Heterogeneity in regional consumption due to current income shocks: Testing the permanent income hypothesis for German panel data

In the light of new theoretical and empirical work on the Permanent Income Hypothesis we tackle earlier findings for German data, which reject its validity given a large fraction of liquidity constrained consumers. Starting from a standard short run approach we do not find evidence for excess sensitivity and thus liquidity constrained households in the context of a dynamic panel data model for German states between 1970 and 2006. Our preferred specification relates changes in consumption to a 'surprise' term in permanent income as well as past values of consumption growth. We can interpret this specification as a solution to a consumer's optimization problem with habit persistence. Different from earlier findings changes of income growth measuring the excess sensitivity of consumption with respect to income changes and thus the degree of liquidity constrained households turn out insignificant. To check the robustness of these results we also combine the long- and short-run perspective in a co-integration model. Here we specify a Panel-ECM and look at its short run adjustment to judge about the share of liquidity constrained households. We find a significant but much lower fraction of constrained agents as in the earlier literature. Since we employ different estimators including (pooled) mean group estimation, we are also able to check for the asymmetry in the income-consumption path for German regions both with respect to the long and short run adjustment dynamics. Here we borrow from the literature on Poolability tests and search for macro regional clusters with similar adjustment paths. The findings show that for the sample of West German states between 1970 and 2006 both for short and long run parameters the assumption of poolability of the data cannot be rejected. However, we find a clear distinction in the short run path of the East German states since 1991 (which nevertheless converge to the same long run co-integration as the Western counterparts). One likely interpretation for the observed lower sensitivity of East German states to income shocks is the smoothening characteristics of income transfers to the East. As a robustness check we also apply the Panel-ECM with spatially filtered variables.