doc daimler only
Regina S. Baker University of Pennsylvania
The Changing Association Among Marriage, Work,
and Child Poverty in the United States, 1974–2010
Marriage and work have long been central to debates regarding poverty and the family. Although ample research has demonstrated their negative association with child poverty, both marriage and work have undergone major transformations over recent decades. Conse- quently, it is plausible that their association with child poverty may have also changed. Using 10 waves of U.S. Census Current Population Survey data from the Luxembourg Income Study, this study examined the relationships among marriage, work, and relative measures of child poverty from 1974 to 2010. The results indicated that both marriage and work still decrease the odds of child poverty. However, time interac- tions showed marriage’s negative association with child poverty has declined in magnitude, whereas work’s negative association with child poverty has increased in magnitude. These findings underscore the historically varying influence of demographic characteristics for poverty. They also suggest the limitations of overemphasizing marriage and the growing importance of work for reducing child poverty in America.
The United States stands out for its failure to significantly reduce child poverty over the past few decades and its unusually high child poverty rates relative to other rich countries (Gornick & Jäntti, 2012; Rainwater & Smeeding, 2004).
Department of Sociology, University of Pennsylvania, 218 McNeil Building, 3781 Locust Walk, Philadelphia, PA 19104 (regbaker@sas.upenn.edu).
Key Words: employment, inequality, marriage, poverty.
Accordingly, there has been vibrant scholarly literature concerning poverty among chil- dren and families (see Edin & Kissane, 2010; Lichter, 1997; Seccombe, 2000). Marriage and work have been central to this scholarship and related policy debates. The continued focus on marriage and work in alleviating child poverty is salient given the major transformations in these institutions. Whereas a married couple with children was once the norm, nonmari- tal births and single-parent households have become commonplace. Moreover, divorce rates remain relatively high; cohabitation is increas- ing; Americans are choosing to marry later, if at all; and marriage has become a more selec- tive institution (Cherlin, 2009; Raley, 2000). The attributes and patterns of work have also changed greatly. Median wages have stagnated, earnings inequality has worsened (Bluestone & Harrison, 2001; Mishel, Birens, Gould, & Shierhotz, 2012), and female labor force partic- ipation has increased dramatically (Lichter & Crowley, 2004). There has also been a decline in well-paid blue-collar work, growth in part-time jobs, and greater job insecurity (Kalleberg, 2007; Mishel et al., 2012). Although it is well documented that marriage and work decrease a child’s odds of being poor, these changes raise the question of whether their associations with child poverty have also changed over time. In this study I addressed this query.
Background
Theories Why Marriage and Work Are Associated With Poverty
There are several causal explanations for why marriage affects child poverty. First, marriage
1166 Journal of Marriage and Family 77 (October 2015): 1166–1178 DOI:10.1111/jomf.12216
Marriage, Work, and Child Poverty 1167
increases the number of potential adult earners, and thus income, in the household. Relat- edly, dual-earner households are more able to absorb the shock of income losses compared to single-earner households. Second, marriage benefits families through economies of scale. By sharing expenses (e.g., rent and other house- hold goods and services), married couples can save more, support a higher standard of living, and invest for the future (Amato & Maynard, 2007). Thus, from an income-to-needs perspec- tive, married households fare better financially (Thomas & Sawhill, 2005). Third, according to Becker’s (1981) theory of household special- ization, married couples have greater flexibility in how they divide their time between home and market production and thus can maximize household earnings (Amato & Maynard, 2007). However, marriage does not benefit all children equally: White children typically experience greater economic benefits from marriage than do Black and Hispanic children (Manning & Brown 2006).
Although marriage may lead to greater levels of income, employment, and other available resources, selection into marriage also con- tributes to the relationship between family structure and child poverty. The adults most likely to form and maintain two-parent married households are more stable, well adjusted, and resource rich (Brown, 2010). Furthermore, unmarried, low-income parents are more likely to marry after experiencing increases in earnings (Gibson-Davis, 2009). Thus, failure to consider preexisting differences that influence selec- tion into marriage can lead to overstating the causal effects of family structure (McLanahan & Percheski, 2008).
Regarding the association between work and child poverty, having paid earners in the household generates earnings, earnings are the dominant source of income for households, and income is the basis for defining poverty. Indeed, Rainwater and Smeeding (2004) concluded that ensuring at least one parent is employed is the most important step to avoiding child poverty, and they found that having multiple earners in the household further lowers this risk. More- over, low employment and low wages translate into large differences in earnings capacities (Sigle-Rushton & McLanahan, 2002), which underscores the necessity of work for avoiding poverty.
Given the role of marriage and work in avoiding child poverty, ample research has examined related trends. Several studies have found that changes in family structure—namely, the rise in nonmarital births—largely explain increases in child poverty from the 1970s to the mid-1990s (Christopher, 2005; Lichter & Crowley, 2004; Nichols, 2013). Accordingly, scholars have shown that poverty rates would have declined greatly had single parents been married (Sigle-Rushton & McLanahan, 2002; Thomas & Sawhill, 2005). Conversely, in more recent decades it has been changes in work, not family structure, that most explain child poverty trends (Chen & Corak, 2008; Lichter & Crowley, 2004; Nichols, 2013). For example, after the 1996 welfare reforms, greater maternal employment accounted for most child poverty in single-parent families, especially among Blacks and Latinos (Lichter & Crowley, 2004). Similarly, Nichols (2013) found that increased parental work effort mainly drove declines in child poverty from 1993 to 2011.
Indeed, these studies illustrate how composi- tional changes in family structure and parental work help explain fluctuations in child poverty. However, one nuanced but key gap in the liter- ature regarding marriage and work is precisely how the magnitude of their association with child poverty has changed. There are plausible reasons to expect these associations could have gotten either weaker or stronger from the 1970s to the present.
The Case for Changes in the Association Between Marriage and Child Poverty
Women’s marriage and fertility trends provide at least one reason why marriage may have a stronger negative association with child poverty. As a whole, women are delaying marriage more, but although delayed childbearing has increased among highly educated women, less educated women still tend to have children earlier (Wilde, Batchelder, & Ellwood, 2010). Delayed child- bearing leads to substantial increases in the earnings and work hours of both mothers and fathers (Miller, 2010; Wilde et al., 2010). More- over, children born to low-skill women tend to come early, when the mother often has few earnings and is more likely to be unmarried, and children born to high-skill women tend to enter a married family during their peak earn- ing years (Wilde et al., 2010). The result is that
1168 Journal of Marriage and Family
married households have increased advantages over unmarried households. Thus, a stronger negative association between marriage and child poverty may have occurred from the 1970s to 2010.
Conversely, there are at least two plausible reasons why the association between marriage and child poverty may have gotten weaker. First, an increasing number of children are living with cohabiting parents (Manning & Brown, 2006). Even in the short term or on a transient basis, these households potentially benefit from economies of scale and having two earners, given that they generally fare better economically than single-parent households (Manning & Brown, 2006; Thomas & Sawhill, 2005). Therefore, any penalty for a child being in an unmarried household may have weakened because of rising cohabitation, which has cre- ated more dual-earner nonmarital households. Second, single parents have generally become less homogeneous. Employment among single mothers has risen dramatically (Lichter & Crow- ley, 2004). Single motherhood has increased even in the middle of the education distribution (Ellwood & Jencks, 2004), and single fathers, who have greater income and are more likely to cohabitate than single mothers, are on the rise (Livingston, 2013). Thus, single parents have become less uniformly disadvantaged, which could have weakened the negative association between marriage and poverty.
The Case for Changes in the Association Between Work and Child Poverty
One factor that may have led to a stronger neg- ative association between work (i.e., number of earners) and child poverty is the long-term stag- nation in median worker earnings that increased pressure on dual earners. Earnings inequality rose dramatically in the 1980s, and by the late 1990s most families had experienced declines in real incomes (Bluestone & Harrison, 2001). In part as a result, female labor participation increased (Bianchi, 2000; Lichter & Crowley, 2004), and women’s incomes have constituted a growing share of family income in all fam- ily types (U.S. Department of Commerce, 2011). This may explain why increases in women’s employment coincided with declines in child poverty (Lichter & Crowley, 2004). In addi- tion, families have had to increasingly rely on paid earnings because of dramatic decreases
in welfare receipt and the value of welfare transfers following the 1996 welfare reforms (Danziger, 2010). Further, the earned income tax credit, which requires that one be employed, has expanded greatly into the largest family assistance program (Danziger, 2010), and con- siderable evidence demonstrates that the EITC reduces child poverty (Handler & Hasenfeld, 2007). Thus, the transition from work-free wel- fare to social assistance contingent upon work may have increased the necessity of work for avoiding child poverty.
Conversely, the changing nature of jobs and greater job insecurity may have weakened the association between work and child poverty. There has been a decline in well-paid blue-collar jobs, abundant low-wage work, and technologi- cal changes that contributed to skill and spatial mismatches that have disadvantaged low-skilled and low-educated workers (Kalleberg, 2007). Despite rising consumer costs, workers with less than a high school education or some col- lege are earning increasingly less, and those with at least a bachelor’s degree are making only slightly more than those in the 1970s (Bureau of Labor Statistics, 2013). Furthermore, the Great Recession has exacerbated job inse- curity, with racial/ethnic minorities and the less educated most adversely affected (Bureau of Labor Statistics, 2012). These changes, in light of the increasing poverty among workers (Brady, Baker, & Finnigan, 2013), suggest work may be less protective against child poverty.
Method
To scrutinize the potentially changing relation- ships of marriage and work with child poverty over time, I use Current Population Survey data from the Luxembourg Income Study (LIS) Database (http://www.lisdatacenter.org). The LIS is advantageous because of its high-quality measure of disposable household income that incorporates taxes, transfers, and tax credits such as the EITC. Because disposable house- hold income provides a more comprehensive and accurate measure of household income than even the underlying Current Population Survey, calculations of poverty that are based on it are more valid and reliable (Brady, 2003; Rainwater & Smeeding, 2004). I used all available U.S. waves: 1974, 1979, 1986, 1991, 1994, 1997, 2000, 2004, 2007, and 2010, and included only households with children under age 18. I
Marriage, Work, and Child Poverty 1169
weighed households using a LIS-constructed child weight accounting for the number of chil- dren in the household. Thus, the unit of analysis was 192,547 individual children.
Variable Measures
The dependent variable included two measures of child poverty. Following recent studies that have used LIS data (Brady et al. 2013; Chen & Corak, 2008; Gornick & Jäntti, 2012; Rainwater & Smeeding, 2004), I employed a standard relative measure of poverty in which the thresh- old is 50% of median equivalized, posttax and posttransfer household income. All children in households below this threshold are poor. My second measure used an anchored threshold based on the 1974 median adjusted for inflation over time (using the Consumer Price Index) so that the measures are fixed (Chen & Corak, 2008). Although the standard relative measure may be less sensitive to the business cycle and improvements in standards of living and economic development, the anchored measure should be more responsive. Compared to the official poverty measure, not only do these measures (and the income definition underlying them) better capture the resources available to families, but they are also more consistent with leading conceptualizations of poverty, such as social exclusion and capability deprivation (Chen & Corak, 2008; Deaton, 2006).
The first key independent variable was a binary measure indicating whether the head of household is married. Although some children resided in households where a parent was not the head, 96% of the heads of household in my sample lived with their own children under 18. My second key independent variable was the total number of earners, which encompassed all persons with positive earnings in the household. In a variety of sensitivity analyses, I experi- mented with alternative work variables: the total weekly work hours, annual full-time hours, and annual part-time hours among all earners in the household. The main conclusions were generally consistent with these alternative work measures. However, because of issues with missing data and model nonconvergence for these alternative work measures, I focused on number of earners in this study.
Following previous research (Brady et al., 2013; Chen & Corak, 2008; Christopher, 2005; Rainwater & Smeeding, 2004), I adjusted
for several variables associated with poverty: age, education, race, and household composi- tion. Age of the head of household and Age2
were measured in years, and a binary vari- able indicated whether the head is under age 25. Two binary variables indicated whether the head of household’s education was “less than high school” or “a college degree” (refer- ence group = high school diploma and some college). “Black,” “Latino,” and “Other” were binary measures of the head of house- hold’s race/ethnicity (reference = White). I also included measures of the number of working-age (18–64) adults in the household, the total num- ber of children under age 18 in the household, and whether adults over age 65 resided in the household.
Analytic Strategy
After presenting descriptive trends, I present logistic regression models to assess the relation- ships of marriage and work with child poverty. The first analysis pooled the 10 waves into one combined sample so I could examine the associ- ations of marriage and work with child poverty when interacted with linear time. I used a count measure for each of the 10 LIS waves based on years (i.e., 1974 = 0, 1979 = 5 . . ., 2010 = 36). Because the results may vary in a nonlinear way, the next analyses included interactions with year binary variables (reference year = 1974). Indi- vidual year measures also control for unobserved year-specific factors (e.g., business cycle).
Results
Descriptive Trends
All variable means, by year, are displayed in Table 1. Both child poverty measures had similar trends, with anchored poverty consistently hav- ing lower rates than relative poverty. From 1974 to 1991, relative and anchored child poverty rates increased from 15% to 25% and 23%, respectively. From 1994 into the early 2000s, child poverty declined slightly and has remained fairly stable since. In 2010, child poverty was 21% (relative) and 15% (anchored).
The percentage of children who lived in a household with a married head of household has steadily declined since 1974 (with the slight exception of 2000). Whereas in 1974, 84% of all children lived in married-headed households, by
1170 Journal of Marriage and Family
T ab
le 1.
V a
ri a
b le
M ea
n s
(a n
d S
ta n
d a
rd D
ev ia
ti o
n s,
in P
a re
n th
es es
) b
y Ye
a r
V ar
ia bl
e 19
74 19
79 19
86 19
91 19
94 19
97 20
00 20
04 20
07 20
10
R el
at iv
e po
ve rt
y .1
5 (.
36 )
.1 9
(. 40
) .2
2 (.
41 )
.2 5
(. 43
) .2
4 (.
43 )
.2 2
(. 42
) .2
2 (.
41 )
.2 1
(. 41
) .2
2 (.
42 )
.2 1
(. 41
) A
nc ho
re d
po ve
rt y
.1 5
(. 36
) .1
8 (.
40 )
.1 9
(. 40
) .2
3 (.
42 )
.2 2
(. 42
) .1
8 (.
39 )
.1 5
(. 36
) .1
5 (.
36 )
.1 5
(. 36
) .1
5 (.
36 )
M ar
ri ed
he ad
.8 4
(. 37
) .7
8 (.
41 )
.7 6
(. 43
) .7
3 (.
45 )
.7 2
(. 45
) .7
1 (.
45 )
.7 2
(. 45
) .7
0 (.
46 )
.6 9
(. 46
) .6
8 (.
47 )
N um
be r
of ea
rn er
s 1.
78 (0
.9 7)
1. 83
(1 .0
4) 1.
76 (0
.9 0)
1. 72
(0 .9
5) 1.
72 (0
.9 2)
1. 75
(0 .8
9) 1.
81 (0
.8 8)
1. 69
(0 .8
7) 1.
69 (0
.8 5)
1. 63
(0 .8
6) A
ge of
he ad
38 .3
2 (1
0. 36
) 38
.0 6
(9 .8
2) 37
.8 6
(9 .3
3) 38
.0 6
(9 .2
7) 38
.4 0
(9 .3
6) 38
.8 5
(9 .5
3) 38
.7 3
(9 .5
3) 38
.9 6
(9 .8
6) 39
.4 6
(1 0.
06 )
39 .6
0 (1
0. 34
) H
ea d
un de
r ag
e 25
.0 7
(. 25
) .0
6 (.
23 )
.0 4
(. 20
) .0
4 (.
20 )
.0 5
(. 21
) .0
5 (.
21 )
.0 5
(. 23
) .0
5 (.
23 )
.0 5
(. 21
) .0
5 (.
22 )
N o
hi gh
sc ho
ol de
gr ee
.3 0
(. 46
) .2
6 (.
44 )
.1 9
(. 39
) .2
0 (.
40 )
.1 9
(. 39
) .1
9 (.
39 )
.1 7
(. 38
) .1
7 (.
37 )
.1 5
(. 36
) .1
5 (.
36 )
C ol
le ge
de gr
ee .1
7 (.
38 )
.2 0
(. 40
) .2
2 (.
41 )
.2 1
(. 41
) .2
3 (.
42 )
.2 3
(. 42
) .2
5 (.
43 )
.2 6
(. 44
) .2
8 (.
45 )
.3 0
(. 46
) B
la ck
he ad
.1 1
(. 31
) .1
5 (.
36 )
.1 4
(. 34
) .1
6 (.
36 )
.1 6
(. 36
) .1
6 (.
36 )
.1 5
(. 36
) .1
5 (.
36 )
.1 5
(. 36
) .1
5 (.
35 )
O th
er ra
ce he
ad .0
1 (.
11 )
.0 2
(. 15
) .0
3 (.
16 )
.0 4
(. 19
) .0
4 (.
19 )
.0 5
(. 21
) .0
5 (.
22 )
.0 7
(. 25
) .0
7 (.
26 )
.0 8
(. 27
) L
at in
o he
ad .0
6 (.
23 )
.0 8
(. 26
) .0
9 (.
29 )
.1 2
(. 33
) .1
3 (.
34 )
.1 5
(. 35
) .1
6 (.
37 )
.1 9
(. 39
) .2
0 (.
40 )
.2 1
(. 40
) N
um be
r of
w or
ki ng
-a ge
ad ul
ts 2.
08 (0
.6 4)
1. 76
(0 .4
5) 2.
01 (0
.6 4)
2. 04
(0 .7
6) 2.
02 (0
.7 3)
2. 01
(0 .7
1) 2.
06 (0
.7 7)
2. 04
(0 .7
6) 2.
04 (0
.7 6)
2. 09
(0 .8
2) N
um be
r of
ch il
dr en
2. 30
(0 .9
5) 2.
57 (1
.3 4)
2. 09
(. 84
) 2.
43 (1
.2 5)
2. 40
(1 .1
7) 2.
42 (1
.2 4)
2. 41
(1 .2
6) 2.
39 (1
.1 9)
2. 40
(1 .2
0) 2.
39 (1
.2 1)
A du
lt s
ov er
ag e
65 .0
3 (.
18 )
.0 2
(. 13
) .0
3 (.
17 )
.0 3
(. 17
) .0
3 (.
17 )
.0 3
(. 18
) .0
4 (.
19 )
.0 4
(. 20
) .0
4 (.
20 )
.0 5
(. 21
) N
4, 61
1 6,
00 7
4, 26
1 22
,2 29
21 ,6
27 19
,0 05
18 ,1
21 33
,7 45
32 ,1
14 30
,8 27
N o
te :
he ad
= he
ad of
ho us
eh ol
d.
Marriage, Work, and Child Poverty 1171
Figure 1. Child Poverty by Head Marital Status, 1974–2010.
0
10
20
30
40
50
60
% C
hi ld
P ov
er ty
Year
Relative Child Poverty by Marital Status of Head
Married Head Unmarried Head
0
10
20
30
40
50
60
% C
hi ld
P ov
er ty
Year
Anchored Child Poverty by Marital Status of Head
Married Head Unmarried Head
2010 only 68% of children did. This decrease in children living in households with a mar- ried head of household is consistent with the declining marriage rates among adults in gen- eral, which fell from 72% in 1970 to 51% in 2011 (Fry, 2012).
The mean number of earners increased from 1.78 in 1974 to 1.83 in 1979, then declined in 1986 and changed little through the 1990s. However, by 2000 the mean number of earn- ers increased to 1.81 and then declined until 2010, when children had only 1.63 earners in the household, on average, the lowest mean of all 10 time points. Of course, the long-term trends in the number of earners combine both a rise in two-earner households and the number of house- holds containing only one working-age adult (e.g., single-mother households)
Given the nature of this study, in addition to descriptive trends of the pooled sample it was also important to consider trends in poverty rates by parental marital status and the number of earners in the household. These descripted trends are depicted in Figure 1 and Figure 2. Child poverty rates for unmarried-headed house- holds were much higher than married-headed households, as expected (see Figure 1). In 1974, child poverty rates for unmarried and married households were 47.8% and 9.3%, compared to 2010 rates of 40.4% and 11.7% (relative mea- sure) and 31.4% and 7.4% (anchored measure),
respectively. As both figures illustrate, the gap in child poverty rates between married and unmar- ried households has narrowed over time.
Figure 2 shows the child poverty rates by number of earners (i.e., no earners, one earner, and two or more earners). In 1974, child poverty rates for children in no-earner, one-earner, and two-or-more-earners households were 83.4%, 19.6%, and 7.2%, respectively. In 2010, the corresponding child poverty rates were 88.5%, 30.2%, and 6.2% (relative measure) and 82.3%, 20%, and 3.5% (constant measure). Despite some fluctuations here and there (e.g., the decline in anchored child poverty to 4.5% in 2010), these graphs illustrate an overall trend that the child poverty gap between number of earners has widened over time.
Pooled Analyses With Linear Time Interactions
Table 2 displays the logistic regression results for marriage and earners on child poverty for the pooled sample. Across all models, mar- riage and the number of earners decreased the odds of both measures of child poverty. For instance, in the baseline Model 1, having a mar- ried head of household reduced relative and anchored child poverty by a factor of 2.71 and 2.74. (N.B. the factor is calculated by dividing 1 by the odds ratio [OR], e.g., 1/.369 = 2.71; 1/.365 = 2.74.) Each additional earner reduced
1172 Journal of Marriage and Family
Figure 2. Child Poverty by Number of Earners in Household, 1974–2010.
0
10
20
30
40
50
60
70
80
90
100
% C
h il
d P
o v e rt
y
Year
Relative Child Poverty by
Number of Earners in Household
No Earners One Earner Two or More Earners
0
10
20
30
40
50
60
70
80
90
100
% C
h il
d P
o v e rt
y
Year
Anchored Child Poverty by
Number of Earners in Household
No Earners One Earner Two or More Earners
relative and anchored child poverty by factors of 3.73 (1/.369) and 3.98 (1/.251).
Model 2 interacted marriage and number of earners with linear year. Both marriage and number of earners still remained significant and decreased the odds of child poverty; how- ever, the interactions suggest a divergence in their relationships with poverty over time. The marriage and year interactions for both relative (OR = 1.022) and anchored (OR = 1.025) child poverty suggest an overall weakening in the negative association between marriage and child poverty. However, the number of earners and linear year interactions for both relative (OR = 0.981) and anchored (OR = 0.976) child poverty suggests a stronger negative associ- ation with work and child poverty. Both sets of interactions are statistically significant, thus indicating a noteworthy change in the magnitude of the association between marriage and work and child poverty over time. It is important to note that these results are consistent with the trends in child poverty by marital status and number of earners displayed in Figure 1 and Figure 2.
Pooled Analyses With Nonlinear Time Interactions
Because the observed trends for marriage and work in Model 2 might vary in a nonlinear way,
Model 3 included binary variables for each year (reference year = 1974) and interactions with marriage and work with each year. As in Model 1 and Model 2, having a married head of house- hold led to reduced odds of relative and anchored child poverty by a factor of 4.61 and 4.57, respectively. However, the main effect of the number of earners had a weaker association than in Model 2; it reduced the odds of child poverty by a factor of only 2.70 and 2.66. Overall, the interactions in Model 3 also suggest change over time. With the exception of 1979 and 1986 in the relative poverty model and 1979 in the anchored poverty model, all interactions were signifi- cant. Compared to the reference year 1974, the association between marriage and child poverty weakened in both the relative and anchored mod- els. By contrast, the interactions for the number of earners illustrated the opposite trend.
For substantive interpretation, Figure 3 graph- ically displays the inverse odds of the interac- tions from Model 3 in Table 2. An increasingly negative association implies a greater reduction in the odds of child poverty. Panel A of the figure displays results for the marriage and time interactions. Over time, the magnitude of the negative association between marriage and child poverty weakened. More specifically, most of this change took place between 1974 and 1991 (as shown in the top four bars in Figure 3, Panel A). From 1991 to 2010, the association between
Marriage, Work, and Child Poverty 1173
Table 2. Logistic Regression Models of Child Poverty, Pooled Sample 1974–2010, Odds Ratios
Relative poverty Anchored poverty
Variables Model 1 Model 2 Model 3 Model 1 Model 2 Model 3
Married (ref.: no) 0.369*** 0.225*** 0.217*** 0.365*** 0.216*** 0.219***
Number of earners 0.268*** 0.381*** 0.371*** 0.251*** 0.386*** 0.376***
Married × year 1.022*** 1.025***
Earners × year 0.981*** 0.976***
Married × 1979 1.084 1.080 Married × 1986 1.381 1.385 Married × 1991 1.890*** 1.897***
Married × 1994 2.023*** 2.007***
Married × 1997 1.947*** 1.827***
Married × 2000 1.897*** 2.008***
Married × 2004 2.050*** 2.095***
Married × 2007 1.795*** 1.732***
Married × 2010 1.896*** 1.867***
Number of earners × 1979 1.061 1.046 Number of earners × 1986 0.759* 0.714*
Number of earners × 1991 0.678*** 0.637***
Number of earners × 1994 0.642*** 0.622***
Number of earners × 1997 0.588*** 0.549***
Number of earners × 2000 0.686** 0.580***
Number of earners × 2004 0.541*** 0.446***
Number of earners × 2007 0.594*** 0.461***
Number of earners × 2010 0.482*** 0.379***
Year 1.005*** 1.016*** 0.986*** 1.002 Year 1979 1.042 0.955 Year 1986 2.186*** 1.893**
Year 1991 2.012*** 1.776***
Year 1994 1.991*** 1.754***
Year 1997 2.073*** 1.568**
Year 2000 1.853*** 1.170 Year 2004 1.857*** 1.250 Year 2007 2.027*** 1.339 Year 2010 1.988*** 1.381*
Age of head 0.924*** 0.997*** 0.918*** 0.935*** 0.928*** 0.927***
Age of head2 1.001*** 1.001*** 1.001*** 1.000*** 1.001*** 1.001***
Head under age 25 (ref.: no) 1.758*** 1.759*** 1.21*** 1.787*** 1.784*** 1.871***
Education (ref.: high school degree/some college)
No high school diploma 2.918*** 2.887*** 2.939*** 2.799** 2.761*** 2.828***
College degree 0.295*** 0.293*** 0.296*** 0.324*** 0.325*** 0.328***
Race of head (ref.: White) Black 1.795*** 1.775*** 1.769*** 1.724*** 1.704*** 1.699***
Latino 1.904*** 1.890*** 1.901*** 1.738*** 1.732*** 1.746***
Other race 1.560*** 1.553*** 1.584*** 1.524*** 1.519*** 1.582***
Number of working age 1.063** 1.123*** 1.125*** 1.045 1.109*** 1.100***
Number of children 1.442*** 1.438*** 1.446*** 1.398*** 1.393*** 1.404***
Adults over age 65 (ref.: no) 0.546*** 0.547*** 0.534*** 0.529*** 0.531*** 0.514***
N 192,547 192,547 192,547 192,547 192,547 192,547 BIC 46,693.3 46,508.52 46,488.86 42,214.75 41,965.97 41.842.87
Note: ref. = reference category; head = head of household; BIC = Bayesian Information Criterion. *p < .05. **p < .01. ***p < .001.
1174 Journal of Marriage and Family
Figure 3. Association Between Marriage, Number of Earners, and Child Poverty Over Time, 1974–2010.
–5 –4 –3 –2 –1 0
1974
1979
1986
1991
1994
1997
2000
2004
2007
2010
PANEL A Married × Time Relative Anchored
–8 –6 –4 –2 0
1974
1979
1986
1991
1994
1997
2000
2004
2007
2010
PANEL B Number of Earners × Time
Relative Anchored
)esrevnI(oitaRsddO)esrevnI(oitaRsddO
Note: The figures illustrate results from Model 3 of Table 2. To calculate the inverse odds ratio, I used the following equation: (e.g., for the year 1979): −1 / (Married79 Odds Ratio * Married Odds Ratio).
marriage and child poverty remained rather sta- ble, with some slight fluctuations. As shown in Panel B of Figure 3, from 1974 to 1979 the strength of the association between number of earners and child poverty was relatively stable. From 1986 onward, however, the number of earners had an increasingly greater reduction on the odds of child poverty (despite small declines in 2000 and 2007).
Especially noteworthy is that, when we com- pare the association between marriage and child poverty in 1974 and 2010, we see that the mag- nitude of the relationship declined by more than half; however, the magnitude of the relation- ship between the number of earners and child poverty from 1974 to 2010 more than doubled. The observed trends are even more pronounced for the anchored poverty measure than that of relative poverty. This offers further support of a noteworthy change in the association among marriage, work, and child poverty over time.
Beyond these results, three additional sensi- tivity analyses warrant mention. First, because the sample size increased more than threefold from 1986 to 1991, I conducted a robustness
check to ensure that changes in sample sizes were not affecting the results by reestimating the analyses with random samples of 4,200 per wave (the minimum N of all years). Second, to address the concern that marriage and number of earners were inherently conflated, I estimated all models without number of earners in the model. Third, I reestimated the analyses as linear probability models. The results for all these sensitivity anal- yses are consistent with the presented results.
Discussion
Motivated by the fundamental changes in mar- riage and work (i.e., household earners) over the past four decades, in this study, I exam- ined the extent to which the magnitude of the associations among marriage, work, and child poverty has changed. The regression results sug- gested that although marriage and work have both maintained a negative association with child poverty, marriage has generally become a weaker defense against child poverty, whereas work has become a stronger defense against child poverty. These results are consistent with
Marriage, Work, and Child Poverty 1175
a narrowing of the child poverty gap in unmar- ried and married households (see Figure 1) and a widening of the child poverty gap among no-earner, one-earner, and two-or-more-earner households (see Figure 2). Furthermore, these findings demonstrate the historically varying relationship of demographic characteristics and poverty over time and illustrate that sources of stratification in modern society are not static.
The overall weaker association between marriage and child poverty may be attributable to either or both of two factors: (a) the increased proportion of children in cohabiting households and (b) the increased heterogeneity among unmarried parents, which makes them less uni- formly disadvantaged. Conversely, the stronger association between work and child poverty may be attributable to either (a) economic changes that have created increased pressure for multi-earner households and/or (b) welfare changes that have made employment more essential for staying out of poverty. Although it is beyond the scope of this study to fully analyze these plausible explanations, additional descriptive characteristics of children in married and unmarried headed households shed some light (see the Appendix). Although the mean number of earners increased from 1.15 to 1.24 for children in unmarried-headed households, it decreased in married-headed households from 1.90 to 1.81. Both households with married and unmarried heads of households had increases in total weekly and annual work hours in the house- hold and the education of the head of household. Particularly noteworthy, however, is that among children in unmarried households, the percent- age of heads of household with less than a high school degree decreased from 45.31% in 1974 to 20.91% in 2010, whereas the percentage with 4 or more years of college increased from 7.29% to 14.55%. These findings illustrate the changing characteristics of both married and unmarried households, providing possible insight into this study’s findings.
There are several areas in which future research would be valuable. First, although this study has provided plausible explanations for the changing associations among marriage, work, and child poverty, future research to test the extent to which these mechanisms account for the observed trends is necessary. For example, longitudinal data would allow for the examination of families with children over time to explore potential causal mechanisms
(e.g., cohabitation, welfare supports, etc.) that may help explain the outcomes observed in this cross-sectional data. (Unfortunately, LIS data do not enable precise over time comparisons of cohabitation because cohabitation data are not available or are not precisely identified in some years.)
Second, this study concerned only 10 time points. Replication of the analyses with data for consecutive years could reveal important nuances in results unobserved in this study. Third, given the distinctively higher poverty among Black and Latino children and racial differences regarding changes in family struc- ture and work patterns (e.g., Lichter & Crowley, 2004; Nichols, 2013), future research to exam- ine possible differences by race is necessary. In a sensitivity analysis, I reestimated the final model for Whites, Blacks, and Latinos. Although the results reflected similar general trends across races, there were notable differences in the non- linear analyses, such as the negative association between work and poverty being strongest for Blacks but not consistently increasing for Lati- nos. Thus, the role of racial/ethnic differences warrants more thorough inquiry.
Finally, this study has implications for antipoverty policies. The observed weakened relationship between marriage and child poverty is notable for a society that has placed a rather strong emphasis on marriage and “family val- ues” in its antipoverty policy (Cherlin, 2009). Although it is clear that marriage still has a strong, positive impact on children’s economic well-being, perhaps policymakers have relied too heavily on marriage alone. As Haskins (2014) noted, “the changes in family composi- tion have been proceeding for more than four decades and show no signs of abating, despite a host of efforts by policy makers.” Moreover, almost half of unmarried parents would have continued to earn below the federal poverty line even if they were to marry (Sigle-Rushton & McLanahan, 2002), and earnings have a positive influence on marriage selection, in particular among low-income mothers (Gibson-Davis, 2009; McLanahan & Percheski, 2008). These findings, coupled with the present study, suggest that focusing on increasing parents’ earn- ings will not only help improve children’s economic well-being, but perhaps also increase the probability of financially stable marriages, which in turn could further reduce child poverty.
1176 Journal of Marriage and Family
That the results indicate the negative asso- ciation between work and child poverty has strengthened should also inform policy debates. The analyses demonstrate how essential employ- ment is for the economic security of fami- lies with children. Rainwater and Smeeding (2004) found that employment, the labor mar- ket, and work supports (e.g., work tax cred- its, child care subsidies) have the largest impact on a country’s child poverty rate, yet this has been relatively neglected in American poverty policy discussions. Instead, emphasis on work has often focused on simply getting people to work and cultivating their work ethic. How- ever, it is equally important to have policies that help families maintain employment. Because single-parent households are more vulnerable to child poverty and such families are less likely to have multiple earners, it is essential to facilitate the gainful and secure employment of single par- ents. Furthermore, even though employment is essential, it does not guarantee an escape from poverty, given that working-poverty rates remain relatively high in the United States (Brady et al., 2013). Therefore, policies designed to boost wages for typical workers and expand work sup- ports (see Sawhill & Karpilow, 2014) would also help to effectively reduce child poverty.
Note
I gratefully acknowledge David Brady and Linda Burton for their valuable feedback and suggestions and Ryan Finnigan for providing helpful statistical advice. I also thank the fol- lowing for comments on previous versions of this article: Amie Bostic, Sandra Danziger, Lane Destro, David Eagle, Lisa Keister, Sancha Medwinter, Cyrus Schleifer, S. Philip Morgan, Feng Tian, and audiences at the annual meetings of the Southern Sociological Society (2012) and the Asso- ciation for Public Policy Analysis and Management (2013). This research was supported by the American Sociological Association Minority Fellowship Program and the Center for Child and Family Policy at Duke University.
References
Amato, P. R., & Maynard, R. A. (2007). Decreas- ing nonmarital births and strengthening marriage to reduce poverty. The Future of Children, 17, 117–141.
Becker, G. (1981). A treatise on the family. Cam- bridge, MA: Harvard University Press.
Bianchi, S. M. (2000). Maternal employment and time with children: Dramatic change or surprising con- tinuity? Demography, 37, 401–411. doi:10.1353/ dem.2000.0001
Bluestone, B., & Harrison, B. (2001). Growing pros- perity: The battle for growth with equity in the
twenty-first century. Los Angeles: University of California Press.
Brady, D. (2003). Rethinking the sociological mea- surement of poverty. Social Forces, 81, 715–751. doi:10.1353/sof.2003.0025
Brady, D., Baker, R., & Finnigan, R. (2013). When unionization disappears: State-level unioniza- tion and working poverty in the U.S. American Sociological Review, 78, 872–896. doi:10.1177/ 0003122413501859
Brown, S. L. (2010). Marriage and child well-being: Research and policy perspectives. Journal of Mar- riage and Family, 72, 1059–1077. doi:10.1111/j. 1741-3737.2010.00750.x
Bureau of Labor Statistics. (2012). BLS spotlight on statistics: The recession of 2007–2009. Retrieved from http://www.bls.gov/cps/cpswom2012.pdf
Bureau of Labor Statistics. (2013). Highlights of women’s earnings in 2012. Report 1045, Bureau of Labor Statistics, Washington, DC. Retrieved from http://www.bls.gov/spotlight/2012/recession/
Chen, W., & Corak, M. (2008). Child poverty and changes in child poverty. Demography, 45, 537–553. doi:10.1353/dem.0.0024
Cherlin, A. J. (2009). The marriage-go-round. New York: Knopf.
Christopher, K. (2005). A “pauperization of mother- hood”? Single motherhood and women’s poverty over time. Journal of Poverty, 9, 1–23. doi:10. 1300/J134v09n03_01
Danziger, S. K. (2010). The decline of cash welfare and implications for social policy and poverty. Annual Review of Sociology, 36, 523–545. doi:10. 1146/annurev.soc.012809.102644
Deaton, A. (2006). Measuring poverty. In A. Baner- jee, R. Benabou, & D. Mookerjee (Eds.), Under- standing poverty (pp. 3–16). New York: Oxford University Press.
Edin, K., & Kissane, R. J. (2010). Poverty and the American family: A decade in review. Journal of Marriage and Family, 72, 460–479. doi:10.1111/j. 1741-3737.2010.00713.x
Ellwood, D. T., & Jencks, C. (2004). The uneven spread of single-parent families: What do we know? Where do we look for answers? In K. M. Neckerman (Ed.), Social inequality (pp. 79–118). New York: Russell Sage Foundation.
Fry, R. (2012). No reversal in decline of mar- riage. Washington, DC: Pew Research Center. Retrieved from http://www.pewsocialtrends.org/ 2012/11/20/no-reversal-in-decline-of-marriage/
Gibson-Davis, C. M. (2009). Money, marriage, and children: Testing the financial expectations and family formation theory. Journal of Marriage and Family, 71, 146–160. doi:10.1111/j.1741-3737. 2008.00586.x
Gornick, J., & Jäntti, M. (2012). Child poverty in cross-national perspective: Lessons from the Luxembourg Income Study. Children and Youth
Marriage, Work, and Child Poverty 1177
Services Review, 34, 558–568. doi:10.1016/ j.childyouth.2011.10.016
Handler, J. F., & Hasenfeld, Y. (2007). Blame welfare, ignore poverty and inequality. New York: Cam- bridge University Press.
Haskins, R. (2014). Poverty and opportunity: Begin with facts (testimony to the U.S. House of Representatives Committee on the Bud- get). Retrieved from http://www.brookings. edu/research/testimony/2014/01/28-poverty- opportunity-begin-with-facts-haskins
Kalleberg, A. L. (2007). The mismatched worker. New York: W.W. Norton.
Lichter, D. T. (1997). Poverty and inequality among children. Annual Review of Sociology, 23, 121–145. doi:10.1146/annurev.soc.23.1.121
Lichter, D. T., & Crowley, M. L. (2004). Welfare reform and child poverty: Effects of maternal employment, marriage, and cohabitation. Social Science Research, 3, 385–408. doi:10.1016/j. ssresearch.2003.09.001
Livingston, G. (2013). The rise of single fathers: A ninefold increase since 1960. Washington, DC: Pew Research Center.
Luxembourg Income Study (LIS). http://www. lisdatacenter.org. (United States; [August 2012– December 2014]). Luxembourg: LIS.
Manning, W. D., & Brown, S. L. (2006). Children’s economic well-being in married and cohabit- ing parent families. Journal of Marriage and Family, 68, 345–362. doi:10.1111/j.1741-3737. 2006.00257.x
McLanahan, S., & Percheski, C. (2008). Family struc- ture and the reproduction of inequalities. Annual Review of Sociology, 34, 257–276. doi:10.1146/ annurev.soc.34.040507.134549
Miller, A. R. (2010). The effects of motherhood tim- ing on career path. Population Economics, 24, 1071–1100. doi:10.1007/s00148-009-0296-x
Mishel, L., Birens, J., Gould, B., & Shierhotz, H. (2012). The state of working America, 12th edition:
An economic policy institute book. Ithaca, NY: Cornell University Press.
Nichols, A. (2013). Explaining changes in child poverty over the past four decades. Low Income Working Families Discussion Paper 2, The Urban Institute, Washington, DC.
Rainwater, L., & Smeeding, T. M. (2004). Poor kids in a rich country: America’s children in comparative perspective. New York: Russell Sage Foundation.
Raley, K. R. (2000). Recent trends and differentials in marriage and cohabitation: The United States. In L. J. Waite & C. Bachrach (Eds.), The ties that bind: Perspectives on marriage and cohabitation (pp. 19–39). New York: Aldine de Gruyter.
Sawhill, I., & Karpilow, W. (2014). A no-cost pro- posal to reduce poverty and inequality. Brief No. 51, Center on Children and Families, Brookings Institute, Washington, DC.
Seccombe, K. (2000). Families in poverty in the 1990s: Trends, causes, consequences, and lessons learned. Journal of Marriage and the Family, 62, 1094–1113. doi:10.1111/j.1741-3737.2000. 01094.x
Sigle-Rushton, W., & McLanahan, S. (2002). For richer or for poorer? Marriage as an anti-poverty strategy in the United States. Population, 57, 509–526. doi:10.3917/popu.203.0519
Thomas, A., & Sawhill, I. (2005). For love and money? The impact of family structure on income. The Future of Children, 15, 57–74.
U.S. Department of Commerce. (2011). Women in America: Indicators of social and economic well-being. Report prepared for the White House Council on Women and Girls, Washington, DC. Retrieved from https://www.whitehouse.gov/ administration/eop/cwg/data-on-women
Wilde, E. T., Batchelder, L., & Ellwood, D. T. (2010). The mommy track divides: The impact of child- bearing on wages of women of differing skill lev- els. Working Paper No. 16582, National Bureau of Economic Research, Cambridge, MA.
1178 Journal of Marriage and Family
APPENDIX Variable Means for Children by Marital Status
Variable 1974 1979 1986 1991 1994 1997 2000 2004 2007 2010
Panel A: Married-headed household Relative poverty (%) 9.34 10.99 12.42 14.76 14.02 12.95 13.56 12.49 12.91 11.66 Anchored poverty (%) 9.34 10.03 10.60 12.84 12.48 9.64 8.62 8.12 7.52 7.41 Number of earners 1.90 2.01 1.95 1.94 1.94 1.95 1.96 1.87 1.86 1.81 Weekly work hours 57.13 64.07 71.44 67.77 67.94 65.26 62.35 63.27 60.25 Annual full-time hours 2441.03 2711.06 3001.40 2998.95 2887.88 2752.26 2820.14 2643.09 Annual part-time hours 254.89 330.35 347.06 322.89 281.53 273.62 266.57 274.96 Number of working-age adults 2.22 1.98 2.19 2.23 2.22 2.21 2.24 2.23 2.24 2.28 Age of head 38.43 38.23 38.04 38.45 39.03 39.42 39.35 39.54 40.10 40.45 Head under age 25 (%) 5.87 4.36 2.56 2.43 2.30 1.96 2.75 2.65 2.23 2.00 No HS diploma. (%) 27.34 22.65 15.91 15.87 15.15 15.07 14.39 13.44 12.52 12.08 College degree (%) 19.17 23.98 25.47 26.24 28.50 28.81 30.48 32.44 34.98 37.23 Black head (%) 7.37 9.00 8.20 8.57 8.38 8.96 9.07 8.36 8.48 7.96 Other race head (%) 1.22 2.32 3.03 3.84 4.16 5.07 5.56 7.52 7.79 8.39 Latino head (%) 5.04 7.40 8.51 11.39 12.51 13.98 15.56 17.91 19.24 19.19 Number of kids 2.31 2.55 2.11 2.42 2.39 2.41 2.45 2.42 2.43 2.42 Adults over 65 (%) 3.01 1.13 2.54 2.41 2.58 2.85 3.11 3.60 3.77 4.23 N 3,844 4,711 3,251 16,138 15,493 13,377 12,747 23,469 22,049 20,772 Panel B: Unmarried-headed households Relative poverty (%) 47.78 50.04 50.92 51.91 49.24 45.49 42.45 40.51 42.47 40.38 Anchored poverty (%) 47.78 48.29 47.50 48.82 46.62 39.74 31.61 31.40 31.49 31.37 Number of earners 1.15 1.18 1.15 1.12 1.17 1.27 1.41 1.29 1.32 1.24 Weekly work hours 26.93 33.70 38.23 35.06 38.95 42.14 39.41 40.04 36.85 Annual full-time hours 1004.76 1299.73 1468.95 1591.56 1761.79 1642.35 1671.02 1485.33 Annual part-time hours 180.18 204.75 220.07 215.36 214.42 197.93 204.16 228.71 Number of working-age adults 1.33 0.97 1.44 1.52 1.53 1.53 1.60 1.60 1.61 1.70 Age of head 37.70 37.42 37.27 37.04 36.84 37.44 37.15 37.60 38.02 37.78 Head under age 25 (%) 11.86 9.95 8.82 9.15 10.68 10.72 12.37 11.84 10.67 11.40 No HS diploma. (%) 45.31 39.44 29.89 31.00 28.73 27.73 23.99 23.61 21.39 20.91 College degree (%) 7.29 7.36 11.17 8.84 8.79 9.58 10.95 12.07 13.62 14.55 Black head (%) 30.54 35.97 30.57 34.52 34.25 31.91 31.52 29.80 29.49 28.57 Other race head (%) 1.13 1.80 1.87 3.03 2.38 3.17 4.07 5.76 6.00 6.54 Latino head (%) 8.03 7.87 12.51 14.08 15.38 16.20 17.55 20.37 21.46 23.45 Number of kids 2.29 2.61 2.01 2.46 2.43 2.44 2.31 2.33 2.34 2.31 Adults over 65 (%) 5.46 3.36 4.13 4.45 4.47 4.68 4.93 5.35 5.53 5.73 N 767 1,296 1,046 6,091 6,134 5,628 5,374 10,276 10,065 10,055
Note. There were no data available for weekly hours in in 1979 and annual hours in 1974 and 1994. This is why the alternative work measures are not included in the main analyses of this article. However, to compare trends among children in married versus unmarried households over time, the work hour data that are available are displayed in the table. HS = high school.