I'm a bit behind on economic data, partly because we know it's terrible and also know that the very short-term data that are we have so far are very imperfect. But I thought it might be worth some wonkish notes on unemployment claims 1/
So, we've had an amazing 30 million initial claims over a very short period. Plus, state offices that process these claims are both overwhelmed and in many cases designed to reject people. So it's natural to assume that true job loss >30M 2/ nytimes.com/2020/04/30/ups…
But if you look at the historical record, initial claims seem to overstate, not understate, job loss in slumps — and that's even if you discount the normal churn that causes several hundred thousand a week even in good times 3/
For example, here's annual sums of initial claims and measured unemployment over the course of the Great Recession. In 2008-2009 there were around 18 M "excess" claims, but "only" around 8 M rise in unemployment 4/ 
The NY Fed blog found similar results for NOLA after Katrina: around twice as many claims as job losses 5/ libertystreeteconomics.newyorkfed.org/2020/04/the-co…
I've been talking to people about this, and one answer may be that people apply multiple times after being rejected or failing to get an answer. Not sure about whether that's right. But some reason to doubt the 30 M job loss figure 6/
On the other hand, if we look at *continued* claims — people actually getting benefits — the relationship goes the other way, with UI understating true unemployment 7/
And continued claims are up by 16M, which is surely an underestimate. So true rise in unemployment somewhere in between. Split the difference and it's 24M, which is 15% of the labor force, implying an unemployment rate of something like 18% 8/
So yes, it's horrific, which you already knew. And already probably worse than CBO projections! Sorry, but this was an itch I needed to scratch 9/
