April 08, 2012
In a NYT Economix blogpost Jason DeParle ponders the fact that government surveys are not showing much increase in poverty, even though we know there are many people experiencing long periods of unemployment and many forms of government assistance have been cut back. One possible explanation is that people in poverty and extreme poverty are less likely to be covered by the survey.
My colleague, John Schmitt, found clear evidence of a coverage problem in comparing employment rates as shown in the 2000 Census and the overlapping months of the Current Population Survey (CPS). This is a useful check on the accuracy of the CPS, the main survey for measuring both unemployment and poverty, since the Census has near universal reach with a response rate of close to 99 percent. By comparison, the coverage rate for the CPS is close to 88 percent.
Even after applying a Census adjustment formula, Schmitt still found a substantial difference in employment rates, with the CPS showing an overall employment rate that was more than a full percentage point higher than the Census. The difference was largest for groups with the lowest coverage rates. In the case of young African American men, who have a coverage rate of close to two-thirds, the CPS showed an employment rate that was 8 percentage points higher than the Census for the same months of 2000.
These results are consistent with a story where the CPS is missing more people through time and the people who it misses are disproportionately at the bottom of the income ladder. If this is true, then there could be a rise in poverty that is largely missed in the standard surveys.
Comments