R-1: Home and Community Accessibility in the
American Housing Survey (AHS)
Sample, Data Collection and Measurement, Data Analysis
The 2009 AHS is the most comprehensive national survey of housing in the U.S. It includes two surveys collected annually: a national survey and a metropolitan area survey. In 2009, it included 62,000 housing units randomly selected across 394 Census Bureau primary sampling units that cover 878 counties and all 50 states. Each observation is weighted to most accurately represent all U.S. housing units. Of all respondents, 17.2% said their units housed someone with a disability. There were 9,591 unweighted observations for people who responded yes to any one of the disability questions.
Data Collection and Measurement
The AHS is a household survey collected using computer-assisted survey methods, either in person or by telephone. The Census Bureau has interviewed residents of the current sample of housing units since 1985. Disability is measured with six disability sequence questions that assess sensory, mental and physical impairments, as well as self-care and leaving the home limitations. Additionally, the dataset includes income source which will allow us to examine disability population by income source (e.g., SSDI).
Independent variables will include income, age, employment status, impairment status, disability status, and metropolitan/non metropolitan. Dependent variables will include home access (i.e., lack of barriers to entrance), availability of public transportation, personal vehicle available for use, and age of multifamily housing unit.
The AHS data and dictionaries are publicly available from the Department of Housing and Urban Development. Data will be imported into and analyzed using SPSS 19.0. Descriptive statistics for all variables will be computed to examine the data for uni- and multivariate normality. When data approximate normality, parametric inferential statistics for detecting sub-population differences and associations will be used (e.g., t-tests, multiple regression) and where normality assumptions are clearly violated, we will use non-parametric statistics (e.g., Mann-Whitney).