Asymmetric persistence in GDP? A deeper look at depth

If economic time series behave asymmetrically, then an interpretation of economic fluctuations based on linear time series models could be misleading. Beaudry and Koop (1993) recently argued that for post war U.S. GDP data there exists a statistically significant difference in persistence between negative and positive shocks. Their finding, if true, would be quite interesting since it would bring a new perspective to the literature on business cycle, which has been dominated by two conflicting views: the trend-reverting view of Blanchard (1981) and the permanent view of Campbell and Mankiw (1987). The purpose of this paper is to reexamine the evidence of asymmetric persistence of GDP by analyzing the statistical properties of BK's test. In particular, we show there are two pitfalls for this test: First, the t-statistic for testing asymmetry in persistence does not have a conventional interpretation. Second, a highly significant t-value may come from sources different from asymmetry. Using international data, we also explore the robustness of the BK result across the G-7 countries and find that the evidence is quite varied. Moreover, there appears to be no simple explanation for why countries display similar types of asymmetric behavior.