Changing gender roles, shifting power balance and long-distance migration of couples

Urban Studies, Vol. 40, No. 3, 603–613, 2003

Changing Gender Roles, Shifting Power Balanceand Long-distance Migration of Couples

Jeroen Smits, Clara H. Mulder and Pieter Hooimeijer

[Paper first received, June 2002; in final form, August 2002]

Summary. Long-distance migration of couples requires joint decision-making within the house-hold. The uneven power balance between men and women and traditional gender roles havegiven rise to the concepts of ‘tied stayer’ (usually the male partner) and ‘tied mover’ (usually thefemale). Since these concepts were introduced, women have gained economic power and genderroles have changed. The paper analyses the effects of these changes on the determinants of thelong-distance migration of couples, using data for the Netherlands. For 1977, the results reflectthe ‘classic’ pattern of family migration: the male’s human capital counted more than thefemale’s and a male age advantage led to more migration. Women seemed to use their powermostly to prevent migration. By 1996, the dominance of the male’s human capital had largelydisappeared and the effect of an age advantage had become more symmetrical. The results pointto the emergence of a new pattern, with a more equal power balance within couples.

1. Introduction

Evidence from various countries has shownthat the incidence of long-distance migrationis lower among two-earner couples thanamong one-earner couples (Lichter, 1982;Mincer, 1978; Mulder, 1993; Wagner, 1989).The classic explanation for this phenomenon(Mincer, 1978) is that in one-earner couplesthe wife is a ‘tied mover’; she moves for thesake of the husband’s career. In two-earnercouples, the labour force participation of thewife may inhibit the move, making the hus-band a ‘tied stayer’. However, several devel-opments are affecting that pattern: the rise inearning capacity of women as a result ofincreasing educational levels and labour

force participation; and the overall shift ingender roles and in the balance of powerbetween men and women. It seems that anew pattern is on the rise in which career-oriented women with good labour marketprospects not only use their power to preventfamily moves for the career of their partner,but also use it to initiate such moves for theirown career, thereby turning the husband intoa ‘tied mover’.

This paper studies the consequences of thechanges in gender roles and in the balance ofpower within relationships for long-distancemigration of couples on the basis of data forthe Netherlands in 1977 and in 1995/96. The

Jeroen Smits is in the Department of Economics, Nijmegen School of Management, University of Nijmegen, PO Box 9108, 6500 HKNijmegen, The Netherlands, Fax: 31 24 361 2379. E-mail: [email protected]. Clara H. Mulder is in the Department ofGeography and Planning, University of Amsterdam, Nieuwe Prinsengracht 130, 1018 VZ Amsterdam, The Netherlands. Fax: 31 20525 4051. E-mail: [email protected]. Pieter Hooimeijer is in the Urban Research Centre Utrecht, Utrecht University, PO Box80115, 3508 TC Utrecht, The Netherlands. Fax: 31 30 254 0604. E-mail: [email protected]. Part of Jeroen Smits’ and ClaraMulder’s research was undertaken while they were employed at the Urban Research Centre Utrecht of Utrecht University.

0042-0980 Print/1360-063X On-line/03/030603–11 2003 The Editors of Urban StudiesDOI: 10.1080/0042098032000053941

JEROEN SMITS ET AL.604

Netherlands is a very interesting country inthis respect, because there was a spectacularincrease in the labour force participation ofmarried women over a rather short time-span. In the early 1970s, only 16 per cent ofmarried women were engaged in paid em-ployment. The traditional family type—ofthe breadwinner husband and the stay-at-home wife—was still very dominant. By themid 1990s, however, more than half of themarried women were gainfully employed.Over the same period, women caught up withmen and even overtook them in terms ofeducational level. People increasingly ac-cepted that married women had a right topursue a career of their own (Hooghiemstraand Niphuis-Nell, 1993; Van der Lippe,1997).

One would expect these changes to in-crease the influence of women on the mi-gration decisions of families. However,recent findings suggest that, in the Nether-lands, most long-distance moves are stillprompted by the career of the male partnerand still affect the female partner’s careernegatively (Smits, 1999, 2001). This classi-cal pattern may persist because women havenot yet caught up completely with men inoccupational achievement and earning ca-pacity. However, it is also possible that the‘tied mover’ and ‘tied stayer’ phenomenonremains gendered. That is, women may beunable to push through a move for the sakeof their own careers, even if their earningcapacity is equal to or higher than that oftheir husband. To explore these dynamics,we analyse the migration decision of coupleswith logistic regression analysis, using char-acteristics of both partners and their house-hold as explanatory variables. This will givean impression of the relative importance ofeach partner’s characteristics and how it haschanged between 1977 and 1996.

An important issue to be addressed in theanalysis is the degree to which the migrationdecision is influenced by age and educationaldifferences between the partners. To a certainextent, the dominance of the male partner inlong-distance migration might reflect the factthat in most couples he is older than the

female. This means that, even if the educa-tional level of the partners is the same, themale partner will be somewhat further alongin his career at any point in time and hencewill earn more or have higher occupationalprestige than the female partner (Markhamand Pleck, 1986). In traditional marriages,the male partner also tended to start themarriage with a higher educational level thanthe female, so that, in many such marriages,the female partner suffered from a doubledisadvantage.

This paper seeks to establish the extent towhich male dominance in the long-distancemigration of couples was caused by either ofthese disadvantages. For this purpose, themigration propensity of the more traditionalcouples is compared with the migrationpropensity among the small number of cou-ples in which the wife is older or has a highereducational level than the husband. This willbe done by adding variables indicating theage and educational difference between thepartners to our logistic regression models.

2. Theoretical Background and Hypo-theses

Three more or less overlapping theoreticalperspectives can be distinguished that lead tosomewhat contradictory hypotheses abouthow the characteristics of the partners andtheir household situation affect a couple’spropensity to move: human capital theory,marital power theory and gender role theory.According to human capital theory, long-distance migration is an investment in humancapital intended to generate returns in theform of increased income, employmentprospects and/or occupational status (Sjaas-tad, 1962; Blau and Duncan, 1967; Green-wood, 1975). The human capital perspectiveyields a number of factors that have a posi-tive effect on migration at the individuallevel: a low age (because wage gains de-crease with age as do the remaining numberof years to recoup the costs), a high level ofeducation (because highly educated workerstend to have faster career development andneed job changes to step up the career lad-

CHANGING GENDER ROLES 605

der) and being unemployed (because of thelack of returns and the depreciation of thehuman capital of the unemployed) (Barteland Lichtenberg, 1987; Simpson, 1992;Topel and Ward, 1992; van Ham, Mulderand Hooimeijer, 2001).

Living together with a partner is expectedto increase the costs of migration and henceto reduce the migration propensity. This isespecially true for partners who are em-ployed and have to give up their job for themove. However, even though the overall spa-tial mobility of couples is presumably lowerthan that of singles because the costs ofmoving are higher, the individual factorssuggested by human capital theory probablyalso stimulate the migration of couples.When both partners’ earning capacity andpower balance are equal and their genderroles non-specific, one would expect thesefactors to contribute equally to a couple’spropensity to move, regardless of whetherthey pertain to the male or the female part-ner. So, the first prediction is that couples inwhich one partner is younger, has a highereducational level or is not employed willhave a higher migration propensity thanother couples. Because the effects of thecharacteristics of the partners are cumulative(the propensity to move is expected to behigher if both partners have a high educa-tional level than if one of the partners does),this is termed the ‘additive hypothesis’.

2.1 Differences between the Partners

In addition to the direct effects of humancapital variables on the couple’s migrationpropensity, the economic theory of ‘tiedmovers’ and ‘tied stayers’ in combinationwith marital power theory leads us to expectthat the difference in earning capacity be-tween the partners will affect their spatialflexibility or inertia (Mincer, 1978; Boyle etal., 1999). According to marital power the-ory, the balance of power within conjugalunits lies with the partner who brings morevalued resources into the marriage (that is,the partner who earns more, has a highereducational level or has a higher occu-

pational prestige). Because the other partneris dependent on him/her, this partner mayexert a disproportionately large influence onmajor family decisions, like the migrationdecision (Blood and Wolfe, 1960; Rodman,1972). In conflicts about whether or not tomigrate, this power may give this partner thepossibility of pushing through a move fortheir own career and thus turning the otherpartner into a tied mover.

Although traditionally the male partnerwas the breadwinner and brought in moreincome than the female partner, the predic-tion of marital power theory is in essencegender aspecific. The balance of gains andlosses favours a move for the sake of thepartner with the highest earning capacity,who also has more power to push through amove. When their earning capacity is aboutequal, on the other hand, the gains of a moveare less likely to outweigh the costs andneither partner has the power to push througha move. The second prediction, which iscalled the ‘power-balance hypothesis’, is thatthe migration propensity of couples will below if the partners are equal with regard tosuch human capital factors as age, educa-tional level or labour market situation. Onthe other hand, if only one of the partners isemployed or if one is clearly older (andhence is on average further in his/her career)or has less human capital, we expect thecouple to be more likely to move (and infavour of the partner with more capital) thanif both partners are equal in these respects.The only exception is the situation in whichboth partners are unemployed, because then amove for either partner’s career is mostlikely to favour both of them. In this situ-ation, it would be expected that the couple ismore likely to move.

2.2 Gender Roles

Whereas the additive and power balancehypotheses do not differentiate between menand women, gender role theory would predictthe direct effects of the characteristics of thefemale partner on the migration decision tobe weaker than those of the male partner.


The reason is that women are socialised toplace family first and personal goals secondwhen it comes to critical household matters(see, for example, Markham and Pleck, 1986;Shihadeh, 1991). This means that in a situ-ation of equal human capital—or even whenthe female partner has more human capitalthan the male partner—the characteristics ofthe male will have a greater influence. Fur-thermore, when there is a difference in hu-man capital, an advantage possesed by themale partner will lead to more migration thanan advantage of the female partner. Thisprediction is called the ‘male-dominancehypothesis’.

2.3 Children

The effects of the human capital, power bal-ance and gender factors may work out differ-ently for couples in different circumstances.An important condition in this respect is thepresence of children in the household. Cou-ples without children can migrate more eas-ily because they face fewer restrictions. Ifthere are children, more persons are involvedin the migration decision. For older children,it may be a problem to change to anotherschool. Moreover, they probably do not wantto give up their social network and start allover again in another place. If there areyoung children, the number of restrictions onthe female partner is high. Even in the mostmodern societies, females still bear most re-sponsibility for the children (Gregson andLow, 1993; van der Lippe, 1997). In additionto the characteristics of the partners them-selves, therefore, the presence and age ofchildren will also be taken into account in theanalyses.

3. Data and Measurement

3.1 Data

The data used in this paper were derivedfrom the Dutch Labour Force Surveys (LFS),which were conducted in 1977, 1995, and1996 by Statistics Netherlands. The LFS arebased on large samples representative of the

Dutch population aged 15 years and over notliving in institutions. The data are gatheredby oral interviews, conducted by speciallytrained interviewers of Statistics Netherlands.The general aim of the survey is to acquiredata on the size, composition, and inflow andoutflow of the working population, and onthe short- and medium-term developments ofthe labour market. As well as information onthe situation at the time of the survey, forsome characteristics (geographical andlabour market) information on the situationone year before the survey is also gathered.This makes it possible to use these data forthe study of family migration.

The data-sets of the 1995 and 1996 LFSwere pooled because the number of respon-dents in the recent 1995 and 1996 LFS islower than in the 1977 LFS and because thenumber of long-distance moves is rathersmall in the Netherlands (Smits, 1999). Be-cause the 1977 and 1995/96 files had to beanalysed at different locations, it was necess-ary to perform separate analyses for bothtime-periods. For reasons of simplicity, wewill speak of (data for) 1996 when referringto the combined 1995/96 data-set.

The analyses are restricted to married andcohabiting persons, aged 22–59 years. Thelower age limit is set at 22 because the focusof this paper is on career migration andtherefore it is desirable to rule out migrationfor reasons of education as much as possible.The upper limit is age 60 because in 1996many persons of that age had already left thelabour market due to retirement or disability.The analyses are restricted to those respon-dents who formed a couple before the poten-tial move. Those who made a move to startliving with a partner are left out of the analy-sis to avoid confounding effects of migrationfor reasons of household formation.

3.2 Method

To determine the effects of the characteris-tics of the partners on long-distance mi-gration, logistic regression analysis was used.The dependent variable in these analyses is a


dummy variable which takes the value of 1for couples who migrated over a long dis-tance in the year before the survey and avalue of 0 for couples who did not move oronly moved over a short distance. The dis-tinction between long-distance migration andother forms of migration is based on thedivision of the Netherlands into 12provinces. All moves in which at least oneprovince boundary was crossed are con-sidered to be long-distance moves. This oper-ationalisation on the basis of geographicalunits instead of in kilometres or miles is notideal, because it will inevitably include someshort-distance moves that just straddle theboundary between two adjacent provinces.However, it was necessary to use it becausein the data no other comparable informationabout the distance of migration was avail-able. This operationalisation issue will bediscussed further in the final section of thepaper.

3.3 Independent Variables

The characteristics of the partners that areused to explain the couple’s migration be-haviour include their age and educationallevel and several characteristics of their worksituation. To indicate the household situation,information on the presence and age of chil-dren is used.

Because of the high correlation betweenthe ages of married persons, the couple’s‘age’ is indicated by the age of the male.This age variable is measured in years. Theeducational levels of the partners are dividedinto four categories: primary education;lower-level secondary education (Dutchnames vbo, mavo; duration until about age16); upper-level secondary education (mbo,havo, vwo; duration until about age 17–18);and, tertiary education (hbo, university; dur-ation until about age 21–22). These variablesare entered in the form of three dummies,with the category ‘primary education’ as thereference category. The presence of childrenin the household is classified according to thefollowing three categories: no children; chil-

dren including under age 6; and, only chil-dren aged 6 and older.

The labour force participation of the part-ners before the move was measured with afour-category variable indicating whether,one year before the survey: only the malewas employed; only the female was em-ployed; both partners were employed; or,both partners were unemployed. The indus-trial sectors of the partners one year beforethe survey were indicated by four categories:manufacturing and construction; agriculture;commercial services; and, non-profit sector.Dummies were used for these variables with‘manufacturing and construction’ as the ref-erence category. The distance over which thepartners had to commute to their work oneyear before the survey was measured withdummy variables indicating whether (1) ornot (0) the place of work was in the sameprovince as the place of residence. We alsoincluded dummy variables to indicate em-ployed individuals with missing values forthe industrial sector and commuting distancevariables.

The occupational prestige of the partnerswas measured with the U&S occupationalprestige scale for the Netherlands (Sixma andUltee, 1983). This variable was only avail-able for the moment of the survey and henceonly applies to males and females who wereemployed at that time. We substituted miss-ing values on occupational prestige with theaverage prestige of the persons of the samesex and included dummy variables to indi-cate the couples for whom the prestige scoreswere substituted. For individuals without ajob one year before the survey, the averagesof the values for individuals of the same sexwith a job were substituted in the dummiesfor occupational sector, commuting distanceand occupational prestige. For reasons ofclarity, the coefficients of the missing valuedummies are not presented in the tables withthe results.

To estimate the effects of age differencesbetween the spouses, we use two differentvariables: the age difference in years (themale’s age minus the female’s age) and theabsolute value of the age difference in years.


The variable for the absolute age differenceindicates the difference regardless of whichpartner is older; it is denoted as the sym-metrical age difference. The variable for theage difference as such (which will be calledthe asymmetrical age difference) indicateshow many years older the male is, pointingto a greater importance of the male. To com-pute the effects of these difference variables,separate models are estimated. We test thecontribution to the baseline model of eachvariable separately and choose the variablewith the greatest contribution. For the effectof an educational difference between thepartners, two variables are also used: thedifference in educational level between thepartners (the male’s level minus the female’slevel; called the asymmetrical educationaldifference) and the absolute value of theeducational difference (called the symmetri-cal educational difference). In the same wayas with the age difference, the contribution tothe baseline model of each of the educationaldifference variables is tested and the onecontributing most is choosen.1

4. Results

4.1 The Number of Long-distance Moves

Table 1 shows that the incidence of long-distance migration among married couples inthe Netherlands is rather low. In 1977, only1.2 per cent of the couples in the data-set hadmoved to a different province during the yearbefore the survey. In 1996, the percentage of

interprovincial moves is even somewhatlower.

In 1977, couples in which only the malepartner was employed tended to move theleast, while (the rather small number of)couples in which only the female partner wasemployed tended to move the most. Coupleswith both partners employed and coupleswith both partners unemployed had anabove-average propensity to move. At firstsight, it might seem surprising that dual-earner couples are more likely to move thanone-earner couples. It is hypothesised thatthis unexpected difference has to do with thefact that dual-earner couples are over-represented among the young, the highlyeducated and the childless couples; it istherefore expected that it will disappear inthe multivariate analysis.

In 1996, still about 1 per cent of coupleshaving only the male partner employed madean interprovincial move. However, this timethe proportion of long-distance migrantsamong the other couples is hardly higher.The couples with only the female partneremployed show the biggest change. Whereasin 1977 almost 5 per cent of these couplesmoved interprovincially, by 1996 this haddecreased to only 1 per cent. The couples inwhich both partners are employed alsomoved less in 1996 than in 1977. The figuresindicate that, in the period under study, theinfluence of the female partner on the mi-gration decision—and especially her like-lihood of preventing migration—hadincreased considerably.

Table 1. Interprovincial migration in the year before the survey among couples aged22–59 years in the Netherlands

1977 1996

Percentage Total Percentage Totalmigrated N migrated N

Only male employed 1.0 44 934 0.9 29 976Only female employed 4.8 944 1.0 3 643Both employed 1.4 12 795 1.1 27 149Both unemployed 1.4 4 102 1.3 7 592

Total 1.2 62 775 1.0 68 360


4.2 Logistic Regression Results

Table 2 shows the results of the logisticregression analyses. The findings for 1977are very well in line with what could beexpected in a rather traditional society—asthe Netherlands was at that time. The male’shuman capital has the expected effects onmigration: a higher educational level andgreater occupational prestige lead tosignificantly more migration; a higher age tosignificantly less. The male’s industrial sec-tor is also important, with higher migrationlevels in commercial services and in the non-profit sector than in manufacturing.

The educational level of the female partnerhas a positive effect on migration, although itis not as strong as the effect of the male’seducational level. Couples in which the fe-male partner has upper secondary or tertiaryeducation migrate significantly more thancouples in which the female has only primaryeducation. However, the female’s occu-pational prestige and industrial sector haveno effect on migration.

With regard to the couple’s employmentsituation, asymmetry is seen in the effects. Inboth situations where the male is not em-ployed (with and without employment of thefemale), the migration propensity issignificantly increased. Thus, the unemploy-ment of the male partner stimulates mi-gration, regardless of the labour marketsituation of the female partner. When bothpartners are employed, the couple’s mi-gration propensity is significantly reduced.So, when the need to migrate is not verystrong (because the male has a job), thefemale partner’s employment does reduce thecouple’s migration propensity.

This finding of a reduced migrationpropensity among dual-earner couples in themultivariate analysis seems to confirm theexpectation that the increased migrationpropensity among these couples found in thebivariate analysis was caused by the overrep-resentation of the young, highly educated andchildless couples among them.

Employment of either partner outside theprovince has a strong positive effect on mi-

gration—although again more so when thispertains to the male than to the female part-ner. This indicates that reducing commutingdistance was an important reason for mi-gration in the 1970s. In contrast, the presenceof children, and especially of school-agedchildren, tended to reduce the migrationpropensity of couples.

The finding for 1977 that the male’s hu-man capital and employment characteristicshave consistently stronger effects thanthose of the female fits in with the maledominance hypothesis rather than with theadditive hypothesis. However, anotherfinding for 1977—that, when both partnersare employed, the migration propensityof the couple is lower than in any othersituation—suggests that male dominance isnot prevalent in all circumstances, which ismore in line with the power balance hypoth-esis.

To test the power balance hypothesis fur-ther, the models were re-estimated with vari-ables for the effects of the age andeducational differences between the partnersadded. This was done separately for the sym-metrical and asymmetrical age and educa-tional difference effects, so that for each yearfour extra models were estimated. The ageand educational difference parameters ofthese models are presented in the lower partof Table 2. The parameters of the othervariables are not presented, because they arelargely the same as in the models without agedifference.

In 1977, the asymmetrical versions of theeducational and age difference variables con-tributed more to the model than the sym-metrical versions of these variables. Withregard to the educational difference effect,neither variable (asymmetrical version orsymmetrical version) is significant. Thus, itseems that educational differences betweenthe partners do not contribute very much tothe explanation of the couple’s migrationpropensity over and above what is alreadyexplained by the main educational effects.For age, the difference effect is quite sub-stantial in 1977. The coefficient of theasymmetrical age difference variable is


Table 2. Logistic regression coefficients (B) for the effects of selected characteristics of the male andfemale partner on interprovincial migration of couples in the Netherlands, and differences between the

coefficients

1977 1995/961995/96–1977

B SE(B) B SE(B) Differenceb

Constant � 5.141*** 0.404 � 3.588*** 0.398 1.553***

Age male � 0.049*** 0.006 � 0.073*** 0.006 � 0.024***

Education malePrimary Reference category Reference categoryLower secondary 0.478*** 0.139 � 0.225 0.195 � 0.703***Upper secondary 0.801*** 0.145 0.088 0.178 � 0.713***Tertiary 0.905*** 0.166 0.549*** 0.193 � 0.356

Education femalePrimary Reference category Reference categoryLower secondary 0.085 0.109 0.118 0.194 0.033Upper secondary 0.468*** 0.127 0.526*** 0.185 0.058Tertiary 0.474*** 0.166 0.724*** 0.202 0.250

Employment situation coupleOnly male employed Reference category Reference categoryOnly female employed 1.401*** 0.199 � 0.036 0.193 � 1.437***Both employed � 0.578*** 0.134 � 0.533*** 0.103 0.045Both unemployed 1.381*** 0.155 0.882*** 0.132 � 0.499**

Occupational prestige malea 0.014*** 0.003 0.004 0.003 � 0.010**

Occupational prestige femalea 0.008 0.006 0.000 0.005 � 0.008

Industrial sector malea

Manufacturing/construction Reference category Reference categoryAgriculture 0.194 0.287 0.245 0.306 0.051Commercial services 0.570*** 0.106 0.341*** 0.117 � 0.229Non-profit sector 0.532*** 0.118 0.327** 0.133 � 0.205

Industrial sector femalea

Manufacturing/construction Reference category Reference categoryAgriculture � 0.384 1.043 0.617 0.479 1.001Commercial services 0.107 0.239 0.161 0.215 0.054Non-profit sector � 0.218 0.233 0.124 0.214 0.342

Workplace male outside provincea 2.259*** 0.093 1.968*** 0.101 � 0.291**

Workplace female outside provincea 1.175*** 0.230 1.052*** 0.131 � 0.123

ChildrenNone Reference category Reference categoryYoungest � 6 � 0.211** 0.107 � 0.350*** 0.093 � 0.139Youngest � � 6 � 0.670*** 0.126 � 0.541*** 0.128 0.129

Difference effects (separate models)Asymmetrical age difference 0.032*** 0.012 0.025** 0.012 � 0.007Symmetrical age difference 0.027 0.015 0.033** 0.014 0.006Asymmetrical educational difference 0.166 0.140 0.081 0.115 � 0.085Symmetrical educational difference 0.064 0.057 0.071 0.058 0.007

N 62 775 68 360N migrated 723 672� 2 Log likelihood 6 443 6 312Model chi-squared/DF 1 449/28 1 238/28

aThe dummy variables for cases with missing values on these variables are not presented.bSignificance of the differences between the years is tested with t-tests for independent samples.***indicates significant at the 1 per cent level; **indicates significant at the 5 per cent level.


significantly positive, whereas the coefficientof the symmetrical age difference variable isnot. This means that there is an asymmetricaleffect to the advantage of the male partner:the more years the male is older than thefemale, the higher the propensity of the cou-ple to move. This result is more in line withthe male dominance hypothesis than with thepower balance hypothesis.

The picture of male dominance that arisesfrom the results for 1977 is hardly discern-ible in 1996. For education, occupationalprestige and employment situation, nostronger effect was found for males than forfemales. Only for working outside the prov-ince of residence and industrial sector does astronger effect for the males remain.

With regard to education and employmentsituation, the changes are rather striking. Theeffect of the male’s education on the couple’smigration propensity has decreasedsignificantly over time (as is clear from thedifference parameters in the last column ofTable 2) and is no stronger than the effect ofthe female’s education. Furthermore, thelarge difference in migration propensity be-tween couples in which only the female part-ner was employed and couples with only themale partner employed has completely disap-peared by 1996.

At the same time, the negative effect ofdual-earnership on migration has notchanged at all in the period under study. Asin 1977, in 1996 dual-earner couples wereless likely to migrate than couples with anyother pattern of labour market participation.After controlling for the other explanatoryvariables, dual-earner couples are onlyexp( � 0.533) � 0.59 times as likely to mi-grate as couples with only a male earner.

The parameters for the difference effectsalso suggest that the dominance of the malepartner has decreased over time. The educa-tional difference variables are again notsignificant in 1996. However, both the asym-metrical and the symmetrical age differenceeffects are significantly positive. The positiveeffect of the asymmetrical age differencevariable would indicate that a larger ageadvantage of either partner increases the mi-

gration propensity. However, this conclusionis not completely unequivocal, for thecoefficient of the symmetrical variable issignificant too. It is therefore concluded thatthere is indeed a trend towards more equal-ity, but that by 1996 the situation of maledominance has not yet completely disap-peared.

5. Conclusions

The migration propensity of couples in theNetherlands has been analysed from a per-spective of gender differences in humancapital, marital power and gender roles.Compared with previous research, this papercontributes to the literature by making acomparison between 1977, when genderroles were quite traditional and the share ofmarried women in the workforce was verylow, and 1996, when the number of two-earner couples was much higher and theNetherlands had become a much more mod-ern society. From human capital theory,power balance theory and gender role theory,three alternative hypotheses were derived:the additive hypothesis (stating that theinfluence of the partner’s labour market re-sources is cumulative); the power balancehypothesis (stating that it is the difference inresources that counts, regardless of whetherthey belong to the male or the female); and,the male dominance hypothesis (stating thatthe male’s resources count more than thefemale’s).

For 1977, the results are mostly in linewith the male dominance hypothesis. Thehuman capital factors of the male partnerwere more important at that time than thehuman capital factors of the female partner.The female’s employment situation onlyplayed a role when the need for migrationwas not very high because the male wasemployed. Only in that case did the female’semployment reduce the couple’s migrationpropensity. The effect of an age difference in1977 was also in line with the male domi-nance hypothesis: the more years the malewas older than the female, the higher theprobability of migration. These findings sug-


gest that married women who were olderthan their husband used their power to pre-vent migration, instead of using it to migratefor their own careers.

By 1996, the picture had changed consid-erably. For education, occupational prestigeand employment situation the effect was notstronger for the male than for the female. Themost remarkable difference, however, wasthat couples in which only the female wasemployed did not migrate any more fre-quently than couples in which only the malewas employed. The results for 1996 seem tobe mostly in line with the prediction of thepower balance hypothesis that an unequalbalance in earning capacity will contribute tothe couple’s migration propensity, regardlessof which partner has this higher capacity.Support for this hypothesis is also given bythe finding that the effect of the age differ-ence has become more symmetrical overtime. In other words, any age difference—regardless of which partner was older—hadmore predictive power than an age advantageascribed to one of the partners. However,with respect to the effect of an age differ-ence, the dominance of the male partner,which was found for 1977, has not yet disap-peared completely. The asymmetrical agedifference variable (age of male minus age offemale) also had a significantly positive ef-fect in 1996 (in a separate analysis). Thissomewhat contradictory result suggests that,when the male is older, the positive effect ofan age difference on migration is strongerthan when the female is older.

The analyses show that in both periodsreducing the commuting distance of the part-ners was an important reason for long-distance migration in the Netherlands. Whenone of the partners was employed outside theprovince of residence, the couple’s propen-sity to migrate was considerably greater. Thiseffect was strongest if the male had a longcommuting distance, but it was also quitesubstantial if the female had a long journeyto work.

The results of the multivariate analysesconfirm earlier findings that, ceteris paribus,dual-earner couples and families are less

likely to migrate than their one-earner coun-terparts. There are no signs that the effect ofdual earnership is changing—the estimatedeffect is about the same for 1996 as for 1977.Combined with the increase in dual earner-ship that has occurred in the Netherlands,this result indicates that, over time, more andmore individuals have become restricted intheir migration possibilities because of thepresence of a working partner. Our findingsalso suggest that the ‘tied mover’ phenom-enon has become less gendered and thatmales with less human capital than theirpartners are becoming ‘tied movers’. How-ever, the likelihood of becoming a ‘tiedstayer’ seems to have increased for bothmales and females, because of the greaterequality in their relationship.

One last point should be made in interpret-ing the findings of this study. As was statedin the method section, it is possible that theoperationalisation of long-distance migrationon the basis of geographical units instead ofdistances has introduced bias in the results.Some of the interprovincial moves may infact have been short-distance moves forhousing reasons of people living close to theprovincial border. No information exists con-cerning the extent of this bias. However, it isknown that the effects of education andlabour market characteristics on short-distance migration in the Netherlands areweaker than their effects on long-distancemigration (Mulder, 1993). A possible bias,therefore, is most likely to have led to anunderestimation of the true effects of thesecharacteristics.

Note

1. We would have preferred to use more (andmore sophisticated) measures of power bal-ance—for example, the number of hoursworked by each partner or the difference inoccupational prestige. However, this was notfeasible. Information about the number ofhours worked was not available for the yearbefore the survey. In the 1977 data, the shareof women with known occupational prestigewas too small to allow inclusion of both theprestige itself and the difference between themale’s and the female’s prestige.


References

BARTEL, A. P. and LICHTENBERG, F. R. (1987) Thecomparative advantage of educated workers inimplementing new technologies, Review ofEconomic Statistics, 69, pp. 1–11.

BLAU, P. M. and DUNCAN, O. D. (1967) TheAmerican Occupational Structure. New York:Wiley.

BLOOD, P. M. and WOLFE, D. M. (1960) Hus-bands and Wives. The Dynamics of MarriedLiving. New York: Free Press.

BOYLE, P., COOKE, T., HALFACREE, K. and SMITH,D. (1999) Gender inequality in employmentstatus following family migration in GB andthe US: the effect of relative occupationalstatus, International Journal of Sociology andSocial Policy, 19, pp. 109–142.

GREENWOOD, M. J. (1975) Research on internalmigration in the United States: a survey.Journal of Economic Literature, 13, pp. 397–433.

GREGSON, N., and LOW, M. (1993) Renegotiatingthe domestic division of labour?, SociologicalReview, 41, pp. 475–504.

HAM, M. VAN, MULDER, C. H. and HOOIMEIJER, P.(2001) Spatial flexibility in job mobility:macro-level opportunities and micro-level re-strictions, Environment and Planning A, 33,pp. 921–940.

HOOGHIEMSTRA, B. and NIPHUIS-NELL, M. (1993)Sociale atlas van de vrouw, Deel 2. Rijswijk:SCP; Den Haag: VUGA.

LICHTER, D. T. (1982) The migration of dualworker families: Does the wife’s job matter?Social Science Quarterly, 63, pp. 48–57.

LIPPE, T. VAN DER (1997) Verdeling van on-betaalde arbeid, in: M. NIPHUIS-NELL (Ed.)Sociale Atlas van de Vrouw. Deel 4Veranderingen in de primaire leefsfeer,pp. 117–157. Rijswijk: Sociaal en CultureelPlanbureau.

MARKHAM, W. T. and PLECK, J. H. (1986) Sexand willingness to move for occupationaladvancement: some national sample results,Social Science Quarterly, 27, pp. 121–143.

MINCER, J. (1978) Family migration decisions,Journal of Political Economy, 86, pp. 749–773.

MULDER, C. H. (1993) Migration Dynamics: ALife Course Approach. Amsterdam: Thesis.

RODMAN, H. (1972) Marital power and the theoryof resources in cultural context, Journal ofComparative Family Studies, 3, pp. 50–57.

SHIHADEH, E. S. (1991) The prevalence of hus-band-centered migration: employment conse-quences for married women, Journal ofMarriage and the Family, 53, pp. 432–444.

SIMPSON, W. (1992) Urban Structure and theLabour Market: Worker Mobility, Commutingand Underemployment in Cities. Oxford:Clarendon Press.

SIXMA, H. and ULTEE, W. (1983). Een beroep-sprestigeschaal voor Nederland in de jarentachtig, Mens en Maatschappij, 58, pp. 109–131.

SJAASTAD, L. A. (1962) The costs and returns ofhuman migration, Journal of Political Econ-omy, 70, pp. 80–93.

SMITS, J. (1999) Family migration and the labour-force participation of married women in theNetherlands, 1977–1996, International Journalof Population Geography, 5, pp. 133–150.

SMITS, J. (2001) Career migration, self-selection,and the earnings of married men and women inthe Netherlands, 1981–1993, Urban Studies,38. pp. 541–562.

TOPEL, R. H. and WARD, M. P. (1992) Job mo-bility and the careers of young men, QuarterlyJournal of Economics, 107, pp. 439–479.

WAGNER, M. (1989) Raumliche Mobilitat imLebensverlauf. Eine Empirische Untersuchungsozialer Bedingungen der Migration. Stuttgart:Enke.

Changing gender roles, shifting power balance and long-distance migration of couples

Documents