Older Workers: Who are the working poor in the U.S.?

Abstract

With data from the 2000 Health and Retirement Study (HRS), the purpose of this study was to provide a profile of older workers who live poverty, and to compare the demographic, financial, employment, and health attributes of such individuals to similar persons not living in poverty. This study found that 3.5% of employed individuals between the ages of 51 and 61 belonged to the class of working poor. The results of the multivariate logistic regression analyses indicated that the older working poor were more likely to be non-White, less educated, non-married, and had lower levels of net worth than the working non-poor. They were more likely to be employed part time and were less likely to be covered by employee-sponsored health insurance.

Introduction & Review of the Literature

Recent evidence suggests that working individuals represent a rapidly expanding segment of the impoverished population (Kim, 1998a, 1998b). In 1999, the working poor numbered 6.8 million in the United States. While demonstrating reasonably active labor force participation—by most definitions, the working poor are employed more than 6 months in a given year—such individuals frequently live in households at or near subsistence levels (Cormier & Craypo, 2000; Davis & Huston, 1991; Klein & Rones, 1989; Mosisa, 2003). As chronic financial deprivation has been linked to a vast number of adverse economic, health, and social outcomes, the rising number of financially dispossessed workers is of great concern to government leaders and policy makers, who recognize the potential costs and obligations of public services.

In general, the economic standing of the working poor is one of instability and insecurity. As Marger (1999) describes, the working poor are caught between the abject poor, below, for whom various forms of government assistance are available, and the economically secure, above. While routinely employed, the working poor frequently work in part-time or temporary jobs, often earning subsistence wages with limited employment benefits, such as health insurance and pension plans; their savings are scare. Yet, despite such economic hardship, the working poor often do not qualify for means-tested government assistance programs intended to reduce the burdens of poverty, leaving them especially vulnerable to injury or unexpected illness (Marger, 1999). Although some research suggests that working poverty may be transient, the transition to economic stability is a monumental challenge for many individuals living in poverty, especially in light of the disproportionately high prevalence of health limitations and substance abuse (Zagorsky, 1999).

The aging of the workforce, as well as legislative changes that result in delayed receipt of government retirement benefits, will likely mean that both the number and proportion of impoverished older workers will grow. Following a decades-long trend of declining labor force participation (Hurd, 1990), resulting from changes in retirement policy, occupational structure, and business organization (DeViney & O'Rand, 1988), older individuals are working later in life; and this trend is expected to continue (Schulz, 2001). Demographic projections suggest that individuals over 55 will comprise 20% of all workers in 2020, compared with just 13% in 2000 (Toossi, 2004). The plight of older workers living in poverty merits particular attention, as economic deprivation in a period of escalating risk of chronic disease and important wealth accumulation could carry substantial health and financial consequences.

A great deal of attention has been given to the negative consequences of poverty. For decades, researchers have investigated the association between economic inequality and a variety of outcomes, predominantly physical and affective well-being. Numerous investigations have documented a link between poverty and somatic health later in life. Studies have demonstrated that members of lower socioeconomic strata have lower self-rated health, higher levels of functional impairment, and greater prevalence of morbidity at mid-life and early late-life than members of higher strata (Hay, 1988; Hayward, Miles, Crimmins, & Yang, 2000; House et al., 1990; House et al., 1994; Lantz et al., 2001; von dem Knesebeck, Luschen, Cockerham, & Siegrist, 2003).

Researchers have moreover provided fairly consistent findings, despite the use of such varied measures of socioeconomic standing as income, education, occupation, and employment attributes. Growing evidence further indicates that hazardous health behaviors, more prevalent in lower socioeconomic groups, have explained only a small part of the differences observed in prevalent health between the indigent and non-poor at middle age, suggesting that cumulative socioeconomic inequality may underlie adverse health outcomes among the disadvantaged. Long-term follow-up studies have supported this notion. For example, one inquiry (Lynch, Kaplan, & Shema, 1997) found that persons who experienced economic hardship in 1965, 1974 and 1983 were significantly more likely to have difficulties with activities of daily living in 1994. A more recent study (Power, Manor, & Matthews, 1999) further highlighted the importance of exposure duration to socioeconomic conditions for adult health. Research (Blackwell, Hayward, & Crimmins, 2001) has also suggested that childhood poverty influences health in later adult life.

A range of investigations has further indicated that economic inequality poses a threat to mental health. Published research has suggested that poverty may influence an array of psychological outcomes, including the prevalence of major psychiatric disorders, major depression, and post-traumatic stress (O'Campo, Eatoner, & Muntaner, 2004; Weich, Churchill, Lewis, & Mann, 1997). One study (Weich & Lewis, 1998), which linked common mental disorders directly with poor material standard of living, suggested that deficiencies in material standard of living could account for nearly one quarter of prevalent cases of common mental disorders in the Great Britain.

Still other studies have explored the effect of psychosocial workplace conditions on health. Job insecurity, common to the working poor, has been negatively associated with self-reported health and mental well-being (Ferrie, Shipley, Stansfeld, & Marmot, 2002; Virtanen, Vahtera, Kivimaki, Pentti, & Ferrie, 2002). Similarly, job conditions common to the working poor, such as job strain, low occupational pride, low stimulation, and inadequate social support, have been related to poor mental health among women (Virtanen et al., 2002). Cheng, Kawachi, Coakley, Schwartz, & Colditz (2000) more broadly found that low levels of employment-based social support were associated with poor health status. A related inquiry linked job satisfaction with health problems for both men and women nurses (Walters et al., 1996).

While health has been the prevailing focus of research on income inequality among middle aged individuals, other related literatures are relevant to our investigation. A considerable corpus of work has, for example, explored matters related to the labor force participation and withdrawal of older workers. In recent decades, older workers’ labor supply has shifted from manual and manufacturing jobs to non-manual employment such as services and finances (DeViney & O'Rand, 1988). Sectorial reallocation also has spelled increased demand for older women, particularly in such white-collar fields as finance, insurance, and real estate. More intensive participation in non-manual employment is anticipated for succeeding cohorts of older workers, as better socioeconomic standing, via higher educational attainment, will mean that future generations of older workers will have a broader range of occupational opportunities (Henretta & Lee, 1996; United States General Accounting Office, 2001).

Research has also suggested that retirement patterns, which have been described as an increasingly complex process of labor force withdrawal (Henretta, 1997), are influenced by socioeconomic factors that define the impoverished, such as pension wealth, health insurance, (Pienta & Hayward, 2002), standard of living (Mutchler, Burr, Pienta, & Massagli, 1997), occupation, education, race, and marital status (Hayward & Grady, 1990). One striking finding is that employment in service occupations, common among the poor, is negatively associated with retirement (Hayward & Grady, 1990), a conspicuous exception to the general trend for blue-collar workers (Henretta & O'Rand, 1983).

In this exploratory, cross-sectional study, we investigate working poverty among individuals between the ages of 51 and 61, workers who are within a decade of early eligibility for Social Security benefits. We use the 2000 wave of the Health and Retirement Study (HRS) to explore 2 objectives: 1) to provide a profile of older workers living in poverty; and 2) to compare socioeconomic, financial, labor and health attributes between the older working poor and older working non-poor.

Methods

Data Source

We use data from the 2000 wave of the HRS, the most recent data available at the time that this research was undertaken. The HRS is a nationally representative, longitudinal survey of individuals over 50 years of age, designed to investigate the dynamic experience of older individuals as they advance from work to retirement, with particular emphasis on health insurance, saving, and trajectories of economic and physical well-being.

In 2000, the HRS included 4 cohorts. The initial HRS cohort, first surveyed in 1992, comprises individuals born between 1931 and 1941; the Asset and Health Dynamics of the Oldest Old (AHEAD) cohort, begun in 1993, consists of individuals born before 1924; the Children of the Depression (CODA) cohort, added in 1998, includes participants born between 1924 and 1930; and the War Babies (WB) cohort, also added in 1998, is represented by individuals born between 1942 and 1947. HRS data are collected from age-eligible individuals and their spouses at two-year intervals.

In 2000, a total of 19,581 individuals participated in the HRS. Blacks, Hispanics, and Florida residents were over-sampled. The HRS is administered by the Institute for Social Research, at the University of Michigan, and is principally funded by the National Institute on Aging. The HRS is described in greater detail elsewhere (Juster & Suzman, 1995).

Study Sample

The eligible sample comprised HRS respondents aged 51 to 61 who, at the 2000 survey, reported working 27 or more weeks per year. Of these individuals, we retained those respondents who contributed data for all relevant variables. Participants with missing values in one or more study variables were eliminated, leaving a final study sample numbering 3,416, including 121 working poor and 3,295 working non-poor. The great majority of our sample members were selected from the HRS (66%) and War Babies (29%) cohorts; the remaining 5% were chosen from the AHEAD and CODA cohorts.

Defining Working Poverty

Working poverty is represented by a dichotomous dummy variable. Working poor status was determined by comparing the sum of household income items, excluding capital gains and losses, to relevant poverty threshold values established by the U.S. Census Bureau, and published in the Current Population Survey (United States Census Bureau, 2003). For one-person households, we compared HRS household income amounts to age-specific, weighted-average poverty thresholds for single individuals; for households of two or more individuals, we compared the HRS income amounts to the corresponding two-person household thresholds, as the HRS summarized household income exclusively for the respondent and spouse dyad. If the HRS household income was less than or equal to the Census Bureau threshold, working poverty was coded 1; if household income exceeded the threshold, the variable was coded 0.

Independent Variables

Independent variables represent four domains: demographic, economic, employment, and health. Univariate statistics, with variable measurement, are presented in Table 1. Demographic characteristics include age, gender, race, educational attainment, and marital status. The economic/financial domain is represented by non-housing wealth, which includes 9 asset types. Employment characteristics comprise occupation, annual weeks worked, job tenure, employment status, and employer-covered health insurance. Health characteristics include self-rated health, affective health (CES-D), and out-of-pocket healthcare expenses. Further detail may be found in Table 1.

Table 1.

Sample Characteristics and Variable Measurement of Older Workers (N=3,416)

Variable	%	Mean (Median)	Measurement
Demographic Domain
Age		56.6 (57.0)	Age in years
Female	54.1%		1 if female, 0 if male
White	81.8%		1 if white, 0 if non-white
Educational attainment		13.0 (12.0)	Respondents’ years of education
Marital Status
Married	78.2%		1 if married, 0 otherwise (o.w.).
Separated/divorced	13.4%		1 if separated/divorced, 0 o.w.
Widowed	5.5%		1 if widowed, 0 o.w.
Never married	2.9%		1 if never married, 0 o.w.
Economic/Financial Domain
Non-housing wealth		$137,309 (53,000)	Household-level sum of the current (2000) value of the following assets: values of checking and savings accounts; certificates of deposit and Treasury bills; stock and mutual funds; bonds; individual retirement accounts; vehicles; business equity; equity in real estate other than respondents' primary assets; and other reported non-housing assets.
Employment Domain
Annual work weeks		50.5 (52.0)	Weeks worked per year
Tenure < 2 years	15.3%		1 if years of service on current job > 2, 0 if service ≥ 2 years
Part-time employment	13.6%		1 if part-time, 0 if full-time
Employer-covered health insurance	69.0%		1 if covered by employer sponsored plan, 0 o.w.
Occupation
Managerial	14.8%		1 if managerial specialty operation, 0 o.w.
Professional	17.8%		1 if professional specialty operation & technical support, 0 o.w.
Sales	9.2%		1 if sales, 0 o.w.
Clerical/administrative	17.7%		1 if clerical, administrative support, 0 o.w.
Health/personal service	9.4%		1 if health services or personal services, 0 o.w.
Mechanics/construction/production	9.8%		1 if mechanics & repair, construction trade and extractors, precision production, 0 o.w.
Farming/armed forces	1.8%		1 if farming, fishing, forestry, Armed services, 0 o.w.
Services	5.1%		1 if private household, cleaning & building services, protection, food preparation, 0 o.w.
Operators	14.4%		1 if machine, transport operator, handler, 0 o.w.
Health Domain
Self-rated health
Excellent	19.7%		1 if excellent, 0 o.w.
Very good	38.1%		1 if very good, 0 o.w.
Good	29.3%		1 if good, 0 if o.w.
Fair/poor	12.9%		1 if fair/poor, 0 o.w.
Mental health (CES-D score)		1.2 (0.0)	Sum of 8-item depression battery (range: 0–8)
Health care expenses		$1,543 (750)	Out-of-pocket medical expenses in previous 2 years

Statistical Analyses

Both bivariate and multivariate analyses were used to compare older workers living in poverty to the working non-poor. In the bivariate analyses, we used t-tests were for continuous independent variables and chi-square tests for dichotomous variables. All 9 non-housing wealth components were compared across poverty status in bivariate analyses; the aggregate measure was used in the multivariable specification. Multivariable analyses were carried out with logistic regression, as the dependent variable is a binary indicator of working poverty. The principal advantage of multivariable specification is that it allows a comprehensive evaluation of the complete set of variables. SAS software, version 9.0, was used to estimate all models.

Results

Sample Characteristics

Table 1 presents descriptive statistics on the full sample (N=3,416) of older workers aged 51 to 61. Overall, the sample averaged 56.6 years of age, was almost equally divided among the sexes, and was more than four-fifths white. Average educational attainment was 13 years, and roughly three-quarters of sample members were married or partnered. Overall, sample members averaged $137,309 in non-housing wealth. About 84% of workers were employed full-time, and nearly 85% reported 2 or more years of service in their current positions. Sixty-nine percent were protected by an employer-based health insurance policy. The occupations with the greatest representation were managerial, professional, and clerical/administrative. Considering global self-assessed health, 12.9% of participants reported fair or poor health, 29.3% reported good health, and 57.8% reported excellent or very good health.

Comparison of Demographic, Employment & Health Attributes

In Table 2, we compare the demographic, employment, and health attributes of the older working poor and working non-poor groups. (Economic/financial differences are assessed in Table 3.) Demographic differences are apparent. The older working poor are more likely to be female, non-white, and non-married than the non-poor. They average fewer years of education. Comparison of the employment attributes of the two groups also yields a number of other outstanding differences. On balance, the older working poor are more likely to be employed part-time, more likely to be employed for fewer than 2 years, and less likely to be covered by employer-sponsored health insurance than the older working non-poor. The older working poor are also concentrated in service occupations and under-represented in white-collar occupations (i.e., managerial, professional, sales, and clerical).

Table 2.

Demographic, Employment, and Health Attributes of Older Workers: A Comparison of Poor and Non-Poor

Variable	Poor (n=121)	Non-Poor (n=3,295)	Test Statistic t-statistic/chi-square
Demographic Domain
Age	56.9	56.6
Female	64.5%	53.7%	χ2=5.42*
White	65.2%	82.4%	χ2=23.09***
Educational attainment	10.8	13.1	t=6.34***
Marital Statusb
Married	Ref.	ref.
Separated/divorced	19.2%	13.2%	χ2=7.87**
Widowed	14.2%	5.2%	χ2=24.79***
Never married	8.3%	2.7%	χ2=19.04***
Employment Domain
Annual work weeks	50.1	50.5
Tenure < 2 years	27.3%	14.8%	χ2=13.93***
Part-time employment	30.6%	13.0%	χ2=30.71***
Employer-covered health insurance	41.2%	70.1%	χ2=44.81***
Occupationb
Managerial	3.3%	15.2%	χ2=15.21***
Professional	11.6%	18.0%	χ2=5.37*
Sales	4.1%	9.3%	χ2=5.82*
Clerical/administrative	9.9%	18.1%	χ2=7.23*
Health/personal service	26.5%	8.7%	χ2=8.04*
Mechanics/construction/production	9.8%	9.9%
Farming/armed forces	1.7%	1.8%
Services	13.2%	4.8%	χ2=4.35*
Operators	Ref.	ref.
Health Domain
Self-rated healthb
Excellent	Ref.	ref.
Very good	23.1%	38.7%
Good	37.2%	29.0%	χ2=5.11*
Fair/poor	26.5%	12.4%	χ2=15.33***
Mental health (CESD score; range 0–8)	1.9	1.3	t=−3.42***
Health expenses	$1,309	$1,552

^aNotes. Unless otherwise noted, variable is not significant at 5% level of significance.

^bChi-square tests on categorical variables reflect comparison of relative (to referent category) proportions between working poor and non-poor.

^*p < .05

^**p < .01

^***p < .001

Table 3.

Household Wealth of Older Workers: A Comparison of Poor and Non-Poor

Variables	Poor (n=121)	Non-Poor (n=3,295)	t-Statistic
Asset Type
Checking/saving accounts	$6,657	$13,902	t=2.35*
Certificates of deposit/treasury bills	$457	$5,427	t=11.35***
Stocks/mutual funds	$14,932	$28,156
Bonds	$8	$2,534	t=5.51***
Individual Retirement Accounts	$15,065	$38,763	t=5.27***
Business equity	$3,579	$16,636	t=4.29***
Equity in real estate (excluding primary housing)	$1,928	$17,957	t=9.49***
Transportation assets (vehicles)	$4,764	$15,423	t=16.07***
Total wealth (excluding primary housing)	$46,482	$140,645	t=8.02***

Note. Unless otherwise noted, variable is not significant at 5% level of significance.

^*p< .05

^**p< .01

^***p< .001

The results described in Table 2 also illustrate potential disparities in individual health attributes between the older working poor and non-poor. A higher proportion of the older working poor reported their perceived health as good or fair/poor (relative to excellent) than that of the older working non-poor. In addition, the older working poor have poorer affective health (i.e., higher depressive symptoms). Nevertheless, no significant difference was found in out-of-pocket medical expenses.

Comparison of Household Wealth Components

Table 3 presents financial portfolios (i.e., economic/financial domain) of older working individuals by poverty status. Disaggregated household wealth data are given for the working poor and non-poor, and differences are assessed using t-tests. Overall, the non-housing wealth of the older working poor is $46,482 and that of the older working non-poor is $140,645; thus, the older working non-poor average more than 3 times the level of wealth of the older working poor. Considering individual wealth categories, we observe statistically significant variations in the mean of every asset category except stock/mutual funds, with the working poor consistently reporting lower values.

Several notable differences warrant comment. The working poor have virtually no secured debt ($8), a low-risk wealth component more commonly found in the portfolios of older individuals. They maintain meager interest-earning deposits and government securities ($457), as well as limited liquid wealth (checking/savings), with mean bank balances ($6,657) that are merely half of those of the working non-poor ($13,902). Similarly, the older working poor have a substantially lower mean level of dedicated retirement savings ($15,065) than do the older working non-poor ($38,763).

Results of Logistic Regression Analysis

Table 4 presents a fully-adjusted, multivariable logistic model of correlates of working poverty among older adults. In this model, we include variables from each of the domains of explanatory variables. The results suggest that a number of factors that are significant in the bivariate analysis remain robust to the inclusion of covariates.

Table 4.

Correlates of Working Poverty among Older Workers: Results of Logistic Regression Analysis (N=3,416)

Variable	Odds Ratio	95% Confidence Interval	Statistical Significancea
Demographic Domain
Age	0.99	(0.93, 1.06)
Female	0.95	(0.57, 1.58)
White	0.56	(0.36, 0.87)	P < .01
Educational attainment	0.86	(0.80, 0.92)	P < .001
Marital Status
Married	ref.	ref.
Separated/divorced	1.90	(1.10, 3.27)	P < .05
Widowed	2.89	(1.55, 5.39)	P < .001
Never married	3.95	(1.81, 8.62)	P < .001
Economic/Financial Domain
Net worth b	0.83	(0.68, 0.99)	P < .05
Employment Domain
Annual work weeks	0.99	(0.94, 1.03)
Tenure < 2 years	1.46	(0.91, 2.35)
Part-time employment	1.95	(1.18, 3.23)	P < .01
Employer-covered health insurance	0.41	(0.26, 0.64)	P < .001
Occupation
Managerial	0.49	(0.16, 1.53 )
Professional	1.34	(0.61, 2.95)
Sales	0.51	(0.18, 1.43)
Clerical/administrative	0.68	(0.31, 1.51)
Health/personal service	1.57	(0.82, 3.00)
Mechanics/construction/production	1.21	(0.55, 2.69)
Farming/armed forces	0.48	(0.10, 2.44)
Services	1.14	(0.53, 2.45)
Operators	ref.	ref.
Health Domain
Self-rated health
Excellent	ref.	Ref.
Very good	0.85	(0.43, 1.64)
Good	1.27	(0.67, 2.41)
Fair/poor	1.27	(0.61, 2.66)
Mental health CESD score	1.06	(0.97, 1.17)
Health care expenses c	1.00	(0.99, 1.00)

^aNotes. Unless otherwise noted, variable is not significant at 5% level of significance.

^bThis variable was rescaled by 100,000.

^cNet worth represents non-housing wealth and is rescaled by $100.

After controlling other factors, our findings indicate that the older working poor are less likely to be White (Odds Ratio [OR] = 0.56; 95% Confidence Interval [CI] 0.36, 0.87) than the working non-poor. They also have lower educational attainment, so that each year of additional education decreases the odds of being in poverty by 14% (OR = 0.86; CI 0.80, 0.92). Marital status is also strongly correlated with working poverty. Relative to married individuals, those who are separated or divorced have a nearly two-fold risk of being among the older working poor (OR = 1.90; CI 1.10, 3.27); widows/widowers have a nearly three-fold increased risk (OR = 2.89; CI 1.55, 5.39); and those who report never having been married have almost four times increased risk of being among the older working poor (OR = 3.95; CI 1.81, 8.62). Wealth is also negatively correlated with working poverty among older individuals. For every $100,000 of additional non-housing wealth, the likelihood of being among the older working poor is reduced by 17% (OR = 0.83; CI 0.68, 0.99).

Two other factors are significant in the multivariable model. Individuals who report working part time are 95% more likely to be among the older working poor (OR = 1.95; CI 1.18, 3.23), and workers with employer-covered health insurance are 59% less likely to be among the older working poor (OR = 0.41; CI 0.26, 0.64). After adjustment, none of the variables in the health domain indicates significant influence on the likelihood of being among the working poor.

Summary

Increasing public attention has been paid to working individuals whose earnings provide a way of life characterized by subsistence. This has been largely prompted by swelling numbers of working poor, resulting from economic conditions and modifications to federal welfare policies in the 1990s. The trend toward later workforce withdrawal could mean that the number of older workers living in poverty will expand. Economic deprivation in the years preceding retirement could result in a multitude of adverse outcomes, from retirement timing and adaptation to physical and emotional well-being. Using data from the year 2000 wave of the HRS, this study provided a profile of working individuals, living in poverty, who are nearing retirement age, comparing the attributes of such individuals to older workers not living in poverty.

Our results suggest that approximately 3.5% of the older workers aged 51 to 61 were categorized as working poor, based on labor force participation and household income. This proportion is consistent with recently reported proportions for members of this age cohort (Mosisa, 2003). Without controlling other factors, the working poor had significantly lower levels of most wealth components than the working non-poor, including such essential income-generating components (Crystal, 1996) as bonds. The working poor also had lower levels of self-rated health. Although out-of-pocket medical expenditures were not markedly different, the relative burden (i.e., relative to income) of medical costs is likely higher for the working poor.

Disadvantage was furthermore observed in factors related to employment. The working poor were overrepresented in blue-collar occupations, were more likely to be employed part-time, and had lower likelihood of having health insurance provided by their employers. Demographic differences also emerged from the data. A greater proportion of the working poor were female, non-white, and unmarried than the working non-poor; the working poor also had lower educational attainment. A number of important factors continued to distinguish the working poor and non-poor in the multivariate specification. Non-white race, lower educational attainment, non-married civil status were all correlated with working poverty status, as were lower net worth, part-time employment, and lack of employer-covered health insurance.

Conclusions and Implications

A number of implications are clear from the findings. In view of the fact that the working poor have, on average, rather low secured debt, and insufficient personal savings and retirement wealth to finance post-employment consumption, consumer educators and financial advisors should be encouraged to target populations identified in this research for potential interventions. Thus, the less educated, females, non-whites, and especially non-married household heads should be guided to accumulating wealth in earlier life cycle stages, which could generate unearned income sources in later life, particularly after retirement. Absent intervention, impoverished individuals could face significant impediments to complete withdrawal from the labor force, and when fully retired, likely encounter a worsened standard of living. Although low-earning workers have a lower opportunity cost (i.e., higher income replacement rate) to retirement than workers who have greater earnings in the immediate pre-retirement period (Clark, Burkhauser, Moon, Quinn, & Smeeding, 2004), the result is nevertheless sustained economic insecurity, whose multiple associated stressors have been linked to a range of adverse effects (Lynch et al., 1997).

That older participants who lack the varied buffers of a spouse or domestic partner, most importantly financial support, are at substantial risk for poverty has particular salience. Once more, our results suggest that unmarried working individuals over 50 are between two and four times more likely to be living in poverty than married participants, with those who have never been married at the highest risk for indigence. Nevertheless, most of the policy concern for the working poor is related to families, especially those with minor children, leaving the unmarried near-elderly acutely vulnerable to the effects of economic hardship. Future strategies could thus be directed toward assistance for unmarried older individuals. The most obvious policy lever is the Earned Income Tax Credit (EITC), an innovative program established in 1975 with the goal of reducing poverty among low-income working families by lessening tax burden and supplementing wages. In its current form, the EITC is disproportionately beneficial to poor families with minor children, and provides very limited benefits, in the manner of minimal federal tax relief, to impoverished people between the ages of 25 and 64 who are not rearing minor children. The EITC could therefore be modified to better address deprivation among our population of interest. Expanding EITC benefits for the near elderly would result in a positive outcome not only for the unmarried, but also for families who no longer have dependent minor children.

Also of concern is that our findings revealed that the poor were nearly twice as likely to be working in part-time positions as the non poor. Part-time employment often excludes important benefits, such as health insurance and pension plans, which are essential to both pre-and post-retirement well-being, and our data do suggest that the working poor are significantly lacking in employer-sponsored health insurance and pension wealth. Previous literature has nevertheless suggested that simply increasing the working hours of the poor is an inappropriate solution to reducing poverty low-income workers in the labor force (Kim, 1998b; Zagorsky, 1999). Rather, comprehensive employment training programs, which could include job search support, on-the-job training and development, and literacy and remedial skills help, are needed to assist low-wage workers to move into positions that promise full-time, full-year employment. Indeed, earlier research has suggested that nearly half of all low-income workers who hold part-time or part-year employment would no longer be classified as indigent if they held in full-time, full-year jobs (Kim, 1998b). Even so, older workers have historically comprised an insignificant proportion of participants in federal and state manpower training programs (Schulz, 2001), a situation which arises from policymakers’ belief that investment in older workers is not cost effective. We must also acknowledge that the efficacy of alleviating poverty via job training programs is debatable (LaLonde, 1995), despite the relatively low cost of implementation.

Disproportionate employment in less-skilled occupations is disconcerting, as well. Both the descriptive and bivariate evidence point to significantly higher proportions of working poor in service and blue-collar occupations, although we should note that no differences were detected in the multivariate specification. All the same, if the older poor are overrepresented in such occupations, a finding that is consistent with the descriptive evidence on low-income workers across the broader age spectrum (Mosisa, 2003), they will, in many cases, be employed in unhealthful conditions or those requiring physically demanding work. Whereas younger workers may be able to meet the physical challenges of lower-skilled occupations, older individuals likely have higher likelihood of workplace injury. Service occupations, in particular, have high rates of disability (Hayward & Grady, 1990) and mortality (Moore & Hayward, 1990). In addition, although the projected trend among workers over 55 is to shift to white-collar occupations (United States General Accounting Office, 2001) which have fewer physical requirements, it is unlikely that the poor will have the luxury to make such an adjustment, leaving them vulnerable to injury and the threat of layoff (Couch, 1998).

There are limitations to this study that must be acknowledged. Foremost is the design. Our investigation is cross-sectional, and therefore provides only a static representation of older workers living in poverty. As such, we have merely identified factors that correlate with working poverty; we cannot, however, assess whether there is any causal relationship between poverty status and the various attributes considered in this research, nor can we comment on the direction of potential causal associations. Disentangling matters of social selection and causation in the context of working poverty must be left to future longitudinal research on older workers. Our findings should moreover be read with some caution, because in studies of age-specific cohorts such as ours, compositional factors may be entwined with explanatory variables, confounding the relationship of the correlate to the outcome. In this study, educational attainment stands out as the correlate most likely to be related to the cohort’s make-up, which may introduce error. That is, since most individuals in the birth cohorts studied are likely to have slightly lower educational attainment than later cohorts, the association between this variable and working poverty should be interpreted with reservation.

Second are limitations resulting from measurement of the variable of interest. For example, our measure of working poverty does not include important historical information, such as the duration of the poverty state (Power et al., 1999), which might be relevant to the assessment of many of the correlates considered in this study and would undoubtedly be important to assessing any causal link to deficits in well-being (Lynch et al., 1997). Moreover, in using the U.S. government’s official poverty threshold, our representation was characterized solely by income deficiency, and thus excludes direct measurement of material deprivation. That the working poor experience restricted standard of living is therefore simply assumed. Some research has argued that definitions which are based on income thresholds do not capture the multifaceted nature of economic insecurity (Budowski, Tillmann, & Bergman, 2002). Nonetheless, the HRS data, albeit lacking in questions regarding deprivation of essential goods and services, remains the best data source for an investigation such as this.

Finally, there is the issue of statistical power. By selecting the most recent wave of HRS data available, we arrive at a working poor number that is sufficient to investigate the direct effect of potential correlates, but not large enough to explore the coexistence of variables or important strata. For example, the joint correlation with working poverty of marital status and gender, two variables of distinction in the bivariate results, could not be explored because of small cell size. Investigators who are interested in exploring concomitant states or relevant strata may have to pool data from the various HRS survey waves or introduce a broader definition of poverty to do so.

The number of persons aged 65 and above is projected to rise to 80 million by 2040 (United States Bureau of Labor Statistics, 2003), and the labor market will reflect this demographic shift. Our findings, which identify a number of factors associated with working poverty status among employed older individuals, provide knowledge that could help professionals who deal with the near-elderly--such as consumer educators, financial educators, and policy makers--to predict and prepare the graying labor force for the coming years. The correlates of working poverty status identified in this study contribute to the literature on the economics of aging as well as to the information available for policy makers that face the increasing needs of older employees.