I found this via The Unbroken Window, where it was used to discuss whether we should care about climate change.
I actually care a lot about whether we should care about climate change, but not right now. Right now, I want to see the places people moved from linked to the places they moved to. The fact that people are moving to Florida is meaningless to me if I don’t know where they’re moving from.
There are 3,144 “county equivalents” (from Wikipedia) . That means the complete undirected graph has 3144(3143)/2 = 4,940,796 possible edges. I’m not sure how this is changes for a directed graph in which each edge appears exactly twice (once in each direction).
Richard Freedman writes (via a former professor’s excellent blog) that
Connections between workers and firms in the United States have more the flavor of a dating game, with workers and firms changing partners frequently, whereas connections between workers and firms in most advanced countries have more the flavor of a stable marriage.
I’d love to see an actual map of (a sample of) the labor market “dating network.” The nodes could be color-coded by length of stay and/or size-coded by salary, and the edges could be color-coded by time spent unemployed between jobs. I’m sure this is doable by scraping LinkedIn and maybe matching salaries from Glassdoor. Better yet, divide the labor force into cohorts by age and see how the networks differ across generations. There are at least two generations of users with significant representation on LinkedIn.
Then again, I’d rather know if the map differs for non-LinkedIn users, and how. And whether demographic differences would be confounding or causal. UPDATE: This and this might help (via Victoria Stodden), with this and other projects.
The paper is also a great, easy read. It lays out a lot of facts about employment and unemployment in the US, and then runs through a few plausible explanations of why this recovery was more jobless than the Great Depression (in terms of aggregate “employment-output elasticity”).