Tilburg University Intermarriage and the risk of divorce in the

Tilburg University

Intermarriage and the risk of divorce in the Netherlands

Kalmijn, Matthijs; de Graaf, P.M.; Janssen, J.

Published in:

Population Studies-A Journal of Demography

Publication date:

2005

Link to publication in Tilburg University Research Portal

Citation for published version (APA):

Kalmijn, M., de Graaf, P. M., & Janssen, J. (2005). Intermarriage and the risk of divorce in the Netherlands: The

effects of differences in religion and in nationality, 1974-94.

Population Studies-A Journal of Demography

(1),

71-85.

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners

and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.

• You may not further distribute the material or use it for any profit-making activity or commercial gain

• You may freely distribute the URL identifying the publication in the public portal

Take down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately

and investigate your claim.

Download date: 01. Oct. 2024

This article was downloaded by:[Ingenta Content Distribution]

On: 22 October 2007

Access Details: [subscription number 768420433]

Publisher: Routledge

Informa Ltd Registered in England and Wales Registered Number: 1072954

Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Population Studies

A Journal of Demography

Publication details, including instructions for authors and subscription information:

http://www.informaworld.com/smpp/title~content=t713689546

Intermarriage and the risk of divorce in the Netherlands:

The effects of differences in religion and in nationality,

1974-94

Matthijs Kalmijn

; Paul M. de Graaf

; Jacques P. G. Janssen

Tilburg University,

Radboud University Nijmegen,

Medtronic Bakken Research Center,

Online Publication Date: 01 March 2005

To cite this Article: Kalmijn, Matthijs, de Graaf, Paul M. and Janssen, Jacques P. G.

(2005) 'Intermarriage and the risk of divorce in the Netherlands: The effects of

differences in religion and in nationality, 1974-94', Population Studies, 59:1, 71 - 85

To link to this article: DOI: 10.1080/0032472052000332719

URL: http://dx.doi.org/10.1080/0032472052000332719

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction,

re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly

forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents will be

complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be

independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings,

demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or

arising out of the use of this material.

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

Intermarriage and the risk of divorce in the

Netherlands: The effects of differences in religion and

in nationality, 1974



/94

Matthijs Kalmijn

, Paul M. de Graaf

and Jacques P. G. Janssen

Tilburg University,

Radboud University Nijmegen,

Medtronic Bakken Research Center

A textbook hypothesis about divorce is that heterogamous marriages are more likely to end in divorce

than homogamous marriages. We analyse vital statistics on the population of the Netherlands,

which provide a unique and powerful opportunity to test this hypothesis. All marriages formed

between 1974 and 1984 (nearly 1 million marriages) are traced in the divorce records and multivariate

logistic regression models are used to analyse the effects on divorce of heterogamy in religion and national

origin. Our analyses confirm the hypothesis for marriages that cross the Protestant 

/Catholic or the

Jewish

/Gentile boundary. Heterogamy effects are weaker for marriages involving Protestants or

unaffiliated persons. Marriages between Dutch and other nationalities have a higher risk of divorce, the

more so the greater the cultural differences between the two groups. Overall, the evidence supports

the view that, in the Netherlands, new group boundaries are more difficult to cross than old group

boundaries.

Keywords: divorce; ethnicity; heterogamy; intermarriage; marriage; nationality; religion

[Submitted September 2003; Final version accepted September 2004]

Introduction

A classic hypothesis about divorce is that when

husband and wife have dissimilar characteristics,

their marriage is more likely to end in divorce.

One reason to expect this relationship is that

differences in religion, ethnicity, and other social

characteristics, are correlated with differences in

tastes, values, and communication styles (Kalmijn

1998). Such differences make it more difficult for

spouses to understand each other, reduce the

number of activities they enjoy doing together,

and limit the degree to which they can confirm

each other’s values and world-views. A second

reason to expect divorce to be more likely is that

marrying someone with different characteristics

implies crossing a social boundary in society.

Because marrying outside the group is often norma-

tively disapproved of, mixed marriages may receive

less support from the social networks of the respec-

tive spouses than other marriages. Although lack

of social support does not necessarily make a

marriage unstable or unhappy, support or disap-

proval from friends and family members probably

does make a difference when the relationship is

troubled.

The notion that heterogamy increases the prob-

ability of the marriage ending in divorce is a ‘text-

book hypothesis’ in the social sciences and is widely

believed to be true in the general public (Glenn et

al. 1974). Several designs have been used to test the

hypothesis. The oldest studies relied on vital statis-

tics and matched divorce records to marriage

records (Monahan and Kephart 1954; Burchinal

and Chancellor 1963). Using information about

brides and grooms that is included on the marriage

record, these studies were able to assess whether

mixed marriages had a higher risk of divorce than

other marriages. Later studies relied on cross-

sectional survey data and used measures of per-

ceived marital stability or marital satisfaction as a

dependent variable (Heaton 1984; Shehan et al.

1990). The most recent studies used prospectiveor

retrospective longitudinal survey data and applied

regression models to compare the probability of

divorce for heterogamous and homogamous couples

Population Studies, Vol. 59, No. 1, 2005, pp. 71

/85

ISSN 0032-4728 print/ISSN 1477-4747 online/05/010071-15 # 2005 Population Investigation Committee

DOI: 10.1080/0032472052000332719

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

(Schwertfeger 1982; Lehrer and Chiswick 1993;

Jones 1996).

What does the evidence show to date? The

heterogamy hypothesis has been studied for a range

of characteristics, including education (Tynes 1990),

social class (Glenn et al. 1974; Jalovaara 2003),

religion, and ethnicity. Our focus is on religion and

ethnicity and we therefore limit our overview to

these two. Studies using actual divorce risks as the

outcome generally find support for the hypothesis,

although most of the studies are now rather old

(Bumpass and Sweet 1972; Becker et al. 1977;

Michael 1979; Lehrer 1996; Bru¨ derl and Engelhardt

1997). An important recent study comes from the

USA (Lehrer and Chiswick 1993). Using a retro-

spective survey with detailed information on de-

nominations, Lehrer and Chiswick show, among

other things, that a marriage between a Catholic

and a Protestant has a higher divorce risk than that

of a marriage between two Catholics or between two

Protestants. In addition, a marriage between mem-

bers of different Protestant denominations also has a

higher divorce risk.

An important recent study of the ethnic dimen-

sion of the heterogamy hypothesis has been con-

ducted in Hawaii (Jones 1996). In this analysis, Jones

analyses two Asian ethnic groups and shows that

there are large differences between these groups in

the risk of divorce. However, in ethnically mixed

marriages, the risk of divorce was in between the risk

for the two types of homogamous marriage between

similar individuals in the same ethnic group. Jones

interprets this as a convergence between groups

rather than a heterogamy effect (Jones 1996). A

recent European study focusing on linguistic hetero-

gamy finds that marriages between a Swedish-

speaking and a Finnish-speaking person havea

divorce risk that is a little above the highest level

of the two language groups, and concludes in favour

of the heterogamy hypothesis (Finna¨s 1997).

In sum, the evidence that religious and ethnic

heterogamy affects the risk of divorce is moderately

positive. The evidence accumulated in the USA is

substantial, at least for religious intermarriage. Little

is known about the relationship in European coun-

tries because less research has been done on the

issue in these countries. In this paper, we present an

examination of the effect on divorce of heterogamy

in religion or nationality, by analysing vital statistics

for the Netherlands. By matching marriage records

and divorce records from the population registers of

all Dutch municipalities, we are able to assess

whether heterogamous marriages are more likely

than homogamous marriages to end in divorce. An

obvious drawback of these data is that the number of

characteristics on the marriage record is limited; the

individual characteristics in our data are religion,

nationality, age, and previous marital status. Despite

this drawback, we believe that vital statistics provide

a powerful opportunity to test the heterogamy

hypothesis. The data cover the entire population of

marriages in a given period rather than a sample of

marriages and the number of marriages we are able

to analyse is therefore quite large (nearly 1 million).

Another advantage of our data is that they are

prospective rather that retrospective. Hence, our

measures of religiosity will not be coloured by recall

bias. In sum, in this paper, we restore one of the

older methods for evaluating the heterogamy effect,

a method we believe was abandoned too soon.

The case of the Netherlands is a particularly

interesting one because of its tradition of pillariza-

tion (the segmentation of Dutch society into four

dominant interest groups based on religion or

ideology and class) that has made the social, institu-

tional, and geographical boundaries between reli-

gious groups quite strong. The period of pillarization

was strongest in the first half of the twentieth

century but the period since the 1950s has been

characterized by rapid secularization. Church mem-

bership declined for all groups except for the most

orthodox Protestant groups, church attendance

among church members declined, traditional reli-

gious beliefs became less common, and religious

intermarriage increased (Hendrickx et al. 1991;

Becker and Vink 1994; Felling et al. 2000). In

comparison with the USA or with Southern Europe,

the Netherlands is now relatively secular and it has

experienced stronger trends in this respect than

other countries (Halman and Riis 1999; Stark

1999). The religious groups we consider in our

work are: (i) Catholics, (ii) ‘Dutch Reformed

Protestants’ (referred to as Reformed ), (iii) ‘Re-

Reformed Protestants’ (referred to as Orthodox),

(iv) Jews, and (v) Unaffiliated persons. These

represent the largest groups in the Netherlands

(with the exception of Jews, who form a very small

group).

The role of nationality in the Netherlands is

different from that in traditional immigrant societies.

The most important immigrant groups in the Nether-

lands are the Moroccans and the Turks. Both these

groups were initially recruited as labour immigrants

during the 1960s and 1970s, and both have since then

grown in size, partly through family reunification in

the 1980s and partly through the marrying of spouses

from abroad in the 1990s. The two groups are

nonetheless small, constituting about 4 per cent of

72 Matthijs Kalmijn et al.

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

the population. Levels of intermarriage remain very

low and many Turks and Moroccans marry a spouse

from abroad (Esveldt and Schoorl 1998; Harmsen

1998; Van Huis and Steenhof 2003). The ethnic

groups we consider in our analyses are: (a) Mor-

occan, (b) Turkish, (c) Western European, (d)

Southern European, and (e) Dutch. We use nation-

ality to measure ethnicity but, recognizing that

nationality is a narrow definition of ethnicity, we

use the term ‘nationality intermarriage’ rather than

‘ethnic intermarriage’ to refer to marriages between

partners of different national origin.

We analyse religion and nationality in one study

because they represent old and new bases for group

identification and group solidarity in society. Owing

to secularization on the one hand, and increasing

immigration on the other, we suspect that in

contemporary times religious boundaries are less

salient than ethnic boundaries. As a consequence,

we also believe that intermarriage across religious

boundaries will have a weaker impact on divorce

than intermarriage across nationality boundaries.

Theoretical background and hypotheses

The general hypothesis we test is that marriages

between individuals who differ in religion or nation-

ality have a higher risk of divorce than homogamous

marriages. The underlying reasoning is that differ-

ences in these characteristics will make it more

difficult for partners to understand each other, will

make it more difficult for them to make joint

decisions (e.g., about childbearing and upbringing),

and will lead to more disapproval from their

immediate social world. In further specifying this

hypothesis, we need to take into account the fact

that religious groups and nationalities also have

different risks of divorce. More orthodox religious

groups tend to have a lower risk of divorce than

more liberal groups and the unaffiliated generally

have more unstable marriages than the various

religious groups (Lehrer and Chiswick 1993; Booth

et al. 1995; Wagner and Weiss 2003; Kalmijn et al.

2004). Similarly, there may be differences among

nationality groups in the risk of divorce, depending,

for example, on the value orientation of the sending

country (Jones 1996).

We therefore introduce two hypotheses. The first

hypothesis is the main-effects hypothesis, which

argues that the more traditional the value orienta-

tion of a religious or national origin group, the lower

the risk of divorce. In the Netherlands, the Orthodox

Protestants are the most traditional, the Unaffiliated

are the most liberal, and the Reformed Protestants

and Catholics are in between these extremes (Felling

et al. 2000). The position of Jews in the list is more

difficult to determine but is probably somewhere at

the more liberal end of the continuum (Van Solinge

and De Vries 2001). Using data from the World

Values Studies and from immigrant surveys in the

Netherlands, we expect that Moroccan and Turkish

persons are more traditional, that Western Eur-

opean and Dutch persons are the most liberal, and

that Southern Europeans are in between (Inglehart

1997; Uunk 2003).

Our second hypothesis concerns the effect of the

spouses’ religion and national origin, and argues that

when the religions or national origins of the two

spouses are dissimilar, the risk of divorce is higher.

We call this the heterogamy hypothesis. Assuming

that the main-effects hypothesis is valid, we need to

decide what constitutes evidence for the heterogamy

hypothesis. If the divorce risk of a mixed marriage

(between, say, a member of group A and a member

of group B) is higher than the divorce risk of AA

marriages but lower than the divorce risk of BB

marriages, we argue that adaptation is taking place.

The behaviour of those couples is in between the two

groups, and one can argue that this is simply the

average of the two group effects and not a hetero-

gamy effect (Jones 1996). To analyse real hetero-

gamy effects, we employ both a strong and a weak

form of the heterogamy hypothesis. According to

the strong heterogamy hypothesis, AB marriages will

haveadivorce risk that is higher than the maximum

divorce risk of AA and BB marriages. For example,

we expect that a marriage between a Catholic and an

unaffiliated person will haveadivorce risk that is

higher than the (already) high risk for unaffiliated

couples. According to the weak heterogamy hypoth-

esis, AB marriages will haveadivorce risk that is

higher than the average risk of AA and BB

marriages. In our example, the risk of the mixed

group will be higher than the average of the low risk

for Catholics and the high risk for unaffiliated

couples.

There are different types of heterogamous mar-

riages and this allows us to formulate two additional

hypotheses (cf., Lehrer and Chiswick 1993). One of

these is that the more dissimilar are two groups in

their value orientation, the higher the risk of

divorce. This implies that the highest risk of divorce

will be observed for mixed marriages when one

partner is Orthodox and the other unaffiliated. The

lowest risk will be observed for a marriage in which

one partner is Catholic and the other Reformed. The

other mixed marriages will be in between these two

Intermarriage and the risk of divorce 73

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

extremes. For nationality groups, we expect the

highest risk for mixed marriages between Dutch

and Moroccan spouses or between Dutch and

Turkish spouses. A somewhat lower risk will exist

for a marriage between a Dutch spouse and one

from Southern Europe, and the lowest risk will exist

for a marriage in which one spouse is Dutch and the

other a Western European immigrant.

A second hypothesis is that there can also be a

social boundary between religious or nationality

groups, regardless of the value orientation they

have. We would expect that the stronger the social

boundary between groups, the higher the risk of

divorce. When social boundaries are strong, the

support for the marriage in the social worlds of the

two spouses will be weaker and this may lead to a

higher risk of divorce. This line of reasoning applies

most clearly to the contrast between Catholics and

Protestants, which has historically been an important

divide both in the Netherlands and elsewhere

(Lenski 1961; Hendrickx et al. 1991; Kalmijn

1991). We would therefore expect marriages be-

tween Catholics and Orthodox or Reformed spouses

(both Protestant) to be more unstable than other

mixed marriages. A similar argument applies to the

position of Jews. When looking at intermarriage in

the Netherlands, there appears to be a clear social

divide between Jews and other groups (Ultee and

Luijkx 1998; Van Solinge and De Vries 2001). Hence

we expect mixed marriages involving Jews to be

more unstable than other mixed marriages.

The hypothesized effects of social boundaries and

value differences may both apply at the same time.

In Table 1, we present our predictions in a more

systematic fashion. The assumed cultural distances

in Table 1 are indicated by the letter A and the

assumed social boundaries by the letter B. In most

cases, the combined implications of the two hypoth-

eses are clear, but there are also cases when

predictions cannot be made. Marriages between

the equally traditional Catholics and Reformed

Protestants may be more or less stable than mar-

riages between Reformed and Orthodox Protestants,

depending on whether it is the social boundary or

the value differences that have the greater effect.

A potential problem of interpretation in the

analysis of heterogamy effects is that heterogamous

marriages may have confounding characteristics that

predispose them to higher divorce rates. This is

probably most likely to occur in the case of religion.

It is likely that religiously heterogamous couples

attach less meaning than homogamous religious

couples to religious norms and values. Since fidelity

to religious norms and values reduces the risk of

divorce, mixed couples may be more likely to

divorce, not because they have dissimilar tastes and

values, but because they are less religious (Shehan

et al. 1990). Although the causal nature of the

influence of heterogamy can best be studied in

longitudinal surveys, our register data contain a

unique variable allowing us to rule out a substantial

part of this competing religious interpretation. More

specifically, we know whether couples married in

church, and this gives us information about the

degree to which couples observed religious norms

and values at the time of their wedding. If the

heterogamy hypothesis is true, we would expect the

effect of religious heterogamy to be present both for

couples who did not marry in church and for couples

who did.

Another potential problem lies in the possibility

of conversion. Heterogamous couples may become

homogamous before their wedding if one spouse

switches to the faith of the other. Religious switching

is a frequent response to intermarriage and it can

occur both before and after marriage (Sherkat

1991). Lehrer and Chiswick (1993) found that

marriages in which conversion had taken place

were equally stable and in some cases more stable

than homogamous marriages. This can be due to

such couples having a stronger religious orientation.

Alternatively, we can regard conversion as an

investment in the relationship and assume that

only someone in a very strong relationship would

be willing to make such an investment. The data we

use do not allow us to examine conversions, but we

Table 1 Sources of possible differences in divorce risks between spouses in mixed marriages in the Netherlands:

value

orientation (A) and social boundary (B)

Unaffiliated Catholic Reformed Orthodox Jewish

Unaffiliated 

Catholic A /

Reformed A B /

Orthodox 2A A/BA /

Jewish B A/BA/B2A/B /

Reformed are ‘Dutch Reformed Protestants’, Orthodox are ‘Re-Reformed Protestants’.

74 Matthijs Kalmijn et al.

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

do need to consider the bias that may stem from

including in the homogamous category couples in

which one spouse converted before marriage. If we

could have classified the convert marriages as

heterogamous*

/which would probably have been

more true to the facts*

/the risk of divorce in the

homogamous group would probably not have chan-

ged but the risk of divorce in the heterogamous

group would have declined. Hence, we are probably

overestimating the heterogamy effect somewhat.

A partly similar problem lies in the effect of

naturalization. We are using nationality to define

groups, and a small segment of immigrants are

naturalized (CBS 1997). Like conversion, naturaliza-

tion can be a response to the entry into a mixed

marriage. If naturalization occurs before marriage,

the marriage will be classified as homogamous when

in fact it is mixed. Naturalization can also occur for

other reasons, however, and in such cases, the

possibility exists that some of our mixed marriages

may in fact be homogamous, that is, between a

naturalized and a non-naturalized immigrant. Be-

cause both misclassifications may occur, it is unclear

what the direction of the bias would be. We are not

aware of studies that investigate the relationship

between naturalization and intermarriage.

Our final hypothesis concerns the timing of

divorce. We compare divorce after the first 5 years

with divorce in the next 5 years for marriages still

intact after the first 5 years. If religious or nationality

differences form an impediment to the viability of

the marriage, we think this will probably become

apparent early on in the marriage. Later in the

marriage, such differences will either have been

resolved*

/people can learn to live with their differ-

ences and the social surrounding can gradually

become more accepting of the marriage *

/or the

marriages with the greatest differences will not

have survived, leaving a less divorce-prone hetero-

gamous group behind (South and Spitze 1986). For

that reason, we expect the heterogamy effect to be

more pronounced in the first 5 years than in the

second.

Data

All municipalities in the Netherlands provide in-

formation about all marriages and divorces regis-

tered in their municipality. For our analysis, we

consider all marriages formed in the period 1974

/84

and we trace these marriages in the divorce records

for the period 1974

/94. This design enables us to

assess the risk of divorce in the first 10 years of

marriage. For the earlier marriages, we could also

examine divorce risks at longer durations, but for the

sake of simplicity, we focus on the 10-year risk only.

Note that divorces that occurred abroad cannot be

studied with the data at hand, which may lead to an

underestimation of the divorce risk for foreign

nationality groups.

To match divorce to marriage records, we used a

combination of characteristics that are available in

both files: the municipality of the marriage, the exact

date of the marriage, and the years of birth of both

spouses. In combination, these characteristics pro-

vide an almost unique key; only 3 per cent of the

marriages in the period 1974

/84 have a duplicate

key. We removed these duplicates before tracing

marriages in the divorce records. After these pre-

parations, the total number of marriages that we

could analyse was 931,198. Of these marriages,

116,269 (12.5 per cent) were divorced within 10

years.

The accuracy of the matching operation was

vulnerable to registration and coding errors. The

only way of assessing the number of errors was by

tracing ‘in reverse’, that is, tracing from divorce to

marriage records. For every divorce, we should have

been able to find a marriage, and the number of

marriages we were unable to find provides a clue to

the quality of our matching procedure. Of the

divorces in the period 1974

/94 (occurring to mar-

riages registered in 1974

/84), we were able to find

92 per cent in the 1974

/84 marriage records. We

conclude that, although it was not 100 per cent

accurate, the matching operation was relatively

successful.

Measurement of religion and national origin

All characteristics of husband and wife are taken

from the marriage registration data and refer to the

time of marriage. With respect to church affiliation,

seven categories are available: no church affiliation,

Roman Catholic, Reformed, Orthodox, Otherwise

Protestant, Jewish, and Other/Unknown. A possible

disadvantage of our measure is that some people

may have abandoned their religious affiliation with-

out officially reporting this on the municipal regis-

tration form on the day of their wedding. We do not

believe this is a problem, however. Survey data of

the 1970s and 1980s indicate that, in the period

under investigation, about 25

/30 per cent of the

population were not church members (Becker and

Vink 1994), and these figures correspond quite well

Intermarriage and the risk of divorce 75

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

with the number of non-church members in the

marriage files.

Our approach to nationality intermarriage is

based on the information available on the marriage

record: the nationality of husband and wife. Nation-

ality is recoded into six categories: Dutch, Western

Europe, Southern Europe (Portugal, Spain, Italy, the

former Yugoslavia, and Greece), Turkey, Morocco,

and Other. As noted earlier, nationality is a rela-

tively narrow definition of ethnicity. In the 1990s,

about a third of first-generation Turks and about a

quarter of first-generation Moroccans had Dutch

nationality (CBS 1997). When someone has both

Dutch and a foreign nationality, Statistics Nether-

lands codes this on the marriage record as Dutch

nationality. For the second generation, the percen-

tages of Turks and Moroccans with Dutch nation-

ality are somewhat higher than for the first

generation.

The ‘Other’ categories of the religion and nation-

ality variables cannot be differentiated further.

Since these categories are heterogeneous, their

marriage parameters are difficult to interpret and

we do not present them in the tables and text.

Because the ‘Other Protestant’ group is also

mixed, containing both liberal and more fundamen-

talist groups, we do not use marriages with a spouse

in this category for testing the heterogamy hypoth-

esis. We do keep all the ‘Other’ categories

in the analyses. We do not have a special category

for Islam in the data. Since virtually all Turks

and Moroccans in the Netherlands are Islamic

(Van Tubergen 2003), analysing these two dimen-

sions simultaneously would not yield more informa-

tion.

The percentage distribution of brides and grooms

married between 1974 and 1984 by religion, nation-

ality, and other characteristics is presented in

Table 2.

Method

In the analyses we look at the observed divorce risks

after 10 years of marriage for every combination of

church affiliation and nationality. Next we look at

divorce risks obtained from a logistic regression

model, in which the influence of the two types of

heterogamy are analysed simultaneously, and in

which we also statistically control for other demo-

graphic characteristics. We initially focus on the

contrast between divorce and no divorce within 10

years. To test our hypothesis about the timing of

divorce, we estimate the model separately for

divorce after the first 5 years and for divorce in

the ensuing 5 years for marriages still intact after the

first 5 years. We abstain from applying event history

techniques, primarily because it was not possible to

estimate such models with the sample size available.

Since we do not have time-varying covariates, we do

not think this is an important disadvantage.

Our logistic regression model contains the two

central independent variables: (i) a set of 48

(7

/7 categories minus 1 reference category)

dummy variables for the combination of husband’s

and wife’s church affiliation, (ii) a set of 35 (6

minus 1) dummy variables for the combination of

husband’s and wife’s nationality. These dummy

Table 2 Percentage distribution of brides and grooms

married in the Netherlands 1974

/84, by religion, nation-

ality, and other characteristics

Husband Wife

Religion None

26.6 25.0

Catholic

41.7 42.9

Reformed

17.9 18.4

Orthodox

9.2 9.5

Other Protestant

0.7 0.7

Jewish

0.1 0.1

Other

3.8 3.4

Nationality Dutch

95.5 96.0

Western

European

1.3 1.2

Southern

European

0.6 0.4

Turkish

0.4 0.3

Moroccan

0.3 0.2

Other

1.8 2.0

Age B

/20

2.0 14.9

20

/24

48.4 58.8

25

/29

32.7 16.9

30

/34

9.4 5.4

35

/39

3.9 2.3

40

/44

2.1 1.1

45

/49

1.4 0.5

Previous marital status Never married

90.3 91.5

Widowed

0.7 0.4

Divorced

9.1 8.0

Married in church Yes 51.0

No 49.0

Population size in 

/100,000 17.6

municipality of 50,000

/100,000 19.9

marriage 20,000

/50,000 20.0

10,000

/20,000 22.7

/10,000 19.0

Source: Marriage and Divorce Files, Statistics Netherlands

(CBS), N

/931,198. Province and year of marriage not

presented.

76 Matthijs Kalmijn et al.

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

variables capture both the main effects and the

interaction effects. The model also contains the

following control variables: (i) a set of 48 dummy

variables for the combination of husband’s and

wife’s age group at marriage, (ii) a set of 8 dummy

variables for the combination of husband’s and

wife’s previous marital status, (iiii) a set of 10

dummy variables for each year of marriage, (iv)a

set of 4 dummy variables for the degree of urbaniza-

tion of the municipality in which the marriage took

place, (v) a set of 11 dummy variables for the

province in which the marriage took place.

Using the parameters of this model, we calculate

predicted divorce risks. To calculate the predicted

risk, all the independent values were set at a mean

value and the marital-status variable was set to the

most common value (never married). The corrected

risks give us the possibility of determining the net

contribution of heterogamy in religion and nation-

ality, which is necessary since these types of hetero-

gamy may be correlated with other variables. It is

also important to consider nationality and religion in

one model because the religious composition of the

nationally mixed marriages will not be random. For

example, it is plausible that the native spouse of

someone of different national origin will be unaffi-

liated, and this alone will already lead to a relatively

high risk of divorce. We note that not all combina-

tions will be present in the data (e.g., there will be

few Catholic Turks). This would be a statistical

problem only if interaction effects between religion

and nationality on divorce were considered, but we

do not consider them.

Note that the data do not allow us to detect

mortality. If one or both spouses die, the marriage is

treated as ‘not divorced’. Although we did not

expect strong biases from mortality in the effects

of heterogamy on divorce, we still decided to limit

our analysis to marriages in which both spouses were

younger than age 50 at the time of marriage.

Because we focus on the first 10 years of marriage,

this reduces mortality effects considerably. Survival

chances between ages 50 and 60 are high in the

Netherlands*

/90 per cent for men and 93 per cent

for women.

Analyses and results

In Table 3 we show the 10-year probabilities of

divorce for marriages representing all combinations

of husband’s and wife’s religious affiliation. Table 7

shows the probabilities for all combinations of

husband’s and wife’s nationality. When testing our

hypotheses, we draw attention to the corrected risks

of divorce. Tables 4

/6 present results from the

models focusing on the effects of religious hetero-

gamy. Tables 8 and 9 present information on the

effects of nationality differences between husband

and wife.

Religion

Table 3 shows the risk of divorce by the religious

affiliation of husband and wife and the frequencies

of occurrence of all combinations of religious

affiliation in marriages in the Netherlands between

1974 and 1984. It is clear that most people marry

within their own group. Since intermarriage is highly

dependent on the relative size of a group*

/smaller

groups being less able to marry endogamously*

/we

need to use odds ratios to measure differences in the

likelihood of marriage within the same group

(Kalmijn 1998). The odds ratios, which are not

presented in Table 3, can be defined as the odds

that, for example, a Catholic marries a Catholic

(rather than a non-Catholic) divided by the odds

that a non-Catholic marries a Catholic (rather than a

non-Catholic). The odds ratios are highest for the

Jews (759) and the more orthodox segment of the

Protestant church, the Orthodox (31). The odds

ratios are also strong for the Catholics (19), perhaps

as a result of geographic segregation. The Reformed,

which is the most liberal religious group in the

Netherlands, have the lowest odds ratio (10), even

lower than the non-church members (12).

Table 3 also allows us to test the main-effects

hypothesis. To do this, we focus on the observed

divorce risks of homogamous couples (on the

diagonal). Homogamous marriages between non-

church members have the highest risk of divorce:

18.6 per cent divorce within 10 years. The 10-year

divorce risks are lowest for the three main religious

groups in the Netherlands: 9.7 per cent for Catholics,

6.7 per cent for the Reformed, and 4.5 per cent for

the Orthodox. The 10-year divorce risk is 16.9 per

cent for marriages among Jews. This order of divorce

risks is in line with our main-effects hypothesis. The

Unaffiliated (and the Jews) are the most liberal in

their value orientations and consequently have the

highest risk, whereas the Orthodox are the most

traditional and consequently have the lowest risk.

The divorce risk of Catholics is somewhat higher

than expected, however.

Is the risk of divorce higher for heterogamous

marriages? To assess the impact of heterogamy,

Intermarriage and the risk of divorce 77

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

we need to compare each mixed combination in two

ways*

/with the maximum risk of divorce in the two

corresponding homogamous groups, and with the

average risk of divorce in the two groups. These

comparisons are made directly in Table 4. The first

column of numbers shows the divorce percentage for

the mixed group, the second shows, as a ratio, the

deviation of this percentage from the maximum

level, and the third shows, again as a ratio, the

deviation of this percentage from the average level.

Thus, a deviation with a value over one indicates a

heterogamy effect. Note that because we have

population data, significance tests for these ratios

are not applicable.

We first look at the average of the ratios of all the

combinations, presented at the bottom of Table 4.

These averages show that support for the hetero-

gamy hypothesis is weak. The overall deviation from

the maximum is 1.06, and the overall deviation from

the average is 1.29. Hence, for all mixed marriages

combined, we find little support for the strong

heterogamy hypothesis but clear support for the

weak heterogamy hypothesis. These figures merely

present a summary of the results and do not reveal

effects that may exist for specific combinations.

When we look at combinations involving two

religious groups, there is clearer evidence for a

heterogamy effect. The strongest effects are ob-

served for marriages between Catholics and Protes-

tants. Marriages between a Catholic and an

Orthodox person havea16

/22 per cent higher risk

than the maximum and a 51 

/57 per cent higher risk

than the average (depending on whether the hus-

band or the wife is the Catholic). Similarly, a

marriage between a Catholic and a Reformed

person (the most common mixed combination

between church members) has a 20

/22 per cent

higher risk of divorce than the maximum level and

about a 34

/36 per cent higher risk of divorce than

the average level. These deviations are clear and

consistent with the heterogamy hypothesis. Less

convincing evidence exists for the combination of

different Protestant denominations. Marriages be-

tween the Reformed and the Orthodox, another

common combination, have a higher risk of divorce

only when they are compared to the average level (a

deviation of 16

/27 per cent). When we compare the

divorce risks of marriages within the Protestant

Churches to the maximum level of divorce, no effect

can be observed. In other words, religiously mixed

combinations of Protestants and Catholics show a

heterogamy effect but combinations within the

Protestant Churches do not. These findings point

more strongly in the direction of the hypothesis

about a social divide than towards a difference in

value orientations. Catholics and Reformed Protes-

tants do not have very different values, but they do

experience a sharp, historically grown, social bound-

ary. Their divorce risk is consequently high. Re-

formed and Orthodox Protestants do not havea

Table 3 Observed 10-year probabilities (percentages) of divorce for marriages representing all combinations of husband’s

and wife’s religion in the period 1974 

/84, the Netherlands (number of marriages in parentheses)

Religion of wife

Religion of husband Unaffiliated Catholic Reformed Orthodox

Other

Protestant Jewish

Unaffiliated 18.6 16.9 13.6 13.2 18.0 32.4

(162,221) (42,680) (27,401) (9,756) (1,875) (204)

Catholic 18.3 9.7 12.0 12.4 18.8 26.3

(34,956) (307,713) (31,575) (8,485) (1,949) (133)

Reformed 14.5 12.4 6.7 7.3 13.2 

(22,230) (33,071) (89,836) (17,734) (1,310) (65)

Orthodox 14.0 12.0 6.7 4.5 11.7 

(7,327) (8,527) (17,791) (50,531) (472) (27)

Other Protestant 17.7 14.0 11.2 10.0 11.7 

(1,285) (1,774) (1,286) (468) (1,081) (3)

Jewish 34.4 27.6 23.1 

/ / 16.9

(337) (196) (130) (60) (12) (278)

Note: For cells in which N B

/100, percentages were not calculated. ‘Other’ categories not presented in the table.

Source: As shown in Table 2.

78 Matthijs Kalmijn et al.

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

sharp social divide. They do have very different

values but this difference does not increase their

divorce risk.

Another very common type of mixed marriage is a

marriage between a religious person and someone

not affiliated to any church. In Table 4, we observe

that these mixed marriages provide less support for

the heterogamy hypothesis. First of all, the level of

divorce in these types of marriages is never higher

than the maximum level. More specifically, when we

compare these couples to marriages between two

unaffiliated persons, the latter always have the

highest risk. Following Jones (1996), this suggests

that some form of adaptation is taking place.

Second, when we compare mixed marriages invol-

ving unaffiliated persons to the average level of

divorce in the two corresponding homogamous

unions, we do find some heterogamy effects but

these are not always very strong: 11

/12 per cent

higher when marrying a Reformed Protestant, 20

24 per cent higher when marrying an Orthodox

Protestant, and 11

/18 per cent higher when marry-

ing a Catholic. The findings are partly in line with

the hypothesis about differences in value orienta-

tions. The sharpest differences in value orientations

are between unaffiliated and Orthodox groups,

and these also have the highest deviation from

the average. We should note, however, that this

conclusion does not apply to deviations from the

maximum.

Because our database is so large, we can also focus

on small religious groups in Dutch society. As is

clear from Table 4, evidence for heterogamy effects

is strongest for the Jewish combinations, especially

when the husband is Jewish. Marriages between a

Catholic wife and a Jewish husband have a 45 per

Table 4 Corrected

10-year probabilities of divorce for religiously heterogamous couples compared with corrected

probabilities for homogamous couples, the Netherlands, marriages 1974

/84

Ratio of corrected divorce risk to

maximum/average risk in

corresponding homogamous

groups

Religious affiliation of spouses Type of mix

Corrected

probability of

divorce (per cent)

Ratio to

maximum

Ratio to

average

Unaffiliated Catholic H

14.3 0.917 1.113

W

15.1 0.978 1.175

Unaffiliated Reformed H

13.1 0.840 1.110

W

13.2 0.846 1.119

Unaffiliated Orthodox H

12.6 0.808 1.200

W

13.0 0.833 1.238

Unaffiliated Jewish H

15.7 1.006 1.189

W

18.8 1.205 1.424

Catholic Reformed H

12.1 1.198 1.337

W

12.3 1.218 1.359

Catholic Orthodox H

12.2 1.208 1.574

W

11.7 1.158 1.510

Catholic Jewish H

12.2 1.130 1.167

W

15.7 1.454 1.502

Reformed Orthodox H

8.5 1.063 1.269

W

7.8 0.975 1.164

Reformed Jewish H 

13.1 1.213 1.394

W



/ / /

Average

1.062 1.285

Probabilities corrected using logistic regression with the following variables: combination of husband’s and wife’s religious

affiliation, national origin, age, and marital status, and the province and urbanization of the place of marriage.

Note: H means husband, W means wife. H

/W means the first column applies to the husband and the second column applies

to the wife.

Source: As shown in Table 2.

Intermarriage and the risk of divorce 79

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

cent higher risk of divorce than the maximum and a

50 per cent higher risk than the average. Even more

striking is the fact that heterogamy effects are also

found for combinations involving unaffiliated per-

sons. Marriages between a Jewish husband and an

unaffiliated wife have a 21 per cent higher risk of

divorce than marriages between two unaffiliated

persons, and a 42 per cent higher risk than the

average for homogamous marriages of unaffiliated

persons and of Jews. These findings are clearly more

in line with the hypothesis about social boundaries

than with the hypothesis about value differences.

Both Jews and unaffiliated persons are liberal in

their values and behaviour, as can also be seen in

their divorce behaviour, but a mix between them

nonetheless increases the chance of divorce.

To test our hypothesis about the duration of

marriage, we re-analyse the model for two divorce

risks: the risk in the first 5 years, and the conditional

risk in the second 5 years. To simplify the inter-

pretation, we use a more basic model for this

analysis. The model focuses only on the four large

religious denominations and ignores the asymme-

tries in the effects of heterogamy.

Table 5 shows small differences in the effects of

heterogamy on the risk of divorce at different points

in the marriage. For example, couples in which one

spouse is Catholic and the other unaffiliated havea

17 per cent higher-than-average risk of divorce in the

first 5 years and a 13 per cent higher risk in the next

5 years. Comparing the percentages for other

combinations leads to similar conclusions: differ-

ences are either absent or small. This evidence

contradicts our hypothesis.

To what extent can the effect of heterogamy be

attributed to mixed couples being less religious? To

answer this question, we incorporate information on

whether couples married in church. Table 6 shows

the effects of religious intermarriage on the risk of

divorce for couples who married in church and for

other couples. We present the results for the first

5 years only.

The results in Table 6 first show that heterogamy

effects also exist for couples who did not marry in

Table 6 Comparison of ratios of corrected

divorce risk to maximum and average risk in corresponding homogamous

groups in religiously heterogamous marriages for couples who married in church and those who did not, the Netherlands,

marriages 1974 

/84

Ratio to maximum Ratio to average

Religious affiliation of spouses In church Not in church In church Not in church

Unaffiliated Catholic 0.808 1.027 1.042 1.018

Unaffiliated Reformed 0.777 0.867 1.117 1.068

Unaffiliated Orthodox 0.769 0.860 1.198 1.111

Catholic Reformed 1.120 0.905 1.310 1.121

Catholic Orthodox 1.104 0.863 1.456 1.122

Reformed Orthodox 1.088 1.183 1.260 1.260

Average 0.944 0.951 1.230 1.117

See footnote to Table 4.

Source: As shown in Table 2.

Table 5 Comparison of ratios of corrected

divorce risk to maximum and average risk in corresponding homogamous

groups in religiously heterogamous marriages in the ﬁrst and second 5 years of marriage, the Netherlands, marriages 1974

Ratio to maximum Ratio to average

Religious affiliation of spouses 0

/5 years 6/10 years 0/5 years 6 /10 years

Unaffiliated Catholic

0.957 0.917 1.174 1.130

Unaffiliated Reformed

0.854 0.827 1.148 1.101

Unaffiliated Orthodox

0.804 0.821 1.261 1.200

Catholic Reformed

1.218 1.243 1.374 1.374

Catholic Orthodox

1.096 1.275 1.526 1.599

Reformed Orthodox

1.007 1.047 1.288 1.207

Average

0.989 1.022 1.295 1.269

See footnote to Table 4.

Source: As shown in Table 2.

80 Matthijs Kalmijn et al.

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

church. For example, couples in which one spouse is

Catholic and the other spouse is Reformed havea

12 per cent higher-than-average risk of divorce if

they did not marry in church. Next, we observe that

the effects of heterogamy are generally higher if

couples married in church. For example, the divorce

risk of a Catholic

/Reformed couple is 31 per cent

higher than average if the couple married in church.

Similar differences are found for other combina-

tions. The unweighted average of the heterogamy

effects is 12 per cent for couples who did not marry

in church and 23 per cent for couples who married in

church. Our conclusion is that the heterogamy

effects are only partially spurious. Part of the overall

heterogamy effect is due to mixed couples being less

religious (as indicated by whether the wedding was

in church). But since, even for couples who did not

marry in church, a heterogamy effect occurs, the

effect of heterogamy cannot be attributed comple-

tely to the selectivity of religiously heterogamous

marriages.

Nationality

In Table 7, we present the absolute numbers of

nationality combinations and their observed divorce

risks. In Table 8, we present our calculations of the

heterogamy effects. The frequencies in Table 7 show

that the percentages of mixed marriages vary greatly

from group to group. The Turkish and the Moroc-

cans have the highest degree of endogamy. Of

the Turkish who married in the Netherlands between

1974 and 1984, 74 per cent of marriages were

endogamous. For Moroccans, the figure is 54 per

cent. Percentages for Western Europeans and South-

ern Europeans are much lower (14 and 21 per cent,

respectively). Odds ratios (not presented in the

table) reveal that all groups are more likely

than expected to marry within their own group,

and that the Turkish group is most closed (8,421),

followed by the Moroccans (1,992), the Southern

Europeans (68), the Western Europeans (15), and

the Dutch (12). Finally, we observe in Table 7 the

well-known tendency toward gender asymmetry:

if there are ethnically mixed couples, it is

usually minority men marrying Dutch women, and

not the other way around (cf., Kalmijn 1993; Qian

1997).

To test the main-effects hypothesis, we first look

at the diagonal in Table 7. The divorce risk of

homogamous marriages between two spouses who

have Dutch nationality is the highest (11.4 per cent).

Marriages of Western Europeans and marriages of

Southern Europeans have a lower risk of divorce

(7.6 and 9.3, respectively). In line with what

one would expect on the basis of traditional value

orientations, we find that Turkish marriages

and Moroccan marriages have the lowest risk of

divorce.

Table 7 Observed 10-year probabilities (percentages) of divorce for marriages representing all combinations of husband’s

and wife’s nationality in the period 1974 

/84, the Netherlands (number of marriages in parentheses)

Nationality of wife

Nationality of husband Dutch

Western

European

Southern

European Turkish Moroccan

Dutch 11.4 22.3 28.1 39.2 63.6

(862,995) (8,572) (2,712) (158) (275)

Western European 15.8 7.6 

/ / /

(9,742) (1,570) (39) (0) (3)

Southern European 24.3 

/ 9.3 / /

(4,708) (63) (1,047) (1) (4)

Turkish 56.0 

/ / 0.7 /

(1,385) (19) (20) (2,288) (3)

Moroccan 52.2 

/ / / 1.6

(1,706) (26) (34) (2) (1,258)

Note: Western European countries are Belgium, France, Germany, Luxembourg, and UK. Southern European countries are

Greece, Italy, Portugal, Spain, and (former) Yugoslavia. Percentages are not presented for cells in which N B

/100. ‘Other’

nationalities not presented in the table.

Source: As shown in Table 2.

Intermarriage and the risk of divorce 81

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

Are there effects of heterogamy on the risk of

divorce? Table 8 shows that the answer is clear: most

mixed combinations have a risk of divorce that is

higher than the highest level of divorce in the two

homogamous groups. The average ratio is 2.02,

indicating that mixed marriages have a risk of

divorce twice as high as that of the maximum level

of divorce in the two corresponding groups. This

effect is quite strong and clearly supports the strong

heterogamy hypothesis.

We also find variations in the magnitude of the

effects that are consistent with our hypothesis about

value orientations. Combinations of Dutch and

Turkish or Moroccan persons reveal a stronger

heterogamy effect than combinations involving

Dutch and Western European persons. The effects

for combinations involving Southern Europeans are

in between the combinations with Turks or Mor-

occans and the combinations with Western Eur-

opeans. When looking at combinations involving

minority men, the differences are quite strong. The

ratio is 4.7 for combinations involving Turkish men,

2.4 for combinations involving Moroccan men, and

1.5 for combinations involving Western European

men. Because European groups are more similar

than Moroccan and Turkish groups to the Dutch in

values and lifestyle, this finding is consistent with

theoretical interpretations of the heterogamy effect

in terms of value similarity.

The observed probabilities of divorce are quite

high for mixed marriages (Table 7). The high divorce

rates of marriages between persons of Dutch nation-

ality and persons with another nationality may be

partly a consequence of marriages conducted solely

for the purpose of securing legal residence docu-

ments. After 3 years of temporary residence while

married to a Dutch person, a foreigner obtains

permanent residence documents. Recent analyses

indicate that the divorce rate of Dutch

/foreigner

couples increases sharply in the third year of

marriage, especially for those in which the foreign

spouse is Moroccan, suggesting that these paper

marriages do indeed occur (Van Huis and Steenhof

2003). Consistent with this, we also find that effects

of nationality heterogamy decrease with the dura-

tion of marriage (Table 9). For example, couples

with a Moroccan husband and a Dutch wife havea

divorce risk that is 4.6 times higher than average in

the first 5 years but only 2.7 times higher in the next

5 years. Similar differences occur in the other

combinations. On average the factor is 3.4 and 2.4,

respectively.

Conclusion

We matched marriage records to divorce records

and calculated the 10-year risk of divorce, using

multivariate logistic regression analyses. Using

Table 8 Corrected

10-year probabilities of divorce for couples heterogamous in nationality compared with corrected

probabilities for homogamous couples, the Netherlands, marriages 1974

/84

Ratio of corrected divorce risk

to maximum/average risk in

corresponding

homogamous groups

Nationality of spouses

Type of

marriage

Corrected

probability

of divorce (per cent)

Ratio to

maximum

Ratio to

maximum

Dutch Western European H 

/W 13.7 1.522 2.108

W

/H 9.8 1.089 1.508

Dutch Southern European H 

/W 18.0 2.000 2.222

W

/H 15.6 1.733 1.926

Dutch Turkish H

/W 19.4 2.156 3.079

W

/H 29.3 3.256 4.651

Dutch Moroccan H

/W 41.7 2.673 3.390

W

/H 29.7 1.904 2.415

Average 2.020 2.614

See footnote to Table 4.

Note: H means husband, W means wife. H

/W means the first column applies to the husband and the second column applies

to the wife.

Source: As shown in Table 2.

82 Matthijs Kalmijn et al.

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

straightforward analyses of nearly a million mar-

riages, we have demonstrated that there is a modest

relationship between religious heterogamy and di-

vorce and a strong relationship between nationality

heterogamy and divorce. This conclusion is based on

registration data for all marriages contracted in the

Netherlands between 1974 and 1984, and registra-

tion data for all divorces between 1974 and 1994.

The effects of religious heterogamy are strongest

for combinations involving Catholics and combina-

tions involving Jews. Mixed combinations within the

Protestant population reveal weaker effects. For

combinations involving the unaffiliated, we also

find somewhat elevated divorce risks, but not when

comparing the risk to that of homogamous unaffi-

liated marriages. Overall, the religious heterogamy

effect is about 6 per cent when using a strict

definition of a heterogamy effect (higher than the

maximum of the two corresponding homogamous

combinations). In many cases, the divorce risk of a

mixed marriage is located somewhere in between

the risks of the two groups. However, this is always

above the average, not below the average. More

specifically, the risk is 29 per cent above the average,

which means that the risk is clearly pulled in the

direction of the most divorce-prone partner. We

consider this a weak form of a heterogamy effect.

Comparing specific types of mixed marriages

yields additional insights into the underlying reasons

for a heterogamy effect. The evidence suggests that

for religion, the social boundaries between groups

have a more important effect than value disagree-

ments on the risk of divorce. Combinations involving

Catholics and Protestants have an increased risk of

divorce, but there is no clear additional tendency

towards instability for mixed marriages involving the

more traditional Orthodox Protestants. In addition,

marriages between Catholics and Reformed Protes-

tants have an elevated risk of divorce, while these

groups have rather similar values, especially in the

important domain of family values (Felling et al.

2000). Hence, the social divide between Catholics

and Protestants dominates the results, rather than

the value differences between these groups. Simi-

larly, Jewish mixed marriages are more unstable, and

this is also true when the spouse is unaffiliated. Since

Jewish persons in the Netherlands are as liberal as

unaffiliated persons, this also points to social bound-

aries rather than to value differences.

For the divorce risk of nationality heterogamy, the

effects are much stronger, and simpler to interpret.

Marriages of Dutch persons to foreigners havean

average divorce risk twice the maximum level of the

two combinations. This effect is much stronger than

the effect of religious heterogamy, suggesting that

new boundaries in society have become more

important than old boundaries. We also found that

the heterogamy effect is stronger for those nation-

ality groups that are culturally more dissimilar from

the Dutch, and this further supports the value

interpretation of the heterogamy effect. Hence, for

religion, the effect of social boundaries seems the

more important, while for nationality, it is the effect

of value differences that seems more important.

We have found nuanced evidence for the hetero-

gamy hypothesis, a hypothesis which has not often

been tested in Europe. Although we find that

heterogamy matters, the question of its theoretical

interpretation is still open. One argument focuses on

selectivity and suggests that heterogamous couples

Table 9 Comparison of ratios of corrected

divorce risk to maximum and average risk in corresponding homogamous

groups for marriages heterogamous in nationality in the ﬁrst and second 5 years of marriage, the Netherlands, marriages

1974

/84

Ratio to maximum Ratio to average

Nationality of spouses 0

/5 years 6/10 years 0/5 years 6 /10 years

Dutch Western European H

/W 1.571 1.490 2.151 2.030

W

/H 1.166 1.082 1.820 1.475

Dutch Southern European H

/W 2.513 1.663 2.760 1.745

W

/H 1.932 1.780 4.107 1.867

Dutch Turkish H

/W 2.454 2.109 3.316 3.108

W

/H 4.015 3.375 3.613 4.972

Dutch Moroccan H

/W 4.086 1.093 5.213 1.451

W

/H 2.119 2.015 4.611 2.675

Average 2.482 1.826 3.449 2.416

See footnote to Table 4.

Source: As shown in Table 2.

Intermarriage and the risk of divorce 83

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

have special traits that tend to make their marriages

more unstable. We have been able to address one

such possibility by looking at information about

church weddings. Using church wedding as an

indicator of religiosity, we find that part but not all

of the heterogamy effect can be attributed to

heterogamous couples being less religious. For the

effect of nationality intermarriage and for other

possible forms of selectivity, we did not have

appropriate controls. While our data source is

unique and powerful, it primarily serves to assess

and describe a possible heterogamy effect in all its

detail. To explain such effects, a more comprehen-

sive set of variables is needed, a set that cannot be

found in vital statistics. Our approach is a precursor

to such an enterprise, not an alternative.

Notes

1 Matthijs Kalmijn is professor of sociology at Tilburg

University. E-mail m.kalmijn@uvt.nl. Paul M. de Graaf

is associate professor of sociology at Radboud Univer-

sity Nijmegen. Jacques Janssen is a researcher at

Medtronic Bakken Research Center, Maastricht. The

analyses reported in this paper are part of the PhD

dissertation of Jacques P. G. Janssen.

2 The authors thank Statistics Netherlands (Sector Popu-

lation) in Voorburg for letting them analyse the data on

site. We also thank the members of the seminar Divorce

in the Netherlands for comments on earlier drafts. The

research reported here is subsidized by the Netherlands

Organization for Scientiﬁc Research (NWO) and is part

of the research programme ‘Causes and consequences

of divorce in the Netherlands since the 1940s’, coordi-

nated by Matthijs Kalmijn and Paul de Graaf.

References

Becker, Gary S., Elisabeth M. Landes, and Robert T.

Michael. 1977. An economic analysis of marital in-

stability, Journal of Political Economy 85: 1141 

/1187.

Becker, J. W. and R. Vink. 1994. Secularisatie in Nederland

1966

/1991. Sociaal-Culturele Studies No. 19 [Secular-

isation in the Netherlands 1966 

/1991]. Rijswijk: Sociaal

en Cultureel Planbureau (SCP).

Booth, A., D. R. Johnson, A. Branaman, and A. Sica. 1995.

Belief and behavior: does religion matter in today’s

marriage?, Journal of Marriage and the Family 57: 661

671.

Bru¨ derl, Josef and H. Engelhardt. 1997. Trennung oder

Scheidung? Einige methodologische U

berlegungen zur

Deﬁnition von Eheauﬂo¨ sungen, Soziale Welt 48: 277

289.

Bumpass, Larry L. and James A. Sweet. 1972. Differentials

in marital instability: 1970, American Sociological

Review 37: 754 

/766.

Burchinal, L. B. and L. E. Chancellor. 1963. Survival rates

among religiously homogamous and interreligious mar-

riages, Social Forces 41: 353

/362.

CBS [Statistics Netherlands]. 1997. Eerste en Tweede

Generatie Turken en Marokkanen naar Nationaliteit

en Leeftijd [First and second generation Turks and

Moroccans by nationality and age], Maandstatistiek van

de Bevolking 45(July): 4.

Esveldt, Ingrid and Jeannette J. Schoorl. 1998. Verander-

ingen in de huwelijkssluiting van Turken en Marokka-

nen in Nederland [Changes in marriage formation of

Turks and Moroccans in the Netherlands], Bevolking en

Gezin 27: 53 

/86.

Felling, A., J. Peters, and P. Scheepers. 2000. Individuali-

sering in Nederland aan het eind van de twintigste eeuw

[Individualization in the Netherlands at the End of the

Twentieth Century] . Assen: Van Gorcum.

Finna¨s, Fjalar 1997. Social integration, heterogeneity, and

divorce: the case of the Swedish-speaking population in

Finland, Acta Sociologica 40: 263 

/277.

Glenn, N. D., S. K. Hope, and D. Weiner. 1974. Social class

heterogamy and marital success: a study of the empiri-

cal adequacy of a textbook generalization, Social

Problems 21: 539 

/550.

Halman, Loek and Ole Riis (eds.). 1999. Religion in

Secularizing Society: The European’s Religion at the

End of the 20th Century. Tilburg: Tilburg University

Press.

Harmsen, C. N. 1998. Naar geboorteland gemengde

huwelijken [Mixed marriages with respect to country

of origin], Maandstatistiek van de Bevolking 46:

13

/15.

Heaton, T. B. 1984. Religious homogamy and marital

satisfaction reconsidered, Journal of Marriage and the

Family 46: 729 

/733.

Hendrickx, John, Jan Lammers, and Wout Ultee. 1991.

Religious assortative marriage in the Netherlands,

1938

/1983, Review of Religious Research 33: 123/

145.

Inglehart, Ronald 1997. Modernization and Postmoderni-

zation: Cultural, Economic and Political Change in 43

Societies. Princeton, NJ: Princeton University Press.

Jalovaara, M. 2003. The joint effects of marriage partners’

socioeconomic positions on the risk of divorce, Demo-

graphy 40: 67 

/81.

Jones, Frank L. 1996. Convergence and divergence in

ethnic divorce patterns: a research note, Journal of

Marriage and the Family 58: 213 

/218.

84 Matthijs Kalmijn et al.

Downloaded By: [Ingenta Content Distribution] At: 13:16 22 October 2007

Kalmijn, Matthijs 1991. Shifting boundaries: trends in

religious and educational homogamy, American Socio-

logical Review 56: 786

/800.

Kalmijn, Matthijs 1993. Trends in black/white intermar-

riage, Social Forces 72: 119 

/147.

Kalmijn, Matthijs 1998. Intermarriage and homogamy:

causes, patterns, trends, Annual Review of Sociology 24:

395

/421.

Kalmijn, Matthijs, Paul M. De Graaf, and Anne-Rigt

Poortman. 2004. Interactions between cultural and

economic determinants of divorce in the Netherlands,

Journal of Marriage and Family 66: 75

/89.

Lehrer, Evelyn L. and Carmel U. Chiswick. 1993. Religion

as a determinant of marital stability, Demography 30:

385

/404.

Lehrer, Evelyn 1996. The determinants of marital stability:

a comparative analysis of ﬁrst and higher order

marriages, in Paul Schultz (ed.), Research in Population

Economics. Greenwich: JAI Press, pp. 91

/121.

Lenski, Gerhard 1961. The Religious Factor: A Sociolo-

gist’s Inquiry. Garden City: Doubleday.

Michael, Robert 1979. Determinants of divorce, in Louis

Levy-Garboua (ed.), Sociological Economics. London:

Sage, pp. 223

/269.

Monahan, T. P. and W. M. Kephart. 1954. Divorce and

desertion by religious and mixed-religious groups,

American Journal of Sociology 59: 454 

/465.

Qian, Z. 1997. Breaking the racial barriers: variations in

interracial marriage between 1980 and 1990, Demogra-

phy 34: 263 

/276.

Schwertfeger, M. M. 1982. Intraethnic marriage and

divorce in Hawaii: a panel study of 1968 ﬁrst marriages,

Marriage and Family Review 5: 49

/59.

Shehan, C. L., E. W. Bock, and G. R. Lee. 1990. Religious

heterogamy, religiosity, and marital happiness: the case

of Catholics, Journal of Marriage and the Family 52:

73

/79.

Sherkat, Darren E. 1991. Leaving the faith: testing socio-

logical theories of religious switching using survival

models, Social Science Research 20: 171 

/187.

South, Scott. J and Glenna Spitze. 1986. Determinants of

divorce over the marital life course, American Socio-

logical Review 51: 583

/590.

Stark, R. 1999. Secularization, R.I.P, Sociology of Religion

60: 249 

/273.

Tynes, S. R. 1990. Educational heterogamy and marital

satisfaction between spouses, Social Science Research

19: 153 

/174.

Ultee, Wout. C and Ruud Luijkx. 1998. Jewish 

/Gentile

intermarriage in six European cities 1900 

/1940. Ex-

plaining differences and trends, Netherlands, Journal of

Social Sciences 34: 165

/195.

Uunk, Wilfred 2003. The cultural integration of immi-

grants in the Netherlands: a description and explana-

tion of modern attitudes of Turks, Moroccans,

Surinamese, Antilleans, and the indigenous population,

in L. Hagendoorn, J. Veenman, and W. Vollebergh

(eds.), Integrating Immigrants in the Netherlands:

Cultural versus Socio-economic Integration . Aldershot:

Ashgate.

Van Huis, Mila and Liesbeth Steenhof. 2003. Echtschei-

dingskansen van allochtonen: Speciﬁeke groepen

[Divorce rates of immigrants: speciﬁc groups],

Bevolkingstrends 54

/57.

Van Solinge, Hanna and Marlene De Vries (eds.). 2001. De

Joden in Nederland anno 2000 [The Jews in the Nether-

lands in 2000 ]. Amsterdam: Aksant.

Van Tubergen, Frank. 2003. Religieuze participatie en

geloof van immigranten in Nederland [Religious parti-

cipation and belief of immigrants in the Netherlands],

Mens & Maatschappij 78: 331 

/354.

Wagner, Michael and Bernd Weiss. 2003. Balanz der

Deutschen Scheidungsforschung: Versuch einer Meta-

analyse, Zeitschrift fu¨ r Soziologie 32: 29

/49.

Intermarriage and the risk of divorce 85