Business cycle and herding behavior in stock returns: theory and evidence

Financial Innovation

Table 3 GDP growth rate, economic uncertainty, and the power law exponent

This table displays the regression results using annual data. Model (1) is a regression of the power law exponent on the US GDP growth rate. Model (2) is a regression of forecaster uncertainty on the GDP growth rate. Model (3) is a regression of the power law exponent on forecaster uncertainty. Models (4)–(6) are regressions of the power law exponent on two dummy variables: Dummy1 is defined as 1 when the GDP growth rate is higher and forecaster uncertainty is smaller than the sample average, and zero otherwise. Dummy2 is defined as 1 when the GDP growth rate is lower and forecaster uncertainty is larger than the sample average, and zero otherwise. The variance inflation factors (VIF) of Models (4)–(6) are less than 5, implying that multicollinearity does not reduce the precision of our estimated coefficients and cannot weaken the statistical power of our regression models (Table 7). The numbers within parentheses are z-statistics calculated with heteroskedasticity and autocorrelation-consistent standard errors, according to Newey and West (1987). *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. PLE denotes the power law exponent.