Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
DATABASES FOR MICROSIMULATION: A COMPARISON OF THE MARCH CPS AND SIPP 44 relationships for tax and transfer program modeling is documented in several studies. For example, Ruggles and Michel (1987) found that a marked drop in the simulated basic AFDC participation rate from 90 percent in 1980 to less than 80 percent in subsequent years was largely due to a seemingly small coding change instituted by the Census Bureau in the March CPS. The coding change added a million subfamilies, who had lower-than-average AFDC participation rates, to the eligible population. SIPP provides much more information with which to determine program filing units. The questionnaire explicitly asks about filing unit composition for major programs such as food stamps and AFDC. It is still not easy to identify potential filing units among households that are eligible but not currently participating in income support programs; nonetheless, the information on filing unit composition in SIPP represents a major step forward. DATA OMISSIONS This section discusses data omitted from the CPS and SIPP, including asset holdings and expenditures, that are necessary for accurate simulation of income support programs. It also discusses omitted data, including data on extended families and other variables, that are necessary for simulations that link income support programs with related programs such as child support enforcement, child care tax credits, and job training and employment programs. Again, the comparison favors the SIPP. Asset Holdings The CPS does not obtain data on asset holdings of households, although income support programs uniformly include some sort of asset test. The models address this problem in a number of ways, for example, by applying an estimated rate of return to reported interest and dividend income to estimate the value of a household's financial assets. SIPP, in contrast, includes a complete battery of asset and liability questions that is administered once to each panel (once a year to the early panels). SIPP also obtains data on income received from assets in each wave. As noted above, reporting of asset holdings is quite complete in the sense of high response rates; it is not so as to their value. The question on value of stock holdings, for example, elicits a 40 percent nonresponse rate. The longitudinal nature of SIPP also poses problems for relating the once-yearly asset portfolio information to changing household composition and income; however, the basic data are there. (The MATH model was recently altered to use SIPP data to impute financial and vehicular assets for simulations of the food stamp program.)