Stars, Wars, and Development

As part of the credibility revolution in applied economics, researchers need to turn away from significance thresholds and stars, and instead focus on economic importance. Yet, existing measures of importance in regression analysis have fundamental flaws. Using an axiomatic approach, I develop two new measures, which decompose the variance of the dependent variable into contributions associated with each regressor. I demonstrate the relevance and complementarity of the two measures in two cases studies. The first shows that the statistically significant relationship between genetic distance and inter-state wars is economically irrelevant. The second identifies the most important factors of long-term development