Conventional wisdom suggests that industry is important for economic development. And while indeed the rise of industry coincided with the time when economic growth took off in the Western world, historically industrial regions (such as the Rust Belt in the US, the Midlands in the UK, or the Ruhr Valley in Germany) are currently struggling economically relative to other regions within their respective countries.
This post not only looks at why this is the case, but also at whether conventional wisdom is wrong: was it really industrialization per se that allowed economic growth to skyrocket in the 19th century? Or did these two events just happen to coincide?
Franck and Galor (2015) attempt to answer these questions in a recent working paper. They take the case of France from the second half of the 19th to the 21st century. The idea is to see how the income gap between historically industrialized and non-industrialized French departments (which are kind of like US counties) evolved over time.
The degree of historical industrialization is measured by the prevalence of steam engines in the 1860-1865 period. The distribution of the total horse power of steam engines across France is shown in the picture below.
Fresnes-sur-Escaut is where the first commercial use of the steam engine took place in France in 1732. It was used to pump water in a mine.
Franck and Galor show that the degree of industrialization in 1860-1865 initially had a positive effect on income per capita, but in the long-term this reversed. The effects of steam engine prevalence in 1860-1865 on income per capita in different years is plotted below with 95% confidence intervals. The plotted values are the regression coefficients of industrialization in IV specifications.
We can see that industrialization was initially very beneficial and its positive effects grew, but then there was a reversal and in modern times early industrialization is negatively associated with income.
Interestingly, in OLS specifications the coefficient of industrialization is positive even in a regression on 2001-2010 income per capita. Using an IV reverses this as seen above. This seems to indicate that once omitted variables are taken into account, industrialization has a negative long-term effect on income. It was thus these omitted variables (that correlate with industrialization) that drove the positive relationship in the OLS case. We can thus conclude that it was likely not industrialization per se that had a positive effect on income per capita, but rather some other forces that also happened to foster industrialization.
The authors also ask through what mechanisms industrialization affects income per capita negatively in the long run. They test three hypotheses.
Low human capital investments. Industrialized areas may have experienced a surge in literacy and schooling at the dawn of industrialization, but more recently these areas have been surpassed by other regions in terms of educational attainment.
Unionization and high wages. Regions with lots of industry were quite conducive to the emergence of organized labor. This in the long run could have made labor markets in these regions rigid and uncompetitive inducing firms to move to other regions.
Trade restrictions and protectionism. It is possible that industrial regions lobbied with the national government to increase import tariffs on goods that were produced in these regions. This could have protected these regions from international competition, but could have fostered inefficiency, which could have in the end led to their demise.
In short, the authors find evidence for low human capital investments but not for the other two. In other words, unionization and protectionism are likely not the main causes as to why industrialization is negatively associated with income per capita in the long run. The main culprit is low human capital investments.
Indeed, the composition of the workforce has changed quite a bit in industrialized areas. As the figure below shows (click to enlarge), industrialization is at first positively associated with the number of executives and intermediary professionals (both mid-to-high human capital occupations), but this trend reverses in recent times.
Let us now look at these two points: (1) why historically industrial regions could have low human capital today, and (2) why the composition of their workforce could have changed. (Note: the following paragraphs are my speculation based on the results of the paper.)
Let us start with (1). My understanding is that in the 19th and early 20th century, working in industry required quite a bit of human capital. This is not to say that industrial workers had to be better trained than today, but relative to the alternatives (e.g. mostly agriculture), industry required more skills such as literacy or numeracy. So it was a relatively high human capital sector of the economy, thus making it lucrative.
But in the second of half of the 20th century, many other sectors requiring human capital started to rise. In fact, these sectors now required more human capital than industry. Returns to human capital rose even further. But now industrial workers were not at the top but rather at the bottom of the human capital distribution.
Put another way, merely being literate/numerate was considered high human capital at the dawn of industrialization, but in recent times it is considered low human capital. It seems that it is not industry that’s changed over time, but the other sectors.
Second, as discussed above another mechanism that could have worsened the human capital of historically industrialized regions is changes in the composition of the workforce (2). At first, industry was small-to-mid sized factories. But as these firms started growing, they built more and more factories, they started expanding. So in the early days, when each firm had a low number of factories, executives and other more intellectual workers were stationed on-site in factories. Later, however, as the number of factories grew, more intellectual tasks were separated from factories (geographically). Office workers moved to separate facilities, usually in some central area like Paris. And the places that were traditionally industrial were left with employees only.
Thus to reiterate, the conclusion of this research is that it was not industry per se that led to a higher income per capita, but forces fostering industrialization such as a relatively high human capital. The more recent decline of historically industrial regions is, however, due to the fact that currently these regions lack human capital relative to other regions. This could have happened because of (i) developments in other sectors as hypothesized above, and (ii) changes in the composition of the workforce in historically industrialized regions (e.g. office jobs moving away).