Our (me, Guillaume Liegey and Grant Lovellette)original motivation was to draw policy implications at the individual level (i.e what can be done to upgrade skills so that prospects for higher productivity and salaries are improved). However, due to time and data limitations we ended up linking occupation codes to industry categories instead of our original idea which was linking career codes (a categorization of the skill-set an individual has)with occupations. As a result our conclusions were at the industry rather than the individual level.
As mentioned in the previous post we borrowed the conceptual framework and methodology from the "Product Space" (see previous post for reference). The Product Space is a view of the world where both countries and products can be represented as vectors of capabilities so a product exported with revealed comparative advantage is the result of the intersection of capability vectors. This means that products connected in the Product Space require the same capabilities. Our Job Space is a view of the world where both jobs and industries are vectors of skills. But in the Job Space an industry results of the union of the skill vectors for the jobs employed in that industry. Therefore the fact that two jobs are connected doesn't mean that they require the same skills but that they are complementary.
I was a bit disappointed with this because the parallel became less straightforward. The Product Space helps think of what capabilities a country has and how upgrading these capabilities results in the country being able to produce more and more sophisticated products ultimately accelerating economic growth and helping countries move up the income ladder. We had hoped to draw a parallel where the Job Space helped think of what skills a worker has and how upgrading those skills could help the worker hold more sophisticated jobs ultimately moving up the income ladder. However, my disappointment vanished when Grant found this article on the New York Times that reports on “Immigration, Offshoring and American Jobs".
According to the Times, this paper proposes that:
We see the job-creating benefits of trade and immigration every day, even if we don’t always recognize them. As other papers by Professor Peri have shown, low-skilled immigrants usually fill gaps in American labor markets and generally enhance domestic business prospects rather than destroy jobs; this occurs because of an important phenomenon, the presence of what are known as “complementary” workers, namely those who add value to the work of others.
The idea that an industry arises when different complementary jobs are joined is very much in line with what we found. As mentioned above, we started with the idea that skills connect occupations with industries:
However, we don't directly observe skills in the data. What we observe is workers that are employed in a particular job in a particular industry. Using the method of reflections (see Product Space paper) we transformed the bipartite network formed by jobs (occupation codes) and industries (North American Industry Classification Codes) into a one mode projection where jobs are connected to other jobs if they are found in the same industry. What emerged was something that looks like this:
In terms terms of network analysis, our most relevant findings were:
- The Job Space is not a random network
- Jobs form industry clusters that that cut across wage and education level
- Jobs strongly connected within clusters, weakly connected across clusters
- Very decentralized, no “core”
- A few very central jobs are key to the connectivity of the network: removing them results in quick disintegration of the network
- The highest paying jobs do not appear to be strongly connected to the network, except for certain types of jobs (e.g. directors, managers and engineers)
- The lowest paying jobs tend to be strongly connected to each other and within certain industry clusters (e.g. agriculture, domestic services, textile, food processing)
No comments:
Post a Comment