Problems in Using Household Survey Data

As mentioned in my post of March 10, the February employment figures from the Household Survey are 152 thousand below where they were last March. However, one needs to be very careful in comparing changes in employment using the Household Survey. One reason is that the Household Survey, involving just 60,000 households, is subject to a relatively large sampling error. Another reason is because population control adjustments are periodically made to the survey data but past data is not revised. This can cause a sudden jump or drop in the data such as the large jump in the year 2000 in the graph in the March 10 post. The following table from page 12 of the BLS summary of recent trends in both surveys shows the effects of population control adjustments made from 2000 to 2007:

Effect on household survey employment from
population control adjustments, 2000-07
(In thousands)
--------------------
January 2000  +1,555
January 2003    +576
January 2004    -409
January 2005     -45
January 2006    -123
January 2007    +153

Page 11 of the trends report states that there was likewise an adjustment in January 2008 which would have changed the reported employed in December of 2007 from the reported 146,334 thousand to 145,736 thousand, a decrease of 598 thousand. I'm not clear why the 146,334 does not quite match the figure on the latest Household Survey (I assume that it was revised) or why there is not a more discernible drop in the data. Still, it is likely better not to compare Household Survey employment numbers over a span that includes an adjustment. The trends report does include tables of smoothed and adjusted Household Survey data. This data can be useful as a check of the Payroll Survey data. Still, to compare employment data over time spans, it is likely best to use the Payroll Survey, both for its smaller sampling error and for the lack of the adjustment problem.

Job Growth Under Bush and Prior Presidents