Survey Bias and Underestimated Unemployment: A Deeper Look at Racial and Age Disparities

This podcast episode discusses a new paper by Dean Baker and Julia Kai, co-sponsored by the Institute for New Economic Thinking and the Center for Economic and Policy Research, which investigates measurement error in the Current Population Survey (CPS) due to non-response bias. The core argument is that the Bureau of Labor Statistics' (BLS) current methodology, which weights non-respondents as similar to respondents, systematically underestimates unemployment, particularly among young Black men. The non-response rate in the CPS has significantly increased from 5% in the 1970s to 15% currently, with rates being considerably lower for Black and Hispanic populations, and lowest for young Black men (below 70%). The paper posits that those who do not respond are systematically different from those who do, drawing parallels to recent political polling misses where non-respondents held different views. The methodology employed leverages the longitudinal aspect of the CPS, where individuals are surveyed for four consecutive months. By analyzing the transition probabilities of labor market states (e.g., unemployed to unemployed) for those consistently in the survey, the authors impute the likely labor market states for missing observations. This approach reveals that the overall unemployment rate is underestimated by approximately 0.7 percentage points, while for young Black men, the underestimation is about 3-4 percentage points. Consequently, the official Black-white unemployment gap is found to be significantly wider, increasing from a reported 6 percentage points to about 8 percentage points, and Black employment rates are shown to be overestimated. The study also notes an upward trend in missing rates for most groups, suggesting that the bias in labor market statistics might be growing over time. Bill Spriggs, chief economist of the AFL-CIO, provides commentary, emphasizing that non-response is a general problem affecting all surveys and has implications beyond unemployment, including educational attainment and age distribution. He raises the possibility of "hidden homelessness" among young Black women, who are also a significantly missing group, and suggests that the current imputation method might benefit from a "within-group" analysis similar to how the Census Bureau handles missing data. Spriggs also discusses the historical trend of the Black-white unemployment rate, noting its apparent decline post-2008 and questioning if this is an artifact of the data bias. He highlights the disproportionately high unemployment rate for Black high school dropouts, suggesting that this group's unemployment is likely even more severely understated. The broader implications of this research are significant for understanding and addressing racial inequality in labor markets. More accurate unemployment figures, especially for marginalized groups, are crucial for effective policy design and resource allocation. The findings challenge the conventional understanding of labor market dynamics and racial disparities, suggesting that the economic struggles of certain communities are more severe than official statistics indicate. The discussion underscores the ongoing challenges in survey methodology, the importance of continuous scrutiny of data collection practices, and the need for professional organizations like the BLS to adapt to evolving societal behaviors and improve the accuracy of vital economic indicators.

"the issue that we're looking at specifically is whether coverage rate the lack of coverage is a problem"

"the problem is worse for black and hispanics that the coverage rate for blacks and hispanics is considerably lower and it's lowest for young black men it's less than 70 percent"

"they're assuming that the people that they don't cover are like the people that they do cover and we're questioning whether that's the case"

"we assumed that the transitions from one month to the next month are the same for people who end up missing as for those who are in there"

"we find that it leads to a higher unemployment rate of roughly seven tenths of a percentage point for the population as a whole and for young black men the group where the coverage rate is lowest it's about three percentage points higher"

"the gap in employment rates was largest for the groups with lowest response rates"

"if someone is unemployed or not in the labor force in the present month their likelihood to be missing from the survey next month would be higher"

"if no considerations are made for the under coverage the buyers might be more salient for black workers this is about 2.6 percentage point difference"

"the racial unemployment gap increased to about eight percent each point which is denoted by this yellow bar from its previous six percentage point difference"

"the unemployment rate for white high school dropouts is always lower than the black unemployment rate"