Conduct of Monetary Policy in an Uncertain World: The Case of Pakistan

Abstract

Most of the research on monetary policy assumes availability of information, regarding current state of the economy, at the time of policy decision. Orphanides and Norden (2002) convincingly put forward that estimates of output gap from real time data misrepresent true state of the economy and decisions taken on the basis of real time noisy data proved to be wrong when true data become available. Therefore, research that is conducted on the basis of revised data may mislead policy makers while choosing monetary policy strategy. Within this context, this thesis sets three objectives in the area of monetary policy of Pakistan assuming uncertainty in the estimated output gap. The first objective is to find evidence of wrong estimates of output gap in real time data. This is done by comparing estimates of output gap based on real time data with that in the revised data. The quasi real time data is also constructed such that the difference between estimates of output gap from real time data and that from quasi real time data reflects data revision and the difference between estimates of output gap from final data and that from quasi real time data portray other revisions including end sample bias. Moreover, output gap is estimated with the help of five methods namely the linear trend method, quadratic trend method, HP filter, production function method, and structural vector autoregressive method. The second objective of the thesis is to estimate the Taylor type rules using different estimates of output gap estimated in the thesis. Linear Taylor rule is estimated with and without lagged interest rate as one of the regressors. For nonlinear Taylor rule, two regimes are identified on the basis of business cycle (boom and recession) and inflation rate (low and high inflationary regimes). The third objective is to assess the optimality of Taylor rule with final as well as real time data. For this purpose historical series of output gap and inflation rate are simulated through estimated macroeconomic model but with rule induced, instead of actual, interest rate series. This process is done for both types of data sets, final and real time data, and both the historical as well as stochastic simulations are done to assess the performance of policy rule on the basis of loss, to the society, of deviations of output from potential level and that of inflation rate from the target. Data on real GDP, inflation rate, interest rate, gross fixed capital formation, unemployment rate, and labor force, over the period 1960 to 2010, for Pakistan economy are used for the analysis. Results indicate that estimates of output gap in real time data are different to what has been found in final data but other revisions, compared to data revisions, are found to be more significant. Moreover, output gap measured with all the methods, except the linear trend method, appropriately portray the state of the economy. It is also found that recessions can be better predicted by real time data instead of revised data and final data show more intensity of recession compared to what is shown in real time data. Regarding monetary policy conduct in Pakistan it is found that there is strong preference of State Bank of Pakistan (SBP) to smooth interest rate as coefficient of lagged interest rate is statistically significant and high in magnitude in all specifications. Moreover, SBP adopts opportunistic monetary policy only with respect to output gap and not with respect to inflation rate. Results of Taylor rule with final data are different to what have been found in real time data and these results are not robust against the use of different measures of output gap. With final data, coefficient of output gap is significant only in recessionary periods and in low inflationary regime. While with the real time data coefficient of inflation rate is significant both in the low and high inflationary regimes. The response to output gap is found insignificant in real time data. Monetary policy reversals are observed in the sample period as parameters in the policy reaction function are found to be unstable. The normative analysis shows that variances of output gap and inflation rate can be reduced if rule based policy is adopted. Results of both the historical and stochastic simulations show that the loss (sum of variances of output gap and inflation rate) based on rule based policy is less than that found in the actual data and this result remains same even in real time data.