Thread by @sahilchinoy

i'm on the job market with a paper about political sorting in the labor market! it's perhaps harder to choose your coworkers than your friends or who you follow on twitter -- could the workplace be less segregated than other places where democrats and republicans tend not to mix?

to find out, we merge all US linkedin profiles with the nationwide voter file to create a panel of 35 million Americans with information about where they work and which party they registered with.

we first show how democrats and republicans sort into distinctive colleges, majors, occupations, industries, and employers. at each stage of their trajectory through the labor market, democrats and republicans are making different decisions. we can see this in great detail.

we show how this leads to segregation at the workplace: a democrat or republican's coworker is 10% more likely to share their party than we would expect based on the local political makeup. that's significant – comparable to ZIP code level partisan segregation.

it's not reducible to education, or industry, or even a very fine-grained measure of where everyone lives. so it's not just reflecting other partisan divides: workplace segregation is a phenomenon in itself.

then, we ask where political segregation comes from. do people adopt the politics of their workplace? we know that people pick up the partisanship of their neighborhood, and if something similar was happening with jobs, that could generate the segregation that we see.

we use a design that compares people moving between the same two workplaces but at different times. we find a causal effect of the workplace on partisanship for unregistered people and independents! but nothing for registered partisans: we don't see any party switching.

jobs can push people towards the democratic or republican party -- but they aren't changing people's minds, so this can't be the main channel driving segregation. instead, we focus on the other direction: whether politics shapes job choice.

we first look at 300,000 company descriptions, hunting for the signaling language that democratic and republican companies tend to use.

then, we plug those signals into ChatGPT to generate 30,000 synthetic job ads that are specific to an occupation and industry, and we ask workers about them in an online survey, showing them ads for titles they're interested in and randomly varying the salary offers.

we find that democrats + republicans are willing to trade off about 3% of their salary for an ideologically compatible version of a similar job (and some people would pay substantially more). that's ~$2,000/year -- this is not a small amount of money!

in a simple calibration, these preferences are strong enough to generate the segregation we see. so for some people, politics plays into job choice, and that can lead democrats and republicans to give up $ to choose different workplaces, even within the same city and industry.

the workplace isn't going to be a site of way more cross-party interaction than other social environments, at least for this group of workers. full draft (with the illustrious Martin Koenen) here: sahilchinoy.s3.us-west-1.amazonaws.com/chinoy_politic…

i like this paper because it has something for everyone: large-scale data and rich descriptive findings, a fun quasi-experiment zooming into job switchers, and a survey that leverages recent advances in LLMs to generate personalized treatments, inspired by our observational data.

i'm excited to keep studying politics + the labor market! this also plugs into my broader research agenda about the determinants of political behavior in the contemporary + 20th-century US.

contemporary: one project demonstrates the sizable influence of where you grow up -- your childhood neighborhood -- on your future partisanship and voting behavior (with Jacob Brown, Enrico Cantoni, Martin Koenen, and @VinPons)

historical: another project shows how being drafted into military service in WWI caused Black veterans to join the NAACP, and documents the pioneering role these men played in the early civil rights movement (with @ProfDesmondAng)

historical: another project examines zero-sum thinking -- the belief that for one group to gain, another must lose -- and shows how it's related to canonical forces in American history: immigration, mobility, and enslavement (with @DrNathanNunn, @SMGSequeira, and @S_Stantcheva)

paper drafts on my website: sahilchinoy.com/research