Read on Twitter
This was great fun to write with superb coauthors (incl. @Ahu_Yildirmaz). We study the patterns of nominal wage adjustments using administrative payroll data from @ADP in order to shed light on questions principally surrounding wage rigidity. Thread 
https://twitter.com/AEAjournals/status/1314546451570200576
https://twitter.com/AEAjournals/status/1314546451570200576
Measuring wage adjustment is difficult. Survey data is full of measurement error, and most administrative datasets have earnings but no measure of hours and thus no wages. Plus compensation is complex. We use anonymized data from paychecks for ~15-20 million workers per month.
First we describe the nature of compensation. For most workers, base pay constitutes a huge share of income. However bonuses are important for the top of the distribution. It turns out that overtime is relatively small for most workers.
Wage rigidity matters in models because it mutes the cyclicality of firms' marginal cost of labor. We show that base wages are the only cyclical component of compensation and bonuses/overtime are almost i.i.d. Thus base wages drive the cyclicality of marginal cost.
Turns out, base wages look pretty rigid for job-stayers: ~35% of workers see no wage change in a given year and just 2% see wage cuts. I expect I will never again create a picture as striking as this one, which shows the incredible asymmetry in the wage adjustment distribution
This is important. The only cyclical piece of compensation looks pretty rigid. The ~35% with no annual change is between the range of Christiano-Eichenbaum-Evans (16.8%) and Christiano-Eichenbaum-Trabandt (43%).
There's strong evidence of time dependence in wage setting. Most wage changes happen 12 or 24 months after a worker's last wage change (though this is only really true for increases). Wage changes are also synchronized within a firm. Off-cycle wage changes tend to be larger.
Wage changes are also state-dependent. We show wage changes are pro-cyclical, but wage cuts are countercyclical. About 6% of workers got a y.o.y. wage cut at worst of 2008-09 Recession. Our BPEA paper also shows lots of wage cuts in COVID era
Job-changers get a lot more wage cuts than job-stayers. But a lot of this is down to selection in 1) who moves and 2) where they move to. If you match job-changers to a similar job-stayer in destination firm, they have similar wage behavior, perhaps for equitable pay reasons.
This suggests that firms can't just "get around" wage rigidity for job-stayers by hiring new people at a low wage. Job-stayer wage rigidity is therefore a "sufficient statistic" for rigidity. This echoes @JADHazell's finding of rigidity in wages on job postings.
Finally, we show that bonuses and overtime are cut idiosyncratically for workers. This makes it look like wages aren't rigid. However, these components don't vary cyclically, so probably not right for NK calibration. More work to be done on allocative consequence of non-base pay.
Overall, we provide a relatively complete picture of wage adjustment in the US which can discipline models. Job-stayer base wages pin down cyclicality of labor costs and look downwardly rigid. However, in severe downturns, even base pay can be cut.
