travel to work and household responsibility: new evidence

23
Travel to work and household responsibility: new evidence TRACY TURNER 1 & DEBBIE NIEMEIER 2, * 1 Department of Economics; 2 Department of Civil and Environmental Engineering, University of California, Davis, CA 95616, USA ( * Author for correspondence) Accepted 4 February 1997 Key words: gender, household responsibility, travel behavior Abstract. A persistent empirical finding in the research on travel patterns is that women tend to engage in shorter work commutes than men. Research evaluating the degree to which this gender differential in commuting may be explained by the division of labor in the household has produced decidedly mixed findings. This paper presents a critical review of the methods and results of recent research on the gender differential in commuting along with the associ- ated implications for the household responsibility hypothesis (HRH). While all of the articles reviewed are informative and unique, not all conclusions are well supported. The paper then tests the HRH using the 1990 NPTS travel data. The new evidence confirms that women continue to exhibit shorter commute times and distances than men and provides support for the HRH. The paper concludes with a summary of findings and recommendations for further research. Introduction An extensive body of research has identified gender as an important pre- dictor of travel patterns. Findings suggest that employed women tend to have shorter commute-to-work distances and times than employed men (Blumen 1994; Hanson & Johnston 1985; Hanson & Pratt 1990; Madden 1981; McLafferty & Preston 1991); 1 that women tend to spend more time in house- hold and family support activity (Niemeier & Morita 1996; Hanson & Hanson 1980) and make more household and family support trips (Hanson & Hanson; Hanson & Johnston; Rosenbloom 1987) and that women make fewer recre- ational trips (e.g., Hanson & Johnston). Among these findings, gender differences in journey-to-work trips is the most studied and is the subject of ongoing debate. Existing empirical research is unified in finding that women have shorter commutes than men, however, there is disagreement on the factors which contribute to the gender differential. The underlying economic theory presented by White (1977, 1986) and Madden (1981) suggests that commute length is a function of wages, housing prices, income and other demographic vari- ables with wages and housing prices endogenously determined. Because Transportation 24: 397–419, 1997 1997 Kluwer Academic Publishers. Printed in the Netherlands.

Upload: tracy-turner

Post on 03-Aug-2016

217 views

Category:

Documents


0 download

TRANSCRIPT

Travel to work and household responsibility: new evidence

TRACY TURNER1 & DEBBIE NIEMEIER2, *1

Department of Economics; 2 Department of Civil and Environmental Engineering,University of California, Davis, CA 95616, USA (* Author for correspondence)

Accepted 4 February 1997

Key words: gender, household responsibility, travel behavior

Abstract. A persistent empirical finding in the research on travel patterns is that women tendto engage in shorter work commutes than men. Research evaluating the degree to which thisgender differential in commuting may be explained by the division of labor in the householdhas produced decidedly mixed findings. This paper presents a critical review of the methodsand results of recent research on the gender differential in commuting along with the associ-ated implications for the household responsibility hypothesis (HRH). While all of the articlesreviewed are informative and unique, not all conclusions are well supported. The paper thentests the HRH using the 1990 NPTS travel data. The new evidence confirms that women continueto exhibit shorter commute times and distances than men and provides support for the HRH.The paper concludes with a summary of findings and recommendations for further research.

Introduction

An extensive body of research has identified gender as an important pre-dictor of travel patterns. Findings suggest that employed women tend to haveshorter commute-to-work distances and times than employed men (Blumen1994; Hanson & Johnston 1985; Hanson & Pratt 1990; Madden 1981;McLafferty & Preston 1991);1 that women tend to spend more time in house-hold and family support activity (Niemeier & Morita 1996; Hanson & Hanson1980) and make more household and family support trips (Hanson & Hanson;Hanson & Johnston; Rosenbloom 1987) and that women make fewer recre-ational trips (e.g., Hanson & Johnston). Among these findings, genderdifferences in journey-to-work trips is the most studied and is the subject ofongoing debate.

Existing empirical research is unified in finding that women have shortercommutes than men, however, there is disagreement on the factors whichcontribute to the gender differential. The underlying economic theory presentedby White (1977, 1986) and Madden (1981) suggests that commute length isa function of wages, housing prices, income and other demographic vari-ables with wages and housing prices endogenously determined. Because

Transportation 24: 397–419, 1997 1997 Kluwer Academic Publishers. Printed in the Netherlands.

commute length measures home-work separation, women’s shorter commutesmay be due to gender differences in job locations or residential locations orboth. Assuming identical households in terms of preferences and skill level,if workers’ job locations are fixed at the city center, individuals will chooselonger commutes if compensated with lower housing prices. Allowing indi-viduals to have differing skill levels, each wage class will prefer to locatewithin a ring about the city center with higher-income households occupyingmore distant rings (White 1986). This implies that lower-income individualswill tend to have shorter commutes to jobs located in the city center.Alternatively, if workers’ residential locations are fixed, then individuals willchoose longer commutes if compensated with higher wages (for a given typeof employment). If workers decide housing and job locations simultaneously,then lower housing prices and higher wages will be associated with longercommutes.

Drawing on White (1986), given identical education, skill level and income,identical preferences and identical household responsibility, for example, wewould expect men and women to exhibit identical commuting behavior.However, a substantial body of research indicates that men and women differwith respect to labor force characteristics and household commitments. In termsof labor-force characteristics, women earn less, have different educationalbackgrounds and work in different occupations (Blau & Ferber 1992; Madden1981). Hanson and Johnston’s (1985) research suggests that female-dominatedjobs (occupations in which women represent at least 70 percent of the laborforce) may be more evenly distributed across an urban area than male-domi-nated jobs.2 If the distribution of residences is uniform, but the distributionof employment opportunities is not, then differently located employmentopportunities may help explain gender differentials in commute lengths.

With respect to household responsibility, time-budget studies indicate that,although men are doing more housework and child care than in the past (Blau& Ferber 1992), women continue to do the greater share (Blau & Ferber1992; Hochschild 1989; Robinson 1988; Shelton & John 1993), even in two-earner households in which the woman is employed full-time (Hersch &Stratton 1994). For example, Hersch and Stratton’s research indicates that inmarried households with children in which both husband and wife areemployed full-time, wives spend approximately 22 hours per week on house-hold tasks (specifically, cooking, cleaning and doing other work around thehouse) while husbands spend approximately 8 hours per week (p. 122). Thenotion that gender differences in the household division of labor mayexplain commuting differentials has been referred to in the literature as thehousehold responsibility hypothesis (HRH) (Johnston-Anumonwo 1992).Specifically, the HRH states that employed women tend to have greater house-hold and child-care responsibilities and, as a result, face greater time constraints

398

and ultimately choose shorter commutes than employed men; that is, in thecontext of White’s theoretical development, women with heavy home respon-sibilities may not be indifferent between job locations – they may be unwillingto undertake more than some minimum commute length.

Empirical testing of the HRH has produced mixed findings. This papercarefully examines and synthesizes previous research findings and providesnew empirical evidence on the HRH. Section 1 identifies the researchapproaches used to formulate and test the HRH; Section 2 provides in-depthreviews of several of the most often-cited articles; Section 3 tests the HRHusing 1990 National personal Transportation Survey data and Section 4provides a summary of research findings and concludes with recommendationsfor future research.

1. The household responsibility hypothesis

The HRH states that women shoulder greater household responsibility thanmen and, as a result, choose shorter journey-to-work commutes; a greater shareof household tasks and child care implies increased schedule rigidity andless willingness to engage in longer commutes. Ideally, to test the HRH, onewould conduct a survey of employed men and employed women and deter-mine the extent to which women’s shorter commutes are due to familycommitments. Hanson and Pratt’s (1990) survey of employed men and womenin Worcester, Massachusetts, is an example of such an approach and indi-cates that, when considering employment opportunities, men and womenexhibit distinct preferences regarding work-home separation: 80% of surveyedwomen report proximity to home as an important job attribute, compared to69% of men (p. 387). According to Hanson and Pratt, primary responsibilityfor children and their welfare may be an important factor reducing women’scommutes. For example, the authors find that 29.5% of women, but only2.8% of men, attribute their relatively short commutes (ten minutes or less)to family responsibilities (p. 387).

Other researchers have used household characteristics as proxies for thedegree of household responsibility to test the HRH. Commonly used proxiesinclude marital status and the presence, number and age of children in thehousehold. The testable hypotheses implied by this approach include

– when a household consists of a married couple, the level of householdresponsibility is higher (particularly if the couple has children) and theworking woman in such a household will tend to have shorter commutelengths than a similarly situated man;

– single, childless men and single, childless women will tend to have similarcommute lengths;

399

– married women will have greater household responsibility and hence shortercommutes than similarly situated unmarried women;

– women with children will have greater household responsibility and henceshorter commutes than similarly situated women without children.

Existing research examining the HRH using the proxy approach has producedmixed findings.

Supporting the HRH, Preston, McLafferty and Hamilton (1993) report thatthe presence of children is associated with shorter commute times for Black,Hispanic and White women.3 Fagnani’s (1987) research suggests that the effectof children on women’s commute times may depend upon their occupationalstatus. Unskilled employed women displayed commute times which variedinversely with the number of children in the household: the greater the numberof children, the closer the proximity of work place to home. However, thenumber of children did not have any significant effect on the travel times ofmanagement, other professionals and office workers.

Madden (1981) found that, in female-headed households, the number ofchildren reduces the woman’s commute distance, whereas the presence ofchildren under three increases it. For two-earner, married households, Madden’sresearch indicates that children significantly affect women’s commute dis-tances, although results do not indicate whether the children variables increaseor decrease women’s commuting distance. In contrast, Hanson and Johnson(1985) find that the presence of children does not affect the woman’s work tripdistance while White’s (1986) research indicates that the presence of childrenincreases women’s journey to work commute times. In section 2 we explorein greater detail several of these research efforts in order to reconcile thecontradictory findings.

Activity-patterns analysis may be used as an indirect test of the HRH.Employed women undertaking a greater number of household-support tripsor spending more time shopping for household needs than comparably situatedmen, suggests that women shoulder greater household responsibility. Theresearch in unified and consistent with respect to the HRH: women tend tomake more household-support trips and spend more time in household supportactivities than men (Hanson & Hanson 1980; Hanson & Johnson 1985;Niemeier & Morita 1996; Rosenbloom 1987).

Having highlighted the major themes in the literature, it is now useful toexamine several notable research efforts in the last decade in order to syn-thesize existing research with respect to the HRH. Two sets of papers whichemploy the proxy approach to test the HRH are reviewed: papers that basetheir results on comparison of means and papers that formally estimate aregression model based on the previous theoretical discussion. The formerincludes papers by Hanson & Hanson (1980), Hanson & Johnston (1985),

400

Fagnani (1987) and Gordon, Kumar & Richardson (1989), whereas the latterincludes papers by Madden (1981), Singell & Lillydahl (1986), and White(1986). We also review Hanson and Pratt (1990) which explores the HRH usingsurvey data.

2. Literature reviews

Hanson and Hanson (1980)

Hanson and Hanson investigate gender differences in urban travel patternsusing data from the 1971 Uppsala Household Transportation Survey conductedin Uppsala, Sweden. Hanson and Hanson examine the mean number of stopsper person by trip purpose and the average number of hours spent in social,personal business (includes household maintenance), shopping, work and recre-ation activities.

The results, shown in Table 1, lend support to the household responsi-bility hypothesis. Although reporting similar amounts of non-work non-hometravel, the composition of such travel varied appreciably between men andwomen. Women reported making significantly greater numbers of stops forshopping than men. Women spend significantly more time attending to personalbusiness, a category that includes household maintenance, whereas men

401

Table 1. Major findings, Hanson and Hanson (1980).

Supports HRH

Yes A No

1 Women undertook significantly more shopping activities than men (p. 296)

d

2 Women made more visits to the supermarket, the small grocery store, the grocery section and other areas of the department store, the dry cleaners, the drug store and schools (p. 296). d

3 The only trip purposes for which men undertook significantly more travel than women were work and recreation (p. 296). d

4 Women spent significantly more time shopping and attending to personal business; Men spent significantly more time in recreation activities (p. 298). d

A: Finding’s implication for the household responsibility hypothesis is ambiguous.

reported significantly more stops for recreation activities and spent moretime pursuing leisure activities. Additionally, women made more stops at thesupermarket, the small grocery store, the grocery section and other areas ofthe department store, the dry cleaners, the drug store and schools. Based onthese travel pattern differences between employed men and employed women,Hanson and Hanson conclude that women bear greater household supportand child-care responsibility.

Hanson and Johnston (1985)

Hanson and Johnston investigate gender differences in work trip length usingthe 1977 Baltimore Travel Demand Data Set. Hanson and Johnston proposethat, if household responsibility does indeed explain women’s shorter journeyto work, then women with more demanding household responsibilities willtend to have shorter work trips than women with less demanding responsi-bilities. Hanson and Johnston use the presence of pre-school or school-agechildren at home and the presence of another nonworking adult in the house-hold as proxies for household responsibility. The authors suggest that thepresence of children will tend to constrain the work-trip times and distancesof working mothers, while the presence of a nonworking adult should relievethe household responsibilities of working mothers. Specifically, the authorspropose that the home-work separation should diminish as individuals tran-sition from the single-person household category to being married withoutchildren and, finally, to being married with children.

Hanson and Johnston’s major findings are shown in Table 2. The authorsfind that women in households with children do not have shorter commute dis-tances or times, on average, than women in households with no children.Comparing men and women in the same household categories, the data indicatethat the average commute distances of male and female single workers arenot statistically different, whereas the commute distances of men and womenin all other categories (except households with preschool children) arestatistically different. However, gender differences are less prevalent withrespect to commute time: men and women’s commute times are not statisti-cally different, except in households with more than one adult and no children.Hanson and Johnston also consider the effect of the youngest child on women’scommute times and distances. Computing means by household category, theyfind that women’s commute lengths do not vary depending upon the age ofthe youngest child in the household.

Hanson and Johnston then examine household responsibility based onactivity patterns. They suggest that women may seek job locations havingshorter work trips as a result of greater non-work and non-home daily activ-ities. They note that this conjecture is not supported by their data: men and

402

women make, on average, the same number of non-work and non-home stopsper day. However, the purpose of the stops are different: men are more likelythan women to make work-related recreational, visiting and pleasure-relatedstops (16% of men’s total stops) while women made more stops for shopping(26% of their non-work and non-home stops).

In summary, controlling for the number and ages of children and thepresence of a nonworking adult, Hanson and Johnston’s findings are mixed.The authors results refute the HRH when comparing commute distances andtimes among women with various levels of household and child care respon-sibilities. However, the authors find support for the HRH when comparingmen’s and women’s commute distances and trip purposes.

Fagnani (1987)

Fagnani investigated the commuting behavior of employed mothers in Franceusing data from a 1981 survey conducted by the Institut National des EtudesDémographiques and the Centre d’Etudes et des Coûts. Table 3 summarizesFagnani’s major findings. In the provinces and in the metropolitan Paris area(MPA), Fagnani finds that women’s average commutes times tend to declinead the number of children in the household increases. Controlling for occu-pation, Fagnani finds that unskilled and service worker’s commute times are

403

Table 2. Major findings, Hanson and Johnston (1985).

Supports HRH

Yes A No

1 Women in households with children have the same commute distances and times as women in households without children. d

2 Single men and single women have the same average commute distances, whereas men travel significantly longer distances than women in all other household categories except households with pre-school children.

3 Gender differences in time only exist for men and women in households with more than one adult and no children. d

4 Men and women make the same number of non-work stops and non-home stops per day. d

5 Men are more likely than women to make work-related, recre-ational, visiting and pleasure-related stops than are women. Women are more likely to make shopping-related stops than men (26% of their stops as opposed to 16% of men’s).

A: Finding’s implication for the household responsibility hypothesis is ambiguous.

reduced as the number of children in the household increases from two to threeor more.

In contrast, the number of children does not affect commute times of man-agement, other professionals and office workers. Occupation and householdresponsibility appear to interact so that the effect of children on women’scommutes varies by occupation. Fagnani’s research also indicates that in theprovinces, the effect of children on women’s commute times depends uponeducational level. Women with at least a high school diploma have commutetimes that do not vary with the number of children. Average commute timeof women without high school diplomas declines as the number of childrenincreases. To summarize, Fagnani’s analysis indicates that children may be afactor in shortening women’s commute times and the degree to which childrenaffect women’s commute times depends upon location, occupation and edu-cation level.

404

Table 3. Major findings, Fagnani (1987).

Supports HRH

Yes A No

1 Controlling for location: in the metropolitan Paris area (MPA), women’s average commute times significantly decline as the number of children increases; in the provinces, women’s average commute times significantly decline as the number of children in the household increase from two to three or more. d

2 Controlling for occupation, number of children significantly reduces the commute time of unskilled and service workers, whereas number of children does not have any significant effect on the travel times of management, other professionals and office workers. d

3 Controlling for education and location in the provinces, the average commute time of women without a high school diploma decreases when the number of children increases, whereas the average commute time of women with at least a high school diploma does not vary with the number of children in the household. d

4 Combining households in the MPA and provinces, average travel time is significantly reduced for women both with and without diplomas as the number of children in the household increases from two to three or more. d

A: Finding’s implication for the household responsibility hypothesis is ambiguous.

Gordon, Kumar and Richardson (1989)

Gordon et al. use 1977 and 1983 National Personal Transportation Survey(NPTS) data to investigate gender differences in metropolitan travel behavior.To investigate the HRH, the authors examine average commute distances andtimes by gender, marital status, number of workers and presence of childrenand suggest that the data give little support for this hypothesis. As shown inTable 4, Gordon et al. report that, “all female groups tend to have shortercommute distances and times than males, but there are negligible differencesbetween married and unmarried women and between women who have childrenand those who have none,” and the authors assert that the presence of childrendoes not, on average shorten, women’s commute distance or time. Gordon etal. conclude that “family commitments do not appear to force women to takejobs closer to home in order to achieve a better balance between work andhome.” (p. 502) However, Gordon et al. do not conduct tests of significanceleaving the reader uncertain as to whether differences in women’s commutetimes by household category are indeed negligible.

Gordon et al. also consider the effects of gender, presence of children andmarital status on work trip lengths using regression analysis. They regress bothcommute distance and commute time on gender, household income, householdsize occupation and location (central city or suburb). As shown in Table 4,Gordon et al.’s results indicate that men have longer commute distances andtimes than women, that married workers tend to have longer work trips thanunmarried workers and that the presence of children does not affect the worktrip.

The author’s, however, do not include an interaction term to test if the effectof children and marital status on commuting length varies by gender nor dothey estimate separate equations based on gender. Thus, it is difficult toconclude from the reported results that marital status and children do not affectwomen’s commuting lengths. The effect on women may be masked by theinclusion of men in the sample, particularly if these variables have a signif-icant effect on men’s commuting lengths or if men are disproportionatelyrepresented in the sample or both.

Madden (1981)

Madden develops a utility maximization framework within which to considergender differences in commuting. The household chooses housing consump-tion, H, other goods, X, and leisure, L, to maximize household utility, U:

U = f (X, H, L)

405

subject to a household budget constraint. Household income depends on thewage rate which is a function of the travel distance from city center, traveldirection and general labor supply and demand conditions. Household incomeis spent on commodities X, H, L and commuting time. Per unit housing (rental)prices depend on the supply and demand conditions in the housing marketas well as the distance and direction of the residence from the city center.The household optimization problem yields optimal work trip distance as afunction of the wage rate, housing prices, the amount of housing consumedand socioeconomic characteristics. Model predictions are consistent withWhite’s (1986, 1977) theoretical development discussed previously: higherwages and lower housing prices will tend to be associated with longercommutes. The model is extended to the two-earner household by adjustingthe optimal work trip distance function to include the spouse’s wage andnumber of work hours.

Madden’s empirical analysis is based on a sample of households in citieswith populations over 100,000 drawn from the 1976 Wave of the Panel Surveyof Income Dynamics. Mean commute distances for men and women, con-trolling for marital status and presence of children, are reported. Calculationsof t-statistics indicate that men and women’s work commute distances are

406

Table 4. Major findings, Gordon et al. (1989).

Supports HRH

Yes A No

1 All female groups tend to have shorter commute distances and times than males. d

a

2 There are negligible differences in commute distances and times between married and unmarried women. d

a

3 There are negligible differences in commute distances and times between women who have children and those who do not. d

a

Regression analysis4 Males undertake longer work trips (both in terms of

commute distance and time) than females. d

5 Married workers have longer work trips than unmarried (except for commute distance of one-worker households). d

6 The presence of children does not have a significant impact on work trip length. d

A: Finding’s implication for the household responsibility hypothesis is ambiguous.a The authors’ conclusions are considered ambiguous due to the lack of statistical confir-mation.

not significantly different for unmarried and two-earner households. In contrast,men’s commute distances appear to be significantly longer than women’s intwo-earner households with children. The lack of distinction between singlemen and women commute times lends support to the HRH. Similarly, genderdifferences in commute distances in married households with children alsolends support to the HRH.

Madden estimates a two-stage least squares model of optimal commutedistance for men and women with commute distance, wages and per unithousing prices jointly determined. Table 5 presents a summary of the key,often-cited results. Beginning with the first finding, the number of childrenand the presence of children under three years old are reported as having asignificant and positive effect on the commute distance of female-headedhouseholds. However, tabular summaries in Madden’s research (p. 187) indicatea contradictory result: that number of children has a significant and negativeeffect on women’s commute distances. Unfortunately, the data analysis is nolonger available to clarify this contradiction,4 thus limiting the interpretationof Madden’s analysis with respect to female-headed households.

With respect to two-earner families, Madden reports that children signifi-cantly affect the work trip distance of mothers and that sex differences inthe household division of labor are partly responsible for the shorter work tripsof mothers with spouse present. This result is of particular interest since itis often cited in the literature (e.g., Johnston-Anumonwo 1992; Hanson &Johnston 1985). Examining Madden’s reported summary statistics, for womenin two-earner households, the coefficient of infants under three present isinsignificant and positive and the coefficient of log of number of children is

407

Table 5. Major findings, Madden (1981).

Supports HRH

Yes A No

1 Increasing number of children and the presence of children under three years old increase the commute distance of parents in one-earner households (men in traditional families with children and unmarried women with children). d

a

2 Children significantly impact the work trip distance of mothers in two-earner families. Sex differences in the household division of labor with respect to child rearing are partly responsible for the shorter work trips of mothers with spouse present. d

a

A: Finding’s implication for the household responsibility hypothesis is ambiguous.a Findings (1) and (2) are ambiguous due to a contradictory text and table presentation.

insignificant and negative (p. 187). The reported F-statistics indicates thatthe children’s variables as a group significantly impact the commute distanceof women in two-earner households. Contrary to past interpretations, however,the opposite signs on the coefficients precludes interpretation of the direc-tion of their combined effect on commute distance. That is, childrensignificantly affect the commute distances of women in two-earner, marriedhouseholds, however, whether they increase or decrease women’s commutedistances is not determinable.

Singell and Lillydahl (1986)

Singell and Lillydahl hypothesize that the male’s job location in two-earnermarried households propels residential location decisions. Using a two-stageleast squares model, based on Madden’s approach, optimal commute time isestimated with commute time, wages and housing prices jointly determined.With respect to household responsibility, Singell and Lillydahl include a binaryvariable, equal to 1 if children are present in the household and zero other-wise. Using 1980 census data, with a subset of approximately 50,000 adultsin two-earner urban area households, the model is estimated for four categories:full-time working men only, all male workers, full-time working women only,and all female workers. Singell and Lillydahl then separate men and womenwho have moved in the previous year from men and women who have notand re-estimate the model (eliminating the “moved” variable). The authorshypothesize that if residential location is determined by the male head of house-hold’s job location, men should have shorter commute distances than womenin households that have moved.

Table 6 presents the key findings with respect to the HRH. Models con-structed by gender and employment status (p. 126) indicate that the presence

408

Table 6. Major findings, Singell and Lillydahl (1986).

Supports HRH

Yes A No

1 The presence of children in the household decreases the commute time for all men and women. d

2 When separate equations are estimated by gender based on migration status (individuals who have recently moved versus those who have not), the presence of children significantly reduces commute times of women only. d

A: Finding’s implication for the household responsibility hypothesis is ambiguous.

of children is associated with significantly shorter commute times for both menand women. That is, when considering the interaction of employment statusand gender and each of the explanatory variables, the authors do not findsupport for the HRH. If the presence of children reduces both men’s andwomen’s commute times, one cannot conclude that women have shortercommutes because women are primarily responsible for household support andchild care. If the presence of children significantly shortens men’s commutetimes, given that the presence of children is a suitable proxy for householdresponsibility, then it appears men also shoulder enough of the householdresponsibility to shorten their commute times, and hence, household respon-sibility cannot necessarily be viewed as explaining the gender differential.

When considering the interaction of gender and migration status and theother predictor variables, Singell and Lillydahl’s results (p. 127) indicate thatthe presence of children significantly reduces only women’s commute times.The findings indicate that migration status and presence of children do notinteract to affect men’s commute distances. In contrast, employment status andpresence of children do appear to interact and significantly shorten men’scommute distance. Singell and Lillydahl’s findings on the impact of childrenon commute lengths are inconclusive with respect to the HRH.

White 1986

As noted in the introduction, White presents a theoretical framework for ana-lyzing commuting behavior with commute length, housing prices and wagesendogenous. The framework predicts that taste and demographic variableswhich differentiate individuals are “the important explanatory variables deter-mining commuting behavior” (p. 370). Using data for New York City fromthe 1980 Annual Housing Survey, White considers the commuting behaviorof heads of household and estimates a reduced-form commute time equationfor both male and female heads as a function of a number of variables includinghousehold income, the number of children under age 18 in the household,and binary variables indicating the presence of a secondary worker, thepresence of pre-school children and whether the household head is Black oris Spanish.

White’s findings, summarized in Table 7, indicate that each child increasesthe commute time of male heads of household by 2.7 minutes if there is nosecondary worker in the household, but by only 0.9 minutes if there is asecondary worker. However, the presence of young children does not signif-icantly affect the commute times of male heads. With respect to female head,White finds that their work trips are not significantly affected by the numberof children, by the presence of a secondary worker or by both at once. Femaleheads’ work trips are significantly lengthened by the presence of pre-school

409

children: young children increase the female head of household’s work tripby 8 minutes. However, some caution should be exercised when generalizingthese results to the US population. White’s results are for New York Cityand the data have a high representation of women from low-income andminority households.

Hanson and Pratt (1990)

Hanson and Pratt examine the spatial characteristics of occupational segre-gation for women using 1980 data collected in Worchester, Massachusetts. Thisresearch is unique because the traditional travel behavior survey data isaugmented with personal in-home interviews. As noted in Table 8, when theauthors examine journey to work trip characteristics they find that womenworking full-time make significantly shorter trips (16.5 minutes) than men(20.5 minutes). Using the in-home interviews, the authors also find that 69%of women report that “proximity to childcare and/or schools” is an impor-tant job attribute, compared to 45% of men. Further, 84% of women indicatethat job hours which “fit with school schedule and/or childcare arrangements”is an “important” or “very important” job attribute compared to 51% of men.Finally, of those respondents working within a 10 minute commute of home,women were much more likely to attribute their short commutes to a desirefor close proximity to children in case of emergencies (29.5%) than weremen (2.8%). Thus, consistent with the HRH, the Hanson and Pratt researchindicates that women’s short commutes are due to their family commitments.

The authors also consider the travel differences between women employed

410

Table 7. Major findings, White (1986).

Supports HRH

Yes A No

1 Each child increases the commute time of male heads of household by 2.9 minutes if there is no secondary worker in the household, but by only 0.9 minutes if there is a secondary worker. d

2 The presence of young children in the household does not significantly affect male heads commute time. d

3 Female heads’ commute times are not significantly affected by number of children. d

4 The presence of young children increases female heads commute times by 8 minutes. d

A: Finding’s implication for the household responsibility hypothesis is ambiguous.

in traditional female-dominated occupations and those women working in otherindustries; they find that the women employed in full-time “female-dominated”5

occupations have significantly shorter travel times than the women employedin other occupations. It is important to note that, if the HRH is germane,then we would expect to find that those women making shorter commutesalso shoulder greater household responsibility. The personal in-home surveysreveal that indeed women in female-dominated occupations place much greateremphasis on proximity to home than do their counterparts. When asked toevaluate the importance of certain job attributes, 86% of those women workingin female-dominated occupations indicated that proximity to home is impor-tant, or very important, compared to 73% of the women in other types ofemployment. For women working within a 10 minute commute of home, 32.3%of those in female-dominated occupations said that they wanted to “be ableto get home quickly for kids, emergencies, family reasons” compared to 15.4%of those women employed in other occupations. Thus, consistent with the HRH,women in female-dominated occupations travel shorter commute times andalso take on more household responsibility than do women in gender-neutralor male-dominated occupations. In summary, Hanson and Pratt’s researchprovides important insight on women’s journey to work commutes and theHRH. Their data suggests that women’s shorter commutes, when comparedwith men, are at least partially attributable to household responsibility, asevidenced by the need for close proximity of work to childcare and/or schools.The commute and household responsibility association is further corrobo-rated when a greater percentage of women in female-dominated occupations

411

Table 8. Major findings, Hanson and Pratt (1990).

Supports HRH

Yes A No

1 Women have shorter commutes and place greater importance on job proximity to childcare and/or schools than do men. d

2 Women report responsibility for more household tasks than do men. d

3 Women in female-dominated occupations have shorter commutes and place greater importance on job proximity to childcare and/or schools than do women in other occupations. d

4 Women in female-dominated occupations report primary responsibility for more household tasks than do women in other occupations. d

A: Finding’s implication for the household responsibility hypothesis is ambiguous.

412

cite the importance of job proximity to schools and childcare, and make shortercommutes, than do women in other types of occupations.

3. The 1990 NPTS data

In this section 1990 NPTS6 data is used to investigate the HRH. The dataanalyzed are a subset of the full NPTS data consisting of 13,074 work tripsincluding all modes of transportation (95% were made by private vehicle);of these work trips, 7,352 (56%) were made by men and 5,722 (44%) weremade by women. Table 9 reports mean commute distances and times for menand women by marital status and presence of children. T-statistics indicate thatmen’s average commute distances are significantly greater than women’s forall household categories.

Reduced-form equations for commute distance and time are estimated forboth men and women as a function of marital status, education level age,household income, number of adults in the household, and binary variablesindicating whether children are present in the household, whether the house-hold is located in an urban area, and whether the household head is Blackor Hispanic (Table 10). Education is a determinant of wages and thus we wouldexpect commute length and education to be positively correlated: highereducation levels imply higher wages and hence longer commutes. Age mayalso be a determinant of wages, in that age may be a proxy for experience,in which case we expect commute length and age to be positively associ-ated. However, we speculate that age also directly affects commute length: onebecome less willing to engage in long commutes as one ages. Thus, the effectof age on commute length may be positive or negative.

The household urban-area indicator is a determinant of housing prices:households in urban areas are assumed to be higher priced per unit of housing.For a given job location, individuals will accept longer commutes if com-pensated with lower housing prices, thus, we expect households in suburbanareas to have longer commutes. Household income is expected to be positivelyassociated with longer commutes.

Regression results are presented in Table 11. The presence of children hasa very small but significant and negative effect on both women’s commute dis-tances and times. Marital status also has a small but significant and negativeeffect on women’s commute times though has no significant effect on commutedistances. Education has a significant and (very small) positive effect onwomen’s commute length and has no significant effect on men’s commutelengths.

Notice that women in black households commute 4.62 minutes longer thanwomen in white households. Since mode choice is endogenous, the longer

413

Tab

le 9

. M

ean

com

mut

e di

stan

ce a

nd t

imes

for

men

and

wom

en b

y m

arit

al s

tatu

s an

d pr

esen

ce o

f ch

ildr

en (

met

ropo

lita

n ar

eas)

.

Cat

egor

yM

enW

omen

t-va

lues

Hou

seho

ldN

Dis

tanc

eT

ime

ND

ista

nce

Tim

eD

ista

nce

Tim

est

ruct

ure

(mil

es)

(min

utes

)(m

iles

)(m

inut

es)

1S

ingl

e17

6511

.11

20.7

112

4508

.98

19.4

005

.39a

02.1

7b

2M

arri

ed15

1113

.77

23.5

214

6610

.02

19.4

007

.43a

06.4

2a

3M

arri

ed &

chi

ldre

n 27

1913

.49

23.0

622

1909

.32

17.6

411

.38a

11.0

3a

4N

ot m

arri

ed &

chi

ldre

n13

5714

.21

24.4

907

9209

.25

19.2

408

.37a

06.3

7a

Not

e:‘N

ot m

arri

ed’

incl

udes

sin

gle

and

divo

rced

;a

Sig

nifi

cant

at

p=

0.0

1.b

Sig

nifi

cant

at

p=

0.0

5.

commute time of women in black households may be accounted for by theirhigher use of public transportation.7 Age has a (small) significant negativeeffect on women’s commute distances and times and men’s commute distances.Further, as expected, higher household income and suburban living are asso-ciated with longer commutes.

With respect to the HRH, though the effect is small, the 1990 NPTS dataindicate that marital status and children tend to reduce women’s commutelengths. That is, where marital status and presence of children affect women’swork trips, they do so in a way that suggests women choose shorter commuteswhen facing greater household responsibility.

4. Conclusion

Although different samples, commute-length measurements and analysistechniques make comparisons somewhat difficult, it is still useful to summa-rize the implications of recent research on the household responsibilityhypothesis. Referring to Table 12, Fagnani’s, Hanson and Pratt’s and ourresearch tends to support the HRH. Hanson and Hanson’s research suggestsa division of labor in the household that is consistent with the householdresponsibility hypothesis. Madden’s, Singell and Lillydahl’s and Gordon et al.’sresearch are inconclusive with respect to the household responsibility hypoth-esis. The research of Hanson and Johnston, using 1977 data, produces mixed

414

Table 10. Variables used in estimation.

Variable Definition

Children A binary variable equaling 1 if children are present in the household, 0 otherwise.

Married A binary variable equating 1 if married, 0 otherwise.

Educ Respondent’s highest education level: 1–12 = 1st–12th grade; 13 = technical school after high school; 21–24 = 1st–4th year of college; 31 = 1 year of graduate school and 32 = 2 or more years of graduate school.

Age Respondent’s age (17–70).

Hispanic A binary variable equaling 1 if the household reference person is Hispanic, 0 otherwise.

Black A binary variable equaling 1 if the household reference person is black, 0 otherwise.

Income Household family income 1–17 = less than $5,000–over $80,000.

Numadult Number of adults in the household (1–8).

Urbnarea A binary variable equaling 1 if the respondent lives in a non-urban area, 0 otherwise.

findings. They report that children are not important in reducing women’scommute times. Examining trip purpose, Hanson and Johnston concur withHanson and Hanson in finding that women tend to make more shopping stopsas part of their network travel and, hence, appear to be shouldering greaterhousehold responsibility. Whites analysis of New York City data indicatesthat women with greater household responsibility engage in longer commutessuggesting evidence against the HRH.

The contradictory nature of these results may be in part explained by theresearchers data and analysis approach. It is important to note that White’sfindings were based on a highly urbanized, transit-dependent location: NewYork City. Moreover, a large proportion of White’s survey respondents were

415

Table 11. Work trip regressions for men and women; 1990 NPTS data.

Men Women

Distance Time Distance Time(miles) (minutes) (miles) (minutes)

Children 0.92 1.20 –0.66 –1.63[0.55, 0.53] (2.65)a (2.63)a (–2.41)b (–3.94)a

Married 0.11 –0.40 –0.20 –1.43[0.58, 0.64] (0.30) (–0.86) (–0.62) (–2.95)a

Educ –0.03 –0.01 0.09 0.11[19.7, 18.8] (–1.42) (–0.427) (4.22)a (3.60)a

Age –0.04 0.01 –0.08 –0.06[39, 38] (–2.53)b (0.46) (–6.64)a (–3.11)a

Hispanic –0.06 1.03 –0.00 1.94[0.06, 0.05] (–0.082) (1.07) (–0.00) (2.07)b

Black –1.12 2.04 0.62 4.62[0.07, 0.10] (–1.68) (2.32)b (1.38) (6.81)a

Income 0.44 0.54 0.20 0.32[9.8, 9.1] (9.48)a (8.86)a (5.48)a (5.74)a

Numadult –0.32 –0.19 –0.17 –0.20[2, 2] (–1.09) (–0.49) (–0.78) (–0.60)

Urbnarea 3.67 2.32 3.32 1.97[0.20, 0.20] (8.62)a (4.15)a (9.74)a (3.82)a

Intercept 10.50 16.77 9.22 17.16(11.28)a (13.72)a (12.30)a (15.20)a

R2 0.03 0.02 0.04 0.02N 7352 7352 5722 5722

Note: Mean values of independent variables for men and women, respectively, are shown inbrackets below each variable; t-values are in parentheses.a Significant at p = 0.01; b Significant at p = 0.05.

416

Tab

le 1

2.

Sum

mar

y: G

ende

r di

ffer

ence

s in

com

mut

e le

ngth

s.

Res

earc

hD

ata

Com

mut

eS

uppo

rt H

RH

(T

ype

of m

easu

re u

sed)

vari

able

Chi

ldre

nM

arri

age

stat

usT

rip

purp

ose

Fag

nani

(19

87)

1981

Nat

iona

l sa

mpl

e (F

ranc

e)T

ime

Yes

NA

NA

Gor

don

et a

l. (1

989)

1977

/198

3 N

PT

SD

ista

nce

Inco

nclu

sive

Inco

nclu

sive

N

A(U

.S.

Met

ropo

lita

n ar

eas)

Tim

eIn

conc

lusi

veIn

conc

lusi

ve

Han

son

& H

anso

n 19

71 U

ppsa

la T

rans

port

atio

n S

urve

yD

ista

nce

NA

NA

Yes

(198

0)(U

ppsa

la,

Sw

eden

)

Han

son

& J

ohns

ton

1977

Bal

tim

ore

Tra

vel

Dem

and

Dat

a S

etD

ista

nce

No

NA

Yes

(198

5)(B

alti

mor

e S

MS

A)

Tim

eN

oN

A

Han

son

& P

ratt

19

80 C

ensu

s D

ata

Tim

eY

esN

AN

A(1

990)

(Spe

cial

run

s fo

r W

orce

ster

, M

A,

SM

SA

)an

d P

erso

nal

In-H

ome

Inte

rvie

ws

Mad

den

(198

1)19

76 W

ave

Pan

el S

urve

y of

Inc

ome

Dyn

amic

sD

ista

nce

Inco

nclu

sive

NA

NA

(US

Urb

an A

reas

)

Sin

gell

& L

illy

dahl

1980

US

Cen

sus

Tim

eIn

conc

lusi

veN

AN

A(1

986)

(Urb

an A

reas

)

Tur

ner

& N

iem

eier

1990

NP

TS

Dis

tanc

eY

esN

oN

A(1

997)

(US

Met

ropo

lita

n A

reas

)T

ime

Yes

Yes

NA

Whi

te (

1986

)19

80 A

nnua

l H

ousi

ng S

urve

y T

ime

No

NA

NA

(New

Yor

k C

ity)

NA

: H

ouse

hold

res

pons

ibil

ity

mea

sure

not

con

side

red.

minority and low income. Previous research has consistently found thatminority and low income transit users tend to travel longer than non-minori-ties (e.g., Taylor & Wachs 1994). By contrast, the NPTS data used in thisanalysis includes a representative random sample of households within theUS Metropolitan areas. Among the studies we reviewed, Hanson’s and Pratt’sresearch, using activity-based survey data, certainly provides fundamentalinsight into the HRH. Both their research and our NPTS study clearly demon-strate that household and family support activities are an important determinantof women’s travel behavior.

Necessary to the development of equitable and informed transportationpolicy is an understanding of individual travel patterns and the degree to whichdifferent groups exhibit distinct travel patterns. Such an understanding willallow policy makers to better assess the differential impacts of proposedtransportation policies. In this paper, we have sought to highlight some ofthe important findings and examine the methods and results of several promi-nent research efforts. While all of the articles reviewed are informative andunique, not all of the conclusions are well supported. In particular, we saw thatMadden’s estimates do not appear to support the household responsibilityhypothesis, while the research of Gordon et al. does not appear to contradictit.

Further research is necessary if we are to better understand the reasonswhy individuals’ choose the commutes they do and why there exists differ-ences in men’s and women’s travel patterns. In particular, Preston et al. (1993)and White (1986) have highlighted the need to consider the gender differen-tial in commuting and race and ethnicity. If, in fact, household responsibilityshortens women’s commute lengths, does it shorten the commute lengths ofwomen of all races and ethnicity? How might household responsibility andrace and income and other variables interact to affect women’s commutingchoice?

Also fundamental to understanding individual’s commuting behavior isunderstanding how commute choice is made. Research to date has indicatedthat household responsibility and education, income and other labor forcecharacteristics interact to affect commuting behavior. Human-capital theoryand bargaining models suggest that women may accept greater householdresponsibility when they are employed in positions that have less potentialfor advancement and/or less financial reward than their spouse’s positions.Hanson and Pratt (1995) also elaborate on the need for greater explicationof household responsibility, noting that traditional measures may not alwayscapture the perceived obligations. Attempting to model household decisionmaking through the use of human-capital theory or bargaining models mayprovide greater theoretical understanding of the empirical finding that womenappear to choose shorter commutes.

417

Acknowledgment

The authors wish to thank Prof. Susan Hanson, John Crespi and the threeanonymous reviewers for their insightful comments and helpful suggestions.

Notes

1. McLafferty and Preston find that women have significantly shorter commute times thanmen (p = 0.05) when dividing their sample based on gender (p. 7). Controlling for raceand industry of employment, they suggest that, whereas White women have shorter commutetimes than White men, Black and Hispanic women have commute times similar to Blackand Hispanic men, respectively (pp. 10–11).

2. In particular, they find that in the Baltimore SMSA, female-dominated administrative supportjobs are more uniformly distributed than male-dominated manufacturing jobs.

3. Reported ANOVA results indicate that children significantly affect women’s commute times(controlling for marital status and race/ethnicity). However, no data is reported to indicatethe direction of this effect.

4. Based on contact with the author.5. Defined by the authors as those occupations in which women represented at least 70% of

the labor force.6. Source: US Department of Transportation.7. Notice that women in Hispanic households also commute significantly longer distances than

women in non-Hispanic households: 1.96 minutes longer. This is consistent with the findingin the 1990 NPTS data that black and Hispanic women have higher transit use rates than whitewomen (1990 NPTS Demographic Special Reports 1995).

References

Blau FD & Ferber MA (1992) The Economics of Women, Men and Work. Englewood Cliffs:Prentice-Hall.

Blumen O (1994) Gender differences in the journey to work. Urban Geography 15(3): 223–245.Gordon P, Kumar A & Richardson HW (1989) Gender differences in metropolitan travel behavior.

Regional Studies 23(6): 499–510.Fagnani J (1987) Daily commuting time: the stakes for working mothers in France. Transportation

Research Record 1135: 26–30.Hanson S & Hanson P (1980) Gender and urban activity patterns in Uppsala, Sweden.

Geographical Review 70: 291–299.Hanson S & Johnston I (1985) Gender differences in work-trip length: explanations and

implications. Urban Geography 6(3): 193–219.Hanson S & Pratt G (1990) Geographic perspectives on the occupational segregation of women.

National Geographic Research 6(4): 376–399.Hanson S & Pratt G (1995) Gender, Work and Space. London, New York: Routledge, 272 p.Hersch, J & Stratton LS (1994) Housework, wages, and the division of housework time for

employed spouses. American Economic Review 84(2): 120–125.Hochschild A (1989) The Second Shift. New York: Avon Books.Johnston-Anumonwo I (1992) The influence of household type on gender differences in work

trip distance. Professional geographer 44(2): 161–169.

418

Madden JF (1981) Why women work close to home. Urban Studies 18: 181–194.Niemeier DB & Morita JG (1996) Duration of trip-making activities by men and women: a

survival analysis. Transportation 23(4): 353–371.Robinson JP (1988) Who’s doing the housework? American Demographics 10(12): 24–28, 63.Rosenbloom S (1987) The impact of growing children on their parents’ travel behavior: a com-

parative analysis. Transportation Research Record 1135: 17–25.Shelton BA & Daphne J (1993) Does marital status make a difference? Housework among married

and cohabitating men and women. Journal of Family Issue 14(3): 401–420.Singell LD & Lillydahl JH (1986) An empirical analysis of the commute to work patterns of male

and females in two-earner households. Urban Studies 2: 119–129.US Department of Transportation, Federal Highway Administration (1995) Demographic Special

Reports, 1990 NPTS Report Series.White MJ (1977) A model of residential location choice and commuting by men and women.

Journal of Regional Science 17(1): 41–52.White MJ (1986) Sex differences in urban commuting patterns. The American Economic Review

76(2): 368–332.

419