A meta-analysis of the investment-uncertainty relationship

In this article we use meta-analysis to investigate the investment-uncertainty relationship. We focus on the direction and statistical significance of empirical estimates. Specifically, we estimate an ordered probit model and transform the estimated coefficients into marginal effects to reflect the changes in the probability of finding a significantly negative estimate, an insignificant estimate, or a significantly positive estimate. Exploratory data analysis shows that there is little empirical evidence for a positive relationship. The regression results suggest that the source of uncertainty, the level of data aggregation, the underlying model specification, and differences between short- and long-run effects are important sources of variation in study outcomes. These findings are, by and large, robust to the introduction of a trend variable to capture publication trends in the literature. The probability of finding a significantly negative relationship is higher in more recently published studies. JEL Classification: D21, D80, E22 1