The Consumption Function: The Permanent Income Versus the Habit Persistence Hypothesis

Abstract

SINCE the publication Friedman's A Theory Consumption Function (1957), a number papers have appeared testing the permanent income hypothesis (PIH). With only a few exceptions (see Singh & Drost, 1971; Singh, 1975), most the researchers have estimated permanent income through the application Cagan's (1956) or Koyck's (1954) distributed lag model.1 As is well known, this leads to a regression current consumption on personal disposable income (PDI) and lagged consumption a form very similar to the consumption function implied by Brown's habit persistence hypothesis (HPH). Indeed, after the Klein-Friedman debate it has become almost a practice to consider HPH to be truly a complete anticipation of PIH notwithstanding the fact that the focus the two hypotheses is quite different. In fact, the application Koyck's model to PIH by its nature violates some the basic assumptions the latter. Not only does it bring about dependence between the transitory and permanent components income, but also yields transitory components income and consumption that may be mutually correlated (see Holmes, 1971; Singh, 1969; Walters, 1968). Thus, the tests PIH that use Koyck's distributed lag model in the time series context (see Holmes, 1970, 1972; Singh and Drost, 1971; Singh, 1975) clearly become dubious value. In addition, there are some builtin problems in the estimation a distributed lag model which although quite well known are often ignored in the empirical research related to PIH. This paper intends to discuss the basic differences in the two hypotheses and thereby show that it is wrong to identify HPH and PIH with each other. In section II, we set up the PIH and HPH models, outline various assumptions and examine in what respect the two models are different. In section III we analyze and compare the properties the estimators in the two models.