Saturday, August 06, 2011

Labor Statistics and Confidence Intervals

According to the BLS establishment survey, non-farm private and public employment rose by 117,000 jobs in July.   http://bls.gov/news.release/empsit.b.htm.  This number was better than the consensus forecast of growth of 85,000 jobs.  It led at least one business economist to say we have "avoided the precipice."

But can a monthly number really tell us that?  The technical notes for the employment report says the 90 percent confidence interval for the establishment survey is 100,000 jobs.  This means the estimate has a standard error of about 100,000/1.64 or about 61,000 jobs.  To the distance between the consensus forecast and the actual number reported is slightly more than 1/2 a standard deviation.  This means we can be only about 60 percent sure that the employment number was better than forecast, which is a little better than a coin flip.



1 comment:

rjs said...

if you get past the optimistic headlines we saw new records in the numbers that really matter: the labor force participation rate declined to 63.9%, and the employment-population ratio declined to 58.1%, both the lowest since before reagan busted the unions & forced women into the workforce, and the average length of unemployment set a new record at 40.4 weeks..