This paper seeks to assess the importance of time‐varying regional patterns to countries' per capita growth rates, and their effect on the conclusions that can be drawn from growth regressions.
Pooled ordinary least squares on a data set of five‐year average growth rates for 101 countries over the period 1960‐1999 are used.
It is shown that time‐varying regional dummies explain more of the variance of per capita growth rates than do commonly used independent variables. This may indicate a problem of omitted variables with a strong regional dimension, or alternatively that growth is highly “contagious” within a region, perhaps through trade. Variables such as the growth of neighbouring countries or trading partners appear to be highly statistically significant when time‐varying regional effects are ignored, but are much less so when they are properly controlled for, and may simply be capturing unobserved regional effects. There is evidence that these variables reflect international business cycle correlation rather than the advantages of trading with, or being close to, faster‐growing countries.
There are strong regional patterns to growth of which one is still far from a full understanding.
Growth of neighbouring countries or trading partners may be much less important for a country's per capita growth than is sometimes claimed.
The findings show the importance of applying appropriate robustness checks to empirical results, and (in a two‐dimensional data set) of exploring whether variables explain the time‐series or the cross‐section dimension.
