Analysis And Estimation Of Output Gap Of Pakistan

Analysis and implementation of Monetary Policy hinges on dynamic behaviorand inter-linkages of many macroeconomic variables. Being part of the ResearchCluster of Central Bank of Pakistan (State Bank of Pakistan-SBP), I got theprivilege of hands–on knowledge of the dynamics that shape the economy of Pakistan.At SBP my area of responsibility was the assessment of Real Sector of Pakistan,therefore the dynamics linking Monetary Policy implementation with the RealEconomy and its growth trajectory in the long run seemed very interesting to me. Inaddition, at SBP I was involved in the analysis and assessment of Business cycles ofEconomy of Pakistan. These two major areas of through provoking analysis led me tothink in terms of research in growth potential of Pakistan Economy.Output Gap and Potential Output are two latent macroeconomic variables linkeddirectly with the assessment of Business cycle analysis as well as affective MonetaryPolicy implementation. Robust estimates of Output Gap provide unique informationabout the trends and cycles of the Real economy. A significant and positive OutputGap indicates existence of inflationary tendencies (i.e. High Aggregate Demand)while a negative Output Gap posits the existence of excess capacity (i.e. excessAggregate Supply) in the economy. I believe this nexus of interrelated researchquestions is the reason of my thesis and research.In this thesis I sample and implement various methods of estimation of theOutput Gap for the case of Pakistan, while previous studies on this subject matteronly aim to estimate the Output Gap on annual frequency. This is due to dearth ofhigh frequency data on National Income Accounts (NIA) for Pakistan. I propose andimplement a theory consistent method for obtaining Quarterly TemporalDisaggregation of Real GDP of Pakistan. In the subsequent analysis, quarterlyestimates of Real GDP are utilized for estimation of Output Gap based on ‘Structural’and ‘Uni-Variate’ methods. Chapter1contains introduction to the concept of Output Gap, the motivation forresearch and discussion for rational of the study on Output Gap of Pakistan.In the first essay (Chapter2) I detail various Uni-Variate methods to estimate thePotential Output and the Output Gap, these methods have sound grounding in theliterature and are in active use at various Central Banks for estimation of Output Gap.The estimation exercise is conducted on Annual and Quarterly frequencies, which isespecially useful for real-time policy input and analysis.In the second essay (Chapter3) I estimate and forecast the Output Gap ofPakistan on Quarterly frequency using Phillips curve based State-Space Methodology.The estimation of Output Gap and anecdotal evidence regarding the business cyclesof Pakistan has analogous interpretation and direction. I also estimate the Output Gapusing Structural VAR in essence of Blanchard and Quah (1979). Also included in thisessay is analysis of Output Gap using Wavelet filtering in essence of Motohiro (2008).Results of all three methods have sound economic interpretation. I also forecast theOutput Gap of Pakistan using a host of Auto regressive (AR), ARIMA and StructuralRegression methods.Literature on Business cycle analysis suggests that estimates of latentmacroeconomic variables like Output Gap and Potential Output have Monetary andFiscal policy significance at high data frequency. Unlike previous, this paper aims tocircumvent the issue of shortage of quarterly National Income Accounts (NIA) forPakistan by estimating a robust proxy for real total output based on Large ScaleManufacturing Index (LSM Index). This paper implements three commonlyemployed methods for estimating the Output Gap, these include; State–Space model,Wavelet filter and Structural VAR model. I conclude that in line with themacroeconomic aggregates; demand pressures in Pakistan have subsided sinceFY2009Q3in addition negative Output Gap since2011Q3is due to slowdown in realeconomy i.e. the Aggregate Supply. Lastly, I forecast that the current trend of lowaggregate demand is expected to last until FY2015Q4Chapter4contains analysis and forecasting of Output Gap and Inflation in astructural Monetary and Credit variable environment. I use a host of Bayesian VARs,differentiated using adequate priors (3different priors), in order to forecast Output Gap and Inflation. Results indicate that the methodology employed in this thesis canarguably circumvent the problem of Multicollinearity, high estimation errors and overparameterization of regression model. In addition, by programming a Metropolis-Hastings Algorithm as in Korobilis (2009), I illustrate how joint posterior density isobtained in cases where such density is not available analytically. Also I use a Gibbssampler for obtaining Predictive Densities and inference.Elaborating, I attempt to forecast inflation and Output Gap of Pakistan usingBayesian VAR Methodology. I implement three different priors for this purpose.Analysis in this paper is conducted using Monetary Aggregates and Credit macrovariables. Output Gap used in this analysis is estimated in a State–Space frameworkusing Kalman filter. Literature suggests that Bayesian shrinkage is an appropriate toolfor forecasting using large number of Macro Economic variables. In addition,appropriate Prior selection is fundamental to robust forecasting in Bayesian VARs; inthis backdrop, the3types of Priors implemented in our analysis are;1: MinnesotaPriors,2: Independent Normal–Wishart Priors and3: Independent Minnesota-Wishart Priors. Estimation and forecasting is conducted in conformity with Koop andKorobilis (2009). Diagnostics of Bayesian VAR models and robustness of forecastestimates show that Bayesian VARs provide robust forecasts and have suitablestructural interpretation, particularly, the results indicate that Independent Normal-Wishart Priors and Independent Minnesota Wishart Priors illustrate better forecastperformance than Minnesota Priors.Essay No.4(Chapter5), illustrates Forecasting Inflation using Bayesian VARmethodology. In contrast to Chapter4I implement7different priors for estimation ofBayesian VARs. Results indicate that SSVS Methods as well as Independent-NormalWishart priors provide better forecasting performance in forecasting horizon greaterthan3months, in addition, other Bayesian VARs exhibit respectable results too.In the last essay I implement Bayesian State-Space Method in line with Planas C.et al (2008), Maximum Likelihood parameter estimates of a similar state-spacemethod are used priors, thereafter posterior density of each parameter is estimated bymodeling them in various distributions. I implement a Metropolis-Hasting algorithmalong with a Gibbs sampler for draws for posterior density. I also apply the Multivariate Filter Method of estimating the output gap as inBenes J. et al (2010). I implement the small macroeconomic model from Benes J. etal (2010), whereby3major equations for Dynamic Output Gap, Unemployment andCapacity Utilization are modeled along with the relevant laws of motion.Concluding; the results from these methods compare well with other methodsutilized in this thesis.