What Do Cross-Sectional Growth Regressions Tell Us about Convergence?
This paper tests the dynamic implications of beta-convergence with time-series data from the 48 contiguous U.S. states. The motivation for this paper rests with the interpretation of results from cross-sectional growth regressions. These results show that poor regions experience faster per-capita income growth than rich regions. This is interpreted as evidence of convergence. However, convergence is a dynamic adjustment process with testable implications in time-series data, while the literature employs cross-sectional data to estimate this dynamic concept. A set of strong assumptions must be made to jump from this cross-sectional correlation to its interpretation as a speed of convergence. We find that the time-series properties of the data appear to be inconsistent with beta-convergence dynamics. Further, our analysis rejects the assumptions necessary to interpret the cross-sectional correlation as a speed of convergence. Therefore, our results call into question the interpretation that has been placed on this important cross-sectional finding.