Paradoxically, researchers using major national surveys often find that participating in the U.S. Food Stamp Program appears to be associated with increased rates of food insecurity and hunger. Of course, the nation's leading anti-hunger program is not itself to blame. Rather, some low-income families face greater food hardship at any given time, while other low-income families may have enough food. A family with greater food hardship is more likely to take the trouble -- and perhaps endure the stigma -- of asking for food stamps.
In a conference paper published next month in the Review of Agricultural Economics, Mark Nord of the Economic Research Service and I for the first time use longitudinal or "panel" data from a major national survey to study the relationship between food stamps and household food security. Panel data have repeated observations on the same households over time -- in our case, thousands of households in the federal government's Current Population Survey were interviewed one year apart in December 2001 and December 2002. Sometimes, under certain statistical assumptions, panel data permit researchers to control for unobserved household characteristics, like being a "hardship" family or a "non-hardship" family in the case of food security. In our research, Mark and I thought a statistical model using panel data might show that food stamps help participants to improve their food security status.
In fact, the study found that using panel data reduced but did not eliminate the paradoxical positive association between food stamp participation and food insecurity. For example, we found somewhat to our surprise that families who entered food stamp participation between December 2001 and December 2002 were more likely to have deteriorating food security status, while families who exited food stamp participation were more likely to have improved food security status. We hold onto our assumptions anyway -- we still don't believe food stamp participation is causing increased hardship. But these results tell us something fascinating about the nature of the unobserved household characteristics that have been confounding this area of research. Those "unobservables" must be largely time-varying. In other words, the unobserved hardships don't just hit some families and not others. Rather, the hardships are very time specific -- many families are fine one year and then for some reason that is not observable in the data they face tough hardships the next year. Interesting.
p.s. The same journal issue includes a critical commentary on Mark's and my article by nationally known poverty economist Jim Ziliak. He liked some things about the article, but also expresses concern that too few of our households may have changed status between the two time periods for our work to be definitive. My time on the research project, and that of Friedman School graduate student Jerusha Nelson Peterman, was supported by the National Poverty Center. The issue also offers related articles from the same conference session by many colleagues and friends working on food assistance and food security issues, including Nader Kabbani, Myra Yazbeck Kmeid, Marianne Bitler, Grace Marquis, and Craig Gundersen.