Brent D. Mast HUD

HUD currently has no administrative data to compare housing quality of public housing units to that of HCVP units. Census data allow neighborhood comparisons Subjective neighborhood ratings may not correlate well with census data. AHS are the only data available to compare housing quality and subjective neighborhood ratings.

Quality comparisons based on AHS data are problematic because the AHS over-represents Public Housing, and under-represents vouchers. The 2009 AHS sum of weights were approximately 1.65 million for either program. Approximately 1 million in public housing, and 2.1 million vouchers Many AHS households using vouchers respond that they live in public housing.

In 2011, the Census Bureau will begin verifying whether AHS households reporting assistance actually receive HUD assistance. However they will not check programs. Furthermore, the information will not be available in the public use file. HUD administrative data, however, are and excellent source of prior information for the expected proportion of households in public housing.

This study explores Bayesian methods for using prior information on variables such income and rents to estimate propensity scores for program participation. Results indicate that after adjusting for program participation propensities, there is little difference in household and neighborhood quality ratings between public housing and voucher households.

Longitudinal studies : follow families using vouchers to move out of public housing (Gautreaux, Moving to Opportunity). Cross Sectional: Newman and Schnare (1997) compare neighborhood quality using census tract measures such as the poverty rate and minority concentration. They find that compared to public housing residents, voucher households are less likely to be located in extremely high poverty neighborhoods.

Buron and Pantrabansh (2007) report that census measures do not correlate well with voucher households’ subjective opinions of their neighborhoods. AHS only data to compare subjective opinions of public housing and voucher households. In general, public housing residents tend to report assistance much more accurately than voucher households. Casey (1992) compared known HUD-assisted addresses to addresses of AHS respondents. 91% of public housing residents correctly identified their type of assistance. 33% of voucher households incorrectly identified themselves as public housing residents.

HUD’s PIC Data System: Most recent transaction at end of ,967,865 voucher households 1,032,239 public housing households. AHS 2009 national: households that self-report receiving voucher rental assistance (702) or living in public housing (720). Total of 1422 responses.

21 Categories based on income and rent burden. Incomes can be higher in public housing 75% of vouchers must go to families with incomes below 30% of the median. Rent burdens tend to be higher in voucher program. HCV Rent burden is supposed to be at least 30%. Flat rent option in public housing makes burdens below 30% possible.

Rent burden category% of HCV households% of PH households Missing10.776%11.642% 10% -19%0.000%6.280% 20% - 27%0.000%6.137% 28% - 31%58.196%72.717% 32% - 40%20.493%0.906% 41% & Above10.535%2.319%

For each category, the proportion in public housing is estimated. Prior proportion in public housing based on HUD administrative data. Sample proportion based on AHS data Bayesian posterior proportion is a weighted average of the prior and sample proportion. Prior proportion given weight 4/5, AHS proportion given weight 1/5.

Using propensity score weighting, the % of AHS assisted households in public housing is.352. The unadjusted AHS estimate is.504. According the HUD administrative data, the percentage is.345.

3 AHS variables: Home rating (scale of 1-10), Neighborhood rating (scale of 1-10), and crime question. 7 binary indicators 3 home rating indicators (>=7,>=8,>=9) 3 neighborhood rating indicators (>=7, >=8, >=9) Indicator for households that responded “yes” when asked if crime was a major problem.

Public housingHCVP VariableMeanStd errorMeanStd error H H H N N N Crime

Public housingHCVP VariableMeanStd errorMeanStd error H H H N N N Crime

Little difference in home ratings, either adjusted or unadjusted Adjusting for program participation 55.1% of HCV families rated their neighborhoods 8 or above,compared to 51.8% of public housing households. 35.9% of HCV households rated their neighborhoods 9 or greater, compared to 32.8% of public housing households.

Unadjusted, the crime indicator is considerably lower for public housing (.257) compared to HCV (.293). Adjusted by propensity scores, 28.2% of public housing households report a major crime problem, compared to 27.1% of HCV households.

UnadjustedAdjusted Variable Chi-square test statisticProbability value Chi-square test statisticProbability value H H H N N N Crime

Propensity scoring may improve reliability of AHS data on assisted households. HUD administrative data are an excellent source of prior information for computing Bayesian propensity scores. After adjusting for program participation, there is no statistical difference in home ratings, neighborhood ratings, or crime perceptions. Raises questions about program equity, because rent burdens tend to be much higher in the voucher program.