2014 3/8 ASEC File Information
The Census Bureau fielded the CPS 2014 ASEC sample using an experimental redesign. All respondents received new health insurance questions, but 3/8ths of the total sample was randomly selected to receive the redesigned income questions. The larger portion of the sample (5/8) was given the existing questions on income. The redesign attempted to address income under-reporting, in particular, retirement, pensions, annuities, and government cash-transfer programs. More accurate income reporting in turn allows for better measurement of poverty statistics. The Census Bureau prepared this version of the income questions through testing in the March 2013 Content Test. In the March 2013 Content Test, respondents received both previous and redesigned income questions for comparison. The preliminary results showed that the previous design captured less sources of income than the redesigned version. Please see the following documentation regarding the redesign.
Previously, an income screen allowed only households with combined earnings of less than $75,000 to answer questions regarding cash-transfer programs such as TANF and SNAP. For the 3/8 sample, that screen has been removed.
The experimental questions regarding asset-based income were clarified to avoid the under-reporting of income such as the interest earned from retirement accounts. The redesigned questions now differentiate between retirement accounts and pension income. There are now specific questions relating to each possible source of asset-based income, such as interest, dividends, and capital gains, where previous surveys did not ask about capital gains income. If a respondent does not know or refuses to answer, there is a follow-up question on the value of an interest-earning asset, and an estimate of the income can be calculated.
The question about disability income was separated into individual questions asking about the different sources or forms of disability pay.
A dual pass approach was introduced so that first, the household is asked whether or not a certain type of income was received for all possible sources of income. Then, the household is questioned on the amount of income received from each source they reported receiving. These changes did not extend to job, unemployment, or worker's compensation income questions.
The redesign rearranged the order of questions to prioritize questions that are more relevant to the respondent based on household composition. Specifically, low income households are asked questions about public assistance and food stamps relatively early, seniors are asked questions about disability and retirement income relatively early, and all other households receive the income questions in the existing order. All households are still asked all the questions, but in different orders.
In order to minimize nonresponse for income questions, enumerators now follow up "don't know" or "refused" responses by asking the respondent for the range in which the income amount falls. The ranges are high, medium, and low, and the values of the bracketed ranges are specific to each income source.
The Census Bureau has released a combined income file which contains all 2014 ASEC respondents, both from the 3/8 and 5/8 files, were certain income variables were imputed for the 5/8 sample respondents to give them responses consistent with the redesigned ASEC. This file also contains a combined supplement weight.
Health Insurance Questions
According to some qualitative research, the current system of asking health insurance questions based on calendar year provides inaccurate coverage estimates. The redesigned ASEC now asks questions on past calendar year and current (point-in-time) health care coverage, health insurance exchanges, and employer-sponsored insurance offers and take-up.
Because the 5/8 file and the 3/8 file are not completely comparable, WTSUPP values have been assigned so that either file is individually representative of the entire US population. If both parts of the file are analyzed together and weighted using WTSUPP, the resulting estimates will be twice the US population. IPUMS-CPS provides a flag variable, HFLAG, for users to identify which respondents were originally in the 3/8 file.