MODELING CONSUMPTION FUNCTI ON USING DISTRIBUTED LAG MODELS

Abstract

Abstract This study utilized the ‘Distributed Lag Models’ which can be defined as the unique equations of a distributed influence upon both dependent and independent variables. This influence is of varying degrees which can last for one prolonged time interval, discreet intervals or infinite intervals. This study is comprised of six chapters: the first four chapters contained the theoretical framework for ‘The Distributed Lag Models‘; besides that, the data of the study. On the hand, the last two chapters were allotted for the empirical aspects of the study. The problem of this study is to identify the methods of applying ‘Distributed lag models‘ with reference to autocorrelation and multicoleanerity, focusing on the influence of the consumer persistence habits and current income upon the current consumption. Therefore, the distributed lag models equations were used to as methodology to estimate the Friedman's Consumption Function which seemed to have lag coefficients of a geometrically distribution. In addition to that Bayesian Analysis method is used to compare and contrast four models of error term employed in this study. I The hypothesis of the study is that, the current consumption is not influenced by the persistence consumption. Hence, through the collection and the analysis of time series data of a seven consumer products from the 1980-2005; this study aimed to estimate consumption functions using statistical methods unlike other economical analysis methods. The results were that the model was significant with regard to the first and the fourth model. It was sometimes significant with regard to the third model; and it was insignificant with regard to the second model. This could be due to the problems of autocorrelation and multicolinearity. The study

concluded that, the factor which influenced consumption during the duration of the study was Habit Persistence. Finally, the most important recommendations of the study were that, when distributed lag models are applied in varied studied it should be expected that problems of autocorrelation and multicolinearity might occur. Therefore, researchers are advised to adopted large sample and increase the lag interval in order to lessen the effect of the lag variables utilizing ARIMA models and transfer functions models. i I n conclusion it could be assumed that studies applying Lag Models are scant regionally and internationally, therefore lag models should be used more often in studies as such because it enables researchers to maintain efficiency, sufficiency and unbiasness in their estimations; on the condition that the problems of the ordinary least square method were redden off. However, it is not advisable to apply lag models unless Bayesain method is used.