I went on Marketplace yesterday talking to Kai about the BLS benchmark revision. Here's the easiest way I think to understand it: How many jobs are there in the US at any given time? The most accurate answer to this question will come from tax data. Employers are required to pay payroll taxes on each employee for Unemployment Insurance. So if we want to know how many jobs there are, we can use the payroll tax headcount. Done. The problem is that data can take 6-18 months to collect and process. It's a census of every employee in the US--over 160 million records! To fill in our understanding before then, the BLS surveys a sample of the tax-paying employers to come up with a monthly estimate, which by contrast comes out just days after the month ends. This is the Current Establishment Survey. The survey has two types of revisions. The first one is the monthly revisions, these relate to collection of the survey. Responses don't all come in on time, and they update that first number twice as more surveys are returned. The second one is the benchmark revisions, these relate to design of the survey's sample. The survey has to make some assumptions about business: how quickly they are opening and how quickly they are closing. When the tax data starts to roll in, they can see how close those assumptions were. Again, they update those numbers twice. We got the first benchmark revision yesterday, and we'll get the second one in the Spring. The journey from that first Jobs Day number to the final number two years later is a remarkable process that reflects timeliness, accuracy, transparency, and the dedication of the Bureau of Labor Statistics to keeping us informed. https://lnkd.in/eW4M-PGU
David Smith
6mo
Thanks Kathryn Anne Edwards for making this straightforward in understanding the survey process. Important work that takes time and refinement to provide accurate and actionable data.
Lu Wang
6mo
I’ve been following both you and Kai. As always you guys are amazing!
I heard the story…👏🏼👏🏼👏🏼