Income effects of divorce in families with dependent children

DOI 10.1007/s00148-005-0029-8

ORIGINAL PAPER

Espen Bratberg . Sigve Tjøtta

Income effects of divorce in familieswith dependent children

Received: 7 November 2002 / Accepted: 19 April 2005© Springer-Verlag 2005

Abstract The Norwegian post-divorce transfer scheme for families with childrenguarantees minimum maintenance payments from the non-custodial to thecustodial parent. We studied the economic effects of divorce in such families.When the mother has custody, she falls behind before maintenance payments. Afteradding these, the outcomes at the median are very similar for both parents,contrasting studies from other countries, but the risk of a drop in equivalizedincome is large.

Keywords Marital dissolution . Income changes . Child custody

JEL Classifications J12 . D31 . I38

1 Introduction

For several years, divorce rates have been rising throughout the developed world:in 1970, the Organization for Economic Cooperation and Development (OECD)average number of divorces per 100 marriages was 14.3, and in 1998, it hadincreased to 41.2. Marital breakups may seriously affect the living conditions ofthose involved, particularly in families with children. In the past few decades, thishas spurred research in the social sciences on the economic consequences ofdivorce, in particular on gender differences, and on comparing the net effects ofdifferent types of social insurance and welfare programs across countries. Thispaper adds a study of the income effects of divorce in Norway, a countrycharacterized by a high degree of female labour force participation, a quiteextensive post-divorce income package for divorcees with child custody and highdivorce rates. Focussing on the interplay of gender and child custody, we use anequivalence scale that takes into account the time spent by non-custodians with

Responsible editor: Junsen Zhang

E. Bratberg . S. Tjøtta (*)Department of Economics, University of Bergen, Herman Fossgt. 6, 5007 Bergen, NorwayFax: +47-55589210, E-mail: [email protected], [email protected]

J Popul Econ 21: (2008)439–461

/ Published online: 10 January 2006

their children. We also suggest an alternative measure of economic well-beingbased on earnings capacity rather than realized earnings.

The main finding of previous research is that women experience seriousreductions in post-divorce income, whereas the effects of the split are less dramaticfor their ex-husbands. Smock (1993) noted that the economic costs of divorce inthe United States were as severe in the 1980s as in the 1960s and 1970s.Burkhauser et al. (1990, 1991) compared the United States with Germany, usingdata from the Panel Study of Income Dynamics and the German Socio-EconomicPanel, respectively. Burkhauser et al. (1990, 1991) found that women and childrenare worse off than men in both countries, and that the inequalities in Germanyfollowing a divorce are at least as great as in the US, despite the more extensivesocial welfare system in Germany. Jarvis and Jenkins (1999), using the BritishHousehold Panel Survey, found that “although the tax-transfer system mitigates thedifferentials in outcomes between husbands, on the one hand, and wives andchildren on the other, significant differentials remain.” In another US study,Bianchi et al. (1999), using data from the late 1980s and early 1990s (the Survey ofIncome and Program Participation), found that wives who retain custody ofchildren after separation have needs adjusted income levels at about 56% of theirformer husbands’ income levels. Other studies suggesting that differing post-divorce outcomes for men and women include Duncan and Hoffman (1985) in theUS and Finnie (1993) in Canada.

Norway, the subject of this study, had an estimated 46.8 divorces per 100marriages in 2000, above the OECD 1998 average of 41.2 and Denmark’s 37, butbelow Finland (57) and Sweden (65).1 Like the other Scandinavian countries,Norway is characterized by a low degree of economic inequality (Atkinson et al.1995; Aaberge et al. 2000). Social insurance and benefits are also generous. As inmany other European countries, there are general child allowances that are notmeans-tested. After divorces, custodians are supported by cash allowances as wellas tax deductions. Moreover, child support is enforced by the public authorities: theminimum level of allowances is law-regulated, and in the absence of payment, thegovernment pays the minimum allowance to the custodial parent and takesresponsibility for collecting the debt from the other party. This feature makes theNorwegian system stand out, although it, in other respects, bears close re-semblances to the post-divorce policies of other European countries. The presentanalysis uses data on more than 60,000 individuals, drawn from a large lon-gitudinal database constructed by combining information from several publicregisters.

We first use these data to compare economic outcomes for divorced familiesaccording to custody status, with intact families as the benchmark. Starting withraw earnings and successively taking into account the transfers described above,we get an impression of how the welfare state affects the economic well-being ofdivorced couples. To assess the full effect of divorce, income dynamics also need tobe addressed. Any household observed at two points in time may improve or

1 Sources: OECD, Society at a glance 2001, Underlying data, Annex G: General contextindicators, http://www.oecd.org, and Statistics Norway, http://www.ssb.no. In 1998, theNorwegian rate was at the OECD average, but in 1980–1995, it was higher. The number ofdivorces per 100 marriages for a given year is the percentage of divorces that would result during60 years of marriage at the current divorce rates.

E. Bratberg, S. Tjøtta440

http://www.oecd.org

http://www.ssb.no

worsen its economic standing, and the question is to what degree such changes areaffected by a marital split. We therefore compare “poverty dynamics”, i.e. the riskof crossing a low-income threshold, by marital and custody status. The tendency toincrease or decrease income in absolute terms, and to move up or down in theincome distribution in relative terms, is also considered.

Common measures of economic well-being are usually based on realizedincomes and transfers, and we follow this approach. However, one weakness withsuch measures is that the value of household production is effectively set to zero.We therefore supplement the analysis by drawing from Haveman and Bershadker(1998, 2001), who suggested using earnings capacity, defined as the earnings ahousehold could obtain if all its adult members were working full time in the labourmarket, when measuring poverty. The underlying concept is that the most needyare those who are incapable of generating a minimum income. Although poverty isnot the focus of this study, the earnings capacity concept has relevance, as a publicpolicy objective towards divorced families is to avoid unfavourable outcomes forthe party with the largest needs.

A substantial part of our analysis is to assess the effect of welfare state in-terventions by comparing pre- and post-transfers incomes. These comparisons areonly suggestive: in the absence of governmental intervention, families and in-dividuals might have chosen other alternatives with respect to labour force par-ticipation, as well as marital status. Furthermore, the probability of obtaining childcustody may be endogenous to the likelihood of divorce. A full analysis of thedistributional effects of the transfers system, with endogenous divorce and custodyprobabilities, is outside the scope of this paper—as in the received literature quotedabove.2

The remainder of this paper is structured as follows. The second sectionprovides an account of the method used in this study. The third section describesthe data and some additional institutional details, and the fourth section containsthe results and discussion. The final section provides some concluding remarks.

2 Method

The aim of the analysis is to assess the distributional effects of divorce on coupleswith children. We focus on the effects of the total tax and transfer system bycomputing and comparing several income measures:

A = Post-tax earnings before any child-related transfers;B = A + ordinary child allowances;C = B + extra advantages for custodial parents (details in Section 3);D = C + child support from the non-custodial to the custodial parent.

2 For analyses of the effects of expected income and transfers on divorce, see, e.g. Johnson andSkinner (1986), Moffitt (1990), Hoffman and Duncan (1995), Konrad and Lommerud (1995),Nixon (1997) and Tjøtta and Vaage (2002).

Income effects of divorce in families with dependent children 441

Thus, we get an impression of the “general effect” of the welfare state (B), theeffects of public transfers directed especially towards divorced families (C), andfinally of transfers between ex-spouses (D). Our income measures C and D arecomparable to Jarvis and Jenkins’ (1999) net income concept and Burkhauseret al.’s (1990, 1991) post-tax, post-transfer incomes.

We perform these computations for divorced and intact families, using grosstaxable earnings as inputs. The outcomes are then compared by gender and maritaland custody status. We also assess income dynamics, summarized as transitions toand from a low-income group and by percentage income change. Finally, theanalysis is extended by computing and comparing A–D based on estimatedearnings capacities.

2.1 Family income and equivalence scales

The purpose of equivalence scales is to make incomes comparable for householdsof different sizes. Generally, expenditures increase by the number of householdmembers, but not linearly. That is, we expect the needs of two adults to be morethan the needs of one, but not twice as much. Accordingly, there are someeconomies of scale in household expenditure. It is also reasonable that the ex-penditure of adding an extra child is not the same as adding an extra adult. We takecare of these aspects by using a parametric scale where household income isdivided by

E ¼ Aþ �wCð Þ�; (1)

where A denotes the number of adults in the household (2 if married, 1 ifdivorced), and C denotes the number of children. w is a weight that allows child

expenditures to be lower than expenditures for adults. θ∈[0, 1] catches economiesof scale. As θ approaches 0, the economies of scale increase; θ=1 indicates noeconomies of scale. As an extension, we also take into account that the children ofdivorced couples spend some time with the parent without custody, by using theadditional weight �, where �=1 for those still married, �=�cust for divorcedindividuals with custody and �=�noncust=1−�cust for divorced individuals withoutcustody.

The scale defined by Eq. 1 is a variant of that suggested by Buhmann et al.(1988), modified by the parameters w and τ. Our default choices of w (0.75) and θ(0.5) are the same as in a Norwegian governmental report on transfers to familieswith children (NOU 1996:13 1996). θ=0.5 (the “square root scale”) is alsocommonly used in international studies. The parameter �, to our knowledge, isunique to this study. Attributing some weight to children for divorced individualswithout custody contrasts sharply to the approach in which all weight is given tothe parent with custody, which makes the custodian seem worse off, and the non-custodian better off. This change should be larger for the non-custodian because ofthe economies of scale parameter. In the analysis, we base our choice of � on a


survey by Statistics Norway (1996), which found that, at the median, children indivorced families spent 7 days per month with non-custodial fathers and 8 withnon-custodial mothers. We converted this to fractions and use τnoncust=0.23 for menand 0.26 for women.3 Thus, we assume a correspondence between this timefraction and expenditures. However, there may be non-linearities due to in-cremental fixed costs, e.g. related to extra space and bedrooms, that are notincluded.

Clearly, equivalence scales cannot offer anything but an approximation to theeffect of household size on real-world expenditures. No universally agreed-uponequivalence scale exists: two broad categories are “real-world” scales, based onactual expenses, and parametric scales. Buhmann et al. (1988) considered in-equality and poverty in 10 countries empirically and investigated the effect of usingdifferent equivalence scales. Coulter et al. (1992) considered the properties ofdifferent scales from a theoretical point of view and recommended three strategies,as there is no consensus about the appropriate scale: (i) checking the resultsempirically, (ii) using “alternate methods to examine income distributions forheterogeneous populations” or (iii) “decomposing overall distributional compar-isons” (p. 1081). Banks and Johnson (1994) and Jenkins and Cowell (1994) tookthe discussion further. Triest (1998), in a discussion of trends in poverty rates in theUnited States, showed that different alternatives “could make a major difference inthe ‘official’ poverty rate and could also have a large impact on the composition ofwho is defined as ‘poor’” (p. 113). It is difficult to find hard evidence for a specificparameterization, and we follow Coulter et al.’s first strategy and check thesensitivity of our main results to changing parameter values. This is similar to theapproach taken by Jarvis and Jenkins (1999), except they used a single-parameter(θ) scale. We also consider the sensitivity to this parameter, but place an emphasison the sensitivity to τ.

2.2 Estimating earnings capacity

The full or potential income of an individual is the income that could be earned byworking full time in the labour market (Becker 1965). Assume that the expectedpotential income (in logs) of individual i is

E Y�i jXi

� � ¼ � 0Xi; (2)

where Xi denotes human capital, β is a vector of coefficients and the asteriskindicates that full income is only observed for individuals who actually work fulltime. With an estimate of β in hand, we may obtain an estimate of E(Yi*|Xi). Wemodel this as a standard sample selection process, where the selection is onworking full time. Xi contains education, experience and experience squared. Thecovariates in the selection equation include X in addition to variables that areassumed to affect the decision to work full time, but not income once the decisionis made.

3 For male non-custodians, there was some discrepancy between reports from the mother and thefather; the (non-custodial) fathers themselves reported 8 days per month. We use the lowernumber in the analysis.


The estimate of β is obtained using Heckman’s (1979) two-step estimator. Aswe are interested in the distribution of income, not just its mean, we followHaveman and Bershadker (1998, 2001) by adding a random component to thepredicted income to account for earnings variation that is not explained by themodel. Thus, the evaluation of the effects of public transfers for those who workless than full time is based on

~Yi ¼ Yi þ �i; (3)

where Yi is estimated potential outcome, and ηi is a pseudorandom draw from thenormal distribution with mean zero and variance equal to the estimated variance inthe income equation. For those who actually work full time, we use the observed Yi.The potential earnings of a household are the sum of the potential earnings of itsindividual (adult) members. We use the same equivalence scale as in the previoussection.

3 Data and institutional details

Our data are drawn from a longitudinal database (“KIRUT”) which links in-dividual level information from several Norwegian public registers for a 10%random sample of the population aged 16–67 years. The database covers the period1989–1996. Records are at the individual (not household) level, but spouses’income and year of birth may be identified. KIRUT does not include surveyinformation.

Our sampling strategy is to define a population at risk of divorce at some pointin time, observe their earnings, and then observe earnings again after a fixedperiod, conditional upon marital and custody status. Thus, the results show earn-ings dynamics over this interval for all marital outcomes.4 We chose 1989, thefirst year in the database, as our starting year and observed the outcomes in 1994.The population at risk is married couples with children in that period. Because thedata are at the individual level, we sampled married individuals in KIRUT withchildren born after 1976, i.e. younger than 18 years in 1994. This restriction waschosen because child-related family transfers are terminated at the age of 18years.

After exclusions due to missing variables, our sample consisted of 29,245women and 29,140 men.5 In 1994, when the income comparisons are made, 2,531women and 2,424 men were divorced. We can readily identify custody statusbecause divorced individuals without custody are recorded as not having children.By comparing post- and pre-divorce child status, we thus indirectly infer status

4An alternative would be to follow individuals until marital breakup and observe earnings atsome fixed interval after divorce. One advantage of our approach is that it allows a comparison ofdivorcees and non-divorcees. A disadvantage is that we may obscure some post-divorce earningsdynamics.5We excluded 290 women and 282 men who subsequently re-married, as they are not entitled tothe same transfers as those who do not and would have to be treated separately. Presumably, theiroutcomes would be most similar to those who stay married, as they do not necessarily suffer thesame loss of spousal income.


with respect to custody. The available income information is gross taxable ownearnings and spouse earnings. As noted in the previous section, we apply the taxrules on these earnings, and add general child benefits and extra benefits forcustodial divorcees where applicable. Finally, child support from the custodian isadded or subtracted. Means-tested social assistance is not taken into account, asonly a binary indicator was available in the data. Compared with survey-basedstudies, we rely somewhat more on imputed transfers: this is discussed further laterin the article. On the other hand, our income data are obtained from the tax registersand, in principle, should have no measurement problems. Moreover, for a countrywith only some 4.5 million inhabitants, the sample is large compared with otherstudies.

For the potential income analysis, a “full-time worker” is defined as havingyearly earnings above NOK 150,000 (women) and 165,000 (men).6 It follows fromthe construction of the database that about 10% of the individuals in the sample aremarried to each other, but we cannot identify who is married to (or divorced from)whom. The available spouse information is age and gross earnings; the earningsinformation is also available for ex-spouses.

Table 1 shows background characteristics by marital status and gender. We notethat the average education of men is slightly higher than that of women, and that theaverage education of those who divorce is lower than that for those who staymarried. As one would expect, the earnings of women in 1989 was lower than thatof men.

The Norwegian law on child custody establishes one of the parents as custodianand the other as non-custodian. Until recently (including the period covered in ourdata), joint custody was ruled out as an outcome flowing from judicial conflict, andonly if both parents agreed. Choice of custodian should be guided by what is in thebest interest of the children. Obviously, in most cases, that judgement is hard tomake for a third party. The law therefore takes into account observable featuressuch as earnings capacity and which parent spends most time at home.

The main features of the child support and maintenance payments system are asfollows. Ordinary child benefits for children under the age of 18 years are paid toall parents. In addition, divorced families get benefits for one additional child (e.g.a divorced mother with custody of two children gets the same child benefits as amarried couple with three children). In married families, the benefits are paid to themother: in divorced families, to the custodial parent. The custodian also gets anextra tax deduction and cash benefits to cover childcare. Moreover, there is an extrabenefit for lone parents without labour income, applicable to custodians.7 Childsupport from the non-custodian to the custodian is negotiated privately based onthe non-custodial parent’s income, but enforced and guaranteed by the publicauthorities. In 90% of cases, the standard (minimum) rate is applied: 11% of grossincome for one child, 18% for two, 24% for three and a maximum of 28% for four

6We had to base the “full-time” definition on earnings due to lack of reliable hours information.The chosen limits approximately correspond to the average earnings of hospital orderlies (men)and unskilled nursery workers (women).7 The purpose of this “transition benefit” is to support lone parents who are unable to uphold workdue to childcare responsibilities, or who are undertaking education to improve their employ-ability. The benefit may be combined with a small amount of labour earnings, which are thendeducted. Child support from the non-custodial parent is also deducted. This benefit is subject totaxation.


Tab

le1Descriptiv

estatistics

Wom

enMen

All

Married

Divorced

All

Married

Divorced

Custody

Not

custody

Custody

Not

custody

Variable

Mean

SD

Mean

SD

Mean

SD

Mean

SD

Mean

SD

Mean

SD

Mean

SD

Mean

SD

Educatio

n,1989

11.2

2.3

11.3

2.3

10.9

2.1

10.9

2.3

11.8

2.7

11.8

2.7

11.4

2.5

11.2

2.4

Experience,1994

12.9

6.2

13.0

6.2

11.7

6.1

12.4

6.2

20.2

5.2

20.3

5.1

20.0

4.9

18.6

5.7

Age,1989

34.2

6.1

34.4

6.1

31.9

6.0

34.4

5.7

36.9

6.6

37.1

6.6

36.5

5.9

34.9

6.6

Spouse’sage,1989

37.0

6.6

37.1

6.6

35.2

6.6

37.5

6.8

34.4

6.1

34.5

6.1

32.8

6.0

32.5

6.2

Children<18,1989

2.1

0.8

2.1

0.8

1.9

0.8

2.0

0.8

2.1

0.8

2.1

0.8

2.1

0.8

1.9

0.8

Children<18,1994

2.0

0.9

2.0

0.9

2.0

0.8

1.8

0.8

2.0

0.9

2.0

0.9

2.0

0.8

1.9

0.8

Gross

earnings,1989

89.5

72.8

89.5

72.6

87.8

72.3

93.7

85.8

233.1

110.6

234.6

109.8

231.7

114.0

212.2

119.3

Spouse’sgrossearnings,1989

231.3

115.6

232.8

114.7

211.7

117.0

232.4

148.0

87.9

74.7

88.0

74.8

89.9

71.7

85.6

73.3

Spouse’sgrossearnings,1994

245.0

265.5

268.3

266.4

––

––

113.2

123.3

123.5

123.7

––

––

Num

berof

individuals

29,245

26,714

2,120

411

29,140

26,716

645

1,779

Educatio

nismeasuredin

years

Experience:

yearswith

incomeabovebasiccountin

gunitin

social

insurancesystem

Earningsandspouse’searnings

in1,0001994

NOK


children or more.8 The paid/received amount is deductable/taxable. According toThe Ministry of Children and Family Affairs (1996), about 85% of non-custodianspay the agreed child support regularly. The public enforcement (with pre-paymentand subsequent debt collection in case of initial non-payment) of the privateagreement between the parents is a quite special feature of the Norwegian system.In 2002, the system was reformed. In the future, the earnings of both parents, alongwith actual child-related costs, are taken into account when calculating the supportlevel. Interestingly, in the UK, the Child Support Pensions and Social Security Act2000 takes steps in the opposite direction, from a cost-based to an income-basedsystem.

4 Results and discussion

4.1 Actual incomes

Table 2 shows different equivalence-scaled income measures for 1994 by genderand family status. The top panel shows household earnings after taxes, but beforeany child-related transfers, adjusted by the equivalence scale. The next two panelsshow the effect of first adding general child benefits, and then extra transfersdirected at custodians. The fourth panel adds the effect of maintenance payments.The last line shows individual pre-tax earnings before adding spousal income forthose married and before applying the equivalence scale. We note that allindividuals are married with children at the outset, but men and women are not(necessarily) married to one another, due to the construction of the database. Whencomputing equivalence-scaled incomes, the “burden” of divorced couples’children is split between the parents according to custody status and the parameterτ, as discussed in Section 2.2.

First, we find that the household incomes for married men and women aresimilar, indicating that using individual data to assess household entities gives afairly good approximation. As expected, Table 2 shows that divorced householdsare worse off than those that stay married. Divorced men have lower pre-taxearnings than married men, whereas divorced women without custody earn morethan married women. The post-tax, pre-transfer equivalence incomes in the toppanel (Income A) show that the burden is typically worsened by the loss of thespouse’s income, particularly for women. The gradual adding of general childbenefits (Income B) and special transfers for divorced families (Income C) showsthat, without these transfers, women with custody would be made worse off thanothers in terms of equivalence income. Standard child benefits are not enough toequalize women with custody to those without. After adding extra advantages forcustodians (cf. Section 3), women with custody are better off than women without,

8 The authorities guarantee a minimum amount of support per year (NOK 11,760 in 1994). If thenon-custodian’s income is insufficient to provide the minimum amount by applying the standardrate, the difference is paid by the government. In these cases, the support payments are decideddiscretionally, typically below the standard rate, with zero payments for those with very lowincomes. In our calculations, we approximate this practice by adjusting the standard rate (q) asfollows. Let a denote the minimum amount and m=a/q. For non-custodial income (y) less than0.5 m, we set payments to 0. For 0.5m≤y<m, we adjust the rate to q(y−0.5m)/0.5m. The receivedsupport is calculated as max(a, qy).


but their median income is still below the median of men without custody.However, the net result after including maintenance payments (Income D) is thatwomen without custody are clearly worse off, even though their pre-tax unscaledearnings (in the first row) are higher than those of married women. The medianequivalence incomes of women with custody and men without are almost identical,indicating the fairly successful re-distribution of income. However, the 10th and90th percentiles reveal larger income dispersion for men. The situation for women

Table 2 Effects of transfers on equivalence scaled 1994 household incomes (1,000 NOK)

Women Men

Married Divorced,custody

Divorced,not custody



Pre-tax Mean 124 123 132 268 242 227Income A Mean 144 62 81 143 111 131

10thcentile

78 0 0 77 30 17

50thcentile

140 67 90 140 110 135

90thcentile

209 116 145 210 178 206

Income B Mean 155 76 81 154 126 13110thcentile

91 14 0 90 46 17

50thcentile

151 81 90 151 126 135

90thcentile

218 128 145 219 189 206

Income C Mean 155 106 81 154 150 13110thcentile

91 67 0 90 86 17

50thcentile

151 107 90 151 147 135

90thcentile

218 149 145 219 213 206

Income D Mean 155 118 73 154 162 11510thcentile

91 79 0 90 100 17

50thcentile

151 118 80 151 159 117

90thcentile

218 160 128 219 223 181

N 26,714 2,120 411 26,716 645 1,779

Pre-tax: Unscaled individual pre-transfer earnings (without spouse)Income A = Post-tax household earnings, no child-related transfersIncome B = A + general child benefitsIncome C = B + extra advantages for custodiansIncome D = C ± maintenance payments from/to ex-spouse


without custody is not as good, with median net income at about two-thirds of themedian for women with custody. The fact that the 10th percentile for this group isestimated to 0 must be seen in connection with social welfare payments not beingincluded.

Table 2 gives a static picture of post-divorce outcomes, without any referenceto the pre-divorce economic situation. Tables 3 and 4 show income dynamicsassociated with marital separations. In both tables, we consider equivalence-scalednet incomes after all transfers (Income D). We first consider poverty, or lowincome, dynamics. Table 3 shows transitions during 1989–1994, where the low-income group is defined as earning less in terms of equivalized net income than theminimum social security benefit (NOK 51,174 for a single person in 1989, NOK60,700 in 1994).9 The 1989 incomes have been computed applying the sameequivalence scale formula as for 1994. The 1994 incomes are after all transfers,corresponding to the final panel in Table 2. Clearly, the inflow to low income islowest for those who stay married. For women who experience a marital split, butretain custody, the inflow is at a similar level, whereas the inflow rate for womenwithout custody is more than 30%. For men without custody, inflow is lower. Thisdifference may be explained by a larger fraction of divorced women having lost theincome of a spouse. We also note the large outflow from poverty among custodialwomen, more than 90%. This may seem surprising but is mostly due to transfersaimed at custodians: if we base the 1994 definition of low income on Income B (notreported), the outflow is reduced to 42%. As demonstrated in Table 2, average pre-tax incomes are very similar for married and custodial women. Thus, the outflowfrom low-income does not result from changes in labour supply.

In Table 4, we take a broader view of income dynamics, addressing the extent ofchange in all parts of the distribution, as well as changes in the relative position ofthe distribution. All results relate to equivalence incomes after all transfers (IncomeD). The upper panel of Table 4 shows percentage net income changes. Amongthose who stay married, more than 80% increase their household earnings. Incontrast, almost 50% of custodial women and non-custodial men experienceincome drops. This is not the case for men with custody, whose income devel-opment is similar to that of intact families. Again, the trend is worse for womenwithout custody, where a large majority have more than a 10% drop. As before, thisis driven by the loss of spousal income. It should also be noted that social assistanceis not included. In the lower panel, we have summed up changes in relative positionwhen the income distribution is split into deciles. Clearly, there is a higher tendencyto move down for divorcees (except custodial men), in particular for womenwithout custody. This group also has the lowest probability of moving up theincome ladder. Furthermore, the risk of a downward movement is larger for womenwith custody (66.9%) than for men without (57%). On the other hand, men withcustody have a more than 50% probability of improving their position in theincome distribution.

9 For countries without an official poverty threshold, half of the median income is often used as a“poverty line”. We have chosen the politically decided minimum pension partly because it givesan idea of what is officially accepted as a minimum, but also because the median income in ourselected sample would not be representative for the population.


4.1.1 Sensitivity analysis

Our results obviously depend on the parameters in the equivalence scale. Forexample, the choice of the economies-of-scale parameter, θ = 0.5, may be con-troversial. No less important is the weight attributed to the non-custodial parent,τ noncust. We therefore perform a sensitivity analysis on these parameters, illustratedin Figs. 1, 2, 3, 4. We show the effect of equivalence-scaled incomes before child-related transfers (Income A in Table 2) and after all child-related transfers andmaintenance payments (Income D).

Figure 1 shows the effect on median incomes of women with, and of menwithout, custody when τ noncust is varied, keeping θ constant at the default value 0.5.Recall from Table 2 that the median net incomes (Income D) of these groups arealmost identical. We see that this result is quite dependent on τ noncust. If non-custodial men spend less time with their children than assumed in our calculations,the picture of the economic outcome is changed in their ex-spouses’ disfavour. Ifτ noncust = 0, custodial women have about 80% of median equivalence income ofnon-custodial men, demonstrating how sensitive these results are to how the timechildren spend with each ex-spouse is treated. Some of the differences in our resultcompared with those of others may be explained by our treatment of this custodialarrangement. The curves cross at about τ noncust = 0.21, which is incidentally veryclose to the value chosen for non-custodial men in our calculations (0.23, based onsurvey information). Moreover, there are no values for τ noncust that equalize menand women before transfers (Income A). Figure 2 shows the effects on the“opposite” constellation, men with and women without custody. Unsurprisingly,τ noncust has no effect on the ranking of men andwomen in terms of the net result (D).

Figures 3 and 4 present a similar exercise for the economies-of-scale parameterθ, keeping τ noncust constant. We find that this parameter, too, is critical for the

Table 3 Income transitions 1989–1994 by marital and custody status 1994

Women Men

All Married Divorced,custody

Divorced,notcustody


Divorced,notcustody

Transitions from low income 1989L–NL 72.9% 70.9% 93.3% 25.0% 65.4% 66.9% 76.5% 51.7%L–L 27.1% 29.1% 6.7% 75.0% 34.6% 33.1% 23.5% 48.3%Observations 1,405 1,180 193 32 1,291 1,110 34 147Transitions from non-low income 1989NL–NL 97.2% 97.6% 96.7% 69.1% 96.9% 97.5% 97.1% 87.0%NL–L 2.8% 2.4% 3.3% 30.9% 3.1% 2.5% 2.9% 13.0%Observations 28,077 25,534 1,927 379 27,849 25,606 611 1,632

NL: Equivalence scaled household income after all transfers (Income D)≥minimum socialsecurity benefit for a single personL: Equivalence scaled household income after all transfers (Income D)<minimum social securitybenefit for a single person


ranking of women with and men without custody. For values of θ indicating lesseconomies of scale (closer to 1) than assumed in the main analysis, the outcome ofmen without custody (Fig. 3) appears better. Actually, the ranking is reversed atabout θ = 0.52. Figure 4 shows that, again, the ranking of net incomes is notaffected by the parameter value when the man has custody; the median woman isalways worse off.

The most critical parameter is perhaps τ noncust because there is probablyconsiderable variation among divorced couples. We have already noted that, forvalues below the one chosen, outcomes appear less favourable for divorcedwomen. This value corresponds to the median reported for men in the StatisticsNorway survey (7 days per month). The 10th and 90th percentiles are 2 and 11days, respectively. To assess the effect on income dynamics on letting τ noncust varyin this range, we have recalculated the figures in Table 4 for divorced families atvalues corresponding to 2 and 11 days per month (0.07 and 0.36).10 The results areshown in Table 6. For example, if men without custody spend only 2 days permonth with their children, the fraction of ex-wives calculated to experience anincome drop increases from about 50 to 57%, and the corresponding number fornon-custodial men is reduced from about 48 to 40%. The fraction moving downin the income distribution increases by about four percentage points for custodialwomen and is reduced by about six percentage points for non-custodial men.

Table 4 Percentage income change and change in relative position in the income distribution1989–1994, by marital and custody status 1994

Women Men


Divorced,notcustody


Divorced,notcustody

Percentage income change50+% drop 2.6 2.0 3.1 36.0 2.9 2.1 2.5 14.510–50% drop 10.6 8.6 30.5 44.5 9.6 8.7 7.8 24.10–10% drop 6.7 5.9 16.0 5.6 6.0 5.7 5.1 9.80–10% increase 10.2 10.0 13.6 4.4 10.0 10.0 10.1 10.310–50% increase 48.2 50.9 22.1 6.3 48.9 50.8 35.8 24.850+% increase 21.8 22.6 14.6 3.2 22.6 22.6 38.7 16.5

100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Change in relative position (decile)Down 36.5 33.5 66.9 81.0 36.5 35.2 32.6 57.0Same 24.7 25.7 13.6 14.4 25.7 26.3 16.3 19.6Up 38.8 40.8 19.5 4.6 37.8 38.5 51.2 23.4

100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0N 29,245 26,714 2,120 411 29,140 26,716 645 1,779

Equivalence scaled household income after all transfers (Income D)

10 Another approach would be to estimate τ as a function of individual characteristics and then useimputed values in the analysis. However, that is not possible as the background information in thesurvey is not the same as in our sample.


Still, the main impression is that divorced couples experience dramatic incomedrops, but the notion of “equalized burdens” (when the woman has custody) ismodified.

Fig. 2 Sensitivity of median household income to the non-custodial time parameter. Men withcustody vs women without custody. 1,000 NOK, equivalence scaled

Fig. 1 Sensitivity of median household income to the non-custodial time parameter. Women withcustody vs men without custody. 1,000 NOK, equivalence scaled


Fig. 3 Sensitivity of median household income to the economies of scale parameter. Women withcustody vs men without custody. 1,000 NOK, equivalence scaled

Fig. 4 Sensitivity of median household income to the economies of scale parameter. Men withcustody vs women without custody. 1,000 NOK, equivalence scaled


4.2 Potential incomes

Wenow turn to comparisons based onpotential incomes.11Asdiscussed inSection 2.2,these are estimates of the household incomes that would be obtained if the parentsworked full time, adjusted by the same equivalence scale as in the previous section.The results may be interpreted as earnings capacities or, alternatively, as placing avalue, namely the market wage, on household production. Thus, we may supple-ment our picture of different family types’ consumption possibilities.

Table 5 shows the same equivalence-scaled income measures as Table 2, butbased on potential earnings.12 The unscaled pre-tax measure in the last line showslower mean earnings capacity for women than for men. Subtracting taxes andrunning the result through the equivalence scale show that, before child-relatedtransfers, women with custody are worse off than men without. However, when theroles are reversed, the “Income A” measure is closer for custodians and non-custodians. Adding child allowances and extra advantages improves the relativestanding of parents with custody, and the final result (Income D) is that the rankingof the median female custodians and male non-custodians is reversed. This followsfrom transfers between the ex-spouses; however, when only child allowances andspecial advantages are included (Income C), the median income of divorced femalecustodians is still NOK 22,000 (15%) below their male counterpart’s withoutcustody. We also note that, according to the potential income measure, femalecustodians are better off relatively to male non-custodians than when thecomparison is based on realized incomes. Sensitivity analysis similar to the onepresented in Figs. 1, 2, 3, 4 (not reported) shows that these results are somewhatless sensitive to the parameters in the equivalence scale than results based on actualincomes (reported in Appendix, Tables 6 and 7).

4.3 Discussion

The primary aim of the Norwegian system of post-divorce transfers is to secure theeconomic well-being of children in divorced families, using public as well asprivate transfers. An important feature of the system is the public enforcement ofprivate transfers within divorced families. To the extent that the welfare of the childis synonymous with the welfare of the parent with custody, the policy is largelysuccessful. In the typical situation, where the mother obtains custody, the medianequivalence-scaled net income (D) is a more than 40% improvement from the post-tax income before extra advantages for custodians (B). For the minority who livewith the father, this is even more the case: after scaling, the net incomes in thesehouseholds are at the same level as in complete families. This comes at the expenseof the parent without custody, however, and particularly if this parent is a woman.

11 The estimated earnings equations are reported in Appendix, Table 7. The selection terms turnedout to be insignificant. However, replacing the predictions with ordinary least significance (OLS)predictions hardly affected the results. For spouses, the potential income is based on averagepredictions grouped by age in 5-year intervals, as we have information on spouses’ age, but nottheir education and experience.12We have also examined the effect of using predicted incomes (including the pseudorandomcomponent) for all individuals, including those who actually work full time by our definition. Theeffect is modest, compressing the distribution of net incomes (D) slightly, mostly for men.


The median net income of a woman without custody is only 53% of the equivalentincome in a complete family. Consequently, the children in these families spendsome of their time with a parent with seemingly severe economic difficulties.

It cannot be claimed that the economic consequences of divorce are totallyequalized between men and women (assuming that the woman has custody). Eventhough the risk for custodial women of a transition into the low-income group is

Table 5 Effects of transfers on equivalence scaled 1994 estimated household earnings capacities(1,000 NOK)

Women Men





Pre-tax Mean 196 196 200 297 282 282IncomeA

Mean 177 97 121 179 128 16010thcentile

136 72 93 138 87 114

50thcentile

174 93 118 172 119 149

90thcentile

224 125 152 227 184 219

IncomeB

Mean 189 110 121 191 143 16110thcentile

150 89 93 151 104 114

50thcentile

184 107 118 182 133 149

90thcentile

233 136 153 236 198 219

IncomeC

Mean 189 131 121 191 166 16110thcentile

150 111 93 151 125 114

50thcentile

185 127 118 182 156 149

90thcentile

233 158 153 236 222 219

IncomeD

Mean 189 148 107 191 179 14110thcentile

150 128 80 151 138 98

50thcentile

185 144 104 182 170 131

90thcentile

233 174 137 236 234 193

N 26,714 2,120 411 26,716 645 1,779

Pre-tax: Unscaled individual potential earnings capacity (without spouse)Income A = Post-tax potential household earnings, no child-related transfersIncome B = A + child benefits transfersIncome C = B + extra advantages for custodiansIncome D = C ± maintenance payments from/to ex-spouse


smaller, the overall risk of a downward movement in the income distribution islarger. Nonetheless, these differences seem less dramatic, particularly whencompared with intact families, where the risk of an income drop is less than 20%.

The potential income measures give an upward adjustment of the consumptionpossibilities for all groups, without changing their relative rankings in terms of netincomes. Women with custody at the median have almost 80% of the equivalentincome of intact families, about the same as when looking at realized incomes. Onthe other hand, the ratio of the potential income of the “untypical” men withcustody to the income of married families is reduced. We interpret this reduction asreflecting the larger value of household production in intact families. In the typicaldivorced family, the ex-husband is the loser, with about 70% of the median incomeof married families, and 90% of the income of divorced women with custody. Oneway to interpret the analysis involving potential income is that measures based onactual incomes somewhat underrate the economic well-being of individuals whodo not work full time. Because women are over-represented among those, therelative standing of custodial women and married couples are improved whencompared, based on potential income. However, there is an additional issue whenpotential incomes are used to indicate economic well-being: parents with childrenbelow school age working full time incur childcare costs. Unless these are splitbetween divorced parents according to the time each one spends with the children(τ), it seems likely a larger fraction of the cost will be borne by the custodial parent.In that case, a marginal increase in equivalent income as calculated by our scalewill be less worth for a custodian than for a non-custodian.

As borne out in the sensitivity analysis, the ranking of outcomes dependscritically on the parameter measuring the time children spend with the parentwithout custody. Our results may be too optimistic regarding these custodialwomen. On the other hand, this sensitivity provides an important insight, as theusual approach when using equivalence scales is to compare divorcees by givingno weight to the child for the non-custodian, thereby, by necessity, making thecustodian seem worse off. However, it is also clear that the outcomes for manywomen without custody are severe, regardless of the equivalence scale parameters.We have also seen that modest changes of the scale parameter (θ) in the direction ofless economies of scale tend to turn around of custodians and non-custodians.

Compared with previous studies, the gender differences found here (again,assuming that the wife obtains custody) are smaller and more in the favour of(custodial) women. More specifically, Burkhauser et al. (1991) found that, in theUS, 10% of divorced men and 22% of divorced women experience drops in post-tax post-transfer income of 50% or more. The corresponding figures for Germanyare 8 and 33%. The results in Table 4 show that 3.1% of custodial women and14.5% of non-custodial men have income drops of that magnitude. Jarvis andJenkins (1999) reported that the median divorced husband has a 2% increase in thecorresponding income measure, whereas divorced wives suffer an 18% drop at themedian. Table 4 indicates a small positive change for custodial women and non-custodial men as well (with 50.3 and 51.6%, respectively, having incomeincreases). The differences in results may have been affected by differences in dataand methods: we base our results on calculated transfers, whereas the survey-basedstudies use reports of actual transfers. We also have a longer observation periodthan these other studies. The sensitivity analysis shows that the assumption thatthere is a cost associated with children also for divorcees without custody is critical.


Nevertheless, we think the results reflect system differences. In particular, we haveseen that the child support, guaranteed by the government, has the effect of almostbringing custodial women on a par with non-custodial men. Some caution must beexercised: as noted in Section 3, 85% of the non-custodians pay support regularly,implying that 15% do not. In these cases, our measure of net income (Income D)underestimates the economic well-being of non-custodians. Jarvis and Jenkins(1999), in their principal calculations, did not deduct fathers’ maintenancepayments, but in a sensitivity analysis, they found that this is of less importance forthe results. The explanation is that few liable non-custodial parents actually paysupport, and the amounts paid are not large. In our analysis, we found thatmaintenance payments play a crucial part in equalizing the burdens of divorce;recall from Section 3 that in the Norwegian system, the payments are sizeable, atleast 11% of the non-custodian’s gross income. Combined with the fact that mostdo pay, we believe that the results in Tables 2, 3, 4 reflect the usual situation.

There are some additional caveats. We have not been able to include socialassistance benefits—thus, the full effect of the welfare state is not reflected in ournet income measure, and the actual outcome in the lower end of the distributionmay be better than our results suggest—nor do we have access to financial wealthand income, housing wealth and expenditures. In the earnings capacity cal-culations, it may be argued that, because childcare expenses are not included, thepotential incomes of people with and without child custody are difficult tocompare. Another problem is that we have no information about cohabitation andcannot rule out that some divorcees have co-habitants, thereby possibly improvingtheir economic position. For the same reason, the effects of breaking up co-habitating unions are not addressed. However, we do not believe it likely that theseshortcomings would alter the study’s principal conclusions.

An interesting question is whether the Norwegian post-divorce transfer systemactually induces divorces. Our results show that the combination of private andpublic transfers and tax deductions dramatically improves the economic perfor-mance of divorced parents with child custody. It is tempting to conclude that thesystem has two side-effects: first, it stimulates marital breakup by reducing theassociated economic risk for the custodial part, and second, it stimulates conflictsover child custody because of the detrimental effects on the non-custodial party,especially in the case of a woman. To draw such conclusions, however, a simul-taneous analysis of divorce decisions, custody probability and labour market par-ticipation would be necessary.

5 Concluding remarks

A marital split may reduce the level of economic well-being of the familysignificantly, especially for couples with children. In this paper, we have studiedthe economic effects of divorce in such families in Norway, taking into account thefull “post-divorce package”, which includes special transfers to the custodialparent, as well as a governmental guarantee for the level and payment of childsupport from the non-custodian. We find that, in the typical case, where the motherhas custody, the mother (and child) falls behind before taking into account main-tenance payments. After taking into account maintenance, the outcomes at themedian are more similar, and certainly to a greater extent than found in previous


studies from the UK, US and Germany. However, we also find that, in the typicalcase, both parties have an almost 50% probability of an income drop, and a morethan 50% risk of moving down in the income distribution. Women without custodyare worse off by all measures, whereas the outcomes of men with custody are moresimilar to intact families.

We also supplement the measures of actual net incomes with an estimate ofearnings capabilities. This does not affect the ranking of the different groups, butthe standings of custodial women and married couples are improved. The reason isthat women have larger unused earnings capacities, whether they are married ordivorced. Thus, measures based on realized incomes may underrate the well-beingof individuals who do not work full time.

As in similar studies, we use equivalence scales to compare the incomes ofhouseholds of different sizes. We depart from these studies by assuming (based onsurvey information) that non-custodial parents spend some time with their childrenand adjust the equivalence scale accordingly. Sensitivity analysis shows that thisassumption is critical to the finding that the median net incomes of custodialmothers and non-custodial fathers are almost equal. On the other hand, it isimportant to point out that studies which do not take into account the time the non-custodial parent spends with the children may underrate the well-being ofcustodians relative to non-custodians. It is also the case that, for values of theeconomies-of-scale parameter slightly larger than our default, the estimatedoutcomes of custodial mothers are worse than those of their ex-partners.

With these reservations, the Norwegian system emerges from the analysisappearing fairly successful in equalizing the economic costs of divorce. At thesame time, the expected number of divorces per 100 marriages in Norway is almostat the upper quartile among OECD countries. One may well ask if there is a causalconnection between these two; however, any such conclusion is unwarranted, asthe current analysis has assumed that the probability of divorce is determinedexogenously.

Acknowledgements The authors would like to thank the Norwegian Research Council forfinancial support, Grant No. 140127/530 (“Velferdsprogrammet”). The paper was completedwhile Bratberg was visiting the University of York, whose hospitality is gratefully acknowledged.We would also like to thank two anonymous referees for their helpful comments. A previousversion of the paper was presented at the European Society for Population Economics (ESPE)2000 conference in Bonn. The data used in this study were provided by the Norwegian SocialScience Data Services (NSD). NSD is not responsible for the authors’ analyses.


Appendix

Tab

le6Sensitiv

ityof

incomedynamicsto

thetim

e-with

-non-custodial-parentparameter

(τ)

Divorcedwom

enDivorcedmen

Custody

Not

custody

Custody

Not

custody

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

Percentageincomechange

50+%

drop

3.6

3.1

2.8

32.6

36.0

37.0

2.6

2.5

2.4

13.8

14.5

15.4

10–50%

drop

37.4

30.5

23.4

40.1

44.5

46.4

10.1

7.8

6.8

16.9

24.1

29.4

0–10%

drop

16.0

16.0

15.7

7.7

5.6

4.6

7.8

5.1

3.7

8.9

9.8

11.1

0–10%

increase

11.8

13.6

14.4

6.3

4.4

3.6

10.7

10.1

8.0

9.2

10.3

8.1

10–50%

increase

17.9

22.1

27.2

7.7

6.3

5.6

39.4

35.8

36.3

27.7

24.8

23.2

50+%

increase

13.2

14.6

16.5

5.6

3.2

2.9

29.2

38.7

42.8

23.5

16.5

12.8

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

Changein

relativ

epositio

n(decile)in

theincomedistributio

nDow

n70.4

66.9

62.7

78.5

81.0

82.1

38.5

32.6

28.4

50.3

57.0

61.6

Sam

e14.5

13.6

14.2

14.0

14.4

14.5

17.8

16.3

16.1

18.3

19.6

19.7

Up

15.1

19.5

23.2

7.5

4.6

3.4

43.7

51.2

55.5

31.3

23.4

18.7

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

100.0

Equivalence

scaled

householdincomes

afteralltransfers(Incom

eD)

Low

:non-custodialparent

spends

2days/m

onth

with

child

ren(τ

noncust=0.07)

Med:non-custodialparent

spends

7/8(m

en/wom

en)days/m

onth

with

child

ren,

correspondingto

Table

4(τ

noncust=0.23/0.26)

High:

non-custodialparent

spends

11days/m

onth

with

child

ren(τ

noncust=0.36)


References

Aaberge R, Björklund A, Jäntti M, Pedersen PJ, Smith N, Wennemo T (2000) Unemploymentshocks and income distribution: how did the Nordic countries fare during their crises. Scand JEcon 102(1):77–99

Atkinson AB, Rainwater L, Smeeding T (1995) Income distribution in OECD countries: evidencefrom the Luxembourg Income Study (LIS). OECD Social Policy Studies, no 18. OECD, Paris

Banks J, Johnson P (1994) Equivalence scale relativities revisited. Econ J 104:883–890Becker G (1965) A theory of the allocation of time. Econ J 75:493–517Bianchi SM, Subaiya L, Kahn JR (1999) The gender gap in the economic well-being of non-

resident fathers and custodial mothers. Demography 36(2):195–203Buhmann B, Rainwater L, Schmauss G, Smeeding TM (1988) Equivalence scales, well-being,

inequality, and poverty: sensitivity estimates across ten countries using the LuxembourgIncome Study (LIS) database. Rev Income Wealth 34(2):115–142

Burkhauser RV, Duncan GJ, Hauser R, Berntsen R (1990) Economic burdens of maritaldisruptions: a comparison of the United States and the Federal Republic of Germany. RevIncome Wealth 36(4):319–333

Burkhauser RV, Duncan GJ, Hauser R, Berntsen R (1991) Wife or Frau, women do worse: acomparison of women and men in the United States and Germany after marital dissolution.Demography 28(3):353–360

Coulter FAE, Cowell F, Jenkins SP (1992) Equivalence scale relativities and the extent ofinequality and poverty. Econ J 102:1067–1082

Table 7 Earnings equations (two-step estimates)

Women Men

Coefficient Standarderror

z Coefficient Standarderror

z

Earnings equationEducation 0.030 0.001 20.19 0.050 0.001 51.59Experience 0.009 0.002 3.60 0.032 0.003 10.51Experience2/10 0.000 0.001 −0.26 −0.007 0.001 −8.67Selection term −0.003 0.013 −0.23 0.010 0.015 0.67Constant 11.702 0.047 251.25 11.570 0.034 335.97Selection equationEducation 0.191 0.004 50.72 0.181 0.004 41.79Experience 0.168 0.007 25.42 0.079 0.012 6.82Experience2/10 −0.025 0.002 −10.97 0.013 0.003 4.04Age −0.020 0.002 −11.58 −0.096 0.002 −38.93No. of children <18 −0.243 0.011 −21.16 −0.052 0.012 −4.45Custody −0.132 0.078 −1.69 0.111 0.070 1.59Married −0.421 0.074 −5.73 0.251 0.039 6.39Social assistance receiver −0.831 0.086 −9.60 −1.021 0.058 −17.73Spouse income/104 −0.003 0.001 −4.82 0.002 0.001 1.69Constant −2.472 0.118 −21.03 0.671 0.138 4.87N 29,245 29,140Selection=1 11,488 24,387


Duncan GJ, Hoffman SD (1985) Economic consequences of marital instability. In: David M,Smeeding TM (eds) Horizontal inequity, uncertainty and well-being. University of ChicagoPress, Chicago, pp 427–467

Finnie R (1993) Women, men, and the economic consequences of divorce: evidence fromCanadian longitudinal data. Can J Sociol Anthropol 30:205–241

Haveman R, Bershadker A (1998) Self-reliance as a poverty criterion: trends in earnings-capacitypoverty, 1975–1992. Am Econ Rev 88(2):342–347

Haveman R, Bershadker A (2001) The ‘Inability to be self-reliant’ as an indicator of poverty:trends for the U.S., 1975–1997. Rev Income Wealth 47(3):335–360

Heckman JJ (1979) Sample selection bias as a specification error. Econometrica 47(1):153–161Hoffman SD, Duncan GJ (1995) The effects of incomes, wages, and AFDC benefits on marital

disruption. J Hum Resour 30(1):19–41Jarvis J, Jenkins SP (1999) Marital splits and income changes: evidence from the British

Household Panel Survey. Popul Stud 53(2):237–254Jenkins S, Cowell F (1994) Parametric equivalence scales and scale relativities. Econ J 104:891–

900Johnson WR, Skinner J (1986) Labor supply and marital separation. Am Econ Rev 76(3):455–

469Konrad K, Lommerud KE (1995) Family policy with non-cooperative families. Scand J Econ

97(4):581–601The Ministry of Children and Family Affairs (1996) Høringsnotat. Forslag til endringer i

bidragsordningen etter barneloven (Note circulated for comment, on changes in child supportregulations). Oslo

Moffitt R (1990) The effect of the US welfare system on marital status. J Public Econ 41(1):101–124

Nixon LA (1997) The effect of child support enforcement on marital dissolution. J Hum Resour32(1):159–181

NOU 1996:13 (1996) Offentlige overføringer til barnefamilier (Public transfers to families withchildren). Norwegian governmental report, Oslo

Smock PJ (1993) The economic costs of marital disruption for young women over the past twodecades. Demography 30(3):353–371

Statistics Norway (1996) Undersøkelser om samværsrett (Surveys on child custody).Documentation report, Oslo

Tjøtta S, Vaage K (2002) Public transfers and marital dissolution. Working paper 0802,Department of Economics, University of Bergen

Triest RK (1998) Has poverty gotten worse? J Econ Perspect 12(1):97–114


Documents

Income effects of divorce in families with dependent children