(DP 1982-11) Truncation Bias in Household Money Demand Tests

Abstract

This paper estimates a demand for money function on household panel data from a national sample in the United States and constructed cross section interest rates. Its principal focus is the problem of truncation bias from a set of observations, which include zero observations of the dependent variable. This problem is treated as one of misspecification, the missing variable being the inverse of the Mills ratio. The auxiliary function which measures the probability of observing a positive value for the dependent variables, is used to estimate the inverse of the Mills ratio. This ratio is a significant explanatory variable in the demand for money equation - indicating the existence of truncation bias when only positive observations are used in the estimation procedure. There is an increase in the income elasticity with this added variable an indication of household economizing on money balances. Other results of the estimation exercise point to the importance of human capital and value of household time consideration in the household demand for money.