Empirical Modelling and Model Selection: An Application of Encompassing Principle on Consumption Function of Pakistan
Author: Shah Abbas

The area of model specification and selection of variables is quite vast in its nature and scope. Several estimation techniques and approaches are designed to select the appropriate model. Its involve comparing competing model by defined goodness of fit or selection criteria. The most commonly used information criteria are Akaike (1973, 1974) information criteria (AIC) and Bayesian Schwarz (1978) information criteria (BIC). The model best-fit criteria are R2, adjusted R2 and student t-statistic. Another important development is the use of encompassing principle in testing non-nested models assuming that one of them is correctly specified. The estimation procedures and approaches are simple to general, general to simple, stepwise regression, forward and backward procedure. The main objective of this study is to select an appropriate consumption model for Pakistan from the set of candidate models. Quarterly time series data is used to estimate different consumption models for the period 1972-2015. Absolute income hypothesis, permanent income hypothesis, and random walk model is estimated by using the traditional econometric methodology. The results are found to be the misleading and improper choice of the specification as several models miss-specifications problems such as autocorrelation, conditional Heteroskedasticity in residuals and omitted variable problem. In order to improve empirical results, Hendry’s general to specific modelling strategy is adopted to select the appropriate model. To develop a general model of consumption several variables are identified from the literature and provide theoretical explanations how private consumption is affected by the change in GDP, remittance inflow, government consumption expenditure, indirect taxes, interest rate, financial wealth, prices and inflation rate. The general model is simplified gradually step down and arrived at a parsimonious model by testing sensible economic ii restrictions. Each simplification steps are checked by diagnostic testing. The several consumption models are estimated and choose general model which encompass pervious estimated consumption models. The empirical results shows that GDP, remittances, price and inflation are positively related with private consumption in short run and long run. The current interest rate is positive and pervious interest rate is negatively related with private consumption. The government consumption expenditure and indirect taxes are negatively associated with private consumption. The empirical results suggest that private consumption is affect through different channels such as interest rate affect private consumption by saving channel and government expenditure effect the private investment through taxes. This study provide some policy insights to policy makers, Pakistan relies mostly on the foreign and domestic loans, it has adversely affect the private consumption as well as overall economic activities and this will lead to increase against the debt services. Therefore reliance on foreign financing should be avoid and generate the domestic resources. The political instability, law and order condition and transparency should be improve which can attract the foreign investors. The state bank of Pakistan should conduct monetary policy efficiently and effectively with regard to its basic object of stabilizing prices under the fiscal control. Supervisor:- Dr Hafsa Hina

Meta Data

Keywords : Encompassing, Error Correction Model, General to Specific Model
Supervisor: Hafsa Hina

Related Thesis​