V Simulations of the Growth Effects of Programs
- Malcolm Knight, and Mohsin Khan
- Published Date:
- November 1985
There are several reasons why cross-country and time-series studies can yield conflicting results for the effects of Fund programs on economic growth. First, the conflict may simply arise from differences in the methodology and empirical criteria employed by the two types of studies. The time-series studies, for example, are based on formal statistical tests, while the cross-country studies rely on a less formal and more judgmental approach to determining the effects of programs. Comparing the results across studies that employ different criteria for testing the effects of policies can, as such, be quite difficult. Second, the time period over which the effects on growth are measured can be an issue. While the time-series evidence in Section III is restricted to comparisons over a one-year period, the cross-country studies often look at longer (three-year) periods. This difference in time horizons for the tests could also explain part of the inconsistency that arises between the results of these studies.
Third (and this is perhaps the most important factor), while the time-series studies are concerned with assessing a single policy, and mainly a demand-oriented policy, the cross-country studies examine the effects of the whole package of policies implemented in the course of a Fund program. Whereas demand-management policies by themselves may cause the rate of growth of output to decline, other policies, particularly those stressing supply-side aspects, could work toward improving the growth picture. Furthermore, the adoption of a Fund-supported program may alter expectations, so that the effects on the ultimate objectives cannot be precisely determined a priori. Expectations are inherently non-quantifiable and thus cannot be easily incorporated into the time-series framework. Nor do these time-series studies reflect the positive growth effects of increased inflows of capital that may result both directly and indirectly from the implementation of a Fund-supported adjustment program, unless they focus explicitly on this question. Since cross-country studies concentrate solely on comparing the outcome of programs (with the historical pattern or with some control group performance) and attribute all changes that occur to the program, they automatically incorporate the effects of all factors, including expectations and increased external finance, that have a bearing on the outcome.
Clearly, it would be useful if the main advantages of the two approaches could be combined in some way. In other words, it would be worthwhile to incorporate a range of policies into the analysis, while at the same time abstracting from factors that influence growth but are exogenous to the program. One way to do so would be to expand the models underlying the time-series studies to handle a variety of both demand-side and supply-side measures and to use the resulting model to perform controlled experiments corresponding to alternative policy combinations. This type of simulation exercise might help to reconcile the observed inconsistency between results and could also provide insight on the main issue of the growth effects of Fund programs.
For this purpose, however, it is necessary to have at hand a structural model that incorporates the relations between various policies and certain macroeconomic variables, including in particular the level or rate of growth of output. No single model can generally cover the whole range of policy measures contained in a typical Fund program, although some small-scale models incorporate certain of the major policy instruments. One such model, developed by Khan and Knight (1981), satisfies the requirement of being able to handle several policies simultaneously, particularly those involving the control of aggregate demand and the exchange rate, and of assessing their effects on the main macroeconomic variables—growth, inflation, and the balance of payments. Furthermore, the parameters embedded in this model, estimated from a sample of 29 developing countries, are broadly consistent with those obtained in the studies cited in Section III. For present purposes, a slightly expanded version of this model, including certain explicit supply-side aspects, is used. With this model, described in the Appendix, the effects of different combinations of policies on growth can be studied by performing hypothetical simulation experiments.
The simulation experiments conducted here are, of course, purely illustrative and are not intended to reflect all the complexities surrounding a program. Formal models of any type are clearly unable to analyze all questions relating to Fund programs, and in particular they do not capture the complex ways in which policy variables are related to ultimate objectives. The model used here is highly aggregative and thus focuses only on what are considered the most important macroeconomic relationships. Also, while expectations are included, they are treated in a very simple fashion. Finally, although the parameters of the model reflect empirical estimates, the changes in policy and the combination of policies studied here are entirely arbitrary. These various reservations should be kept in mind in considering the simulation results.
The simulations conducted with this model start with the assumption that the authorities wish to achieve an (arbitrarily defined) increase in the stock of international reserves in a period of one year. To hit this target, it is possible to use demand-management policies alone or some combination of demand-side and supply-side measures. Since the time horizon is restricted to one year, supply-side measures alone cannot be used, as they tend to operate with a significant lag. This lag in the effect of supply-side policies is built explicitly into the model.
Specifically, the simulations trace out the effects on the growth of real GDP of the following:48
A set of demand-management policies, defined as a once-for-all 10 percentage point reduction in the rates of growth of nominal domestic credit and nominal government expenditures, and a 10 percent devaluation.49 Since prices do not adjust immediately in the model, these policies translate into real changes in the short run.
The above demand-management policies, combined with a set of supply-side policies that would raise the rate of growth of capacity output by 0.5 percentage point a year over a period of four years. In the context of the model this requires an increase in the investment-income ratio of about 2–3 percentage points a year for the same four-year period.50 No attempt, however, has been made to specify exactly the measures that would produce this result. As discussed in Section III, the increase in the investment-income ratio would have to be achieved by a combination of supply-side measures geared to the productive structure of the country under consideration.51
The results of the simulations are presented in Chart 1. Assume that the economy is initially growing at some arbitrary constant rate (equal to 5 percent a year) and that a stabilization program is introduced in the third period (year).52 If the policy package consisted solely of demand-side measures (Simulation a), the rate of growth would decline at the beginning of the program, as the tighter credit and fiscal policies reduced aggregate demand. The expansionary effect of devaluation is insufficient to offset these developments, and altogether the growth rate in this simulation falls by about 1.5 percentage points in the first year, but then starts to rise as inflation declines, raising real domestic credit and real government expenditures. The general improvement persists for about two years, and eventually the growth rate approaches its original level. The international reserves target is achieved with a pure demand-oriented policy package, but at the price of a transitory drop in the growth rate.
Chart 1.Effect on Growth Rate of Demand-Management and Supply-Side Policies
The output costs can be reduced significantly if appropriate supply-side measures are introduced simultaneously with the demand-management package. Assuming that these supply-side policies raise investment and thereby the economy’s trend growth rate of capacity output (in the present illustration by 0.5 percentage point a year), the actual growth rate would also start to rise.53 The combined package of demand- and supply-oriented policies (Simulation b) would still result in a decline in the growth rate in the first period, but now the negative impact of the demand-management policies is partially offset by the positive effect of the increase in the growth of capacity. Since the supply-side policies are assumed to raise the growth rate permanently by increasing capacity growth, the overall package succeeds in putting the economy on a higher secular growth path.
Despite being only illustrative, these simulation experiments yield three particular insights into the effects of Fund programs, as well as into reconciling the conflicting evidence produced by the time-series and cross-country studies. First, the combined effect of several policies implemented simultaneously turns out to be different from the effect of such policies enacted individually. Second, it is clear that suitable supply-side policies can help to offset, at least partially, any adverse short-term effects on growth that may result from demand restraint. Since the time-series studies look mainly at the demand-side policies, it is understandable why they reach the conclusion that certain policies included in Fund programs reduce the growth rate in the short run. Cross-country studies are arguably more ambiguous on this issue because they implicitly incorporate supply-side policies into the analysis. Third, it is clear that in judging the effects of programs, care has to be exercised with regard to the time period in question. In the present example, if one makes only one-year comparisons, the results in Figure 1 would imply that the stabilization program had significant costs, irrespective of the of policies it contained. In contrast, if the period of comparison were extended to, say, three years, the conclusion would be quite different.