Abstract

Relative importance of the explanatory variables in linear regression is very crucial for implementing a policy by policy makers. In India demand for money is very important in determining the effectiveness of Government policy in changing the level of income, interest rate and price. In this paper we observe from the ANOVA model that partially interest rate is the most significant variable to explain the variability of the dependent variable, M1. But this is not the correct scenario because of the presence of multicollinearity. There is a strong multicollinearity between income and price. But while we do an average of orthopartial correlation and simple correlation, then the importance of the interest rate becomes the least. If the policy makers identify the actual reason for which it is happening and which variable is most responsible and accordingly proceed the outcome will be maximum. But in the existing literature we cannot correctly estimate the relative importance of explanatory variables in the presence of multi-collinearity - multicollinearity with or without enhancement synergism and with or without change in sign. The existing problem is not completely solved. Based on the work of Mondal (2008) this paper tries to prescribe a solution for finding the relative importance of explanatory variables in linear regression. The methodology is illustrated with the help of a time series estimation of demand for money function in India.

Authors and Affiliations