Civilrights.org

Introduction and Key Findings

In earlier research, we established that the nation's most important survey of labor-market activity― the Current Population Survey (CPS) ― may be systematically missing a large share of nonemployed adults.1 According to our estimates, based on a comparison of responses to the 2000 Decennial Census and corresponding months of the CPS, the undercounting of non-employed workers in the CPS raises the measured employment rate for adults in the CPS by about 1.4 percentage points. If our estimate is correct, the official employment rate for June 2006, for example, would have been 64.8 percent rather than the 66.2 percent reported by the BLS (2006: Table A-1). Since employment typically falls 1.5 to 2.0 percentage points in a recession, the magnitude of this measurement problem is of substantial economic significance.

In this paper, we provide additional estimates of the impact of undercounting in the CPS. For the most recent period where the analysis is possible, we produce estimates of the impact of the undercounting of the non-employed on national poverty rates and health-insurance coverage. More importantly, since the problems with undercounting appear to have become more severe over time, especially over the last decade, we also report simple estimates of the impact on employment rates of this deterioration in the representativeness of the CPS over time.

Our findings suggest that undercounting in the CPS has a substantial impact on our national measures of employment, poverty, and health-insurance coverage, and that the extent of the impact is likely to be growing over time.

• According to our earlier estimates, in 2000, the CPS appeared to miss about 1.4 percent of the adult population, or over 2.5 million non-working adults.
  
  
• If we assume that the non-workers who are not represented in the CPS have the same likelihood of being in poverty and have the same family structure as the non-working adults that do appear in the CPS, the official national estimate of poverty would have underestimated the actual number of adults and children in poverty by about 600,000 people (about 0.2 percentage points).
  
  
• If we assume that the non-workers who are not represented in the CPS have the same likelihood of being without health insurance and have the same family structure as the nonworking adults that do appear in the CPS, the official national estimates of the population lacking health insurance coverage would have underestimated the number of adults and children without health insurance by about 350,000 people (about 0.1 percentage points).
  
  
• The impact on poverty estimates for blacks and Hispanics are proportionately much greater. In 2000, the CPS underreported the poverty rate for blacks by 0.5-0.7 percentage points and for Hispanics by about 0.4 percentage points.
  
  
• Since the undercounting has become more severe in the CPS in recent years, estimates of employment rates from the CPS are biased and the bias is growing over time. For all adults, we estimate that the CPS overstated employment by about 1.1 percentage points in 1986, growing to 1.3 - 1.4 percentage points in 2000, and about 1.7 percentage points by 2005.
  
  
• The size and the increase over time in the bias in the CPS are largest for black men. We estimate that the CPS overstated black male employment by about 2.5 percentage points in 1986, rising to 3.0 percentage points in 2000, and reaching 3.5 percentage points in 2005.