Are U.S. regions converging? Using new econometric methods to examine old issues

Abstract

Are di¨erent regions of the United States experiencing convergence in levels of GDP? Carlino and Mills (1993) examined this question through time-series techniques, and found some evidence in favor of regional convergence. This paper checks the robustness of their results by using new econometric methods proposed by Vogelsang (1998). Our results, together with results from Loewy and Papell (1996), suggest there is stronger evidence in favor of convergence than previously thought based on the results of Carlino and Mills (1993).

Panel data stochastic convergence analysis of the Mexican regions

“…Country-by-country analysis considering the presence of structural breaks can be found in Loewy and Papell (1996), and Tomljanovich and Vogelsang (2002), who confirm the evidence of convergence obtained in Carlino and Mills (1993) for the US regions. Smyth and Inder (2004) for the Chinese regions, DeJuan and Tomljanovich (2005), and Rodríguez (2006) for the Canadian regions, and Strazicich, Lee and Day (2004), and Dawson and Sen (2007) for some OECD countries, and Carrion-i-Silvestre and German-Soto (2007) Recently, panel-data-based unit root tests have been used to conduct stochastic convergence analysis.…”

“…As before, we base our inference on the Pesaran (2004) To sum up, after the analysis has been controlled for the presence of multiple structural breaks and cross-section dependence, the results that have been obtained indicate that relative regional Mexican per capita incomes show stationary fluctuations around a broken trend, i.e., the study has found evidence of stochastic convergence. However, this sole evidence does not warrant the existence of convergence, since according to Carlino and Mills (1993) and Tomljanovich and Vogelsang (2002), actual convergence exists if stochastic convergence and E-convergence are verified.…”

“…To do so, we follow Tomljanovich and Vogelsang (2002) Table 4 reports the estimated coefficients for each of the m i + 1 regimes, along with the corresponding statistics of significance computed using the Newey and West (1994) robust estimator of the covariance matrix. Looking at these estimates we can conclude that there has been E-convergence when the coefficients of the parameters of each regime are significant at least at the 10% level of significance and have opposite sign, i.e., either when T i,k < 0 and J i,k > 0, or when T i,k > 0 and J i,k < 0 -using the notation in Tomljanovich and Vogelsang (2002), this situation is denoted as C. Similarly, we conclude that divergence has occurred when the coefficients of the parameters of each regime are statistically significant at least at the 10% level of significance and have the same sign, which is denoted in Tomljanovich and Vogelsang (2002) as D. It is possible to distinguish other interesting situations. Firstly, it would be possible that one of the parameters of the regime is not significant at the 10% level, but the other one is.…”

“…Secondly, Tomljanovich and Vogelsang (2002) characterize the situation in which both parameters are non-significant as the case where E-convergence has occurred so that we have achieved the equilibrium growth. This case is denoted by E. Table 5 summarizes the different situations corresponding to each regime.…”

European Integration, Productivity Growth and Real Convergence

“…(A.1) as follows: (A.4) where DU 1t ¼ 1 if tpT b (the break date) and zero otherwise, DU 2t ¼ 1 if t4T b and zero otherwise, DT 1t ¼ t if tpT b and zero otherwise, and finally DT 2t ¼ t À T b if t4T b and zero otherwise. His analysis, using Maddison's data (Maddison, 1991), and a later application by Tomljanovich and Vogelsang (2002) 35 that focuses mainly on convergence issues, provide interesting exploitations of this methodology.…”