I have blogged before about a rather surprising result (to me at least) regarding productivity convergence in manufacturing. Despite the lack of convergence among economies as a whole, there is apparently quite strong convergence within manufacturing industries. What this means is that manufacturing sectors that are further away from the technological frontier tend to experience more rapid productivity growth. And -- this is the really interesting part -- this happens regardless of the quality of domestic policies, institutions, or geography. Manufacturing productivity converges even if you are in a remote country with lousy policies and institutions. In economics jargon, manufacturing industries are subject to unconditional convergence.
The earlier results were produced using UNIDO data at the 4-digit level of disaggregation. The problem was that these data are patchy, and forced me to work with only few developing countries, over short-term horizons, and with only post-1990 data.
In the meantime, UNIDO issued its 2-digit data base (INDSTAT2), which covers many more countries and goes back to the 1960s. So I redid everything with the new data, not without some trepidation -- for you never know what new data will produce. But the results were equally striking and robust. Here is the scatter plot that illustrates the central finding, using the baseline sample (which looks for each country at the latest decade that has data):
Each dot in this scatter plot represents a 2-digit manufacturing industry in a particular country. The variable on the vertical axis is the growth of labor productivity in that industry, controlling for period, industry, and period X industry fixed effects.
The estimated rate of convergence (in my baseline sample) is 2.9 percent. This means that industries that are, say, a tenth of the way to the technology frontier (roughly the bottom 20 percent of the industries in my sample) experience a convergence boost in their labor productivity growth of 6.7 percentage points per annum (0.029 × ln(10)).
Here is the pattern for specific industries, where the convergence is also readily observable:
The new data allow me also to look at what economists call “sigma-convergence,” that is convergence in productivity levels. Even though I have a smaller sample (of 63 countries) to work with for this exercise, I still find substantial sigma convergence. Here is the picture for manufacturing as a whole, over the 1995-2005 decade:
Notice how the two peaks have moved closer over the decade. The difference of 10 log-points represents roughly a 10 percent reduction in dispersion.
All this begs the question why economies as a whole do not convergence, if manufacturing experiences strong convergence. The answer turns out to have three components.
First, non-manufacturing does not exhibit convergence. Second, manufacturing’s impact on aggregate convergence is curtailed by its very small share of employment, especially in the poorer countries. Third, the growth boost from reallocation – the shift of labor from non-manufacturing to more productive manufacturing – is not sufficiently and systematically greater in poorer economies. Taken together, these three facts account for the absence of aggregate convergence.
Almost all of the growth miracles of the last 60 years were based on rapid industrialization. Today, technological changes and global competition are fostering rapid de-industrialization (in terms of employment shares) almost everywhere. This makes me wonder whether the kind of rapid growth experienced by countries like South Korea, Taiwan, and China will ever become possible elsewhere.
The new paper can be downloaded here.
Recent Comments