*Names in bold indicate Presenter
Recent calculations by Sherman suggest that the combination of food stamps and unemployment insurance prevented incomes to fall between 2008 and 2009 for those in the bottom third of the income distribution. Another line of research points to a large and growing share of people who do not work and are not reached by the safety net (Blank and co-authors). A third line of work finds that those raised above poverty by food stamps and other non-cash benefits as reported in major U.S. household surveys appear to be still worse off than many below the poverty cutoff (Meyer and Sullivan, 2012). A long line of research points to incomplete takeup of benefits for those with low incomes (e.g. Currie 2006). However, all of these analyses rely on survey data that is known to suffer from severe underreporting of welfare programs (e.g. Meyer, Mok, Sullivan 2009). Our previous work using administrative data on food stamp receipt in IL and MD linked to survey data found that false negative rates in major surveys ranged from 23% to 53%. We found strong evidence that misreporting is systematically related to observable characteristics. The extent of misreporting and its relation to observable characteristics is likely to lead to significant biases in studies that examine the determinants of participation in welfare programs, the distributional consequences of these programs, and other program effects.
For this project, we will have administrative data on food stamps, TANF and General Assistance linked to the CPS, ACS and the SIPP for New York State from 2005 to 2009. The accuracy of the administrative data will allow us to re-examine the questions raised above regarding the effectiveness, targeting and takeup of these programs and examine how underreporting biases the conclusions we draw from survey data. Since the data contains both participation and amounts received, we will be able to assess the distributional consequences of these programs and their effectiveness at shielding those at the bottom from the negative income shocks they incurred during the most recent recession. Comparing administrative and survey data will also allow us to learn more about the determinants of misreporting. We will develop methods that incorporate the insights from administrative data to improve estimation based on the more available survey data.