This study used as its primary data source the 2001 SIPP, a large, multipanel, longitudinal survey that collected demographic and socioeconomic information on a nationally representative sample of U.S. households. The data cover the period from late 2000 through 2003. SIPP provides detailed monthly measures on labor force participation (for those age 15 and older), income, participation in public programs, and household composition.(1) This section provides a brief overview of some of the key data and sample decisions made for this study; the Appendix provides a more detailed description of these decisions.
Selecting the Poverty Measure. We used the official U.S. Census Bureau poverty measure as the primary one for the study. This measure has several well-documented shortcomings, but it is the one most commonly used in research examining poverty-related issues, and it offers a straightforward and easily understood method for gauging poverty. The use of the official poverty measure also facilitates comparisons of our study results with those of previous research.
Under the official measure, a family is "poor" if its total family income is less than its money income threshold (based on the Office of Management and Budget's Statistical Policy Directive 14). Money income includes earnings; cash assistance (such as TANF benefits, unemployment compensation, or Supplemental Security Income [SSI]); child support; educational assistance; pension income; and interest and dividends. Income does not include non-cash benefits such as food stamps; Medicaid; public housing subsidies; and the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC).(2) There are 48 possible poverty thresholds that vary according to the size of the family and the ages of its members. These thresholds do not vary geographically, but they are updated annually for inflation using the Consumer Price Index for All Urban Consumers (CPI U).
We measured poverty at the family level largely because the family is the basis for the official definition of poverty. A SIPP family is a group of two or more persons related by birth, marriage, or adoption and living together; a SIPP family does not include cohabiters. We also considered using the subfamily — a nuclear family unit — rather than the family in defining poverty and found that this change made little difference in the size of our sample (see Appendix).
Analysis Sample and Unit of Analysis. The primary sample for the analysis consists of single mothers who, during the first 12 months of the panel period (roughly 2001), were in poverty in one month and exited poverty during the following month. If a sample member had more than one poverty exit during the one-year window, we used the first exit for the analysis. We defined single mothers as those women who, during the month before spell exit, were (1) older than 15 (so that employment information is available), (2) living with a related child younger than 18, and (3) unmarried. We included single mothers who were cohabiting with a partner, whose spouses were absent, or who were in school during the panel period.(3) This definition of a "single mother" included all single female family heads with related children under 18 in the family, rather than only single female family heads living with their own children under age 18. Thus, for instance, our sample included a small number of single grandmothers living with their grandchildren, and single aunts living with their nieces and nephews. Overall, we have a sample of 615 single mothers who exited poverty during the first 12 months of the panel period.
A key decision for the analysis was how to deal with changes in family composition. For example, a woman might move out of the family she is living in and move in with others, or other people might join her family. In our analysis, we treated the single mother as the unit of analysis and tracked her poverty status in whatever family she was in during any given month using the income of that family, even if its members changed from month to month. Thus, our analysis focused on obtaining estimates of, for example, the duration of non-poverty spells of single mothers who exited poverty rather than the duration of non-poverty spells of the families that contained these women.
Defining Poverty and Non-Poverty Spells. Another key decision for the study was how to define poverty spells. In data sets that try to capture monthly income, "transitional" poverty spells are sometimes observed — that is, in a given month, a family can have a sudden income dip due either to real changes in income or to measurement error. Such dips can lead to more people being "ever" poor than "consistently" poor. It is also possible that for those with incomes near the poverty threshold, small deviations in income could lead to frequent changes in poverty status. We conducted analyses to determine the extent to which such churning is present in our data and what steps could be taken to measure the "true" income volatility of single mothers.
We observed a high degree of poverty churning in the sample and were concerned about "noise" in the monthly poverty timelines. Thus, we explored several strategies to smooth poverty spells: (1) doing nothing, (2) closing "near-threshold" spells, (3) closing one-month spells, and (4) using a three-month moving average of family income. Based on these explorations, we elected to smooth volatility by closing all near-threshold spells, where a near-threshold spell is defined as a spell in which income is within 10 percent of the poverty threshold for the duration of the spell. Our main results, however, are not sensitive to the income-smoothing strategy.