Measurement errors skew our understanding of deep poverty

The Brookings Institute released a brief this week on deep poverty. The brief repeated two often cited claims:

Recently, both extreme and deep poverty in the US have become more prevalent. As shown below, this holds true for most definitions of income, even those including non-cash benefits. Although there is some disagreement about these figures and their causes, many scholars suggest that the rise is a consequence of the 1996 welfare reforms. Strengthened work requirements and newly implemented time limits meant that many low-income families entered work, received higher tax credits and improved their situation. But those unable or unwilling to move from welfare to work as intended received less cash assistance from the government and became reliant on a hodgepodge of government support, such as SNAP, housing subsidies, Medicaid/CHIP.

There is reason to be highly skeptical of both of these claims. When looking at consumption poverty measures (how much people consume rather than how much income they report), deep poverty rates have improved over time. Experts attribute the difference in trends using income versus consumption measures to the underreporting of income and public benefits in census surveys. According to Bruce Meyer of the University of Chicago and James Sullivan of the University of Notre Dame (page 169):

There is considerable evidence that changes in measurement error are important for families with few resources. First, transfer income, which is particularly relevant for these families, is significantly underreported in surveys, and the extent of this underreporting has grown over time. Meyer and others (2009) find that nearly half of food stamp benefits and TANF dollars are not reported in the CPS in recent years. Second, reported expenditure exceeds reported income among families with few resources (Meyer and Sullivan 2011a).

This underreporting of public benefits helps explain why deep poverty seems to have gotten worse following welfare reform, when in fact, it likely has more to do with measurement errors than the actual deterioration of the financial situation of the country’s poorest families. The Brookings brief reported that consumption among those in deep poverty was virtually identical to those in poverty, further suggesting that the concept of “deep poverty” is misrepresented due to measurement error.

There is no doubt that some families in the US are in deep poverty and attention needs to focus on them. But accurately reflecting the trends and correlates of deep poverty are important for this broader policy discussion.

Follow AEIdeas on Twitter at @AEIdeas.