Access to federal individual income tax return data is vital to the U.S. Census Bureau's income and poverty estimates. Each year, the Census Bureau receives selected items of information from all Form-1040 tax returns. The Census Bureau uses the street address on each tax return to assign a state and county code and often a school district code. Subsequently, the resulting "geocoded" data are aggregated to provide state- and county-level variables for use as predictors in the models and for school-district shares in the Minimum Change approach. These data are obtained and kept in the strictest confidentiality, consistent with the requirements in IRS Publication 1075, "Tax Information Security Guidelines For Federal, State and Local Agencies", as well as with the Census Bureau's confidentiality standards.

The total number of exemptions attributed to a return includes 1) the filer, 2) the spouse of the filer, and 3) the number of child exemptions for the household. If the adjusted gross income (AGI) on a return is below the official poverty threshold for a family of the size implied by the number of exemptions, all the exemptions (people) on the return are treated as poor (see Poverty Thresholds). AGI is not the same conceptually as family income, which the Census Bureau uses to determine official poverty status. We use exemptions for people aged 65 and over to make the distinction between people under 65 years old and people 65 and older in the matrix of poverty thresholds for one and two-person families. We do not use the related child delineation in the official poverty matrix. Instead, we compare AGI to the published weighted average threshold appropriate to the size of the family unit. Where we require age groups, we assume child exemptions to correspond to the population under 18 years, and we obtain exemptions for the 65 and over age group directly. We do not use returns which claim no exemptions. These are filed by persons reported as dependents on some other tax return. We also do not use estate returns, which correspond to people who are deceased.

Key variables constructed from these data include 1) the number of tax returns, 2) the number of tax returns filed by people below the relevant poverty threshold, 3) median and mean AGI on all returns from an area, 4) the total number of exemptions, i.e., the number of persons on tax returns, 5) the total number of poor exemptions, i.e., the number of persons on returns with AGI below the relevant poverty threshold, 6) the total number of child exemptions, 7) the number of poor child exemptions, 8) the number of exemptions for people age 65 and over, and 9) the number of poor exemptions for people age 65 and over.

Note that some variables refer to tax returns and some to people on tax returns. We use different variables for different models. For median family income, we use median AGI as a predictor. For the number poor in each age group, our county-level models use the number of poor exemptions (people), the number of child exemptions, and total number of exemptions. Our state-level poverty models use:

• the ratio of exemptions under age 65 on tax returns with AGI less than the poverty threshold (for a family of the size implied by the number of exemptions on the return) to the population under age 65;
• the ratio of child exemptions on returns with AGI below the poverty threshold (for a family of the size implied by the number of exemptions on the return) to the number of child exemptions on all returns;
• the ratio of exemptions for people age 65 and over on returns with AGI below the poverty threshold (for a family of the size implied by the number of exemptions on the return) to the population age 65 and over;
• the complement of the ratio of the number of exemptions under age 65 to the population under age 65; and
• the complement of the ratio of the number of exemptions age 65 and over to the population age 65 and over.