An alternative approach is to use the data from two or more surveys to produce a combined estimate. Unbiased estimates can be produced for each state from the CPS and NHIS. State weights are being produced for SIPP that will hopefully have minimal bias. These can be combined to produce a single estimator. While there are a number of methods for producing such a combined estimator, the most logical procedure is to weight the three estimators in inverse proportion to their mean square errors. This gives greatest strength to the estimate from the survey with the most precise estimate for that state.
For specific characteristics of interest, there may be surveys other than the CPS, SIPP, and NHIS that collect the desired information. For example, Westat is currently collecting the National Survey of America's Families for the Urban Institute in 14 states and the remainder of the country. This study will be redone at the end of the decade to measure the change with devolution of programs to the states.
When combining data from multiple surveys it is very important to examine nonsampling errors. Data from one survey may not be asked in quite the same manner as another survey, or may only be asked of a subset of the population. For example, definitions of disability are not exactly consistent across the three surveys (Kalton and Mohadjer, 1994). The sequence in which questions are asked can also affect the survey estimates. Also, work disability questions are asked of all elderly on the CPS, but only for those under age 70 on SIPP. Income definitions can vary dramatically from one survey to another, for example, by whether or not, and how, they attempt to include non-cash income. The CPS asks for income for an entire year, while SIPP combines reported income from multiple interviews covering one year. On the NHIS, many questions are only asked of a single member of the household, while others are asked for all household members. Each survey has its own rules regarding proxy respondents as well. If proxy responses are allowed the response rates will be higher, but an additional possible source of bias is introduced.