Thread by @annastansbury

What do we know about occupational mobility in the US?
As you may know, there isn’t very good existing data on it.

So @gregorschub, @Bledi_Taska & I construct new occupational mobility data, using an amazing new data set of 16 million U.S. resumes from @Burning_Glass. [1/N]

Resumes are snapshots of workers’ career histories. Assigning occupation codes to jobs, we can calculate transition probabilities from each occ to each other occ, from one year to the next.

We use this data to document 6 facts about occupational mobility. [2/N]

Fact 1: Occupational mobility is high. When workers leave their job, about 25% of them *also* leave their SOC 6-digit occupation. (6-digit codes are pretty narrowly defined…) [3/N]

Fact 2: Outward mobility varies a lot across occupations. So, the SOC 6-digit occupation is actually a pretty good description of the entire labor market for some workers (particularly highly specialized occs), but does a terrible job for others. [4/N]

Fact 3: Occupation switches don’t necessarily tend to be to other occupations within the same SOC occupation group… so we can’t capture workers’ labor markets effectively by just aggregating up to a broader occupational definition. [5/N]

Fact 4: The occupation transition matrix is sparse: most occupation-to-occupation transition cells have infinitesimally small probabilities. This means that for most workers, we can create a good representation of their labor mkt with only a relatively small number of occs. [6/N]

Fact 5: Occupational transitions are highly asymmetric. The probability of transitioning from a generalist to a specialist occupation is much lower than vice versa. (This means that symmetric measures of occupational similarity (e.g. task-based) should be used w/ caution!) [7/N]

Fact 6: occupational transitions capture a range of measures of task, skill, and amenity similarity across occupations – and can capture these, and other aspects of feasibility and desirability of occupational moves, in a non-parametric way. [8/N]

In our new working paper, we argue that you can use these occupational transitions to construct “probabilistic” labor markets for workers out of a cluster of related occupations. [9/N]

Thread:

https://twitter.com/annastansbury/status/1260568713830506500

Paper: papers.ssrn.com/sol3/papers.cf…

While the underlying resume data stays with @Burning_Glass, we’re working on making the occupation transition data available, including creating a web interface for researchers & policymakers. Please *be in touch* if you think this wd be helpful or have ideas of use cases! [10/N]

And thanks so much to @Bledi_Taska & @Burning_Glass for taking the plunge to work with two PhD students excited about occupational mobility - and for making so much data accessible to researchers more broadly! [/End]