This study re-evaluates the validity of the joint rational expectations-permanent income hypothesis under the framework of seasonal cointegration using seasonally unadjusted quarterly data from Austria, Canada and Taiwan. Evidence is found that the consumption change only depends on the innovations of the income and the unemployment rate changes, and that agents are rational in forming their expectations, i.e., the joint hypothesis is supported by the data used. However, with the same data set, a similar test based on non-seasonal cointegration tends to reject the joint hypothesis, since the test ignores completely the possible stochastic seasonalities that may contain important information, as has been pointed out by Wallis (1974), embodied in the data.
I . In t r o du c t i o nThe current study re-examines the permanent income hypothesis (PIH), which was developed by Friedman (1957) and has figured prominently in consumption theory over the past forty years, under the additional hypothesis of rational expectations. The rational expectations-permanent income hypothesis (RE-PIH) postulates that agents are forward looking and make consumption plans on the basis of their permanent income rather than their current income. This hypothesis implies that temporary changes in income should have less impact on consumption than changes in permanent income. Hall (1978) showed theoretically that, under rational expectations with fixed real interest rates, consumption changes are unpredictable since the desired consumption evolves as a random walk process.According to Flavin (1985), the optimal consumption in period t, C t , under the PIH is expressed in first differences aswhere D denotes the first difference operator, y p t is the permanent income in period t, y t is current income, and y t À y p t is transitory income. Equation (1) is quite a general form that nests special cases of both the PIH, when b T ¼ 0, and an extreme form of the Keynesian consumption function, when b p ¼ b T . To avoid any potential misspecifications, equation (1) often extends to incorporating extra exogenous variables affecting consumption. Similar to the treatment of income, both the expected and unexpected components of the extra variables are included in equation (1). If the coefficients of the unexpected part of the extra variables are significantly different from zero, then the optimal consumption plan (1) is incorrect, which invalidates the PIH.