tQPS -4'?7 POLICY RESEARCH WORKING PAPER 2487 Finance and Macroeconomic Countries with more developed financial sectors Volatility experience fewer fluctuations in real per capita output, Cevdet Den izer consumption, and investment Murat . lyigun growth. But it matters how Muran L. Iyigun the financial sector develops: the proportion of credit provided to the private sector is important in explaining volatility. The World Bank Europe and Central Asia Region Poverty Reduction and Economic Management Sector Unit H November 2000 PotLfI(CY RlSFARCH WORKING PAPER 2487 Summary findings Countries with more developed financial sectors Denizer, Iyigun, and Owen generate their main results experience fewer fluctuations in real per capita output, using fixed-effects estimation with panel data from 70 consumption, and investment growth. But the manner in countries for the years 1 956-98. which the financial sector develops matters. Their general findings suggest that the risk The relative importance of banks in the financial management and information processing provided by system is important in explaining consumption and banks may be especially important in reducing investment volatility. The proportion of credit provided consumption and investment volatility. The simple to the private sector best explains volatility of availability of credit to the private sector probably helps consumption and output. smooth consumption and GDP. This paper-a product of the Poverty Reduction and Economic Management Sector Unit, Europe and Central Asia Region-is part of a larger effort in the region to understand the links between finance and macroeconomic volatility. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Irina Partola, room H4-346, telephone 202-473-5759, fax 202-522-2751, email address ipartolaC@Lworldbank.org. Policy Research Working Papers are also posted on the Web at www.worldbank.org/research/workingpapers. The authors may he contacted at cdenizer@Cworldbank.org, murat.iyigun@;frb.gov, or aowen@'hamilton.edu. November 2000. (28 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An ohjective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should he cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the vieu of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center Finance and Macroeconomic Volatility Cevdet Denizer The World Bank cdenizer@worldbank.org Murat F. Iyigun University of Colorado murat.iyigun@stripe.colorado.edu Ann L. Owen Hamilton College aowen@hamilton.edu Keywords: Financial Development, Economic Fluctuations, Business Cycles JEL Classification Numbers: D31, E32 We have benefitted significantly from Tim Troha and David Trzepacz's competent research assistance. We would like to thank Alan Ahearne, Thorsten Beck, David Bowman, Bill Easterly, Jon Faust, Hali Edison, Caroline EYeund, Jane Ihrig, Jonathan Wright, and seminar participants at the Federal Reserve Board and the World Bank for very helpful comments, and Asli Demirguc-Kunt and Luigi Zingales for making some of their data available. This paper represents the views of the authors and should not be interpreted as reflecting those of the World Bank or its executive directors. 1. Introduction The Southeast Asian turmoil of the late 1990s brought to the forefront the role finance can play in propagating and dampening macroeconomic fluctuations. In this paper, we investigate this role, providing empirical evidence on the link between the depth and structure of a country's financial sector and the severity of its business cycles. We have two important findings: First, countries with more developed financial sectors experi- ence less fluctuations in output, consumption, and investment growth. And second, the type of finance matters. Private sector finance is particularly important in reducing macroeconomic volatility. Specifically, we show that the relative supply of credit from banks is effective in reducing consumption and investment volatility, while a greater supply of credit to the private sector in general reduces consumption and per capita in- come volatility. Overall, these results suggest that the risk management and information processing provided by banks may be particularly important in reducing consumption and investment volatility, while the simple availability of credit to the private sector helps to smooth consumption and GDP. We find evidence for these conclusions in two different cross-country data sets. Our main data set is a panel of 70 countries spanning the years 1956 to 1998, which we divide into four time periods. We observe the four different measures of financial development employed by King and Levine (1993a, 1993b) at the beginning of each time period and then examine their effects on the subsequent variability of real consumption, investment, and per capita GDP growth using fixed-effects estimation and controlling for other factors that may be associated with the level of macroeconomic volatility, such as inflation and foreign exchange rate variability, government spending, income per capita, and average growth rates. In addition, we find supporting evidence in a cross-sectional data set with which we examine macroeconomic fluctuations over a greater period of time, controlling for various characteristics of countries that have been linked to financial development, such as the strength of accounting standards, contract enforcement, and the level of corruption. We also show that our main results are robust to different estimation techniques. Our analysis is related to three different strands in the literature. The first relevant 1 line of work is the one on finance and development which conjectures that well-developed financial systems should strengthen an economy's ability to absorb shocks, and therefore, help to reduce cyclical fluctuations. Most of the papers in this strand are theoretical, however, and in contrast to our empirical work below which examines consumption and investment volatility in addition to GDP variability, their focus is generally on aggregate output fluctuations. In a recent paper, Aghion, Banerjee, and Piketty (1999) develop a macroeconomic model based on micro-foundations which combines financial market imperfections and unequal access to investment opportunities. They show that economies with less developed financial systems will tend to be more volatile and experience slower growth. In their model, low levels of financial development and the separation of savers from investors result in macroeconomic fluctuations with the economy converging to a cycle around its steady-state growth path. In contrast, when there exists a fully functioning capital market, the economy converges to a stable growth path along which fluctuations are only due to exogenous shocks. Intuitively, their result derives from the fact that when the financial sector is not as well developed, the supply of and the demand for credit is more cyclical. In particular, investors are more likely to get locked out of credit markets when the economy faces a bad shock, only to rush back in when the economy sustains a good shock. Aghion et al. point out that the financial sector tends to be less developed in non-industrialized countries and that this may be one reason why thAse economies experience more volatility. Acemoglu and Zilibotti (1997) also imply an important link between financial development and volatility by highlighting the role that diversification plays in reducing risk. They demonstrate that when there are indivisible investment projects, in the early stages of development, diversification is not possible. As wealth accumulates, however, diversification becomes possible, investment increases, and investment risk and volatility is reduced. In another related paper, Aghion, Bachetta, and Banerjee (2000) show that volatility is most likely to occur for open economies with intermediate levels of financial development. The second relevant line of work has studied the effects of financial market im- perfections and the underlying informational asymmetries on output fluctuations. For example, Bernanke and Gertler (1989, 1990), Greenwald and Stiglitz (1993), and Kiy- 2 otaki and Moore (1997) develop dynamic general economic equilibrium models where asymmetric information in financial markets exacerbates output volatility. While the role financial market imperfections play in generating macroeconomic volatility has not been empirically analyzed directly, the empirical evidence at the micro level is support- ive of the idea that asymmetric information has real effects. Fazzari, et al. (1998) show that fixed investment depends on firms' cash flow, which would not be the case if capital markets were perfect. Microeconomic evidence of the importance of asymmetric infor- mation has also been found by examining the behavior of firms that are more likely to be subject to information asymmetries. For example, Gertler and Gilchrist (1994) confirm this and find that the impact of monetary policy changes is larger for smaller firms compared to the large ones. Kashyap and Stein (1995, 2000) find similar results for banks and show that monetary tightening affects smaller banks more than large banks. These findings are most relevant to what we discuss below to the extent that the degree of asymmetry of information varies from one country to another. It may well be that low or high levels of financial development are related to the ability of economies to generate and process information.' Put differently, financial development indicators may proxy for the effects of financial imperfections arising from information asymmetries or other structural constraints. Thus, a negative relationship between financial development and volatility would generally support the hypothesized role of asymmetric information in propagating business cycles. The third strand of the literature to which our work is more generally related is the large and growing body of work-both empirical and theoretical-on the role of financial de- velopment in long run macroeconomic performance. According to these papers, financial intermediaries reduce the costs of acquiring information and help to lower transactions costs (Gertler, 1988 and Levine, 1997). In doing so, they help ameliorate information asymmetries, improve corporate governance, and lead to better resource allocation. As a result, higher levels of financial market development lead to faster economic growth (Boyd and Prescott, 1986, Fischer, 1993, Goldsmith, 1969, Greenwood and Jovanovic, 1990 and King and Levine, 1993a, 1993b). Furthermore, financial institutions improve 'See, for examnple, Dewatripoint and Maskin (1995) and Diamond (1984). 3 risk management, provide liquidity and channel funds to most productive uses (Ben- civenga and Smith, 1991, and Greenwood and Smith, 1997). The empirical literature is supportive of this view,2 and cross-country findings are complemented by those at the industry level. For example, Rajan and Zingales (1998) and Carlin and Mayer (1999) find that in countries with developed financial markets, industries that rely relatively more on external financing tend to grow faster. In contrast, in countries with poorly developed financial systems, industries that depend on external financing grow relatively more slowly. At the firm level, new research shows that strong legal and market institu- tions are associated with firm growth at rates faster than that which could be realized using internal funds (Demirguc-Kunt and Maksimovic, 1998, 1999). While this liter- ature does not focus on the macroeconomic role of credit markets, it conjectures that well developed financial systems are likely to better absorb shocks and that banks may influence the magnitude of cyclical fluctuations. For example, Levine (1998) notes that banks may affect the magnitude of cyclical fluctuations-a view that is consistent with all existing empirical studies. Finally, Ramey and Ramey (1995) show that countries with higher volatility grow slower and allow us to make a more direct connection to the growth literature. Our findings support the idea that an additional channel through which financial development may affect growth is by reducing volatility. In summary, existing literature implies several routes through which finance can affect macroeconomic cycles. First, more developed financial markets and institutions may more efficiently match savers and investors, allowing the economy to absorb shocks more easily. The financial sector may also facilitate diversification (at both the micro- economic and macroeconomic level) which would reduce risk and volatility. Or third, financial development may be a proxy for the extent of information asymmetries which may themselves cause increased volatility. Initial empirical evidence is suggestive of a role for financial development in dampening macroeconomic fluctuations. lyigun and Owen (1999) present some preliminary evidence that the amount of credit provided to the pri- vate sector is particularly important in smoothing consumption and output variability in economies in which there are relatively more low-income individuals. In addition, 2See, for example, King and Levine (1993a, b), Levine (1997, 1998) and Levine and Zervos (1998). 4 Easterly, Islam, and Stiglitz (2000) use a cross-country framework to show that credit allocated to the private sector can reduce output fluctuations but flexible labor market institutions do not. This paper contributes to the works mentioned above by building on these initial empirical results to more thoroughly investigate the effects of both the extent and nature of financial development on output, consumption, and investment growth variability. Thus, it offers a combined test of the empirical implications of all of the hypotheses discussed above. The remainder of our paper is organized as follows: In section 2, we discuss our estimation strategy and data. In section 3, we present our main results and conduct a sensitivity analysis. And in section 4, we conclude. 2. Estimation Strategy and Data As discussed above, existing studies provide the theoretical motivation for our work. While each contribution relies on different mechanisms, the literature asserts that, in general, financial development should reduce macroeconomic volatility. A test of this prediction at the macro level can be accomplished using the following empirical frame- work: i,t= pi + At + O1FINDEVi,t_1 + 2Xi,t + Vi,t (1) where Vi t is the standard deviation of real per capita consumption, investment, or income growth at time t for country i, pi is a country-specific effect, At is a time specific effect, FINDEVi,t_1 is a measure of financial development in country i in the preceding period, Xi,t includes additional control variables that may help to explain volatility and vi,t is the variability in consumption, output, or investment growth not explained by the regressors.3 We assume that vi,t is uncorrelated with the regressors and is distributed 3We estimate equation (1) using a fixed-effects estimation. Random effects estimation, if appropriate, would provide more efficient estimates. However, one can argue on theoretical grounds that random effects is not appropriate because it is likely that the country specific effects are correlated with the regressors. Results from Hausman tests are mixed but, overall, suggest that a fixed effects strategy is appropriate-the random-effects model is rejected at the 5 percent significance level for 6 of the 12 5 normally with a mean of zero and a variance of 't.4 The control variables in Xt include the average growth rate of real per capita income, consumption or investment at time t, GROWTHt, the average level of real per capita income, consumption or investment at time t - 1, MEANt-1, the mean and standard deviation of inflation over the period, INFMEANt, INFSTDEVt, the mean and standard deviation of government spending as a share of GDP, GOVMEANt, GOVSTDEVt, the degree of openness of the economy as measured by the ratio of ex- ports plus imports to GDP, OPENt, the standard deviation of exchange rate changes, FXVOLt, and an index of the type of political regime, POLITYt. GROWTHt and MEANt-, are included because the standard deviation of growth rates may be corre- lated with the initial level and the growth rate of the dependent variables, with more developed or faster growing economies exhibiting less variability.5 GOVMEANt and GOVSTDEVt, are taken into account to control for the effects of changes in govern- ment spending on macroeconomic fluctuations. Similarly, the political environment, POLITY, may also affect economic stability. We include OPENt and FXVOLt to con- trol for the effects of external shocks on domestic macroeconomic volatilities and interact the two variables on the theory that exchange rate volatility should affect investment, consumption and output volatility differently in economies that are more open. Finally, we also consider measures of inflation and its variability because they will be correlated with output growth variability when the aggregate supply curve is upward sloping.6 To measure financial development, FINDEV, we use four measures proposed by King and Levine (1993a): LLY, which is M2 divided by GDP; PRIVY, the ratio of claims on the nonfinancial private sector to GDP; PRIVATE, the ratio of claims on the specifications in Table 2. (In the cases in which the random-effects model is not rejected, results of random effects estimations are similar to those reported in Table 2.) Of course, use of the fixed-effects estimations restricts the interpretation of our results to analysis of within-country effects. 4In addition to adopting this assumption on the distribution of errors because of its intuitive appeal for cross-country data, we also confirmed it with a Cook-Weisberg test for heteroscedasticity. 5See, for example, Ramey and Ramey (1995). Also, we measure MEAN at time t - 1 because it is the stage of development at the beginning of the period that is the most economically meaningful. However, our results are not sensitive to the time period in which we measure MEAN. 6For a recent survey of the vast literature on the output and inflation tradeoff, see Erceg, Henderson and Levin (forthcoming). 6 nonfinancial private sector to total domestic credit (excluding credit to money banks); and BANK, the ratio of deposit money bank domestic assets to deposit money bank domestic assets plus central bank domestic assets.7 While all of these measures are positively associated with financial development, each one captures a slightly different aspect of it. LLY is a measure of the overall size of the financial system, while BANK measures the relative importance of banks within the financial system. PRIVY and PRIVATE measure the extent to which financial services are provided to the private sector, with PRIVATE being a more direct measure of how credit is allocated to the private vs. the public sector. Thus, LLY and PRIVY are more general measures of overall development, while PRIVATE and BANK attempt to gauge the nature of the development. We use annual data from 70 countries for the period 1956 to 1998. The majority of our data is extracted from the IMF's International Financial Statistics. The exceptions are the average level of real GDP, real consumption and real investment per capita (MEAN), which are calculated from the Penn World Tables 5.6, and the political regime and stability variables (POLITY), which come from the 1996 version of the POLITY III Dataset.8 In order to calculate our volatility measure, the standard deviation of the growth rate of GDP and its components, we need to collapse several years of data into one time period. An ideal characterization of the amplitude of the business cycle in each country and each time period would require a large number of annual observations to capture both the upturns and downturns of the business cycle. However, we face a tradeoff: as we increase the number of years in each period, perhaps increasing the accuracy with which we characterize volatility, we reduce the number of periods we can use in our fixed effects estimation, reducing its efficiency. Nonetheless, we are able to create 4 time periods while still measuring volatility over a relatively long period of time- 7See King and Levine (1993a) for a detailed description of the calculation of these measures. 8The POLITY III dataset includes ten different variables which aim to measure the level of institu- tional democracy and autocracy across countries over time. The results presented here use the variable XRCOMP, the extent to which executives are chosen through competitive elections. Our results are robust to using alternative measures from this dataset: The POLIY variable was never statistically significant in any of the estimations possibly because, for many countries, these variables do not change much over time. Therefore, it is difficult to identify their impact in a fixed-effects estimation. The complete data can be accessed at http://garnetl.acns.fsu.edu/ phensel/intlpoli.html. 7 9 years. Specifically, to obtain our measure of financial development at the beginning of the period, we use average values of BANK, LLY, PRIVATE, and PRIVY over the periods 1956-60, 1966-1969, 1975-1978, and 1984-1987.9 The four time periods for all other variables except MEAN correspond to 1960-1968, 1969-1977, 1978-1987, and 1987-1998. In summary, our main data set contains a panel of 70 countries and 4 time periods. For each period we observe financial development at the beginning of the period and then observe subsequent fluctuations. [See Table 1.] Table 1 presents summary statistics from the main data set and shows that the volatility of real per capita GDP growth, real per capita consumption growth, and real per capita investment growth are all positively associated, with the strongest correla- tion being that between consumption and GDP. In addition, with one exception, all financial development indicators are negatively correlated with macroeconomic volatil- ity. However, despite the fact that each of the 4 indicators attempt to measure financial development, the correlations between these variables are rather low, suggesting that each indicator is in fact capturing a slightly different feature of the financial system. 3. Results 3.1. Initial Estimates Table 2 presents the results of the estimation of equation (1) for the standard deviation of real per capita GDP, consumption, and investment growth, using each of the four financial development indicators. [See Table 2.] In general the results in Table 2 suggest that financial development is associated 9We average the financial development measures over 4 years to smooth through any temporary events that might be affecting the financial system in any given year. 8 with less volatility in GDP, consumption, and investment; however the importance of specific indicators varies across dependent variables. Only BANK, the relative impor- tance of the banking industry, is significant for all three estimations. PRIVATE is significant in the GDP and consumption estimation (though the coefficient retains the negative sign in the investment regression). PRIVY is not significantly related to any of the dependent variables and flips signs across the estimations. At this point we will only note that the relative size of the entire financial sector, LLY, is significant only in the consumption regression, but it has an inconsistent sign. While we defer a more comprehensive discussion of this result until Section 3.2 when we discuss the robustness of all our results, we will note at this time that we do not find this relationship to be robust. Interestingly, of all our other control variables, only two consistently enter the estimation in Table 2 in a statistically significant way. The standard deviation of the inflation rate is positively related to GDP and consumption variability and exchange rate volatility is positively related to all three dependent variables. However, our failure to find a statistically significant relationship between either MEAN or GROWTH and the standard deviation of output, consumption or investment does not contradict earlier work that has shown that economies with less fluctuations grow faster in the long run. Each of our time periods consists of only 9 years. Thus, our results do not allow us to make inferences about long-run relationships. Tables 1 and 2 present relatively low correlations between financial development indicators and varying levels of explanatory power for each different indicator in our 3 main regressions. Taken together, this suggests that different aspects of financial development may affect volatility of GDP and its components differently. To determine which financial development indicator has the strongest relationship with volatility, we conduct a "horse race" and put all four measures of financial development into each regression. The results of this experiment are displayed in Table 3. [See Table 3.] 9 These results show that BANK (the relative importance of banks) and PRIVATE (the percentage of credit that is funneled to the private sector) both retain some explana- tory power in the GDP and consumption regressions. Furthermore, results in the third column of Table 3 show that BANK is negatively related to investment volatility, but the positive coefficient on PRIVY now also becomes significant. The PRIVY result, however, will not be robust in subsequent estimations. One interpretation of the results in Table 3 is that it is not just the overall size of the financial sector but the way in which the financial sector develops that matters in reducing fluctuations. Higher values of PRIVATE suggest that a larger percentage of credit is finding its way to the private sector and into the hands of households, allowing them to better smooth consumption. Similarly, higher values of BANK associated with greater importance of the banking sector may also indicate greater availability of credit for households because banks may be an important source of consumer loans. Because consumption is such a large component of GDP, it is not surprising that this effect is then funneled through to GDP. On the other hand, the results in column 3 of Table 3 show that the relative availability of bank credit is the aspect of the financial system that reduces investment volatility. Perhaps banks may be in the best position to reduce information asymmetries and develop longer term relationships with borrowers, thus reducing the volatility of investment. The effect of financial development on macroeconomic fluctuations is economically meaningful. In our sample, the standard deviation of PRIVATE is .368 and the stan- dard deviation of BANK is .200. Using the results in Table 2, one can calculate that a one standard deviation increase in PRIVATE reduces the standard deviation of real per capita GDP growth by .85 (about 14 percent) and reduces the standard deviation of real consumption growth by a similar magnitude (0.87 or about 14 percent). A one standard deviation increase in BANK reduces GDP volatility by 1.64 (about 28 per- cent), consumption volatility by 1.86 (about 32 percent), and investment volatility by 1.99 (about 14 percent). Although our measures of financial development may indirectly incorporate in- ternational capital flows, our results suggest that the domestic financial system is still 10 important in determining volatility. Specifically, BANK is a measure of the importance of banks residing in that country. One possible interpretation of this result is that if international capital flows exacerbate investment volatility, a stronger banking sector reduces their relative importance. 3.2. Sensitivity Analysis The previous section establishes that higher values of PRIVATE are associated with lower values of consumption and GDP volatility, that higher values of BANK are as- sociated with lower values of output, consumption and investment volatility, and that higher values of LLY are associated with higher values of consumption volatility. In this section we examine how robust these results are to different data selection and estimation techniques. As explained in Section 2, our main data set contains a measure of financial devel- opment at the beginning of a time period and a measure of subsequent macroeconomic variability. However, the years we use to calculate initial financial development imme- diately precede those years which we use to calculate subsequent volatility. In order to determine if our results are sensitive to the time period in which we measure financial development, in Table 4 we repeat the estimations presented in Table 2, but this time we 'measure financial development with a five year lag between the last year we use in the financial development measure and the first year we use in the variability measure (i.e., financial development is measured over the periods 1951-55, 1961-1964, 1971-1973, and 1979-1982). These results confirm our initial findings for consumption and output vari- ability, however, in this case BANK becomes insignificant in the investment regression while PRIVATE becomes significant. [See Table 4.] Although our data set is a relatively large macro data set, our analysis still may be sensitive to outliers. We therefore provide another estimate of the relationship between financial development and fluctuations this time using a robust regression technique that 11 identifies outliers and removes them or places a smaller weight on them in the estimation. Essentially, our procedure excludes observations for which Cook's D is greater than 1 and then iteratively selects weights for the remaining observations based on the size of the absolute residuals, with large-residual observations receiving the lowest weights.10 Results of this procedure are presented in Table 5. [See Table 5.] Table 5 shows that our major conclusions are also robust to the new estimation procedure: PRIVATE reduces fluctuations in GDP and consumption, and BANK re- duces fluctuations in consumption and investment. However, the investment regressions provide conflicting evidence. Now, PRIVATE enters positively and significantly in the investment regressions. Taken together with the results in Table 4, this result suggests that the relationship between investment volatility and financial development may be more complex than that which is captured by our approach.11 As we alluded to earlier, the robust regression technique reduces the size of the LLY coefficient in the consumption equation considerably, rendering it statistically in- significant. One reason why the LLY result may be particularly sensitive to outliers is that, in addition to identifying countries with large financial sectors, high values of M2/GDP might also be signaling a central bank that has not been successful in stabiliz- ing the economy. Thus, we can observe higher than average M2/GDP being associated with larger than average fluctuations. In some of our observations this type of instability may be extreme. Since these few observations are not representative of the entire data set, when they are removed, the initially positive effect we identified becomes insignifi- cant. While it is difficult to test this conjecture directly, it is supported by the fact that '°See Hamilton (1991) for more details on this procedure. "1With robust regression techniques, both the negative association of BANK and the positive asso- ciation of PRIVATE survives in a horse race similar to that displayed in Table 3. However, the effect of BANK on investment volatility remains the most important of the two indicators. The results in Table 5 indicate that a one standard deviation increase in BANK reduces the standard deviation of investment by 1.72, while a one standard deviation increase in PRIVATE increases it by .96. 12 when we square LLY and include it in the regressions, we find a U-shaped relationship between LLY and fluctuations in output, consumption, and investment, indicating that at lower levels of LLY, an increase reduces volatility, but at very high levels, it increases it. We do not find this same quadratic relationship with our other measures of financial development. 12 PRIVY may also be subject to the criticism that it is linked to stabilization policy implemented by the central bank, but the link may not be as direct. While total credit to the private sector is likely to increase during a monetary expansion, the central bank has only indirect control over it through use of its interest rate or money supply policy levers. Even so, our inability to disentangle these two measures of the overall size of the financial system from monetary policy actions could be a reason why the results for these two measures do not consistently associate financial development with smoother business cycles. However, the two measures that we find to be the most robustly related to volatility, BANK and PRIVATE are ratios that should not vary systematically in the short-run with a change in monetary policy. We also considered if our results were sensitive to an alternative means of measuring volatility. In the results presented in Tables 1 through 5, volatility of our macroeconomic variables is calculated using the standard deviations of annual growth rates over a period of years. While this technique characterizes volatility over approximately a decade, it requires us to collapse over 40 years of data into 4 time periods. An alternative measure of volatility that would potentially offer more efficient estimation of the relationship of interest is to use the variability of the error term from a GARCH process to calculate the variance of each of our macroeconomic series on an annual basis.13 Of course, since the GARCH process uses data from all time periods to estimate the variance in any given year, we are no longer able to get a "clean" look at initial levels of financial development 12A full set of results for all estimations not detailed in the text is available from the authors upon request. '3Specifically, for all variables in our estimation for which we previously calculated standard deviations (both dependent and independent variables), we now calculate the variance of the error term from a GARCH(1,1) process for that series. Using the coefficients on the estimated GARCH(1,1) equation, we transform the variance of the error term to obtain the variance of the dependent variable in the GARCH estimation. Results from this procedure are available upon request. 13 and subsequent volatility as we do in our main estimations. Thus, we present these results only as supplementary to our initial findings. Nonetheless, the results from this procedure confirm many of our initial findings-PRIVATE is negatively related to the variability of GDP, consumption, and investment growth and BANK is negatively related to GDP variability. Consistent with the mixed nature of results we have reported for investment variability, BANK is positively associated with the variability of investment calculated in this manner. Finally, an additional possibility that we consider here is that our finding of a negative association between financial development and reduced volatility is being driven by the developed countries in our sample which have both developed financial sectors and lower macroeconomic volatility. Although we do not report the results here, we confirmed that our major conclusions still hold even when we drop high income countries from our sample."4 Our results are also robust to using a logarithmic specification for either of our three dependent variables or for the measure of financial development. Thus, the relationship we identify is not likely to be the result of a an omitted variable common to high income countries that also are likely to have more developed financial sectors. 3.3. Further Discussion So far, our results indicate that the manner in which the financial sector develops could have important implications for the severity of the business cycle. Others have sug- gested that the financial system is in fact influenced by several important underlying characteristics of the economy (e.g., La Porta et al., 1997, 1998, Levine, 1998, 1999, and Demirguc-Kunt and Levine, 1999). These characteristics may be linked to macroeconomic fluctuations and it would be desirable to separate them out from the country-specific fixed effect we estimate above. VVhile panel data is not available to allow us to do this in our main estimation, we are able to implement a cross-country regression for a subset of the countries in our panel. The additional variables we add to the cross-country regressions are from La Porta 14We considered countries that had income in the top quartile to be high-income countries. These countries had real per capita incomes greater than $11,000 in the last period. 14 et al. (1998). In all, we add 5 different variables. ACCOUNT is an index of the strength of accounting standards ranging from a low of 0 to a high of 90. High val- ues of ACCOUNT indicate that companies in that country are required to have more comprehensive financial statements. ENFORCE is an indicator of the enforceability of contracts. It is actually an average of two separate indices-one which assesses the law and order tradition of an economy and the second which assesses the chances that a given government will change the provisions of a contract once it has been signed. ENFORCE ranges from 1 to 10 with higher values indicating stronger enforcement of contracts. CORRUPT is an index of corruption ranging from one to ten with lower val- ues indicating a greater incidence of government officials demanding special payments. STRUCTURE is an index of the extent to which the financial system is based on the stock market rather than banks. Higher values of STRUCTURE imply that the market is relatively more important than banks. Finally, our data contains dummy variables for the legal origin of a country, ORIGIN (i.e., French, English, Scandinavian or German)."5 To implement our cross country estimation, we collapse our data into two time periods and modify equation (1) to estimate Vi,t = pi + At + 0lFINDEVi,t,j + 02Xi,t + f3Zi,t-1 + Vi,t (2) where Zi,t_j includes ACCOUNT, ENFORCE, CORRUPT, and STRUCTURE. We measure Z2,t_j in the initial time period (1976-1980) and then measure subse- quent volatility and Xi,t over the period 1980 to 1998.16 Thus, our cross-country data set gives us some advantages. It allows the inclusion of additional variables and gives us the ability to measure volatility over a much longer period of time. Moreover, the addi- tional variables in our cross-country data set also allow us to assess whether or not our main results are affected by endogeneity. Reverse causation may also be an issue in this 15For a more detailed discussion of the definitions of these vaxiables, see also Demirguc-Kunt and Levine (1999). '6Since we view these additional variables included in Z as country characteristics that are essen- tially fixed over the time period of the data, the exact timing of when we measure them should be inconsequental. 15 relationship-more stable economies may promote greater financial sector development."7 Clearly, the legal origin of a country, which has been linked to financial development, is a predetermined and exogenous variable in the above estimations.'8 Hence, we are able to use it to instrument for financial development in our cross-country data set. Our cross-country data set also has some important disadvantages compared to the panel we used for the main results. Not the least of which is the small size of our data set; data availability restricts our analysis to only 27 countries. In addition, while we believe that the additional variables we add are important components of the country-specific fixed effect, we do not believe they include all important country characteristics. As a result of these disadvantages, we are cautious in our interpretation of the cross-country results and present them only as supplementary to our main results. Table 6 presents the results of this estimation for the variability of per capita consumption growth.19 [See Table 6.] Our results for the per capita consumption growth regressions are generally sup- portive of the conclusions we drew above. Higher levels of BANK, PRIVATE and PRIVY are associated with lower volatility of per capita consumption. Per capita GDP and investment regressions, not reported here, generate negative but insignificant coeffi- cients in three of the four GDP regressions and all of the investment regressions. The coefficients on the additional country characteristics we were able to include in the cross-country regressions also suggest some interesting conclusions. Stronger ac- counting standards are associated with lower variability of consumption in two regres- sions, while exchange rate volatility is robustly and positively related to consumption variability in three out of four specifications. Surprisingly, less corrupt governments are 17While our use of initial financial development and subsequent volatility somewhat addresses this issue, it is interesting to reverse this estimation and examine the effect of initial volatility and subsequent financial development. When we use the two non-overlapping periods of financial development and volatility in the panel data, however, we find no evidence for this reverse causality. '8For more detailed discussions, see La Porta et al (1998) and Beck, Levine and Loayza (2000a, 2000b). 19We were not able to obtain statistically significant results for the financial development variables in the GDP and investment estimations, and we do not report them here. 16 associated with higher variability of consumption in all four regressions. One possible explanation for this counter-intuitive result is that the elements of corruption that cre- ate volatility manifest themselves in enforceability of contracts and in the strength of accounting standards. Thus, the positive relationship we find between corruption and stability is only a partial effect after controlling for the most damaging effects of cor- ruption. This conjecture is supported by the fact that the raw correlations between corruption and volatility indicate that more corrupt economies have greater volatility. A possible interpretation of our finding is that, after controlling for these variables, consis- tent and predictable corruption has a stabilizing component. Finally, STRUCTURE, the reliance of the economy on the stock market relative to banks, enters positively and significantly in two of the four consumption regressions. These results would be con- sistent with a wealth effect that stimulated consumption through appreciation in stock market wealth. However, it is doubtful that this effect would be identifiable in aggregate data for most of the countries in our sample. 4. Conclusion There exist at least three different but related strands in the economic literature which assert or imply that financial development should reduce macroeconomic volatility. Panel data from 70 countries covering the years between 1956 through 1998 reveal evidence in support of this hypothesis. In particular, we have shown above that countries with more developed financial sectors experience less fluctuations in real per capita output, consumption and invest- ment growth. However, we have also demonstrated that the manner in which the financial sector develops matters. The importance of banks in the financial system is the most ro- bust in explaining consumption and investment volatility, whereas the relative amount of credit supplied to the private sector has the most explanatory power for the volatility of consumption and output. These results appear to be robust to different estimation tech- niques and data selection strategies. Although our data and methodology do not allow us to distinguish between the different mechanisms through which financial development 17 affects volatility, our general findings suggest that the risk management and information processing provided by banks may be particularly important in reducing consumption and investment volatility, while the simple availability of credit to the private sector helps to smooth consumption and GDP. 18 5. References Acemoglu, D. and F. Zilibotti. (1997). "Was Prometheus Unbound by Chance? Risk, Diversification and Growth," Journal of Political Economy, 105(4), August, 709-51. Aghion, P., A. Banerjee, and T. Piketty. (1999). "Dualism and Macroeconomic Volatil- ity," Quarterly Journal of Economics, 114, November, 1359-97. Aghion, Philippe, P. Bacchetta, and A. Banerjee. (2000). "Capital Markets and the Instability of Open Economies" in The Asian Crisis: Causes, Contagion, and Conse- quences, P.R. Agenor, M. Miller, D. Vines, and A. Weber, eds. (Cambridge University Press, Cambridge, England.) Bachetta, P. and R. Camninal. (2000). "Do Capital Market Imperfections Exacerbate Output Fluctuations?," European Economic Review, 44 (1), 449-468. Beck, T., R. Levine, and N. Loayza. (2000a). "Financial Intermediation and Growth: Causality and Causes," Journal of Monetary Economics, forthcoming. Beck, T., R. Levine, and N. Loayza. (2000b). "Finance and the Sources of Growth," Journal of Financial Economics, forthcoming. Bencivenga, V. R. and B. D. Smith. (1991). "Financial Intermediation and Endogenous Growth," Review of Economic Studies, 195-209. Bernanke, B. and M. Gertler. (1989). "Agency Costs, Net Worth, and Business Fluctu- ations," American Economic Review, 79 (1), 14-38. Bernanke, B. and M. Gertler. (1990). "Financial Ragility and Economic Performance," Quarterly Journal of Economics, 105 (1), 87-114. Boyd, J. H. and E. C. Prescott. (1986). "Financial Intermediary Coalitions, " Journal of Economic Theory, 38 (2), 211-232. Carlin, W. and C. Mayer. (1999). "Finance, Investment and Growth," unpublished manuscript. Demirguc-Kunt, A. and V. Maksimovic. (1998). "Law, Finance, and Firm Growth," Journal of Finance, December, 53(6), 2107-37. 19 Demirguc-Kunt, A. and V. Maksimovic. (1999). "Institutions, Financial Markets, and Firm Debt Maturity," Journal of Financial Economics. Demirguc-Kunt, A. and R. Levine. (1999). "Bank-Based and Market-Based Financial Systems: Cross-Country Comparisons," unpublished manuscript. Diamond, D. W. (1984). "Financial Intermediation and Delegated Monitoring," Review of Economic Studies, 51(3), 393-414. Easterly, W., R. Islam, and J. E. Stiglitz. (2000). "Shaken and Stirred: Explaining Growth Volatility," World Bank Economic Review, forthcoming. Erceg, C., D. W. Henderson, and A. T. Levin. (forthcoming). "Optimal Monetary Policy with Staggered Wage and Price Contracts," Journal of Monetary Economzcs. Fazzari, S.M., R. G. Hubbard and B. C. Petersean. (1988). "Financing Constraints, and Corporate Investment," Brookings papers on Economic Activity, 1, 141-195. Fischer, S. (1993). "The Role of Macroeconomic Factors in Growth," Journal of Mone- tary Economics, 32(3), 485-511. Gertler, M. (1988). "Financial Structure and Aggregate Economic Activity: An Overview," Journal of Money, Credit, and Banking, 20 (3, Part 2), 25-50. Gertler, M., and S. Gilchrist. (1994). "Monetary Policy, Business Cycles, and the Be- havior of Small Manufacturing Firms," Quarterly Journal of Economics, 109, 309-340. Goldsmith, R. (1969). Financial Structure and Development. (New Haven, CT: Yale University Press.) Greenwald, B., and J. Stiglitz. (1993). "Financial Market Imperfections and Business Cycles," Quarterly Journal of Economics, 108, 77-114. Greenwood, J., and B. Smnith. (1997). "Financial Markets in Development, and the Development of Financial Markets," Journal of Economic Dynamics and Control, 21, pp. 145-186. Greenwood, J. and B. Jovanovic. (1997). "Financial Development, Growth, and the De- velopment of Financial Markets," Journal of Economic Dynamics and Control, 21 (1), 20 145-186. Hamilton, L. C. (1991). "How Robust is Robust Regression?," Stata Technical Bulletin, 2, 21-26. International Monetary Fund. "International Financial Statistics," 1969 - 1992. lyigun M. F. and A. L. Owen. (1999). "Income Inequality and Macroeconomic Fluctu- ations," unpublished manuscript. Kashyap, A. K. and C. J. Stein. (1995). "The Impact of Monetary Policy on Bank Balance Sheets," Carnegie-Rochester Conference Series. Kashyap, A. K. and C. J. Stein. (2000). "What Do a Million Banks Have to Say About the Transmission of Monetary Policy?," American Economic Review, forthcoming. King, R. G. and R. Levine. (1993a). "Finance and Growth: Schumpeter Might Be Right," Quarterly Journal of Economics, August, 108, 715-37. King, R. G. and R. Levine. (1993b). "Finance, entrepreneurship, and growth," Journal of Monetary Economics, 32, 513-42. Kiyotaki, N. and J. Moore. (1997). "Credit Cycles," Journal of Political Economy, 105, 2, 211-48. La Porta, R., F. Lopez-de-Silanes, A. Shleifer and R. W. Vishny. (1997). "Legal Deter- minants of External Finance," Journal of Finance, 52, 1131-1150. La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny. (1998). "Law and Finance," Journal of Political Economy, December, 106(6), 1113-55. Levine, R. (1997). "Financial Development and Economic Growth: Views and Agenda," Journal of Economic Literature, June, 35(2), 688-726. Levine, R. and S. Zervos. (1998). "Stock Markets, Banks, and Growth," American Eco- nomic Review, 88, 3, June, 537-58. Levine, R. (1998). "The Legal Environment, Banks, and Economic Growth," Journal of Money, Credit, and Banking. 21 Levine, R. (1999). "Law, Finance and Economic Growth," Journal of Financial Inter- mediation, January. Rajan, R. G. and L. Zingales. (1998). "Financial Dependence and Growth," American Economic Review, 88, 3, June, 559-86. Ramey, G. and V. Ramey. (1995). "Cross-Country Evidence on the Link Between Volatility and Growth," American Economic Review, 85 (5), 1138-51. Rousseau, P. L., and P. Wachtel. (1998). "Financial Intermediation and Economic Per- formance: Evidence from Five Industrial Countries," Journal of Money, Credit, and Banking. Street, J. O., R. J. Carroll, and D. Ruppert. (1988). "A Note on Computing Robust Regression Estimates Via Iteratively Reweighted Least Squares," The American Statis- tician, 42, 152-54. Summers, R. and A. Heston, 1991, "The Penn World Table: An Expanded Set of Inter- national Comparisons, 1950-1988," Quarterly Journal of Economics, 106, 327-443. 22 Table 1: Descriptive Statistics and the Correlation Matrix Mean Std. Dev. SDGDP SDCONS SDINV BANK LLY PRIVATE PRIVY SDGDP 5.88 4.93 1 .85 .59 -.349 -.004 -.210 -.265 SDCONS 6.20 5.81 ... 1 .65 -.377 .08 -.197 - .249 SDINV 14.0 8.93 ... ... 1 -.313 -.116 -.066 -.293 BANK .727 .200 ... ... ... 1 .048 .375 .410 LLY .147 .077 ... ... ... ... 1 -.004 .253 PRIVATE .714 .368 ... ... ... ... ... 1 .170 PRIVY .282 .219 ... ... ... ... ... ... 1 Table 2: Fixed Effects Estimation Dependent Variable: STANDARD DEVIATIONS OF GDP, CONSUMPTION, AND INVESTMENT GROWTH (Annual Rates) GDP C (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) BANKt-1 -8.18* ... ... ... -9.30* ... ... ... -9.97* (2.89) (3.02) (4.82) LLYt-1 ... 8.40 ... ... ... ll.1* ... ... ... 9.92 ... (5.56) (5.20) (8.95) _ PRIVATEt-1 _ -2.31* ... ... -2.37* ... ... ... -1.32 (.643) (.662) _ (1.84) PRIVYt- ... ... ... -2.06 ... ... ... -.968 ... ... ... 4.96 (2.74) (2.66) (4.77) GROWTHt .219 .174 .193 .170 .330 .234 .280 .252 .083 .117 .084 .098 (.212) (.216) (.216) (.220) (.310) (.319) (.311) (.316) (.123) (.122) (.122) (.116) MEANt-± .151 .014 -.008 .003 .237 .066 - .144 -.052 -.246 - .517 -.559 - .569 (.230) (.255) (.250) (.240) (.404) (.429) (.431) (.410) (.588) (.593) (.571) (.546) INFMEANt -.016 -.030 -.044 -.029 -.029 -.025 -.062 -.023 -.011 -.009 -.043 -.009 (.044) (.045) (.046) (.045) (.048) (.049) (.050) (.049) (.079) (.081) (.081) (.080) INFSTDEVt .046 .056** .067* .053** .074* .066** .099* .064** .049 .043 .076 .043 (.031) (.032) (.032) (.032) (.034) (.034) (.037) (.034) (.076) (.079) (.079) (.078) FXVOLt 10.8* 11.7* 13.0* 11.7* 9.20* 10.0* 12.0* 10.3* 17.9* 18.9* 20.1* 20.0* (4.73) (5.07) (5.12) (5.26) (4.10) (4.47) (4.43) (4.66) (6.73) (7.01) (7.07) (7.05) No. of obs. 198 192 197 198 199 192 198 198 198 191 197 197 R2 .56 .53 .55 .54 .62 .59 .61 .59 .68 .67 .68 .67 Note: Colntry-specific and time-specific fixed effects estimate. Heteroskedasticity-corrected standard errors are in parentheses. *, ** respectively denote significance at the 5 per- cent and 10 percent levels. The variables GOVMEAN, GOVSTDEV, POLITY, OPEN, OPEN * FXVOL are included but not shown. Table 3: Fixed Effects Estimation with all Finance Variables Dependent Variable: STANDARD DEVIATIONS OF GDP, CONSUMPTION, AND INVESTMENT GROWTH (Annual Rates) GDP C I (1) (2) (3) BANKt-1 -6.88* -7.23* -9.81** (2.91) (2.94) (5.44) LLYt-I 5.26 7.38 7.59 (5.69) (5.62) (8.96) PRIVATEt-1 -1.54* -1.49* -.354 (.583) (.539) (1.66) PRIVYt-1 1.25 2.78 9.18** (2.73) (2.55) (5.40) GROWTHt .216 .270 .085 (.209) (.309) (.124) MEANt-1 .153 .254 -.403 (.222) _(.379) (.617) INFMEANt - .031 - .026 .004 (.046) (.049) (.080) INFSTDEVt .057** .068** .038 (.031) (.035) (.075) FXVOLt 12.1* 10.5* 18.8* (5.12) (4.45) (6.90) No. of obs. 190 190 189 J_R2 56__ .61 .68 Note: Country-specific and time-specific fixed effects estimate. Heteroskedasticity-corrected standard errors are in parentheses. *, ** respectively denote significance at the 5 per- cent and 10 percent levels. The variables GOVMEAN, GOVSTDEV, POLITY, OPEN, OPEN * FXVOL are included but not shown. Table 4: Fixed Effects Estimation (5 year lags) Dependent Variable: STANDARD DEVIATIONS OF GDP, CONSUMPTION, AND INVESTMENT GROWTH (Annual Rates) GDP C (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) BANKt-U -6.47** ... - -7.33* . .. ... -5.06 ... ... ... (3.59) (3.43) (4.43) LLYt-1 ... 5.00 ... ... ... 7.22* ... ... ... -3.49 ... ... (3.59) (3.48) (8.08) i PRIVATEt-1 ... . .. -3.27* .. . . .. . .. -2.84** . .. . .. . .. -3.55* .. . (1.47) (1.49) (1.65) PRIVYt-l . .. . .. . .. -.010 . .. .. .. - .704 . .. . .. ....1 (2.19) (2.21) _ (4.97) GROWTHt .187 .169 .186 .175 .229 .234 .252 .251 -.012 .115 .024 .101 (.234) (.213) (.222) (.220) (.350) (.312) (.333) (.318) (.123) (.122) (.120) (.121) MEANt-1 .136 -.020 .168 -.036 .238 .023 .113 -.105 -.574 -.688 -.615 -.501 (.227) (.269) (.247) (.253) (.414) (.436) (.438) (.433) (.560) (.617) (.548) (.581) INFMEANt -.034 -.025 -.021 -.028 -.051 -.016 -.053 -.023 -.015 -.016 -.025 -.010 (.047) (.045) (.048) (.045) (.051) (.049) (.053) (.050) (.081) (.081) (.086) (.081) INFSTDEVt .072** .055** .055 .055** .106* .063** .096* .064** .072 .045 .066 .042 (.036) (.032) (.035) (.032) (.037) (.034) (.040) (.034) (.081) (.079) (.080) (.080) FXVOLt 10.5* 11.0* 12.2* 11.8* 9.23* 8.74** 10.7* 10.5* 17.6* 19.4* 17.6* 19.5* (4.97) (5.14) (5.01) (5.13) (4.38) (4.60) (4.55) (4.54) (7.06) (6.68) (7.13) (7.04) No. of obs. 190 198 188 195 191 198 189 195 190 197 188 194 F`2 .55 .53 .57 .54 .62 .59 .61 .59 .69 .66 .69 .67 Note: Country-specific and time-specific fixed effects estimate. Heteroskedasticity-corrected standard errors are in parentheses. *, ** respectively denote significance at the 5 per- cent and 10 percent levels. The variables GOVMEAN, GOVSTDEV, POLITY, OPEN, OPEN * FXVOL are included but not shown. Table 5: Fixed Effects Robust Regressions Dependent Variable: STANDARD DEVIATIONS OF GDP, CONSUMPTION, AND INVESTMENT GROWTH (Annual Rates) GDP C (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) BANKt-1 .426 . .. . .... .. -5.38* . .. . .. . .. -8.58* . .. . .. .. . (1.41) (1.56) (3.95) LLYt-1 -.844 ... ... ... .676 ... ... ... -.460 ... (4.50) (2.93) (9.83) PRIVATEt-1 ... ... -1.68* ... ... ... -1.78* ... ... ... 62.60* .. (.497) (.512) (.924) PRIVYt-1 ... ... ... -2.32 ... ... ... -.577 ... ... ... 4.24 __ (1.70) _ (1.68) (4.67) GROWTHt .135* .016 .025 .122** -.047 -.103** .034 -.062 -.063 -.176* -.262 .049 (.063) (.074) (.070) (.066) (.082) (.056) (.083) (.075) (.082) (.072) (.350) (.082) MEANt-1 .060 .059 .027 .107 .322 .005 .125 .196 -.098 -.015 .265 .097 (.120) (.139) (.130) (.123) (.227) (.150) (.223) (.203) (.495) (.430) (.349) (.488) INFMEANt .024 .028 -.030 .024 -.043** -.016 -.079* -.005 -.045* .105* -.057 -.046 (.021) (.025) (.024) (.022) (.024) (.016) (.024) (.022) (.605) (.053) (.043) (.062) INFSTDEVt .051* .043** .078* .050* .074* .078* .088* .058* .179* -.016 .089* .134* (.021) (.024) (.024) (.022) (.023) (.016) (.024) (.021) (.059) (.051) (.043) (.060) FXVOLt .437 2.84 6.53* 1.90 6.37* 2.65* 13.4* 2.20 7.13 9.31* 18.2* 10.9* (1.71) (1.97) (1.91) (1.80) (1.93) (1.29) (1.97) (1.78) (5.00) (4.23) (3.60) (4.99) No. of obs. 198 192 197 198 199 192 198 197 198 190 197 197 Note: Country-specific and time-specific fixed effects estimate. *, ** respectively de- note significance at the 5 percent and 10 percent levels. The variables GOVMEAN, GOVSTDEV, POLITY, OPEN, OPEN * FXVOL are included but not shown. Table 6: Two-Stage Least Squares Estimation Dependent Variable: STANDARD DEVIATION OF CONSUMPTION GROWTH (Annual Rates) (1) (2) (3) (4) BANKt-1 -6.23** ... ... ... (3.45) LLYt-1 ... 1.78 ... ... (5.16) _ PRIVATEt-1 _3.41** ... (1.56) PRIVYt-1 ... ... ... -2.25* (1.24) ACCOUNTt-1 -.049 -.079 -.076* -.097* (.029) (.060) (.024) (.022) ENFORCEt-1 -.334 -.474 .073 - .174 (.582) (.356) (.368) (.303) CORRUPTt-1 .395* .500* .434* .386* (.195) (.230) (.123) (.140) STRUCTUREt-1 .442 - .105 .530* .738* (.313) (.430) (.241) (.359) FXVOLt 2.80 7.74* 7.71* 9.27* (5.46) (3.41) (3.35) (2.81) No. of obs. 27 25 26 26 R________________ .91 .96 .97 .97 Note: Country-specific and time-specific fixed effects estimate. Heteroskedasticity-corrected standard errors are in parentheses. *, ** respectively denote significance at the 5 percent and 10 percent levels. The variables MEAN, GROWTH, INFMEAN, INFSTDEV, GOVMEAN, GOVSTDEV, POLITY, OPEN, and OPEN * FXVOL included but not shown. Policy Research Working Paper Series Contact Title Author Date for paper WPS2468 Pricing, Subsidies, and the Poor: Ian Walker November 2000 S. Delgado Demand for Improved Water Services Fidel Ordohez 37840 in Central America Pedro Serrano Jonathan Halpern WPS2469 Risk Shifting and Long-Term Mansoor Dailami November 2000 W. Nedrow Contracts: Evidence from the Robert Hauswald 31585 Ras Gas Project WPS2470 Are Larger Countries Really More Stephen Knack November 2000 P. Sintim-Aboagye Corrupt? Omar Azfar 38526 WPS2471 Validating Operational Food Luc J. Christiaensen November 2000 L. Christiaensen Insecurity Indicators against a Richard N. Boisvert 81463 Dynamic Benchmark: Evidence John Hoddinott from Mali WPS2472 Uzbekistan and Kazakkhstan: Asad Alam November 2000 L. Henson A Tale of Two Transition Paths Arup Banerji 84026 WPS2473 Banking Risks around the World: Luc Laeven November 2000 R. Vo The Implicit Safety Net Subsidy 33722 Approach WPS2474 Exports and Information Spillovers Alessandro Nicita November 2000 L. Tabada Marcelo Olarreaga 36896 WPS2475 Industrial Growth and the Quality of David A. Grigorian November 2000 A. Nelson Institutions: What Do (Transition) Albert Martinez 37117 Economies Have to Gain from the Rule of Law? WPS2476 Measuring Banking Efficiency in the Cevdet Denizer November 2000 M. Ding Pre- and Post-Liberalization Mustafa Ding 36233 Environment: Evidence from the Murat Tarimcilar Turkish Banking System WPS2477 Picking the Poor: Indicators for Norbert R. Schady November 2000 T. Gomez Geographic Targeting in Peru 32127 WPS2478 Institutions, Politics, and Contracts: Lorena Alcazar November 2000 P. Sintim-Aboagye The Attempt to Privatize the Water Lixin Colin Xu 37644 and Sanitation Utility of Lima, Peru Ana Maria Zuluaga WPS2479 Estimating the Elfects of Corruption: Aminur Rahman November 2000 S. Powell Implications for Bangladesh Gregory Kisunko 33526 Kapil Kapoor Policy Research Working Paper Series Contact Title Author Date for paper WPS2480 Productivity Growth and Resource Mubarik Ali November 2000 D. Byerlee Degradation in Pakistan's Punjab: Derek Byerlee 87287 A Decomposition Analysis WPS2481 Foreign Direct Investment in Africa: Jacques Morisset November 2000 N. Busjeet Policies Also Matter 33997 WPS2482 Can Institutions Resolve Ethnic William Easterly November 2000 K. Labrie Conflict? 31001 WPS2483 The Credit Crunch in East Asia: Pierre-Richard Agenor November 2000 M. Gosiengfiao What Can Bank Excess Liquid Joshua Aizenman 33363 Assets Tell Us? Alexander Hoffmaister WPS2484 Banking Crises in Transition Helena Tang November 2000 A. Carcani Economies: Fiscal Costs and Edda Zoli 30241 Related Issues Irina Klytchnikova WPS2485 Are Corruption and Taxation Really Raymond Fisman November 2000 R. Bonfield Harmful to Growth? Firm-Level Jakob Svensson 31248 Evidence WPS2486 Who Must Pay Bribes and How Much? Jakob Svensson November 2000 R. Bonfield Evidence from a Cross-Section of 31248 Firms