Cross-section regressions, so-called "Barro regressions," have been widely
used in world and regional income data to test whether convergence takes pl
ace. This paper applies alternative methodologies, a time series test and a
Markov chain model, to Japanese prefectural data and reexamines the result
s of J. Barro and X. Sala-i-Martin (1992, J. Japan. Int. Econ. 6, 312-346).
These methodologies show that the hypothesis of convergence as "catching u
p" does not hold for Japanese prefectural data. Through analyses similar to
those of Barro and Sala-i-Martin, Markov chain models are shown to be more
informative than cross-section regressions about the evolution of cross-se
ction data. J. Japan. Int. Econ., March 1999, 13(1), pp. 61-72, ECO/CS3, OE
CD, 2 rue Andre-Pascal, 75775 Paris Cedex 16, France. (C) 1999 Academic Pre
ss.