Open economies review 15: 5–22, 2004 c 2004 Kluwer Academic Publishers. Printed in The Netherlands.
Do IMF and IBRD Cause Moral Hazard and Political Business Cycles? Evidence from Panel Data AXEL DREHER
[email protected] ROLAND VAUBEL
[email protected] Fakultat ¨ fur ¨ Volkswirtschaftslehre, Universitat ¨ Mannheim, D-68131 Mannheim, Germany
Key words:
IMF programs, IBRD programs, political business cycles, moral hazard
JEL Classification Numbers:
D72, F33, F34
Abstract Using panel data for 94 countries in 1975–97, we estimate OLS, 2SLS and GMM regressions to explain IMF and IBRD lending as well as monetary and fiscal policies in the recipient countries. With respect to moral hazard, we find that a country’s government budget deficit and its rate of monetary expansion are higher the larger its borrowing potential in the Fund. New net lending of the Bank (relative to GDP) raises monetary expansion but lowers budget deficits of the recipient countries while new net credit from the Fund is associated with less expansionary policies. As for political business cycles, our evidence indicates that new net credits from the IMF are significantly larger prior to elections and that borrowing from the IBRD is significantly smaller after elections.
The International Monetary Fund (IMF) and the International Bank for Reconstruction and Development (IBRD) have come under increasing scrutiny and attack. The IMF has been shown to be an almost continuous provider of aid to a few dozens of developing countries and emerging market economies. Its policy conditions have frequently been criticized as inappropriate and ineffective.1 Its forecasts are comparatively poor (see Brunner and Meltzer, 1990, Table 4.5) and biased in favor of optimism.2 The growth of IMF and IBRD staff does not seem to be related to the “need for balance of payments credits” as defined by the Fund (see Vaubel, 1996). The IBRD has been shown to provide aid to many emerging market economies enjoying easy access to private capital markets. These countries received 70 percent of the IBRD’s resources between 1990 and 1999. Indeed, their share increased from 60 percent in 1993 to 99 percent in 1999. With respect to outcome, institutional development impact and sustainability, 30 to 40 percent of these projects are judged to be failures. Still worse is the Bank’s performance in the poorest countries: its projects had a failure rate between 65 and 70 percent (see IFIAC, 1999).
6
DREHER AND VAUBEL
In this paper, we test for the validity of two additional criticisms of IMF and IBRD policy. Both are derived from public-choice theory, and both imply that the availability of subsidized credit has undesirable incentive effects on the governments which are eligible for these loans. The first of these criticisms is the “moral-hazard hypothesis” which, with respect to the IMF, was originally proposed in Vaubel (1983). IMF and IBRD lending may be interpreted as a (subsidized) income insurance against adverse shocks. The insurance cover induces the potential recipients to excessively lower their precautions against such damages (or even to intentionally generate a crisis). It is easy to show that balance of payments crises “can be produced at will, virtually overnight” by an inappropriate monetary or exchange rate policy (see Niehans, 1985, p. 67). There is also a considerable body of evidence that the balance of payments problems of IMF borrowers have been largely of their own making3 and that macroeconomic performance during inter-program years has been deteriorating as the number of past programs increased.4 However, the true test of moral hazard is whether the policies causing these crises are at least partially due to the influence of these international organizations. Accordingly, we explain fiscal and monetary policy by the amount of subsidized credit available or received (Section 2). What we are looking at may be called “direct moral hazard’’ because we are analyzing the behavior of the direct recipients of insurance payments—the governments of the member states. This ought to be distinguished from indirect moral hazard effects on the lending behavior of their creditors, i.e. the “bail-out’’ of foreign banks, etc.5 The second criticism to be tested is that IMF and IBRD lending facilitates political business cycles in the recipient countries. This hypothesis, too, has been suggested for the IMF in an earlier paper: “The ruling politicians try to influence the domestic business cycle in their favor by generating a boom at the time of election and low popularity and by reversing the impulse thereafter. IMF lending facilitates the expansion. IMF conditionality facilitates the contraction. In this way, the IMF tends to contribute to the generation of political business cycles” (Vaubel, 1991, p. 213). Thus, there are two parts to this hypothesis: one relating to the pre-election boom and one to the post-election recession. Both criticisms could also apply to the IBRD. Monetary and fiscal policy can be used to generate a pre-election boom even if the exchange rate is immutably fixed. As the portfolio balance approach to exchange rates shows, the rate of monetary expansion compatible with a given exchange rate parity is higher, the larger the central bank’s sales of foreign exchange.6 In other words, foreign exchange interventions financed with credits from the international organizations permit a higher rate of monetary expansion
EVIDENCE FROM PANEL DATA
7
at a given exchange rate. Fiscal expansion, in an open economy, tends to induce a real appreciation because the increase in the budget deficit raises the real interest rate, attracts foreign capital and thereby shifts demand from foreign to domestic goods. Thus, an expansionary monetary policy is possible in combination with sales of foreign exchange or with an expansionary fiscal policy even if the exchange rate is fixed. Official borrowing from the IMF and the IBRD can be used to finance the central bank’s sales of foreign exchange and the government’s budget deficit. However, if IBRD lending to private borrowers replaces government expenditure, the budget deficit may fall. If our hypothesis is correct, we would expect that IMF and IBRD lending is larger before elections than otherwise. This presupposes that the government of the borrowing country may be more attracted by the subsidized credit than put off by the policy conditions attached to it. After the election, borrowing may decline because the available credit lines have been drawn before the election. But the recipient governments may wish to use IMF or IBRD conditionality as a scapegoat for the unpopular corrective measures that are now required.7 This part of the hypothesis has recently been tested by Vreeland (1999) and Przeworski and Vreeland (2000) for the IMF. They found that the conclusion of an IMF program is significantly more likely in a postelection year.8 As such programs are also likely to be associated with additional lending, we check whether current net credits are larger or smaller than usual after the election. Both parts of the political business cycle hypothesis will be tested in Section 2. Section 3 draws conclusions for reform. 1.
The moral hazard of IMF and IBRD lending
Our regressions are pooled time-series cross-section analyses (panel data). Our annual data cover the years 1975–97 and extend to 94 countries that have obtained IMF and IBRD credit during this period.9 Since some of the data are not available for all countries or years, the panel data are unbalanced and our number of observations depends on the choice of explanatory variables. We found significant fixed country effects in all specifications. We also tested for fixed time effects and included them where appropriate. However, the coefficients of the country and time effects are not reported in the tables. All variables, their precise definitions and data sources are listed in the appendix. In Table 1, the government budget deficit relative to GDP10 is regressed on three variables checking for moral hazard: – the amount of IMF credit outstanding at the beginning of the year relative to the country’s quota (“exhaustion of quota”), – new (net) non-concessional IMF credit relative to GDP (t − 1) and – new (net) IBRD loans relative to GDP (t − 1).
8
DREHER AND VAUBEL Table 1. Government budget deficit in percent of GDPa , 1975–97. Explanatory variables
(1)
(2)
(3)
−0.004 (−2.43∗∗ )
−0.006 (−2.73∗ )
−0.004 (−2.20∗∗ )
−0.18 (−0.64)
4.54 (3.62∗ )
−0.17 (−2.46∗∗ )
New net IBRD credit in percent of GDP (t − 1)
0.33 (1.44)
1.92 (2.48∗∗ )
−0.30 (−3.42∗ )
Part of year falling within 18 months prior to an election
0.48 (1.83o )
0.55 (1.42)
0.37 (3.24∗ )
Real GDP growth (t − 1)
0.02 (0.84)
0.07 (2.12∗∗ )
0.02 (1.77o )
Rate of inflation (t − 1)
−0.002 (−0.79)
−0.001 (−2.70∗ )
−0.0002 (−3.45∗ )
Lagged endogenous variable
0.40 (5.51∗ )
0.42 (9.58∗ )
0.33 (10.05∗ )
Exhaustion of IMF quota (t − 1) New net (non-concessional) IMF credit in percent of GDP (t − 1)
Number of countries
77
72
74
Number of observations
1013
856
956
Method of estimation
OLS
2SLS
GMM
Time dummies
Yes
No
Yes
R2
0.53
0.24
(overall)
Sargan Test ( p-level)
0.99
Arellano-Bond-Test ( p-level)
0.17
Notes: a If there is a budget deficit, the dependent variable has a positive value. In column 2, IMF and IBRD loans are instrumented with all variables included in Tables 3 and 4. The regressions in levels include a different intercept for each country. The coefficients of the country and time dummies are not reported. (robust) t-statistics in parentheses: ∗ : Significant at the 1 percent level; ∗∗ : Significant at the 5 percent level; o : Significant at the 10 percent level.
The first variable measures the quantity dimension of moral hazard generated by the IMF: as the country’s quota is increasingly exhausted and, by implication, the quantity of additional credit available from the IMF diminishes, the incentive to run excessive budget deficits declines. Conversely, moral hazard increases if the country repays its credit to the IMF or if IMF quotas are raised. The term “moral hazard’’ is sometimes also used in a wider sense describing an incentive to abuse the claim to an indemnity once the accident has occurred or an incentive to abuse a loan which ought to be, but may not be, repaid. To allow for the possibility of such abuse, the budget deficit is also regressed on the amount of new non-concessional credit (net of repayment) which the country has received from the IMF during the previous year relative to its GDP (“new net IMF credit’’).11 Officially, IMF conditionality is supposed to prevent the recipient country from embarking on overexpansionary policies. But the Fund
EVIDENCE FROM PANEL DATA
9
may not be successful because the conditions are too weak or because they are violated.12 With respect to the World Bank it is not possible to measure the amount a country expects to be eligible for. Thus, we are confined to testing moral hazard in the broader sense.13 To control for the influence of elections, an index is used which measures the share of the year which is within eighteen months prior to a national (executive or legislative) election.14 We also include the rate of real GDP growth, the inflation rate (both lagged one year) and the lagged endogenous variable. Table 1, column 1, contains the OLS estimate. The reported standard errors are heteroscedasticity-consistent. As can be seen, budget deficits fall as the country’s quota with the Fund is exhausted. This effect is significant at the 5-percent level. At the 10-percent level, budget deficits are higher prior to elections.15 Credits from the Fund or the Bank, inflation and real GDP growth do not have a significant influence. Even though all explanatory variables are lagged one year, some of them raise obvious endogeneity problems. For example, we shall argue in Section 2 that monetary and fiscal policies may affect Fund and Bank lending. The institutions’ loans are thus endogenous in our moral hazard regressions and we cannot be sure that lagging them by one period is an adequate solution. To account for the endogeneity problem we instrument the endogenous right-hand variables. We employ two different methods. In column 2, the regression is estimated with Two-Stage Least Squares (2SLS). IMF and IBRD loans are instrumented with all variables determining them in Section 2. (A detailed analysis of these determinants is provided there.) The results of column 2 are more in line with our expectations. Once more, the budget deficit falls significantly as the country’s quota with the IMF is increasingly exhausted. New net credit (relative to GDP) from the IMF and from the IBRD now has a significantly positive effect on the budget deficit.16 However, if we do not instrument the lagged endogenous variable, this variable is correlated with the error term and the OLS and 2SLS estimates of columns 1 and 2 are biased and inconsistent. For this reason, we proceed to the Generalized Methods of Moments (GMM) estimator suggested by Arellano and Bond (1991). This estimator removes the fixed country effects by firstdifferencing the equation. Lagged levels of the dependent variable and differences of the exogenous right hand side variables are then used as instruments. Since there are more instruments than right-hand side variables, the equations are over-identified and the instruments must be weighted. The Arellano-Bond one-step estimator uses the identity matrix as a weighting matrix. The two-step estimator weighs the instruments asymptotically efficiently using the covariance of the one-step estimates. We prefer the two-step estimator because it allows for heteroscedasticity of the errors. We employ a Sargan test to ensure that the instruments are not correlated with the error term, and we use
10
DREHER AND VAUBEL
the Arellano-Bond test for second-order autocorrelation of the error term because the estimator would not be consistent in the presence of second-order correlation. Column 3 reports the results. Both the Sargan test and the Arellano-Bond test clearly accept this specification. The significantly negative effect of quota exhaustion remains. Loans from the organizations now significantly reduce deficits. This may be due to policy conditionality or simply the pre-dominance of post-election lending.17 The positive pre-election effect and the negative inflation effect are now significant at the 1 percent level. Table 2 repeats the analysis for monetary policy as measured by the rate of monetary expansion (money and quasi money). When estimated with OLS (column 1), quota exhaustion, IMF credit and World Bank loans are all insignificant at conventional levels. Only real growth has a significant (negative) effect. Once we account for the endogeneity of IMF and World Bank loans, however, the results are closer to our expectations. The exhaustion of IMF quotas significantly reduces monetary growth rates, and IBRD credits have a significantly positive effect. This is true regardless of whether the regression is estimated with 2SLS (column 2) or GMM (column 3).18 However, in column 3 monetary Table 2. Monetary expansion, money and quasi-money, 1975–97. Explanatory variables
(1)
(2)
(3)
Exhaustion of IMF quota (t − 1)
−0.03 (−0.25)
−0.62 (−2.24∗∗ )
−0.69 (−6.13∗ )
New net (non-concessional) IMF credit in percent of GDP (t − 1)
−21.75 (−1.77o )
238.41 (1.32)
−30.86 (−6.09∗ )
New net IBRD credit in percent of GDP (t − 1)
16.09 (0.58)
348.92 (3.11∗ )
107.12 (11.42∗ )
Part of year falling within 18 months prior to an election
37.83 (1.62)
28.74 (0.76)
86.82 (14.76∗ )
Real GDP growth (t − 1)
−7.40 (−2.61∗ )
0.20 (0.06)
−10.55 (−20.96∗ )
Rate of inflation (t − 1)
−0.10 (−1.08)
0.93 (8.96∗ )
−0.10 (−16.53∗ )
Lagged endogenous variable
0.39 (2.12∗∗ )
−0.62 (−6.12∗ )
0.33 (37.60∗ )
Number of countries
94
74
94
Number of observations
1443
877
1400
Method of estimation
OLS
2SLS
GMM
No
Yes
Yes
0.15
0.18
Time dummies R 2 (overall) Sargan Test ( p-level)
0.99
Arellano-Bond-Test ( p-level)
0.31
Notes: See Table 1.
EVIDENCE FROM PANEL DATA
11
expansion is still negatively affected by IMF (net) credit in the previous year. Monetary expansion is significantly higher prior to elections. As was the case for the budget deficit, inflation has a significant negative effect on monetary growth. Monetary expansion also reacts negatively to real GDP growth. To summarize, we find evidence that the availability of IMF credit generates significant moral hazard (in the narrower sense) with respect to both fiscal and monetary policy. To some extent, such moral hazard may be optimal because insurance can be less costly than precautions against damage.19 In other words, if multilateral lending is a more efficient way of dealing with country risk, it may be efficient that governments take more risk. However, moral hazard is unlikely to be optimal if the insurance cover is subsidized as is the case with IMF lending. As for moral hazard in the broader sense, we have found that IMF credit leads to less expansionary monetary and fiscal policies while loans from the IBRD are associated with higher monetary growth but smaller budget deficits. 2.
Do IMF and IBRD facilitate political business cycles in the recipient countries?
In testing for political business cycles, we use the same data base as in Section 2. Again, we report results from OLS, 2SLS and GMM. Once more, the regressions in levels include fixed country effects and, where appropriate, a dummy for each year. All regressions include the lagged endogenous variable. Once more, all quantitative explanatory variables are lagged one year. The first dependent variable is new net non-concessional IMF credit relative to GDP. Ideally, we would like to distinguish between changes in demand for, and supply of, IMF resources. Unfortunately, the distinction between demand and supply effects cannot be drawn. Almost all regressors may be interpreted at the same time as determinants of the governments’ demand for credit and as criteria by which the Fund judges the creditworthiness of its applicants. Thus, a meaningful simultaneous or two-stage estimation is not feasible.20 However, for our purpose of testing for election effects, a reduced-form estimate is entirely sufficient. To account for electoral cycles, two time indices are employed. They measure that part of the year which falls within eighteen months before or after national elections, respectively. Basically, we use two groups of additional variables. All of them have been proposed in previous studies.21 The first group of variables can be rather closely controlled by the economic policy makers of the borrowing countries. Thus, they may also be indicators of moral hazard. In addition to the overall budget deficit and the rate of monetary expansion introduced in Section 2, we include government consumption (in percent of GDP). The expected effect of these variables on IMF credits is not obvious. On the one hand, they might be taken to indicate a country’s need for credit, so demand and supply could be higher. On
12
DREHER AND VAUBEL
the other hand, the Fund might regard over-expansionary policies as imprudent and be less willing to provide resources. Moreover, the demand for loans might fall as monetary expansion raises the revenue from seigniorage.22 The second group contains variables that are not current policy instruments but clearly affected by them: – – – – – – –
the rate of real GDP growth, the rate of inflation, the share of foreign short-term debt in total foreign debt, a country’s total debt service relative to GDP, international net reserves relative to money supply, the current account balance as a percent of GDP and GDP per capita.
A decline of real GDP growth probably increases the demand for, and supply of, IMF credit because it is used as an indicator of need. Accelerating inflation, a growing share of short-term foreign debt and increasing total debt service may raise the demand for, and supply of, IMF credit because they might be interpreted as signs of financial crisis.23 However, they may reduce IMF lending if they violate the Fund’s performance criteria. Falling international reserves or increasing current account deficits are likely to raise both supply and demand because they would be considered to indicate need.24 If GDP per capita is low, the need for credits may be high but such countries may have little influence in the Fund’s Executive Board and are less dangerous sources of contagion (see Dreher 2003a). Since some researchers have claimed that the Fund uses its credit to support undemocratic regimes (see Assetto, 1988; Edwards and Santaella, 1993; Bandow, 1994), we also include an indicator of democracy. Column 1 of Table 3 presents results from OLS. Except for the lagged endogenous variable, only two coefficients are significant: real GDP growth and international reserves reduce the amount of (net) IMF loans disbursed. This is in line with our expectations. The coefficients of the election indices are both insignificant.25 Since the overall budget deficit and the rate of monetary expansion might be endogenous to new credits, we instrument these variables with the variables employed in Section 2. Column 2 reports the results (two-stage least squares). Real GDP growth, international reserves and the lagged endogenous variable keep their significance. Moreover, monetary expansion and inflation now have significant effects. However, as explained in Section 2, the OLS and the 2SLS estimators are inconsistent and biased in the presence of the lagged endogenous variable. In column 3, therefore, we report results estimated with GMM. Both the Sargan test and the Arellano-Bond test accept this specification. Credits are now significantly higher prior to elections.26 This effect is significant at the 1 percent level. At the 10 percent level, there is also a positive post-election
13
EVIDENCE FROM PANEL DATA
Table 3. New net non-concessional credit from the IMF in percent of GDP, 1975–97. Explanatory variables
(1)
(2)
(3)
Part of year falling within 18 months prior to an election
0.04 (0.56)
−0.001 (−0.02)
0.05 (3.09∗ )
Part of year falling within 18 months after an election
0.06 (0.97)
0.06 (1.12)
0.03 (1.85o )
Index of democratic regime
−0.002 (−0.28)
0.001 (0.09)
−0.02 (−2.42∗∗ )
Monetary expansion (t − 1)
−0.00001 (−0.06)
0.001 (1.70o )
0.0001 (2.41∗∗ )
Budget deficit in percent of GDP (t − 1)
0.01 (0.84)
0.01 (0.60)
0.004 (0.61)
Government consumption in percent of GDP (t − 1)
−0.01 (−0.67)
−0.01 (−1.27)
−0.02 (−2.95∗ )
Real GDP growth (t − 1)
−0.01 (−2.38∗∗ )
−0.01 (−2.37∗∗ )
−0.01 (−3.72∗ )
Rate of inflation (t − 1)
0.00004 (0.36)
−0.001 (−1.69o )
−0.0001 (−1.75o )
Foreign short-term debt/ foreign debt (t − 1)
−0.003 (−0.86)
0.0004 (0.15)
-0.005 (−4.34∗ )
0.003 (0.45)
0.01 (1.00)
0.001 (0.37)
−0.46 (−3.19∗ )
−0.32 (−2.04∗∗ )
−0.51 (−5.89∗ )
Current account balance in percent of GDP (t − 1)
0.004 (1.02)
0.001 (0.27)
0.004 (1.99∗∗ )
GDP per capita (t − 1)
0.0001 (1.05)
0.0001 (0.71)
0.0001 (1.17)
Lagged endogenous variable
0.21 (1.86o )
0.22 (5.49∗ )
0.15 (11.09∗ )
Total debt service in percent of GDP (t − 1) International reserves /money supply (t − 1)
Number of countries
74
74
71
Number of observations
878
878
784
Method of estimation
OLS
2SLS
GMM
Time dummies
Yes
No
Yes
R 2 (overall)
0.09
0.04
Sargan Test ( p-level)
0.99
Arellano-Bond-Test ( p-level)
0.25
Notes: In column 2, monetary growth and overall budget deficits are instrumented with all variables included in Tables 1 and 2. The regressions in levels include a different intercept for each country. The coefficients of the country and time dummies are not reported. (robust) t-statistics in parentheses: ∗ : Significant at the 1 percent level; ∗∗ : Significant at the 5 percent level; o : Significant at the 10 percent level.
14
DREHER AND VAUBEL
effect. This is in line with the finding of Vreeland (1999) and Przeworski and Vreeland (2000) that the conclusion of IMF arrangements is significantly more likely after elections. The amount of credit is negatively affected by government consumption, GDP growth, short-term debt, international reserves and current account deficits but positively by monetary expansion. The results confirm that, even when allowing for GDP per capita, more democratic countries tend to receive less credit. Table 4 replicates the analysis for the World Bank with new net IBRD loans relative to GDP as the dependent variable. The justification for including the explanatory variables is the same as in the case of IMF lending. Like IMF loans, credits from the Bank can be used by member governments to stimulate the economy in order to be re-elected. This is true for project lending and Structural Adjustment Loans. In both cases, the recipient country decides when to draw. Since Structural Adjustment Loans are paid out directly to the government (if certain pre-conditions are met), they can be used for election purposes most easily—even without concluding a new arrangement. Once more, columns 1 and 2 contain the OLS and 2SLS regressions, respectively. As can be seen, all coefficients except for the lagged dependent variable are insignificant in the OLS regression. When, in column 2, we account for the endogeneity of budget deficits and monetary growth and drop the time dummies which are insignificant, five variables (and the lagged endogenous variable) significantly influence new IBRD loans. Lending is larger, – – – –
the higher the budget deficits, the lower government consumption, the higher short-term foreign debt and the higher the debt service is.
At the 10 percent level, democracies tend to receive less credit. There is still no significant electoral effect. Finally, column 3 presents the GMM estimates. The economic variables which were found to be significant in column 2 keep their influence. In addition, higher monetary growth now leads to significantly larger IBRD credits27 while inflation and real GDP growth significantly reduce them. At the 10 percent level, IBRD lending is negatively associated with current account deficits. Democracy no longer has a significant effect. While there is no significant pre-election effect, IBRD loans are now significantly lower in post-election periods. Since the recipient governments decide when to draw, their incentive to borrow declines after the election. To summarize, only IMF lending is significantly higher 18 months prior to elections. Since interest on these loans is lower than what the borrowing governments would have to pay in the world capital market (if they have access at all), IMF credit reduces the opportunity cost of over-expansionary macroeconomic policies and government transfers to marginal voters prior to elections. The post-electoral effect is significantly negative for the World Bank but
15
EVIDENCE FROM PANEL DATA Table 4. New net credit from the IBRD in percent of GDP, 1975–97. Explanatory variables
(1)
(2)
(3)
Part of year falling within 18 months prior to an election
0.05 (1.17)
0.02 (0.48)
0.01 (0.35)
Part of year falling within 18 months after an election
0.01 (0.19)
−0.03 (−0.57)
−0.04 (−2.42∗∗ )
Index of democratic regime
−0.001 (−0.07)
−0.02 (−1.77o )
0.01 (0.55)
Monetary expansion (t − 1)
−0.0002 (−0.73)
−0.001 (−1.47)
0.0004 (2.00∗∗ )
Budget deficit in percent of GDP (t − 1)
−0.01 (−0.83)
0.04 (2.81∗ )
0.01 (3.53∗ )
Government consumption in percent of GDP (t − 1)
−0.01 (−0.71)
−0.03 (−2.74∗ )
−0.04 (−4.49∗ )
Real GDP growth (t − 1)
−0.004 (−1.01)
−0.002 (−0.58)
−0.01 (−4.41∗ )
Rate of inflation (t − 1)
0.0001 (0.39)
0.001 (1.02)
−0.0001 (−3.23∗ )
Foreign short-term debt/ foreign debt (t − 1)
0.0002 (0.12)
0.01 (2.18∗∗ )
0.04 (2.50∗∗ )
Total debt service in percent of GDP (t − 1)
0.01 (1.18)
0.01 (2.34∗∗ )
0.02 (4.98∗ )
International reserves /money supply (t − 1)
−0.14 (−1.13)
0.10 (0.68)
0.02 (0.33)
Current account balance in percent of GDP (t − 1)
−0.002 (−0.74)
0.0003 (0.07)
0.002 (1.69o )
0.000001 (0.15)
−0.0001 (−0.67)
0.0001 (1.35)
0.42 (5.57∗ )
0.43 (10.64∗ )
0.17 (7.19∗ )
GDP per capita (t − 1) Lagged endogenous variable Number of countries
74
74
70
Number of observations
878
878
783
Method of estimation
OLS
2SLS
GMM
Time dummies
Yes
No
Yes
R2
0.36
0.19
(overall)
Sargan Test ( p-level)
0.99
Arellano-Bond-Test ( p-level)
0.27
Notes: See Table 3.
positive and only marginally insignificant for the IMF. Apparently, the World Bank loans which governments may draw under existing programs are exhausted after the election whereas post-electoral borrowing from the IMF comes from new arrangements with new policy conditions facilitating the required unpopular measures. Thus, macroeconomic policy conditionality seems to be more
16
DREHER AND VAUBEL
important in IMF than in IBRD lending. In this respect, the scapegoat role after elections is confined to the IMF. 3.
Implications for reform
Section 2 has shown that a country’s government budget deficit and its rate of monetary expansion are higher the less it has exhausted its borrowing potential with the Fund. Moreover, IBRD credits were found to raise monetary expansion in the recipient country. These are signs of moral hazard. They have implications for the reform of the international financial institutions. Moral hazard could be reduced, first, by establishing conditionality on an ex-ante basis (as suggested by Vaubel, 1991; and the International Financial Institutions Advisory Commission, chaired by Allan Meltzer, 1999). For example, all member states in which monetary expansion exceeds an n-year moving average of real GDP growth by more than x percent could be excluded from credits. With respect to fiscal policy, a limit for the budget deficit relative to GDP could be set (as is now in force in the European Union).28 Secondly, the moral hazard problem could be addressed by raising the opportunity cost of borrowing from the institutions. The interest rate subsidy could be eliminated for non-project lending (see Vaubel, 1991). Indeed, as recommended by the IFIAC for the Fund (1999), it could be replaced by a penalty so that the financial institutions become a lender of last resort. In the past, they have rather been a lender of first resort. Thirdly, it is possible to reduce moral hazard associated with IMF lending by strictly limiting the period over which a country may obtain credit (see Vaubel, 1983; IFIAC, 1999). While these solutions to the moral-hazard problem are relatively straightforward, it is much more difficult to prevent the IMF from contributing to political business cycles. This is because such lending is merely temporary and because past performance may not be a reliable guide to the future. Ex-ante conditionality can prevent governments from turning to the Fund after having embarked on overexpansionary policies before the election. But ex-ante conditionality does not prevent governments that have behaved well in the past from obtaining loans, even at an interest penalty, and then spending the proceeds to finance a pre-election boom. The conditions have to relate to the subsequent use of the loan, prohibiting a future increase of monetary expansion and the budget deficit. The Fund has imposed such conditions in the past. However, as our results show, this type of conditionality has not stopped pre-election borrowing. It seems to be necessary to improve the enforcement of such conditions by introducing more effective sanctions. The IFIAC (1999) has suggested that borrowers should have to submit some sort of collateral. Moreover, governments which have violated the agreed conditions, notably prior to their reelection, could be excluded from further borrowing for at least one term of office.
EVIDENCE FROM PANEL DATA
4. 4.1.
17
Definitions and data sources Section 2
“Overall Budget Deficit in percent of GDP’’, IBRD (2000): Overall budget deficit is total expenditure and lending minus repayments less current and capital revenue and official grants received. Data are for central government only. “Exhaustion of IMF quota’’, IMF (2000): The amount of IMF credit outstanding at the beginning of the year relative to the country’s quota in the IMF. “New net (non-concessional) IMF credit in percent of GDP’’, IBRD (2000), IMF (2000): Denotes net changes in repurchase obligations to the IMF for all uses of non-concessional IMF resources. “New net IBRD credit in percent of GDP’’, IBRD (2000): Disbursements of loans and credits less repayments of principal. Election Indices, Beck et al. (2001). “Real GDP growth’’, IBRD (2000): Annual percentage growth rate of GDP at market prices based on constant local currency. “Inflation’’, IBRD (2000): Consumer price index in percent. “Monetary Expansion’’, IBRD (2000): Average annual growth rate in money and quasi money for end-of-year data. 4.2.
Section 3 (additional variables)
“Index of Democracy’’, Marshall and Jaggers (2000). “Government consumption as a share of GDP’’, IBRD (2000). “Foreign short-term/ foreign debt’’, IBRD (2000). “Total debt service in percent of GDP’’, IBRD (2000). “Gross international reserves relative to money supply’’, IBRD (2000): Gross international reserves comprise holdings of monetary gold and holdings of foreign exchange under the control of monetary authorities. They are net of transactions with the IMF. “Current account balance in percent of GDP’’, IBRD (2000). “Real GDP per capita’’, IBRD (2000): Gross domestic product converted to international dollars using purchasing power parity rates. Notes 1. E.g., IFIAC (1999) and Vaubel (1991). Several recent studies claim that IMF programs reduce growth in the borrowing countries (see Barro and Lee, 2001; Hutchison, 2001; Przeworski and Vreeland, 2000). Boockmann and Dreher (2003) show that they did not lead to changes in structural policies. 2. For a summary see Vaubel (1991, p. 235f).
18
DREHER AND VAUBEL
3. See the sources quoted in Vaubel (1991, p. 205, p. 207) and Evrensel (2002, Table 2). 4. Evrensel (2002) shows that budget deficits, inflation rates and domestic credit, among others, are higher in the second inter-program-period compared with the first. According to Conway (1994), participation in IMF programs is more likely, the more frequently the country participated in the past. A strong tendency for recidivism has also been reported by Bird (1995). Bird, Hussain, and Joyce (2002) analyze some of the causes. The experience of having received IMF credit in the past thus seems to raise the probability of going for another IMF credit in the future. This might be called a “dependency trap” or an “hysteresis effect”. 5. For an empirical analysis of this issue see, e.g., Lane and Phillips (2000). 6. The argument assumes that bonds denominated in different currencies are imperfect substitutes (or that the reserve currency country does not sterilize the intervention). For a verbal explanation see Vaubel (1991, p. 212). 7. The scapegoat interpretation has also been proposed by Spaventa (1983), Vaubel (1983, 1986, 1991), Remmer (1986), Putnam (1988), Stein (1992), Edwards and Santaella (1993), Bjork (1995), Dixit (1996) and Przeworski and Vreeland (2000). 8. Killick (1998, p. 158) concludes that no systematic evidence in favor of the scapegoat role of the Fund is available. He does, however, also state that the IMF has been used as a scapegoat by a “good number of governments’’. Killick does not present an econometric analysis. 9. In theory, the availability of subsidized credit could also have caused moral hazard with nonborrowing countries. In practice, however, such effects are likely to be negligible. 10. If there is a deficit, the dependent variable has a positive value. 11. This variable is only weakly correlated with the exhaustion of IMF credit (r = −0.028) which is a stock variable. 12. It is also possible that the IMF is more lenient prior to elections in order to support the incumbent (Dreher, 2003). 13. The borrowing governments are encouraged to accept programs and loans without being willing to implement the attached conditions because they do not have to fear effective sanctions when they fail to comply with agreed policy requirements (see Hermes and Schilder, 1997). In fact, World Bank programs are almost never cancelled—even if non-compliance is evident (see Dollar and Svensson, 2000; Nash, 1993, p. 24; Mosley et al., 1991, p. 166). 14. For example, if an election would be in February, the index would take the value of 1/18 in that year, 12/18 in the previous year and 5/18 the year before. 15. This is in line with Schuknecht (1996, 2000). 16. Borrowing from the IMF may enable the recipient government to obtain (more) loans in the world capital market and thereby increase its budget deficit but most studies of this issue (e.g. Bird and Rowlands, 1997; Killick, 1995) show that IMF lending does not serve as a catalyst for private lending. 17. If the post-election effect dominates, IMF lending may on balance be found to precede lower investment and growth. This could explain the empirical evidence of a negative influence of Fund credits on growth reported above. However, results on economic performance are mixed. For example, contrary to the results reported above, Killick, Malik, and Manuel (1992), Schadler et al. (1993) and Dicks-Mireaux, Mecagni and Schadler (2000) found a positive correlation between IMF programs and growth. Ergin (1999) reports that IMF programs result in an (insignificant) reduction in the rate of real GDP growth and an increase in the current account. Inflation seems to be unaffected by the Fund. A negative influence of the Fund on investment has been reported by Conway (1994). With respect to the World Bank, Harrigan and Mosley (1991, p. 83) report a weak influence of structural adjustment loans on GDP. 18. Since the Sargan test rejects the over-identifying restrictions when all variables are included as strictly exogenous, the IMF and World Bank variables are treated as predetermined. This means that they are instrumented with their lagged levels instead of their lagged differences. 19. For a formal demonstration of this point see Sinn (1978). 20. Almost all empirical studies of IMF lending are confined to reduced-form estimates: Bird and Orme (1981), Officer (1982), Cornelius (1986), McDonald (1986), Joyce (1992), Edwards and
EVIDENCE FROM PANEL DATA
21. 22.
23. 24.
25.
26.
27.
28.
19
Santaella (1993), Conway (1994), Rowlands (1994), Bird (1995), Thacker (1999) and Bird and Rowlands (2000). The only exceptions are Knight and Santaella (1997) and Przeworski and Vreeland (2000). However, the separation of demand from supply factors in these studies is rather dubious. For example, Knight and Santaella classify the level of international reserves exclusively as a determinant of demand even though it also affects the Fund’s willingness to lend. Sturm, Berger, and de Haan (2001) provide a summary of these studies since 1990. Przeworski and Vreeland (2000, Table 1) present evidence of a positive relationship between budget deficits and IMF credits, while Joyce (1992, Table 2) reports a positive influence of government consumption on credits. Thacker (1999) does not find a significant effect of money supply growth on the probability of program conclusion. Regarding short term debt, it is, however, not obvious which is the cause and which the effect (Diamond and Rajan, 2000). Evidence to this effect is reported by McDonald (1986, p. 96), Joyce (1992, Table 2), Edwards and Santaella (1993, p. 427), Rowlands (1994, Table 1), Knight and Santaella (1997, Table 5), Thacker (1999, Table 4), Bird and Rowlands (2000, Full Sample, Table 3) and Przeworski and Vreeland (2000, Table 1). This is in line with Sturm, Berger, and de Haan (2001) who conducted an Extreme Bounds Analysis using 41 explanatory variables. There are some differences between their OLS analysis and ours. They do not allow for the precise date of the election within the election year but employ simple election year dummies. Moreover, they use two separate dummies for executive elections and legislative elections which are likely to be collinear so that the t-statistics are biased downward. They also have a somewhat different sample and report results for gross rather than net IMF credit. A more complete analysis of the relationship between elections and IMF programs is provided by Dreher (2003, 2003b). He shows that conclusion of an IMF program is less likely within six months prior to an election. Within twelve months prior to elections the paid-out share of the credit line rises in the case of democratic governments but falls in the case of autocratic regimes (which probably indicates non-compliance). Stone (2002), who reports results from a duration analysis, finds that the probability of turning to the Fund falls as the date of the election approaches. This is in line with our finding that IBRD loans raise monetary expansion (Table 2). IBRD loans increase monetary expansion the more so as the recipient governments know that higher monetary expansion leads to more loans from the IBRD (moral hazard in the broader sense). Of course, these proposals are controversial (see, e.g. the minority view in IFIAC, 1999).
References Arellano, Manuel and Stephen Bond (1991) “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations.’’ Review of Economic Studies 58:277–297. Assetto, Valerie (1988) The Soviet Bloc in the IMF and the IBRD. Boulder: Westview. Bandow, Doug (1994) “The IMF: A Record of Addiction and Failure.’’ In Doug Bandow and Ian Vasquez ´ (eds.) Perpetuating Poverty; The World Bank, the IMF, and the Developing World. Washington, DC: Cato Institute, pp. 15–36. Barro, Robert J. and Jong-Wha Lee (2001) “IMF-Programs: Who is Chosen and What are the Effects?” Paper presented at the IMF Annual Research Conference Nov. 29–30, 2001. Beck, Thorsten et al. (2001) “New Tools in Comparative Political Economy: The Database of Political Institutions.’’ World Bank Economic Review 15(1):165–176. Bird, Graham (1995) IMF Lending to Developing Countries: Issues and Evidence. London and New York: Routledge.
20
DREHER AND VAUBEL
Bird, Graham, Mumtaz Hussain, and Joseph P. Joyce (2002) “Many Happy Returns? Recidivism and the IMF.’’ Journal of International Money and Finance (forthcoming). Bird, Graham and Timothy Orme (1981) “An Analysis of Drawings on the International Monetary Fund by Developing Countries.’’ World Development 9(6):563–568. Bird, Graham and Dane Rowlands (1997) “The Catalytic Effect of Lending by the International Financial Institutions.’’ World Economy 20:967–991. —— (2000) IMF Lending: How is it Affected by Economic, Political and Institutional Factors?, mimeo. Bjork, James (1995) IMF Lending to Developing Countries, Issues and Evidence. London: Routledge. Boockmann, Bernhard and Axel Dreher (2003) “The Contribution of the IMF and the World Bank to Economic Freedom.’’ European Journal of Political Economy 19:633–649. Brunner, Karl and Allan H. Meltzer (1990) Money and the Economy—Issues in Monetary Analysis. Cambridge. Conway, Patrick (1994) “IMF Lending Programs: Participation and Impact.’’ Journal of Development Economics 45:365–391. Cornelius, Peter (1986) Theoretische und okonometrische ¨ Betrachtungen zum Ziehungsverhalten der Entwicklungslander ¨ in den Finanzierungsfazilitaten ¨ des Internationalen Wahrungsfonds, ¨ Diskussionsbeitrage ¨ aus dem Volkswirtschaftlichen Seminar der Universitat ¨ Gottingen ¨ 26. Diamond, Douglas W. and Raghuram G. Rajan (2000) “Banks, Short Term Debt and Financial Crises: Theory, Policy Implications and Applications.” NBER Working Paper Series 7764, Cambridge, MA. Dicks-Mireaux, Louis, Mauro Mecagni, and Susan Schadler (2000) “Evaluating the Effect of IMF Lending to Low Income Countries.’’ Journal of Development Economics 61:495–526. Dixit, Avinash (1996) The Making of Economic Policy: A Transaction-Cost Politics Perspective. Cambridge: MIT Press. Dollar, David and Jakob Svensson (2000) “What Explains the Success or Failure of Structural Adjustment Programs?’’ Economic Journal 110:894–917. Dreher, Axel (2003) “The Influence of Elections on IMF Program Interruptions.’’ The Journal of Development Studies 39:101–120. —— (2003a) “A Public Choice Perspective of IMF and World Bank Lending and Conditionality,” Public Choice (forthcoming). —— (2004) “The Influence of IMF Programs on the Re-Election of Debtor Governments.” Economics and Politics 16(1). Edwards, Sebastian and Julio Santaella (1993) “Devaluation Controversies in the Developing Countries: Lessons from the Bretton Woods Era.’’ In Michael Bordo and Barry Eichengreen (eds.) A Retrospective on the Bretton Woods System. The University of Chicago Press, pp. 405–455. Ergin, Evren (1999) “Determinants and Consequences of International Monetary Fund Programs.” Ph.D. dissertation, mimeo, Stanford University. Evrensel, Ayse (2002) “Effectiveness of IMF-Supported Stabilization Programs in Developing Countries.’’ Journal of International Money and Finance 21(5):565–587. Harrigan, Jane and Paul Mosley (1991) “Evaluating the Impact of World Bank Structural Adjustment Lending: 1980–87.’’ The Journal of Development Studies 27(3):63–94. Hermes, Niels and Ard Schilder (1997) “Setting Priorities: The IMF and World Bank Structural Adjustment Programmes.’’ In Peter B. Boorsma, Kees Aarts, and Albert E. Steenge (eds.) Public Priority Setting: Rules and Costs. London: Kluwer, pp. 39–60. Hutchison, Michael M. (2001) “A Cure Worse than the Disease? Currency Crises and the Output Costs of IMF-Supported Stabilisation Programs.’’ In Michael Dooley and Jeffrey A. Frankel (eds.) Managing Currency Crises in Emerging Markets. Chicago: University of Chicago (Chapter 10). International Bank for Reconstruction and Development (2000) “World Development Indicators, CD-ROM.” Washington, DC. International Financial Institutions Advisory Commission, (IFIAC), Chairman: Allan Meltzer (1999) Report, http://phantomx.gsia.cmu.edu/IFIAC/USMRPTDV. html, 1.9.2000. International Monetary Fund (2000) “International Financial Indicators, CD-ROM.” Washington, DC. Joyce, Joseph P. (1992) “The Economic Characteristics of IMF Program Countries.’’ Economics Letters 38:237–242.
EVIDENCE FROM PANEL DATA
21
Killick, Tony (1998) Aid and the Political Economy of Policy Change. London: Routledge. —— (1995) IMF Programmes in Developing Countries: Design and Impact. London: Routledge. Killick, Tony, Moazzam Malik, and Marcus Manuel (1992) “What Can We Know About the Effects of IMF Programmes?’’ The World Economy 15(5):575–597. Knight, Malcolm and Julio A. Santaella (1997) “Economic Determinants of IMF Financial Arrangements.’’ Journal of Development Economics 54:405–436. Lane, Timothy and Steven Phillips (2000) “Does IMF Financing Result in Moral Hazard?” IMF Working Paper 00/168, Washington, DC. Marshall, Monty G. and Keith Jaggers (2000) “Polity IV Project: Political Regime Characteristics and Transitions, 1800–2000.” http://www.cidcm.umd.edu/inscr/polity/. McDonald, Judith A. (1986) “Factors Influencing Countries’ Decision to Use Credit from the International Monetary Fund.” unpublished Ph.D. dissertation. Mosley, Paul, Jane Harrigan and John Toye (1991) Aid and Power—The World Bank & Policy-based Lending Vol. 1. New York: Routledge. Nash, John (1993) “Implementation of Trade Reform in Sub-Saharan Africa—How Much Heat and How Much Light?” Policy Research Working Paper 1218, World Bank, Washington, DC. Niehans, Jurg ¨ (1985) “International Debt with Unenforceable Claims.’’ Economic Review, Federal Reserve Bank of San Francisco, Winter: 65–79. Officer, Lawrence H. (1982) “The Differential Use of IMF Resources by Industrial, Other Developed, and Less Developed Countries.’’ The Journal of Development Areas 16:401–420. Przeworski, Adam and James Vreeland (2000) “The Effect of IMF Programs on Economic Growth.’’ Journal of Development Economics 62:385–421. Putnam, Robert (1988) “Diplomacy and Domestic Politics: The Logic of Two-Level Games.’’ International Organization 42(3):427–460. Remmer, Karen L. (1986) “The Politics of Economic Stabilization: IMF Standby Programs in Latin America, 1954–1984.’’ Comparative Politics 19(1):1–24. Rowlands, Dane (1994) “Political and Economic Determinants of IMF Conditional Credit Agreements: 1973–1989.” The Norman Paterson School of International Affairs Working Paper, Carleton University. Schadler, Susan et al. (1993) Economic Adjustment in Low-Income Countries: Experience Under the Enhanced Structural Adjustment Facility. IMF Occasional Paper 106, Washington, DC. Schuknecht, Ludger (1996) “Political Business Cycles and Fiscal Policies in Developing Countries.’’ Kyklos 49(2):155–170. —— (2000) “Fiscal Policy Cycles and Public Expenditure in Developing Countries.’’ Public Choice 102:115–130. Sinn, Hans-Werner (1978) “The Efficiency of Insurance Markets.’’ European Economic Review 11: 321–341. Spaventa, Luigi (1983) “Two Letters of Intent: External Crises and Stabilization Policy, Italy, 1973– 77.’’ In John Williamson (ed.) IMF Conditionality. Washington, DC: Institute for International Economics, pp. 441–474. Stein, Howard (1992) “Economic Policy and the IMF in Tanzania: Conditionality, Conflict and Convergence.’’ In Horace Campbell and Howard Stein (eds.) Tanzania and the IMF: The Dynamics of Liberalization. Boulder: Westview, pp. 59–83. Stone, Randall W. (2002) Lending Credibility, The International Monetary Fund and the Post-Communist Transition. New Jersey and Woodstock, Oxfordshire: Princeton University Press. Sturm, Jan-Egbert, Helge Berger, and Jakob de Haan (2001) “IMF Credit: How Important are Political Factors?” A Robustness Analysis, mimeo. Thacker, Strom C. (1999) “The High Politics of IMF Lending.’’ World Politics 52:38–75. Vaubel, Roland (1983) “The Moral Hazard of IMF Lending.’’ In Allan H. Meltzer (ed.) International Lending and the International Monetary Fund: A Conference in Memory of Wilson E. Schmidt. Washington, DC: Heritage Foundation, pp. 65–79.
22
DREHER AND VAUBEL
—— (1986) “A Public Choice Approach to International Organizations.’’ Public Choice 51:39–57. —— (1991) “The Political Economy of the International Monetary Fund: A Public Choice Approach.’’ In Roland Vaubel and Thomas D. Willett (eds.), The Political Economy of International Organizations. Boulder: Westview, pp. 205–245. —— (1996) “Bureaucracy at the IMF and the World Bank: A Comparison of the Evidence.’’ The World Economy 19:185–210. Vreeland, James R. (1999) The IMF: Lender of Last Resort or Scapegoat?, Yale University.
