Friday, January 23, 2009

Any time you survey, you will revise 

Steve Perry seems to think he's discovered something new: economic data can be revised.
As I noted in my earlier post this morning, the number of jobs lost in Minnesota for 2008 that was being bandied about the Capitol just last week was roughly 39,000. The new number, 55,000+, underscores just how rapidly the economy is losing ground beneath the feet of bean-counters and prognosticators.
Rapidly losing ground? Indeed the pace of recession, at least in terms of job loss, quickened in the second half of 2008 both nationally and in Minnesota. But go back to 1982, when job loss was in a 12-month period averaging 3% a year. Even for the second half of 2008, the average rate of job loss was 0.9% (Data.)

Data revisions are nothing new, and I noted last spring one that turned a negative outlook for the end of 2007 to a positive one. The data come from surveys, which like the polling data we all read last year in the elections are subject to standard errors. In Minnesota those bands on the unemployment rate, for example, are +/- 0.6%, so that when we say the unemployment rate this morning is 6.9%, we are using an estimate for which the 90% confidence interval is 6.3% to 7.5%. A revision like the one noted by Perry is no cause for gasps (the difference is less than .6% of total employment; here's the estimates of standard errors for employment by state.) It's part of the normal process of reporting and revising survey data.

Labels: ,