Impact of changes in U.S. economic and demographic composition on the poverty rate, selected periods, 1979–2010
|Actual change in poverty rate||1.2||-1.5||1.2||0.8||2.6|
|Change due to demographic shifts||-0.2||-0.6||-0.1||-0.8||0.3|
|Change due to economic shifts||1.4||-0.9||1.2||1.7||2.3|
Note: Data may not sum due to rounding and the effect of the interactions between demographic characteristics, explained in the table note.
Source: Authors' analysis of Current Population Survey Annual Social and Economic Supplement microdata. Analysis based on Danziger and Gottschalk (1995)
Documentation and methodology
Underlying data are from Current Population Survey Annual Social and Economic Supplement microdata; see Appendix A for details. The methodology for this decomposition is taken from Danziger and Gottschalk (1995, Chapter 5), which explores the role of changes in socioeconomic characteristics (e.g., changes in average income, changes in income inequality, and demographic changes such as the change in racial groups’ shares of the overall population) on the poverty rate (using the official poverty rate) between any two years. We focus specifically on the 1979–1989, 1989–2000, 2000–2007, 2007–2010, and 1979–2007 periods. To examine the impact of average income of the U.S. population on the poverty rate, we assign the average real income growth across the period to be the growth for all individuals between years t0 and t1 and simulate a new poverty rate. This procedure holds the shape of the distribution (inequality) constant in t0 while allowing incomes to grow equally for all individuals. This simulated poverty rate for t1 is then compared to the actual poverty rate in year t0, and the percentage-point difference is the change in the mean, i.e. the impact of income growth. The change due to income inequality is the percentage-point difference between the simulated poverty distribution in t1 and the actual poverty rate in t1.
We repeat this exercise using the demographic composition of each variable of interest to see the effect of these demographic changes on the overall poverty rate. First we calculate the weight of each demographic factor (such as individuals with college degrees) by its population share and simulate the poverty rate in t1 for all persons between t0 and t1, allowing for income to grow equally among all families and holding the demographic composition of the population in t0 constant. Then, we calculate a second simulated rate that incorporates both the mean income growth and the demographic changes across the period. The difference between these two simulated rates in t1 is the percentage point-change in the poverty rate due to demographic changes.
The interaction, or error, term states to what degree the demographic variables are conflated, which could lead to bias in measurement of a factor’s impact. Since our error term is negative and relatively small (-.4 from 1979–2007), the reported relationship might slightly overstate the degree to which the simulated income decreases the poverty rate for each demographic group, but it is not enough to change the story.