SLIDE 1 1
Working Poverty Risk Factors for U.S. Elders in the Great-Recession Era
Jo Mhairi Hale, Christian Dudel, & Angelo Lorenti
Max Planck Institute for Demographic Research
Objectives: Many more Americans experience working poverty than unemployed poverty, yet little research in the U.S. focuses on this group. The Great Recession only exacerbated working poverty, especially for older workers. We hypothesize older workers with lower human capital would be even more at risk. We test the hypothesis that those who experience greater life-course disadvantage (e.g., lower early-life SES, People of Color) will have a higher risk of working poverty and longer expectancy in working poverty after the Great Recession than those who are less disadvantaged. Method: Analyzing the U.S. Health and Retirement Survey, we use for causal mediation analysis to decompose the effect of early-life SES on risk of working poverty into the direct effect and the indirect effect that operates through the association between early-life SES and educational and
Results: Preliminary results indicate an increase in working poverty for those aged 50-70 at the time of the Recession, especially for POC. Causal mediation analysis suggests those in the lowest early-SES quartile compared to the higher quartiles were 22% more likely to experience working poverty, net of race/ethnicity/sex, education, occupation, and age. Only 35% of the association between early-SES and working poverty is mediated through education/occupation, evidence for
- ur hypothesis that early-SES has a long-lasting impact on working poverty. In general all
subgroups who originate in lower-SES are predicted to spend a greater percent of their working expectancy in working life poverty than their high-SES counterparts, net of education and
- ccupation. Women in the lowest three early-SES quartiles are estimated to spend significantly
longer share of their working life expectancy in working poverty during the Recession than before. Discussion: When people experience economic hardship (e.g., the Great Recession), lack of familial wealth and low human/social capital places people who originate in lower socioeconomic statuses, especially People of Color (POC) (Bloome 2014; Conley 1999; Neckerman and Torche 2007; Western et al. 2012), in more precarious economic positions.
SLIDE 2 2
Working Poverty Risk Factors for U.S. Elders in the Great-Recession Era
Although most people transition in and out of poverty spells, a small subset of Americans are chronically poor (Magnuson and Votruba-Drzal 2009; Stevens 2012) despite working (Brady et al. 2010). In the U.S., there are four times more people in working poverty than in poor households where no one is working (Brady et al. 2010). Older displaced workers experienced the largest earnings declines in the wake of the Great Recession (Farber 2011). Researchers have paid less attention to these working poor (Burgard and Kalousova 2015; Crettaz 2013) even though they represent an increasing share of the U.S. population and are almost a hallmark of the Great Recession era (Brady et al. 2010). Lee and colleagues (2005) use earlier waves of the Health and Retirement Survey to develop a sociodemographic profile of middle-aged working poor, defined as aged 51-61. However, they consider primarily contemporary predictors, e.g., age, sex, race, marital status, and net worth, only including respondent’s own education as an early-life predictor, and, of course, the Great Recession had not yet become a factor. We draw on three main bodies of literature to motivate our study: social mobility, intersectionality, and life course studies. Research on social mobility and the intergenerational transmission of poverty ranges from analyses of individual-level data to determine which factors are most predictive of life outcomes to cross-national studies of social mobility (Blau and Duncan 1967; Hout and DiPrete 2006; Sewell et al. 1970). Intersectionality theory reminds us that in studying socioeconomic status (SES), race/ethnicity, and sex, we must always consider their interactive effects (Collins 2015). Cumulative advantage/disadvantage and cumulative inequality theories predict that disadvantage accumulates over the life course (Ferraro et al. 2016; Schafer et
- al. 2011; Willson et al. 2007). Despite these extensive bodies of research, less work has focused
SLIDE 3 3
- n the juncture of these fields, and, to our knowledge, no one has studied them in the context of
working poverty. Read together, we can derive from these theories that a lack of familial wealth and human/social capital places people who originate in lower socioeconomic statuses, especially People of Color (POC) (Bloome 2014; Conley 1999; Neckerman and Torche 2007; Western et al. 2012), in more precarious positions when disaster strikes, whether personal (a health calamity or divorce) or during macroeconomic shock (Great Recession). We take a life-course approach in hypothesizing that early-life socioeconomic status (SES) will intersect with race/ethnicity/sex and the Great Recession to predict both higher risk of mid- to later-life working poverty and also a longer share of working life expectancy spent working poor.
Methods
To study the risk factors for mid- to later-life working poverty, we use the U.S. Health and Retirement Survey, a nationally-representative, population-based, biennial, longitudinal survey (n≈30,000). We include those aged 50 to 70 during the years 2002-2014 to focus on the pre- and Great Recession effects and reduce problems with cohort confounding1 (n=19,426). Working Poverty. Since the ambiguous measurement of working poverty has resulted in conflicting findings, we clearly define our terms here: 1) The sample is restricted to those who self-report working full-time or part-time (but not unemployed), and who report working at least 27 weeks of the previous year2; 2) We use the household as the unit of analysis; 3) Household income includes before-tax income from earnings, unemployment, social security, social security
1 Wave 6 (2002) also marks the first wave at which RAND calculates the ratio of income to poverty threshold. 2 Taking into consideration results from Thiede et al.’s (2015) analysis, in alternative analyses, we define working
poor as above, but not restricted by reports of hours or weeks worked. We also restrict to those who report averaging at least 17 hours/week. Results are consistent.
SLIDE 4 4 insurance, pensions, but not non-cash benefits or capital gains (see RAND documentation for more details); 4) RAND uses the Bureau of Labor Statistics’ definition of the poverty threshold, adjusted for household composition; 5) For those who meet the definition of working, we count as ―poor‖ those whose household income is below 200% of the Federal Poverty Line (FPL)3. We take into consideration average wealth and debt over the study period (1992-2014) to account for elders who have, for example, less income, but greater housing wealth.
- Predictors. Social factors include quartiles of early-life SES, which is a composite
variable that includes parents’ education, whether the family required financial help or had to move due to financial difficulty, whether the father contributed financially to the household, self- report of family’s SES, and childhood health (eigenvalue=2). Education is a categorical variable (less than high school, high school diploma/GED, some college, or college +). We use longest
- ccupation as a binary ―Office/Professional‖ versus ―Manual, Service, Farms/Forestry/Fishing,
Construction, Extraction.‖ We create a categorical variable for race/ethnicity and sex (male or female – non-Hispanic White, Black, and non-Black Latinx). We define pre-Recession as 2002- 2008 because although the Great Recession technically began in 2007 with the most critical crisis in the last quarter of 2008, income and financial assets listed in the HRS is for the previous year. In the first part of the analysis, age is measured as age in years as of 2002. In the second part of the analysis, age is measured as age in years. Analytic Strategy: In the first stage of the analysis, we test Hypothesis 1 that individuals who were disadvantaged in early life will have higher risk of working poverty, net of educational and occupational attainment. When studying the association between early-life factors and
- utcomes later in life, intermediate confounding becomes particularly problematic. Early-life SES
3 The FPL is widely accepted to be an inaccurate measure of hardship because it relies on an outdated formula
(Thiede et al. 2015). We also examine those 150% above the FPL; results are consistent.
SLIDE 5
5 affects educational and occupational attainment, which also affects risk of working poverty. To determine whether there is also a direct effect of early-life SES on later-life working poverty, we use a counterfactual framework—one method through which we can reduce this source of bias (cf. Nandi et al. 2012; VanderWeele 2015; VanderWeele and Robinson 2015). Therefore, we first use causal mediation analysis to decompose the effect of early-life SES on risk of working poverty into the direct effect and the indirect effect that operates through the association between early-life SES and educational and occupational attainment. We perform the decomposition for the whole period, and then split the period into 2002-2008 and 2008-2014. We next test hypothesis 2 that individuals aged 50-64 who have more life course disadvantage will spend a greater share of their working lifetime in working poverty than their more advantaged counterparts. To calculate the remaining lifetime spent in working poverty and the remaining lifetime spent in unemployment, we use a multistate approach (Hoem 1977). The multistate approach allows us to model transitions between labor force states and is based on the probabilities of transitioning from one state to another; e.g., the probability of moving from regular employment to working poverty. Given estimates of transition probabilities, many useful quantities can be calculated, including the remaining lifetime in a state. We use a multistate model with four states: regular employment (full time or part time but not poor); working poverty; unemployment; and one category capturing retirement, disability, or otherwise not being in the labor force. In addition to transitions between labor force states the model also captures mortality. To estimate the transition probabilities between states we use discrete event history analysis (Allison 1982). We let transition probabilities depend on age, gender, period (2002-2008 vs. 2009-2014), early-life SES, and either education or race/ethnicity. This allows us, for instance, to
SLIDE 6 6 estimate the remaining lifetime in working poverty of Latinas before the recession hit, and to compare it to post-recession estimates. Preliminary Results: In descriptive Figure 1 we can see there was, indeed, an increase in poverty, unemployment, and working poverty for those aged 50-70 at the time of the GR, especially for
- POC. Differences by five-year age categories (not pictured) are in levels, not in impact (e.g., the
eldest have lowest risk of working poverty, but all age categories experience similar spikes). Figure 1 Proportion of poverty, unemployment, and working poverty by race/ethnicity/sex (ages 50-70) We next use causal mediation analysis to decompose the marginal total effect (2008-2014: OR 1.33, p<.001) into natural direct effects of early-SES (2008-2014: OR 1.22, p<.001) and natural indirect effects through education/occupational attainment (2008-2014: OR 1.09, p<.001). In the Recession era, we find that those in the lowest early-SES quartile compared to the higher quartiles were 22% more likely to experience working poverty, net of race/ethnicity/sex,
SLIDE 7
7 education, occupation, and age (Figure 3). 35 % of the association between early-SES and working poverty is mediated through education/occupation, evidence for our hypothesis that early-SES has a long-lasting impact on precarity. The pre-Recession era shows the same pattern, but results for the controlled direct effect and the natural direct effect are non-significant. provides some evidence for our hypothesis of differences before and after the Recession, but the differences are not strong. Figure 2. Effect decomposition for the exposure (early-SES) and the mediator (educational/occupational attainment) on risk of working poverty pre- and post-Great Recession Finally, we estimate the remaining lifetime in working poverty by early-SES and then race/ethnicity/sex. Women in the lowest three early-SES quartiles are estimated to spend significantly longer in working poverty (relative to their working life expectancy) during the Recession than before. The other differences are non-significant. 2002-2008 2009-2014
SLIDE 8 8 Figure 3 Difference in working poverty before and after great recession by sex & early-ses (net education) Over the whole period, almost all subgroups who originate in lower-SES are predicted to spend a greater share of their working lifetime in working poverty than their high-SES
- counterparts. Low-SES Latinos before the Recession are an exception. However, Latinas from
low-SES families spent the longest in working poverty relative to their work expectancy.
5 10
Males Females
Lowest Early-SES 2nd Q. 3rd Q. Highest Early-SES
Graphs by sex
SLIDE 9
9 Figure 4 Expected share of working life expectancy spent in working poverty by race/ethnicity, sex, and early-SES (net education and occupation)
Conclusion
In conclusion, preliminary results show early-life SES interacts with race/ethnicity and sex to affect risk of and expected time spent in working poverty later in life. The Great Recession increased low-SES elders and POC elders’ risk of working poverty. As in the UK (Giesselmann 2015), at-will employment may put older workers at particular risk of working poverty in their final years of employability. Next steps in the analysis will be to investigate multiple mediators and competing risks.
10 20 30 40 White Male White Fem Black Male Black Fem Latino Latina Pre-GR Low SES GR Low SES Pre-GR Higher SES GR Higher SES
SLIDE 10 10 References Allison, P. D. (1982). Discrete-Time Methods for the Analysis of Event Histories Author ( s ): Paul D . Allison Published by : Wiley Stable URL : http://www.jstor.org/stable/270718 Accessed : 20-05-2016 10 : 10 UTC DISCRETE-TIME METHODS FOR THE ANALYSIS OF EVENT HISTORIES. Sociological Methodology, 13(1982), 61–98. Blau, P. M., & Duncan, O. D. (1967). The American occupational structure. Bloome, D. (2014). Racial Inequality Trends and the Intergenerational Persistence of Income and Family Structure. American Sociological Review, 79(6), 1196–1225. doi:10.1177/0003122414554947 Brady, D., Fullerton, A. S., & Cross, J. M. (2010). More than just nickels and dimes: a cross- national analysis of working poverty in affluent democracies. Social problems, 57(4), 559–
- 585. doi:10.1525/sp.2010.57.4.559
Burgard, S. A., & Kalousova, L. (2015). Effects of the Great Recession: Health and Well-Being. Annual Review of Sociology, 41(1). doi:10.1146/annurev-soc-073014-112204 Collins, P. H. (2015). Intersectionality’s Definitional Dilemmas. Annual Review of Sociology, 41, 1–20. doi:10.1146/annurev-soc-073014-112142 Conley, D. (1999). Being black, living in the red: Race, wealth, and social policy in America. Univ of California Press. Crettaz, E. (2013). A state-of-the-art review of working poverty in advanced economies: theoretical models, measurement issues and risk groups. Journal of European Social Policy, 23(4), 347–362. doi:10.1177/0958928713507470 Farber, H. S. (2011). Job loss in the Great Recession: Historical perspective from the displaced workers survey, 1984-2010. Nber Working Paper. http://www.nber.org/papers/w17040.pdf
SLIDE 11
11 Ferraro, K. F., Schafer, M. H., & Wilkinson, L. R. (2016). Childhood Disadvantage and Health Problems in Middle and Later Life: Early Imprints on Physical Health? American Sociological Review, 81(1), 107–133. doi:10.1177/0003122415619617 Giesselmann, M. (2015). Differences in the Patterns of in-work Poverty in Germany and the UK. European Societies, 17(1), 27–46. doi:10.1080/14616696.2014.968796 Hoem, J. M. (1977). A Markov chain model of working life tables. Scandinavian Actuarial Journal, 1977(1), 1–20. doi:10.1080/03461238.1977.10405621 Hout, M., & DiPrete, T. A. (2006). What we have learned: RC28’s contributions to knowledge about social stratification. Research in Social Stratification and Mobility, 24(1), 1–20. doi:10.1016/j.rssm.2005.10.001 Magnuson, K., & Votruba-Drzal, E. (2009). Enduring influences of childhood poverty. Focus, 26(2), 32–37. Nandi, A., Glymour, M. M., Kawachi, I., & VanderWeele, T. J. (2012). Using Marginal Structural Models to Estimate the Direct Effect of Adverse Childhood Social Conditions on Onset of Heart Disease, Diabetes, and Stroke. Epidemiology, 23(2), 223–232. doi:10.1097/EDE.0b013e31824570bd.Using Neckerman, K. M., & Torche, F. (2007). Inequality: Causes Consequences. Annual Review of Sociology, 33, 335–357. Schafer, M. H., Ferraro, K. F., & Mustillo, S. A. (2011). Children of misfortune: early adversity and cumulative inequality in perceived life trajectories. American Journal of Sociology, 116(4), 1053–1091. doi:10.1086/655760 Sewell, W. H., Haller, A., & Ohlendorf, G. W. (1970). The Educational and Early Occupational Status Attainment Process: Replication and Revision. American Sociological Review, 35(6),
SLIDE 12 12 1014–1027. Stevens, A. H. (2012). Chapter 15 Poverty Transitions. In The Oxford handbook of the economics
Thiede, B. C., Lichter, D. T., & Sanders, S. R. (2015). America’s Working Poor: Conceptualization, Measurement, and New Estimates. Work and Occupations, 42(3), 267–
- 312. doi:10.1177/0730888415573635
VanderWeele, T. J. (2015). Explanation in causal inference: methods for mediation and
- interaction. Oxford University Press.
VanderWeele, T. J., & Robinson, W. R. (2015). Confounding and Mediating Variables, 25(4), 473–484. doi:10.1097/EDE.0000000000000105.On Western, B., Bloome, D., Sosnaud, B., & Tach, L. (2012). Economic Insecurity and Social
- Stratification. Annual Review of Sociology, 38, 341–359. doi:10.1146/annurev-soc-071811-
145434 Willson, A. E., Shuey, K. M., & Elder Jr., G. H. (2007). Cumulative advantage processes as mechanisms of inequality in life course health. American Journal of Sociology, 112(6), 1886–1924. doi:10.1086/512712
