds and the probability that traders make decisions on the basis of speculative rumors increases.12 We contend that it is the propensity of currency traders to verify these political rumors that distinguishes currency regimes. The tendency to switch to a decision rule based on unverified political rumors during periods of political uncertainty is the mechanism whereby politics in young and incipient democracies spawns exchange rate volatility. 13

New work by Calvo and Mendoza (2000) and others is suggestive here. Calvo and Mendoza argue that the periods of high volatility in emerging financial markets are caused by rational contagion, a situation in which traders choose not to pay for costly information to verify rumors about future returns on investments.1 When traders base their expectations on unverified rumors rather than fundamentals, their behavior is more erratic. Consequently, rumor driven investment decisions produce large capital flows into and out of emerging market economies and this generates significant volatility in their currency markets. If it is more costly for

1 See "Happy Anniversary, Mrs. President," The Economist, January 19, 2002, p. 36-37. '2 The Economist article goes on to note that the government fears the "talk of political instability is scaring off

investors" making it difficult to grow the economy. Of course, coups are not the only type of politics that generates rumors. During periods of conflict over the future course of economic policy, uncertainty about the government's ability to manage its opposition and generate popular support and (or) uncertainty about the government's survival

prospects more generally, the political rumor mill is likely to be in operation. 13 The "news" approach to modeling exchange rates is popular in financial economics (for a discussion see

Hallwood and MacDonald, 1994:247-252). News, which drives exchange rate dynamics in these models, is

unanticipated changes in fundamentals. In our model, rumors lead to changes in the exchange rate. Rumors are similar to news in that they are unanticipated. News, however, is always factual, while rumors may not be.

14 Calvo and Mendoza (2000) argue that globalization combined with short selling constraints and performance based incentives that punish poor portfolio performance more than they reward good performance makes

contagion more likely. That is, globalization, short selling constraints, and performance based incentives increase the likelihood that it will be an optimal strategy for traders to make changes in their portfolio allocations on the basis of unverified rumors. Short selling constraints place limits on the ability of traders to profit from knowing the true value of future returns and performance based incentives make contagion equilibria stable by increasing the cost of

deviating from the herd's behavior. These incentives make it less likely that the effects of rumors will be undone by price corrections in the market.




traders to verify political than economic rumors, then rumor driven investment decisions will likely prevail during periods of political uncertainty when these rumors are in abundant supply.

Are political rumors more difficult to verify? The small literature in economics that examines the politics of exchange rates treats political and economic information in a way that suggests the former is unanticipated or slowly processed by markets. Currencies and other economic variables are treated as random walks suggesting that currency traders process economic information quickly and efficiently. In contrast, when they do appear in models, political variables are represented as dummies or level variables (Bachman, 1992; Blomberg and Hess, 1997; Christodoulakis and Kalyvitis, 1997), which suggests that political information is not anticipated and processed much more slowly than economic information. Moreover political information arrives in discrete quantities causing one-shot changes in behavior of currency traders. While economists rarely offer a rationale for this distinction, we believe it is a reasonable assumption.

There is a large body of research in political science that suggests it is easier for traders to forecast market prices than political outcomes. This literature argues that politics is inherently ill-conditioned and complex and that there are no predictable political equilibria that are stable over the long term (McKelvey, 1976; Schofield, 1993; Richards, 1994).15 This complexity implies that confirming (or denying) the validity of political rumors will be problematic and costly, even in countries with established democratic institutions where political information is readily available. These problems will be magnified in new democracies and particularly acute in countries with nondemocratic political institutions and opaque political processes.16 Therefore, we posit that it is more difficult for traders to manage and reduce political uncertainty than economic uncertainty.

Thus, one can think of the regimes in the Markov switching model as being driven by separate rumor mills, which specify the politico-economic informational environment in which traders buy and sell currencies. A schematic of our version of the model is presented in Figure 2. There are two regimes: a "fundamentals" and a "contagion" regime. The exchange rate follows a random walk in each regime, but the drift and volatility of the random walk or its mean (y) and variance (2), respectively, differ across regimes. The variance is determined by the optimal trading rule, specifically whether or not traders verify rumors, which is a function of the informational environment.

The first rumor mill is in operation during periods of political certainty (stability). It generates rumors that are largely economic in nature and, given the type of resources available to traders and their economic expertise, relatively cheap to verify. As a result, traders confirm the veracity of these rumors and only respond to those that are true. This informational environment generates the "fundamentals" regime in which the observed log difference in the exchange rate has a low variance. During periods of potential political instability, the second rumor mill goes into operation. It generates a much higher quantity of political rumors. Because these rumors are more costly to verify, traders do not check them.

15 Of course, the degree to which a country's politics is unpredictable may be a function of its institutions. 16 If, for example, there were a rumor that several pro-reform ministers in a government were going to be

sacked, this might generate pessimistic expectations about the future health of the economy. Regardless of the

transparency of a country's polity, because of the complexities of the cabinet dissolution and formation processes, it would be difficult to verify the long-term political and policy consequences of this rumor were it true. But in countries with more opaque political processes that have limited credible political information available, it would be either very costly or impossible for traders to determine whether there was or was not any intragovernmental conflict that threatened the stability of the government in the first place. In a more transparent democracy, a rumor like this one could be readily disconfirmed by the (free) press, for instance.




Switching Mechanism: Political (Un) certainty

"Fundamentals" r i "Fundamentals" Regime Regime

P(st =

Flst_ = F,

xt_;F) Political

x( )Random Walk Random Walk exp ( ) Certaintyndom Walk

Rumor Mill: iiiii Rumor Mill: Economic Rumors pc =(1- pcc) Economic Rumors

P(st = Fst-,

= C, x,_,; c)

l+exp(xt ,c)

(exp (x1- ))

1

+exp(xt_,) 'Contagion" 1+expixPF

..i '"Contagion"

Random Walk Pt c Political 4 I Random Walk

(fctr) Aist =CJst

tCcx

) [ Uncertainty [iI (,c ...) ..........ex

p (x.( . .

) . . . . . .......

Rumor Mill: RumorMill:

Political Rumors ::Ni::i +iiiii: [i Political Rumors

t-1 t

Note: x,_1 = (1,xj_

,..x,,_,,,_ ) and fi = (fi.0 , .... ~

,_o)-)',i = F, C. When the last (k-1) terms of the parameter vectors

Fr and Pc are set to zero, the time-varying transition probability model collapses to the constant transition probability model. The transition probability notation above is based on Diebold, Lee, and Weinbach (1994, 285).

FIG. 2. Markov Regime Switching Model of Exchange Rates for Emerging Market Economies

Consequently, their buying and selling is more erratic and herd-like. Hence we observe greater volatility in the exchange rate (o2c > U2F).

The model's dynamics are represented in Figure 2 with solid and broken line arrows. The solid line arrows illustrate the evolution of currency markets during times of political certainty. A market will either remain in or switch to the "fundamentals" regime, and the transition probabilities will reflect this dynamic. More specifically, p

t will be close to one and (or) pCt will be close to zero, which

implies a long expected duration in the "fundamentals" regime and (or) a short expected duration in the "contagion" regime. The broken line arrows show market dynamics during periods of political uncertainty. A market will either switch to or remain in a "contagion" regime in these periods. The transition probabilities are modeled using a logit specification. Proxies for political instability and uncertainty (denoted in Figure 2 as xt- 1)-like a large drop in presidential approval, a party defection from the governing coalition, or student riots in the capital city-should help identify which rumor mill is in operation and therefore "explain" the over time variation in the model's transition probabilities. In other words, some of the fl's in the logit specification should be statistically significant and correctly signed. (For a more technical and complete discussion of the Markov regime switching model, see Hamilton, 1994:ch. 22; and Diebold et al., 1994.)

When recast in these terms, the political science literature on new and emerging democracies yields the following propositions:

Because, in the face of political uncertainty, it is not optimal for traders to verify political rumors, the probability of currency markets remaining in or switching to high volatility "contagion" regimes increases

PROPOSITION 1: during electoral periods--especially for elections in which there is a high degree of uncertainty about the outcome




PROPOSITION 2: the more opinion polls and other sources of political information show declining support for incumbent executives

PROPOSITION 3: the less stable incumbent coalitions are

PROPOSITION 4: the higher the degree of overt popular opposition to the government and its policies (e.g., frequency of political riots)

To the extent that they produce less political uncertainty, certain political institutions will be associated with a higher probability of currency markets remaining in or switching to a fundamentals regime. In particular,

PROPOSITION 5: Among young democracies, the more consensual the form of government, the less impact political uncertainty will have on the probability of currency market switching to a high volatility regime. [Propositions 1-4 will be less likely to hold the more consensual the young democracy.]

In incipient democracies, where institutions are just taking root, extragovernmental factors will be the primary source of political uncertainty; the markets in these countries will switch from fundamentals regimes to contagion regimes during periods of social unrest. Intragovernmental factors like cabinet negotiations will have comparatively less impact on regime switches:

PROPOSITION 6: Only proposition 4 will hold for emergent democracies. It is to a test of these propositions that we now turn.

Research Design We focus on the US dollar exchange rates for Indonesia, the Philippines, South Korea, and Thailand between March 1998 and September 2000. Thailand, South Korea, and the Philippines all had democratic institutions in place throughout our sample period; Indonesia, on the other hand, did not. As noted above, however, there are important differences between the young democracies. First, they are not generally viewed as equally democratic. Of the three, Thailand is typically considered to have the most democratic processes and institutions, followed by South Korea and the Philippines.17 Additionally, their democracies differ in kind. The Philippines has a strong presidential political system; South Korea has a semipresidential system with both a popularly elected president and a prime minister approved by the parliament; and Thailand has a pure parliamentary system with electoral institutions that typically produce large multiparty coalitions. Again, we expect that electoral and, in the cases of South Korea and Thailand, intra-governmental politics will be important in determining the probability of currency regime switching in the young democracies (propositions 1-4). The somewhat consensual nature of Thai political institutions may mitigate this relationship to some extent (proposition 5). And, because politics is more opaque, variables like social unrest will have an impact on the probability of currency regime switches in Indonesia (proposition 6).

Daily exchange rates over the period 3/16/98 - 9/29/00 were used to estimate the models.18 Over the entire period in which the four countries' currencies have floated, there are clearly three currency market regimes: a regime for the initial crisis period of extreme volatility, a moderately volatile regime, and a low volatility regime. Because we are estimating a two-regime model, it would be inappropriate

17 The Polity IV scores for Indonesia, the Philippines, South Korea, and Thailand in 1998 are 0, 7, 8, and 9, respectively, with the low (high) numbers representing the absence (presence) of democratic institutions and

procedures (Polity IV Project, 2000). 18 The exchange rate series are from the PACIFIC Exchange Rate Service's Historic FX Rates Database.




TABLE 1. Extreme Currency Volatility in the Floating Period

Country 1-Day Currency Average 1-Day Currency Ratio Return (US$ Exchange Rate) Return (Absolute Value)

Indonesia 1/08/98 30.188 1.876 16.091 1/09/98 - 21.122 1.876 - 11.258 1/27/98 - 19.330 1.876 - 10.303 2/10/98 - 23.316 1.876 - 12.428 2/13/98 21.677 1.876 11.554

Philippines 12/15/97 7.176 0.577 12.447 1/06/98 6.170 0.577 10.702 1/22/01 - 12.518 0.577 - 21.713

South Korea 11/20/97 10.105 0.689 14.675 11/21/97 - 8.080 0.689 - 11.735 12/08/97 8.971 0.689 13.028 12/09/97 8.714 0.689 12.654 12/11/97 8.954 0.689 13.003 12/15/97 - 8.687 0.689 - 12.616 12/16/97 - 9.340 0.689 - 13.565 12/22/97 7.955 0.689 11.553 12/23/97 13.444 0.689 19.524 12/24/97 - 6.888 0.689 - 10.004 12/29/97 - 25.705 0.689 - 37.330 12/30/97 14.996 0.689 21.778 1/30/98 - 9.700 0.689 - 14.088

Thailand 3/13/98 - 6.574 0.645 - 10.187

NOTE: Dates of extreme volatility are identified by one-day currency returns that are more than ten times the size of the average one-day currency return during the floating periods for each country.

to use the full sample. We therefore focused on the post-crisis period and the last two regimes.19 The starting point was determined by comparing the one-day currency returns with the average daily currency return over the floating period. Days when the log difference in the exchange rate was more than ten times the size of the average absolute value of the log difference over this period were excluded from the sample. These dates, which are listed in Table 1, are all concentrated in the period before March 16, 1998.20

Several types of political variables were employed in the analysis. First, machine- coded events data from the IDEA project (Bond et al., 1997) were used to construct measures of social conflict. In particular, these data were used to create three variables for the degree of social violence and coerciveness: maximum violence, maximum coercion, and Bond et al.'s conception of conflict carrying capacity

19 The other option was to estimate a three-regime Markov switching model with time varying transition

probabilities. Because of the inherent difficulties in estimating these models and the amount of programming it would have required to do so, we chose to shorten the sample.

20 Our sample period thus distinguishes our research from that of Haggard and others, which has tended to focus only on the crisis period. There is, of course, contagion and high volatility in the post-crisis period, but this

volatility is not of the same magnitude as that of the crisis period.




(CCC), which measures a regime's ability to regulate contentious interaction without violence. Each event that is coded receives a score of 0, 1, 2, 3, or 4, with 0 representing low violence/coercion and 4 representing high violence/coercion. In the case of violence, a score of 0 is given to events that involve no significant physical force either threatened or used on another. At the other end of the scale, a score of 4 is given to events that involve explicit threats or the use of major physical force. Noncoercive events, given a score of zero, include those that involve no social, economic, or political costs imposed on another. Highly coercive events, which receive a score of 4, involve the use of major and explicit costs, either imposed or threatened on another. Bond et al. (1997) define conflict carrying capacity as one minus the proportion of events that are violent. Thus, the measure ranges from 0 to 1.21 Altogether, these variables are indicative of the kind of political uncertainty that exists about the abilities of governments to manage opposition and implement their policies.

In addition to these variables we also included election period dummies as well as some country-specific variables discussed below. As proposition 1 suggests, ceteris paribus, there will be political uncertainty in the run-ups to elections. Hence traders are more likely to receive political rumors in these periods.22

Because of the importance of the presidency in Philippine politics, we included net presidential approval in our analysis of that case-that is, the percentage of people who approve of the president's performance in office less the percentage that disapprove. There is substantial movement in this variable over our sample period. In the early part of 1998, the net approval for Ramos stays at or below 20% until the election in May won by Estrada. In that month, net presidential approval jumps to 60% and stays at that approximate level for more than a year. In July of 1999, Estrada's net approval begins a steady decline, reaching a low in December of 1999 of 5%. Drops in presidential approval should cause political uncertainty and hence a shift to currency regimes in which political rumors are not verified.-3

In South Korea, there are two important executive officers: the president and the prime minister. The former is popularly elected while the latter is approved by the parliament. In this semipresidential system, the potential for intragovernmental conflict is always present. We posit that the stronger the president's party in the parliament, the less likely there will be conflict between the president and prime minister or between the president and the parliament more generally. Under these conditions, presidents will be better able to appoint prime ministers who share their political views (and dismiss those who do not). There will be little doubt about the political influence of the president vis-a-vis parliament. Therefore, we included the percentage of parliamentary seats held by the president Kim Dae Jung's party, which during our sample was the National Congress for New Politics (NCNP), until January 20, 2000, and the New Millennium Democratic Party (MDP) thereafter. This seat percentage gradually increases over the course of our sample. When Kim Dae Jung was inaugurated as president on February 25, 1998, the NCNP held 26.4% of the seats in the National Assembly. By March 30, 1999, as a result of by- elections and party defections, the NCNP increased its seat percentage to 35.1%. In the April 2000 elections, Kim Dae Jung's party (now the MDP) increased its seat share to 42.1%. This should have created less political uncertainty and therefore

21 The CCC measure that we use is discussed in detail in Bond et al. (1997:559-562). This measure is refined in

Jenkins and Bond (2001). 22 The dummies are eighty-day periods, forty days up to and including the day of the election and forty days

after. 23 The data is from the Social Weather Station's special media release (12/22/2000) "Erap Hangs on to + 9 net-

Satisfaction, As Classes and Regions Polarize," which is available online at http://www.sws.org.ph/pr122200.htm. In our analysis, missing monthly values were filled in using linear interpolation.




lowered the probability that traders would use decision rules that connote the contagion regime.24

Because coalition politics is a central feature of the Thai system we modeled the transition probabilities for the baht as a function of the percentage of seats held by the governing coalition of parties. The Chavalit government whose coalition members held 65% of the seats in parliament collapsed in November of 1997. It was replaced with a bare majority six-party coalition (53%) under the leadership of Chuan Leekpai of the Democratic Party. A little less than a year later, a seventh party joined the coalition thereby increasing its seat share in the parliament to 66%. The Social Action Party withdrew from the coalition in July of 1999 dropping the coalition's percentage of seats in the parliament to 61%. This may have created enough uncertainty to increase the probability of traders switching to the contagion regime. But, once more, Thailand's somewhat consensual institutions may be a mitigating factor.

Results

The switching models were constructed in a couple of steps. First, fixed transition probability models were estimated using maximum likelihood. Then several pretests were performed to establish the model structure for each country. The pretests included Garcia's test for one versus two currency regimes, Wald tests for the equality of means and variances across regimes as well as simple versus Markov switching, and finally the D'Agostino z-test for excess kurtosis. 25 Next, the time varying transition probability models were estimated using our political variables. Because of the nonlinear nature of the Markov regime-switching model, it is often difficult to converge on a solution with more than one variable in the transition probability vector. Therefore, we started by estimating our time-varying transition probability models with one political variable at a time. If we found evidence that more than one variable was statistically significant, we attempted to reestimate the model with all the statistically significant variables included in the transition probabilities simultaneously.26

If the propositions hold, our results should indicate when political uncertainty causes traders to alter their decision rules and choose not to verify political rumors, shifting the market to a high volatility currency regime. Table 2 contains the results for Indonesia. The constant transition probability estimates and pretests are reported in the first column. The pretests indicate that there are two currency regimes. The Garcia test statistic is statistically significant. Neither drift parameter is statistically significant so the two regimes are distinguished by their degree of volatility. In fact, the second regime is more than twelve times as volatile as the first. The test for simple switching (Ho: pCC= 1 - pFF) is rejected. Currency rates in

24 Some observers interpreted the 2000 elections as a setback for Kim Dae Jung because the opposition Grand National Party (GNP) gained seats at the expense of the UDL, the NCNP/MDP's coalition partner. For an alternative

interpretation see Haggard (2000:107). 25 Simple switching models are those in which the probability of being in a particular state is the same regardless

of the previous state. Garcia (1998) provides a method for testing the null hypothesis of a single regime by deriving the nonstandard asymptotic null distribution of the LR statistic to obtain critical values. The z-test for excess kurtosis is described in D'Agostino et al. (1990). Following Kim and Nelson (1999), these tests were performed on the standardized residuals from the model.

26 There is some research that suggests good and bad news (i.e., unanticipated changes in economic variables) have asymmetrical effects on currency market volatility. For example, Leblang and Bernhard (2000a) argue that

unanticipated negative shocks may produce larger increases in volatility than unanticipated positive shocks. If we find that political uncertainty shifts currency markets into higher volatility regimes, then our results are entirely consistent with the spirit of this argument: political uncertainty increases currency market volatility by much more than political certainty. Additionally, as long as there is persistence in the two regimes (0o > 0) the nonlinear functional form of our transition probabilities guarantees that political certainty and uncertainty will have the

anticipated asymmetrical effects on currency market volatility.



TABLE2. Markov Switching Models for the Indonesian Rupiah (US$ Exchange Rates, 3/16/98-9/29/00)

Parameter Constant Transition Maximum Daily Maximum Daily CCC Election** Probabilities Coercion Score Violence Score Period Dummy

f/t 0.044 (.066) 0.053 (.092) 0.046 (.112) 0.046 (.064) 0.046 (.096) Pc - 0.183 (.340) - 0.199 (1.027) - 0.188 (.535) - 0.188 (.413) - 0.184 (.365)

a' 1.292 (.232) 1.279 (.257) 1.301 (.243) 1.303 (.225) 1.274 (.249)

a(: 16.563 (3.550) 16.187 (3.359) 16.528 (3.593) 16.596 (3.446) 16.396 (3.623)

PFO 2.500 (.315) 2.858 (2.913) 2.905 (1.161) - 1.330 (5.739) 2.462 (.318)

FIl - 0.106 (.742) - 0.125 (.300) 4.199 (6.350) 0.100 (1.10)

flco 1.470 (.310) - 1.530 (1.282) 1.372 (1.122) 0.472 (1.441) 1.549 (.353)

flc1 0.881 (.418) 0.050 (.425) 1.114 (2.360) - 0.900 (1.10)

log(L): - 1370.690 - 1367.408 - 1370.514 - 1370.295 - 1370.058

Wald Tests

Ho: pC(:( = 1 - p"F 164.39 [.000] Ho: 4v = -N 0.421 [.516] Ho: aF = a: 20.069 [.000]

Kurtosis

Currency Returns 10.701 [.000] Stan. Residuals 3.241 [.196]

Likelihood Ratio Tests Garcia 302.055* Constant vs. Time-Varying 6.564 [.038] 0.352 [.839] 0.790 [.674] 1.264 [.532]

NOTES: Standard errors in parentheses. Brackets contain p-values. *p-value <.01 **Parliamentary Elections, 6/7/99.

1\3 r

~sl

Vq m

O c4

~3C. O X3

N. X3

C/p

N.

C4~ O




Indonesia are governed by a Markov switching process where the probability of being in each regime depends on the previous state of the market. Finally, the model removes all of the excess kurtosis from the unconditional distribution of currency returns. A normal distribution has a kurtosis of three. The raw distribution of rupiah-dollar currency returns for our sample had a kurtosis of 10.701. After estimating the model, however, we are unable to reject the hypothesis that its standardized residuals have a kurtosis of three.

Columns 2-6 of Table 2 report the estimated impact of our political variables on the transition probabilities for the rupiah-dollar currency market. The coefficients PFO and 3co are the constants in the logistic expression for pFF and pCC, respectively, while ~F1 and Pct are coefficients for the political variable (xt - 1) listed at the top of each column (cf. Figure 2). Indonesian politics clearly affects the behavior of currency traders. Individual coefficient and likelihood ratio tests for the maximum daily coercion score are statistically significant. Moreover, the sign of the coefficient is consistent with our expectations. The exchange rate and coercion series are plotted in Figure 3a. It shows that on days when there are highly coercive events, the market is more likely to persist in the high volatility regime. For example, when the maximum daily coercion score is 2, the probability of staying in high volatility currency regime is .56, giving an expected duration in this regime of a little over two days.27 When the maximum daily coercion score is increased to 4, the probability of staying in the high volatility regime increases to .88 and the expected duration rises to over eight days (see Figure 3b). No such results are found for the maximum daily violence or for Bond et al.'s CCC measure. Nor did the Indonesian election on June 7, 1999 have an impact on currency trading. Rather, in this case, it is social coercion that causes the change in traders' decision rules.

The results for the Philippines are contained in Table 3. Again, there are two regimes distinguished by their volatility, the estimated variance in the second regime being almost sixteen times as great as the variance in the first regime. The model removes most but not all of the excess kurtosis in the currency returns. The kurtosis goes from 13.093 in the unconditional distribution to 4.296 in the standardized residuals. Individual coefficient and likelihood ratio tests show that domestic politics has an impact on regime switching. The election period dummy is statistically significant. During election periods, in the Philippines, the market is more likely to switch out of the low volatility regime. The coefficient on Bond et al.'s conflict carrying capacity variable is positive and also statistically significant in the low volatility regime. (The exchange rate series and conflict carrying capacity are plotted in Figure 4a.) As conflict carrying capacity rises, the market is more likely to persist in this regime. For example, when CCC is .85, the expected duration in the low volatility depreciation regime is a little over 3 weeks. When CCC rises to 1, the expected duration increases to a little less than three months (see Figure 4b). Interestingly, government coercion, maximum daily violence, and presidential approval do not affect the dollar-peso market. Since both the election dummy and the CCC variable are statistically significant, we finish by estimating a model with both of these variables in the transition probabilities. When we do this, both variables remain statistically significant and correctly signed in the "fundamentals" regime. This final model clearly outperforms the others in terms of identifying the currency regime in our sample period.

Table 4 provides the results for South Korea. Again, the Garcia test indicates that there are two regimes. Since neither drift parameter is statistically significant, the regimes are distinguished by their volatility with the variance in the second regime more than 17 times that in the first. Again, the model removes most but not all of the excess kurtosis initially displayed in the won-dollar currency returns. The

27 The expected duration in a regime is 1/(1 - p"), where i = F,C.



TABLE 3. Markov Switching Models for the Philippine Peso (US$ Exchange Rates, 3/16/98-9/29/00)

Parameter Constant Maximum Maximum CCC Election Net CCC / Election Transition Coercion Violence Period Presidential Period Dummy**

Probabilities Score Score Dummy** Approval

1F 0.034 (.013) 0.035 (.013) 0.035 (.013) 0.033 (.013) 0.035 (.013) 0.034 (.013) 0.033 (.014)

Pc 0.008 (.048) 0.007 (.543) 0.007 (.323) 0.010 (0.052) 0.006 (.028) 0.008 (.051) 0.009 (.046)

F2 0.062 (.008) 0.063 (.008) 0.063 (.010) 0.062 (.007) 0.063 (.008) 0.062 (.008) 0.063 (.008) ac 0.978 (.174) 0.995 (.178) 0.989 (.179) 0.978 (.164) 1.045 (.206) 0.982 (.182) 1.022 (.196)

3F0 3.577 (.368) 3.497 (2.254) 4.367 (.947) - 4.193 (2.961) 3.562 (.358) 4.005 (.549) - 4.295 (3.138)

3PF 0.025 (.800) - 0.315 (.385) 8.650 (3.500) - 5.100 (1.600) - 0.011 (.015) 8.759 (3.681) PF2 - 5.153 (1.888)

13co 3.052 (.544) 1.648 (1.264) 4.334 (6.227) - 3.608 (5.970) 2.483 (.511) 2.356 (.751) - 3.170 (9.972) BPc1 0.525 (.525) - 0.525 (1.96) 7.112 (6.500) - 0.600 (.900) 0.014 (.022) 6.099 (10.905 Pca - 0.142 (2.265)

log(L): - 418.651 - 418.082 -417.748 - 416.144 - 412.178 - 417.815 - 409.731

Wald Tests

Ho: pC" = 1 - pF" 1008.6 [.000] Ho: VF = 1k: 0.262 [.609] Ho: o2F = oY2 28.483 [.000]

Kurtosis

Currency Returns 13.093 [.000] Stan. Residuals 4.296 [.000]

Likelihood Ratio Tests Garcia 447.560* Constant vs. Time-Varying 1.138 [.566] 1.806 [.405] 5.014 [.082] 12.946 [.002] 1.672 [.433] 17.840 [.001]

NOTES: Standard errors in parentheses. Brackets contain p-values. *p-value <.01 **Presidential Election, 5/11/98.

00

It

ca

eQ.




18000 5

4.5

16000 -4

3.5 14000

12000 I 2.5

-2 10000 1.5 8000

0.5 9

0.45

8

0.4

0.35 Transition Execteda

SProbability D

0.3 - 0 6 >

0.25

0.15--

0.1

3

0.05

0 I I I I I tI I I I I I I I 2

2.5 2.6 2.7 2.8 2.9 3 3.1 32 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4

(b) Maximum Daily Coercion Scores

FIG. 3.(a) Indonesia, Exchange Rate, and the Maximum Daily Coercion Score, 3/16/98-9/27/00 (b) Maximum Daily Coercion Scores and Regime Switching in the Rupiah/Dollar Exchange Rate

kurtosis drops from 19.556 to 5.051. The estimated coefficients on the maximum daily violence score and the NCNP/MDP parliamentary seat percentages are both statistically significant and the estimates have the expected sign. The higher the maximum daily violence score the more likely the market is to switch out of the low volatility currency regime. The higher the parliamentary seat percentage of the president's party the longer the market will persist in the low volatility currency regime: the expected duration in the low volatility regime increases from 2.7 days to more than 70 days as the seat percentage of the NCNP/MDP increases from 25 to 40 (see Figures 5a and 5b). Neither social coerciveness nor Bond et al.'s CCC variable has any impact on regime switching in the Korean case. Since both the




46 1.2

44t 1.15

42- 1.1

40 -1.05

38-- 1

36 0.95

34-- 0.9

32 I I I I I I I I I I 0.85 CD 0 0 0000 0 o D0O0O0 a

0 00 0 0 0 0 00 0

(a)

0.05 90

80

: 0.04

Transition t Probability

70 -70

.C 0.03--

60> S

0.02

40

,Di

030

0 I I I I I I 20 0.85 0.86 0.87 0.88 0.89 0.9 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1

(b) Conct C ying Capaclty

FIG. 4.(a)Philippines, Exchange Rate, and Conflict Carrying Capacity, 3/16/98-9/27/00 (b) Conflict Carrying Capacity and Regime Switching in the Peso/Dollar Exchange Rate

maximum daily violence and the parliamentary seat percentage variables are statistically significant, we estimate a model with both in the transition probabilities. This time only the seat percentage variable remains statistically significant. Once we control for this variable, maximum daily violence does not improve our ability to identify the currency regimes, which suggests social violence does not generate the kind of political uncertainty in South Korea that causes traders to alter their decision rules.

The results for Thailand are presented in Table 5. As in all of the previous cases, there are two regimes. The Garcia test statistic is significant at the .01 level. Neither



TABLE 4. Markov Switching Models for the South Korean Won (US$ Exchange Rates, 3/16/98-9/29/00)

Parameter Constant Maximum Maximum CCC** Parliamentary Seat Max Violence / Transition Probabilities Coercion Score Violence Score Percentage (NCNP/MDP) Seat Percentage

4, - 0.018 (.016) - 0.017 (.017) - 0.019 (.016) - 0.018 (.018) - 0.018 (.015) - 0.019 (.017)

l'c - 0.083 (.080) - 0.086 (.108) - 0.087 (.089) - 0.078 (.082) - 0.099 (.101) - 0.098 (.105)

C2 0.070 (.020) 0.072 (.025) 0.076 (.029) 0.068 (.020) 0.081 (.031) 0.081 (.033)

cr 1.207 (.426) 1.241 (.625) 1.313 (.571) 1.147 (.639) 1.465 (.647) 1.476 (.673)

3FO 2.849 (.382) 3.098 (8.898) 6.823 (5.359) 16.840 (57.366) - 5.594 (2.607) - 4.209 (3.708)

P3HF - 0.100 (2.600) - 1.292 (1.500) - 14.570 (58.80) 0.246 (.092) - 0.336 (.389)

F2 0.231 (.105)

Pco 2.218 (.759) 3.545 (6.001) 1.941 (1.494) 58.496 (286.17) 4.783 (2.672) 4.764 (2.706)

3ci --

0.400 (1.900) - 0.007 (.700) - 58.555 (294.00) - 0.117 (.096) 0.052 (.466)

PC2 -0.121 (.102)

log(L): - 471.128 - 470.999 - 468.457 - 470.469 - 458.510 - 457.595

Wald Tests

Ho: pCC = 1 - pFF 111.19 [.000]

Ho: 1F = PC 0.580 [.446] Ho: cr = C72 7.648 [.006] Kurtosis

Currency Returns 19.556 [.000] Stan. Residuals 5.051 [.000] Likelihood Ratio Tests Garcia 432.899* Constant vs. Time-Varying 0.258 [.879] 5.342 [.069] 1.318 [.517] 25.236 [.000] 27.066 [.000]

NOTES: Standard errors in parentheses. Brackets contain p-values. *p-value <.01 **Three-period moving average.

to

cj?

t~o

H

1"




1550 43

1500 41

1450 39

1400 37

1350 35

1300 33

1250 31

1200 --29

1150 -27

1100I9 I

'9

' ;25 00 90 000CO 0CO CO CD 0CO (a

(a)

0.4 80

0.35 70

= 0.3 60

Transition > Probability

c . Expected

0.25 / Duration 50

E 0.2 40

C a S 0.15 30

0.1 20 0

0

0.05 -10

0 I I I 0 I 0 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

(b) PrnIamentry Sean Percentage

FIG. 5.(a) South Korea, Exchange Rate, and Parliamentary Seat Percentage of NCNP/MDP, 3/16/98-9/ 27/00 (b). Seat Percentage of President's Party and Regime Switching in the Won/Dollar Exchange

Rate

drift parameter is statistically significant. The currency regimes, therefore, are distinguished by their volatility with the variance in regime two being eleven times as great as the volatility in regime one. The regime-switching model removes enough of the excess kurtosis from the initial distribution of currency returns that we are unable to reject the null hypothesis of residual kurtosis equal to three at the .05 level of statistical significance. In the Thai case, there is no evidence that extragovernmental variables cause shifts in traders' decision rules. But there is evidence that intragovernmental variables cause these shifts. The coefficient on the coalition seat percentage variable is statistically significant in the low volatility



TABLE5. Markov Switching Models for the Thai Baht (US$ Exchange Rates, 3/16/98-9/29/00)

Parameter Constant Transition Probabilities Maximum Coercion Score Maximum Violence Score CCC** Coalition Seat Percentage

t F 0.019 (.018) 0.019 (.019) 0.019 (.022) 0.016 (.034) 0.020 (.019)

?lC - 0.022 (.146) - 0.024 (.134) - 0.021 (.338) - 0.022 (.997) - 0.055 (.151)

C2F 0.136 (.118) 0.141 (.080) 0.135 (.180) 0.144 (.026) 0.162 (.017) 2 1.491 (1.414) 1.562 (1.047) 1.486 (2.153) 1.787 (.573) 2.143 (.543)

3FO 3.259 (1.864) 3.393 (1.496) 3.300 (2.676) 39.930 (45.830) - 21.040 (8.773)

3FI - 37.311 (46.656) 0.418 (.161)

k3co 2.031 (.614) 1.278 (1.263) 1.701 (2.457) 2.179 (.740) - 5.362 (7.015)

c3 0.225 (.425) 0.150 (.058) 0.116 (.124)

log(L): - 524.918 - 524.883 - 524.810 - 524.495 - 513.853

Wald Tests

Ho: p": = 1 - p" 60.835 [.000] Ho: lF = ?c 0.074 [.785] Ho: c = C2 1.088 [.297] Kurtosis

Currency Returns 12.728 [.000] Stan. Residuals 3.399 [.053] Likelihood Ratio Tests Garcia 30.034* Constant vs. Time-Varying 0.070 [.791] 0.216 [.642] 0.846 [.358] 22.130 [.000]

NOTES: Standard errors in parentheses. Brackets contain p-values. *p-value <.01 **Three- period moving average.

cj

H1

EN




68

49

66

47 64

45 62

43

58

41

56

39

37 52

0 0 0-- 0 0 00000 080000

(a)

0.6 500

o -450

C 0.5

400

Zo 0 350lo . 0.4

Transition Probability 300

E 0.3 2Duration 250 c

- 2 0 0

Coo0.1

X-.--loo

50

0 I I I I I 0 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

(b) Pwllmentary ueat Percentage

FIG. 6.(a) Thailand, Exchange Rate, and Seat Percentage of the Governing Coalition, 3/16/98-9/27/00 (b). Seat Percentage of Governing Coalition and Regime Switching in the Baht/Dollar Exchange Rate

currency regime and has the anticipated sign. Because of the fluidity of coalitional politics in Thailand, this variable may serve as a good proxy for the probability of cabinet dissolution. The withdrawal of the Social Action Party from the government coalition in July of 1999 is an example of this fluidity. Because the government had a supermajority in this period, the defection was largely inconsequential. However, had the SAP withdrawn during the first Chuan government, the impact would have been much more significant: the coalition would have lost its majority status in the parliament.

Our results suggest that as the governing coalition's seat percentage drops, the market is more likely to switch out of the low volatility regime. For example, when




the percentage of parliamentary seats held by the governing coalition is 65, the average of the two periods of supermajority coalitions, the market is essentially locked into the low volatility currency regime with an expected duration of close to 460 days. When the percentage of seats drops to 53, which is the percentage of parliamentary seats held by the first Chuan government from November 1997 to November 1998, the expected duration in the low volatility currency regime falls to just 4 days (see Figures 6a and 6b). It should be noted that the likelihood ratio statistic for this model is highly significant (p-value <.001).

Discussion

Overall, our results provide support for propositions 1, 3, 4, and 6. With regard to proposition 1, we found that during the period surrounding the Philippine presidential election, the peso-dollar currency market was more likely to switch out of the low volatility currency regime. Given that this election was "strongly contested but highly uncertain" (Haggard, 2000:130), this result is consistent with the idea that electoral politics causes currency traders to adopt decision rules whereby they do not verify political rumors (proposition 1). We also found that during the periods when the Thai coalition government's future would have been most uncertain-that is, periods when the government had a minimal share of the legislative seats and therefore was most vulnerable to defection-the baht-dollar market was less likely to persist in the low volatility currency regime; this evidence supports proposition 3 but refutes proposition 5. Proposition 5 was derived from previous research that focused on the advanced industrial democracies, specifically Germany (Freeman et al., 2000). The different impacts that coalition politics has on the Thai and German currency markets may reflect differences between young and long-standing democracies. Or it may be attributable to differences in their respective party systems, like the presence of strong parties and government coalitions that are dimension-by-dimension medians with empty win sets in the latter country but not in the former (Laver and Shepsle, 1996; MacIntyre, 2001:118).

In the cases of Indonesia and the Philippines we found support for proposition 4: during periods of high social violence and/or coercion, traders use decision rules whereby they do not verify political rumors, hence currency markets are more likely to remain in or switch to high volatility regimes. The fact that our social conflict variables did not affect regime switching in the baht-dollar and won-dollar markets most likely reflects the degree of democratic development in Thailand and South Korea. Conversely, in the incipient democracy Indonesia, we found that maximum daily coercion and not the more familiar political variable, in this case an election dummy, had a statistically significant impact on currency market switching. This result is consistent with proposition 6. Finally, our results provide mixed support for proposition 2. Net presidential approval did not have a statistically significant impact on regime switching in the peso-dollar market. But changes in the legislative seat percentage of Kim Dae Jung's party-which rose over the period of our sample partly as a result of his popularity--did have a statistically significant impact on currency regime switching in the won-dollar market.

Conclusion

Our research both reinforces and develops the existing arguments about democratization and globalization. We agree with, and our results support, Haggard's (2000) conclusions about the linkages between transparency, political certainty, and financial crises. Because politics generates substantial uncertainty in countries with less democratic institutions and more opaque policy processes, these countries suffer more from volatility in international financial markets. In fact, our




estimate of the variance of Indonesia's high currency volatility regime was more than ten times larger than the estimated variances for the other countries. In this sense, then, globalization and democratization in emerging market countries are compatible processes.

At the same time, however, democracy does not eliminate political uncertainty completely. And because it is difficult to verify the political rumors that circulate during times of uncertainty, even in young democracies with consolidated institutions and processes, there is a possibility that the globalization of financial markets will impose some costs. Our empirical results suggest this is true. Some form of routine democratic politics affected currency market volatility in all of the young democracies that we studied.

In this respect, our analysis is consistent with MacIntyre's (2001) argument that most developing democracies are more susceptible to financial crises because they have either widely dispersed or extremely concentrated veto authority as opposed to the developed democracies which have distributions of veto authority falling somewhere between the two extremes. Similarly, our empirical results lead us to conclude that most developing democracies are susceptible to currency volatility and that there are significant differences between the mature and young democracies in this regard (cf. Freeman et al., 2000). Most important, because the nature of coalition politics seems to be more fluid in developing countries, PR electoral systems do not necessarily stabilize democratic politics in a way that reduces the uncertainty elections and cabinet dissolutions generate; consequently PR systems do not lessen the potential for democratic politics to create currency market volatility in developing democracies.

MacIntyre's framework is illuminating. But it is also too simple. Our research shows the value from adding political context to his veto points logic. MacIntyre may be right that the presidential system in the Philippines allows a mix of policy flexibility and restraint that, most of the time, is reassuring to international investors. However, we found that presidential elections create significant currency volatility, which suggests that because of the importance of the presidency in this system these elections can make traders very nervous. Our analysis of the Thai case highlights the important role that the size of governing coalitions plays in determining the amount of political uncertainty generated by coalition politics. The veto power of small parities in governing coalitions (a very important source of political uncertainty) is maximized when governments are bare majorities rather than supermajorities. Finally, the South Korean case demonstrates the importance of political context in semipresidential systems. Simply put, intragovernmental conflict, a source of political uncertainty and hence currency market switching, is more likely in these systems when the president's party is underrepresented in the legislature because parliament is more likely to challenge the president under these conditions.

To conclude, we note that, by focusing on the post-crisis period, we have expanded the scope of research on domestic politics in emerging market countries and the behavior of international financial markets. Currency traders monitor and react to politics continuously. Therefore, research that goes beyond the politics of financial meltdowns, speculative currency attacks, and other "rare events" is clearly needed (Haggard, 2000; Leblang, 2001; MacIntyre, 2001). We addressed this need by exploring the politics of currency market volatility. While periods of volatility have serious welfare consequences, they are, at the same time, much more frequent and "normal" than the events that consume much of the existing scholarly research.

References

ALESINA, A., N. ROUBINI, AND G. COHEN (1997) Political Cycles and the Macroeconomy. Cambridge, MA: MIT Press.




BACHMAN, D. (1992) The Effect of Political Risk on the Forward Exchange Rate Bias: The Case of Elections. Journal of International Money and Finance 11:208-219.

BERNHARD, W., AND D. LEBLANG (2001) Polis and Pounds: Exchange Rate Volatility and Domestic Political Competition in Britain. Unpublished manuscript. Champain Uibara, IL: University of Illinois.

BERNHARD, W., AND D. LEBIANG (2002) Democratic Processes, Political Risk, and Foreign Exchange Markets. American Journal of Political Science 46:316-333.

BLOMBERG, S.B., AND G. D. HESS (1997) Politics and Exchange Rate Forecasts. Journal of International Economics 43:189-205.

BOND, D., J. C. JENKINS, C. L. TAYLOR, AND K. SCHOCK (1997) Mapping Mass Political Conflict and Civil Society. Journal of Conflict Resolution 41:553-579.

CALVO, G.A., AND E. G. MENDOZA (2000) Rational Contagion and the Globalization of Securities Markets. Journal of International Economics 51:79-113.

CHRISTODOULAKIS, N.M., AND S. C. KALYVITIS (1997) Efficiency Testing Revisited: A Foreign Exchange Market with Bayesian Learning. Journal of International Money and Finance 16:367-385.

D'AGOSTINO, R. B., A. BELANGER, AND R. B. D'AGOSTINO, JR. (1990) A suggestion for Using Power and Informative Tests for Normality. The American Statistician 44:316-321.

DEVRIES, C.G. (1994) "Stylized Facts of Nominal Exchange Rate Returns." In The Handbook of International Macroeconomics, edited by F. Van der Ploeg, pp. 348-389. London: Blackwell.

DIEBOLD, EX., J. LEE, AND G. C. WEINBACH (1994) "Regime Switching Models with Time-Varying Transition Probabilities." In Nonstationary Time Series Analysis and Cointegration, edited by C.

Hargreaves, pp. 283-302. Oxford: Oxford University Press. ENGEL, C., AND J. HAMILTON (1990) Long Swings in the Dollar: Are They in the Data and Do the

Markets Know It? American Economic Review 80:689-713. FRANKEL, J.A., AND D. ROEMER (1999) Does Trade Cause Growth? American Economic Review 89:

379-399. FREEMAN, J.R., J. C. HAYS, AND H. STIX (2000) Democracy and Markets: The Case of Exchange Rates.

American Journal of Political Science 44:449-468. FRIEDEN, J. (2002) Real Sources of European Currency Policy. International Organization 56:831-860. GARCIA, R. (1998) Asymptotic Null Distribution of the Likelihood Ratio Test in Markov Switching

Models. International Economic Review 39:763-788. HAGGARD, S. (2000) The Political Economy of the Asian Financial Crisis. Washington DC: Institute for

International Economics. HALLWOOD, C. P., AND R. MACDONALD (1994) International Money and Finance. 2nd Edition. London:

Blackwell Publishers. HAMILTON, J.D. (1994) Time Series Analysis. Princeton, NJ: Princeton University Press. HAMILTON, J.D. (1989) A New Approach to the Economic Analysis of Nonstationary Time Series and

the Business Cycle. Econometrica 57:357-384. HERRON, M.C. (2000a) The Source of Gridlock: Divided Government or Preference Conflict? Paper

presented at the Annual Meeting of the American Political Science Association, Washington, D.C. HERRON, M.C. (2000b) Estimating the Economic Impact of Political Party Competition in the 1992

British Election. American Journal of Political Science 44:326-337. HERRON, M.C., J. LAVIN, D. CRAM, AND J. SILVER (1999) Measurement of Political Effects in the Stock

Market. Economics and Politics 11:51-82. HOWELL, L.D. (2001) The Handbook of Country and Political Risk, 3rd ed. Syracuse, NY: The PRS Group. INTERNATIONAL MONETARY FUND (1997) International Capital Markets: Developments, Prospects, and Key

Policy Issues. Washington, DC: IME

JENKINS, J.C., AND D. BOND (2001) Conflict Carrying Capacity, Political Crisis and Reconstruction: A Framework for the Early Warning of Political System Vulnerability. Journal of Conflict Resolution 45:3-31.

KIM, C.J., AND C. R. NELSON (1999) State-Space Model with Regime Switching: Classical and Gibbs-Sampling Approaches with Applications. Cambridge, MA: MIT Press.

LAVER, M., AND K. A. SHEPSLE (1996) Making and Breaking Governments. New York: Cambridge University Press.

LEBLANG, D.A. (2001) To Devalue or to Defend? The Political Economy of Exchange Rate Policy. Manuscript. Boulder, CO: University of Colorado.

LEBLANG, D.A., AND W. BERNHARD (2000a) Parliamentary Politics and Exchange Markets: The World According to GARCH. Paper presented at the Annual Meeting of the Midwest Political Science Association, Chicago.

LEBLANG, D., AND W. BERNHARD (2000b) The politics of Speculative Attacks in Industrial Democracies. International Organization 54:291-324.




LIJPHART, A. (1999) Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries. New Haven, CT: Yale University Press.

LIN, T., AND B. ROBERTS (2001) Markets and Politics: The 2000 Taiwanese Election. Paper presented at the Annual Meeting of the Midwest Political Science Association, Chicago.

LOBO, B., AND D. TUFTE (1998) Exchange Rate Volatility: Does Politics Matter? Journal of Macroeconomics 20:351-365.

MACINTYRE, A. (2001) Institutions and Investors: The Politics of the Economic Crisis in Southeast Asia. International Organization 55:81-122.

MCKELVEY, R.D. (1976) Intranswitivities in Multidimensional Voting Models and some Implications for Agenda Control. Journal of Economic Theory 12:472-482.

MEESE, R. (1990) Currency Fluctuations in the Post Bretton Woods Era.Journal of Economic Perspectives 4:117-134.

MISHKIN, F.S. (1999) Global Financial Instability: Framework, Events, Issues. Journal of Economic

Perspectives 13:3-20. PANTZALIS, C., D. A. STRANGELAND, AND H. J. TUTTLE (2000) Political Elections and the Resolution of

Uncertainty: The International Evidence. Journal of Banking and Finance 24:1575-1604. POLITY IV PROJECT (2000) Polity IV Dataset. [Computer file; version p4v2000] College Park, MD:

Center for International Development and Conflict Management, University of Maryland. RICHARDS, D. (1994) Intransitivities in Multidimensional Spatial Voting: Period Three Implies Chaos..

Social Choice and Welfare 11:109-119. ROBERTS, B.E. (1990) Political Institutions, Policy Expectations, and the 1980 Election: A Financial

Market Perspective. American Journal of Political Science 34:289-310. RODRIK, D. (2000) How Far Will International Economic Integration Go? Journal of Economic

Perspectives 14:177-186. ROSE, A.K. (2000) One Money, One Market: Estimating the Effect of Common Currencies on Trade.

Economic Policy 30:7-46. SCHOFIELD, N. (1993) "Political Economy: A Personal Interpretation and Overview." In Political

Economy: Institutions, Competition and Representation, edited by W. A. Bennett, M. J. Hinich, and N. Schofieldpp. 1-22. New York: Cambridge University Press.

USCINCPAC (2000) Asia-Pacific Economic Update. Honolulu. WEINBACH, G.C. (1993) Regime Switching with Time Varying Transition Probabilities: Methodological

Issues and Applications to Exchange Rates. Ph.D. diss., University of Pennsylvania.



exchange rate volatility and democratization in emerging market countries

Documents