Since weekly unemployment claims are out this morning, I thought we might revisit an odd permutation — call it an amusing hedonic adjustment — in last month’s Non-Farm Payroll, courtesy of the Penobscot Princess:
"Let’s get this bit of business out of the way before we proceed to dissect Friday’s NFP. Ahem. Each month and without much fanfare, the BLS releases a survey which shows the number of jobs reported by each of the 50 states and the District of Columbia. It’s got a one-month lag on it. Thursday, they put out the February report.
So I waited for Friday’s Non-Farm payroll to see the February revision. It dropped from 262k to 243k. Then I went here, and did all the arithmetic, backing out February from January, state by state. Guess how many jobs were reported by the states +DC in total? How about 197,300?
Now compare that to the 243k reported in the NFP Friday revision. That’s a difference of 18.8%, nothing to sneeze at, eh? So how the heck does this happen?
This calls for a little background check, directly from the BLS website: … “The CES survey is a Federal-State cooperative endeavor in which State employment security agencies prepare the data using concepts, definitions, and technical procedures prescribed by the Bureau of Labor Statistics.”
The BLS then goes on to caution not to try to add up the individual states and get them to jibe with NFP. They call this part “Caution on aggregating state data”. The BLS discourages summing up the individual states’ figures because the state series “is subject to larger sampling and nonsampling errors than the national series ….”
Is that right? How convenient is that anyhow? You be the judge as to why these numbers never jibe. Next case.
That’s right, kids: The aggregate BLS data is not equal to the sum of all 50 States, plus DC. (Imagine that)