Accounting for Trends in Child Support∗ Urvi Neelakantana a

Department of Agricultural and Consumer Economics University of Illinois at Urbana-Champaign 1301 W. Gregory Dr., 421 Mumford Hall Urbana, IL 61801 E-mail: [email protected] Tel: 1-217-333-0479 Fax: 1-217-333-5538 Abstract Beginning in 1975, the U.S. government passed several reforms to improve the child support system. Contrary to expectations, real child support income declined following the reforms, falling by nearly 38% between 1978 and 1987. This paper provides a simple method to account for historic trends in child support. Results show that 69% of the post-reform decline was due to a fall in fathers’ incomes. Subsequent increases can be attributed to improvement in compliance with awards. The results can be used to evaluate the success of the reforms.

Keywords: child support, children, public policy

This paper is adapted from a chapter of my Ph.D. dissertation at the University of Minnesota. I am especially grateful to Larry Jones for his advice. I thank Andrea Beller, Michele Boldrin, V.V. Chari, Partha Chatterjee, Zvi Eckstein, Mary Arends-Kuenning, Deborah Levison, Han Ozsoylev, Malik Shukayev, Shino Takayama and participants in the Growth and Development workshop at the University of Minnesota for helpful comments and suggestions. ∗

1

1

Introduction

Child support is a payment from the non-custodial parent to the custodial parent for the economic maintenance of their children.2 In the USA, each state has uniform guidelines to determine the child support award, i.e., how much child support the non-custodial parent owes. Guidelines to calculate child support awards first appeared in two states— Illinois and Maine—in 1975 (Beller and Graham, 1993) prior to which judicial discretion determined child support awards. Judicial decisions were often criticized for being inadequate, inconsistent, and unpredictable. Poor enforcement was blamed for the growing number of divorced mothers dependent on welfare programs. In response, Congress passed a series of laws to improve the child support system. Legislation in 1975 added Title IV-D to the Social Security Act, creating the Child Support Enforcement (CSE) program. The Office of Child Support Enforcement was subsequently established. The use of guidelines spread to all states with the passage of the CSE amendments of 1984, which required each state to formulate advisory guidelines. The amendments also aimed at improving enforcement by requiring child support to be withheld directly from the wages of delinquent payers. The 1988 Family Support Act made the guidelines presumptive; if judges deviated from the guidelines in awarding child support, they had to 2

Five out of six custodial parents in the USA are mothers (Grall, 2003). In the text, I use “mother” interchangeably with “custodial parent” and “father” with “non-custodial parent.”

2

provide written reasons for doing so. The Act also called for making wage withholding universal instead of restricting it to delinquent payers.3 Surprisingly, child support awards and income decreased in the period immediately following the reforms. Data from the April Child Support Supplement (CSS) of the Current Population Survey (CPS) shows that child support awarded to newly-divorced mothers fell from $3988.22 in 1978 to $2643.41 in 1987, a decline of 33.7%. Over the same period, average child support income fell by 37.6% from $2927.31 to $1826.90.4 Several papers have analyzed the reasons for the post-reform decline in child support income. Hanson, Garfinkel, McLanahan and Miller (1996) attribute the decline to a fall in awards. They find that awards in turn declined for three reasons. First, the demographic composition of the population changed to include a higher proportion of never-married mothers. Their partners usually had lower incomes and hence lower child support obligations. This brought down the average award amount. Hanson et al. (1996) find that a decrease in fathers’ incomes and inflation also played a role in reducing awards while mothers’ incomes did not. In contrast, Robins (1992) argues that since most states took the income of both parents into account while determining child support, an increase in mothers’ incomes was responsible for reducing child support awards. He also finds that inflation was 3

See Garfinkel, McLanahan and Robins (1994) for a detailed evaluation of the child support system as it evolved. 4 Unless otherwise specified, all amounts are expressed in real dollars using the Consumer Price Index with 1982-84=100.

3

responsible for reducing the real value of existing awards. Graham (1995) suggests that it was the real value of new awards that fell, probably due to the increase in mothers’ incomes relative to fathers’. Finally, analyzing trends between 1968 and 1997, Case, Lin and McLanahan (2003) find that child support declined when states shifted from bilateral to unilateral divorce laws, which reduced mothers’ bargaining power. Inflation, fertility decline, and decreases in men’s earnings may have also contributed. This paper introduces a simple method to account for the historical trends in child support and shed further light on the question. Using data from the CSS, results show that a fall in fathers’ incomes was responsible for over half the total decline in child support. Finally, applied to court-record data from Wisconsin, results show that improvement in compliance following the reforms contributed significantly to the increase in child support in that state.

2

Determinants of Child Support Income

The child support income that the custodial parent receives depends on the child support award and the extent to which the non-custodial parent complies with it. Since 1984, awards have been based on child support guidelines that each state has implemented. Although they differ across states, the guidelines follow one of three models. The Income Shares model, adopted by 35 states, estimates expenditures on children based on the sum of the parents’ incomes and the number of children. The expenditure is divided

4

between the parents in the ratio of their incomes. The non-custodial parent pays the custodial parent his share as child support. The Percentage of Income model, used in 13 states, computes the child support award as a percentage of the non-custodial parent’s income alone. The percentage increases as the number of children increases. The remaining five states use either the Melson formula or a hybrid model; both take the income of both parents into account. The models suggest the following relationship between the parents’ income, the number of children, and the child support award. Fathers’ Income: Child support awarded increases with the father’s income, but may do so at a decreasing rate. Mothers’ Income: In some states, the award also depends on the mother’s income. Given the parents’ combined income, the higher the mother’s share, the less child support the father owes her. However, an increase in the mother’s income also implies an increase in the parents’ total income. Since parents with higher total income are assumed to spend more on their children, the father may owe more child support. The net effect of an increase in the mother’s income on child support is hence ambiguous. Number of Children: Child support awarded increases with the number of children, but usually at a decreasing rate. Mothers with more children are thus likely to have higher total child support awards, but smaller awards per child. The paper proceeds under the assumption that judicial awards before 1984 were also based on the parents’ income and the number of children, but 5

Figure 1 about here

allows for the possibility that awards did not depend on the mother’s income.

3

Data

The data used in this paper comes from two sources. The pre-1984 period is analyzed using data from the April Child Support Supplement (CSS) of the Current Population Survey. Court Record Demonstration Project (CRD) data from Wisconsin is used to study the post-1984 period. Before turning to these data sets, the IPUMS-CPS (King et al., 2004) is used to construct Figure 1, which provides an overview of the long-term national trend in child support income of divorced mothers. It shows that real child support income declined during the reform period, reaching in lowest point in the mid-1980s before recovering. Note that the picture excludes mothers who may have been awarded child support but received zero payments because there is no variable in the data to identify such mothers. The April CSS began in 1979 and was repeated in 1982 and every alternate year thereafter. The CSS is one of the few data sets in which the child support award is known. In each survey year, respondents were asked to report how much child support they were supposed to receive and how much they actually received the previous year. To study new divorce cases, the sample is therefore restricted to mothers who divorced exactly two years

6

Figure 2 about here

prior to the survey year. The sample consists of an average of 130 divorced mothers per survey year. As Figure 2 shows, between 1978 and its lowest point in 1987, child support income fell by 37.6% from $2643.41 to $1826.90.5

3.1

Fathers’ Incomes

The child support award depends on the father’s income at divorce, about which information is not available because the mothers in the sample are divorced. Following previous work, (Miller, Garfinkel and McLanahan, 1997; Garfinkel and Oellerich, 1989), information on the mother’s characteristics is used to estimate the father’s income. The papers cited estimate divorced fathers’ incomes because their focus was on the father’s continuing ability to pay child support after divorce. The analysis here requires data on the father’s income prior to divorce to find its relationship to child support awarded. The procedure used to estimate fathers’ incomes is detailed below. Following Mincer (1974), given a male’s years of education, ef , and work experience, xf , earnings, yf , can be estimated using the following equation.

ln(yf ) = α + ρef + ζ1 xf + ζ2 x2f 5

(1)

New divorce cases were chosen for two reasons. First, when mothers who have been divorced longer were included in the sample, the decline in child support income was only 18%, suggesting that new cases drove the decline. Second, choosing new cases allows us to assume that the amount the mother was supposed to receive the previous year is the original court-ordered award.

7

As is standard, experience is derived from age, af , setting xf = af − ef − 6. The regression in (1) is run on data on married males from the IPUMS-CPS to derive the estimated value of the coefficients α, ρ, ζ1 , and ζ2 . Table 1 reports the values. Table 1 Coefficients of Mincer Regression (IPUMS-CPS data) Survey Year 1979 1982 1984 1986 1988

α (constant) 8.7390 8.7650 8.6440 8.4860 8.4200

ρ (education) 0.0709 0.0676 0.0725 0.0799 0.0869

ζ1 (experience) 0.0390 0.0382 0.0355 0.0392 0.0384

ζ2 (experience sq.) -0.0006 -0.0006 -0.0005 -0.0006 -0.0006

All coefficients are significant at the 1% level.

Assuming that matching is assortative on age and education, the mother’s age and education are used to estimate the father’s age and education at the time of divorce.6 Equation (1) can be written as

ln(yf ) = α + ρE(ef |em , d) + ζ1 [E(af |am , d) − E(ef |em , d) − 6] + ζ2 [E(af |am , d) − E(em |ef , d) − 6]2

(2)

where d signifies that the expectation is conditional on the couple divorcing. It thus allows for the possibility that age and education relationships could be different for couples who divorce. E(af |am , d) and E(em |ef , d) are estimated using a sample from the Panel 6

Mare (1991), for example, provides empirical evidence of assortative matching.

8

Study of Income Dynamics (PSID) consisting of females who were divorced in a given year but married in the prior survey year. Information on the husband is available for the prior year. The sample consists of 529 new divorce cases and yields

E(af |am , d) = 5.235 + 0.925am

(3)

E(ef |em , d) = 4.514 + 0.645em .

(4)

Substituting the values from (3) and (4) in (2) and using the coefficients in Table 1 yields the estimate of fathers’ incomes.

3.2

Mothers’ Incomes

Because of how the sample is chosen, mothers’ reported income is their income the year after their divorce. It is assumed that this is equal to their income the previous year, i.e., the year of their divorce.

3.3

Number of Children

Around 88% of the sample consists of mothers who were household heads. In such cases, the mother’s relationship to all the other members of the household is known and her own children can be identified. It is assumed that all her children who were under 18 were covered by the child support award. If the mother is not the head of the household, it is assumed that all children under 18 present in the household were covered by the award. 9

4

Accounting Method

For each individual i in year t, child support income s can be computed as

i sit = θti gt (yfi t , ymt , nit ) = θti gti .

(5)

Here θ is the ratio of amount paid (child support income) to amount due (child support award) and is referred to as “compliance” following the literature. The law at time t is denoted by gt , a function that calculates the child support award in terms of the father’s income, yf , the mother’s income, ym and the number of eligible children, n. Since there were no uniform child support guidelines between 1977 and 1984, two possibilities are considered: 1) gt was a function of only the father’s income and 2) gt was a function of both parents’ incomes. Table 2 reports the averages of fathers’ incomes, mothers’ incomes, number of children, child support awards, and child support incomes for the two years. All the variables except the mother’s income declined; the percentage changes are shown in the table. Equation (5) can now be used to determine the impact of each of these changes on child support. Appendix A describes how the particular functional form was chosen under each assumption. First assume that the child support award depended only on the father’s income and the number of children:

gti = β1 + β2 yfi t + β3 nit 10

(6)

Table 2 Average Values for Divorces in 1977 and 1984 Variable Father’s Incomea Mother’s Income Number of Children Child Support Award Child Support Income Compliance a

1977 1984 Percentage Change $23,756.67 $21,210.38 -10.72% $9,688.43 $10,722.23 10.67% 1.92 1.77 -9.38% $3,988.22 $3,231.45 -18.98% $2,927.31 $2,460.58 -15.94% 0.66 0.67

Father’s income is derived from the CSS using the method detailed in the text.

Regressing the above on the CSS data from 1979–1986 yields the coefficients in Table 3. Table 3 Coefficients of Regression (6) to Determine Child Support Award β1 Intercept -2662.68 (-3.014)

β2 Father’s Income 0.21 (5.516)

β3 Children 765.92 (4.729)

Numbers in parentheses are t-statistics.

Next, each year’s sample is divided into 20 roughly equal-sized groups.7 Denote the average earnings and average number of children within each group by yfj and nj respectively, with j = 1, 2, . . . , 20. The average child support award at time t is estimated to be

gt =

20 X

ηtj [β1 + β2 yfj t + β3 njt ]

(7)

j=1

7

The size of each group may differ slightly if the number of cases in that year was not divisible by 20.

11

Here ηtj is the fraction of people in each group j with

j j=1 ηt

P20

= 1. Using

equation (7), the change in child support between two periods is

∆g =

!

20 X

∆ηtj [β1 + β2 yfj t + β3 njt ]

j=1

+

20 X

ηtj β3 ∆njt

j=1

!

+

20 X

ηtj β2 ∆yfj t

j=1

!

(8)

+ higher order terms.

The terms in brackets measure the change in the award that would have taken place if, respectively, the weight of each group, the father’s income, and the number of children had changed. “Higher order terms” refers to the sum of the changes in child support that would result from changing the variables two at a time and three at a time. Average child support income at time t is estimated by

st =

20 X

ηtj θtj [β1 + β2 yfj t + β3 njt ]

(9)

j=1

where θtj is average compliance within each group. From equation (9), the change in child support income can be accounted for as follows.

∆s =

20 X

∆ηtj θtj [β1

+

β2 yfj t

+

!

β3 njt ]

j=1

+

20 X

ηtj ∆θtj [β1

+

β2 yfj t

+

!

β3 njt ]

j=1

(10) +

20 X j=1

ηtj θtj β2 ∆yfj t

!

+

20 X

ηtj θtj β3 ∆njt

j=1

12

!

+ higher order terms.

Now suppose that the child support award depended on both parents’ incomes. i gti = β1 + β2 yfi t + β3 nit + β4 ymt

(11)

Running the regression in (11) on the CSS data yields the coefficient values reported in Table 4. Table 4 Coefficients of Regression (11) to Determine Child Support Award β1 β2 β3 β4 -2535.13 0.19 790.92.51 0.02 (-2.85) (4.715) (4.845) (1.2) Numbers in parentheses are t-statistics.

The average child support award at time t in this case is

gt =

20 X

i ηtj [β1 + β2 yfj t + β3 njt + β4 ymt ]

(12)

j=1

The decomposition of the change in child support due is

∆g =

20 X

!

∆ηtj [β1 + β2 yfj t + β3 njt ]

j=1

+

20 X j=1

ηtj β3 ∆njt

!

+

20 X j=1

+ higher order terms.

13

+

j ηtj β4 ∆ymt

20 X j=1

!

ηtj β2 ∆yfj t

!

(13)

In this case average child support income is

st =

20 X

i ηtj θtj [β1 + β2 yfj t + β3 njt + β4 ymt ]

(14)

j=1

and the change in child support income is measured by

∆s =

20 X

!

∆ηtj θtj [β1 + β2 yfj t + β3 njt ]

j=1

+

20 X

!

ηtj ∆θtj [β1 + β2 yfj t + β3 njt ]

j=1

(15) +

20 X

ηtj θtj β2 ∆yfj t

j=1

!

+

20 X

ηtj θtj β3 ∆njt

j=1

!

+

20 X j=1

j ηtj θtj β4 ∆ymt

!

+ higher order terms.

5

Results

Table 5 summarizes the results from the accounting exercise under the assumption that the child support award depended only on the father’s income. Women who divorced in 1984 received, on average, $756.77 dollars less than women who divorced in 1977. Equation (8) estimates that the changes in parents’ income and the number of children would have caused awards would have fallen by $620.76. It thus accounts for 82% of the actual decline in awards. Similarly, equation (10) accounts for 80% of the actual decline in child support income. The results show that 64% of the fall in child support awards and 69% of the fall in child support income can be attributed to the

14

fall in fathers’ incomes. The next largest contributing factor is the decline in the number of children, which accounts for 15% of the decline in awards and 16% of the decline in payments. Table 5 Accounting—Award Depends on Father’s Income (1978-1984) Change in Award Change in Payments Due to change in Amount Percent Amount Percent Weight (η) 47.09 -6% 12.36 -3% Compliance (θ) N/A N/A 19.51 -4% Father’s Income (yf ) -485.39 64% -322.34 69% Number of Children (n) -115.19 15% -74.44 16% Higher Order Terms -67.26 9% -7.75 2% Estimated Change -620.76 82% -372.66 80% Unaccounted Change -136.01 18% -94.07 20% Actual Change -756.77 -625.59 The results under the assumption that awards depended on both parents’ incomes are reported in Table 6. The exercise accounts for 74% of the actual decline in awards and 70% of the decline in payments. The results show that the fall in fathers’ incomes contributed to 60% of the decline in awards and 63% of the decline in payments, while a fall in the number of children contributed to 15% of the decline in awards and 16% of the decline in payments. Though the specific magnitude varies slightly, under either assumption about the type of law, the decline in fathers’ incomes thus accounts for most of the decline in child support awards and payments and the decline in the number of children played the next biggest role. After 1984, each state adopted its own guidelines to determine child sup15

Table 6 Accounting—Award Depends on Both Incomes (1978-1984) Change in Award Change in Payments Due to change in Amount Percent Amount Percent Weight (η) 41.72 -6% 9.37 -2% Compliance (θ) N/A N/A 24.52 -5% Father’s Income (yf ) -444.61 60% -295.25 63% Number of Children (n) -118.95 15% -76.87 16% Mother’s Income (ym ) 31.94 -4% 24.02 -5% Higher Order Terms -66.80 -3% -10.53 2% Estimated Change -559.36 74% -324.75 70% Unaccounted Change -197.41 26% -141.98 30% Actual Change -756.77 -625.59 port awards. This implies that g, the function that estimates child support awards, varied by state. To study the period after 1984, court-record data from Wisconsin is used. Since the data is from a single state, its guidelines can be used to estimate a functional form for g.

5.1

Accounting with Wisconsin Data

Data on child support in Wisconsin is available from the Court Record Demonstration Project (CRD) sponsored by the Institute for Research on Poverty. The sample is from 21 counties in Wisconsin and consists of divorce cases involving minor children. Information from each court appearance is recorded for up to six years from the date of the initial petition. The advantage of the CRD over the CSS is that it reports both parents’ incomes at divorce. The sample chosen from the CRD consists of cases in which the father 16

was the payer and the mother had physical custody of all the children. Each case is followed for one year from the date of divorce or the date child support payments were ordered to commence (if this was after the divorce date). The average values of parents’ incomes, number of children, and child support income are reported in Table 7. Between 1985 and 1992, average child support income increased from $3162.25 to $3771.55. Wisconsin pioneered the child support reforms of 1984 so it is important to ask whether the increase in child support that followed was indeed due to changes in the law. Table 7 Average Values in Wisconsin, 1985 and 1992 Variable Father’s Income Mother’s Income Number of Children Child Support Payments Compliance

1985 1992 Percentage Change $18,008.47 $18,924.90 5.09% $10,952.54 $11,244.06 2.66% 1.71 1.85 8.19% $3,162.25 $3,771.55 19.27% 0.87 0.93

According to Wisconsin guidelines, child support is based only on the income of the non-custodial parent. The award is 17% of the non-custodial parent’s income for one child, 25% for two, 29% for three, 31% for four, and 34% for five or more children. Figure 3 shows a continuous estimate of the guidelines, whose functional form is derived from the line that best fits the points. The child support award in Wisconsin can thus be estimated by

gti = [β1 + β2 ln(nit )]yfi t

17

Figure 3 about here

where β1 = 0.1732 and β2 = 0.1032. The estimate of average child support income at time t is

st =

20 X

ηtj θtj [β1 + β2 ln(njt )]yfj t

j=1

Decomposing the change in child support income yields

∆s =

20 X

∆η j θj [β1 + β2 ln(nj )]yfj +

j=1

+

20 X

20 X

η j ∆θj [β1 + β2 ln(nj )]yfj

(16)

i=1

η j θj β2 ∆ ln(nj )yfj +

j=1

20 X

η j θj [β1 + β2 ln(nj )]∆yfj

j=1

+ higher order terms

Table 8 summarizes the results. The total increase in child support income was $609.30. The improvement in compliance accounts for the largest share at 47%. The improvement in compliance seen in the data is likely to have come from the amendments of 1984, which introduced automatic withholding of child support from the wages of delinquent payers. It appears that child support policy changes did have a significant role to play in the growth in child support income in Wisconsin. As other states adopted the amendments, such changes are likely to have been responsible for the growth in child support in the rest of the country as well. Data for each state from

18

1984 onwards is required to confirm this. Table 8 Accounting for Changes in Wisconsin (1985-1992) Due to Father’s Income (yf ) Number of Children (n) Compliance (θ) Weight (η) Higher Order Terms Estimated Change Unaccounted Change Actual Change

6

Change in Payments Amount Percent 124.29 20% 129.76 21% 288.27 47% 13.68 2% 82.72 14% 638.71 105% -29.41 -5% 609.30

Conclusion

What determinants of child support income have been responsible for the trends observed in the past 30 years? The question has often been asked in the literature and several reasons have been proposed. This paper provides a method to answer the question and to measure what fraction of the observed changes are due to each factor. The decline in awards between 1978 and 1984 can largely be attributed to the decrease in fathers’ incomes while improvements in compliance are likely to have contributed to the subsequent increase. The results have positive implications for child support policy, suggesting that the reforms were not responsible for the early decrease in child support and may in fact have contributed to the subsequent increase by improving 19

Figure 4 about here

compliance. A limitation of the exercise in this paper is that it excludes certain factors. The first of these is the change in divorce laws from bilateral to unilateral. Data shows, however, that child support declined even in states that continued to have bilateral divorce laws, so the question still remains.8 The exercise also does not account for the growing proportion of never-married mothers and the lack of inflation updating for existing awards. The data has shown, however, that child support declined even among divorced mothers with new awards. The results are able to account for much of the decline, at least for this group of mothers.

8

Figure 4 uses the IPUMS-CPS to plot average child support income in both groups of states—those that allowed unilateral divorce before 1976 and those that continued to have bilateral divorce laws. For a list of states in each group, see Gruber (2004). It shows that the decline in child support income took place under both types of laws and average child support was not systematically lower in states with unilateral divorce.

20

References Beller, A. H. and Graham, J. W. (1993), Small Change: The Economics of Child Support, Yale University Press, New Haven. Case, A., Lin, I.-F. and McLanahan, S. (2003), ‘Explaining Trends in Child Support: Economic, Demographic and Policy Effects’, Demography 40, 171–189. Garfinkel, I., McLanahan, S. S. and Robins, P. K., eds (1994), Child Support and Child Well-Being, Urban Institute Press, Washington, DC; Lanham, MD. Garfinkel, I. and Oellerich, D. (1989), ‘Noncustodial Fathers’ Ability to Pay Child Support’, Demography 26, 219–233. Graham, J. W. (1995), ‘A Comment on ‘Why Did Child Support Award Levels Decline From 1978 to 1985?’ by Philip K. Robins’, Journal of Human Resources 30, 622–632. Grall, T. S. (2003), Custodial Mothers and Fathers and their Child Support, Current Population Reports P60-225., U.S. Census Bureau. URL: http://www.census.gov/prod/2003pubs/p60-225.pdf Gruber, J. (2004), ‘Is making divorce easier bad for children? the long-run implications of unilateral divorce’, Journal of Labor Economics 22(4), 799–

21

833. URL: http://www.journals.uchicago.edu/doi/abs/10.1086/423155 Hanson, T., Garfinkel, I., McLanahan, S. and Miller, C. (1996), ‘Trends in Child Support Outcomes’, Demography 33, 483–496. King, M., Ruggles, S., Alexander, T., Leicach, D. and Sobek, M. (2004), ‘Integrated public use microdata series, current population survey: Version 2.0.’, [Machine-readable database]. Minneapolis, MN: Minnesota Population Center [producer and distributor]. Mare, R. D. (1991), ‘Five Decades of Educational Assortative Mating’, American Sociological Review 56, 15–32. Miller, C., Garfinkel, I. and McLanahan, S. (1997), ‘Child Support in the U.S.: Can Fathers Afford to Pay More?’, Review of Income and Wealth 43, 261–281. Mincer, J. (1974), Schooling, Experience, and Earnings, National Bureau of Economic Research; distributed by Columbia University Press, New York. Robins, P. K. (1992), ‘Why Did Child Support Award Levels Decline from 1978 to 1985?’, The Journal of Human Resources 27, 362–379.

22

A

Child Support Award Function

Figures 5–7 are used to make assumptions about the form of the function, g, which estimates how child support was awarded prior to 1984. They show how the child support award varied with one factor, controlling for the other two. To construct the figures, mothers and fathers are divided into income groups, where each group is an interval $5,000 in length. The largest group consists of fathers with between $15,000 and $20,000 in income and mothers with income less than $5,000. Figure 5 shows that the child support award increased linearly with the number of children in this group. In families with two children where mothers have less than $5,000, Figure 6 shows how the child support award changes with the father’s income. Each point on the graph plots the average income of fathers within an income quintile against the average child support award in that group. Similarly, Figure 7 is drawn for families with two children and fathers’ incomes in the $15,000–$20,000 range. It plots the average income of mothers in an income quintile against average child support income for that group. It appears that child support due increased linearly with the father’s income and the number of children, but did not vary systematically with the mother’s income. However, based on the previous discussion that the effect of the mother’s income on child support may have two components that cancel each other out, the exercise also allows for the possibility that the mother’s income influenced the award.

23

Figure 5 about here

Figure 6 about here

Figure 7 about here

24