‘ The response of consumption to income : A cross - country investigation ’

757 Comments 'The response of consumption to income: A crosscountry investigation' The central facts driving this paper are that consumption growth is slightly forecastable, and the variables that forecast consumption growth also forecast income growth. The paper reviews a very clever interpretation of these facts in terms of 'rule of thumb' consumers that devote a fixed fraction of income to non-durable consumption. I only have two small criticisms of the paper itself. First, the basic facts are a little tenuous. The permanent income model predicts, and the data show, a strong association between ex post consumption growth and income growth. Thus, if by fishing one were to find a variable that spuriously forecast consumption growth, that variable is practically guaranteed to also forecast income growth, and thus spuriously indicate the presence of 'rule of thumb' consumers. And the R2 of the consumption growth forecasting regressions is low-as low as 0.01 for the U.S.-suggesting that this may be an important problem. Similarly, if one can account for the slight predictability of consumption growth by small changes in the model specification (non-quadratic or non-time separable utility, non-geometric depreciation of durables , seasonal adjustment, measurement or aggregation errors), this would also eliminate the basic facts behind the results. Second, I have to mention the obvious methodological complaints about adopting correlations in the data as decision rules. If we can posit that some consumers follow the empirically convenient rule of thumb ct=yI, why not just estimate a vector-autoregression of consumption, income and other variables, and call the consumption equation the 'rule of thumb'?