analyze the social security administration’s national beneficiary survey (nbs) with r

above and beyond its namesake roosevelt-era retirement safety net, the social security administration oversees two federal programs – social security disability insurance and supplemental security insurance.  currently covering more than ten million disabled americans (many far younger than retirement age), the quants in the woodlawn disability research office thought it might be smart to learn a little something something about the people who they serve.  so they went ahead and asked them: the national beneficiary survey interviews the populations covered by these two programs with one giant heap of questions.

this microdata is for the serious disability researcher, not the dabbler.  although some of the variables in this data should reasonably compare to questions fielded by other federal surveys, the instrument goes deep into the inner-workings of these public programs.  lemme say that another way: the current population survey (cps), a survey representative of the entire non-institutionalized united states population, includes a flag indicating ssi receipt as well as some disability payment values.  if you want to answer a broad question – “how does the self-reported health status of americans on disability compare to the self-reported health status of all other americans?” – use the two hundred thousand-respondent cps and not the two thousand-respondent nbs.  if you want to the answer to a pinpointed question – “what share of disabled americans currently on a physical therapy regiment are out of work because they fear that they will lose their benefits should they gain employment?” – you’re in the right place.  once again, for the serious disability researcher.  this new github repository contains three scripts:


download all microdata.R

• loop through and read.csv from all of the available rounds into working memory
• save each data.frame object to the local disk

analysis examples.R

• load an example single year of data
• construct a taylor-series linearization survey design
• rekindle your love affair with the r language.  i promise not to tell your partner

replication.R

• load the round four data file
• construct a taylor-series linearization survey design
• precisely match the percent and standard error generated from this sudaan block saved here

click here to view these three scripts



for more detail about the national beneficiary survey, visit:

• the social security administration’s one-paragraph description of this microdata (read it)
• the social security administration’s one-page description of this microdata (yes, read it too)

notes:

the social security administration’s public use files exclude about half of all sampled individuals, specifically those targeted because of their enrollment in the ticket-to-work (ttw) programs.  if you compare the unweighted record counts in the public use file to this table, you’ll hit the representative beneficiary sample row rather than the total.


confidential to sas, spss, stata, and sudaan users: your statistical languages of choice are about as huggable as a cactus.  time to transition to r.  😀