New Access to Experimental Data Announced

New Access to Experimental Data Announced

Mark Shroder, Associate Deputy Assistant Secretary for Research, Evaluation, and Monitoring

In January, HUD announced a new competition for access to the data from two of the most important randomized social experiments ever conducted, the Moving to Opportunity for Fair Housing (MTO) demonstration and the Family Options Study. HUD has partnered with the U.S. Census Bureau to make the experimental data more available to qualified researchers and more readily matched with other data.

The MTO demonstration ran from 1994 to 1998 and involved 4,600 families from high-poverty public housing projects in Baltimore, Boston, Chicago, Los Angeles, and New York. These families were randomly assigned to one of three groups: those who received a housing voucher that they could use only in low-poverty neighborhoods, those who received a housing voucher without that limitation, or those who could continue to live in public housing. Participating families were tracked directly through 2010, and researchers have conducted some matching exercises from later data. The power of the experimental design permits researchers to derive the impact of neighborhood on the lives of low-income families with children without bias from self-selection or other omitted variables.

Researchers have shown considerable interest in these experiments. In his book Evicted: Poverty and Profit in the American City, Matthew Desmond writes, “According to Google Scholar, there are more than 4,800 scholarly articles and books in which the phrase ‘Moving to Opportunity’ appears in the text…. In other words, by now every family who benefited from Moving to Opportunity could have their own study in which their program was mentioned.”¹

Of course, most of this work is secondary or tertiary literature; fewer than 100 analysts have ever been permitted to work with the data directly. However, a database prepared by the National Bureau of Economic Research, which conducted the final evaluation of MTO, will soon be available to qualified researchers at the U.S. Census Bureau’s Center for Administrative Records Research and Applications (CARRA), where it can be matched with a variety of government records stored there under strong security to ensure the personal privacy of MTO participants.

The goal of the Family Options Study was to understand which housing and services interventions work best for families with children experiencing homelessness and learn whether or not certain types of families benefit more or less from different interventions. Between 2010 and 2012, a total of 2,282 families in 12 communities nationwide, including more than 5,000 children, were recruited for the study after staying at least 1 week in emergency shelter. Families were randomly assigned to one of four possible interventions that differed in the duration of housing assistance provided or in the intensity of social services offered: (a) a long-term deep housing subsidy, usually in the form of a housing choice voucher, with no specialized services; (b) community based rapid re-housing, which is temporary rental assistance for up to 18 months with limited housing-related services; (c) project-based transitional housing, which is temporary housing for up to 24 months in an agency-controlled building coupled with intensive supportive services; and d) usual care, which was whatever mix of housing and services that homeless families could obtain on their own outside of the demonstration.

The information provided by and about the Family Options Study participants represents, to our knowledge, the largest and richest dataset on homeless families in the United States. The dataset also includes the most rigorous assessment of the impact of offering various interventions typically available within a community to families experiencing homelessness. Families were followed for 3 years after random assignment, with extensive surveys of families conducted at baseline and again approximately 20 and 37 months after random assignment. Although the primary outcome of interest was housing stability, additional outcomes of interest included family preservation, adult well-being, child well-being, and self-sufficiency. In addition to collecting data about the well-being of families and children at different points in time following random assignment, researchers collected extensive cost data on each of the interventions to calculate the fiscal costs to government of achieving the documented outcomes. Other valuable and unique data files were created over the course of the study, including detailed housing histories and family composition records for each household, both made possible due to frequent and regular communication with families over the full three years of study implementation.

Until now, access to the Family Options Study data has been limited to the original study team and less than half a dozen other analysts. HUD has transferred the full volume of data collected by Abt Associates over the course of the Family Options Study to the U.S. Census Bureau. These data will also become available to qualified researchers at CARRA in spring 2017.