Government Quality Improvements and Sudden Stops Incidence (Part 1)

Do Improvements in Government Quality Necessarily Reduce the Incidence of Costly Sudden Stops? Department of Economics April 2007 Abstract Sudden stops have been linked to a number of financial crises in emerging market countries. While a large literature has developed emphasizing the importance of institutions and governance in reducing economic volatility, this paper finds that the effect of government quality on the incidence of sudden stops is non-linear. Initial improvements in governance actually increase the incidence of costly sudden stops. A possible explanation is that improved governance encourages capital inflows that can overwhelm banking systems in countries with weak institutions. What is striking is that this result holds for a large number of countries including those with average levels of institutional quality that already receive considerable inflows. Eventually, however, improving institutions does reduce the frequency of sudden stops, allowing countries to enjoy the benefits of financial globalization with fewer risks. JEL Classification: F32, F36, F41. Keyword(s): Sudden stops, capital flows, institutions. _____________ 1. Introduction The question of why some societies suffer from severe volatility, financial crises and contagion is one of the most critical in international macroeconomics, and there is a growing literature emphasizing the importance of institutions and governance in determining these macroeconomic outcomes (Acemoglu, et al., 2003; Tommasi, 2002). A central part of this question is the effect of institutions on “sudden stops,” or large and sudden reversals in net capital inflows caused by foreigners’ sudden refusal to hold liabilities of the stricken country. Sudden stops are generally associated with large exchange rate depreciations and financial crises, leading to sharp contractions in investment and output. While international investors are partially to blame, it seems reasonable that institutional quality is a key determinant of the incidence of sudden stops. First, the quality of fiscal and monetary policy is in part determined by a country’s fiscal and monetary institutions. Reckless fiscal policy has been implicated in a number of sudden stops (Calvo, 2003) as it results in unsustainable levels of public debt, especially when excessive borrowing from international capital markets is channeled into unproductive government expenditure. Second, emerging markets vary in their degree of domestic liability dollarization, i.e. the domestic component of “Original Sin” (Eichengreen, et al., 2002; Hausmann and Panizza, 2003). Unhedged foreigncurrency-denominated liabilities are a major source of vulnerability for both firms and banks because large depreciations can lead to significant reductions in net worth (Aghion, et al., 2001; De Nicoló, et al., 2005; Mishkin, 1996). Domestic liability dollarization has been shown empirically to increase the probability of a sudden stop (Calvo, et al., 2004) as well as the cost associated with it (Cavallo, 2004; Guidotti, et al., 2004). De Nicoló, et al. (2005) and Honig (2002) find that institutional quality is a key determinant of domestic liability dollarization as it 2 influences policy credibility and expected relative volatilities of the real exchange rate and domestic inflation. Therefore, countries with good institutions are less likely to experience a recession after a reversal in capital flows and thus have a low frequency of costly sudden stops. Finally, better regulation and supervision of the banking system improves credit allocation and reduces the probability of a capital flow reversal. There is also evidence to suggest that improved governance reduces vulnerability to crises. Corruption has been linked to the share of foreign direct investment in inflows, a more stable source of funding (Wei and Wu, 2002). Good institutions augment the positive effect of capital account liberalization on both financial development (Chinn and Ito, 2005) and growth (Alfaro, et al., 2003; Arteta, et al., 2001; Klein, 2005). Prasad et al. (2004) find that institutional quality has a similar effect on the impact of financial globalization on growth and consumption volatility. It is possible, however, that improved governance can increase the risk of a costly sudden stop, defined as a reversal in capital flows accompanied by a contraction in output. One reason is that improved governance encourages capital inflows. Alfaro, et al. (2003) find that institutional quality was a key determinant of capital inflows in the period 1971-1998. Countries with better institutions, such as secure property rights and non-corrupt governments, invest more in physical and human capital and use these factors more efficiently, leading to higher returns to lending. In addition, as Fernandez-Arias and Montiel (2003) argue, institutional reforms reduce the perceived risk on domestic investment. As institutional quality improves and foreign lending becomes a more important source of funding, a capital flow reversal, when it does occur, involves a greater cost in terms of reduced investment and production. 3 This paper finds that for low levels of government quality, this effect dominates the positive effects of improved governance so that initial improvements increase the frequency of sudden stops. A likely explanation is that when institutional quality in the form of bank regulation and supervision is still weak, increased capital inflows will be channeled to firms with weak fundamentals or that assume excessive risk, exacerbating existing problems in the financial system (Ishii and Habermeier, 2002; Bakker and Chapple; 2002). This poses serious risks to bank balance sheets as worsening macroeconomic conditions result in a large share of nonperforming loans, thus sowing the seeds of a sudden stop. Greater inflows could also exacerbate fiscal problems by allowing excessive borrowing that is channeled to unproductive government spending. At the other end of the spectrum, countries with good institutions are better able to handle inflows. Their regulation and supervision of the banking system is far superior and, as mentioned previously, countries with better governance receive more FDI relative to total inflows. They are also less likely to enter a recession if a reversal in capital flows does occur because there is less liability dollarization (Honig, 2002), reducing the probability that they will experience a sudden stop (Calvo and Mishkin, 2003). Denmark, for example, experienced six episodes of significant capital flow reversals between 1982 and 2004, but none caused a contraction in output. Thus there is an inverted U-shaped relationship between government quality and the incidence of sudden stops.1 1 It is important to distinguish this result from previous work that has found a non-linear relationship between institutions, liberalization and growth/volatility. Ranciere, et al. (2006a) argue that for countries with a moderate degree of contract enforceability, financial liberalization increases the risk of crises although at the same time leads to higher mean growth by allowing financially constrained firms to increase their leverage and increase their investment. Ranciere, et al. (2006b) finds that financial liberalization leads to faster long-run growth, even though it has increased the incidence of twin banking and currency crises. Klein (2005) finds that the effect of capital account openness on growth depends on the level of institutional quality with the same inverted U-shaped relationship. Specifically, capital account liberalization promotes growth when a country has good institutions, but there is not a significant effect of capital account liberalization on growth for countries with either very poor institutions or those 4 It is important to note that the initial positive slope is not an artifact of the fact that countries with very poor institutions do not receive large capital inflows relative to GDP and therefore cannot have a costly sudden stop. It is certainly true that this observation is consistent with the non-linear relationship found in this paper and is in part an explanation for it. Specifically, countries with very poor institutions that do not receive large inflows relative to GDP are unlikely to experience a recession-inducing sudden stop, even if there is a drop in inflows. Meanwhile, advanced countries receive more inflows but are better able to handle them. “Intermediate” countries may therefore be the most vulnerable to sudden stops that cause contractions in economic activity, implying a non-linear relationship. This implies that institutional development in intermediate countries is sufficient to attract capital inflows, but not to prevent large capital outflows or their contractionary effects. This reasoning, however, is only part of the story. What is more striking is that the non-linear relationship holds even when countries with the worst institutions are excluded. The initial positive relationship between government quality and the incidence of sudden stops holds for a large number of countries in the sample, including those with average institutional quality that receive considerable inflows. Examples include Thailand, Mexico, the Philippines and Brazil. Thus, for a significant range of countries that are already receiving significant foreign capital, improvements in government quality can increase the risk of a sudden stop. The intuition is that even in countries that receive capital inflows and are therefore eligible to have sudden stops, improvements in government quality that generate additional inflows can still increase the risk of sudden stops if the risks inherent in those with the very highest quality institutions. Prasad, et al. (2004) show that countries that have undergone financial liberalization at low to moderate levels of governance have experienced an increase in consumption volatility. But none of these papers find that improvements in government quality can increase risk in the short run. Rather, improvements in government quality are unambiguously welfare-enhancing as they unleash the growth-enhancing and volatility-reducing effects of liberalization. 5 increased inflows outweigh the positive effects of institutional improvement. In other words, institutional improvements have two effects, to increase inflows but also improve how those inflows are allocated, and for low levels of government quality, this paper finds that the first effect dominates. Thus, while it is interesting that a non-linear relationship exists, it is far more significant that the location of the “hump” falls to the right of institutional quality levels of many emerging market countries.2 The rest of the paper is organized as follows. Section 2 presents the empirical methodology and the data. Section 3 discusses the results of the estimation while Section 4 describes robustness checks. Section 5 summarizes the findings and draws policy implications. 2. Empirical Methodology 2.1 The effect of institutions on the incidence of sudden stops3 To estimate the effect of government quality on the frequency of sudden stops, I estimate the following Poisson model for count data using annual data from 1984-2004: 2 # SuddenStopsi = β 0 + β1 GovQual i ,t + β 2 GovQual i ,t + γ ' MacroControls i ,t + ε i ,t (1)4 The dependent variable is the number of sudden stops that country “i” experiences during the sample period. The Poisson regression model has been used to analyze discrete data with 2 This result introduces an important caveat to the finding that improving institutions reduces economic volatility (Acemoglu, et al., 2003). 3 In order to conclude that improvement in institutions can, for some countries, increase the incidence of sudden stops by encouraging capital inflows, it is first necessary to establish a causal link from institutions to inflows. This result has been established by Alfaro, et al. (2003). I nevertheless confirm their results by regressing capital inflows, defined as foreign direct investment and portfolio investment (% of GDP), on a measurement of government quality and a number of control variables from the literature on the determinants of capital inflows (see for example Alfaro, et al., 2003). Because the results are similar, I do not report them here although they are available upon request. 4 Data on capital flows necessary to define sudden stops are available before 1984, but the measures of government quality from the International Country Risk Guide are only available beginning in 1984. However, because a relatively large number of sudden stops associated with the Latin American debt crises occurred in the early 1980’s, I include any sudden stops occurring in 1982-1983 in the count. Given the lack of variation over time in the government quality variables, this is unlikely to be problematic. 6 large numbers of zeros and a limited number of values (the largest number of sudden stops during the sample period is four). Because this variable is time-invariant, I average the observations for each country over time to focus on cross-country variation. I choose to analyze the number of sudden stops over a twenty year period as opposed to a simple binary dependent variable indicating whether country “i” experienced a sudden stop in year “t”. There are a number of reasons for doing this. First, the second approach ignores cross-country variation in the frequency of sudden stops. It is important to know why some countries experienced three or four sudden stops while others experienced only one. Second, given the problems with standard models of crises and the inability of most leading-indicator variables to accurately predict events, the explanatory power of a probit regression is not high. This is especially true of the government quality variable that shows little variation over time and is therefore unlikely to explain individual crisis episodes. For example, variation in government quality cannot explain why Argentina suffered a crisis in 2001 but not in 2000. This paper, therefore, focuses on crosscountry variation in the frequency of sudden stops.5 Although I emphasize the Poisson regression, I also estimate a probit model where the observations vary by country and year. The probit specification avoids the problem with the Poisson mean-regression, namely that a country with a high level of reserves, for example, in one decade and a low level in the next is treated the same as a country with constant reserves throughout the sample period (although this is not a significant problem for government quality which displays little variation over time). Not surprisingly, the government quality variable is 5 The goal of this paper, therefore, is not to add a variable to the literature on predicting financial crises, (DemirgüçKunt and Detragiache, 1999; Eichengreen and Rose, 1997; Frankel and Rose, 1996; Kaminsky and Reinhart, 1999) as government quality does not display the kind of variation necessary for an early warning signal. It should be noted of course that a significant cross-sectional relationship does not necessarily imply the same relationship will hold within countries over time. 7 less able to explain both within and cross-country variation in sudden stops than it is able to explain solely cross-country variation, although the results are still generally significant. 2.1.1 Defining Sudden Stops Defining what constitutes a sudden stop is obviously key to this analysis. A sudden stop is a large and unexpected fall in net capital inflows (i.e. a reduction in the financial account surplus) that is accompanied by a reduction in the current account deficit and a contraction in output. To operationalize this notion, I follow the algorithm in Frankel and Cavallo (2004) to classify observations as sudden stops, updating their data through 2004. A sudden stop is a situation in which at a year “t,” the fall in the financial account surplus (from period “t-1”) of country “i” exceeds twice the standard deviation of the financial account surplus; the current account deficit falls by any amount either in “t” or in “t+1”; and GDP per capita falls by any amount either in “t” or in “t+1.”6 The requirement that GDP per capita fall is necessary because in some cases, a decline in the financial account surplus may be the natural result of a positive terms of trade shock that provides an additional source of funding (Calvo, et al., 2004). More importantly, however, this restriction limits attention to costly sudden stops, which is of far greater interest.7 Using this definition, there are 85 sudden stops between 1982 and 2004 in a sample of 154 countries, which constitutes 2.7% of the yearly observations in the dataset. A list of sudden 6 Requiring that the fall in the financial account surplus is at least two standard deviations below its sample mean did not affect the classification. 7 If I drop the restriction that the reversal in capital inflows must be accompanied by a contraction in output, the non-linear relationship no longer holds. The reason is that without this restriction, episodes in which advanced economies have experienced reversals, but without recessions, will be counted as sudden stops. As shown in Table 1, eliminating the GDP restriction yields a far larger number of identified sudden stops. The mean is now slightly more than one sudden stop per country during the sample period. Using this definition of sudden stops, for example, Denmark has six sudden stops during the sample period but zero using the default definition. In addition, countries with weak institutions that do not receive many capital inflows now have more classified episodes of sudden stops as a result of the low volatility of their inflows. I do not find this definition particularly interesting since analyzing the determinants of reversals in capital flows that cause recessions, as opposed to those that do not, is more relevant for policy-makers in emerging market countries that have suffered from these episodes. 8 stops is provided in the appendix. Not surprisingly, these episodes occur during crisis periods such as the Latin American debt crisis of the early 1980’s, the 1992-1993 European Monetary System crises, the 1997-1998 Asian crises, and the spate of developing country crises in the late 1990’s and onward.8 As shown in Table 1, 19% of all sudden stops occurred in Asia; 9% in Europe (none in Eastern Europe); 33% in Latin America and North America; and 14% in the Middle East; 25% in Africa respectively. 18% of the sudden stops occurred in industrial countries. 2.1.2 Institutional Quality Variable The composite government quality variable, GovQual, is based on several variables from the International Country Risk Guide. These variables were chosen as reasonable proxies for aspects of government quality that might affect capital inflows. They proxy for the degree to which contracts are enforced, corruption, risk of expropriation and government effectiveness. Data are available from 1984-2004. The composite variable is a simple average of these variables. Bureaucracy Quality (range 0-4) measures institutional strength and quality of the bureaucracy as well as autonomy from political pressure. Higher scores also indicate that the bureaucracy has the ability to operate without drastic changes in policy when governments change. Corruption (range 0-6) within the political system measures the extent to which government officials are able to assume positions of power through patronage rather than ability 8 This paper focuses on sudden stops as indicators of crises as opposed to currency crises (Frankel and Rose, 1996; Kaminsky and Reinhart, 1999) or current account reversals (Milesi-Ferretti and Razin, 2000; Edwards, 2004). Sudden stops have become more frequent as capital has become more mobile. Calvo (1999) argues that many of the recent crises were caused by credit shocks in international financial markets and therefore crises should be defined by sharp swings in capital flows as opposed to large exchange rate depreciations or current account reversals that may be more affected by policy choices than sudden stops. Hutchison and Noy (2006) define sudden stops as joint currency crisis and capital flow reversal episodes and find that only half of the currency crises in their sample occurred simultaneously with capital flow reversals. As a robustness check, I estimate the Poisson model for the number of currency crises over the sample period, using the procedure developed in Frankel and Rose (1996) to define currency crises. The results are quite similar. The same non-linear relationship holds and is significant in most of the specifications. 9 and to which they can be influenced by illegal payments. Finally, Law and Order (range 0-6) assesses the strength and impartiality of the legal system and popular observance of the law. Higher scores also indicate well functioning political institutions, implying that this variable should be able to account for sound policy as well. As a robustness check, I look at the effect of the variables individually.9 Whereas institutional quality is a measure of the quality of the framework supporting financial (and other) markets, financial depth is a direct measure of the quality of financial markets themselves. More developed financial markets should channel resources more efficiently, leading to a lower probability of a sudden stop for a given level of capital inflows. Again, however, more developed financial markets should encourage capital inflows indicating a possible non-linear relationship between financial market development and the probability of a sudden stop. I therefore estimate equation (1) using indicators of financial market development, in particular the development of the banking sector and the stock market (Beck, et al., 2000). These include the ratio of M2 to GDP, the stocks-traded turnover ratio defined as the total value of shares traded during the year divided by the average market capitalization for the year, the total value of stocks traded (% of GDP), domestic credit to the private sector (% of GDP) and credit provided by the banking sector (% of GDP). 9 I also used the database built by Kaufmann, et al., (2003) which contains similar variables that measure government effectiveness, the rule of law and corruption. While the coefficients indicate the same non-linear relationship, they are not significant at standard confidence levels. One explanation is that their database only includes data from 1996, 1998, 2000, and 2002 whereas the ICRG data goes back to 1984. Because the paper attempts to explain sudden stops starting in 1982, this more recent data may be inappropriate. Although institutional quality does not vary significantly over time, there is still within-country variation. Examples include countries in Asia and Latin America that have ended dictatorships in the 1990’s as well as moderate improvements in the rule of law starting in the mid-1980’s in East Asia, Latin America, the Middle East, and Africa. Furthermore, because the Kaufmann, et al., (2003) data begin only in the late 1990’s, there is a greater risk of reverse causality than in the ICRG data. 10