The Evidence Clearly Shows That Deep Poverty Has Worsened

There has been a lively discussion of late about the effect of welfare reform on the poor in this country. I have weighed in on the side of saying that welfare reform has made life worse for the very poor in our society. I've gotten some pushback on that, so I thought it would be helpful to briefly list the evidence that deep poverty (defined as 50% of the poverty line) has gotten worse.

Sherman and Trisi (CPS + TRIM; Persons)

Arloc Sherman and Danilo Trisi, of CBPP, released figures on deep poverty dynamics in July 2014 (and again in May 2015). They derive their figures from a modified version of the Current Population Survey (CPS) that corrects for the underreporting of social benefits. To correct for underreporting, they use the Urban Institute's Transfer Income Model (TRIM), which is a microsimulation model that uses administrative data on program participation and benefit eligibility rules to impute benefit receipt to households in the CPS.

Using this data, and a comprehensive income definition, Sherman and Trisi find that deep poverty for children went up after welfare reform, though it got somewhat better in 2010 because of a major expansion of SNAP.

Here is the graph showing change in deep poverty of children (up from 2.1% in 1995 to 3% in 2005):

Here is the even more telling graph that breaks it down by family type. If welfare reform was going to drive up deep poverty, you'd expect to find it most pronounced in unmarried families with children:

Finally, here is the graph showing SNAP coming in to save the day in 2010:

Moffitt (SIPP; Families)

Robert Moffitt has published figures about changes in deep poverty using the Survey of Income and Program Participation (SIPP). SIPP is a series of longitudinal surveys beginning in 1984. Unlike the annual surveys of the CPS, SIPP surveys its participants every four months. This leads to better recall and less underreporting. SIPP also oversamples lower income populations and, as the name suggests, is specifically designed to track participation in social benefit programs. For these and other reasons, SIPP is generally considered to be a better source of income data for poor people and especially poor people who participate in social benefit programs. Even though the level of underreporting in SIPP is far lower than the CPS, there still is some, and Moffitt adjusts the SIPP data to account for that underreporting.

Using this data, and a comprehensive income definition, Moffitt finds:

When we examine trends in the benefit system over the twenty years prior to 2004, we find that the overall contours of the system are unchanged and that the benefit system has continued to have a large impact on poverty. However, we also find major shifts in the distribution of transfers within and across demographic groups. Within single-parent and two-parent families, as well as those with nonemployed members, we find a notable shift in transfers away from those in deep poverty toward those at higher income levels, both below and above poverty. These trends reflect primarily the rise of the EITC and decline of AFDC/TANF and Food Stamps for the very poor, which is a regressive combination. We find that the post-transfer deep poverty rates for these groups have actually risen over time as a result.

In particular, the percentage of families who were in deep poverty rose from 4.5% in 1984 and 1993 to 6.6% in 2004. Broken down by family type, the percent of single-parent families in deep poverty rose from 7.9% in 1984 to 11.8% in 2004. Nonemployed families saw their deep poverty rate rise from 27.3% in 1984 to 52.5% in 2004.

Shaefer and Edin (SIPP + Survey Month Only; Households)

Luke Shaefer and Kathryn Edin have data from the Survey of Income and Program Participation (SIPP) that focuses on extreme $2-a-day poverty (so not deep poverty, but people even poorer than that). They restrict their observation period to the incomes reported on the same month that the survey was administered. Recall that SIPP contacts survey participants every four months and asks them to report their income for the prior four months. Naturally, the income they report for the same month that they answer the call is known to be the most accurate, as it requires less backwards recall. So, using only the month that the survey is conducted is a method for solving (or reducing) the underreporting problem. They don't do anything else to adjust for underreporting.

Using this data, and a number of income concepts, Shaefer and Edin find that extreme $2-a-day poverty has gone up in households with children. Based on cash income only (not counting tax credits), the extreme poverty rate went up from 1.7% in 1996 to 4.3% in 2011 (increase of 152%). Based on cash income plus SNAP, the extreme poverty rate went up from 1.3% to 2.2% (increase of 69.2%). Based on a cash income plus SNAP plus tax credits plus housing subsidies, the extreme poverty rate went up from 1.1% to 1.6% (increase of 45.5%).


So we've got three different methods here, all of which use comprehensive income definitions and correct for underreporting in one way or another, pointing to the basic conclusion that those at the very bottom of the income scale took a hit after welfare reform. This is exactly what you would expect given the nature of welfare reform, which reduced benefits for the very poor and increased them for those around the poverty line.