Article
Journal of European Social Policy
2022, Vol. 0(0) 1–16
Weathering the storm together: © The Author(s) 2022
Does unemployment insurance help Article reuse guidelines:
couples avoid divorce? sagepub.com/journals-permissionsDOI: 10.1177/09589287221141363
journals.sagepub.com/home/esp
Dorian Kessler
Debra Hevenstone
Bern University of Applied Sciences, Bern, Switzerland
Leen Vandecasteele
University of Lausanne, Lausanne, Switzerland
Samin Sepahniya
University of Applied Sciences and Arts Northwestern Switzerland, Olten, Switzerland
Abstract
This study examines whether unemployment insurance benefit generosity impacts divorce, drawing on full
population administrative data and a Swiss reform that reduced unemployment insurance maximum benefit
duration.We assess the effect of the reform by comparing the pre- to the post-reform change in divorce rates
among unemployed individuals who were affected by the reform with the change in divorce rates among a
statistically balanced group of unemployed individuals who was not affected by the reform. Difference-in-
differences estimates suggest that the reform caused a 2.8 percentage point increase in divorce (a 25%
increase). Effects were concentrated among low-income couples (+58%) and couples with an unemployed
husband (+32%) though gender differences are attributable to men’s breadwinner status. Female main
breadwinners were more strongly affected (+78%) than male main breadwinners (+40%). Results confirm the
‘family stress model’ which posits that job search and financial stress cause marital conflict. Policymakers
should consider a broad array of impacts, including divorce, when considering reductions in unemployment
insurance generosity.
Keywords
Unemployment insurance, divorce, income, unemployment benefits, family stress, quasi-experiment
Introduction
Corresponding author:
Unemployment insurance was designed to alleviate Dorian Kessler, Department of SocialWork, Bern University of
the hardship of unemployment. Until the recent Applied Sciences, Hallerstrasse 8, Bern 3012, Switzerland.
COVID-19 pandemic there was a trend towards Email: dorian.kessler@bfh.ch
2 Journal of European Social Policy 0(0)
reducing unemployment insurance generosity across be biased given different trends in family formation
the OECD (Biegert, 2017; Gschwind, 2021; OECD, and dissolution by social strata (Cherlin, 2020).
2018). This was partly premised on findings that Second, a number of studies from the social policy
generous unemployment benefits extend unem- literature examine policy changes that include si-
ployment duration (Mortensen, 1976; for an over- multaneous variations on multiple policy dimensions
view see: Schmieder and Von Wachter, 2016). (for example, generosity and activation requirements)
In contrast, the literature on the effects of un- (Bitler et al., 2004; Halla et al., 2016; Van Winkle,
employment benefits has paid little attention to 2020). For these studies, it is difficult to attribute
outcomes such as health (Kuka, 2020; Shahidi and effects to specific policy changes. Third, some studies
Parnia, 2021) and subjective wellbeing (Carr and use relatively small experiments (Goñalons-Pons and
Chung, 2014; Rose, 2019; Wulfgramm, 2014). With Calnitsky, 2021), finding increases in discord due to
respect to families, there is just one study examining cash transfers but not divorce. This could simply be
divorce and fertility (Lindo et al., 2020). From the due to the small sample, as divorce is a rare event. In
comparative social policy literature we know that studies using survey data, sample size problems are
other welfare state programmes impact family sta- exaggerated due to panel attrition (Vandecasteele and
bility (Bitler et al., 2004; Halla et al., 2016; Van Debels, 2007).
Winkle, 2020), suggesting a likely, though insuffi- The study improves on each of these shortcomings.
ciently explored, relationship between unemploy- First, both our treated group and control group include
ment insurance and divorce. unemployed individuals, which allows for more re-
This article takes an important step in filling this liable comparisons of changes in divorce rates than
gap in our knowledge, using full population adminis- comparisons with continuously employed individuals.
trative data from Switzerland and a quasi-experimental Second, our design examines a policy change in
design to examine whether a reduction in maximum maximum benefit duration allowing us to isolate ef-
unemployment insurance benefit duration impacted fects of benefit generosity. Third, we use full pop-
the likelihood of divorce. We exploit a Swiss policy ulation linked register data that is not prone to panel
reform which reduced maximum benefit duration attrition, allowing us to detect even small changes in
from 1.5 years to one year. The reform increased divorce and examine heterogeneous effects by income
pressure to find a job and shortened unemployment group and gender. Given Switzerland’s high levels of
duration for individuals shortly before or after ex- average wealth (Kuhn, 2020) and generous social
hausting one year of benefits and reduced benefit assistance (Obinger, 1999), these results can be seen as
income for those exhausting one year of benefits. a ‘least likely’ case, meaning that if unemployment
We measured the effect of the reform on divorce by benefit generosity affects divorce in this context, it is
comparing the pre to post reform change in divorce likely relevant in any context.
rates among unemployed individuals affected by the
reform (treated) with the change in divorce rates Income shocks and divorce
among a statistically balanced group of unemployed
individuals who were not affected by the reform The literature on how income shocks, such as those
(controls). caused by the reform, affect divorce can be broadly
Our study offers key improvements over the lit- summarized into two groups: a) findings that are
erature on the positive effects of unemployment in- interpreted in support of the family stress model and
surance and the literature on social policy and family b) findings that are interpreted in support of the cost
stability. First, Lindo et al. (2020), the only other study and independence hypotheses.
on the effects of unemployment benefits on divorce,
examine the effect of changes in unemployment The family stress model
benefit generosity on divorce in the US, but compare
changes in divorce among the unemployed to those First, there is the argument that economic stability
without unemployment spells, a comparison that could contributes to marital stability. This is often
Kessler et al. 3
summarized as the family stress model and linked to unemployment benefit generosity, but not those of
the work of Rand Conger and Glen Elder (Conger unemployed women.
et al., 1990, 1994, 2010). The central argument is that
a lack of economic resources induces a chain of The cost/independence hypothesis
negative events for couples. With an income loss,
households might reduce expenditures, have diffi- A second set of arguments can be summarized as the
culties paying bills, or be unable to meet con- cost/independence hypothesis which stresses the role
sumption needs. This economic stress then translates of economic resources in a) helping couples bear the
into psychological stress (Inanc, 2018) through economic cost of separation and in b) fostering
conflicts related to finances (Goñalons-Pons and economic independence from a partner.
Calnitsky, 2021) eventually leading to low rela- The high cost of divorce is one potential reason
tionship satisfaction, and ultimately divorce (Blom why a negative income shock might depress divorce.
et al., 2019a, 2019b; Conger et al., 2010; Di Nallo Divorce is costly as the dissolution of the household
et al., 2022). into two new households leads to the loss of econ-
A gender-focused extension of this argument is omies of scale. Furthermore, divorce comes with
that conflict increases not only in response to income considerable direct costs such as legal fees (Boertien
loss but depending on whose income has been lost. and Lersch, 2021). A decrease in benefit income
Stress and conflict might be higher when gendered might thus lead couples who were considering di-
expectations are not met. An income loss of the male vorce to delay it.
partner indicates a failure to follow the male A second reason a negative income shock might
breadwinner norm (Cherlin, 2020; Townsend, 2010; depress divorce is explained by the economic in-
West and Zimmerman, 1987), whereas an income dependence hypothesis (Becker et al., 1977; Özcan
loss of the female partner is not at odds with prev- and Breen, 2012; Sayer and Bianchi, 2000). The key
alent norms. Thus male income loss, more than fe- argument is that marriage is attractive for those with
male income loss, translates into increased conflict limited individual resources as they can depend on
and marital instability (Bertrand et al., 2015; Blom the income support of a spouse. With no employment
et al., 2019b; Goñalons-Pons and Gangl, 2021; income, a decrease in benefit income could decrease
Inanc, 2018; Killewald, 2016; Kim and Luke, 2020). the unemployed individuals’ economic indepen-
The family-stress model and its gender-focused dence and reduce divorce (Bitler et al., 2004; Halla
extension are supported by some empirical evidence. et al., 2016). However, it must be noted that, when
Goñalons-Pons and Calnitsky (2021) show that cash seen from the perspective of a spouse who depends
transfers in the Manitoba Negative Income tax ex- on the unemployed spouse, a decrease in benefits
periments reduced money-related conflict in couples. could also increase divorce as the attractiveness of
Several longitudinal studies have shown that re- marriage is reduced.
ductions in men’s employment incomes lower rela- Several experimental and quasi-experimental
tionship satisfaction and increase divorce (Bertrand studies are interpreted as supporting the cost and
et al., 2015; Blom et al., 2019b; Di Nallo et al., 2022; independence arguments. Analysing the impact of
Killewald, 2016), while women’s incomes are less the US welfare reform in the 1990s, Bitler et al.
clearly related to divorce (Killewald, 2016; Özcan (2004) concluded that the introduction of policies
and Breen, 2012). This association is more pro- restricting access to benefits and promoting work
nounced in contexts with stronger gender norms (work requirements, sanctions, time limits, and
(Goñalons-Pons and Gangl, 2021). Boertien (2012) lowered income tax rates while on welfare) reduced
finds that income from lottery wins increases rela- the risk of divorce in states adopting these policies.
tionship satisfaction and reduces the chances of Groeneveld et al. (1980) showed that income
separation, but only if the male partner won. Simi- transfers in the Seattle–Denver Negative Income Tax
larly, Lindo et al. (2020) showed that divorce rates experiments led to an increase in the divorce rate.
of unemployed men decrease with increased Bobonis (2011) analysed the conditional cash
4 Journal of European Social Policy 0(0)
transfer programme in Mexico (Progresa) and found covered and income is insured up to 148,200 CHF/yr.
small increases in divorce rates related to more In terms of eligibility requirements and replacement
transfer income for women. Francesconi et al. (2009) levels, the Swiss unemployment benefit system is
showed that working tax credits in the UK (in-work rather generous compared to other OECD countries.
benefits) significantly increased divorce rates in In a reform implemented in April 2011, the
couples with low-earning men. Finally, using an maximum benefit duration was reduced for the sub-
instrumental variable design, Halla et al. (2016) have population of unemployed aged 25–55 with in-
shown that in OECD countries, increases in national complete contribution histories.3 Prior to March
social spending are related to higher divorce rates. 2009 benefit recipients with contributions of 12–17
Hence, the literature on how income shocks im- months in the 24 months before claiming benefits
pact divorce does not lend uniform support to either were entitled to a maximum of 400 daily allowances
the family stress model or the cost/independence (∼1.5 years of benefit receipt).4 Following the re-
hypotheses (see Table 1A (appendix) for an over- form, in April 2011, the same group was entitled to a
view of the most relevant studies). One possible maximum of only 260 daily allowances (∼1 year of
explanation for these inconsistent findings could be benefit receipt) (SECO, 2013). This means that the
that both mechanisms operate simultaneously with reform reduced maximum unemployment benefit
one dominant for specific subgroups. For example, a entitlement by 6 months for prime age workers with
higher ability to bear the cost of divorce after a incomplete contributions. Unemployed individuals
positive income shock likely only translates into in the same age group with longer contributions were
more divorce among couples who were already entitled to 400 days of benefits (∼1.5 years of benefit
unhappy (Goñalons-Pons and Calnitsky, 2021). receipt), both before and after the reform.
Unemployment insurance in Switzerland and The expected effect of the reform on divorce
the 2011 reform
The reform reduced benefit income for those ex-
Unemployment benefit systems can be described by hausting one year of benefits and increased pressure to
various characteristics, including eligibility re- find a job even earlier than that. The family stress
quirements, income replacement levels, and maxi- model predicts that these changes should lead to more
mum benefit duration (Schmieder and Von Wachter, marital conflict and, ultimately, divorce. Arguments on
2016). In Switzerland, eligibility depends on reason economic dependence also anticipate more divorce
for job loss, willingness to work, and paid unem- when taking the perspective of the spouse of the
ployment insurance contributions. Benefits are paid unemployed. Due to benefit cuts, previously depen-
for both involuntary and voluntary job loss, with a dent spouses can depend less on their unemployed
waiting period in the latter case. Recipients must partner. This reduces the attractiveness of the marriage
write a minimum number of job applications per relative to finding a new partner.
month and potentially participate in employment
programmes, both determined by the recipient’s Hypothesis 1a: The reform increased divorce.
caseworker. Furthermore, the recipients must have
been employed in a job that paid social contributions From the viewpoint of the economic dependence
into the Swiss unemployment benefit system for a of the unemployed spouse, however, the reform
minimum of 12 of the last 24 months.1 The re- should increase economic dependence, as he or she
placement level is stable throughout the unemploy- must rely on their spouse after their benefits are cut,
ment spell at 70% of the previous salary for those reducing divorce. A reduction in divorce is also
without dependents and 80% for those with depen- anticipated by the cost hypothesis as a reduction in
dents or an income below 3797 CHF/mo.2 Those benefit income makes it more difficult for the couple
with annual earnings below 6000 CHF are not to afford a divorce.
Kessler et al. 5
Hypothesis 1b: The reform reduced divorce. (Goñalons-Pons and Gangl, 2021; Killewald, 2016).
This would mean that couples experience greater
The literature also suggests testable hypotheses increases in conflict and divorce when husbands were
about how these mechanisms play out among certain affected by the reform, even after considering their
subgroups. The family stress mechanism should be greater income contribution to the household.
more pronounced for couples with lower household
incomes who already suffer economic stress. Hypothesis 3a: The reform increased divorce more for
couples in which the unemployed spouse was the
Hypothesis 2a: The reform increased divorce more husband, irrespective of men’s greater relative income
for couples with a low household income before contribution.
unemployment.
In contrast, the cost hypothesis would anticipate
The cost hypothesis also suggests that couples that the reform suppresses divorce more when the
with fewer economic resources were more strongly unemployed spouse is the husband, because of the
affected by the reform. A low-income household is greater income loss. From the viewpoint of cost,
more likely to drop below the threshold at which they gender norms play no role, meaning that any gender
can shoulder the cost of divorce and maintain two effect should be explained by men’s greater incomes.
separate households. Thus, if the cost hypothesis
holds, it should also be stronger for low-income Hypothesis 3b: The reform decreased divorce more for
couples. couples in which the unemployed spouse was the
husband.
Hypothesis 2b: The reform decreased divorce more for
couples with a low household income before If changes in economic dependence are the
unemployment. dominant mechanism, however, the reform should
suppress divorce more when the unemployed spouse
We also expect the reform’s effect to differ ac- is the wife. Given husbands’ higher earnings, wives
cording to the unemployed spouses’ gender. On the losing benefits become even more dependent. Their
one hand, there are arguments supporting the expec- husbands, not relying on the unemployment benefits
tation that increases in divorce caused by the reform of their wives anyhow, likely won’t react either way.
are stronger for men than for women. Gendered norms In contrast, for an unemployed man losing benefits,
on breadwinning lead to differences in the relative their wives are less able to support them, implying
income contributions to the household by unemployed that unemployed men do not become more depen-
men versus unemployed women. Unemployed men dent when they lose benefits. Their wives, however,
have more insured income, lose more when benefit as described earlier, are more likely to seek a new
eligibility is reduced, and can rely less on their part- partner. Jointly these facts suggest that if the reform
ners’ income – meaning more stress caused by the suppresses divorce, it should do so more for un-
reform for men. In addition, seen from the perspective employed wives.
of a dependent spouse, the attractiveness ofmarriage is
more strongly reduced by men’s versus women’s loss Hypothesis 3c: The reform decreased divorce more for
of benefits. couples in which the unemployed spouse was the wife.
Beyond the fact that men contribute more income
to the household, gender norms could exacerbate the
stress caused by the reform. There is a strong cultural Methods
expectation that men contribute to the household. The Data
stress associated with benefit loss is thus likely greater
for unemployed men, irrespective of the amount of To test our hypotheses, we used linked full pop-
benefit they lose or how much their wives earn ulation register data from Switzerland. Civil status
6 Journal of European Social Policy 0(0)
register data provided us with information about likelihood of divorce.7 Also, simply comparing the
marriages5, births, deaths, and divorce from 2000 to change in divorce rates of the treated before the
2018 (FSO, 2019).6 Unemployment insurance register reform to the treated after the reform would lead to
data (2000–2015) added information on the timing of biased estimates of the reform’s effect, as the like-
unemployment spells, the characteristics of the un- lihood of divorce changes across cohorts, for ex-
employed (pre-unemployment income, working ample due to changing norms.
hours, education, occupation) and unemployment Given that the timing of the reform was unrelated
benefit receipt (contribution history, maximum num- to divorce (see Footnote 4), however, it seems
ber of benefits, sanctions, actual duration of benefit plausible that the pre-to-post-reform change in di-
receipt) (Federal Council, 2006). Finally, social se- vorce rates among the treated would have been
curity data on individual accounts included earned identical to the change in divorce rates among the
employment income, self-employment, and social controls, had the reform not taken place. If this holds,
insurance income (unemployment benefits, maternity, any observed difference in the change in divorce
wage replacement during mandatory military service) rates between the two groups reflects the causal effect
for both spouses from 2000 to 2016 (Central of the reform. We measure differences in the change
Compensation Office (CCO) 2018). using a repeated cross-sectional difference-in-dif-
ferences research design, which, due to the random
Study design timing of the reform, can be called quasi-
experimental. Figure 1 (upper panel) displays
Measuring the effect of the reform by simply com- the definitions of treated and controls in the pre-
paring divorce rates between those affected by the and post-reform cohorts, based on contributions
reform (treated, 12–17 of 24 months of contribu- and corresponding maximum benefit durations.
tions) versus those not affected by the reform Due to the design of the reform, both treated and
(controls, 18–23 of 24 of months of contributions) controls were limited to individuals aged 25 to 53
would lead to biased estimates, as individuals with whose unemployment spells lasted at least nine
more interrupted employment histories have a higher months.8
Figure 1. Study design and sample selection. Selection of treated and controls, benefit duration among treated and
controls in the pre- and the post-reform period (upper panel), and selected unemployment spells and marriages (lower
panel). The effect of the reform is measured with (B-A) – (D-C).
Kessler et al. 7
Adverse selection. A potential threat to the validity of individuals married between October 2006 (3 years
the research design is adverse selection, that is, the and 6 months between the first marriages and the first
possibility that following the reform more capable unemployment spells, as for the pre-reform sample)
individuals postponed quits and layoffs until they and December 2015 (6 years and 3 months between
had 18 months of contributions. Such strategic be- first and last marriages, as for the pre-reform sample)
haviour would create differences in divorce rates due and whose unemployment spells started after April
not to the causal effect of shorter benefit eligibility, 2010 (the first date unemployment spells started in
but due to compositional shifts between treated and the post-reform regime). With this definition, there
controls. were 21,269 couples. After removing the 5.2%with a
To test for strategic behaviour, we plotted the missing value on at least one of the study variables,
share of individuals with a tertiary degree among all the sample was 20,161. (Due to the small amount of
individuals registering for unemployment with at missing information, we simply dropped these ob-
least 18 contribution months for all monthly cohorts servations). We ended up with 4456 treated and 6331
between January 2009 and December 2011 (see controls before the reform and 3124 treated and 6250
Figure 1A (appendix)). If there was strategic be- controls after the reform.
haviour among more capable individuals, then we Table 1 presents the descriptive statistics for the
would have seen a sudden increase in the share of four groups. By definition the treated group had been
individuals with a tertiary degree after the reformwas employed for fewer months in the 2 years prior to
accepted in the popular vote in September 2010 or unemployment and had lower levels of education and
after it was introduced in April 2011. The data does occupational status. The last column of Table 1
not show a sudden increase, but rather the contin- presents the coefficient of the respective character-
uation of a steady trend in increased tertiary degrees. istics regressed on an interaction term between the
There was similarly no sudden increase using al- treatment and period (pre-reform vs post-reform)
ternative measures of selection such as earnings in variables with statistically significant effects of this
the last job before unemployment (not reported). DiD estimator for several variables, suggesting dif-
ferences between treated and controls that changed
Sample following the reform. We adjust for these differences
by including all variables in Table 1 as control
We restricted observations to those individuals whose variables in our multivariate models and by bal-
marriage took place prior to or in the first year of ancing the four comparison groups to match the
unemployment and remained married throughout the characteristics of the treated post-reform (see the
first year of unemployment. Among couples with section on Analytical strategy).
multiple unemployment spells after marriage, we se-
lect the first spell (taking a random spell if both spouses Variables
registered for unemployment on the same day).
For the pre-reform sample, we selected all indi- The outcome variable is whether the couple was
viduals married between January 2000 (the earliest legally divorced. Using this measure, the effects of
date that we could measure unemployment after the reform will have a substantial lag. Using data
marriage) and March 2009 (the last date unem- from the Swiss Families and Generations Survey
ployment spells started that were not affected by the (Federal Statistical Office, 2021) we find an av-
reform). Among these, we selected those for whom erage lag of one year and 9 months between the
the first unemployment spell after marriage started year of self-reported separation and the year of
between July 2003 (the first date unemployment divorce in the general population in our study
spells started in the pre-reform regime)9 and March period. Given that unemployment is associated
2009 (the last date unemployment spells started that with even longer time lags between separation and
were not affected by the reform, see Figure 1, lower divorce (Tumin and Qian, 2017), this can be seen
panel). For the post-reform samples, we selected all as a lower bound for our sample. Second, the
8 Journal of European Social Policy 0(0)
Table 1. Characteristics of treated and controls and trends from before to after reform.
Treated Controls DiD
Pre Post Pre Post
UI contribution months 14.01 14.18 21.58 21.48 0.27***
Education: Tertiary 0.23 0.28 0.24 0.32 
0.03*
Upper-secondary 0.45 0.35 0.52 0.41 0
Less than vocational 0.32 0.37 0.24 0.26 0.03*
Occupation: Managerial/professional 0.16 0.15 0.21 0.24 
0.04***
Clerical/Services/Sales/Skilled agricult 0.48 0.43 0.53 0.48 
0.01
Craft/Plant and machine operators/Elementary 0.35 0.42 0.26 0.27 0.05***
Citizenship: Swiss 0.29 0.2 0.47 0.39 
0.01
Foreign: Neighbouring countries 0.11 0.13 0.15 0.17 
0.01
Foreign: Non-neighbouring countries 0.6 0.67 0.38 0.44 0.02
Age at marriage 32.19 31.66 32.94 32.52 
0.11
Partner age at marriage 34.42 33.55 33.59 33.51 
0.79**
First marriage 0.73 0.77 0.72 0.77 0
Age at unempl. Start 33.98 33.81 35.05 34.98 
0.1
Employed y. Before unempl 0.96 0.96 0.97 0.97 0
Income y. Before unempl. (CHF 1000/yr.) 35.43 31.08 53.66 57 
7.69***
Working hours before unempl. (100 = 42 hrs./wk.) 89.61 86.5 88.91 88.47 
2.67***
Desired working hours (100 = 42 hrs./wk.) 92.33 92.89 88.47 89.48 
0.44
Sanctions 4.94 5.33 5.87 6.81 
0.56*
Waiting days 3.35 3.11 3.75 3.97 
0.45***
Partner employment y. Before unempl 0.79 0.8 0.81 0.8 0.02
Partner income y. Before unempl. (CHF 1000/yr.) 47.65 48.66 53.71 59.08 
4.36**
Number of children: < Age 4 0.4 0.42 0.5 0.55 
0.03
Age 4–6 0.12 0.13 0.14 0.16 
0.01
Age 7+ 0.17 0.13 0.23 0.18 0
Hh. Income year. Before unempl. (CHF 1000/yr.) 83.08 79.74 107.37 116.08 
12.05***
Region: Lake geneva 0.3 0.31 0.29 0.33 
0.03*
Central lowlands 0.21 0.21 0.19 0.17 0.02
Northwest 0.1 0.1 0.11 0.12 
0.01
Zurich 0.19 0.19 0.2 0.18 0.01
East 0.08 0.09 0.09 0.1 0.01
Central 0.06 0.05 0.06 0.05 0
Ticino 0.05 0.05 0.05 0.05 0
N (couples) 4456 3124 6331 6250 —
Treated: 12–17 months contributions in the 24 months before start of the UI spell. Controls: 18–23 months contributions. Pre ob-
servations: Married between Jan 2000 and Mar 2009, start of unemployment between Jul 2003 and Mar 2009. Post: Married between Sep
2006 and Dec 2015, start of unemployment between Apr 2010 and Dec 2015. DiD = interaction term between treated versus controls
and pre- versus post-reform. p-value thresholds DiD: * = 5%, ** = 1%, *** = 0.1%
divorce process itself (that is, the time between the granted in 2008, there was a 9-month lag between
moment couples file for divorce and the moment filing for divorce and the divorce being legally
the divorce is legally granted) can take a long time. granted.10 Together, the total lag between sepa-
Analysing divorce register data, for divorces ration and the legal enactment of divorce is
Kessler et al. 9
between 2 and 3 years. The effects of the reform group. βDiD expresses the difference between treated
would thus be expected in years 3 and 4 after the and control group in the change from the pre- to the
year the unemployment spell began. Given this post-reform period – the difference-in-differences
time lag, we measure cumulative divorce up to estimate of the effect of the reform. To remove co-
4 years after the start of unemployment. variate imbalance, we use entropy balancing (see
We define household income as income from Table 2A (appendix)). Entropy balancing generates
employment and social insurance from both partners weights that adjust means in all comparison groups to
in the calendar year before unemployment. In the those in the treated group post-reform. Compared to
presented analysis, we examine income terciles, but adjustment using linear control variables, entropy
smaller categories (quartiles and quintiles) were also balancing reduces the dependence of treatment effect
tested.11 estimates on model specification (Hainmueller,
We measure the relative income contribution of 2012). We report results using entropy balancing
the unemployed spouse using the share of the total in the main article and results using regression in the
household income in the calendar year before un- Online appendix. To further strengthen conclusions
employment. To estimate simultaneous differences on the causal effect of the reform, we consider the
by gender and relative income contribution (Hy- trends in divorce among treated and controls before
pothesis 3a), we used a binary indicator dis- the reform: if divorce trends in the two groups were
tinguishing between ‘primary earners’ (relative parallel before the reform, then a divergence in di-
income contribution greater than 50%) versus ‘sec- vorce trends between the two groups just when the
ondary/non-earners’ (50% or less). reform happened suggests that this is due to the
causal effect of the reform.
Analytical strategy. To capture the effects of the re-
form on divorce, we use linear probability models
(LPM) on the stock of divorced couples in each year
after unemployment start with the dependent var- Results
iable of whether a couple has had a divorce yet. Overall effect of the reform
Compared to event-history analysis of person-year
data, our approach is less prone to statistical noise Figure 2 displays the share of couples who were
(present in annual separation rates). Given that divorced in the years following an unemployment
divorce is a relatively rare event impacting about spell in the four groups (treated/control, pre/post-
10% of the couples in our sample, we also compare reform). The lower panel includes difference-in-
results using a logit model (presented in the Online differences estimates and 95% confidence intervals.
appendix). For all four groups, the share of couples who are
We measure the effect of the reform with an in- divorced increases relatively steeply in the years
teraction term between a dummy variable for after an unemployment spell. Also, there is a clear
whether the unemployment spell took place in the increase in divorce rates between the two periods,
pre-reform or post-reform period and a dummy illustrating an increasing trend in divorce in
variable for the unemployed spouse belonging to the Switzerland over this period. What is important is
treated or the control group. Our model is: that divorce rates before the reform are identical
for treated and control couples (gray lines) while
Divorced ¼ β0 þ βT  XT þ βP  XP þ βDiD  XT  XP after the reform divorce rates for the treated ac-
celerate. The treated group’s relatively higher di-
β0 denotes the intercept, βT the coefficient that vorce rate increases steadily in the years after
estimates the effect of being in the treated versus unemployment starts, only reaching statistical
control group in the pre-reform period and βPost the significance in the fourth year. This confirms
coefficient that estimates the effect of the change Hypothesis 1a, showing an increase in divorce due
from the pre- to the post-reform period in the control to the reform.
10 Journal of European Social Policy 0(0)
Figure 2. Effect of the reduction of unemployment benefit duration on divorce. Levels in treated/controls and pre/post-
reform and cumulative difference-in-differences of the share of couples divorced by year after unemployment start
(95% confidence intervals). Results based on entropy-balanced samples.
The estimated effect of the reform is a 2.8 per- and the controls before the reform. Before the re-
centage point increase in the share of couples who form, trends in divorce were parallel between
divorce (Figure 1, bottom panel). Relative to the treated and controls (with covariates unbalanced,
counterfactual level of divorce in the treated group left hand panel) or even steeper among the controls
(the observed post-reform level of divorce minus the (with covariates balanced, right hand panel). Hence,
treatment effect = 14.1–2.8 percentage points = 11.3 extrapolating this trend to the post-reform period
percentage points) this is a 25% increase in divorce without the reform’s effect, one would expect a
due to the reform.12 parallel increase in divorce between the two groups
It is possible that the greater difference in di- (resp. A stronger increase among the controls).
vorce among the treated after the reform is not due Instead, we find stronger increases in divorce
to the causal effect of the reform, but due to di- among the treated just after the reform (cohort
verging trends in divorce between the two groups 2010–2012), in line with the expectation that the
that already existed before the reform. To test this, difference between treated and controls in the post-
Figure 3 shows trends in divorce among the treated reform sample is due to the reform.
Kessler et al. 11
Figure 3. Trends in the level of divorce before and after reform, divorce stock measured in the fourth year after
unemployment start, balanced and unbalanced control groups, 95% confidence intervals.
From a policy perspective, it is important to consider following unemployment. The divorce rate among
the effect of the reform on divorce (+25%) in the context couples in the lowest tercile increased by 5.8 per-
of other reform effects including losses in benefits and centage points (a 56% higher divorce rate than in
increases in employment and income. Those reaching at the absence of the reform). In the middle tercile
least 9 months of unemployment received CHF 9770 there is an increase in divorce of 2.8 percentage
less in benefits, which is a 23% loss (concentrated in the points that is not statistically significant, while the
second year of unemployment,
55%). Benefit loss was reform has no impact on divorce rates among the
partially compensated by faster reemployment (about upper tercile.
3 weeks,
4.3% in unemployment duration) and higher Hypotheses 3a-c considered gender-specific ef-
earnings (about CHF 2,880, +11.3%, see Table 2A fects. We do not find support for the gendered ex-
(appendix)). pectations of any of these hypotheses (Figure 4,
middle panel). In the fourth year after unemployment,
Heterogenous effects of the reform the reform had a statistically significant impact on both
men’s and women’s divorce rates. While absolute
In addition to main effects, we hypothesized dif- effects for men (3.7 percentage points) are somewhat
ferential effects across subgroups. Figure 4 presents stronger than for women (2.8 percentage points), they
the results from the stratified analyses by terciles of are more similar when expressed relative to coun-
household income, gender and gender combined terfactual levels of divorce (32% for men, 28% for
with relative income contribution. women).
Confirming Hypothesis 2a, we find a greater Considering the results for gender and relative
increase in divorce due to the reform for couples in income contribution simultaneously suggests that
the bottom income tercile (Figure 4, upper panel), stronger effects for unemployed men are entirely due
with reform effects, and the difference in reform to their greater contribution to household income
effects by income, significant only in the fourth year (Figure 4, bottom panel). The reform impacted
12 Journal of European Social Policy 0(0)
Conclusion
Although there is literature showing unemployment
insurance generosity impacts health and that unem-
ployment causes divorce (e.g. Di Nallo et al., 2022;
Kuka, 2020), there is only one study examining un-
employment insurance and divorce (Lindo et al.,
2020). Further, from the social policy literature, we
know that social programme generosity is associated
with family stability, but these studies have not ex-
amined how unemployment insurance, the most im-
portant social safety net for the working age
population, affects divorce. Using full-population
administrative data and a policy change that re-
duced maximum benefit duration in Switzerland, in
this study we provide new evidence for how unem-
ployment insurance generosity impacts divorce.
We found that the reform, reducing entitlement by
6 months, was associated with a 2.8 percentage point
increase in the share of couples who divorce, which
is a 25% relative increase. As could be expected from
the lag between marital conflict, household separa-
tion and formal divorce, we found statistically sig-
nificant effects only 4 years after unemployment
start. The reform’s effect on divorce was mediated by
income and employment effects, with the reform
leading to a 23% drop in total benefit income, a small
reduction in unemployment duration (
4.3%) and an
increase in earned income (+11.3%) for those un-
Figure 4. Difference-in-differences estimates of the effect employed for more than nine months.
of the reduction of unemployment benefit duration on The increase in divorce was largely concentrated
cumulative divorce after unemployment start by among poor households, with upper income
household income, gender and gender combined with households unaffected. Unlike Lindo et al. (2020),
relative income contribution, 95% confidence intervals. we also found substantial effects of benefit gener-
osity on divorce rates of unemployed women.
divorce only among breadwinners with no statisti- Slightly higher average effects for men are explained
cally significant effects for unemployed individuals by men’s higher relative contribution to household
contributing less than half of the couples’ income income before job loss. Among primary earners, the
before unemployment. It increased divorce among reduction in benefits had a stronger impact on di-
unemployed female breadwinners in the third year vorce for unemployed women.
after unemployment start by + 8.9 percentage points These results largely lend support for a gender-
which is a 117% increase; in the fourth year, it in- neutral version of the family stress model in which
creased divorce by + 9.2 percentage points, a 78% the stress caused by financial difficulties and in-
increase. In contrast, for male breadwinners the re- creased pressure to find a job leads to marital conflict
form’s effect is only statistically significant in the and divorce. Taking the viewpoint of the unem-
fourth year, increasing divorce by + 5.2 percentage ployed spouse, these findings could also be related to
points, or 40%. economic dependence: due to benefit loss, the spouse
Kessler et al. 13
can depend less on the financial support of their be generalizable to the overall population receiving
unemployed partner, reducing the attractiveness of unemployment benefits, as most unemployment
this marriage. spells are short or might not apply to different
In contradiction to interpretations put forward in changes in unemployment insurance, like a reduction
many studies on the effect of income shocks on in the replacement rate. More knowledge is needed
divorce (Bertrand et al., 2015; Goñalons-Pons and on how other changes to unemployment benefits
Gangl, 2021; Killewald, 2016), our results do not impact divorce (Lindo et al., 2020). Also, it might
support the notion that the male breadwinner norm seem that the Swiss context might be very specific
(which is still dominant in Switzerland) alters the and not generalizable to other countries. That said,
interpretation of income loss. Rather, the high effects given high savings rates and wealth, as well as
among (female and male) primary earners suggest generous social assistance, it seems likely that
that stress is greater when the income loss is more Switzerland is a ‘least-likely’ case, that is, if un-
financially significant for a couple. It is possible employment insurance is key for marital stability in
earlier studies finding stronger effects for male in- Switzerland, it is likely even more so in other
come loss did not sufficiently account for the countries. Still, additional studies from other coun-
household’s dependency on lost income. In fact, our tries are necessary to confirm the results.
results combining gender and relative income con- Assuming these estimates reflect the causal effect
tribution suggest that women’s income loss is more of shortened maximum benefit duration on divorce,
important than men’s. our findings have important policy implications.
Results cannot, however, conclusively reject the When deciding on shorter benefits, policymakers
cost/independence hypotheses. It is possible that un- must not only consider the immediate savings from
happy couples or unemployed individuals who were fewer benefits being paid out and small positive re-
considering divorce might have postponed it in the face employment effects, but also potential impacts on
of reduced benefits. Since our data did not allow us to other outcomes like divorce. Social insurance gen-
identify dissatisfied couples (as compared to Goñalons- erosity has a broader impact than is often considered
Pons and Calnitsky, 2021), we can only conclude that when policymakers cut programmes.
this was not the dominant mechanism in our overall
sample, nor among the subgroups we explored. Acknowledgements
The study also has several limitations. In principle, our We thank the Swiss National Science Foundation for
results could reflect differences in divorce trends between funding for this project (grant nr. 176371), Ursina Kuhn, as
more (treated) and less (controls) disadvantaged groups well as workshop participants of the 18th Meeting of the
that existed independently of the reform. Or they could
‘European Network for the Sociological and Demographic
express adverse selection, with more able, and deter- Study of Divorce’ at University of Cologne and of the
mined (and less divorce-prone) individuals self-selecting Ingrid Expert Workshop ‘Experimental designs in social
into the control group. However, our robustness checks policy research: strengths, weaknesses, opportunities, and
reject these assumptions in favour of a causal interpre- threats’ at the University of Stockholm for comments on
tation of our findings: the higher divorce rate among prior versions of this manuscript.
treated versus controls emerges just after the reform,
which could not be predicted looking at how di- Declaration of Conflicting Interests
vorce evolved among the two groups before the
reform. Also, we find no evidence that individuals The author(s) declared no potential conflicts of interest
strategically adapted to the new eligibility rules with respect to the research, authorship, and/or publication
after these became known to the public. of this article.
In addition, our results might have limited gen-
eralizability. We only studied long-term unemployed Funding
individuals and only a single dimension of benefit The author(s) disclosed receipt of the following financial
generosity, that is, benefit duration. Results might not support for the research, authorship, and/or publication of
14 Journal of European Social Policy 0(0)
this article: This work was supported by the Schweizerischer do not see this data limitation as a threat to our study
Nationalfonds zur Förderung der Wissenschaftlichen For- design since it affected all our study groups (see
schung (176371). Study design) alike.
7. We excluded the unemployed who had paid 24
months of contributions to restrict the control
Data availability statement sample to unemployed with disrupted employment
Due to Swiss data protection laws, authors are not al- histories, that is, to select controls that are more
lowed to share research data directly. Data can, however, comparable to the treated, but were still not af-
be requested from the Swiss Federal Statistical Office fected by the reform.
(https://www.bfs.admin.ch/bfs/en/home/services/data- 8. Due to the design of the reform (reducing benefits
linkages/for-third-parties.html). Code for data prepa- from 1.5 to one year), only those who were at the
ration and analysis is available from authors upon least approaching one year of benefits were affected
request. by it. We assumed that anticipation of benefit ex-
haustion after 12 months already affects individuals
ORCID iD
after 9 months. We therefore restricted the sample to
Dorian Kessler  https://orcid.org/0000-0003-0495-0858 those with at least 9 months of unemployment
duration.
Supplemental Material 9. There were different rules to benefit eligibility before
Supplemental material for this article is available online. July 2003.
10. Information on both divorce filing and the date the
divorce was granted was last available in 2008 as the
Notes
Federal Statistical Office stopped collecting this var-
1. Workers are exempted from contribution requirements iable to reduce the administrative burden of the courts.
if they recently finished education, gave birth to a 11. Overall conclusions remain unaffected using quartiles
child, were sick, had a divorce or separation, or fin- versus terciles.
ished a prison sentence. 12. Alternative estimates using logistic regression (see
2. Figure refers to the level in 2021. Figure 2A), illustrating effects as average marginal
3. The reform was motivated by budget shortfalls fol- effects have similar results. Covariate adjustment with
lowing an underestimate of the anticipated unem- linear controls (versus entropy balancing) also leads to
ployment rate from 2004 to 2010 (Devaud and Keller, similar but less pronounced effects (statistically sig-
2012). nificant at the 10% level in the third year after
4. Starting in July 2003, when a reform last changed unemployment).
benefit generosity.
5. The observation period of the study (2000–2018)
refers to the period before the introduction of same-sex References
marriage in Switzerland (July 2022). Therefore, same- Becker, G, Landes, EM and Michael, RT (1977) An
sex marriages could not be considered in the analysis economic analysis of marital instability. The Journal
even if the theoretical arguments apply equally to of Political Economy 85(6): 1141–1187.
opposite-sex and same-sex marriages. Bertrand, M, Kamenica, E and Pan, J (2015) Gender
6. Data were linked with pseudonymized social security identity and relative income within households. The
numbers. Social security numbers of individuals who Quarterly Journal of Economics 130(2): 571–614.
married/divorced before 2010 were missing in the DOI: 10.1093/qje/qjv001
original data. Therefore, we identified them via Biegert, T (2017) Welfare benefits and unemployment in
matching with 2010 population registers using affluent democracies: the moderating role of the in-
unique combinations of marriage and birth dates, stitutional insider/outsider divide. American Socio-
leading to a loss of observations of individuals who logical Review 82(5): 1037–1064. DOI: 10.1177/
were no longer resident in Switzerland in 2010. We 0003122417727095
Kessler et al. 15
Bitler,MP, Gelbach, JB, Hoynes, HW, et al. (2004) The impact development. Journal of Marriage and Family 72(3):
of welfare reform on marriage and divorce.Demography 685–704. DOI: 10.1111/j.1741-3737.2010.00725.x
41(2): 213–236. DOI: 10.1353/dem.2004.0011 Devaud, L and Keller, D (2012) AVIG-Revision: Aus-
Blom, N, Kraaykamp, G and Verbakel, E (2019a) wirkungen auf die Versicherten und die Finanzen der
Current and expected economic hardship and sat- ALV. Available at: https://dievolkswirtschaft.ch/de/
isfaction with family life in Europe. Journal of 2012/09/devaud-3/
Family Issues 40(1): 3–32. DOI: 10.1177/ Di Nallo, A, Lipps, O, Oesch, D, et al. (2022) The effect of
0192513X18802328 unemployment on couples separating in Germany and
Blom, N, Verbakel, E and Kraaykamp, G (2019b) Couples’ the UK. Journal of Marriage and Family 84(1):
job insecurity and relationship satisfaction in the 310–329. DOI: 10.1111/jomf.12803
Netherlands. Journal of Marriage and Family 82(3): Federal Council (2006) Verordnung vom 1. November
875–891. DOI: 10.1111/jomf.12649 2006 über das Informationssystem für die Arbeits-
Bobonis, GJ (2011) The impact of conditional cash vermittlung und die Arbeitsmarktstatistik (AVAM-
transfers on marriage and divorce. Economic Devel- Verordnung). Available at: https://www.admin.ch/
opment and Cultural Change 59(2): 281–312. DOI: opc/de/classified-compilation/20061964/index.html
10.1086/657123 Francesconi, M, Rainer, H and Van Der Klaauw, W (2009)
Boertien, D (2012) Jackpot? Gender differences in the The effects of in-work benefit reform in Britain on
effects of lottery wins on separation. Journal of couples: theory and evidence. The Economic Journal
Marriage and Family 74(5): 1038–1053. DOI: 10. 119(535): F66–F100. DOI: 10.1111/j.1468-0297.
1111/j.1741-3737.2012.01003.x 2008.02225.x
Boertien, D and Lersch, PM (2021) Gender and changes in FSO (Federal Statistical Office) (2019) Vital statistics
household wealth after the dissolution of marriage and (BEVNAT). Available at: https://www.bfs.admin.ch/
cohabitation in Germany. Journal of Marriage and bfs/de/home/statistiken/bevoelkerung/erhebungen/
Family 83(1): 228–242. DOI: 10.1111/jomf.12705 bevnat.html
Carr, E and Chung, H (2014) Employment insecurity and FSO (Federal Statistical Office) (2021) Families and
life satisfaction: the moderating influence of labour generations Survey. Available at: https://www.bfs.
market policies across Europe. Journal of European admin.ch/bfs/en/home/statistiken/bevoelkerung/
Social Policy 24(4): 383–399. DOI: 10.1177/ erhebungen/efg.html
0958928714538219 Goñalons-Pons, P and Calnitsky, D (2021) Exit, voice, and
Central Compensation Office (CCO) (2018) Individual loyalty in the family. Socio-Economic Review 0(0):
accounts. Available at: https://www.zas.admin.ch/zas/ 1–29.
de/home/services-en-ligne/particuliers/extrait-du- Goñalons-Pons, P and Gangl, M (2021) Marriage and
compte-individuel.html (accessed 15 February 2018). masculinity: male-breadwinner culture, unemploy-
Cherlin, AJ (2020) Degrees of change: an assessment of the ment, and separation risk in 29 countries. American
deinstitutionalization of marriage thesis. Journal of Sociological Review 86(3): 465–502. DOI: 10.1177/
Marriage and Family 82(1): 62–80. DOI: 10.1111/jomf. 00031224211012442
12605 Groeneveld, LP, Tuma, NB and Hannan, MT (1980) The
Conger, RD, Elder, GH, Lorenz, FO, et al. (1990) Linking effects of negative income tax programs on marital
economic hardship to marital quality and instability. dissolution. The Journal of Human Resources 15(4):
Journal of Marriage and Family 52(3): 643–656. 654–674. DOI: 10.2307/145406
DOI: 10.2307/352931 Gschwind, L (2021) When free choice turns into a pitfall:
Conger, RD, Ge, X, Elder, GH, et al. (1994) Economic conditional social protection for immigrants in vol-
stress, coercive family process, and developmental untary unemployment insurance systems. Journal of
problems of adolescents. Child Development 65(2): European Social Policy 31(1): 3–14. DOI: 10.1177/
541–561. DOI: 10.1111/j.1467-8624.1994.tb00768.x 0958928720973912
Conger, RD, Conger, KJ and Martin, MJ (2010) Socio- Hainmueller, J (2012) Entropy balancing for causal effects:
economic status, family processes, and individual a multivariate reweighting method to produce
16 Journal of European Social Policy 0(0)
balanced samples in observational studies. Political Rose, D (2019) The impact of active labour market
Analysis 20(1): 25–46. policies on the well-being of the unemployed.
Halla, M, Lackner, M and Scharler, J (2016) Does the Journal of European Social Policy 29(3): 396–410.
welfare state destroy the family? Evidence from DOI: 10.1177/0958928718792118
OECD member countries. The Scandinavian Jour- Sayer, LC and Bianchi, SM (2000) Women’s economic in-
nal of Economics 118(2): 292–323. DOI: 10.1111/ dependence and the probability of divorce: a review and
sjoe.12144 reexamination. Journal of Family Issues 21(7):
Inanc, H (2018) Unemployment, temporary work, and 906–943. DOI: 10.1177/019251300021007005
subjective well-being: the gendered effect of spousal Schmieder, JF and von Wachter, T (2016) The effects of
labor market insecurity. American Sociological Review unemployment insurance benefits: new evidence and
83(3): 536–566. DOI: 10.1177/0003122418772061 interpretation. Annual Review of Entomology 8(1):
Killewald, A (2016) Money, work, and marital stability 547–581. DOI: 10.1146/annurev-economics-080614-
assessing change in the gendered determinants of di- 115758
vorce. American Sociological Review 81(4): 696–719. State Secretariat for Economic Affairs (SECO) (2013)
Kim, J and Luke, N (2020) Men’s economic dependency, Arbeitslosigkeit in der Schweiz 2012. Neuchâtel:
gender ideology, and stress at midlife. Journal of Federal Statistical Office.
Marriage and Family 82(3): 1026–1040. DOI: 10. Shahidi, FV and Parnia, A (2021) Unemployment insur-
1111/jomf.12615 ance and mortality among the long-term unemployed:
Kuhn, U (2020) Augmented wealth in Switzerland: the a population-based matched cohort study. American
influence of pension wealth on wealth inequality. Journal of Epidemiology (kwab144). DOI: 10.1093/
Swiss Journal of Economics and Statistics 156(1): aje/kwab144
19–35. DOI: 10.1186/s41937-020-00063-9 Townsend, N (2010) Package Deal: Marriage, Work and
Kuka, E (2020) Quantifying the benefits of social insur- Fatherhood in Men’s Lives. Philadelphia, PA: Temple
ance: unemployment insurance and health. The Re- University Press.
view of Economics and Statistics 102(3): 490–505. Tumin, D and Qian, Z (2017) Unemployment and the
DOI: 10.1162/rest_a_00865 transition from separation to divorce. Journal of
Lindo, J, Regmi, K and Swensen, I (2020) Stable income, Family Issues 38(10): 1389–1413. DOI: 10.1177/
stable family. NBER Working Paper. Available at: 0192513X15600730
https://www.nber.org/system/files/working_papers/ VanWinkle, Z (2020) Family polices and family life course
w27753/w27753.pdf complexity across 20th-century Europe. Journal of
Mortensen, DT (1976) Unemployment insurance and job European Social Policy 30(3): 320–338. DOI: 10.
search decisions. Industrial and Labor Relations 1177/0958928719880508
Review 30(4): 505–517. Vandecasteele, L and Debels, A (2007) Attrition in panel
Obinger, H (1999) Minimum income in Switzerland. data: the effectiveness of weighting. European So-
Journal of European Social Policy 9(1): 29–47. DOI: ciological Review 23(1): 81–97. DOI: 10.1093/esr/
10.1177/095892879900900102 jcl021
OECD (2018) Unemployment-benefit Coverage: Recent West, C and Zimmerman, D (1987) Doing gender.
Trends and Their Drivers.OECDEmployment Outlook Gender & Society 1(2): 125–151. DOI: 10.1177/
2018. Paris: OECD Publishing, 185–210. Available at: 0891243287001002002
https://doi.org/10.1787/empl_outlook-2018-9-en. Wulfgramm, M (2014) Life satisfaction effects of unem-
Özcan, B and Breen, R (2012) Marital instability and female ployment in Europe: themoderating influence of labour
labor supply. Annual Review of Sociology 38(1): market policy. Journal of European Social Policy
463–481. DOI: 10.1146/annurev-soc-071811-145457 24(3): 258–272. DOI: 10.1177/0958928714525817