Poor countries have access to new technologies already developed elsewhere so should grow more rapidly than richer economies. This is one of the implications of standard growth models, as well as of common sense.
But in reality, there is no automatic tendency for economic "convergence" among countries at different levels of income. Convergence depends instead on a number of additional determinants. It is only those developing nations with the "appropriate" preconditions – for example, adequate schooling or physical investment – that manage to absorb new technologies sufficiently rapidly and therefore to catch up. In the language of growth economics, there is conditional convergence, but not unconditional convergence.
When we look at the same question at the level of individual industries rather than countries a surprising finding emerges. Suppose we focus on, say, plastics, furniture, or the auto industry in developing countries. Does productivity in these (and other) industries experience automatic convergence with the technological frontier? Or is convergence once again conditional, depending on a host of country-level variables?
The interesting (and I think new) finding is that productivity convergence appears to be unconditional at the industry level – at least for manufacturing industries and for the period since the 1980s. Here is the picture:
The picture shows unconditional labor productivity convergence for 4-digit (ISIC) manufacturing industries during 1996-2006. (More details: This is the partial correlation between labor productivity growth and initial labor productivity from a regression that includes industry dummies. The data are from UNIDO.) The rate of convergence is relatively rapid (the estimated coefficient on the log of initial labor productivity is -0.03, compared to the -0.02 that one typically finds in conditional convergence regressions using country-level data). (The t-statistics are very high, even when standard errors are clustered at the country level.)
The conditional rate of industrial convergence (obtained by adding country fixed effects to the above regression) is naturally even higher at -0.07. Very similar results are obtained from samples from other recent time periods and from pooled regressions over different time periods.
I add that a precursor to this result was produced by Jason Hwang in his Harvard PhD dissertation. Hwang showed that there is an unconditional tendency for product unit values to converge in exports. The present results show that this is true for production (in manufactures) as well, and for a direct measure of productivity (labor productivity, or value added divided by employees) as well.
These results are important because they underline once again the importance of structural change – or, of getting into the "right" industries. The reason that unconditional convergence fails at the country level seems to be that not all industries (agriculture?, informality? many services?) are equally adept at absorbing technologies from abroad.
The trick seems to be moving labor into the manufacturing (and other) industries where there is automatic convergence to the global productivity frontier.