The Excess Smoothness of Consumption: Identification and Interpretation

The paper investigates the implications of the omitted information problem -- that is, the econometric problem which arises because an econometrician cannot explicitly include the complete set of variables potentially used by agents -- in the context of the "excess smoothness" phenomenon posed by Deaton 11987]. The paper shows that an econometrician who fails to take into account the effects of omitted information will incorrectly conclude that an empirical finding of excess smoothness of consumption implies that the income process is nonstationary. By contrast, with a more thorough understanding of the omitted information problem, the finding of excess smoothness of consumption is easily explained with two assumptions: a) the consumption data is generated by the excess sensitivity alternative hypothesis, in which consumption is a weighted average of current income and permanent income, and b) agents are forecasting on the basis of a larger information set than the econometrician. Further, excess smoothness is revealed to be consistent with a wide range of stationary income processes as well as nonstationary income processes. Thus the common presumption that the excess smoothness phenomenon is linked in an essential way to the stationarity or nonstationarity of the income process evaporates when omitted information is taken into consideration.