Benchmark revisions to nonfarm payrolls

Fans of the many animated .gifs that have appeared here in recent months will surely not be surprised by what appears below - the nonfarm payrolls data before and after the benchmark revisions that were incorporated into today's labor report.
As would be expected, there was a huge 385,000 downward revision to the 2008 data but, surprisingly, the 2006 data was revised up by 40,000 and the 2007 data was revised up by 56,000, neither of which would seem to make sense (particularly the changes for 2007) given that the economic expansion was coming to a conclusion during these years.

In fact, nonfarm payrolls were revised upward by 155,000 in November 2007, the month before the recession began, and, at the official start of the recession in December, there was another upward revision of 79,000.

The revisions are a combination of updating both the birth-death model along with seasonal adjustment factors so, it's possible that the negative changes in former are being more than offset by positive changes in the latter.

Recall that, by the Bureau of Labor Statistics' own admission, the birth-death model does poorly at estimating job growth in real-time during economic turning points, underestimating job growth when a contraction turns into an expansion and vice versa.

ooo