DEMOCRATIC THREATS

REGIME TYPE AND SOVEREIGN DEBT CRISIS RESOLUTION

Henrik Enderlein, Laura Müller, and Christoph Trebesch1 Hertie School of Governance

Paper to be presented at the IPES-Conference in Philadelphia, November 14-15, 2008

Second Draft Please do not cite or quote without permission.

Comments welcome!

Abstract This paper analyzes government behavior in sovereign debt crises between 1980 and 2004. We test the relationship between regime type and crisis management on a new data set of government behavior during crises. Looking at distress episodes in 31 emerging countries, we find that democratic debtors differ significantly from non-democratic sovereigns. Democracies generally take a more aggressive stance towards lenders in economic crisis than autocracies. Aggressive government behavior increases with the level of democracy. Finally, more experienced democracies show more coercive behavior than recently democratized debtors. The results hold across tests with different democracy indicators and using different estimation techniques.

1 The authors gratefully acknowledge financial support of the German Research Foundation (DFG) under the Special Research Area 700 on Governance in Areas of Limited Statehood (www.sfb-governance.de). We thank Alexander Agronovsky, Henrik Maihack and Said-Khalid Scharaf for excellent research assistance in preparing this draft. We are also grateful to Sebastian Saiegh for sharing data.

1

1. Introduction

The paper is concerned with a crucial type of international economic relations – sovereign

borrowing.2 Since the 1980s the role of private-to-public lending has increasingly been gaining

importance. As emerging countries continue to rely heavily on access to private capital markets

one would expect their governments to be interested in good relations with financial market

participants. Why then, do we still see aggressive policies towards international creditors during

financial crises as in Argentina in 2001? In the paper we try to answer this and a set of related

questions. The main goal is to test if differences in debtor behavior can explained by domestic

institutions. Going back to a large literature on macroeconomic policy making outlined below,

we find very compelling arguments on how regime type could affect economic outcomes and

economic policy making in emerging countries undergoing political transition. However, to this

date the empirical evidence concerning the effect of democracy remains largely inconclusive. The

literature on the institutional origins of economic performance and policies is far too voluminous

to summarize adequately in this paper. Important, albeit partly contradicting, recent contributions

are provided by Barro (1996), Hall and Jones (1999), Przeworski et al. (2000), Acemoglu et al.

(2001), Glaeser et al. (2004), and Epstein et al. (2006). Likewise, it remains an open question if

democratic institutions increase debtor compliance with international debt contracts or not.

To gain a more systematic understanding of the actual policy choices we study sovereign

debt crises between 1980 and 2004 in 31 emerging economies. Specifically, we aim to explain

policy choices with regard to external private creditors. We thereby contribute to what Panizza et

al. (2008: 41) in a new review of the sovereign debt literature call the study of “private incentives

of policymakers to default or fight a crisis.” As we argue here, those private incentives - to stay in

2 Sovereign debt is commonly defined as the total outstanding financial liabilities of a country, including all public and publicly guaranteed debt incurred by other debtors, denoted in foreign and local currency. We focus on the part of sovereign debt owed to external private actors. This is the standard definition of external Sovereign Debt used by the rating agency Standard and Poors (S&P) which is employed in most of the literature.

2

power - are likely to depend on the regime type of the debtor, i.e. whether it is democratic or

autocratic.

We try to contribute to a more nuanced discussion on sovereign debt crises introducing a

newly coded index of government behavior that captures the degree of unilateral, non-

cooperative policies imposed on creditors in sovereign default situations and in the run-up to debt

restructuring agreements. By and large, previous empirical tests of the role of various economic

and institutional factors in default relied on a simple binary definition of default versus non-

default. Although useful for a number of research questions the binary definition fails to display

the variation in policy choices falling under this definition. It gives no account of the variation in

government behavior towards private creditors.

Sovereigns over the last decades have, however, approached their creditors in very

different ways during crises, reaching from low key and conformable behavior to outright

repudiation of debt and arrogance on the side of government officials. A comparison between an

autocracy like Romania and a democracy like Peru in the early 1980s illustrates this point. The

totalitarian Romanian government from 1980 onwards, imposed extremely high costs of

continued debt service on the population to convince creditors of its continued willingness to

repay their debt. Expenditures were cut to a minimum. At the peak of the crisis during the winter

of 1985 Romanian President Nicolae Ceaucescu put a ban on use of private vehicles and ordered

drastic limitations on citizens’ use of light and heating. 3

In stark contrast, Peru – after initial attempts to implement austerity had failed – declared

a unilateral debt ceiling on its external debt servicing in June 1985 (Financial Times on 29 June,

3 At the same time when power supply was cut short, debt payments to external creditors continued (WSJ, 20 Dec. 1985, FT, 15 July 1986, WSJ, 24 July 1986). Despite rising poverty rates and chronic foot shortages Ceausescu continued to follow his drastic policy of debt reduction through 1987 (FT, 17 Nov. 1987, FT, 24 April 1987 FT, 20 Nov. 1987).

3

1985). After years of democratic transition with political turmoil, failed IMF programs and

frequent general strikes, newly elected President Alan Garcia opted for a very different set of

debt policies than the Romanian government at the time..

Another striking example of the importance of domestic political pressures on democratic

leaders is the situation in Argentina in December of 2001 and in the following years. After a three

year-period of economic recession and political deadlock, the government of Argentina declared

a complete halt on all external debt payments in one of the most prominent cases of unilateral

default in recent history. It is without question that mass protests of the population were a major

factor in the decision to declare a default.

The empirical analysis yields several important insights on what explains this variability

in debt crisis handling. First of all, looking at the distribution of our “Index of Government

Behavior” we find that the degree of aggressiveness of debtors during sovereign debt crises is

relatively constant over time. We found no evidence that governments behaved significantly

more or less aggressively towards their private creditors in the era of syndicated bank lending of

the 1980s as compared to the Post-Brady-Deal-Era characterized by highly dispersed bond

holders.

As concerns the determinants of government behavior, there are three main results: First,

evidence shows that democratic leaders applied significantly more coercive debt policies towards

creditors on average.4 Second, the ability to take an aggressive stance towards creditors increases

significantly with the level of democracy as reported by the Polity index. Third, looking only at

the sub-sample of democratic countries in our data we find that years of democratic experience

seem to explain differences in debtor behavior among democratic governments. Countries with

4 In various test using the standard S&P-default indicator as dependent variable we also find a significant positive effect of democracy on the likelihood of a default.

4

more than five years of democratic experience turn out to be significantly more aggressive

towards external creditors. We ascribe this outcome to the fact that voters in democratic systems

are generally in a better position to force their governments into aggressive external debt policies.

The remainder of the paper is structured as follows: Section 2 presents the related

literature on sovereign default and government behavior in external crisis while Section 3 departs

from the literature to develop the theoretical foundations of our argument and provide testable

hypotheses. Section 4 presents our new indicator, the “Index of Government Behavior” which is

followed by a brief presentation of stylized facts and descriptives on democracies in default

(Section 5). In Section 6 we describe regression techniques, data sources and variables. Results

are discussed in Section 7, followed by concluding remarks (Section 8).

2. What are Sovereign Defaults? And what explains external debt policy decisions?

Concerned with explanations of government debt policies during financial distress this

section starts out with a brief discussion of the appropriate dependent variable to use (section

2.1). We then go on to discuss three different strands of literature relevant to our research

question: the macroeconomic literature on debt crises (section 2.2), institutional explanations of

default (2.3), and contributions concerned with the effects of democratic institutions on

macroeconomic policy making in newly democratized countries (2.4).

2.1 Measuring debt crises and default

To this date, the literature treated government choice mainly as a binary variable. Most

researchers employ a dummy default indicator based on data from Standard and Poor’s5 or from

5 The S&P definition of default takes into account any payments missed on scheduled bond debt, notes or bills and on bank loan interest or principal (Standard & Poor’s 2006). Also, any exchange of new debt that contains less

5

the World Bank’s lists of restructuring events (World Bank 2002, 2003).6 Some researchers have

also combined these two key sources with additional data and definitions in an attempt to provide

more concise measures of debt crisis.7 Nevertheless, standard indicators of default applied in

most of the literature remain dichotomous variables. We find the dichotomy to be too

constraining and propose a more nuanced picture of the “choice of weapons” for governments in

crisis.

In fact, there is a large body of more qualitative accounts of sovereign default events since

1980 concerned with government behavior during crisis.8 They include attempts to categorize

different types of debt crises and government behavior during crises. Authors agree in that crises

vary on a spectrum from soft to hard or from voluntary to more involuntary cases (examples are

Aggarwal 1996, Cline 2004, Frankel and Roubini 2001, Roubini 2004 and Roubini and Setser

2004). Cline (2004) and Roubini (2004) are particularly concerned with the categorization of past

crises as a function of the degree of private sector burden sharing. In their approaches, cases vary

on a continuum between extremely friendly and very coercive cases.9 Inspired by these

favourable terms than the original bond issue and any rescheduling of principal and/or interest at less favourable terms than indicated in the original contract counts as a default. 6 See e.g., Borensztein and Panizza (2006),Gelos, Sahay and Sandleris (2004), Manasse, Roubini and Schimmelpfennig (2003), , Reinhart, Rogoff and Savastano (2003), van Rijckeghem and Weder (2004), and Kohlscheen (2007) 7 As an example, Detraigache and Spilimbergo (2001) consider arrears of principal or interest towards commercial creditors, while 2003 define a country in default whenever the S&P criterion holds or if it receives a large non-concessional IMF loan meaning in excess of 100 % of quota. Pescatori and Sy (2007) suggested a further debt crisis measure, which explicitly takes into account the increase in bond financing since the early 1990s. In addition to default cases as of S&P they regard a country in severe financial distress, whenever the sovereign bond spread surpasses a critical threshold, such as 1000 basis points above U.S. Treasuries. Lastly, authors such as Reinhart, Rogoff, Savastano (2003), Gelos, Sahay and Sandleris (2004) and Fostel and Kaminsky (2007) have supplemented the S&P list of defaults with information of the qualitative list of debt crises by Beim and Calomiris (2001: 32-36). 8 Historical accounts of sovereign debt crises going back into the past centuries of sovereign lending include Stasavage (2007), Suter (1992), Tomz (2007a), Reinhart and Rogoff (2008b). See Cline (1995), Aggarwal (1996) and Boughton (2001) for a discussion of 1980s debt crisis cases. Roubini and Setser (2004), Andritzky (2006) and Sturzenegger and Zettelmeyer (2006) are mainly devoted to more recent cases of financial distress. 9 Also, Beim and Calomiris (2001) and Manasse, Roubini and Schimmelpfennig (2003) both hint at the variation of debtor behavior reaching from “outright repudiation” or “outright default” to default in a more “conciliatory fashion”.

6

approaches we create a categorical “Index of Government Behavior” which will be discussed in

detail in section 3.

2.2 Literature on macroeconomic determinants of sovereign default

Much of the literature in Economics has focused on the macroeconomic and financial

determinants of debt crises. Panizza, Sturzenegger and Zettelmeyer (2008) provide an

encompassing “history of economic thought” on sovereign debt covering the past 30 years of

research including theoretical and empirical contributions.

Recent empirical contributions identified a set of economic factors to be important

macroeconomic determinants of default (see, for example, Kraay and Nehru 2004, Manasse,

Roubini and Schimmelpfennig 2003 or Detraigache and Spilimbergo 2001). They include the

pre-crisis level of indebtedness, measures of liquidity, the level of output, trade openness and past

default history (see Reinhart and Rogoff 2008).10 Additionally, the presence of programs by the

IMF or other official rescue lending institutions can be important (a.o. Marchesi 2003 and

Roubini and Setser 2004).

2.3 Institutional explanations for default

There is still relatively little empirical work on the role of political and institutional

factors. In recent quantitative studies Manasse, Roubini and Schimmelpfennig (2003) and Kraay

and Nehru (2004) include a number of political and institutional factors in their analysis.11 Their

results indicate political economy variables to play a role in debt distress but the articles lack an

explanation why and how these variables matter in detail.

10 Reinhart, Rogoff and Savastano (2003) give an explanation how countries’ default history affects market perceptions of the likelihood of default. 11 Other examples are Reinhart, Rogoff and Savastano 2003 or Detraigache and Spilimbergo 2001. For an overview see also Hatchondo, Martinez and Sapriza (2007).

7

More in-depth studies on the role of institutions highlight the importance of regime type in

explaining differences in debt repayment behavior, however with inconclusive empirical results.

Schultz and Weingast (2003) set the stage for a deeper exploration of the effect of democratic

institutions on sovereign expropriation and sovereign default.12 In a widely quoted article they

find more liberal political systems to enjoy better borrowing conditions providing them with a

“democratic advantage”. Because liberal political institutions allow for credible commitments to

contracts, democracies are predicted to have a lower default risk, reflected in lower risk premia

on government bonds.13

Jensen (2003), van Rijckeghem and Weder (2004), Saiegh (2005), Kohlscheen (2007),

Stasavage (2007), Alichi (2008), and McGillivray and Smith (2008) also look at the effects of

regime type (democracies versus autocracies) in more detail.14 Their predictions, however, are

inconclusive. Saiegh (2005) and Alichi (2008) find democracies to have a higher likelihood of a

debt rescheduling and default. In contrast, Schultz and Weingast (2003), van Rijckeghem and

Weder (2004), McGillivray and Smith (2008), and Jensen (2003, 2008) conclude that

democracies have lower probabilities of default and lower country risk premia.15

12 Earlier contributions had already hinted at the possible relationship between democracy and default. However, authors predicted the relationship to have different fore signs than Schultz and Weingast (2003). Kaufmann (1985), sees democratic rule as a major driving force for explaining debt policies. In a similar vein, Frieden (1988) looks at the interaction between government policy choice and private sector attitudes towards governments to explain debt and other macroeconomic management in the 1980s Latin American Debt Crisis. He focusses on the labor - capital relation in a country and finds the level of democracy has an important intervening function as it affects the capacity to suppress demands from either interest group.

13 The hypothesis is based on an earlier argument by North and Weingast (1989) on the relationship between liberal rule and the protection from expropriation in 17th Century England. The article predicted property rights to be more respected in more liberal polities. The assumption of a positive causal effect of “checks” on the respect of property rights and on borrowing terms in liberal political systems has been criticised by authors including Clark (1996), Epstein (2000), Sussman and Yafeh (2006). 14 Kohlscheen (2007) gives further insights on democratic regimes. In his sample parliamentary democracies turn out to default less often than presidential systems. The main reason seems to be the high responsiveness of parliamentary systems to their constituency. . 15 The studies by Saiegh (2005) and Van Rijckeghem and Weder (2004) use different dependent variables and data sources which makes comparison and interpretation a bit difficult. Saiegh (2005) uses 1999 data on private and public external reschedulings and restructurings in 80 between 1971 to 1997. Van Rijckeghem and Weder (2004) test the effect of democracy on domestic and external default, finding democracies to default less often on external debt then

8

Jensen (2003, 2008) argues that democratic leaders are more likely to respect property

rights. Not only do they expropriate less, their increased property rights stability and credible

commitment to international debt contracts is rewarded by lower default risk premia in capital

markets. McGillivray and Smith (2008) conclude that systems ruled by larger winning

coalitions – which are highly correlated with democracies – are less likely to default on debt. The

main reason is that democratic leaders can be replaced more easily if they follow unpopular

policies.16 According to McGillivray and Smith (2008), the anticipation of a punishment effect is

what drives governments to cooperate with their creditors.17

On the more skeptic side of the “democratic advantage” debate are Saeigh (2005) and

Stasavage (2007). Saiegh (2005) states an important critique of the unconditionally positive

effect of democratic institutions. Looking at data for developing countries, he finds democracies

to have a higher propensity to reschedule than non-democracies and brings up two necessary

conditions for the democratic advantage-argument to hold. First, lenders would have to play an

important role in the domestic economy and be able to exert electoral power on the government.

Second, populations are not necessarily in favor of debt repayment at all times. Contrary to the

prediction of Schultz and Weingast (2003) creditors may not be represented in sufficient numbers

in the debtor country to deter contract breach. 18

Finally, Stasavage (2007) revisits the democratic advantage argument using historical data

on 27 European states between 1274 and 1765. In the article he provides evidence on a

significant impact of liberal institutions on states’ market access. However, what matters is

autocracies. They look at 73 low and middle income countries between 1970 and 2000 using annual data from S&P for 2003 covering default on domestic and external debt denoted in foreign and local currency. 16 Authors base their analysis on work by Bueno de Mesquita et al. (2003) who studied conditions of political survival of leaders in small versus large winning coalition systems. 17 In their analysis of changes in debtors’ credit worthiness in 70 countries between 1993 and 2003, McGillivray and Smith (2008) use bond indices as dependent variables They also test their model on data for default rates as measured by Standard&Poor’s between 1824 and 2003 and find similar results.

18 Saiegh derives this idea from a borrowing model by Drazen (1998). He claims that governments will not comply if the median voter’s saving is less than the economy’s average (Saiegh 2005: 370).

9

apparently the interaction of institutions and the type of state – city state or territorial state.

While democratic institutions in city states have historically had great impact on borrowing

conditions, institutions were less decisive in territorial states.

2.4 Literature on macroeconomic policy decisions in new democracies

The “democratic advantage”-approach was noteworthy because it provided an explanation

why investors could prefer democracies to authoritarian regimes. In contrast, much of the earlier

literature on macroeconomic policy making in developing and emerging countries during the

1980s emphasized the relative advantage of authoritarian governments to grow economically and

attract capital. The main reason was seen in the ease with which autocratic leaders could suppress

popular demands for distributive policies for the sake of macroeconomic adjustment and

compliance with international business contracts (O’Donnell 1978, Kaufman 1986, Haggard and

Kaufman 1992 and 1995, Williamson and Haggard 1994).

Kaufmann (1985: 473) describes the situation at the onset of the 1980s Debt Crisis:

“[G]overnmental elites, if they remain in power, must also answer to (or repress)

their own populations. And the price to be paid for external help with ‘liquidity problems’

has typically involved politically dangerous stabilization measures (devaluations, wage

and credit restrictions, and fiscal deficit reductions) - measures that often arouse the

strong opposition of major social forces.”19

His main conclusion is that, faced with such political conflicts, autocratic regimes had

“apparent advantages” over completely elected governments. They could repress opposition from

19 In a similar vein, Frieden (1989:24) notes that foreign debt is often “a source of domestic political conflicts for it can raise important distributional issues”.

10

unions and popularly based party organizations (Kaufman 1986:195). Furthermore executive

authorities could operate “outside a framework of institutionalized accountability”. From a

powerful position technocrats could thus promote policies “aimed at winning the general

‘confidence’ of local and international capital”. By and large, authors in the early 1990s agreed

that distributional pressures were reform-deterring in democracies.

Predictions were particularly gloomy for new democracies as a separate category of

regimes distinguished from autocratic regimes and from established democracies (Kaufman 1986,

Frieden 1989, Williamson and Haggard 1994, Haggard and Webb 1994, Haggard and Kaufman

1995). Nevertheless, the empirical ground for these negative predictions was relatively weak as

was shown in the methodological critique of the “conventional wisdom” by Stallings (2004) and

Remmer (1990) among others.20

Few authors found possible advantages of the newly created democratic systems although

Hirschman (1987) hints at one obvious advantage of new democracies. He ascribes new

democratic governments a “special reserve of trust and goodwill” allowing them to take

unpopular reform steps. Other authors have described this phenomenon as the “honeymoon

effect” (Haggard and Kaufman 1995 and Linz and Stepan 1996 a.o.). The high level of popularity

based in a high legitimacy provides new governments with the - temporary - freedom to

implement even the most comprehensive and painful adjustment policies (for an overview on the

debate see Bernard et al. 2001). Likewise, all responsibility for economic hardship and negative

consequences of necessary reforms can be blamed on the hated old regime (Stokes 2001).

Tomz (2002) provides a new approach which allows combining the opposing views on

the effect of democratic institutions on macroeconomic policymaking to some extent. He draws

20 In their view, the selection of only successful reform initiatives in autocratic systems and/or unsuccessful examples of reform under democratic regimes lead early commentators of transition in Latin America, Southern Europe and parts of Africa to undue generalizations. Unlike many of the earlier studies, authors found that chances for successful reform implementation in new democracies were not completely off.

11

on audience cost theory to explain how changes in domestic preferences translate into external

policies. Like Kaufman (1985) and Frieden (1989) earlier, Tomz (2002) stresses the distributional

consequences of international agreements. Those parts of the population most negatively affected

by an international agreement can turn into “champions of noncompliance” (Tomz 2002: 2). In

good times, when the economic situation is stable, citizens in democracies are likely to take a

pro-market position and punish their government if it takes steps that are detrimental to the

reputation of the country (McGillivray and Smith 2008). However, as the economic situation

deteriorates, populations change their preferences. In the context of a debt crisis, this would lead

them to prefer default to hurtful domestic adjustments.21

Tomz’ argument that domestic audience costs are behind default decisions seems intuitive

but requires further testing. In the context of international debt negotiations audience cost theory

gives the whole story of coercive debt policy decisions by governments an important twist. One

could argue that the capacity to raise credible audience costs provides an incentive for democratic

governments to take a tough position towards external creditors.

Another central finding from audience cost theory is that democratic leaders are thought

to be better at choosing the “right” conflict when they defect as suggested by Lake (1992).22

According to this logic democratic leaders go through a more thorough internal reasoning before

risking external conflict. If wining a conflict seems unlikely democratic governments are backing

down. However, have they decided to fight they are thought to expend significantly higher

resources on winning (Fearon 1994: 585).

21 Tomz (2002) was first to show in an impressively detailed account of the Argentine crisis around 2001 that changes in preferences are likely if the economic circumstances deteriorate. A precondition for this change is that voters have access to and can process information on the economic and financial situation.

22 Bueno de Mesquita, Morrow, Siverson and Smith (1999), Partell and Palmer (1999), and Eyerman and Hart (1996), testing the argument using different proxies for audience costs find democracies to be better able to prevail in disputes. For a comprehensive overview see Gelpi and Griesdorf (2001). Critical accounts of the argument are presented by Desch (2002) and Slantchev (2004).

12

What was described as a major disadvantage of democracies turns into an advantage on

the international stage. The domestic political pressure lends strong credibility to leaders’ default

threats or other aggressive actions. Given the condition of economic crisis, democratic leaders

can credibly claim that they will soon have no other option to avoid further hardship on the

population than by defaulting on external debt. This strategy is reflected in a quote by President

Garcia at the time of the Peruvian Moratorium declaration in 1985, which made it clear that the

choice for him was “either debt or democracy” (Drake 1989: 53). A threat by an autocratic leader

in the same position would seem less credible.

A democratic government that decides to engage in a conflict with its external creditors

can be seen as highly resolved to take all means to reach debt relief of some sort. Threats to

repudiate are backed by a majority of the population, lending high credibility to a leader’s

aggressive external policy.

The audience cost argument has been used in a number of related analyses of economic

policymaking and international economic cooperation.23 How audience costs play out in the

context of new democracies seems to be a blind spot, however, at least in the field of economic

policymaking.

3. Hypotheses

In this section we briefly present our three main hypotheses on the relationship of regime type,

experience with democratic institutions and sovereign default behavior of governments. The key

underlying question is whether regime type is what affects external debt polices? And if there is a 23 Mansfield, Milner and Rosendorff (2002) find that more democratic regimes have a higher likelihood to cooperate via international trade agreements. Leblang (2005) argues that democracies face higher audience costs than autocracies when abandoning a de jure exchange rate regime. Incumbent governments are found to be heavily penalized in terms of re-election chances. Also Lektzian and Souva (2003) finds regime type to affect foreign economic policy goals, in particular the choice of sanctions governments impose on other states.

13

difference, does it matter how experienced citizens and policymakers of a country are with

democratic institutions? Our main arguments are based on domestic audience cost theory and on

theories describing the role of political liberalization for macroeconomic policymaking presented

in section 2.

In line with the preceding paragraphs, we argue that democratic leaders generally face a different

trade-off during crisis than autocrats (O’Donnell 1978, Kaufman 1985, Frieden 1988, 1989,

Haggard and Kaufman 1992, 1995, Tomz 2002 a.o.). In democracies, audience costs, in this case

the political costs of implementing hurtful economic adjustment against resistance from ‘loser

groups’, are likely to be higher and increasing over the crisis period. Therefore, the government

constantly has to chose between keeping its external reputation and holding on to power

internally, driving it at times to jeopardize the former for the sake of the latter given that the

economy is in deep crisis (Tomz 2002).

A first testable implication of this view would be that a state’s behavior towards creditors

is driven by the type of regime - democratic or autocratic. We predict the following hypothesis to

be supported by the data:

Hypothesis 1: Democratic governments behave more aggressively than autocracies in debt crisis.

The first hypothesis states a positive relationship between democracy and coercive debtor

behavior on average. To gain a better understanding of the relationship of democratic institutions

and domestic audience costs of policy adjustment it seems necessary to look also at different

levels of democracy. From existing theories presented above we conclude that audience costs

should be high in countries with advanced democratic institutions (Fearon 1994, McGillivray and

14

Smith 2008). As an extension to this argument, we expect countries that rank high on the

democracy-scale to take on more aggressive debt policies given that the country is experiencing

an economic crisis.

The main reason could be that leaders in more developed democracies face larger

difficulties in implementing unpopular economic policies (Tomz 2002). Citizens in countries

with very advanced democratic institutions are likely to be more experienced with the policy

making process. In more established democracies we also assume a higher likelihood that

opposition groups are more institutionalized and active. Such groups have longer experience with

their constitutionally guaranteed veto powers, preventing the implementation of government

policy proposals. Concretely, they may be better able to exert veto power through parliamentary

decision making procedures and/or through presence in the media and on the streets engaging in

strikes and protests. We would therefore expect the following hypothesis to hold:

Hypothesis 2: The degree of government aggressiveness towards external creditors increases

with the degree of institutionalization of democratic decision making processes over

macroeconomic policy. Accordingly, higher levels of democracy correspond to more coercive

government behavior in debt crisis resolution.

Clearly, not all democracies are the same in terms of experience with the political system.

In a sample of emerging economies the issue of political stability seems to be central to crisis

handling. Looking at the literature the evidence on reform effectiveness for newly established

democracies is mixed. Many of the early approaches were skeptic of leaders’ capacities to

implement reforms. This view is echoed in a recent study on militarized interstate conflict by

Jessica Weeks (2008) which concludes that newly established and unstable democracies show

15

less resolve to stick to a contentious policy. One reason could be that weaker institutional

capacities and less decision making routine would render threats to act aggressively less credible

in the eyes of the opponent.

If this argument holds, new democracies would also seem less likely to take a tough

stance towards external creditors. They can hardly claim that institutionalized democratic

opposition to adjustment is what forces them into contract breach. Thus, creditors can easily

argue that a new democratic government can a) put all the blame of economic hardship on the

previous ousted regime (Stokes 2001) and b) exploit the “honeymoon effect” after its

inauguration to implement a quick and comprehensive economic reform package even if it

includes unpopular austerity measures (Hirschman 1987, Haggard and Kaufman 1995).

A second argument is that new democracies have strong incentives to attract foreign

investors to enhance growth and development. In the early years of political transition, when it is

not clear that democracy is irreversible, governments are seen as highly dependent on growth and

increase in revenues to provide public goods (Linz 1978). Public goods provision is an essential

part of democratic government that makes it distinct from autocratic rule (Bueno de Mesquita et

al. 2003; for a more nuanced view on autocratic provision of public goods see Clark 2005). Both

arguments, the enhanced capacity to implement comprehensive adjustment and the high reliance

on good relations with foreign investors make cooperative behavior much more likely. This leads

us to the third prediction:

Hypothesis 3: Leaders of newly established and unstable democracies face less resistance against

painful adjustment policies. They are therefore expected to act more cooperatively towards their

external creditors than established democracies.

16

Our three hypotheses will be tested on data for 31 emerging countries between 1980 and

2007 using a new Index of Government Behavior as the main dependent variable. We also test

the effect of democracy on the simple standard default variable by S&P which is widely used in

the existing literature. Results are presented in section 7 below. The index allows us to create a

more nuanced picture of the institutional effects on actual government policy choices during

crises.

4. Measuring Government Behavior: The Index of Government Coerciveness

In this section we briefly present the “Index of Government Coerciveness” which serves

as the dependent variable in the empirical analysis. The index is coded on a yearly basis and debt

policy changes can be tracked over the course of multi-year crisis episodes. 24

Construction of the Index

For the construction of the index we relied partly on previous attempts to categorize government

behavior by Cline (2004) and Roubini and Setser (2004). Additionally, we could draw on two

key policy documents listing “best practices” for debt crisis resolution that found wide

acceptance among policy makers and private sector representatives. These are the IMF’s criteria

of good faith efforts in debt negotiations with creditors (IMF 1999; 2002) and the criteria outlined

in the so called Principles of fair debt restructuring by the Institute of International Finance (IIF

2006).

24 While we acknowledge the need to identify government actions affecting various creditor groups, the focus throughout this paper is solely on government behavior towards private international creditors. Negotiations with official creditors, IFIs or domestic banks or investments funds are not taken into account.

17

The “Index of Government Coerciveness” during debt crises consists of 9 sub-indicators. These

are measures of observable government behavior and of rhetoric of crucial members of

government towards international creditors.25 Each sub-indicator is a dummy, which is coded 1 if

the respective action by the government is observed and zero otherwise. The final index is simple

additive measure with a lower bound of 1. The highest possible score is 10 and indicates the

highest level of government coerciveness. A score of 1 - on the other extreme - indicates that the

debt problem was resolved in a fully cooperative way without missed payments. Accordingly,

none of the 9 sub-categories of coercive behavior was coded as 1.

The Index and sub-indicators were coded based on a systematic and standardized evaluation of

main policy reports, all standard reference books on debt crises and more than 19,000 pages of

articles from the financial press. A detailed description of the coding process, the data and

stylized facts on many debt crisis cases from the 1980s until today can be found in Enderlein,

Müller and Trebesch (2008).

The 9 binary sub-indicators are the following:

1) Payments missed (yes/no)

The category “payments missed” is coded 1 if the government misses a due interest or

principal payment on bonds or loans owed to private international creditors. This includes

cases in which the government arranged a temporary roll-over of debt payments, but it does

not include missed payments that occurred within the grace period foreseen in the respective

debt contract.

25 We only consider actions and statements by core members of the government, including the president or prime minister, the economy and or finance ministry and the presidency of the central bank.

18

2) Unilateral payment suspension (yes/no)

Payments that are withheld unilaterally and without warning creditors are a clear sign of non-

cooperative, unilateral behavior. For this reason, we include the sub-indicator “unilateral

payment suspension”. It is coded 1 whenever the government incurs arrears unilaterally,

without agreeing with its creditors on a payment deferral and/or if creditors are not notified of

payment delays ahead of time.

3) Suspension of interest payments (yes/no)

The suspension of interest payment is coded as a separate indicator of payment behavior. The

reason is that a government that fully suspends interest payments sends a strong signal of its

unwillingness to service its debt, even at a reduced pace. A series of debtor governments have

explicitly ignored creditor demand for partial debt payments even rejecting to make symbolic

token payments on interest. The sub-indicator is coded 1 in case the government suspends

interest payments on sovereign bonds or public syndicated bank loans for more than 90 days

in a given year.

4) Freeze on assets of non-residents (yes/no)

In a series of crisis cases, governments issued emergency decrees that lead to an effective

freeze of creditor assets in the country, which should certainly be regarded as coercive

government behavior. We therefore also include a sub-indicator “freeze on assets of non-

residents”. It is coded 1 for any kind of additional capital or exchange controls that are

enacted during crisis years and that directly affect debt flows to foreign private creditors

(including private to private debt repayment).

19

5) Explicit moratorium or default declaration (yes/no)

The sub-indicator is coded 1 in case a key government actor, i.e. the President, the Prime

Minister, the chief debt negotiator or Ministers of Finance, Economy or Planning, or the

President of the Central Bank officially proclaims the decision to default. It is interesting to

note that most de facto moratoria were actually not officially declared. In most cases

governments have avoided to declare default publicly falling into arrears or starting debt

renegotiation without an official proclamation. Therefore, an official declaration of default

can be seen analogous to a declaration of war, and usually only takes place in an already very

conflictive situation.

6) Explicit threats to repudiate on debt (yes/no)

The sub-indicator is coded 1 if a key government actor publicly threatens to repudiate on debt

or to impose a unilateral moratorium. Such threats, often issued by populist governments, are

normally widely reported in the press and are generally thought to have a major public

impact.

7) Breakdown or refusal of negotiations (yes/no)

We coded 1 if either one of the following criteria applied: (i) the refusal of governments to

engage in early negotiations with creditors and (ii) delays or even breakdowns of debt

negotiations of more than 3 months that are caused by unilateral government behavior.

Delays that are caused by creditor coordination failure or outright inter-creditor disputes are

not coded.

20

8) Data disclosure problems (yes/no)

As a further proxy for negotiation behavior, we include a sub-indicator that is coded 1 if the

government explicitly refuses to provide timely information on crucial negotiation related

issues or if there is a dispute with creditors due to the provision of grossly inaccurate data.

The provision of accurate macroeconomic and financial data by debtor governments is of

high importance for private creditors seeking to understand the debtor government’s financial

stance and repayment capacity, in order to evaluate any potential restructuring offers.

9) Forced and non-negotiated restructuring (yes/no)

This indicator considers whether the restructuring was ultimately negotiated or not. It

captures instances (i) where the government enforced a fully unilateral restructuring or (ii)

where the government issued a non-negotiated offer on a final agreement. While most

modern-type bond restructurings involve a final, unilateral offer that is usually not amended

after it is launched, even those offers can be the result of a coordination and negotiation

process. The sub-indicator thus aims to differentiate between cases of close creditor

consultations and other restructurings, e.g. in Argentina in 2001 or 2005, where the

government rejected to engage in close negotiations before putting the offer to the market.

Additionally, we aim to capture cases of forced restructurings. This includes cases where

governments unilaterally decide to lower the interest rate on debt, or to restructure debt owed

by the private sector without any prior consultations (e.g. Argentina 1982).

21

Case selection and coding

Generally, we started to code cases from 1980 on.26 Regarding the selection of countries,

our list initially included all 136 developing and emerging economies in the Global Development

Finance database. First, we excluded countries which did not feature a default since 1980. Given

our focus on disputes between debtor governments and private creditors, we further excluded the

poorest, least developed countries (LDCs).27 The rational for excluding low income countries is

that they usually have very limited access to private financing and government lending tends to

be heavily dominated by debt to official creditors such as donor governments or the IMF.28

Clearly, the information base for coding was more comprehensive for large countries and

default cases. We excluded a number of countries from our dataset for which there was only

insufficient information available, even though they are not regarded as LDCs.29 In the absence

of a broader information base we decided to provide no information rather than risking to present

wrong or incomplete information on the behavior of governments. While this incurs the

possibility of bias we tried to control for possible effects of size and region in the regression

analyses later on. Nevertheless, we believe our data-set of government behavior can serve as an

adequate base of analysis.

26 The main reason for this is the difficulty of gathering sufficient information on government behavior in debt crises before 1980. 27 The main selection criterion was the United Nations definition of Least Developed Countries. Further non-LDC defaulters not considered were the low-income countries of Cameroon, Congo, Ghana, Guyana, Honduras, Kenya, Mongolia, and Zimbabwe and countries of former Yugoslavia (Bosnia and Herzegovina, Croatia, Macedonia, Serbia and Montenegro, Slovenia). 28 The debt restructuring process in these countries is mostly dominated by Paris Club and IMF talks while commercial creditors play a less important role. Moreover, negotiations with private creditors usually cover only small debt amounts and receive little attention in the press and in the literature. This makes it extremely difficult to draw any meaningful conclusions about public-private negotiations. Besides, Lex Rieffel notes that private financing follows a very different logic than financing by governments or IFIs: “commercial bank lending and bilateral donor agency lending are functionally quite distinct. The daily business of commercial banks is to make profit by pricing and managing credit in a huge global market place. (…) By contrast bilateral donor agencies make loans to developing country borrowers to advance various foreign policy objectives: economic growth, alleviation of poverty, regional stability, civil order, and the like” (Rieffel 2003:105). 29 These are Côte d’Ivoire, Gabon, Iran, Nicaragua, Trinidad und Tobago and Vietnam. The information available from the press and other sources on the 1980s and 1990s debt crises in these countries was not comprehensive enough to allow for objective and reliable coding. In contrast, small country debt restructurings of more recent years (e.g. Dominica, Grenada, and Moldova) were much better covered by the press and by detailed official policy reports.

22

Ultimately, we ended up coding cases in 31 countries, which featured a sovereign debt

crisis since the year 1980. Table 4 in the Appendix provides an overview on the default periods

covered, corresponding to 251 yearly events.30 The dataset used here is coded year-by-year. It

provides indicator values for every debt crisis year since 1980, i.e. capturing year-by-year

fluctuations in the 9 sub-indicators of government behavior during a crisis. An overview of the

development of the index since 1980 is provided in Figure 1 in the Appendix. As can be seen, the

average degree of coerciveness in crises has stayed relatively constant over the three decades,

although the index appears to have become more volatile in the more recent period.

5. Some Stylized Facts on Democracies in Default

Before going into the details of our estimation approach (section 6), it is useful to sketch

out some stylized facts on the behavior of democratic versus autocratic governments in default.

Looking at the coding results presented in Enderlein et al (2008), it is striking that all but one of

the most conflictive debt restructuring episodes since 1980 took place in democracies31, namely

in Argentina 2002-2005, Brazil 1987 and 1989, the Dominican Republic 1989 to 1990 and Peru

1985 to 1989. All of these crises are characterized by very coercive behavior of governments

towards external private creditors, indicated by an index value of 7 or higher. It is also notable

that, at the peak of the dispute with creditors, Argentina, the Dominican Republic and Peru all

were established democracies with at least 5 years of democratic rule. This gives further

motivation to test whether established democracies tend to behave more coercively in crisis

resolution (hypothesis 3). Brazil in 1987 and 1989 was still a young democracy, when its

government imposed exceptionally coercive actions on the country’s external creditors.

30 This includes the Republic of Yugoslavia but the defaults of the early 1990s in its follow-up Republics. 31 The exception is Nigeria in 1990-1991, which was not democratic at the time.

23

Nevertheless and despite the fact that democratic transition occurred only in 1985, domestic

audiences and opposition forces were already very strong in the late 1980s. As a result, electoral

pressure played a crucial role for the unilateral debt policies of President Sarney from 1987 to

1989.32

Autocracies, in contrast, have often behaved particularly cooperatively in past crises. The

debt crises in autocracies such as Algeria, Chile, Morocco and South Africa during the 1980s and

1990s were all resolved in a very consensual manner, with average index values of 3 or lower.

The example of autocratic Romania in the 1980s discussed in the introduction represents a

particularly extreme case. The descriptive statistics tell a similar story. The correlation between

democracy and the index of coercive government behavior is positive for all democracy measures

employed in the econometric analysis. Higher levels of democracy clearly coincide with higher

degrees of coerciveness.33 Figure 3 in the Appendix provides further visualization, by plotting the

index distribution grouped by regime type. As can be seen at first sight, democracies have a

considerably larger proportion of observations in the “high coerciveness” range of index values 5

or higher. In contrast, the distribution for non-democracies is rather hump-shaped, with most

country-year observations in the “low coerciveness” range. It is also insightful to distinguish

between particularly conflict-riddled and particularly cooperative country-year observations. The

average Polity Score for conflictive episodes - with an index value of 5 or higher - is 5.01 (79

yearly obs.). Contrarily, the average polity score is only 1.13 for episodes with a low level of

disputes, i.e. with an index value lower than 3 (72 yearly obs.).

32 The 1987 moratorium coincided with a deep political crisis, with mass protests and strong opposition against Sarney by the Congress. Many analysts at the time interpreted the conflictive debt policies as aimed to improve the government’s popularity. In 1989, President Sarney again adopts a conflictive stance in his debt policies towards international creditors, after having cooperated in 1988. This novel policy shift took place at a time of fastly eroding popularity and right after a devastating election result for his party. (Aggarwal 1996, p. 478, FT, 23 Febr. 1987, Reuters 16 Nov 1987, FT, 17 Nov 1988, Reuters, 25 July 1989) 33 As an example, the correlation of our index and the Polity IV Score is 0.22 for the sample of debt crisis years

24

All of this indicates that there seems to be a positive relation between the degree of

democracy and the degree of conflict that governments incur. However, it is necessary to validate

this relationship in a more systematic way by controlling for additional determining factors,

which we do in the following section.

6. Empirical Test

The aim of this section is to explore the relationship between regime type and government

behavior during crisis econometrically. We explain the estimation techniques and describe the

main explanatory variables and controls.

Econometric Technique

To test the role of democracy in debt crisis resolution, we construct a yearly panel dataset

for 27 developing countries that have defaulted on sovereign debt obligations held by private

creditors between 1980 and 2004 (a list is provided in Table 4 in the appendix).34

The above described index is used as proxy for coercive government behavior and taken

as dependent variable. One issue arising with this approach is certainly weighting. It is not

straightforward to decide which weight each of the 9 sub-indicators of government behavior

should have in the overall index. In our view, an additive index of each of the 9 dummy variables

34 Note that, as argued above, we consider only countries that had reasonable access to finance. Accordingly, LDCs and a number of other countries are excluded. The main selection criterion was the United Nations definition of Least Developed Countries. Further non-LDC defaulters not considered were the low-income countries of Cameroon, Congo, Ghana, Guyana, Honduras, Kenya, Mongolia, and Zimbabwe and countries of former Yugoslavia (Bosnia and Herzegovina, Croatia, Macedonia, Serbia and Montenegro, Slovenia). The main reason for this is the difficulty of gathering sufficient information on government behavior in debt crises before 1980. Moreover small countries with a population below 1,5 million are excluded due to notorious problems of data reliability in such countries, particularly in the 1980s.

25

is a good choice as it is straightforward and transparent. However, it might be appropriate to

employ more objective weighting methods when it comes to data analysis. Here, we resort to the

standard technique of principal component analysis (PCA) to derive weights of each sub-

indicator.35 The key idea behind principal component analysis is to summarize the information of

a set of variables in a smaller set of newly created continuous variables (principal components),

while retaining as much information as possible. For the data at hand, the first principal

component contains more than 30% of the variation of the original 9 sub-indicators. The

correlation between our simple additive index value (from 1 to 10) and the first principal

component is a high 0.99. It can thus be seen as a valid dependent variable for our purposes.36

Formally, we estimate variants of the following reduced form equation:

itititit uXDemocracyCOERC +++= γβα 1

where itCOERC is our measure of coercive government behavior in crisis years, α is a constant

or a vector of country fixed effects, itDemocracy is a measure of democracy, itX is a set of

economic and political control variables and itu are robust standard errors. All specifications

further include dummies for world regions (following the World Bank classification) and time

dummies (capturing the three decades since 1980).37

The baseline estimation methodology chosen is standard ordinary least squares (OLS) for

which all crisis-year observations are pooled in a cross-section. Given our aim to identify true

underlying effects, the results are further validated by employing additional estimation

35 Jolliffe (2002) provides an encompassing discussion of PCA techniques. 36 Note that, principal component analysis provides a dependent variable of government behavior with large variation in parameter values. As a result, standard regression techniques may be employed. 37 The dummy is included to account for changes in the structure of emerging market borrowing. While in the early 1980s syndicated bank loans dominated sovereign lending, the post-Brady Deal era is characterized by an increasing share of bond debt in total external debt, traded in deep secondary markets.

26

techniques, in particular panel and limited response models. Details on the robustness analysis

are discussed in section 7.2.

Main Explanatory Variables: Democracy and Democratic Experience

The main explanatory variable of interest here are measures of democracy. 38 To test

hypothesis 1 on the general role of regime type, we employ the democracy dummy developed by

Przeworski, Alvarez, Cheibub and Limongi (2000) and updated up to the year 2000 by Saiegh

(2005). After rescaling, it takes the value of 1 for democracies and the value of 0 for autocratic

regimes. As an alternative binary measure of democracy, we construct dummy variables based on

the widely used PolityIV composite index which ranges from +10 (very democratic) to -10 (very

undemocratic).39 Concretely, we build a dummy with a value of 0 (non-democracy) for Polity

scores of smaller or equal to 0, and a value of 1 (democracy) for Polity scores of 1 or higher. As a

further differentiation, we construct three additional dummy variables in the spirit of Jaggers and

Gurr (1995), Mansfield and Snyder (2002) and others, capturing three separate categories of

democracies. We look at “pure democracies”, “new democracies” and “pure autocracies” again

based on the Polity score.40

To test hypothesis 2, of whether higher degrees of democracy imply higher coerciveness,

we rely on continuous measures of democracy instead of the binary dummies used before. As

baseline continuous measure, we use the original Polity IV score from -10 to 10. In a second step,

38 We decided to test the regime type hypotheses using the three most widely used democracy indicators, to guarantee that results are not dependent on the variable and measurement approach employed. For a detailed comparison of institutional measures see for example Munck and Verkuilen (2002). 39 We employ the 2004 release of the Polity IV database. 40 We code polity scores from -6 to -10 as “pure autocracies”, values of 6 to 10 are coded as “pure democracies” and scores in between as “new democracies”. In doing so our categorization differs slightly from the initial classification of “coherent regimes” (democratic or autocratic) and “incoherent regimes” suggest by Jaggers and Gurr (1995). The main reason is that applying their categories we lost out a large number of cases in the category of pure autocracies. However, it should be noted that the direction of effects were not different when we applied the Jaggers and Gurr- categorization. They can be made available upon request.

27

we use an interaction term of the Polity score and the rescaled Przeworski et al dummy. This

variable excludes variations of autocratic rule but captures the level of democracy in those states

that fulfill the minimum criteria of Przeworski et al (2000). Thirdly, we also use a continuous

measure, based on the Freedom House (2006) data and coding criteria. Precisely, we take the

average of the indicators on “Political Rights" and ”Civil Liberties", which both range from 1 to

7. After rescaling, higher values of the aggregate Freedom House index indicate a more

democratic system.

The role of democratic experience (hypothesis 3) is tested in a sub-sample of democracies

only. To distinguish between young and more established democracies, we construct three

measures of experience with democratic institutions inspired by Vollmer and Ziegler (2008).

First, we include a dummy for “infant democracies” taking the value of 1 if a country

experienced less than 5 years of consecutive democratic rule (i.e. 1, 2, 3 or 4 years) and a related

dummy for “experienced democracies” with 5 or more years of consecutive democratic rule.

Additionally, we follow the example of Besley and Kudamatsu (2006) and create a variable

capturing contemporaneous measure denoting the fraction of democratic years between year t -4

and t.41

Control Variables

To avoid omitted variable bias and to better identify the immediate effect of democracy

for crisis resolution, we control for economic and financial conditions domestically, as well as

externally. The variables chosen are fairly standard and derived from the large empirical

literature on the determinants of default and debt distress that was summarized above (e.g.

41 Further empirical studies that employed similar categorizations to test the effect of different sub-categories of democracy and autocracy on the likelihood of militarized conflict and on growth and economic policies are Mansfield and Snyder (2002), and Rodrik and Wiarczig (2005) and Epstein et al. (2006) respectively.

28

Manasse, Roubini and Schimmelpfennig 2003 for an overview). First, we include two key debt

indicators capturing potential solvency and liquidity problems, namely the ratio of total external

debt to gross national income (GNI) and short term debt to reserves. The ratio of debt to GNI can

be seen as a good proxy for debt service pressure, while a high short term debt to reserves ratio

can indicate severe although temporary constrains in being able to repay foreign currency debt.

Higher levels of both variables are thus expected to increase the likelihood of coercive behavior.

To control for macroeconomic conditions in a given year, we include the log of inflation as well

as a variable capturing the GDP’s deviation from trend (in %).42 We also include a variable

capturing the overall weight of private creditors in government finances, namely the share of

government debt owed to private creditors in total public debt. All of these variables are taken

from the World Bank’s GDF and WDI databases.

Lastly, it is meaningful to control for the role of external factors, namely the global

interest rate (LIBOR) and the size of total capital flows towards developing and emerging

economies.43 While higher interest rates are expected to increase debt payment pressure and, thus,

coerciveness, higher capital flows towards developing countries are believed to lower constraints

and raise the opportunity costs for governments to behave non-cooperatively towards financial

market participants.

7. Results

7.1. Main Results

42 Calculated by using the Hodrick-Prescott filter. 43 Total flows to all countries included in the World Bank’s Global Development Finance dataset.

29

Results for Hypothesis 1: Are Democracies More Coercive Towards Private Creditors?

Table 1 summarizes the results regarding hypothesis 1. We tested whether, ceteris

paribus, democracies act more aggressively towards their private external creditors when they

enter a sovereign debt crisis. The estimations indicate that this is the case, underlining the validity

of the hypothesis. The coefficient for the democracy dummy is positive, highly significant and

has a sizable quantitative effect, even after controlling for a large number of economic variables

and when using different econometric techniques (see section 7.2.for details on the robustness

analysis). Regime type and democracy appears to be a main explanation of why governments

behave coercively towards creditors in periods of financial distress. As can be seen in Table 1,

this overall result is the same when using our baseline proxy of democracy from Przeworski et al.

(2000) or the dummies based on the Polity IV score.

Regarding the economic control variables, they are broadly in line with the literature and

theoretical predictions. The debt/GNI ratio, the proxy for liquidity constraints (short term debt to

exports) and the variable capturing high rates of inflation is significant and positive, indicating

that heavier financial constraints and economic instability are associated with more conflictive

government behavior. The variable capturing a higher share of debt owed to private creditors is

negatively signed. Governments appear to act more cooperative with private creditors when these

are crucial for the countries’ overall access to capital. The only surprising result is the negative

(albeit insignificant) coefficient of the global interest rate, as we expected higher interest rates to

increase financial and political pressure.

30

Table 1: Democracies vs. Autocracies in Debt Crisis Resolution

Model 1 Model 2 Model 3 Model 4with Dummy by Przeworski et al.

with Dummy based on Polity>0

with Dummy for Pure

Democracies

with Dummy for Pure

Autocracies

1.109***(0.222)

1.227***(0.259)

0.692***(0.240)

-0.958***(0.312)

0.013*** 0.013*** 0.012*** 0.013***(0.002) (0.003) (0.003) (0.003)

0.170*** 0.154*** 0.141*** 0.150***(0.028) (0.029) (0.030) (0.033)

-1.140*** -0.578 -0.891** -0.908**(0.415) (0.432) (0.430) (0.421)

0.123*** 0.085*** 0.082** 0.077**(0.033) (0.032) (0.032) (0.032)-0.214 -0.203 -0.096 -0.134(0.943) (1.012) (1.058) (1.043)-0.026 -0.038 -0.054 -0.065(0.046) (0.048) (0.049) (0.051)0.000 0.000 0.000 0.000

(0.000) (0.000) (0.000) (0.000)-2.106*** -1.396 -0.736 0.033

(0.682) (1.094) (1.151) (1.044)

Regional Dummies YES YES YES YES

Decade Dummies YES YES YES YES

Observations 202 206 206 206

Adj. R² 0.304 0.269 0.233 0.234

Democracy Dummy (Przeworski)Democracy Dummy (Polity>0)Pure Democracies Dummy (Polity ≥ 6)Pure Autocracies Dummy (Polity ≤ -6)

Note: Pooled OLS Regression. Dependent variable is the index of coercive government behavior weighted through PCA. ***/**/* denotes significance at a 1/5/10 % respectively. Robust standard errors in parentheses.

External debt / GNI

Short-term Debt / ReservesShare of Debt to Private CreditorsInflation (log)

GDP (% deviation from trend)Global Interest Rate (LIBOR)Total Capital Flows to Developing World

Constant

Results for Hypothesis 2: Do Higher Degrees of Democracy Imply Higher Coerciveness?

Departing from these baseline results, we go on to test whether the degree of democracy matters.

With a view to the theoretical arguments made above, we expected higher values on a democracy

scale to be associated with more conflictive crisis resolution patterns. The results in Table 2 give

strong support to this assertion. The continuous Polity variable is highly significant,

quantitatively important and positive. This is the case when using the original index score and

when interacting with the democracy dummy by Przeworski et al. Also the average Freedom

31

House score is a significant and important predictor for the degree of coercive actions imposed by

governments. Apparently, higher degrees of democracy amplify the effect in debt crisis

resolution.

Table 2: Degrees of Democracy and Coerciveness

Model 1 Model 2 Model 3with

Continuous Polity IV

Score

with Democracy Interaction

Term

with Continuous Freedom House

Score

0.063***(0.019)

0.087***(0.027)

0.227***(0.080)

0.012*** 0.010*** 0.014***(0.003) (0.003) (0.003)

0.151*** 0.165*** 0.145***(0.030) (0.032) (0.032)-0.836* -1.245*** -0.869**(0.428) (0.424) (0.428)0.082** 0.118*** 0.078**(0.032) (0.034) (0.032)-0.182 -0.402 -0.121(1.049) (0.997) (1.057)-0.051 -0.057 -0.058(0.049) (0.048) (0.048)0.000 0.000 0.000

(0.000) (0.000) (0.000)-0.463 -1.132 -1.164(1.090) (0.713) (1.102)

Regional Dummies YES YES YES

Decade Dummies YES YES YES

Observations 206 196 212Adj. R² 0.237 0.272 0.226

Polity x Dem.Dummy by Przeworski

Freedom House

Poility IV Score

External debt / GNI

Short-term Debt / ReservesShare of Debt to Private Creditors

Inflation (log)

GDP (% deviation from trend)Global Interest Rate (LIBOR)Total Capital Flows to Developing World

Constant

Note: Pooled OLS Regression. Dependent variable is the index of coercive government behavior weighted through PCA. ***/**/* denotes significance at a 1/5/10 % respectively. Robust standard errors in parentheses.

32

Results for Hypothesis 3: Are Established Democracies More Conflictive?

The third hypothesis predicted more experienced democracies to act more aggressively

compared with less experienced or unstable democracies. The results for the sub-sample44 of

democracies shown in Table 3, give support to the argument that experience with democracy

matters. All three measures capturing experience with democracy are significant with a sizable

quantitative effect (see table 3). Infant democracies, for which the transition to democracy lays

less than 5 years back, behave significantly less coercive towards external private creditors

during crises. In contrast, experienced democracies and countries with a higher share of

democratic years in recent history exert systematically more conflictive behavior.45

Table 3: The Role of Democratic Experience

44 The sub-sample only includes observations for which the yearly binary measure by Przeworski et al indicates a democratic regime. The results for hypothesis 3, however, are unaffected when the sub-sample of democracies is identified via dummies based on the Polity IV score, even when we include only observations of “pure democracies” with a Polity score of 6 or higher

45 Note that regional dummies are not included in the specifications of Table 3. The reason is that, in the smaller sub-sample of democracies, some of the regional dummies have a high correlation with the binary variables for young or established democracies thus leading to a potential bias in the estimations.

33

Model 1 Model 2 Model 3

Accounting for Infant

Democracies

Accounting for Established

Democracies

With Recent Democratic

History

-0.731**(0.336)

0.807**(0.330)

1.774***(0.512)

0.013*** 0.012*** 0.012***(0.004) (0.003) (0.003)

0.293*** 0.297*** 0.308***(0.074) (0.075) (0.064)-0.625 -0.524 -0.402(0.548) (0.554) (0.553)

0.127*** 0.129*** 0.129***(0.037) (0.037) (0.037)0.519 0.392 0.094

(1.018) (0.985) (0.999)-0.110* -0.107* -0.089(0.064) (0.063) (0.062)0.000 0.000 0.000

(0.000) (0.000) (0.000)-0.285 -1.086 -2.230**(0.857) (0.923) (1.008)

Regional Dummies NO NO NO

Decade Dummies YES YES YES

Observations 132 132 132

Adj. R² 0.196 0.203 0.223

Share of Democratic years in previous 5 years

Established Democracies (Dummy)

Total Capital Flows to Developing World

Note: Pooled OLS Regression. Dependent variable is the index of coercive government behavior weighted through PCA. ***/**/* denotes significance at a 1/5/10 % respectively. Robust standard errors in parentheses.

Constant

External debt / GNI

Short-term Debt / Reserves

Share of Debt to Private Creditors

Inflation (log)

GDP (% deviation from trend)Global Interest Rate (LIBOR)

Infant Democracies (Dummy)

7.2. Robustness Analysis

To verify the validity of our results, we conducted a series of robustness checks. We first

estimated the above models with a set of alternative specifications and additional variables.

Altogether, our main finding - the significant and positive coefficients of the democracy and

democratic experience measures - is very stable under different model specifications. In

particular, results are not significantly affected when including key institutional measures such as

constitutional system type, a proxy for veto players, or government polarization along ideological

34

lines.46 Also adding further economic variables such as GDP per capita, the balance of the current

account to GDP, the degree of openness (imports+exports/GDP) and several other standard

variables widely employed in the sovereign debt literature does not change our main findings, but

often decreases the sample due to missing values. The same is true for variables capturing the

scope of IMF involvement (IMF disbursement as % of quota or annual net contributions by the

IMF). We also included a variable that controls for duration dependence (number of years a

country is already in default) and a dummy for past defaults (in the previous 5 or 10 years).

Again, results remain very similar.

As a further main robustness check, we investigated in how far our results are driven by

the composition of our dependent variable, in particular if any of the 9 sub-indicators is particular

important for our findings. For this purpose, we used a PCA weighted dependent variable based

on 8 instead of 9 of the sub-indicators and excluded each of the individual indicators stepwise

from the overall index. We then ran all of the above regressions with this new set of dependent

variables. It turned out that none of the individual sub-indicators are crucial for our results.

Overall, the findings were highly robust to alterations in the dependent variable. In a similar vein,

we checked in how far the weighting approach via PCA affects the results. We thus used the

simple additive index of government behavior with numerical values from 1 to 10 as dependent

variable, instead of the continuous PCA weighted measure. Given the ordinal character of the

additive index, we resort to a standard ordered probit model estimated in the cross section. As

46 Constitutional system type is approximated by a dummy for purely presidential systems taken from the Database of Political Institutions (DPI) (Beck et al. 2001). The variable on government ideological polarization is also taken from DPI. We use a dummy for left government, which takes the value of 1 if the government is left oriented and zero otherwise. Political constraints are approximated by Henisz’s (2000) Polcon III measure.

35

can be seen in Table 6 of the Appendix, results are not affected, as all three hypotheses are

confirmed in an ordered probit estimation framework.47

In a next step, we evaluated whether the results change when standard random and fixed

effects panel data models are employed.48 Given the focus on democracy, it is more appropriate

to apply random effects estimation, as regime type shows only limited variability over time.

Nevertheless, we also check our results using fixed effects estimation. As shown in table 7 in the

Appendix, the results regarding hypotheses 1 and 2 hold in both, a random and fixed effects panel

estimation framework. However, the result regarding hypothesis 3 cannot be confirmed, as the

dummy for “established democracies” turns clearly insignificant.

Finally, we aim to account for the possibility that selection into default might not be

exogenous, i.e. that the group of defaulting countries shares some unobserved characteristics that

also affect government behavior.49 We thus set up a Heckman selection model, with “sovereign

debt distress”50 as binary dependent variable in the selection equation, and the pca weighted index

of government behavior as dependent variable in the primary equation. In order to account for

selection effects, we expand the sample by adding additional observations from 23 countries that

did not default in the period 1980 to 2004. To identify the model, we choose the share of bond

debt in total public debt and include it in the selection equation only. The choice of this

identifying variable appears valid, as the share of bond debt is a highly significant determinant of

47 We get similar results when estimating the ordered probit model in a panel framework, following the routine developed by Rabe-Hesketh et al. (2000) and Frechette (2001a, 2001b).

48 The panel is highly unbalanced given that our dependent variable is observed in crisis years only.

49 For obvious reasons, we can only observe government behavior during crises for crisis years, i.e. for the sub-sample of years in which countries actually default. Heckman (1979) pointed out that such incidental data truncation can lead to sample selection bias. 50 To identify distress years in the first stage (selection equation) we use our definition above, i.e. the standard binary default measure as of Standard & Poor's (2006) and add observations in which debt negotiations took place.

36

default, but an insignificant factor for the degree of coercive government behavior.51 The two-step

Heckman regression results are reported in Table 8 of the Appendix. As can be seen, the

estimations yield no evidence for selection effects. Lambda is clearly insignificant, even when

changing country sample or model specification, indicating that the error terms of the selection

and primary equation are not correlated. It thus seems appropriate to apply standard regression

techniques with no need to correct for selection bias. A comparison of the regression results also

shows that the coefficients of the sample selection models closely resemble those of the ordinary

OLS, panel and ordered probit estimations shown above.

Note, that the regressions for the Heckman selection model also reveal an interesting side

finding related to the existing body of literature on regime type and sovereign debt distress. In

our panel of 50 developing countries, we find that democracy, measured via Polity or the

Przeworski Dummy increases the probability of debt distress significantly. This finding in the

cross-section is in line with Saiegh (2005), Alichi (2008) and others who also found democracies

to have a higher likelihood of sovereign default or debt restructurings, but stands in contrast to

the democratic advantage hypothesis of Schultz and Weingast (2003). However, the picture

becomes less clear when running additional regressions on the determinants of debt distress. In

fact, the democracy dummy turns insignificant once the equation is estimated individually in a

panel probit or logit framework. The result is also not overly robust to changes in specification

and sample. We thus conclude that the link between regime type and sovereign default remains

unclear and somewhat contradictory. We can confirm, however, that democracies tend to behave

more conflictive towards their external creditors during default. This overall finding is very stable

to the choice of the estimation technique and to a large number of robustness checks.

51 Once a country defaults, the share of bond debt is unlikely to affect the general stance of governments towards its private creditors. However, the share of bond debt does matter for the likelihood of default. In fact, since World War II and until the late 1990s countries which had a large share of bond financing were more unlikely to default.

37

9. Conclusion

In early theoretical models of sovereign lending, governments weigh the costs of

defaulting on their foreign debt against the expected political gains. However, these models did

not differentiate between different types of government. The “democratic advantage”- approach

showed that government behavior towards lenders is likely to depend on the regime type. This

seems reasonable given that the “power calculus” for leaders to remain in office is different for

different types of political systems (Tullock 1987, Bueno de Mesquita et al. 2003 a.o.). Political

leaders in democracies are more constrained in their policy choices. From the “democratic

advantage”-perspective, the main channels through which preferences of a population are

translated into policy are free elections, direct checks on the government through legislative

controls and an independent judiciary (Schultz and Weingast 2003). Although useful to describe

constraints on government policies in general, this concept falls short of explaining why there are

still numerable occasions of noncompliance in democracies.

The puzzle can be solved if a more dynamic view of domestic audience costs of

adjustment policies are taken into consideration. As Tomz (2002) showed for the Argentine

crisis, citizens of democratic countries can force their political leadership into taking an

aggressive stance towards external creditors. Mass protests and general strikes can increase the

pressure on governments to extremes. Therefore, international debt contracts are unlikely to be

respected at any given time. Precisely in a setting where voters can constrain governments, they

may use their power to punish international commitments. At the core of this argument is the

recognition of time variant voter preferences following redistribution of income in the context of

economic crisis. For this to hold, governments have to be opportunistic in the sense that they

prefer short-term gains of power to long-term benefits to the economy. Second, the group of

38

voters that gains from default has to be more influential than the group with a stake in contract

compliance as was shown by Saeigh (2005).

Our empirical analysis of government behavior during debt crises yields three main

results. First, we find democratic governments to act more aggressively towards their creditors on

average. The existing literature which used a binary default variable remains inconclusive on the

effect of regime type on the likelihood of a default. We find unambiguous evidence that regime

type matters for the degree of coerciveness of external debt policies. 52

The second finding is that the level of democracy is important. The degree of

aggressiveness of government policies towards creditors is significantly higher at high levels of

democracy. The third finding we report is that experience with democratic institutions matters for

government behavior. Less experienced democracies act significantly less aggressive towards

creditors.

A significant positive effect on coerciveness can only be reported for countries which

have reached the threshold of five years of consecutive democratic rule. The finding for new

democracies contrasts with the “conventional wisdom” which claimed that leaders in new

democracies were unable to take unpopular reform measures at an early stage of democracy. Our

results hint at the effectiveness with which new democratic leaders exploited “honeymoon

effects” to implement unpopular adjustment rather than looking for debt relief from external

sources.

An important finding is finally that the empirical results show no difference in

government behavior over time. Government aggressiveness remains at the same average level

despite significant changes in the type and composition of investor groups from the 1980s until

52 We also tested the effect of regime type on default as measured by the widely used S&P-indicator and found a significantly positive effect of democracy on the likelihood of a default.

39

today. Apparently, it made little difference for political leaders whether they knew they had to

face a “banking committee” or if their opponents were dispersed bondholders.

Our findings on the high degree of aggressiveness in democratic countries are not to say

that investors should generally avoid democracies. Nevertheless, they should follow government

behavior and rhetoric closely; especially where the economic situation deteriorates significantly

creating new “champions of noncompliance” (Tomz 2002). Given that threats to engage in

external conflict are a means that democratic leaders rely on very rarely, they should be taken as

a credible sign of the resolve to “fight through” the crisis.

To get a better understanding of the effects of different ruling systems on sovereign debt

policy, further analysis of the variance among democracies and autocracies could yield important

insights. If there is variance in democracies it seems logical that there is also variance in

autocracies.53 According to by Bueno de Mesquita et al. (2003) and Weeks (2008), for example,

some types of autocracies have the capacity to raise audience costs and make credible threats.54

As this paper tried to highlight, further analysis should also take into account the effects of

experience with particular regime types.

53 This argument finds further theoretical foundation in the work of Hannah Arendt (1951) on the elements of totalitarism and differences between types of totalitarian systems and in Tullock (1987). 54 In their theoretical approach Bueno de Mesquita, Smith, Siverson and Morrow (2003) show that powerful “selectorates”, i.e. interest coalitions on which autocrats rely to stay in power, can serve an equivalent function to the electorate in a democratic system. If preferences of the “selectorate” are visible to the outside world, autocrats could be able to raise audience costs much in the same way as democrats.

40

Appendix

Table 4: Crisis Periods Covered in the Debt Disputes Database

Albania 1991-1995 Nigeria 1982-1991Algeria 1991-1996 Panama 1983-1996Argentina 1982-1993 Pakistan 1998-1999 2001-2005 Peru 1983-1997Belize 2006-2007 Philippines 1983-1992Bolivia 1980-1993 Poland 1981-1994Brazil 1983-1994 Romania 1981-1983Bulgaria 1990-1994 1986 Chile 1983-1990 Russia 1991-2000Costa Rica 1981-1990 South Africa 1985-1987Dominica 2003-2005 1989 Dom. Rep. 1982-1994 1993 2004-2005 Turkey 1981-1982Ecuador 1982-1994 Ukraine 1998-2000 1999-2000 Uruguay 1983-1991Grenada 2004-2005 2003 Jordan 1989-1993 Yugoslavia 1983-1988Mexico 1982-1990 Venezuela 1982-1990Moldova 2002 Morocco 1983-1990

41

Figure 1: The Index of Coerciveness Over Time

Figure 2: The Index of Coerciveness by Regime Type

02

46

810

Deg

ree

of C

oerc

iven

ess

0 1Non-Democracy Democracy

Index Distribution by Regime Type

Note: Scatterplot with index values grouped by regime type. The width of the horizontal bars represents the number of observations for each index value. In line with our definition, higher index values indicate more conflictive crisis resolution. The categorization of democracy versus non-democracy is based on the Dummy by Przeworski et al.

42

Table 5: Summary Statistics and Data Sources (Sample of Debt Crisis Years)

Variable Obs Mean Std. Dev. Min Max Data Source

Coerciveness Index (weighted by PCA) 238 0.00 1.67 -2.33 4.45 Enderlein, Müller and

Trebesch (2008)

Coerciveness Index (additive) 238 3.63 1.90 1.00 9.00 Enderlein, Müller and

Trebesch (2008)

Democracy Score (Polity)228 3.42 6.41 -8.00 10.00 Polity IV

Democracy Score (Freedom House) 226 0.61 0.49 0.00 1.00

Freedom House (average of “Political Rights" and ”Civil

Liberties"), invertedDemocracy Dummy (Pzerowski et al) 238 4.71 1.46 1.00 7.00 Przeworski et al (2000) and

Saiegh (2005), inverted

External debt / GNI 216 81.88 38.81 17.16 253.21 GDF (World Bank)

Short-term Debt / Reserves 221 1.99 2.90 0.05 24.00 GDF (World Bank)

Share of Public Debt to Private Creditors 227 0.52 0.23 0.02 0.90 GDF (World Bank)

Inflation (log) 227 2.49 4.00 -11.51 9.42 WDI (World Bank)

GDP (% deviation from trend) 230 -0.03 0.12 -0.51 0.36 GDF (World Bank),

authors calculationGlobal Interest Rate (LIBOR) 238 7.57 2.97 1.22 16.87 IFS (IMF)

Total Net Capital Flows to Developing World 236 127701.90 107291.30 -12104.42 674427.00 GDF (World Bank)

43

Table 6: Results in Ordered Probit Models

Model 1 Model 2 Model 3

with Dummy by Przeworski

with Polity IV Score

Accounting for Established

Democracies

0.792***(0.180)

0.043***(0.014)

0.623**(0.254)

0.010*** 0.008*** 0.009***(0.002) (0.002) (0.003)

0.138*** 0.116*** 0.233***(0.025) (0.023) (0.057)

-0.837** -0.595* -0.319(0.327) (0.305) (0.411)

0.099*** 0.060** 0.097***(0.029) (0.024) (0.032)-0.174 -0.083 0.357(0.713) (0.734) (0.732)-0.022 -0.042 -0.080*(0.036) (0.036) (0.047)0.000 0.000 0.000

(0.000) (0.000) (0.000)

Regional Dummies YES YES YES

Decade Dummies YES YES YES

0.336 -0.781 0.223(0.516) (0.721) (0.612)0.843 -0.284 0.573

(0.522) (0.730) (0.621)1.448*** 0.251 1.176*(0.520) (0.732) (0.618)

2.000*** 0.776 1.793***(0.518) (0.733) (0.616)

2.769*** 1.461** 2.625***(0.515) (0.738) (0.628)

3.367*** 2.005*** 3.174***(0.513) (0.720) (0.626)

3.741*** 2.363*** 3.584***(0.531) (0.728) (0.678)

4.199*** 2.956*** 3.985***(0.588) (0.765) (0.754)

Regional Dummies YES YES NO

Decade Dummies YES YES YES

Observations 202 206 132

Adj. R² 0.101 0.080 0.080

Share of Debt to Private Creditors

Inflation (log)

Established Democracies (Dummy)

Democracy Dummy (Przeworski)

Poility IV Score

/cut2

GDP (% deviation from trend)Global Interest Rate (LIBOR)Total Capital Flows to Developing World

/cut1

External debt / GNI

Short-term Debt / Reserves

/cut7

/cut8

Note: Ordered Probit Regression. Dependent variable is the ordinal index of coercive government behavior ranging from 1 to 10. ***/**/* denotes significance at a 1/5/10 % respectively. Robust standard errors in parentheses.

/cut3

/cut4

/cut5

/cut6

44

Table 7: Results in Panel Data Models

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Random Effects (Dem. Dummy)

Fixed Effects (Dem. Dummy)

Random Effects (Polity Score)

Fixed Effects (Polity Score)

Random Effects (Dem. Experience)

Fixed Effects (Dem. Experience)

0.932*** 0.835***(0.273) (0.320)

0.043** 0.041*(0.019) (0.021)

0.041 -0.229(0.294) (0.290)

0.018*** 0.022*** 0.022*** 0.026*** 0.023*** 0.031***(0.003) (0.004) (0.004) (0.004) (0.006) (0.007)

0.180*** 0.177*** 0.152*** 0.151*** 0.175*** 0.108*(0.034) (0.038) (0.034) (0.036) (0.062) (0.061)-0.906 -0.046 -0.426 0.369 -2.058** -4.637***(0.580) (1.020) (0.636) (1.133) (0.914) (1.506)

0.091*** 0.071** 0.046 0.030 0.075** 0.060*(0.032) (0.029) (0.030) (0.027) (0.038) (0.034)-0.039 0.294 0.121 0.523 1.382 2.051**(0.822) (0.860) (0.866) (0.856) (0.960) (0.992)-0.020 -0.023 -0.030 -0.035 -0.056 -0.004(0.039) (0.044) (0.040) (0.044) (0.056) (0.063)0.000 0.000 0.000 0.000 0.000 0.000

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Regional Dummies YES NO YES NO NO NO

Decade Dummies YES YES YES YES YES YES

-2.191*** -2.618*** -1.831* -2.236* -1.051 -0.151(0.615) (0.620) (1.023) (1.163) (1.022) (0.756)

Observations 202 202 206 206 132 132Adj. R² 0.271 0.270 0.264

Short-term Debt / Reserves

Share of Debt to Private Creditors

Inflation (log)

Democracy Dummy (Przeworski)

Poility IV Score

Established Democracies (Dummy)

External debt / GNI

Constant

Note: Fixed and Random Effects Panel Regression. Dependent variable is the index of coercive government behavior weighted through PCA. ***/**/* denotes significance at a 1/5/10 % respectively. Robust standard errors in parentheses.

GDP (% deviation from trend)Global Interest Rate (LIBOR)Total Capital Flows to Developing World

45

Table 8: Results in Heckman Selection Models

Primary Eq. (Coerciveness)

Selection Eq. (Default)

Primary Eq. (Coerciveness)

Selection Eq. (Default)

1.025*** 0.423***(0.261) (0.158)

0.064*** 0.037***(0.023) (0.012)

0.009* 0.021*** 0.012** 0.021***(0.005) (0.002) (0.006) (0.002)

0.149*** 0.113** 0.153*** 0.119***(0.040) (0.044) (0.045) (0.043)

-1.471*** 1.596*** -0.805 1.472***(0.534) (0.282) (0.585) (0.265)

0.104*** 0.071*** 0.083*** 0.053***(0.032) (0.022) (0.032) (0.019)-0.049 -0.987 -0.209 -1.119*(0.983) (0.699) (1.050) (0.641)0.006 -0.090*** -0.054 -0.090***

(0.059) (0.030) (0.064) (0.028)0.000 0.000 0.000 0.000

(0.000) (0.000) (0.000) (0.000)-1.976*** -1.442***

(0.447) (0.386)

Regional Dummies YES YES YES YES

Decade Dummies YES YES YES YES

-1.976*** -1.442***(0.447) (0.386)

-1.260 -8.552 -0.587 -3.136***(0.926) (1.666) (0.332)

Lambda

Observations

Share of Bonds in Total Debt

Share of Public External Debt in Bond Form

Constant

Heckman Selection Model with Polity Score

1,002(0.416) (0.504)

GDP (% deviation from trend)Global Interest Rate (LIBOR)Total Capital Flows to Developing World

Heckman Selection Model with Democracy Dummy

Democracy Dummy (Przeworski)

Poility IV Score

External debt / GNI

Short-term Debt / ReservesShare of Debt to Private Creditors

Inflation (log)

Note: Two-Step Heckman Selection Model. The binary dependent variable in the selection equiation is debt distress, based on the S&P criterion of default and details on the start of negotiations. The dependent variable in the primary equiation is the index of coercive government behavior weighted through PCA.The country sample included in the selection equation is the following: Albania, Algeria, Argentina, Armenia, Azerbaijan, Belarus, Bolivia, Brazil, Bulgaria, Chile, Colombia, Costa Rica, Zech Rep., Dominican Rep., Ecuador, Estonia, Georgia, Hungary, India, Kazakhstan, Kyrgyz Rep.,Latvia, Lithuania, Malaysia, Mexico, Moldova, Morocco, Nigeria, Oman, Pakistan, Panam, Papua New Guinea, Philippines, Poland, Romania, Russia, Sri Lanka, Tajikistan, Thailand, Tunisia, Turkey, Ukraine,Uruguay and Venezuela. ***/**/* denotes significance at a 1/5/10 % respectively.

-0.437 0.043

1,200

46

Literature

Acemoglu, Daro/Johnson, Simon/Robinson, James A. 2001: The Colonial Origins of Comparative Development: An Empirical Investigation, in: American Economic Review 91:5, pp. 1369-1401.

Aggarwal, Vinod K. 1996: Debt Games: Strategic Interaction in International Debt Rescheduling, New York: Cambridge University Press.

Alichi, Ali 2008: A Model of Sovereign Debt in Democracies. IMF Working Paper. 08/152, International Monetary Fund.

Andritzky, Jochen 2006: Sovereign Default Risk Devaluation - Implications of Debt Crises and Bond Restructurings, Berlin, Heidelberg, New York: Springer.

Arendt, Hannah 1951: The Origins of Totalitarianism, New York: Harcourt: Brace Jovanovich.

Barro, Robert J. 1996: Democracy and Growth, in: Journal of Economic Growth 1:1, pp.1-27.

Beck, Thorsten/Clarke, George/Groff, Alberto/Keefer, Philip/Walsh, Patrick 2001: New Tools in Comparative Political Economy: The database of political institutions, in: The World Bank Economic Review 15:1, pp. 165-176.

Beim, David O./Calomiris, Charles W. 2001: Emerging Financial Markets, New York: McGraw-Hill.

Besley, Timothy/Kudamatsu, Masayuki 2006: Health and Democracy, in: American Economic Review 96:2, pp. 313-318.

Borensztein, Eduardo/Panizza, Ugo 2006: Do Sovereign Defaults hurt Exporters? IADB Working Paper 1018.

Boughton, James M. 2001: Silent Revolution: The International Monetary Fund 1979-1989, Washington, D.C.: The International Monetary Fund.

Bueno de Mesquita, Bruce/Smith, Alastair/Siverson, Randolph M./Morrow, James D. 2003: The Logic of Political Survival, Cambridge, MA: The MIT Press.

Clark, Gregory 1996: The Political Foundations of Modern Economic Growth, England 1540-1800, in: Journal of Interdisciplinary History 26:4, pp. 563-588.

Clark, William Roberts/Flores, Thomas /Kaufman, Robert 2005: What’s so good about democracy? Working Paper. University of Michigan.

Cline, William R. 1995: International Debt Reexamined, Washington, D.C.: Institute for International Economics.

47

Cline, William R. 2004: Private Sector Involvement in Financial Crisis Resolution: Definition, Measurement, and Implementation, in: Haldane, Andrew (Ed.): Fixing Financial Crises in the Twenty-first Century, London: Routledge, pp. 61-94.

Desch, Michael C. 2002: Democracy and Victory: Why Regime Type Hardly Matters, in: International Security 27:2, pp. 5-47.

Detraigache, Enrica/Spilimbergo, Antonio 2001: Crises and Liquidity: Evidence and Interpretation, in: IMF Working Paper 01:2.

Drake, Paul 1989: Debt and Democracy in Latin America, 1920s-1980s. In: Barbara Stalling/Robert Kaufman (Eds.); Debt and Democracy in Latin America. Boulder, Co: Westview Press, pp. 39-58.

Drazen, Allan 1998: Towards a Political Economic Theory of Domestic Debt, in: Calvo, Guillermo/King, Mervyn (Eds.): The Debt Burden and Its Consequences for Monetary Policy, London: Macmillan, pp. 159-176.

Enderlein, Henrik/Müller, Laura/Trebesch, Christoph 2008: Sovereign Debt Disputes. Testing the Role of Politics and Institutions in Financial Crisis Resolution (Political Economy of International Finance Workshop, Claremont, CA, May 2008).

Epstein, Stephan R. 2000: Freedom and Growth: The Rise of States and Markets in Europe, 1300- 1750, London: Routledge.

Epstein, David L./ Bates, Robert /Goldstone, Jack /Kristensen, Ida/O’Halloran, Sharyn 2006: Democratic Transitions, in: American Journal of Political Science 50:3, pp.551-569.

Eyerman, Joe/Hart, Robert A. Jr. 1996: An Empirical Test of the Audience Cost Proposition: Democracy Speaks Louder than Words, in: The Journal of Conflict Resolution 40:4, pp. 597-616.

Fearon, James D. 1994: Domestic Political Audiences and the Escalation of International Disputes, in: The American Political Science Review 88:3, pp. 577-592.

Fostel, Ana /Kaminsky, Graciela L. 2007: Latin America’s Access to International Capital Markets: Good Behavior or Global Liquidity?, in: Cowan, Kevin/Edwards, Sebastian /Valdés, Rodrigo (Eds.): Current Account and External Financing, Santiago de Chile: Central Bank of Chile.

Frankel, Jeffrey A./Roubini, Nouriel 2001: The Role of Industrial Country Policies in Emerging Market Crises, in: NBER Working Paper Series 8634, pp. 1-110.

Freedom House 2006: Annual Survey of Freedom Country Scores, 1972-73 to 1999-06, retrieved from: http://www.freedomhouse.org (2008-08-07).

Frechette, G. 2001a: Random-effects ordered probit, Stata Technical Bulletin 59, 23-27. Reprinted in Stata Technical Bulletin Reprints 10, 261-266.

48

Frechette, G. 2001b: Update to random-effects ordered probit, StataTechnical Bulletin 61, 12. Reprinted in Stata Technical Bulletin Reprints 10, 266- 267.

Frieden, Jeffrey 1988: Classes, Sectors, and Foreign Debt in Latin America, in: Comparative Politics 21:1, pp. 1-20.

Frieden, Jeffrey 1989: Winners and Losers in the Latin American Debt Crisis: The Political Implications. In: Barbara Stalling/Robert Kaufman (Eds.); Debt and Democracy in Latin America. Boulder, Co: Westview Press, pp. 39-58.

Geddes, Barbara 2004: Authoritarian Breakdowns. NEW Working Paper. University of California Los Angeles.

Gelos, Gaston/Sahay, Ratna /Sandleris, Guido 2004: Sovereign Borrowing by Developing.

Countries: What Determines Market Access?, in: IMF Working Papers 04:221.

Gelpi, Christopher F./Griesdorf, Michael 2001: Winners or Losers? Democracies in International Crisis, 1918-94, in: The American Political Science Review 95:3, pp. 633-647.

Glaeser, Edward L./La Porta, Rafael/Lopez-De-Silanes, Florencio/Shleifer, Andrei 2004: Do Institutions Cause Growth?in: Journal of Economic Growth , pp. 271-303.

Haggard, Stephan/Kaufman, Robert (Eds.) 1992: The Politics of Economic Adjustment+ Princeton, N.J.: Princeton University Press.

Haggard, Stephan/Kaufman, Robert1995: The Political Economy of Democratic Transitions+

Princeton, N.J.: Princeton University Press. Haggard, Stephan/Webb, Steven (Eds.) 1994: Voting for Reform: Democracy, Political

Liberalization, Distributive Conflicts, and Economic Adjustment. New York: Oxford University Press.

Hatchondo, Juan Carlos/Martinez, Leonardo/ Sapriza, Horacio 2007: The Economics of Sovereign Defaults, in: Economic Quarterly 93(2), pp.163-187.

Heckman, James J. 1979: Sample Selection Bias as a Specification Error, in: Econometrica 47:1, pp. 153–161.

Henisz, Witold J. 2000: The Institutional Environment for Economic Growth, in: Economics and Politics 12:1, pp. 1-31.

Hirschman, Albert O. 1987: The political economy of Latin American development: Seven exercises in retrospection. Latin American Research Review 22: pp.7-36.

IIF 2006: Principles for Stable Capital Flows and Fair Debt Restructuring In Emerging Markets. Report on Implementation by the Principles Consultative Group, Washington, D.C.

IMF 1999: IMF Policy on Lending into Arrears to Private Creditors, Washington, D.C.

49

IMF 2002: Fund Policy on Lending into Arrears to Private Creditors—Further Consideration of the Good Faith Criterion, Washington, D.C.

Jaggers, Keith/ Gurr, Ted Robert 1995: Tracking Democracy’s Third Wave with the Polity III Data. Journal of Peace Research 32: 4, pp. 469 -82.

Jensen, Nathan M. 2003: Democratic Governance and Multinational Corporations: Political Regimes and Inflows of Foreign Direct Investment, in: International Organization 57:3, pp. 587-616.

Jensen, Nathan M. 2008: Political Risk, Democratic Institutions, and Foreign Direct Investment, forthcoming in: Journal of Politics 70: 4.

Jolliffe, Ian T. 2002: Principal Component Analysis, New York: Springer.

Kaufman, Robert R. 1985: Democratic and Authoritarian Responses to the Debt Issue: Argentina, Brazil, Mexico, in: International Organization 39:3, pp. 473-503.

Kaufman, Robert R. 1986: Democratic and Authoritarian Responses to the Debt Issue: Argentina, Brazil, Mexico. In: Miles Kahler (Ed.) The Politics of International Debt. Ithaca: Cornell University Press, pp. 187-217.

Kohlscheen, Emanuel 2007: Why are there Serial Defaulters? Evidence from Constitutions, in: The Journal of Law and Economics 50:4, pp. 713-730.

Kraay, Aart/Nehru, Vikram 2004: When Is External Debt Sustainable?, in: World Bank Policy Research Working Paper 3200.

Lake, David A. 1992: Powerful Pacifists: Democratic States and War, in: The American Political Science Review 86:1, pp. 24-37.

Leblang, David 2005: Pegs and Politics. Paper presented at the 2005 Meeting of the American Political Science Association, Washington, DC.

Lektzian, David/Souva, Mark 2003: The Economic Peace Between Democracies: Economic Sanctions and Domestic Institutions. Journal of Peace Research 40(6), pp. 641–660.

Linz, Juan 1978: Crisis, breakdown, and reequilibration. In: Juan Linz/Alfred Stepan (Eds.),The breakdown of democratic regimes. Baltimore: Johns Hopkins University Press, pp. 3-124.

Linz, Juan J./Stepan, Alfred 1996: Problems of democratic transition and consolidation: Southern

Europe, South America, and post-communist Europe. Baltimore: Johns Hopkins University Press.

Manasse, Paolo/Roubini, Nouriel/Schimmelpfennig, Axel 2003: Predicting Sovereign Debt Crises, in: IMF Working Paper 03:221.

50

Mansfield, Edward D./Milner, Helen V./Rosendorff, B. Peter 2002: Why Democracies Cooperate More: Electoral Control and International Trade Agreements, in: International Organization 56:3, pp. 477-513.

Mansfield, Edward D./Snyder, Jack 2002: Democratic Transitions, Institutional Strength, and War, in: International Organization 56: 2, pp. 297-337.

Marchesi, Silvia 2003: Adoption of an IMF Programme and Debt Rescheduling. An Empirical Analysis, in: Journal of Development Economics 70:2, pp. 403-423.

McGillivray, Fiona/Smith, Alastair 2008: Punishing the Prince: A Theory of Interstate Relations, Political Institutions, and Leader Change. Princeton: Princeton University Press.

Munck, Gerardo L./Verkuilen, Jay 2002: Conceptualizing and Measuring Democracy: Evaluating Alternative Indices, in: Comparative Political Studies 35:1, pp. 5-34.

North, Douglass C./Weingast, Barry R. 1989: Constitutions and Commitment: The Evolution of Institutional Governing Public Choice in Seventeenth-Century England, in: The Journal of Economic History 49:4, pp. 803-832.

O’Donnell, Guillermo 1978: Reflections on the Patterns of Change in the Bureaucratic- Authoritarian State, in: Latin American Research Review 13:1, pp. 3-38.

Oneal, John R. 1994: The Affinity of Foreign Investors for Authoritarian Regimes, in: Political Research Quarterly 47:3, pp. 565-588.

Olson, Mancur 1991: Autocracy, Democracy, and Prosperity, in: Zeckhauser, Richard (Ed.): Strategy and Choice, Cambridge, MA: MIT Press., pp. 131-157.

Panizza, Ugo/Sturzenegger, Federico/Zettelmeyer, Jeromin 2008: The Economics and Law of Sovereign Debt and Default. Washington, D.C.: IMF Working Paper.

Partell, Peter J./Palmer, Glenn 1999: Audience Costs and Interstate Crises: An Empirical Assessment of Fearon's Model of Dispute Outcomes, in: International Studies Quarterly 43:2, pp. 389-405.

Pescatori, Andrea/Sy, Amadou N. R. 2007: Are Debt Crises Adequately Defined?, in: IMF Staff Papers 54:2, pp. 306-337.

Przeworski, Adam/Alvarez, Michael E./Cheibub, José Antonio/Limongi, Fernando 2000: Democracy and Development. Political Institutions and Well-Being in the World, 1950–1990, Cambridge: Cambridge University Press.

Remmer, Karen L. 1990: Democracy and Economic Crisis: The Latin American Experience. World Politics, 42:3, pp. 315-335.

Rabe-Hesketh, Sophia/Pickles, Andrew/Taylor, Colin 2000: Generalized linear latent and mixed

models, in: Stata Technical Bulletin 53, pp. 47-57.

51

Reinhart, Carmen M./Rogoff, Kenneth S. 2008: This Time is Different: A Panoramic View of Eight Centuries of Financial Crises, in: NBER Working Paper 13882.

Reinhart, Carmen M./Rogoff, Kenneth/Savastano, Miguel A. 2003: Debt Intolerance, in: Brookings Papers on Economic Activity 1, pp. 1-70.

Rieffel, Lex 2003: Restructuring Sovereign Debt: The Case for Ad Hoc Machinery, Washington, D.C.: Brookings Institution Press.

Rodrik, Dani/Wacziarg,Romain 2005: Do Democratic Transitions Produce Bad Economic Outcomes? In: American Economic Review 95: 2, p.50-55.

Van Rijckeghem, Caroline/Weder, Beatrice 2004: The Politics of Debt Crises, in: CEPR Discussion Paper Series 4683, pp. 1-42.

Robinson, James A. 1998: Debt Repudiation and Risk Premia: The North-Weingast Thesis Revisited, in: Harvard University, Weatherhead Center for International Affairs, retrieved from: http://www.wcfia.harvard.edu/node/785 (2008-08-07).

Roubini, Nouriel 2004: Private Sector Involvement in Crisis Resolution and Mechanisms for Dealing with Sovereign Debt Problems, in: Haldane, Andrew (Ed.): Fixing Financial Crises in the Twenty-First Century, London: Routledge.

Roubini, Nouriel/Setser, Brad 2004: Bailouts or Bail-ins? : Responding to Financial Crises in Emerging Economies, Washington, D.C.: Institute for International Economics.

Saiegh, Sebastian M. 2005: Do Countries have a Democratic Advantage? Political Institutions, Multilateral Agencies, and Sovereign Borrowing, in: Comparative Political Studies 38:4, pp. 366-387.

Schultz, Kenneth A./Weingast, Barry R. 2003: The Democratic Advantage: Institutional Foundations of Financial Power in International Competition, in: International Organization 57:1, pp. 3-42.

Slantchev, Branislav L. 2004: The Watchful Eye: Information Transmission and Political Failure, in: Univ. of California San Diego Working Paper September 9.

Standard & Poor's 2006: Default Study: Sovereign Defaults At 26-Year Low, To Show Little Change In 2007 (Sep 18, 2006).

Stasavage, David 2007: Cities, Constitutions, and Sovereign Borrowing in Europe, 1274–1785, in: International Organization 61:3, pp. 489-525.

Stokes, Susan (Ed.) 2001: Public Support for Market Reforms in New Democracies New York: Cambridge University Press.

Sturzenegger, Federico/Zettelmeyer, Jeromin 2006: Debt Defaults and Lessons from a Decade of Crises, Cambridge, MA: The MIT Press.

52

Sussman, Nathan/Yafeh, Yishay 2006: Constitutions and Commitment: Evidence on the Relation Between Institutions and the Cost of Capital, in: Journal of Economic History 66:4, pp. 906-935.

Suter, Christian 1992: Debt Cycles in the World-Economy. Foreign Loans, Financial Crises, and Debt Settlements, 1820-1990, Oxford: Westview Press.

Tomz, Michael 2002: Democratic Default: Domestic Audiences and Compliance with International Agreements (Annual Meeting of the American Political Science Association, Boston, August 29 - September 1, 2002).

Tomz, Michael 2007a: Reputation and International Cooperation: Sovereign Debt across Three Centuries, Princeton, NJ: Princeton University Press

Tomz, Michael 2007b: Domestic Audience Costs in International Relations: An Experimental Approach, in: International Organization 61, Fall 2007, pp. 821-40.

Tullock, Gordon 1987: Autocracy, Dordrecht, The Netherlands: Kluwer Academic Publishers.

Vollmer, Sebastian/Ziegler, Maria 2008: Political Institutions and Human Development Does Democracy Fulfill its “Constructive” and “Instrumental” Role?, unpublished manuscript, Georg-August-University Göttingen.

Williamson, John/Haggard, Stephan 1994: The Political Conditions for Economic Reform. In: J.ohn Williamson (Ed.): The Political Economy of Policy Reform. Washington, D.C.: Institute for International Economics, pp.527-596.

World Bank 2002: Global Development Finance, Washington, D.C.

World Bank 2003: Global Development Finance, Washington, D.C.