Using a VECM to characterize the relative importance of permanent and transitory components of U.S. total disposable income and total consumption

Abstract

This paper uses cointegrated two-variable autoregressions (Vector Error Correction Model—VECM) to characterize the transitory components in U.S total disposable income and total consumption. The results indicate that almost all the variance of consumption change is explained by the trend, while the variance of the cyclical component accounts for an estimated 64% of the total variance of income change, consumption is the trend in income, but is not a random walk; furthermore, there is evidence of excess sensitivity and excess smoothness in consumption. While multivariate trend-cycle decomposition isolates large cyclical components in income, the persistence is high. This outcome suggests that the persistence of total income is not a useful indicator of the size of trend in the series confirming two well-known results: (i) the random walk specification for the trend, common in the univariate analysis, is biased towards establishing the permanent component as important; (ii) the importance of the cyclical component depends on the information set and is higher when the multivariate estimate is performed. The interpretation of these features of the data cannot be performed in terms of the traditional permanent income model.