Borrow, Adjust, Repeat: The Dynamic Economics of Pakistan's IMF Recidivism and its Cost to the Poor

Abstract

Pakistan’s quarter century under IMF programs since 1958 provides perhaps the most instructive, and certainly the most salient, laboratory in international political economy for its citizens. This study subjects the Pakistani laboratory to the full force of contemporary multivariate time series econometrics. The results, obtained using the techniques of Vector Autoregression (VAR), Johansen’s multivariate cointegration, Vector Error Correction Modelling (VECM), Impulse Response Functions (IRFs), Forecast Error Variance Decompositions (FEVDs), and Distributional Regressions, covering forty-two annual observations (FY1981/82–FY2023/24), reveal a system which is at once self-correcting and recidivist, capable of finding its own equilibrium but forever thwarted in its efforts by the stop-go nature of Fund conditionality. The Johansen procedure confirms the presence of two cointegration vectors in the GDP-Investment- Inflation-Exchange Rate system, thus establishing the presence of stable long-run equilibriums which are periodically upset by Fund programs. The results of the VAR estimation reveal exchange rate depreciation as a significant and negative Granger cause of GDP growth (F=7.169, p=0.008), while the Fund program dummy suppresses GDP growth by 0.641 percentage points (p=0.042) when the full set of conditioning variables is employed. The VECM estimation reveals a GDP self- correction coefficient which yields a GDP self-correction half-life of 2.4 years. The recidivism phenomenon is thus not a reflection of a lack of self-adjusting capabilities, but a constant disruption of those capabilities. The impulse response functions reveal the exchange rate shock as the largest negative influence on GDP in the system (-0.949 percentage points in Year 2, cumulative -0.806 percentage points). The estimates of the distributed lag provide a striking and somewhat troubling pattern of when the entries of the Fund programs occur. Growth falls right away by an average of 1.694 percent, but then rises slightly in the second and third years. However, this growth is short- lived as growth slows down again in the fourth year, declining by 1.412 percent. The distributional regression provides insights into the social impact of these programs. For instance, in IMF program years, poverty rises by 1.767 percent on average, and job creation falls by 0.450 million. All these findings indicate that although these programs are designed to promote economic stability, they often come at a significant short-run cost for growth as well as social welfare.