A Cautionary Tale of Using Data From the Tail
- PDF / 284,001 Bytes
- 8 Pages / 439.37 x 666.142 pts Page_size
- 44 Downloads / 251 Views
A Cautionary Tale of Using Data From the Tail James X. Sullivan 1 # Population Association of America 2020
Introduction Understanding the levels and trends in deep and extreme poverty in the United States is of great interest to both policymakers and researchers. Social safety net programs are designed in part to prevent such extreme deprivation, and evidence that individuals and families slip through the cracks informs debates on how to improve social policy. To measure deep and extreme poverty, Brady and Parolin (in an article published in this issue of Demography) use data from the Current Population Survey (CPS) to estimate the fraction of individuals in the United States that live in households with income below 20% of median income (about $7,300 in 2016), which they call deep poverty, and below 10% (about $3,600 in 2016), which they call extreme poverty. They estimated that 5.2 to 7.2 million Americans (1.6% and 2.2%) were in deep poverty and 2.6 to 3.7 million (0.8% and 1.2%) were in extreme poverty in 2016, and that the rates of deep and extreme poverty have risen sharply over the past 20 years. In addition, they conclude that the expansion of Supplemental Nutrition Assistance Program (SNAP) benefits has led to declines in deep and extreme poverty for households with children. Brady and Parolin make an important contribution by bringing greater attention to this issue of extreme deprivation in the United States. However, a large and growing literature using linked survey and administrative data has shown quite convincingly that income is significantly underreported in large national surveys, particularly for individuals and families in the far left tail of the reported income distribution.1 This
1
Other studies that relied on extreme outliers of survey income data include Fox et al. (2015) and Edin and Shaefer (2015).
* James X. Sullivan [email protected]
1
Department of Economics and the Wilson Sheehan Lab for Economic Opportunities (LEO), University of Notre Dame, Notre Dame, IN 46556, USA
J.X. Sullivan
underreporting leads to an overestimation of extreme and deep poverty and an underestimation of the impact of government programs. Brady and Parolin acknowledge concerns about measurement error and address them by imputing some income from government programs. Efforts to address underreporting using microsimulation models, however, do not accurately allocate imputed benefits to true recipients. Moreover, recent studies that relied on linked survey and administrative data have shown that many income sources besides government program income are underreported for those at the bottom of the reported income distribution (Meyer et al. 2019). Accounting for these other income sources would lead to lower estimates of extreme poverty.
Evidence on Underreporting of Transfer Income Many studies over the past 25 years have documented with both indirect and direct evidence that survey income is significantly underreported. We know from these studies that reported family income is often far below re
Data Loading...
