The Internet and the gender division ofhousehold labour

TIM SCHWANEN*, MEI-PO KWAN†‡ AND FANG REN§*Transport Studies Unit, School of Geography and the Environment, University of Oxford, Oxford

E-mail: tim.schwanen@ouce.ox.ac.uk†Department of Geography and Geographic Information Science, University of Illinois at

Urbana-Champaign, Illinois, USA‡Department of Human Geography and Spatial Planning, Utrecht University, Utrecht,

The Netherlands§MS GIS Program, University of Redlands, Redlands, California, USA

This paper was accepted for publication in November 2012

Geographers and others have long since examined the distribution of paid work, childcare andhousework within households and have more recently begun to explore the various effects ofInternet use on everyday life and digital inequalities along lines of gender and other social markers.These lines of inquiry have so far remained largely separate and this paper brings them together byanalysing the interrelations between the Internet and the gender division of household labour.Multi-group structural equation modelling is applied to dedicated survey data collected amongheterosexual couples in Columbus (Ohio, USA). The results demonstrate that Internet use isgendered in many ways. Variations in Internet use are explained by a broader range of factors forwomen than for men, and an unequal division of domestic responsibilities within the householdconstrains women’s Internet use but not men’s. Overall, however, the relations between Internet useand the gender division of household labour are modest. The latter is related in different ways forwomen and men to the residential location, the household situation, the employment situation andgender-specific interactions among paid work, childcare and housework.

KEY WORDS: Columbus (Ohio), Internet, gender, activity patterns, structural equation modelling

Introduction

This paper seeks to bring two hitherto uncon-nected strands of geography literature together –the longstanding interest in the gender division

of household labour in feminist geography and beyond(Hanson and Hanson 1980; Tivers 1985; Kwan 1999;Schwanen 2007), and the more recent literature onhow the Internet mediates everyday life (Kwan 2002;Valentine and Holloway 2002; Madge and O’Connor2006; Gilbert and Masucci 2011). This focus reflectstwo observations. First is that in the past the interactionof new technologies and women’s daily lives did notnecessarily lead to the beneficial changes often toutedby the promoters of these technologies. For instance,such domestic technologies as the washing machine,freezer and microwave oven did not reduce the timewomen spent on housework or significantly change thegender division of labour within the household (Cowan1983; Kwan 2003; Bittman et al. 2004). This is prob-lematic given the tenacity of inequalities in the genderdivision of labour and women’s more extensive

space–time constraints at a time when women areexpected to be in paid employment and primary care-givers to children and other dependants (Kwan 1999;McDowell et al. 2005).

Second, although it is well recognised that theeffects of Internet use on activity participation differaccording to gender, race/ethnicity and so forth(Crang et al. 2006; Schwanen and Kwan 2008; Gilbertand Masucci 2011), few studies to date have explicitlyand systematically compared different social groupsregarding how the Internet mediates everyday life.Studies using qualitative methods frequently focus ona particular (disadvantaged) social group but tend notto contrast the experiences of its members with thoseof other categories. Authors using quantitativemethods often concentrate on general relationshipswithin a sample of respondents (but see Ren andKwan 2009a). Like race/ethnicity and class, gender isoften taken into consideration but as an exogenousvariable; how the relations of such factors as Internetskills, experience and access with everyday activitiesin the online and offline realms differ by gender (or

bs_bs_banner

The Geographical Journal, Vol. 180, No. 1, March 2014, pp. 52–64, doi: 10.1111/geoj.12014

The information, practices and views in this article are those of the author(s) and do notnecessarily reflect the opinion of the Royal Geographical Society (with IBG).© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society(with the Institute of British Geographers)

race/ethnicity and class) is often not considered. As aresult, our knowledge of how the Internet mediateseveryday activities differently for various social groupsis still rather limited. Such knowledge nonetheless hasconsiderable policy relevance and can inform policiesto reduce digital divides (Gilbert and Masucci 2011),improve employed adults’ work–life balance (Huberset al. 2011), and reduce carbon emissions from pas-senger transport (Banister 2011).

This paper asks whether and how the interrelationsamong internet use, housework, care-giving to chil-dren and paid labour differ between male and femalepartners in couples. Answering this question demandsthat consideration is given to the nature of domesticlabour, each partner’s employment situation, thehousehold situation, and characteristics of the house-hold’s residential location. We apply multi-groupstructural equation modelling to data collected amongwhite, middle-class, heterosexual couples in Colum-bus (Ohio, USA). In this article we use ‘domesticresponsibilities’ and ‘domestic work’ as catch-allphrases for different forms of housework and care-giving to children, and ‘gender division of householdlabour’ as shorthand for the division of housework,childcare and paid labour on the understanding thatall are essential to the household’s reproduction andwellbeing.

Study background

Gender division of household labour

Across the Western world men’s role in domesticlabour has increased over time, but women continueto spend more time on domestic and employmentresponsibilities combined (Fisher et al. 2007) andexperience more stringent space–time constraints(Kwan 2000; Schwanen and Kwan 2008). There none-theless exist notable differences in the gender divisionof household labour across households at a singlepoint in time. Sociologists, economists and geogra-phers have proposed multiple theoretically informedexplanations for those differences, most of which con-sider the gender division of labour the outcome of arational choice process (Shelton and John 1996;Bianchi et al. 2000; Kroska 2004; Ettema and Van derLippe 2009). Suffice it to say that explanations revolvearound five key factors. One pertains to men’s andwomen’s employment situation – their income andtype of occupation – and another to the space–timeconfiguration of paid work as expressed throughweekly employment hours, employment scheduleand commute length (South and Spitze 1994; Bianchiet al. 2000; Presser 2003; Schwanen 2007). FollowingBecker (1965) and new home economics, much lit-erature in economics and sociology assumes that acentral mechanism in the production of the genderdivision of household labour is that the partner withthe most economic capital (income and occupational

prestige) and/or least time available (longest employ-ment hours or commute time) negotiates his/her wayout of domestic responsibilities and especially themore repetitive and unattractive tasks.

A third factor is the household situation as reflectedin the presence, number and age of children (Bianchiet al. 2000; Presser 2003); it has been suggested that inhouseholds with one or more young children the divi-sion of domestic labour is particularly unequal (DeMeester andVan Ham 2009). Fourth, differences in thedegree to which men and women hold egalitarianattitudes toward gender roles have been shown toexplain differences in the gender division of domesticresponsibilities across households (Huber and Spitze1983; Ettema and Van der Lippe 2009). Finally, geog-raphers have highlighted the importance of space, asthe gender division of household labour is more bal-anced in urban, higher-density locations, at least innorth-western Europe (Ettema et al. 2007; De Meesterand Van Ham 2009). Whilst this difference may followfrom more egalitarian attitudes about gender roles(Karsten 2003), it also reflects variations in accessibilityand space–time constraints among urban households(Ettema et al. 2007). In locations offering good acces-sibility to jobs, childcare providers, shops and otherfacilities, it is easier for women to combine longeremployment hours with domestic demands; house-holds elsewhere may cope with poor accessibilitythrough role specialisation with one partner – usuallythe woman – conducting more domestic work and theother specialising in paid work.

Whilst useful, the above explanations can be criti-cised. It is not evident whether attitudes towardsgender roles drive or follow from the actual division ofhousehold labour, and discrepancies between atti-tudes and actual practices often exist (Vincent et al.2004; Schwanen 2007). Additionally, the distributionof domestic responsibilities is sometimes considered tobe a function of employment-related factors. However,it is critical to consider the two-sided nature of employ-ment and domestic relationships. Housework andchildcare also impose significant constraints on indi-viduals’ – especially women’s – paid work, commutingand activity space (Tivers 1985; Kwan 1999; Schwanen2007). Moreover, previous research has not alwaysconsidered differences between domestic responsibili-ties. This is nonetheless important as different respon-sibilities vary in the rigidity of space–time constraintsthey impose (Schwanen et al. 2008) and becausemen’s participation varies considerably between typesof housework and childcare (Schwanen et al. 2008;South and Spitze 1994; Kroska 2004): such activities asrepair, gardening and other homework outside thedwelling are often more male-oriented or masculine.

Finally, the literature discussed so far tends to side-line the non-rational aspects of the gender division ofhousehold labour. For cultural geographers and soci-ologists housework, childcare and paid labour are notneutral activities but symbolic performances through

The Internet and the gender division of household labour 53

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society

(with the Institute of British Geographers)

which ‘proper’ gender identities are enacted (Westand Zimmerman 1987; Valentine 1997; McDowellet al. 2005). Hence, housework and childcare are notnecessarily to be allocated or avoided. They can beexpressions of love and care, allowing men andwomen to craft identities that are in keeping withwider societal norms.

The Internet and gender

The studies discussed in the previous subsection dis-regard the increased importance of the online realmfor the gendering of everyday activities. However, adiversified literature about differences in Internet usehas come into existence across the social sciences.Much of this work focuses on the gender dimensionsof the digital divide – disparities in access to and useof the Internet and related technologies (Hargittai andShafer 2006; Helsper 2010; Gilbert and Masucci2011). Research has indicated that across geographi-cal contexts marked differences between men andwomen and among the latter exist in technical appa-ratus (hardware and connection speed), perceivedskills and the type of activities undertaken online.With 2004 German data, Zillien and Hargittai (2009)showed that women used the Internet more than menfor information on travelling and health and for e-mailbut less to access economic, political, computer andsports news or information on stock prices and prod-ucts. Dholakia (2006) reports comparable results forthe USA. Using UK data from 2007, Helsper (2010)found that women used the Internet more than menfor health information but less for online shopping,finance, social networking, personal networking, per-sonal communication and entertainment and access-ing sexual material. These differences depended,however, on employment status, marital status andage. Dholakia (2006) also showed that in the USAwomen tend to spend fewer hours online than men.

A variety of explanations for the above gender dis-parities have been put forward, including differencesin economic capital (monetary resources) and techno-logical capital, with women having poorer access tofast Internet connections and less online skill andexperience (Dholakia 2006; Hargittai and Shafer2006). Gendered design and the social constructionof technologies and online activities have also beenidentified as relevant (Singh 2001; Dholakia 2006;Kelan 2007): men tend to see the Internet as a tool andwomen consider it more as a toy, although both some-times ‘do’ gender differently and subvert stereotypicalgender–technology relations.

Geographers have extended the literature bygrounding gendered Internet use explicitly in activityparticipation in the offline world (Crang et al. 2006;Gilbert and Masucci 2011; Ren and Kwan 2009a; Renet al. 2012). Kwan and colleagues have explored howoffline maintenance activities (satisfying personal orhousehold needs) and offline leisure activities are

related to their online counterparts; people withextensive space–time constraints imposed by house-work and childcare – among whom women wereclearly overrepresented – tended to conduct feweronline activities, which lasted for shorter periods andwere undertaken later during the day (Ren et al.2012). In addition to these time displacement effects,substitution between time use online and offlineoccurred, with online maintenance replacing offlinemaintenance activities for women and online leisuresubstituting for offline leisure participation amongmen (Ren and Kwan 2009a). However, more timespent on online maintenance activities also impliedlonger travel times to access offline maintenanceactivities for women but not for men. Hence, theoverall effect of online maintenance activities on thegender division of household labour was rather small.

Geographers have also highlighted the spatiality ofdisparities in Internet access and use according togender and other social markers (Graham 2002; Cranget al. 2006; Gilbert 2010). Gilbert (2010) outlines amultiscalar model of digital inequalities from the inter-national scale to the individual, arguing that the met-ropolitan and neighbourhood levels are critical to theexplanation of inequalities. It is at these levels thatprocesses of the gendering and racialisation of thelabour market, residential segregation and the avail-ability of public and non-profit services operate,shaping people’s socio-technical resources for andconstraints on Internet use in important ways. Addi-tionally, geographers have shown location and acces-sibility to affect Internet use in the context of domesticlabour. Ren and Kwan (2009b) found the adoption ofonline shopping and frequency of online purchases tobe higher in white neighbourhoods and at locationswhere accessibility to brick-and-mortar shops is low.Farag et al.’s (2007) findings are more complex: inlocations offering good access to physical shops,online searching for product information is less fre-quent but online purchases are more common. Thispattern may reflect that people prefer to experience aproduct in-store and then buy it online. Thus, onlineand offline domestic activities may not simply substi-tute but also complement each other. Feminist geo-graphical research of how mothers use chat rooms andmothering websites point towards a similar conclusion(Madge and O’Connor 2006; Chan 2008): the Internetis not so much used to conduct activities that wouldotherwise take place in the offline world but rather torelieve concerns regarding parenting and home/workarrangements. It mobilises emotional and informa-tional support allowing mothers to carry on, therebycontributing to the perpetuation of existing inequalitiesin the gender division of household labour.

Towards the empirical analysis

Based on the above review we suggest that an inves-tigation bringing together the literatures on the gender

54 The Internet and the gender division of household labour

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society(with the Institute of British Geographers)

division of household labour and on gender and otherinequalities of Internet use is in order. Such an analy-sis should focus on the following eight key conceptsor variables, measured separately for men and womenin (heterosexual) couples: their Internet use, conductof housework, participation in childcare, engagementin paid work, as well as their technological capital,employment situation, household situation and thecharacter of their residential location. That analysisshould meet various requirements.

Caution should be exercised regarding the assump-tion that the observed relations among housework,childcare and paid labour are the outcome of arational choice process based on consensual deci-sionmaking between the partners and governed by alogic of instrumentality: power relations between part-ners can be complex, and domestic and employmentactivities are also sites for identity construction andreconfiguration (Hanson and Pratt 1995; McDowellet al. 2005). That is not to say that there are no trade-offs between partners’ activities but rather that theseare outcomes of complex negotiations within thehousehold and involve space–time constraints anddifferential access to resources, such as money, timeand skills.

It is also important to not a priori assume that aperson’s paid labour structures his/her domestic activi-ties – especially for women – or that Internet usesubstitutes for offline employment and domesticresponsibilities. Online activities may relate in muchmore complex ways to offline activities and it is alsopossible that no interactions exist at all. Additionally,the residential location may be integral to the genderdivision of household labour: households may selectthemselves into neighbourhoods that facilitate a par-ticular division or reduce the space–time constraints on(women’s) offline activities (Karsten 2003). Finally, it isimportant to distinguish between domestic responsi-bilities and to differentiate, at the very least, betweenchildcare and housework; the latter might well bedifferentiated in more feminine and masculine types.

We believe it is important to conduct empiricalanalyses with a fairly open mind about the nature anddirection of the interrelations of Internet use, house-work, childcare and paid work. We expect womenundertaking more childcare to also engage more inhousework, and this greater domestic responsibility toresult in less Internet use and less time spent in paidlabour (including commuting to work). But we alsowant to keep open the possibility that greater Internetuse and more paid work reduce women’s housework.For men we expect greater trade-offs between paidwork and domestic activities, and between childcareand housework. We anticipate that the situation ofInternet use displacing time spent on childcare andespecially housework is more likely for men than forwomen. For both genders, however, we expect theinterrelations among Internet use, housework, child-care and paid work to be shaped by the household

situation, the employment situation, and the characterof the residential location. Finally, we anticipate Inter-net use to be directly dependent on a person’s tech-nological capital, which might in turn be affected bythe household situation, employment situation andresidential location.

Research design

Study area and data

The study is situated in Columbus (Ohio), a mid-sizedmetropolitan area with a population of �1.62 millionin 2000. According to the 2000 US Census, white andAfrican-American groups account for �95% of thepopulation in the region’s most populated county(Franklin County). Columbus is particularly suitablefor our study because it has been called the ‘averagecity’ of the USA (England 1993). Over half of its popu-lation have computers and use the Internet – 57.1%have computers while 52.7% use the Internet in 1999,according to the 2000 US Census. The digital divide inOhio parallels national trends: African-Americans,poorer and less educated people use computers andthe Internet less.

The dataset used for the current study was collectedin 2003–2004 using an activity-Internet diary and aquestionnaire about Internet use and householdresponsibilities. Obviously the age of the data comeswith limitations: we cannot explore the recent growthin the use of mobile devices and in social media. Thedata nonetheless continue to have considerable value.There is no other dataset available with the same levelof detail on the gender division of household labourand Internet use; our data provide unique opportuni-ties to address the issues outlined above. Further,Internet use is a fast changing social practice that hasbeen transformed significantly since 2003 but thisdoes not necessarily imply that theoretical notionsand results from the pre-social networking and pre-tablet and smart phone era are useless (cf Hogan andQuan-Haase 2010). It can reasonably be assumed thatover time the relations between Internet use and thegender division of labour have both changed andremained stable. We therefore propose to understandour empirical analysis as exploratory and as offering areference point for the generation of new ideas forfuture research on people’s Internet use and thegender division of household labour.

To recruit respondents we sent screening packagesinviting couples with and without child(ren) andsingle parents with child(ren) to participate in thestudy to 32 000 randomly selected households livingacross Franklin County. A total of 875 householdsagreed and received a survey package with activity-Internet diaries and questionnaires for all adult house-hold members. Complete data from 420 adults in 270households were obtained. There were 272 women(65%) and 247 individuals with at least one

The Internet and the gender division of household labour 55

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society

(with the Institute of British Geographers)

co-resident child (59%). Most adults were white(93%), highly educated (80% were college or gradu-ate degree holders) and relied on their own automo-bile (97%). Apart from female, highly educated andwhite adults, respondents with a high householdincome were also overrepresented: About 14% hadan annual income under US$40 000 and 42% anannual income over US$80 000. Clearly, then, therespondents are not representative of the wider popu-lation, and we do not intend to generalise the resultsbelow to the entire population of the study area. Wehave excluded single parents from the analysis andutilise data from 405 of the original 420 participants.

Measuring key concepts

For this paper we employ the questionnaire data tomeasure the eight key concepts of Internet use, house-work, childcare, paid labour, technological capital,employment situation, household situation and theresidential location. Table 1 lists the variables drawnfrom the questionnaire to operationalise the key con-cepts and shows if and how scores on these variablesdiffer between women and men.

Internet use is described through measures of thetime spent per week on online activities in generaland for shopping and banking/paying bills. The femaleparticipants on average used the Internet for some 50min more than did their male counterparts and theyparticipated slightly longer in online shopping andonline banking/paying bills. Yet, the difference is onlystatistically significant (at p < 0.05) for banking/payingbills. Participation in housework and childcare isoperationalised using questions about the share ofspecific tasks undertaken by participants. Table 1shows that women’s contribution exceeds men’s formost tasks, although cleaning outdoors, gardening/yard work and repairs are male-dominated activities.Women’s role is much greater than men’s for suchroutine activities as cooking, in-home cleaning andgrocery shopping. Overall the observed patternconcurs with previous findings (South and Spitze1994; Kroska 2004) and underlines the importance ofconsidering different types of housework and child-care. The average shares for men and women do notadd up to 100% in Table 1. This reflects that in couplesin which both adults participated in the study, others(children, cleaners, professional carers, etc.) mayperform certain duties and/or partners may have dif-ferent perceptions of their own role in the household.A second reason is that the analysis includes 60couples for which we had only data on one adult –typically the female – available. We preferred toinclude only one person from these households in theanalysis rather than excluding them altogether; thelatter would have made it more difficult to reliablyestimate the coefficients in the structural equationmodel described below. The space–time configurationof paid labour is described using weekly employment

hours and the home-to-work distance. Only theformer of these observed variables varies significantly(at p < 0.05) between men and women.

Technological capital is operationalised using indi-cators of experience, skills and the technical appara-tus – i.e. the connection speed – for accessing theInternet. In contrast to previous studies (Dholakia2006; Hargittai and Shafer 2006; Zillien and Hargittai2009), there are no differences (at p < 0.05) betweenwomen and men for these indicators. The employ-ment situation is described through personal income,occupation level and the industry of employment. Thewomen in the recruited sample earn significantly (at p< 0.05) less than the men, but differences in occupa-tion and industry are very small. Table 1 also lists thevariables that describe the participants’ householdsituation and residential location. We only considerthe residential location as this is the place where mosthousework and childcare and online activities wereundertaken.

Method of analysis

We employ structural equation modelling (SEM),which enables us to construct latent variables forInternet use, housework, childcare, paid labour, tech-nological capital, employment situation, householdsituation and the residential location out of theobserved variables listed in Table 1; and to examinehow those latent constructs are interrelated. Theadvantage of SEM is that it allows for more complexinteractions between variables than standard regres-sion analysis. Causation cannot be established withcross-sectional data but within a larger set of correla-tive associations SEM enables comparison of the effectof latent construct A on B with the effect of B on A interms of model fit. SEM also allows the relationsbetween two latent constructs to be separated intodirect and indirect effects. In the former case, A has aneffect on B without the involvement of other con-structs. With an indirect effect, A affects B via one ormore other constructs. The total effect is the sum of thedirect and all indirect effects from one construct toanother.

In SEM a distinction is made between the measure-ment model and the structural model. The formerdescribes how the various latent variables are con-structed out of the observed variables; the latter therelations between the latent constructs. Coefficientsare estimated via the covariance method, which canbe done in various ways. Since many of our observedvariables are ordinal in nature, we estimated poly-choric correlations and asymptotic covariance matri-ces (Jöreskog 2005), and used maximum likelihood(ML) as estimation method. Weighted least squares isoften used when many variables are ordinal in nature,but results can be severely biased when the numbersof observations are low (Jöreskog and Sörbom 2001);ML or other estimation methods are then preferred.

56 The Internet and the gender division of household labour

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society(with the Institute of British Geographers)

As the relations among Internet use, housework,childcare and paid labour may differ between womenand men, we have specified a multi-group structuralequation model. This means that different measure-ment and structural models and correlation and cov-ariance matrices are estimated for each gender

(Jöreskog 2005). Multi-group models are usuallyemployed to confirm whether the same model struc-ture (the total set of relations among measured andlatent variables), coefficients, residuals, and/or thesame correlation and asymptotic covariance matricesapply to each group (Bollen 1989). But it was evident

Table 1 Observed variables in the analysis

Women Men Total

Mean Share Mean Share Mean Share

Internet use: hours spent per week on online . . .Activities in general 10.9 10.1 10.6Shopping 0.4 0.3 0.4Banking/paying bills 0.5a 0.3 0.4

Housework: share of . . . in householdGrocery shopping 76.1a 27.0 58.2Cooking/baking/preparing meals 74.9a 28.1 57.8In-home cleaning 72.1a 28.5 56.2Cleaning outdoors 39.1a 63.9 48.2Gardening/yard work 36.1a 64.2 46.4Repairs 26.0a 78.1 45.0Banking/paying bills 64.7a 41.0 56.0

Childcare: share of . . . in householdCare for older child(ren) 34.3a 16.6 27.8Transporting child(ren) 38.9a 16.8 30.8Helping child(ren) with homework 33.4a 13.1 26.0

Paid labourWeekly employment hours 29.8a 36.8 32.4Home to work distance (miles) 7.1 8.0 7.5

Technological capitalYears of internet experience 6.8 7.0 6.9Internet ability (1 = poor, 5 = excellent) 3.8 3.9 3.8Internet connection Slow 45.7 45.3 45.5

Fast (DSL/cable) 54.3 54.7 54.5Residential location

Number of stores within 5 min from home 133.1 122.3 129.1Population density, census block group area 156.7 153.3 155.5Median household income, census block group area 60.4 61.6 60.9Share of higher education persons, census block group area 48.2 49.1 48.5Share of whites, census block group area 85.5 86.3 85.8

Employment situationPersonal income Less than US$20 000 31.6a 11.5 24.3

US$20 000–39 999 27.3 34.5 30.0US$40 000–59 999 19.5 23.0 20.8More than US$60 000 21.5 31.1 25.0

Occupation level Managerial/professional 43.6 46.6 44.7Other 56.4 53.4 55.3

Sector of the economy GovernmentOther 14.8 13.5 14.3

Household situationChild 0–6 years in household Yes 26.5 31.8 28.3

No 73.5 68.2 71.7Household size (1 = 1, 5 = 5 or more) 3.0 3.1 3.0

aStatistically different from the average or distribution for men at p < 0.05.

The Internet and the gender division of household labour 57

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society

(with the Institute of British Geographers)

that assuming equal matrices, residuals, coefficientsand/or model structures would obscure critical genderdifferences in our data. We therefore developeda three-step modelling approach that is more explora-tory and iterative than most multi-group SEManalyses:

1 We developed well fitting model specifications forwomen and men in completely separate analyses.

2 In an iterative process in which the original speci-fications changed to some extent, we merged thedata for women and men and re-estimated the cor-relation and asymptotic covariance matrices andmodel specifications.

3 We made the step 2 model, in which all coefficientswere allowed to differ between men and women,more parsimonious by constraining coefficients tobe equal across genders wherever possible. Westarted with those coefficients for which the differ-ence in magnitude between women and men wassmallest and worked towards those with greaterdifferences in estimated coefficients, using c2 teststo avoid that the model as a whole became statis-tically inferior. Imposing constraints induced further(minor) changes to the model specification but inthe end we arrived at a model that is conceptuallyplausible and maximises the extent of gender dif-ference in a statistically robust manner: constrain-ing any further coefficient to be equal for womenand men will produce a model that fits the datasignificantly (p < 0.05) worse than at present.

At the heart of our approach is a reversal of theanalytical logic proposed by Bollen (1989) andJöreskog (2005). Instead of working from similarityacross groups towards differentiation, we started fromdifference and worked towards similarity to reducethe risk of overseeing potentially relevant gender dif-ferences. The adopted modelling approach alsoenabled us to make sure that the final set of coefficientestimates is plausible and robust given the samplesize.

During the SEM analysis we have considered therequirements and expectations regarding the interre-lations among the latent constructs for Internet use,housework, childcare, paid labour, technologicalcapital, employment situation, household situationand the residential location outlined in ‘Towards theempirical analysis’. We also considered direct effectsfrom childcare to housework more plausible than thereverse, because childcare activities impose morestringent space–time constraints and more often act asanchors around which other activities are scheduled(Schwanen 2007; Schwanen et al. 2008). However, incases where reverse links proved statistically equiva-lent or superior, we have correlated the residuals. Tokeep the model identifiable, we have specified thelatent constructs of respondents’ employment situa-tion and household situation as exogenous latent

variables, meaning they cannot be affected by any ofthe other latent variables.

Results

The final model fits the data reasonably well. There isno unanimity regarding goodness of fit measures forSEM but the most widely used indicator is the rootmean square error of approximation (RMSEA), whichis in our case 0.0428 and reliably (p < 0.01) below thecut-off criterion for acceptable models of 0.05. Thestandardised root mean square residual (SRMR) is0.0843 for women and 0.0847 for men, which is alsobelow the common 0.10 criterion. However, theSatorra–Bentler scaled c2 statistic differs clearly from 0(c2 = 953.8, df = 690) but this is rather common whenhundreds of observations are used and not necessarilyproblematic. While our model does not fully replicatethe data, it is adequate and acceptable.

Measurement models

The most important conclusion from the measurementmodels is that the relations between the observedvariables and the latent constructs are both genderedand complex. This is especially true of the latent con-structs for housework, childcare, paid labour andrespondents’ employment situation. Table 2 includestwo latent constructs for housework – one called mas-culine and the other feminine housework. We havespecified two constructs because it became clear earlyon that distinguishing between male-dominated andfemale-dominated tasks significantly increased themodel’s fit to the data. Table 2 further includes latentconstructs for Internet use, technological capital,household situation and the residential location. Weintended to have a single construct for all observedlocation variables but specifying two partially con-nected latent constructs provided more plausibleresults and a statistically superior model. One con-struct represents the population composition of theresidential neighbourhood and especially its socio-economic status (because the share of whites has alower coefficient than the other variables). The othermeasures land use and accessibility, although theshare of higher educated persons in the neighbour-hood also loads onto this construct (albeit lessstrongly than onto the population composition con-struct). The double loading of the share of highereducated persons reflects the geography of Columbus:highly educated individuals are concentrated not onlyin middle-class, almost exclusively white neighbour-hoods but also in relatively dense and mixed neigh-bourhoods in the vicinity of the Ohio State Universitycampus and Columbus’ CBD.

The latent constructs are gendered because theobserved variable loadings for feminine housework,masculine housework, paid labour and employmentsituation differ significantly (at p < 0.05) between

58 The Internet and the gender division of household labour

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society(with the Institute of British Geographers)

women and men (Table 2). Most differences are minorbut there are some exceptions. First, working in thegovernment sector loads considerably stronger ontothe latent construct for women’s employment situa-tion. More than for men being a governmentemployee is associated with high levels of occupa-tional prestige and income for women. Second, theslightly smaller coefficient for weekly employmenthours for women – 2.386 versus 2.913 for men –indicates that home-to-work distance and the numberof hours worked per week are related slightly strongerfor women. Our model thus expresses that the lengthof women’s commute is more strongly dependentthan men’s on how many hours they work. This mayfollow from a greater tendency to economise on com-

muting relative to the total employment time amongthe participating women (Schwanen and Dijst 2002),which might reflect more extensive domestic respon-sibilities and/or lower pay (see, for instance, Hansonand Pratt 1995).

Third, Table 2 also shows that there are no overlapsbetween the observed variables loading on the latentconstructs of feminine and masculine housework formen. These latent constructs are more clearly identi-fiable among the men than the women, for whom theconstructs are partially connected (Strathern 2004)and the shares of in-home cleaning and repairs loadon both feminine and masculine housework. Thestronger separation of housework into two categoriesfor men may reflect cultural scripts regarding men’s

Table 2 Measurement models

Name Observed variableCoefficient

womenCoefficient

men

Latent constructs varying between women and menFeminine housework Grocery shopping 1.000 1.000

Cooking/baking/preparing meals 0.945 0.861In-home cleaning 0.790 0.671Banking/paying bills 0.654 0.431Repairs 0.283Duration of online banking 0.122

Masculine housework Gardening/yard work 1.000 1.000Cleaning outdoors 1.218 1.250Repairs 0.590 0.640In-home cleaning 0.218

Childcare Transporting children 1.000 1.000Care for older children 1.125 1.182Helping children with homework 1.051 1.040

Paid labour Home-to-work distance 1.000 1.000Weekly employment hours 2.386 2.913

Employment situation Personal income 1.000 1.000Managerial/professional occupation 0.834 0.773Employed in government sector 0.924 0.498

Name Observed variable Coefficient

Latent constructs that do not vary between women and menInternet use Time spent on online shopping 1.000

Time spent on online banking/paying bills 0.928Time spent online 1.472Banking/paying bills 0.368

Technological capital Years of Internet experience 1.000Internet ability 1.275Internet connection 0.673

Residential location, population Median household income, census block group area 1.000Share of higher educated, census block group area 1.059Share of white people, census block group area 0.558

Residential location, land use andaccessibility

Population density, census block group area 1.000Number of stores within 5 min from home 0.925Share of higher educated, census block group area 0.605

Household situation Child 0–6 years in household 1.000Household size 5.774

The Internet and the gender division of household labour 59

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society

(with the Institute of British Geographers)

and women’s role in the household (Vincent et al.2004; Schwanen 2007): whereas men’s involvementin domestic responsibilities is considered more volun-tary and laudatory, women’s engagement in the fullrange of tasks is much more taken for granted inWestern culture. Because of such scripts the womenin our study may, on average, have experienced lesschoice in determining which types of housework toundertake.

Partial overlap between latent constructs can alsobe observed between Internet use and femininehousework; the share of banking/paying bills loads onthe construct for Internet use for both genders andduration of online banking also loads on that for femi-nine housework among women. With the growth ofInternet use for banking/paying bills the managementof the household’s finances has become so digitallymediated that drawing a clear-cut distinction betweenoffline housework and online activity is increasinglydifficult, particularly for women. We have thereforedefined the latent construct of Internet use for bothgenders and feminine housework for women alongthe lines of Haraway’s (1991) cyborg metaphor asonline/offline hybrids. These hybrid definitionsimproved the model fit without clearly affecting therelations among the latent constructs in the structuralmodels.

Structural models

Table 3 summarises the final versions of the structuralmodels for women and men. A comparison showsthat the model for women is more complex, withmore direct effects included.

Internet use Contrary to expectations, Internet use isrelated to the gender division of household labour forwomen but not men. Women undertaking more femi-nine housework tasks like grocery shopping andcooking spend less time online, which suggests thatspace–time constraints associated with offline com-mitments prevent women from using the Internet to amuch greater extent than men. We also exploredreverse effects from Internet use to housework andchildcare but none of these rendered statistically sig-nificant effects (at p < 0.05). Other latent constructsassociated with the gender division of householdlabour have indirect effects on Internet use becausewomen’s feminine housework responsibilities are afunction of their childcare responsibilities; the landuse and accessibility of their residential location; andtheir employment situation (Table 3). Perhaps surpris-ingly, more childcare responsibilities for women mar-ginally increases rather than displaces time online.This positive indirect effect results from the multipli-cation of the negative direct effect of feminine house-work on Internet use and a second negative directeffect from childcare to feminine housework. Thelatter suggests that with more childcare responsibili-

ties, women (like men) to some extent substitute care-giving for feminine housework (which is reallocated totheir partner, to children or to others); this substitutionhelps to conserve some time for Internet use amongthe female participants. Supplementary analysis of theactivity-Internet diaries indicated that mothers alsoused the Internet to search for information related tochildrearing and their children’s development. Theindirect effect of land use and accessibility is positive:women living in higher-density locations with betteraccessibility to stores use the Internet more exten-sively, because the more equal division of (feminine)housework between partners frees up time for Internetuse.

The women’s Internet use is affected by both thedivision of household labour and their employmentsituation – and by implication their socioeconomicposition – but men’s use is only related to the latter.The effect of the employment situation is identical forboth genders. A higher income and position at theoccupational ladder result in more technologicalcapital, which leads to more time online. However,because the effect of technological capital on Internetuse is significantly stronger for men than for women,the effects of employment situation on Internet use isslightly larger for the former. Contrary to expectations,technological capital is not affected by men’s orwomen’s household situation and residential location.

Overall there are only modest relationshipsbetween Internet use and the gender division ofhousehold labour: they can only be observed forwomen and the effects run from offline housework toInternet use rather than the reverse. Internet use didnot appear to have induced a more equitable genderdivision of labour among middle-class households inColumbus in 2003. Factors beyond Internet use andtechnological capital were more clearly associatedwith that division.

Paid labour, childcare and housework The interac-tions of domestic and employment activities differ bygender and are more complex for women. Morechildcare responsibilities imply that fathers carrygreater responsibility for masculine and especiallyfeminine housework, and that they work andcommute longer. The latter effect may reflect thatlarger households with young children need morefinancial resources and a cultural script assigning theprovider role more strongly to fathers (Aitken 2000).Relations are different for the recruited women: child-care responsibilities are not associated with weeklyemployment hours and commute length, but there aretrade-offs between paid work and feminine house-work. During the modelling it did not become clearwhether employment hours and commute distanceshaped feminine housework or the reverse; the coef-ficients associated with each effect were comparablein size. We have therefore not specified direct effectsbetween paid labour and feminine housework but

60 The Internet and the gender division of household labour

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society(with the Institute of British Geographers)

Tabl

e3

Stru

ctur

alm

odel

s

Expl

anat

ory

cons

truc

ts

Wom

en–

cons

truc

tsto

beex

plai

ned

Men

–co

nstr

ucts

tobe

expl

aine

d

INTU

SEFH

WM

HW

CC

AR

EPL

TCA

PR

LPO

PR

LLU

AIN

TUSE

FHW

MH

WC

CA

RE

PLTC

AP

RLP

OP

RLL

UA

FHW

D-0

.129

**0.

339*

*0.

127*

T-0

.129

**0.

399*

*0.

127*

CC

AR

ED

-0.1

38**

0.00

1-0

.138

**0.

057*

*T

0.01

8**

-0.1

38**

-0.0

55**

0.00

1-0

.138

**-0

.017

*0.

057*

*TC

AP

D0.

734*

*0.

935*

*T

0.73

4**)

0.93

5**

RLP

OP

D-0

.059

**-0

.059

**T

-0.0

59**

-0.0

59**

RLL

UA

D-0

.183

**-0

.179

**-0

.383

**T

0.02

1**

-0.1

58**

-0.0

63**

-0.1

790.

000

0.05

3**

0.00

7-0

.383

**-0

.022

*)EM

PLO

YD

-0.1

50**

-0.1

20**

0.31

0**

0.20

2**

0.06

5-0

.087

*-0

.150

**0.

185*

*0.

202*

*0.

649*

*-0

.360

**T

0.16

4**

-0.1

19**

-0.0

48**

-0.1

04**

0.30

6**

0.20

2**

0.06

5-0

.087

*0.

189*

*-0

.169

**-0

.021

*0.

138*

*0.

154*

*0.

202*

*0.

649*

*-0

.360

**H

HD

0.71

0**

0.23

9**

-0.3

11**

0.95

3**

-0.3

11**

T0.

006

-0.0

49-0

.019

0.76

6**

-0.0

130.

239*

*-0

.311

**-0

.148

**-0

.019

1.07

2**

0.06

1*-0

.311

**C

orre

late

dre

sidu

als

FHW

and

CC

AR

E0.

275*

*M

HW

and

CC

AR

E0.

091*

*M

HW

and

CC

AR

E0.

091*

*R

LPO

Pan

dR

LLU

A-0

.188

**FH

Wan

dPL

-0.0

43*

HH

and

EMPL

OY

0.04

0**

RLP

OP

and

RLL

UA

-0.3

96**

HH

and

EMPL

OY

-0.0

04

Bol

dte

xtre

pres

ents

dire

ctef

fect

orco

rrel

ated

resi

dual

sw

ithco

effic

ient

diffe

ring

from

othe

rge

nder

atp

<0.

05;

*sta

tistic

ally

sign

ifica

ntfr

omze

roat

p<

0.10

;**

stat

istic

ally

sign

ifica

ntfr

omze

roat

p<

0.05

.C

CA

RE,

child

care

;D

,di

rect

effe

ct;

FHW

,fe

min

ine

hous

ewor

k;IN

TUSE

,In

tern

etus

e;M

HW

,m

ascu

line

hous

ewor

k;PL

,pa

idla

bour

;R

LLU

A,

resi

dent

ial

loca

tion,

land

use

and

acce

ssib

ility

;R

LPO

P,re

side

ntia

llo

catio

n,po

pula

tion;

T,to

tal

effe

ct;T

CA

P,te

chno

logi

cal

capi

tal.

The Internet and the gender division of household labour 61

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society

(with the Institute of British Geographers)

instead correlated the residuals. As mentioned andcontradicting expectations, for many of the womenmore childcare meant less housework. Yet, the posi-tive correlation of the residuals for housework andchildcare indicates that this does not hold for allwomen; there appears to be a number of householdswhere all domestic responsibilities are concentratedin the female partner. On balance our analysis offerslittle support for the expectation of greater trade-offsbetween paid labour and domestic activities andbetween childcare and housework among therecruited men.

The model indicates that the household situation,employment location and residential location are allrelated to the gender division of household labour, andthe effects are generally in line with expectations andprevious studies. Of particular interest is that the totaleffects of the household situation on childcare andfeminine housework show that, in relative terms, thegender balance in domestic responsibilities is moreequal in larger households or when a young child ispresent. The opposite holds, however, regarding paidlabour in those households. The model also indicatesthat a higher income and better occupational positionhas stronger effects on women’s weekly working hoursand commute length than on men’s. The effects onchildcare responsibilities differ by gender: a higherincome and better occupational position increasemen’s but reduces women’s childcare responsibilities.

Further, our analysis confirms that the gender divi-sion of household labour is geographically differenti-ated within Franklin County in Columbus, Ohio. Indenser neighbourhoods providing better accessibilityto stores and a better educated population, the divisionof housework is more balanced: women undertake lessmasculine and particularly feminine housework andtheir partners do more. Men and women also under-take less childcare in those neighbourhoods but this isprimarily because households with children are lesslikely to live there. Because the negative effect onchildcare is stronger for men, the division of thoseresponsibilities is less balanced and more disadvanta-geous to mothers in denser neighbourhoods. Finally,the residential location is indeed integral to the genderdivision of household labour; the model fit improvedsignificantly once residential location was treated asinterwoven and endogenous to the relations betweenthe household and employment situation on the onehand and the gender division of labour on the other.Hence, the model confirms the idea that having onebut preferably two well paid, high-status jobs enablescouples to both live in a socioeconomically goodneighbourhood and keep commuting and employmenthours within bounds.

Conclusions and discussion

Our analysis contributes to the existing literature infour ways. It demonstrates the importance of consid-

ering gender differences in the factors that are associ-ated with Internet use. The time spent on variousonline activities and Internet use in general wasrelated to technological capital and the employmentsituation for both the women and the men in ourstudy. In addition, however, Internet use was also con-strained by the extent of responsibility for certaintypes of housework and for childcare and associatedto the location of residence for women. There wasthus a richer and more varied set of factors shapingwomen’s Internet use.

Secondly, for the households considered the inter-relations between the gender division of householdlabour and Internet use were rather limited. Amongwomen Internet use was shaped by domestic respon-sibilities, but reverse effects whereby the gender divi-sion of household labour is more balanced inhouseholds where women spend more time onlinecould not be detected. This may have changed since2003–2004 when the data for the current study werecollected, and further analysis using more recent datais much desired. Nonetheless, all too optimistic nar-ratives about the Internet’s potential to reconfigure thegender division of labour should be treated withcaution; previous research has suggested thatinequalities according to gender and other socialmarkers can be consolidated and even exacerbatedwith the growth and development of Internet use(Graham 2002; Schwanen and Kwan 2008). Addition-ally, much housework and childcare entails physicaleffort and face-to-face contact, and the effects of Inter-net use can only be indirect here. It can raise aware-ness and confidence among women that the timedevoted to physical housework and childcare can berenegotiated with their partner, but qualitativeresearch on chat-room use among mothers has not(yet) offered support for such change (Madge andO’Connor 2006; Chan 2008). Finally, additionalanalysis using the collected activity-Internet data(available from the authors upon request) indicatesthat the recruited women undertake online activities,such as paying bills, e-commerce, and searching forrecipes and health information, more frequently thando men. Only helping children online with homeworkor games and searching for online information orsupport regarding parenting were done equally oftenacross both genders. For housework the offline genderrelations tended to be replicated online among studyparticipants.

Our analysis also extends the literature about thegender division of household labour more generallyby confirming that different mechanisms are impli-cated in the organisation of housework, childcare andpaid labour among women and men. For men agreater contribution to childcare led not only togreater responsibility for housework but also to moretime spent on paid labour and commuting; for womenthere were trade-offs between childcare and house-work and between the responsibility for housework

62 The Internet and the gender division of household labour

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society(with the Institute of British Geographers)

and employment hours and commuting distance. Ourfindings challenge the notion that the partner with thelongest employment hours or commute time alwaysnegotiates his/her way out of domestic responsibili-ties. They are more consistent with the notions thatpower relations between partners are complex; thatdomestic and employment activities are also plat-forms for identification processes; and that culturalscripts matter to how women and men organise thegender division of household labour (Hanson andPratt 1995; Vincent et al. 2004; McDowell et al. 2005;Schwanen 2007). Nevertheless, respondents’ employ-ment situation was of key importance to how paidwork, childcare and housework were organised forboth genders. From a policy perspective our analysisconcurs with the idea that improving women’s oppor-tunities to obtain well paid, high-status jobs is a veryeffective strategy to increase the gender balance inemployment and domestic responsibilities.

Finally, the analysis shows that quantitativemethods can make important and unique contribu-tions to critical geography (Kwan and Schwanen2009). Advanced methods such as SEM can capturemany complexities of everyday life and its geographi-cal context; challenge dichotomous, hard-and-fastcategories as our measurement models have done;and allow for rigorous comparisons of the experiencesof different social groups and so shed new light on thedifferential mechanisms through which socio-spatialinequalities in everyday life come about.

Acknowledgements

The comments of the editor and anonymous reviewerson an earlier version of the manuscript are gratefullyacknowledged. The data used in this paper were col-lected with a grant from the US National ScienceFoundation’s Information and Technology Research(NSF/ITR) Program (BCS-0112488) to Mei-Po Kwan.

References

Aitken S C 2000 Fathering and faltering: ‘Sorry, but you don’thave the necessary accoutrements’ Environment and PlanningA 32 581–98

Banister D 2011 Cities, mobility and climate change Journal ofTransport Geography 19 1538–46

Becker G S 1965 A theory of the allocation of time The EconomicJournal 68 571–83

Bianchi S M, Milkie M A, Sayer L C and Robinson R P 2000 Isanyone doing homework? Trends in the gender division ofhousehold labor Social Forces 79 1991–2028

Bittman M, Rice J M and Wajcman J 2004 Appliances and theirimpact: the ownership of domestic technology and time spenton household work The British Journal of Sociology 66 401–23

Bollen K A 1989 Structural equations with latent variables JohnWiley & Sons, New York

Chan A H-N 2008 ‘Life in happy land’: using virtual space anddoing motherhood in Hong Kong Gender, Place and Culture15 169–88

Cowan C S 1983 More work for mother: the ironies of householdtechnology from the open hearth to the microwave BasicBooks, New York

Crang M, Crosbie T and Graham S 2006 Variable geometries ofconnection: urban digital divides and the uses of informationtechnology Urban Studies 43 2551–70

De Meester E and Van Ham M 2009 Symmetry and asymmetryin working and commuting arrangements between partners inthe Netherlands: Does residential context matter? Environmentand Planning A 41 2181–200

Dholakia R R 2006 Gender and IT in the household: evolvingpatterns of Internet use in the United States The InformationSociety 22 231–40

England K V L 1993 Suburban pink collar ghettos: the spatialentrapment of women? Annals of the Association of AmericanGeographers 83 225–42

Ettema D, Schwanen T and Timmermans H 2007 The effects oflocational, mobility and socio-demographic factors on taskand time allocation in households Transportation 34 89–105

Ettema D and Van der Lippe T 2009 Weekly rhythms in task andtime allocation of households Transportation 36 113–29

Farag S, Schwanen T, Dijst M and Faber J 2007 Shopping onlineand/or in-store? A structural equation model of the relation-ships between e-shopping and in-store shopping. Transporta-tion Research Part A 41 125–41

Fisher K, Egerton M, Gershuny J I and Robinson J P 2007 Genderconvergence in the American Heritage Tim Use Study(AHTUS) Social Indicators Research 82 1–33

Gilbert M R 2010 Theorizing digital and urban inequalities Infor-mation, Communication & Society 13 1000–18

Gilbert M R and Masucci M 2011 Information and communica-tion technology geographies: strategies for bridging the digitaldivide Praxis (e)Press, E-edition (www.praxis-epress.org/ICT/ictgeographies.pdf) Accessed 1 December 2012

Graham S 2002 Bridging urban digital divides? Urban polariza-tion and information and communication technologies (ICTs)Urban Studies 39 33–56

Hanson S and Hanson P 1980 Gender and urban activity patternsin Uppsala, Sweden Geographical Review 70 291–9

Hanson S and Pratt G 1995 Gender, Work, and Space Routledge,London

Haraway D J 1991 Simians, cyborgs, and women: the reinventionof nature Routledge, New York

Hargittai E and Shafer S 2006 Differences in actual and per-ceived online skills: the role of gender Social Science Quar-terly 87 432–48

Helsper E J 2010 Gender Internet use across generations and lifestages Communications Research 37 352–74

Hogan B and Quan-Haase A 2010 Persistence and change insocial media Bulletin of Science, Technology & Society 30309–15

Huber J and Spitze G D 1983 Sex stratification: children, house-work and jobs Academic Press, New York

Hubers C, Schwanen T and Dijst M 2011 Coordinating everydaylife in the Netherlands: a holistic quantitative approach to theanalysis of ICT-related and other work-life balance strategiesGeografiska Annaler Series B: Human Geography 93 57–80

The Internet and the gender division of household labour 63

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society

(with the Institute of British Geographers)

Jöreskog K G 2005 Structural equation modeling with ordinalvariables using LISREL (www.ssicentral.com/lisrel/techdocs/ordinal.pdf) Accessed 15 June 2011

Jöreskog K G and Sörbom D 2001 LISREL 8: User’s referenceguide Scientific Software International, Lincolnwood, IL

Karsten L 2003 Family gentrifiers: challenging the city as a placesimultaneously to build a career and to raise children UrbanStudies 40 2573–84

Kelan E K 2007 Tools and toys: communicating gendered posi-tions towards technology Information, Communication &Society 10 358–83

Kroska A 2004 Divisions of domestic work: Revising andexpanding the theoretical explanations Journal of Family Issues25 900–32

Kwan M-P 1999 Gender, the home–work link and space–timepatterns of nonemployment activities Economic Geography 76370–94

Kwan M-P 2000 Gender differences in space–time constraintsArea 32 145–56

Kwan M-P 2002 Time, information technologies, and the geog-raphies of everyday life Urban Geography 23 471–82

Kwan M-P 2003 Gender troubles in the Internet era FeministMedia Studies 3 371–4

Kwan M-P and Schwanen T 2009 Quantitative revolution 2: thecritical (re)turn The Professional Geographer 61 283–91

Madge C and O’Connor H 2006 Parenting gone wired: empow-erment of new mothers on the Internet? Social & CulturalGeography 7 199–220

McDowell L, Ray K, Perrons D, Fagan C and Ward K 2005Women’s paid work and moral economies of care Social &Cultural Geography 6 219–35

Presser H B 2003 Working in a 24/7 economy: challenges forAmerican families Russell Sage Foundation, New York

Ren F and Kwan M-P 2009a The impact of the Internet on humanactivity-travel patterns: analysis of gender differences usingmulti-group structural equation models Journal of TransportGeography 17 440–50

Ren F and Kwan M-P 2009b The impact of geographic context one-shopping behavior Environment and Planning B: Planningand Design 36 262–78

Ren, F, Kwan M-P and Schwanen T 2012 Investigating the tem-poral dynamics of Internet activities Time and Society in press

Schwanen T 2007 Gender differences in chauffeuring childrenamong dual-earner families The Professional Geographer 59447–62

Schwanen T and Dijst M 2002 Travel-time ratios for visits to theworkplace: the association between commuting time andwork duration Transportation Research Part A: Policy and Prac-tice 36 573–92

Schwanen T and Kwan M-P 2008. The Internet, mobile phoneand space–time constraints Geoforum 39 1362–77

Schwanen T, Kwan M-P and Ren F 2008 How fixed is fixed?Gendered rigidity of space-time constraints and geographies ofeveryday activities Geoforum 39 2109–21

Shelton B A and John D 1996 The division of household laborAnnual Review of Sociology 22 299–322

Singh S 2001 Gender and the use of the Internet at home NewMedia & Society 3 395–416

South S J and Spitze G 1994 Housework in marital and non-marital households American Sociological Review 59 327–47

Strathern M 2004 Partial connections, updated edition Rowman& Littlefield Publishers, Walnut Creek, CA

Tivers J 1985 Women attached: the daily lives of women withyoung children Croom Helm, London

Valentine G 1997 ‘My son’s a bit dizzy’, ‘My wife’s a bit soft’:gender, children and cultures of parenting Gender, Place andCulture 4 37–62

Valentine G and Holloway S L 2002 Cyberkids? Exploring chil-dren’s identities and social networks in on-line and off-lineworlds Annals of the Association of American Geographers 92302–19

Vincent C, Ball S J and Pietikainen S 2004 Metropolitan mothers:mothers, mothering and paid work Women’s Studies Interna-tional Forum 27 571–87

West C and Zimmerman D H 1987 Doing gender Gender andSociety 1 125–51

Zillien N and Hargittai E 2009 Digital distinction: status-specifictypes of Internet usage Social Science Quarterly 90 274–91

64 The Internet and the gender division of household labour

The Geographical Journal 2014 180 52–64 doi: 10.1111/geoj.12014© 2013 The Authors. The Geographical Journal © 2013 Royal Geographical Society(with the Institute of British Geographers)