Appendix 1: Data and Methodology
Conway, P., 1998: ‘Evaluating Fund Programs: Methodology and Empirical Estimates’, mimeo, department of Economics, University of North Carolina at Chapel Hill (May).
De Melo, M., C. Denizer and A. Gelb, 1996: ‘Patterns of Transition from Plan to Market’, The World Bank Economic Review, Vol. 10 No. 3 (September), pp. 397-424.
De Melo, M., C. Denizer, A. Gelb, and S. Tenev, 1997: ‘Circumstance and Choice: The Role of Initial Conditions and Policies in Transition Economies’, mimeo (October), The World Bank.
Goldstein, M. and P. Montiel, 1986: ‘Evaluating Fund Stabilization Programs with Multicountry Data: Some Methodological Pitfalls’, IMF Staff Papers 33, 1986, pp. 304-344.
Havrylyshyn, O., T. Wolf, M. Castello-Branco and J. Berengaut, 2000 (forthcoming), ‘Growth Experience in Transition Economies’, IMF Occasional Paper.
Havrylyshyn, O., I. Izvorski and R. Van Rooden, 1998, ‘Recovery and Growth in Transition Economies 1990-97: A Stylized Regression Analysis’, IMF Working Paper No. WP/98/141.
IMF, 1997: ‘The ESAF at Ten Years: Economic Adjustment and Reform in Low-Income Countries’, IMF Occasional Paper No. 156, prepared by the staff of the IMF, International Monetary Fund, Washington D.C. (December).
Killick, T., M. Malik and M. Manuel, 1995: ‘What Can We Know About the Effects of IMF Programmes?’, World Economy, 15, 1992, pp. 575-597
Mecagni, M., (1999), ‘The Causes of Program Interruptions’, in Economic Adjustment and Reform in Low-Income Countries: Studies by the Staff of the IMF, edited by H. Bredenkamp and S. Schadler (Washington, International Monetary Fund)
Pastor, M. 1987, ‘The Effects of IMF Programs in the Third World: Debate and Evidence from Latin America,’ World Development, Vol. 15, pp. 249-62 (February)
Polak, J., 1991: ‘The Changing Nature of IMF Conditionality’, OECD Technical Paper No. 41, OECD Development Center, Paris (August)
Schadler, S., F. Rozwadowski, Siddharth Tiwari and D. Robinson, 1993: ‘Economic Adjustment in Low-Income Countries–Experience Under the Enhanced Structural Adjustment Facility’, IMF Occasional Paper No. 106, Washington D.C.: International Monetary Fund.
Ul Haque, N. and M. Khan, 1998: ‘Do IMF-Supported Programs Work? A Survey of the Cross-Country Empirical Evidence’, IMF Working Paper No. WP/98/169
Zettelmeyer, J. And Taube, G., 1998: ‘Output Decline and Recovery in Uzbekistan - Past Performance and Future Prospects’, IMF Working Paper No. WP/98/132.
We would like to thank Oleh Havrylyshyn for guidance and comments; Peter Uimonen for providing the data; and Robert Christiansen, Marta Castello-Branco and Ron van Rooden for their contributions. All errors are our own.
SB incorporate actions the country must undertake as part of its structural policies in order to ensure structural adjustment and long-term macroeconomic viability.
PC usually consist of a set of numerical floors or ceilings placed on various macroeconomic policy instruments or outcomes. Abiding by them within the stated levels is expected to allow the country to achieve macroeconomic stabilization. Whereas SB vary depending on the country’s specific needs for structural adjustment, PC tend to be generic across programs.
For our purposes ‘sustained’ growth is defined here as three or more consecutive years of positive real GDP growth. With less than a decade of transition and, therefore, only a few years of positive growth for any country, the study cannot focus on longer-term horizons.
MONA (Database for Monitoring Fund Arrangements) contains detailed information about programs for all countries, and has been compiled by the IMF’s Policy Development and Review Department since 1993. See Appendix 1 for a description of the MONA database.
For an excellent analytical survey of studies that assess the impact of IMF adjustment programs on economic outcomes see Ul Haque and Khan (1998).
During this period, the growth rate was significantly lower in program countries than in non-program countries, but the difference became smaller once a longer time horizon was considered.
Given the approach used in the study, however, this result does not necessarily imply that a Fund program by itself hinders growth.
However, we are able to ‘grade’ countries depending on how well they performed on the program. In other words, having a Fund program per se in our study is not sufficient information to make inferences about the likelihood of economic growth. This differs from the method used in several other studies, which essentially divide the sample to be tested into two possible states of nature: having a Fund program and not having a Fund program.
In the early 1990s most countries were at early stages of transition. Some countries, however, were at a more advanced stage (for example, Poland and Hungary experienced an earlier start of reforms), while other countries (such as Armenia, Georgia and Turkmenistan) could not proceed with reforms because they were experiencing internal conflicts. In the analysis of this paper the differences within transition countries can be controlled by using dummy variables in regressions, as in Havrylyshyn, Izvorski and Van Rooden (1998). Note that none of the Asian transition economies, and only some of the former republics of Yugoslavia are considered in the paper.
For an explanation of well-known caveats that apply to growth data for transition economies see Havrylyshyn, Izvorski and Van Rooden (1998).
Two interesting types of patterns emerge (Table 2): (i) A large difference on average between the number of actions requested under SBAs on the one hand, and under ESAF or EFF arrangements on the other. In general terms, the latter two arrangements require a greater number of actions, (ii) Many programs are quite comprehensive and require actions across several areas of structural reform. It is interesting to note that, on average, EFF programs tend to require more actions on financial sector reform than other arrangements, whereas ESAF programs tend to require more actions on tax and expenditure reforms.
A small negative but statistically insignificant correlation of -0.036 was found between the number of actions and a structural performance index constructed by the authors (the so-called Structural Benchmark Index explained below).
The amount of points subtracted was loosely based on the compliance scoring units (see appendix 1 section 4).
Belarus and Uzbekistan each had only one Fund program, which went off track within the first six months. In the case of Ukraine, two out of its three programs went off track.
For example, Bulgaria adopted an SBA in July 1996 which went off track almost immediately, obtaining an IFI score of 0. At the same time, a subsequent program (the only one for which data are available) received a perfect IFI score. Accordingly, the average score is somewhat weak. Since then, stabilization has been successful and positive growth of4 percent was recorded in 1998.
In this matrix, each column represents a region and each row represents the number of years of continuous positive real GDP growth the country has experienced. Studies have shown that, in general, Central and Eastern European countries recovered faster and sooner than the Baltics, which in turn recovered faster than the CIS countries. Research continues on how much of this was due to initial conditions, speedy implementation of reforms, or proximity to Western Europe.
For an early analysis of the effects of initial conditions on growth in transition economies, see De Melo et al. (1997). Initial conditions are defined as a series of variables measuring the extent of macroeconomic distortions before transition and the degree of overindustrialization.
However, by end-1998 the growth prospects for Russia had considerably worsened, after barely growing during 1997.
See Taube and Zettelmeyer (1998) for the case of Uzbekistan, and Havrylyshyn et al. (1999, forthcoming) for explanations of why these two countries have experienced positive growth despite lagging behind in reforms.
See Havrylyshyn et al. (1999, forthcoming).
These indicators measure on a scale from 1 to 4 the progress achieved in areas such as price liberalization, trade and foreign exchange system, small and large scale privatization, governance and restructuring, banking reform and interest rate liberalization and the securities market. The indicators are available every year since 1994 and are calculated for all transition countries.
Havrylyshyn, Izvorski and Van Rooden (1998) employed a methodology similar to that of De Melo et al. (1996) who constructed a so-called ‘liberalization index’ using EBRD transition indicators. De Melo et al. divided indicators into three major areas: (i) price liberalization, (ii) foreign trade liberalization, and (iii) privatization and banking reform, which they found crucial for achieving growth.
These regressions are taken from a study by Havrylyshyn, Izvorski and Van Rooden (1998), in which panel regressions were conducted to examine various determinants of growth in transition economies. The study found that between 1990 and 1993, the relationship between reform and growth has a U-shape, but that after 1993 it becomes positive. The authors argue that this may be due to the initial ‘destructive’ effects of price liberalization. Once the adverse effects of initial conditions ‘wear off’, reforms begin to have a positive, ‘constructive’ effect for the region as a whole.
A fixed-effect panel regression model with these equations was also run with country-specific constants. Although the goodness of fit increased, it was difficult to give an objective interpretation of the factors explaining the variation in economic performance across countries. For this reason, and noticing that the statistical significance of the coefficients generally remained intact, the regressions without constants were preferred. Tests for stability of the equations for the period 1993 to 1997 yielded good results. See Havrylyshyn, Izvorski and van Rooden (1998) for a more detailed discussion of the econometric tests.
Provided the gaps in the data are random, assigning the score 5 to cases when information is lacking should not create a systematic bias in the SBI. Another equally valid option would be to exclude such cases from the observation sample used in calculating the SBI.