divcapsr.html
ABSTRACT
SPECIFIC INVESTMENTS, SEARCH, PREFERENCES AND THE ECONOMICS OF
MARITAL EXIT.
The present paper uses a large national survey, from the UK
to estimate a logit equation for the probability of ever having
been divorced within a population of those who have ever been
married. The survey utilised permits the construction of several
variables, not hitherto deployed in econometric work, which
measure aspects of partner search and relationship capital
formation. The results indicate that these things are important,
along with tastes, religion, risk disposition and general human
capital in determining divorce probability.
SPECIFIC INVESTMENTS, SEARCH, PREFERENCES AND THE ECONOMICS OF
MARITAL EXIT.
I.Introduction.
Divorce is one of the major features of modern life with
important implications for public policy; indeed the Social
Security Administration in the United States routinely attempts
to predict the number of divorces. Research in the USA [Martin &
Bumpass(1989) cited in Lillard & Waite(1990)] suggests that 2/3
of all first marriages will end in divorce and that remarriage
with new partners are no less prone to dissolution. On current
projections, the figure for first marriages in the U.K. is 40%;
the annual divorce rate in England and Wales in 1989 was 13 per
1,OOO marriages, six times the 1960 rate. Even in a less divorce
prone nation such as Denmark [Jensen & Smith(1990)] the divorce
rate increased substantially in the 1980's. The crude divorce
rate has shown a general upward trend in most industrialized
nations since the mid-1960's.
There have been many econometric studies of divorce
with an initial Becker-inspired focus on the role of general
human capital in choice, and a later preoccupation with the
impact of legislation (specifically no-fault laws) on divorce
rates [see e.g. Peters(1986,1992), Smith(1997),Clark(1999)].
Little attention has been paid to the role of partner search and
relationship specific capital formation with the exception of
Chiswick and Lehrer(1990) which uses some measures of marriage
specific human capital in examining re-marriage likelihoods. A
more general discussion, without econometric estimates, by Frey &
Eichenberger(1996), infers that current high rates of divorce are
an index of a faulty matching process in the partner search
market in contrast to the picture painted in various formal
treatments of the matter [Gale & Shapley(1962),Bergstrom &
Bagnoli(1993), Burdett & Coles(1997)]. Apart from these papers,
not much has been said by economists about the effect of the
search process on the stability of marriage although scholars in
other fields have made many relevant contributions [La Gaipa
(1982), Karney & Bradbury(1995), Shackelford & Buss(1997)].
The present paper uses a large national survey, from the UK
to estimate a logit equation for the probability of ever having
been divorced within a population of those who have ever been
married. The survey utilised permits the construction of several
variables, not hitherto deployed in econometric work, which
measure aspects of partner search and relationship capital
formation. The results indicate that these things are important
with there also being notably different results for men and
women.
II. Background.
Economists would seek to explain movements in the divorce
rate in terms of a rational choice to dissolve a partnership
which was initially entered into on the basis of maximizing
expected utility. The specified regression equations are derived
from models of utility maximising individuals operating with
discounted lifetime profiles. The rationale for marriage lies in
gains from trade through specialization, economies of scale in
consumption and sharing public goods [Becker(1991),
Grossbard-Schectmann(1982,1984, 1995)] . The role of the search
market is to match partners optimally [Becker et al. (1977)]. The
main burden of search related explanation falls on the earnings
variables for the partners. Becker (1981,p.231) explains the
pattern of earnings coefficients in divorce regressions as due to
the fact that [ceteris paribus, of course]:
"women with higher earnings gain less from marriage than other
women do because the higher earnings reduce the demand for
children and the advantages of the sexual division of labour in
marriage".
Higher wages for women also make it easier for them to
survive divorce and therefore make it more likely that an
unsatisfactory union will be brought to an end. Education would
tend to display a smilar sign to wage rates as it will proxy
unmeasured elements of permanent income.
Following Becker et al.'s [1977] lead, most authors have
looked for the impact of search in terms of the human
capital/earnings coefficients. That is, the woman who has a
higher expected future wage may be more likely to divorce
because, in search terms, she is better able to support herself
whilst re-searching the partner market. There are potentially two
conflicting elements here: exogenously given higher
earnings/human capital facilitate more efficient search and thus
reduce the likelihood of a mis-match but at the same time they
make exit easier if there was some exogenous shock in terms of
the available supply of partners 'outside' the relationship. This
neglects the effect of capital formation whilst searching
III. Sketch of a model.
In this section we extend the traditional marital formation
model to encapsulate the impacts of pre-marital search and
capital formation. The usual place to start in any economic model
is with the assumption of utility maximisation, under given
tastes, although considerations of pervasive uncertainty, which
is costly to eradicate, may lead to a shift to a satisficing
model. Satisficing models have not been found in the marital
search literature with the exception of Todd(1997).
Thus the basic individual function is:
(1) U = U(X,P)
Utility is derived from private goods (X), such as food,
and semi-public goods (P), such as consumer durables like
televisions, in which there may be scale economies from sharing
in a household context. The scale economy aspect has long been
recognised in the old cliche that 'two can live as cheaply as
one'. None of this satisfactorily motivates the formation of
mixed sex households consisting of a single dyad plus their own
offspring viz. the conventional notion of a family. There is thus
an issue of 'why does the family exist' analogous to the
provocative question posed by Coase (1937) on the theory of the
firm and carried into the modern day by Williamson (1975,1976).
The broad ultimate answer in the family case is, of course, the
same as in the firm context viz. transaction costs. But, costs of
transaction over what exactly?
In the standard model the key motivation for family
formation is supplied by treating offspring as local public goods
[Weiss & Willis(1985)]. An investment in the child by either
parent generates external benefits for the other and this may go
on even after separation or divorce. One incentive to stay
married then must be the wish to capture consumption benefits of
investment in children which might otherwise leak out to the
uncompensated gain of the other parent. The specific legal
institutional forms of marriage can therefore be rationalised as a
response to the transactions costs issues surrounding personal
investments in family set-ups [Allen(1990)]. From an economic
point of view, the absence of polygamy and polyandry in developed
economies [cp. Bergstrom (1996)] is thus a manifestation of
transactionally efficient law making rather than merely the
result of an exogenous moral/social curb. The Chicagoan models of
marriage and divorce have tended to neglect the effect of capital
formation, that goes on during partner searching and thus ignore
the insights contained in Becker and Stigler's wide ranging 1977
paper on tastes. Certain investments may increase the discounted
utility of a marriage. This becomes more important when we add
some key arguments to that of 'progeny as public goods' in the
utility function. These are: recreational sex (S) and intrinsic
needs for comfort, security, companionship or whatever (M) as
conceptualised in the papers of Cameron & Collins(1997, 1999) on
'dating' as search behaviour. These goods can be obtained without
recourse to marriage [1] and yet many people marry without
inmediate production of progeny. Part of this could be a case of
'locking in' for future use the genetic input of a suitable
partner. Yet some partnerships are surely formed because the
intensive and extensive costs of searching for S and M are
thereby reduced.
The augmented utility function is then:
(2) U=f(X,P,S,M,H)
where X and P are as before with P including children along with
tv sets, refrigerators and so forth. S = sexual services defined
broadly to include any sexual arousal or gratification received
from interaction with a partner and M = companionship broadly
defined to include non-sexual exchange. We assume continuous
substitutability and convexity between S and M, at least over
certain ranges of the utility function. The H variable is a
reference group comparison which is a symbolic value attached to
being married because of the status it may confer. Conceptually
there is no reason why this should not be, in some cases a
negative valued index. That is, in concrete terms, an individual
could purposefully get married to experience S,M and economies of
scale because the marginal net gains in those areas offset the
private distaste or referential stigma of marriage which might be
felt in a particular place or time.
Despite increasing rates of divorce, and a considerable
drift up over time in the age at marriage most heterosexuals
still enter into the marital state at some time. Indeed, so
ingrained is the symbolic value of marriage that many homosexuals
campaign for the right to it up to the level of being allowed to
adopt children.. So assuming that most individuals expect to
consume at least one marriage, how do they engage in search and
investments which will enhance its productivity.? Firstly they
may 'practice' on would-be partners. Entering the marital state
with no experience in coducting the relevant negotiations,
allocating the division of household labour and so on may
greatly jeopardize stability. From this perspective, increased
rates of cohabitation [see Ressler & Waters(1995). Smock &
Manning(1997)] might be indicative of a higher value placed on
marriage revealed in a large degree of intensive search over the
present partner prospect combined with capital formation for
potential future liaisons. Given a generally favourable outlook
on marriage, it follows that a relatively high value
placed on H might lead to rapid re-marriage, following
dissolution, as indeed seems to the the case, rather than marital
stability.
In addition to such 'learning by doing' in relationships,
individuals may also be engaged in a process of discovery of
their own preferences over the S and M arguments in their
utility function. It seems unreasonable to regard tastes for
these things as perfectly known, even if they are given at the
outset of partner search as they are 'experience goods' where
tastes may be acquired, cultivated and even discarded after an
exploratory phase. In a situation where one's tastes are evolving
along with a growth in capital, which is general to any
relationship rather than specific to the present one, it may be
inefficient to become locked in to the present relationship
through the contract of marriage. It can be argued that the
'shocks' facet of the original Becker,Michael, Landes(1977)
handles this to some extent but it does not spell out the search
and capital formation that might influence the risk of a 'shock'
shortfall in returns to marriage from those expected.
There are some fairly clear intuitive notions about the
profile of returns to investment and search [broadening search to
include the discovery of one's true preferences in the
definition] in the partnership market although it is hard to
reduce all such activity to a single valued index. There may be
positive, if diminishing, returns to any future established
relationship from the prior number of partners sampled and to the
amounts of time spent with individual partners. Any such sampling
does of course generate idisoyncratic relationship capital, of no
value outside of dealings with the person at hand, along with
generalised capital of relevance to all relationships. The cut
off point where it is optimal to discard the samplee (assuming
that the individual treats the probability of being a discard as
beyond control) will depend on the balance of these elements
along with the risk attitude of the sampler in view of the fact
that the utility stream of future prospects is more difficult to
evaluate than that of present prospects. After some time interval
there will be no growth in the transferrable skills/assets being
accumulated in a trial/training relationship. The nearest that
the economic literature gets to considering this issues is in
looking at the influence of age at marriage as instanced in the
following remarks by Sander & Ferber(1989,p.19) [cp.
Lemennicier(1988,p.66)]:
"Economic theory suggests that those who take more time to
search are likely to be more successful in finding a compatible
mate. Additionally, one might recognize that individuals become
more mature and perhaps make wiser choices for that reason.
Whatever the reason, age at marriage has been found to be related
to the probability of marital disruption" age at marriage
It is not necessarily age at marriage that is the crucial
factor: rather age broadly correlates with experience of
exploratory partner search and capital investment. Further, early
marriage tends to be associated with early child bearing which
can be a very disruptive shock element in a scenario where there
has been a poverty of investments in relationship conduct such as
bargaining over the allocation of time to tasks. Accordingly this
paper eschews the age at marriage variable in favour of variables
for the age at first child birth plus variables to measure the
degree of 'lock in' to early relationships which might cause an
individual to 'miss the boat' in terms of relationship specific
capital formation I am ignoring throughout any possibility of a
"grass is greener on the other side" externality which may also
show up in some of the variables unique to this study.
IV. Data and expectations.
The estimated equations use the National Survey of Sexual
Attitudes and Lifestyles which has a large sample and a wide
range of variables. The NSSAL survey was carried out, by
professional market researchers, in the UK in 1990/1. This was a
rigorously conducted survey of 18,876 adults of ages from 16 to
59. The questions were extensively piloted and a lengthy in
person interview was combined with separate self completion
schedules in an attempt to ensure greater accuracy in response to
the more 'sensitive' questions. Although the primary purpose of
the study was to examine whether there had been discernible
behavioural responses to the threat of HIV/AIDS, it was conducted
in the context of a thorough examination of the overall
lifestyles and socio-economic circumstances of the individuals.
The sexual content was revised carefully in order to ensure the
maximum participation, across all demographic groups, by avoiding
fall out through embarrassment or distaste. Two samples are
available. The 'long' version comprises 4,548 individuals and
contained a detailed investigation of religious beliefs and
sexual attitudes. The 'short' version is used here in order to
get a larger sample of data. The data were provided, on a CD-rom,
by the ESRC Data Archive at the University of Essex.
There have been many econometric studies of the
determinants of divorce [see Cameron(1995) for a detailed
review]. The majority of these are American and many use large
samples of individual level data. The precise nature of the
model tends to vary with the exigencies of the sample used. At
the most sophisticated level hazard modelling is used [Lehrer
(1988),Starkey(1996)] to examine the sequential probability of
dissolution at all stages of a marriage. The hazard rate is
usually restricted to a five year period over which couples
legally married at the start of the period are followed.
Obviously this requires panel data on couples who were once
married whilst the data in the present study is a one off cross
section for only one person from a dyad. This is less than ideal
but we persevere because it gives us access to variables not
found in other work. Some studies have used only male or female
data to fit the estimating equations with the other partner's
characteristics featuring as independent variables in the
equation. This paper estimates separate male and female
equations.
The present sample is for 'ever married' men and women,
aged 16-59 by 1990-1, giving samples of 6060 women and 6934 men.
The names and definitions of the variables are shown in Table 1.
The data used do not contain any explicit measures of wages or
earnings. However education will form a good proxy given its
relationship to lifetime permanent income. If we follow Becker's
original argument then the expectation is of a positive sign for
women and a negative sign for men based on specialization in
household production. However there may be offsetting negative
effects if higher permanent income results in more efficient
search.
Turning now to the capital and search variables. One
measure of potential validity in a UK context is the nature of
the school environment. Proximity to the opposite sex whilst
being educated may generate a joint product with schooling of
easing the facility with which one interacts with them. This
would lead one to expect a negative coefficient on the variable
SINGSKL which is a dummy for having been at a single sex school.
Unfortunately the measurement of 'training' in relationship
capital formation has to be focused heavily on the sexual element
as the measures available require this. The major variable used
is FSEXUN3 which is to capture the case of brevity of the
relationship developed with the first sexual partner. The number
of partners is obviously a candidate for use but this is not
provided in the 'short' questionnaire. A 'gap' variable NSEXUN3,
representing brevity or otherwise of the transition to the second
sexual partner, is included as a control for errors in
interpreting FSEXUN3. That is, there may be some 'taste'
difference between people who have a short first sexual
relationship and move quickly to the next and those who move
slowly to the next.
As this is an eclectic approach, we make some attempt to
measure the strength and variety of tastes for S. Scitovsky(1976)
argued that the need for variety is a neglected factor [2] in the
study of demand in general. The pure strength of tastes for S is
married by the age at first intercourse which means this set of
dummy variables should have positive coefficients for the
youngest age group diminishing to negative ones as the reference
group age is passed. It is widely accepted in the social biology
field [ Wellings et al. (1994)] that a good proxy for variety in
tastes for S is ever having had any homosexual encounters.
Although this may indicate a latent preference for a homosexual
sexual orientation which might precipitate marital dissolution in
any case it is also reflective of an exploratory set of
preferences. Hence the relevant variables are expected to have a
negative sign.
We use various measures of risk taking disposition which
might would probably be regarded as taste measurement from a
mainstream perspective as such a view would tend to make a
blanket assumption of risk neutrality of risk aversion without
any explicit stipulation of any thing such as variations
in the Arrow-Pratt measure. It is expected that AIDSHI, CONDO1ST
and CONTROL will have negative coefficients as they are all
dummies for which a 1 value reflects greater risk aversion
[although CONTROL may also capture other personality dimensions
than risk taking per se].
It is generally found in the typical study of recent
transitions from marital to dissolved states that the presence of
own children is a negative influence on dissolution probability.
As this study is of the cumulative likelihood of all those who
have ever been in the marital state ever having experienced a
dissolution then it is not necessarily the case that the same
result will be anticipated. For example, in the case of a second
marriage, the presence/existence of children from a previous
relationship may contribute positively to the dissolution
probability of the current union.
As mentioned in the last section, the common focus on the
age at marriage conflates search and relationship specific
capital factors. Age at first marriage is likely to correlate
with age at first sex and age of first child born. As early child
birth is expected to be problematic due to insufficient
relationship specific capital formation it is expected to be
positively related to dissolution probability,
In the literature, measures of the frequency of religious
observation indicate a negative correlation with divorce
probability [Lemmenicier(1979) Lehrer(1988), Starkey(1991)] and
there are also denominational differences [Sander &
Ferber(1989)]. For example, Roman Catholics display a lower
propensity to divorce reflecting observance of the teachings of
their faith. Clearly the way in which religion works is through
lowering 'tastes' for divorce via the possession of higher levels
of 'god appreciation capital' [Cameron (2000)]. It is possible
however that mis-specification in past work has over estimated
the direct effect of religion on the probability of dissolution
within a union. As religion will influence search behaviour in
the market for marital partners and attendant marriage related
capital formation it will also have an indirect effect. Religion
may also improve the efficacy of search by eliminating
'fuzziness' in product search and imposing a large amount of
cheap pre-screening of the pool of candidates.
A number of additional dummies for ethnic origin and region
are included to control for any residual variation.
V. RESULTS.
The results of estimating the logit equations for ever
having had a divorce are shown in Tables 2 and 3.
First we look at the religion variables. These confirm the
finding, in the few econometric studies that consider religion,
that it is an important inhibitor of divorce probability.
However, there is little sign that being Roman Catholic, or Jewish
per se has any impact. The inhibition comes through level of
observance of ceremonies rather than denomination. There is some
support for the ASIAN dummy as a divorce inhibitor and this to a
considerable degree reflects religious factors. It should also be
borne in mind that support of south Asian religions requires
higher stipulated attendance and thus there will be an additional
impact through the GOCHURCH term having a larger mean.
There is no impact of the education variables for women
but the usual negative impact is found for men at higher levels
of education [OLEVEL is not significant and DEGREE has a larger
coefficient than ALEVEL]. These results contrast with the
suspicion of a positive education-adultery relationship in these
data [Anon. (1997)] suggesting that adultery and marital
dissolution may be gross substitutes for the more highly educated
man.
Broadly speaking the older people are, the more likely
they are to have ever been divorced suggesting that the
cumulative risk factor is dominating any secular trend towards
greater mis-matching. The only exception to this is found for
men in the 50-59 group who are no more likely to have ever been
divorced than men in the 30-39 group.
The risk and variety seeking variables conform to
expectations with the exception of the AIDSHI and CONDO1ST
variables being insignificant for women.
Single sex school is not significant in any of the
equations. For the early sexual experience variables, the gap
between the first and second relationships dummies is not
significant in any case whilst the dummy for brevity of the
relationship with first sexual partner is only significant in the
male equation. Notably it is negative suggesting that this
strategy of capital formation is beneficial to the longevity of
contractually established unions. All of the variables for age at
first sexual intercourse are significant in all cases. For both
men and women, delay until being over 21 is a deterrent to
divorce risk compared with the reference group of 18-20.
The age at birth of first child is a significant
contributor to both male and female dissolution risk with an
interesting time pattern. The later first birth is delayed the
less is the risk of divorce which conforms to expectations.
However the dissolution risk reduction is gone by age 29 for
women but not for men.
Along with delayed production of progeny, the delay of
sexual intercourse per se also contributes to a reduction in
dissolution probability. All of the dummies are highly
significant showing an uninterrupted declining profile. A gross
decline with age would hardly be surprising as there are many
factors due to culture and religion which might jointly determine
low divorce probability and low a delayed incidence of first
sexual intercourse. Yet, these logit results show the net impact
of delayed virginity loss holding the other covariant factors
constant. It is likely, as argued above, that this indicates
taste variation (in terms of the marginal utility of substitution
of S for M, marginal utility of child 'quality' etc.) within the
core population rather than any kind of deliberate strategy to
optimize search.
VI. CONCLUSION.
There has been a steady flow of econometric papers on
marital dissolution on a variety of data sets but most with
broadly the same foci on legal provision for divorce and the
impact of relative non-specific human capital endowments of the
partners. This paper is unique in the literature in employing a a
large data set that is much more oriented to the formation of
sexual partnerships than those normally used by economists.
Within this data set, we were able to find questions that could
be used to construct measure of early investment and search
activity in the market for relationship partners. These were
shown to be significantly related to the accumulated likelihood
of ever having had a marriage dissolved. Measure of risk taking
disposition and religiosity were also significantly related to
observed dissolution probability.
REFERENCES
Allen,D.W. 1990. An inquiry into the state's role in marriage.
Journal of Economic Behavior and Organization. 13(2) June
171-191
Anon. 1997. Degrees in adultery, Cherwell. Trinity 1997 Issue 5.
Becker,G.S.1991. A Treatise on the Family. Cambridge. Harvard
University Press.
Becker,G.S. & Stigler,G.J.1977. De gustibus non est disputandum.
American Economic Review 68(1) 76-90.
Becker,G.S.,MIchael,R.T. & Landes,E.M.1977. An economic theory of
marital instability. Journal of Politcal Economy. 85(6) 1141-1187
Bergstrom,A. 1996. Economics in a family way. Journal of Economic
Literature. 34(4) 1903-1934
Bergstrom,A. & Bagnoli,M.1993. Courtship as a Waiting Game.
Journal of Political Economy. 101(1) 185-203
Burdett,K. & Coles,M.G.1997. Marriage and class. Quarterly
Journal of Economics. 112(1) 141-168
Cameron,S. 1995. A review of economic research on determinants
of divorce. Brtish Review of Economic Issues. 17(41), 1-22
Cameron,S. & Collins,A.1997. Estimates of a hedonic ageing
equation for partner search. Kyklos 50(3),409-418.
Cameron,S. & Collins,A. 1999. Random utility maximizer seeks
similar: an economic analysis of commitment level in personal
relationships. Journal of Economic Psychology. FORTHCOMING
Cameron,S. 2000. Faith, frequency and the allocation of time: a
micro level study of religious capital and participation. Journal
of Socio-Economics, 29(4) FORTHCOMING
Coase,R.H. 1937. The nature of the firm. Economica. 4. November
386-405
Chiswick, C.U. & Lehrer,E.J. 1990. On marriage-specific human
capital: its role as a determinant of remarriage. Journal of
Population Economics. 3(3) 193-214
Clark,S.1999. Law, property and marital dissolution. Economic
Journal 109(454) March C41-C54
Ferber,M. & Sander,W.1989. Of women,men and divorce: not by
economics alone. Review of Social Economy 47(1),15-26
Frey,B.S. & Eichenberger,R.1996. Marriage paradoxes. Rationality
and Society. 8(2) 187-206
Gale,D. & Shapley,L.1962. College admissions and the stability of
marriage. American Mathematical Monthly. 69. 9-15.
Grossbard-Shechtman,A. 1982. A theory of marriage formality.
Economic Development and Cultural Change. 30. July 813-830
Grossbard-Shechtman,A.S.1984. A theory of the allocation of time
in markets for labour and marriage. Economic Journal. 94.
863-882.
Grossbard-Shechtman,A.S.1995. Marriage market models.pp.92-112 in
Ierulli,K. & Tomassi,M.(eds.) The new economics of human
behavior. Cambridge University Press. Cambridge,England.
Harrison,G.W. & McCabe,K.A.1996. Stability and preference
distortion in resource matching. An experimental study of the
marriage problem. In Isaac,R.M. (ed.). Research in Experimental
Economics. Vol.8. Greenwich. Jai Press.
Keeley,M.C. 1977. The economics of family formation. Economic
Inquiry. 15. 238-248
La Gaipa,J.J.1982. Rituals of disengagement. in Duck,S.(ed)
Personal Relationships 4: Dissolving Personal Relationships.
London. Academic Press.
Lancaster,K.O.1971. Consumer demand: a new approach. New York.
Columbia University Press.
Lane, R.E. 1994. Friendship or commodities? The Road not taken:
friendship, consumerism and happiness. Critical Review 8(4)
521-554
Lehrer,E.1988 Determinants of marital instability: a
Cox-regression model. Applied Economics 20, 195-210
Lemennicier,B.1988. Le Marche du Mariage et de la Famille.
Presses Universitaires de France.
Lillard,L.A. & Waite,L.J.1990. Determinants of divorce.Social
Security Bulletin 53(2), 29-31
Lutz,W.,Wills,A.B. & Nieminen,M. 1991.The demographic dimensions
of divorce: the case of Finland. Demography. 45 437-453
Peters,H.E.1986. Marriage and divorce: informational constraints
and private contracting. American Economic Review.76(3),437-454
-----------1992. Marriage and divorce: reply. American Economic
Review 82(3), 686-693
Ressler,R.W. & Waters,M.S.1995. The economics of cohabitation.
Kyklos. 48(4) 577-592
Sander,W.1986. On the economics of marital instability in the
United Kingdom. Scottish Journal of Political Economy 33(4),
370-381
Scitovsky,T.1976. The Joyless Economy. Oxford: Oxford University
Press.
Shackelford,T.K. & Buss,D.M.1997. Marital satisfaction in
evolutionary psychological perpective. in Sternberg,R.J. &
Hojjat,M.(eds). Satisfaction in close relationships.pp.7-25.
New York. Guildford.
Smith,I. 1997. Explaining the growth of divorce in Great Britain.
Scottish Journal of Political Economy 44(5) 519-544.
Smock,P.J. & Manning,W.D. 1997. Cohabitating partners' economic
circumstances and marriage. Demography. 34(3) 331-341.
Starkey,J.L.1991 The effect of a wife's earnings on marital
dissolution: the role of a husband's interpersonal competence.
Journal of Socio-Economics 20(2),125-154
Starkey,J.L.1996 Race difference in the effect of unemployment on
marital instability: a socioeconomic analysis Journal of
Socio-Economics 25(6), 683-720
Todd,P.M. 1997. Searching for the nest best mate. in Conte,R.,
Hegelsmann,R. & Terna,P. 1997. Simulating Social Phenomena.
Springer-Verlag. Berlin.
Trivedi,M.1999. Variety Seeking Behaviour. in Earl,P.E. & Kemp,S.
1999. The Elgar Companion to Consumer Reesearch in Economic
Psychology.
Weiss,Y. & Willis,R.J.1985. Children as collective goods and
divorce settlements. Journal of Labor Economics. 3(2) 268-292
Wellings,K.,FieLd,J.,Johnson,A.M. & Wadsworth,J. 1994
Sexual Behaviour in Britain. Penguin Books.
Williamson,O.E.1975. Markets and Hierarchies: Analysis and
Antitrust Implications. New York.
Williamson,O.E.1976. The Economics of Internal Organisation: Exit
and Voice in Relation to Markets and Hierarchies. American
Economic Review. 66(2) 369-377
FOOTNOTES
1. Lane(1994) argues that the rising marginal value of time and
other factors such as the high variance of returns to investing
in friendship with people compared to other investments are
creating a rising tendency to loneliness. Under certain
circumstances this may create an added inducement to attempt to
buy into a quasi-permanent technology of loneliness abatement
such as marriage.
2. See Trivedi(1999) for a more modern treatment.
TABLE 1: VARIABLE NAMES AND DEFINITIONS
NAME DEFINITION
EVERSEP =1 if individual has seperated from a legal marriage
GOCHURCH frequency of atendance at church converted to
annual equivalents from interval answers
JEW =1 if Jewish
RCATH =1 if Roman Catholic
BLACK =1 if gives ethnic origin as black
ASIAN =1 if gives ethnic origin as Asian
OTHRACE =1 if non white,black or Asian
AGEL20 =1 if aged under 20
AGEL30 =1 if aged 20 to 29
AGEL50 =1 if aged 40 to 49
AGEL60 =1 if aged 50 to 59
OLEVEL =1 if highest educational qualification is 'O'levels
ALEVEL =1 if highest educational qualification is 'A'levels
DEG =1 if highest educational qualification is degree
OTHQL =1 if highest educational qualification is a other
SINGSKL =1 if attended single sex schools
NATKIDS =1 if has any natural children
FBL18 =1 if first child was born at age<18 FBL21="1" if first child was born at age 18-21 FBLM29="1" if first child was born at age 30+ AIDSHI="1" if thought that multiple partners caused a high risk of catching HIV/AIDS CONDO1ST="1" if used a condom during first sexual experience CONTROL="1" if respondent feels they are in control of their lives HOMONCE="1" if had ever had one homosexual experience HOMMORE="1" if has had more than one homosexual experience FSU16="1" if first intercourse was at age<16 FSU18="1" if first intercourse was at age 16-18 FSU25="1" if first intercourse was at age 20-24 FSO25="1" if first intercourse was at age 25+ [omitted category 18-19] FSEXUN3="1" if first sexual intercourse did not lead to more than a three months relationship with the partner. NSEXUN3="1" if there was a gap of less than three months between first and second sexual partners The following dummies are="1" for the name of the region stated EANGL East Anglia EMIDS East Midlands GLOND Greater London NWEST North West SCOTL Scotland SOUTHE South East SOUTHW South West WALES Wales WMIDS West Midlands YANDH Yorkshire and Humberside [omitted category: NEAST="North" East] TABLE 2: LOGIT ESTIMATES OF FEMALE DIVORCE EQUATION Variables in the Equation Variable B S.E. Wald df Sig R Exp(B) FSEXUN3 .0655 .1144 .3276 1 .5671 .0000 .9366 NSEXUN3 .3131 .2382 1.7270 1 .1888 .0000 1.3676 FBL18 1.1188 .1666 45.1183 1 .0000 .0846 3.0612 FBL21 .7750 .0931 69.3174 1 .0000 .1056 2.1706 FBM29 .1443 .1310 1.2131 1 .2707 .0000 1.1552 AGEL20 1.7273 .6334 7.4373 1 .0064 .0300 .1778 AGEL30 1.1812 .1207 95.7780 1 .0000 .1247 .3069 AGEL50 .5395 .0861 39.3102 1 .0000 .0787 1.7152 AGEL60 .4569 .1105 17.0844 1 .0000 .0500 1.5791 AIDSHI .0456 .0789 .3339 1 .5633 .0000 1.0467 ALEVEL .0208 .1012 .0424 1 .8369 .0000 .9794 ASIAN .6171 .3687 2.8022 1 .0941 .0115 .5395 BLACK .0752 .3142 .0572 1 .8109 .0000 .9276 CONDO1ST .0874 .0730 1.4311 1 .2316 .0000 .9164 DEG .1418 .1596 .7896 1 .3742 .0000 .8678 EANGL .2297 .2295 1.0023 1 .3167 .0000 1.2583 CONTROL .2033 .0704 8.3434 1 .0039 .0324 .8160 EMIDS .3527 .1935 3.3205 1 .0684 .0148 1.4228 FSO25 1.5735 .2897 29.5071 1 .0000 .0675 .2073 FSU16 .7240 .1481 23.9135 1 .0000 .0603 2.0627 FSU18 .3543 .0896 15.6562 1 .0001 .0476 1.4252 FSU25 .3631 .0979 13.7407 1 .0002 .0441 .6956 GOCHURCH .0088 .0025 12.5884 1 .0004 .0419 .9912 GLOND .5349 .1949 7.5291 1 .0061 .0303 1.7072 HOMMORE .8211 .4849 2.8675 1 .0904 .0120 2.2729 HOMONCE 1.0517 .2871 13.4186 1 .0002 .0435 2.8626 JEW .7119 .4973 2.0488 1 .1523 .0028 .4907 NATKIDS .7926 .1901 17.3846 1 .0000 .0505 2.2091 NWEST .3614 .1812 3.9774 1 .0461 .0181 1.4354 OLEVEL .0623 .0851 .5367 1 .4638 .0000 .9396 OTHQL .1856 .2884 .4143 1 .5198 .0000 .8306 OTHRACE .0029 .3620 .0001 1 .9936 .0000 .9971 RCATH .0755 .1227 .3784 1 .5384 .0000 1.0784 SCOTL .0882 .1998 .1947 1 .6591 .0000 .9156 SINGSKL .0587 .0759 .5979 1 .4394 .0000 1.0604 YANDH .3811 .1869 4.1593 1 .0414 .0189 1.4639 WALES .1697 .2064 .6758 1 .4110 .0000 1.1849 SOUTHW .3481 .1923 3.2787 1 .0702 .0146 1.4164 SOUTHE .2801 .1729 2.6227 1 .1053 .0102 1.3232 WMIDS .2406 .1900 1.6037 1 .2054 .0000 1.2720 Constant 2.6443 .2717 94.7415 1 .0000 n="6060" 2 Log Likelihood 5391.637 Goodness of Fit 5975.565 Chi-Square df Significance Model Chi-Square 639.821 40 .0000 Improvement 639.821 40 .0000 TABLE 3: LOGIT ESTIMATES OF MALE DIVORCE EQUATION Dependent Variable: EVERSEP Variables in the Equation Variable B S.E. Wald df Sig R Exp(B) FSEXUN3 .3912 .0635 37.9133 1 .0000 .0626 .6762 NSEXUN3 .0497 .0985 .2549 1 .6136 .0000 1.0510 FBL18 1.6165 .2227 52.6935 1 .0000 .0743 5.0357 FBL21 1.0039 .0905 123.0555 1 .0000 .1149 2.7290 FBM29 .3180 .0841 14.2983 1 .0002 .0366 .7276 AGEL20 1.6107 .8871 3.2966 1 .0694 .0119 .1998 AGEL30 .7126 .0884 64.9506 1 .0000 .0828 .4904 AGEL50 .1923 .0690 7.7723 1 .0053 .0251 1.2120 AGEL60 .0888 .0755 1.3843 1 .2394 .0000 .9150 AIDSHI .2304 .0560 16.9546 1 .0000 .0404 1.2591 ALEVEL .2357 .0719 10.7390 1 .0010 .0309 .7900 ASIAN .8734 .2541 11.8124 1 .0006 .0327 .4175 BLACK .1377 .1909 .5206 1 .4706 .0000 1.1477 CONDO1ST .2550 .0573 19.7873 1 .0000 .0440 .7749 DEG .3269 .1053 9.6390 1 .0019 .0289 .7212 EANGL .0289 .1629 .0316 1 .8590 .0000 1.0294 CONTROL .3086 .0533 33.4626 1 .0000 .0586 .7345 EMIDS .0611 .1390 .1935 1 .6600 .0000 .9407 FSO25 .5890 .1347 19.1282 1 .0000 .0432 .5549 FSU16 .1807 .0902 4.0123 1 .0452 .0148 1.1980 FSU18 .2260 .0709 10.1485 1 .0014 .0298 1.2535 FSU25 .2256 .0767 8.6438 1 .0033 .0269 .7980 GOCHURCH .0073 .0020 13.6763 1 .0002 .0357 .9928 GLOND .3047 .1350 5.0941 1 .0240 .0184 1.3562 HOMMORE .4406 .2924 2.2711 1 .1318 .0054 1.5536 HOMONCE .3079 .1987 2.4025 1 .1211 .0066 1.3606 JEW .4697 .4022 1.3633 1 .2430 .0000 .6252 NATKIDS 1.4238 .2286 38.8026 1 .0000 .0633 4.1529 NWEST .2225 .1278 3.0300 1 .0817 .0106 1.2492 OLEVEL .0231 .0698 .1099 1 .7403 .0000 .9771 OTHQL .3723 .1796 4.2987 1 .0381 .0158 .6891 OTHRACE .1827 .2421 .5692 1 .4506 .0000 1.2004 RCATH .0148 .0949 .0244 1 .8758 .0000 .9853 SCOTL .0062 .1370 .0021 1 .9637 .0000 1.0062 SINGSKL .0499 .0582 .7353 1 .3912 .0000 1.0512 YANDH .0980 .1345 .5306 1 .4663 .0000 1.1029 WALES .0254 .1504 .0285 1 .8658 .0000 .9749 SOUTHW .1833 .1364 1.8057 1 .1790 .0000 1.2012 SOUTHE .0364 .1185 .0943 1 .7588 .0000 .9643 WMIDS .1724 .1322 1.7010 1 .1922 .0000 1.1882 Constant 1.7319 .2665 42.2497 1 .0000 n="6934" 2 Log Likelihood 8493.592 Goodness of Fit 6945.751 Chi-Square df Significance Model Chi-Square 682.753 40 .0000 Improvement 682.753 40 .0000