The Effect of State and Local Taxes on Economic Development: A Meta-Analysis

Abstract

Economists and policy makers have long questioned the effect of state and local taxes on economic development. According to Schmenner [55], economists have long argued that taxes had little impact on business location decisions, while development practitioners and elected officials ignored this advice and aggressively pursued development by using tax incentives. More recent research using improved data and methodologies has begun to show that taxes can make a difference, and the emerging consensus among economists now says that taxes do matter. This changing view among researchers is perhaps best exemplified in a 1991 study by Timothy J. Bartik [2]. Bartik examined 84 econometric studies completed since 1979 which assess the impact of state and local taxes on economic growth in the U.S. He concluded that the long run elasticity of business activity with respect to state and local taxes was between -.10 and -.60 for studies focusing on intermetropolitan or interstate business economic activity and between -1.0 and -3.0 for intrametropolitan areas. However, Bartik's study did not control for a variety of study characteristics that may have influenced the measured elasticity. This study builds upon Bartik's analysis by performing a meta-analysis of the studies Bartik reviewed. Meta-analysis is a method for statistically analyzing results across empirical studies. Though it has been little used by economists, researchers in psychology, education, and health sciences have used meta-analysis extensively. This study uses a meta-analytic technique suggested by Stanley and Jarrell [57] to analyze the studies reviewed by Bartik. The technique yields increased precision for our understanding of the relationship between state and local taxes and economic development. The objective of this study is two-fold: first, to derive a more precise estimate of the tax elasticity and, second, to determine how the inclusion and omission of key variables in the estimated equations influences the elasticity estimates.