Have the gender differences in commuting been shrinking or persistent? Evidence from two-earner households in the U.S.

Abstract

This study explores gender differences in commute behavior with a focus on two-earner households using data from the 2001, 2009 and 2017 U.S. National Household Travel Surveys (NHTS). To understand whether gender differences are shrinking or persistent in terms of commute distances, we first analyze these differences by assessing descriptive statistics and t-test across multiple population sub-categories and trip purposes. We then employ Seemingly Unrelated Regression (SUR) models on the pooled data in order to analyze the determinants of the household total commute distance and share of women’s commute distance. Our study reveals: (1) gender gap in commute distances has narrowed over the years, however the magnitude of change is small; (2) women continue to have shorter commute distances; (3) commute mode, presence of children, and occupation-related characteristics affect gender gaps in commuting.

Keywords:

• Commute behavior
• gender difference
• two-earner households
• seemingly unrelated regression (SUR)
• U.S. National Household Travel Survey (NHTS)

1. Introduction

The subject of gender differences in travel behavior has become the focus of studies in the transportation, urban planning, and geography fields because patterns in how men and women share responsibilities as well as their choices related to workplace and residential locations may result in different travel outcomes (Gimenez‐Nadal & Molina, 2016; Hu, 2021; Turner & Niemeier, 1997). These studies date back almost 40 years with Rosenbloom (1978) bringing up the importance of studying women’s travel needs and outcomes. In this article, the author (Rosenbloom, 1978) stated that the ways women and their families make decisions about their trips, as well as their residential and work locations will affect future land-use patterns and transportation outcomes.

The literature on the links between gender and commute lengths is inconclusive. Some studies state women travel shorter distances to get to work (Gimenez‐Nadal & Molina, 2016; Rosenbloom, 2004; Shaw et al., 2020), while others find that commute lengths are converging (Doyle & Taylor, 2000; Vandersmissen et al., 2004), and commute durations are comparable for men and women (Gossen & Purvis, 2005). Earlier studies report that the short commute trips for women are due to their social roles (housework, children, and family care) and secondary labor status within a family (Hanson, 2010; MacDonald, 1999). However, many signs indicate that women’s mobility patterns and their underlying factors have changed (e.g., increases in vehicle ownership rates, incomes and having professional occupations) (Fan, 2017; Iwata & Tamada, 2014; Plaut, 2006). Some studies state the short commute trips of women are considered as the outcomes of the constraints on women’s movement, mobility, and opportunity (Crane, 2007; Hanson, 2010; Wheatley, 2013). Others state that a shorter commute for women might be considered a positive, for instance, longer commutes are associated with poor health outcomes such as stress (Roberts et al., 2011; Lancée et al., 2017; Chatterjee et al., 2020 ), in particular for pregnant women (Wang & Yang, 2019). In addition, recent studies have pointed out that the shorter commutes mean lower carbon emissions, and thus, it might be more sustainable (Hanson, 2010; Miralles-Guasch et al., 2016).

While a considerable amount of research has been done on the commute patterns of men and women (for instance, McLafferty (1997), Doyle and Taylor (2000) and Crane (2007)), little or no effort has been made to directly examine gender differences in two-earner households, and its trends over the years.

Women’s entry in the workforce increased the share of two-earner households. In 1967, the percentage of single- and two-earner households among married-couples were 37.3% and 43.6%, respectively. In 2017, while the percentage of single-earner households decreased to 26.2%, the percentage of two-earner households increased to 48.3% (U.S. Department of Labor, 2014, 2019). The reason differences in commute behavior in two-earner households requires special attention is twofold. First, the share of this population group is on the rise (Chakrabarti & Joh, 2019; Sultana, 2005). Second, the resulting behavior in two-earner households is generally the outcome of interrelated decisions based on various factors, such as multiple workplaces, household responsibilities and roles (Deding et al., 2009; Rosenbloom, 2004). The links between these factors need to be disentangled to assess their impacts of future travel outcomes. A few attempts have focused on two-earner households and their commutes (Deding et al., 2009; Plaut, 2006; Sultana, 2005). These studies reveal that workers in two-earner households face constraints when it comes to decision making about residence and workplace. However, they do not consider the reasons that generate different outcomes for men and women, and they do not take into account its changes over time.

This study explores whether the findings of earlier research still hold true when it comes to gender differences in commuting, with a focus on two-earner households. This study seeks answers to the following questions:

1. Did the commute distances of men and women in two-earner households change over time? If so, did these changes result in increases or decreases in gender differences?

2. How do two-earner households compromise on their commute decisions? What factors are associated with the increases (or decreases) in gender differences when it comes to commute distances?

To understand whether gender differences are growing, persistent, or shrinking when it comes to commute distances nationwide, we utilize the detailed individual level data from 2001, 2009 and 2017 U.S. National Household Travel Survey (NHTS). We first analyze these differences by simply accessing the descriptive statistics and t-test results across multiple population sub-categories (e.g., age groups, work status, mode choices, race, household children, home ownership and residential location) and trip purposes. We then employ Seemingly Unrelated Regression (SUR) models on the pooled data (data covering all three survey years) in order to analyze the determinants of household total commute distance and share of women’s commute distance.

This study is organized as follows. The next section reviews the literature on gender differences in travel behavior. In section 3, we present our data, the descriptive statistics, and analysis methods. We also compare commute distances across multiple population sub-categories and trip purposes over the past two decades. In section 4, we present model estimations and interpretations. In section 5, we conclude with important findings, policy implications, and limitations of this study.

2. Literature review

Empirical studies have shown that women’s commute patterns differ substantially from men’s. Women tend to have shorter work-trip (Chidambaram & Scheiner, 2020; Fan, 2017; Hanson, 2010; Hirte & Illmann, 2019; Hu, 2021; Kwan & Kotsev, 2015), but tend to make more non-work trips (Duncan, 2016; Lee et al., 2007; McGuckin et al., 2005). In a detailed study using panel data (American Housing Survey data from 1985 through 2005), Crane (2007) confirms that the average commute distances are different for men and women and concludes there are no big changes when it comes to gender differences in commute patterns across this time period.

We initially give a brief summary of gender differences in the U.S. context because gender roles may vary significantly across countries and cultures. England (1993) states that suburban women’s daily lives in the U.S. were oriented toward housework and caregiving for children and other household members, while men’s stereotypical role was the income-earning breadwinner. The 1960s and 1970s witnessed, the women’s movement known as women’s right movement or women’s liberation movement. Women fought for equal rights, and more personal freedom in all aspects of life such as politics, work, family, and sexuality (Beebeejaun, 2017; Sandercock & Forsyth, 1992). According to Fan’s (2017) research, today’s men and women in the U.S. have more similar roles at work and home than ever before. For instance, there have been significant increases in women’s labor force participation rates, education levels, driver license attainment and vehicle ownership since the 1970s. These continuous changes in social and economic conditions have broadened society’s attitudes toward the multiple roles women play and have had a profound effect on family and household structures such as smaller families and rising divorce rates (England, 1991; Gauchat, 2012).

Existing empirical research is unified women having shorter commutes as compared to men; however, there is mixed and inconclusive evidence on the factors associated with this outcome (Crane, 2007; Gimenez‐Nadal & Molina, 2016; Sermons & Koppelman, 2001). Several explanations have been offered as to why women have shorter commute trips. The primary explanation is related to the need for women to coordinate dual roles as mothers and wage-earners, which means married women with children need to reconcile the demands of family life with those of work (Gimenez‐Nadal & Molina, 2016; Hanson, 2010; MacDonald, 1999; McGuckin et al., 2005; McLafferty, 1997; Turner & Niemeier, 1997). The space-time constraints that women face regarding their dual roles result in shorter commutes (Kwan, 1999).

Several researchers mention the Household Responsibility Hypotheses (HRH) as the main underlying reason for women’ shorter commutes (Turner & Niemeier, 1997). The HRH argues married women have greater household and child-care responsibilities, and hence have shorter commutes than married men as well as single women. Existing research examining the HRH has used household characteristics (i.e., marital status and the presence, number of age of children in households) as proxies for the levels of household responsibilities (Fan, 2017; Gimenez‐Nadal & Molina, 2016; Sermons & Koppelman, 2001; Shelton & John, 1993). For instance, Sermons and Koppelman (2001) findings support the HRH. They argue that the presence of children plays a key role in gender differences when it comes to commute attributes. Gimenez‐Nadal and Molina (2016) examine the relationships between commute durations and time spent for household chores and childcare using the Dutch Time Use Survey for the years 2000 and 2005. They conclude that time spent for household chores and childcare are correlated with women’s shorter work-trips. Fan (2017) examines the effects of household structures and gender differences on travel outcomes using the 2003-2010 American Time Use Survey (ATUS) data, and demonstrates marital status and number of children affect gender differences in commute durations.

Turner and Niemeier (1997) test HRH based on their review of previous studies focusing on gender differences in commute outcomes. They find that although marital status and presence of children tend to reduce women’s commute lengths, not all studies support the HRH. Although the role of women in the family, and their household responsibilities have been discussed as the major factors, some studies indicate that men and women differ with respect to labor force characteristics. That is, lower wages and occupational segregation of women are the reasons for differences in commute patterns. For instance, MacDonald (1999) provides a review of literature on women’s commuting and labor force participation and argues that women make decisions about commute behavior under a different set of constraints than men. She states these constraints include relatively lower wages, greater home responsibilities, and preferences for jobs in sales or service industries, as well as part-time job positions. Assaad and Arntz (2005) argue that differences in labor force participation patterns across men and women explain the gender gap in geographical mobility rates, and women’s commute rates are much lower than those of men. Iwata and Tamada (2014) use Japanese Panel Survey of Consumers (JPSC) data (1993-2002) and find that married and employed women’s commute mainly vary by their wage rates, and in contrast previous studies, they find that the presence of children rather leads to increases in the commute durations of married women. Beck and Hess (2016) research indicates that women’s lower incomes lead to their shorter work trips, and women’s willingness to accept longer commutes appear when their salary increases.

Part of the existing literature focuses on the commute behavior of two-earner households specifically. The empirical evidence in this regard also shows that women have shorter commute distances and durations than men. One of the dominant explanations is that women have more household responsibilities than men, even in two-earner households (Chidambaram & Scheiner, 2020; Deding et al., 2009; Fan, 2017; Lee et al., 2007; Plaut, 2006; Sultana, 2005). For example, Hersch and Stratton (1994) research indicates that in two-earner households where both individuals work full-time, women’s household related work averages around 17 hours per week. This number is only about 7 hours for men. Chidambaram and Scheiner (2020) examine the relationships between intra-household arrangements within two-earner households and the gender gap in commute distances using the German National Time Use Survey. They find that an increase in time spent on household work by men decreases differences in commute distances.

The literature on the interplay between commute patterns among spouses in two-earner households shows that the outcomes of commute decisions differ from those of single-earners (Hirte & Illmann, 2019; Sultana, 2005). For example, Sultana (2005) presents empirical evidence that average commutes in two-earner households are shorter as compared to single-earners, despite facing more constraints in balancing home and work locations. Gender difference in commute patterns is also important when it comes to understanding family commute decisions (Beck & Hess, 2016; Hirte & Illmann, 2019; Plaut, 2006). For instance, Deding et al. (2009) find that two-earner households tend to tradeoff between two workplaces and residential locations to optimize total commute distance. While Deding et al. (2009) findings point to substitution effects, Plaut (2006) finds a complementary relationship between spouses’ commute distances and durations. Based on their analysis, spouses are likely to increase or decrease their commute distances and durations together.

The existing literature presents conflicting evidence when it comes to the relationships between women, household responsibilities and commute patterns. The findings of this study shed light on the underlying factors that result in gender differences in commute distances. Moreover, analyses results based on three years (2001, 2009 and 2017) of national data significantly improve our understanding of gender differences in commuting since we can observe these trends over time.

3. Data, descriptive statistics, and methods

We utilize the detailed individual level data from 2001, 2009 and 2017 U.S. National Household Travel Surveys (NHTS). The NHTS dataset is the source of the nation’s information about travel by US residents in all 50 States. The NHTS datasets include data on individuals, households, their travel behavior, and vehicle ownership (U.S. Department of Transportation, 2017).

The analysis is based on two-earner households where both partners (husband and wife) work in the labor market with paid-work. In this study, we only focus on opposite-sex married-couple household.¹ Table 1 shows the summary statistics for the explanatory variables for three U.S. NHTS datasets used in this study. These give us a sample size of 10,121 households in 2001, 19,795 households in 2009, and 20,260 household in 2017. The total sample used in this study becomes 50,176 two-earner households (50,176 men and 50,176 women, respectively).

Table 1. Summary statistics for the explanatory variables by year and gender.

Independent Variables		2001		2009		2017
		Mean^a (Std. Dev.)		Mean^a (Std. Dev.)		Mean^a (Std. Dev.)
		Men	Women	Men	Women	Men	Women
Age (year)		44.5	42.5	49.5	47.4	46.6	44.6
		(10.2)	(9.8)	(10.3)	(10.0)	(12.2)	(11.9)
Job types	Sales/service	20.2%	21.6%	22.0%	20.1%	17.8%	16.7%
	Professional	46.0%	46.6%	52.3%	53.0%	56.6%	60.9%
	Clerical/administrative	2.8%	25.2%	2.9%	23.3%	3.3%	19.2%
	Manufacturing/Farming	31.0%	6.6%	22.8%	3.6%	22.3%	3.2%
Work status	Full-time	95.2%	73.1%	92.8%	76.0%	92.9%	79.7%
	Part-time	4.8%	26.9%	7.2%	24.0%	7.1%	20.3%
Commute mode	Walk/bicycle	2.3%	2.1%	1.7%	1.1%	2.6%	1.9%
	Public transit	1.5%	1.5%	1.7%	1.4%	2.8%	2.8%
	Automobile	96.2%	96.4%	96.6%	97.4%	94.6%	95.3%
		Households		Households		Households
Household race	Non-Hispanic White	91.3%		86.4%		80.5%
	Black/African American	2.3%		3.8%		4.4%
	Others (Asian, American Indian, Hispanics and others)	6.4%		9.8%		15.1%
Number of adults		2.2 (0.5)		2.2 (0.5)		2.2 (0.5)
Number of children (age < 16)		1.1 (1.3)		0.7 (1.0)		0.7 (1.0)
Household vehicle count		2.6 (1.1)		2.7 (1.1)		2.6 (1.1)
Household Income ($)		92,278.7 (32601.2)		90,453.8 (26260.3)		108,716.2 (51914.6)
House Ownership	Own	90.5%		94.7%		83.0%
	Rent	9.5%		5.3%		17.0%
Employment density of home location		1059.0 (1292.7)		824.2 (1214.2)		1378.7 (1501.1)
Residential Location	Urban	4.5%		7.5%		11.3%
	Suburban	41.6%		40.6%		43.1%
	Rural	53.9%		51.9%		45.6%
MSA	1 million/more with rail	15.2%		17.3%		15.8%
	Others	84.8%		82.7%		84.2%
Number of Observations (Households)		10,121		19,795		20,260

^a Mean or Percentage. For continuous variables, we report the mean values. Standard deviations are reported in parenthesis. We report the percentages for categorical (dummy) variables.

For both men and women, the dominant commute mode continues to be the automobile. This includes auto use as a driver and a passenger. There are slight increases in the usage of public transportation over the last two decades. In general, we see an increase in individuals holding professional-jobs as compared to service-related ones for both genders in our data samples. The number of children (aged less than 16) have gradually decreased between 2001 and 2017, indicating shrinking household sizes for two-earner families in our data samples. Based on these data, we observe a higher percentage of two-earner families living in urban areas in 2017.

For household income, we use Consumer Price Index (CPI) inflation rates for adjusting monetary values, since we use pooled data in our empirical models. The CPI inflation rates between 2001 and 2017 and between 2009 and 2017 are 38.69% and 15.01%, respectively (US Bureau of Labor Statistics). We report the adjusted values of household income in Table 1. Household income in two-earner households has slightly increased between 2001 and 2017.

One-way ANOVA tests are conducted to examine differences in commute distances in all three survey years (Table 2). Commute distances used in this study refer to one-way road network distances between respondents’ home (origin) and work locations (destination) generated by the Google Map API (U.S. Department of Transportation, 2017).

Table 2. One-way ANOVA test results.

Dependent Variable	Average Commute Distance (miles)
	2001	2009	2017	N	F-value
Men	14.0	15.6	15.2	50,176	39.15***
Women	10.5	11.9	12.3	50,176	90.82***

Note:.

* p < 0.10.

** p < 0.05.

*** p < 0.01.

The results show statistically significant differences in commute distances across survey years. In 2001, the average commute distances for men and women were 14.0 and 10.5 miles, respectively, indicating a 3.5 mile difference. In 2017, the average commute distances increased to 15.2 miles for men and 12.3 miles for women. We find that average commute distances have increased from 2001 to 2017, with women having shorter commute distances in all three survey years.

We conduct t-tests to examine the differences in commute distances between men and women. Table 3 presents the average commute distances for men and women in two-earner households using different population sub-categories based on age groups, work status, mode choices, race, household children, home ownership and residential location.

Table 3. T-test results.

	Average Commute Distance (miles)
	2001		2009		2017		% Change^a
	Men	Women	Men	Women	Men	Women	Men	Women
Age groups
HH^b aged 15 to 24	14.04	11.01	15.59	11.66	12.23	11.09	−12.87	0.71
HH aged 25 to 54	14.33	10.69	16.24	12.40	15.38	12.43	7.33	16.25
HH aged 55 to 64	12.31	9.22	14.51	11.19	15.48	12.38	25.73	34.33
HH aged 65 +	8.93	7.74	11.44	8.95	12.61	11.31	41.18	46.15
Work status
Both full-time	14.25	11.19	15.88	12.59	15.60	12.94	9.45	15.61
Men: full-time, Women: part-time	14.03	8.81	15.92	10.13	14.95	10.08	6.55	14.37
Men: part-time, Women: full-time	10.16	9.25	12.03	12.06	11.97	12.09	17.76	30.66
Both part-time	10.27	7.82	11.63	8.85	10.96	10.65	6.73	36.22
Mode choice sets
Both automobile	14.12	10.54	15.66	12.03	15.44	12.44	9.39	18.05
One^c: auto, Other^c: transit	21.02	15.47	24.02	15.58	18.99	16.12	−9.66	4.15
One^c: auto, Other^c: walk/ bicycle	6.88	5.90	6.03	6.12	7.27	6.63	5.73	12.39
One^c: transit, Other^c: walk/bicycle	10.65	1.74	10.19	5.33	6.98	6.86	−34.44	294.29
Both transit or Both walk/bicycle	10.51	8.88	7.57	6.56	9.65	9.94	−8.20	12.00
HH race
Non-Hispanic White	13.90	10.34	15.54	11.88	15.19	12.27	9.26	18.69
Black or African American	15.68	13.50	16.36	13.21	15.88	14.29	1.28	5.81
Others	14.85	11.39	15.85	12.01	15.06	12.10	1.43	6.27
Household children
No children	13.48	10.48	14.85	11.83	14.83	12.49	9.99	19.13
Youngest child 0-5	14.56	11.30	16.26	13.20	15.06	12.24	3.47	8.34
Youngest child 6-15	14.31	10.03	16.47	11.69	16.06	12.06	12.26	20.28
Youngest child 16-21	14.09	9.99	16.14	11.29	15.97	12.12	13.36	21.39
Home ownership
Own	14.06	10.60	15.74	12.01	15.72	12.59	11.78	18.70
Rent	13.45	9.25	13.08	10.87	12.69	11.09	−5.65	19.92
Residential location
Urban	12.18	8.68	12.17	9.26	12.22	10.01	0.33	15.33
Suburban	11.91	8.57	13.64	10.33	13.54	10.79	13.74	25.83
Town and Rural	15.77	12.09	17.62	13.60	17.51	14.37	11.04	18.86

Note: Mean differences by gender which are statistically significant at 95% level are presented in bold (e.g., t-test results in each year).

^a Percentage change from 2001 to 2017.

^b Household Head.

^c Men or Women. For example, “One: auto, Other: walk/bicycle” means that if men use auto, women use walk/bicycle, or vice versa.

Overall, the two-sample mean tests in each survey year indicate that men have longer commute distances for most of the population sub-categories. We find gender gap exists in all survey years for households aged between 25 and 64. We consider four work status based on full- and part-time status. We find that the gender gap is more likely to increase if the man holds a full-time job, while woman has a part-time one. We observe gender differences among couples with different travel mode preferences, if one or both use the automobile.

Gender gaps exist among couples with or without children. Gender gap is more likely to increase if the couple has a child between the ages of 6 to 15 (school-aged children). Homeowners are likely to have longer commute distances as compared to renters, because renters may be more flexible in terms of job-related moves (Plaut, 2006). We observe that the differences in commute distances are more likely to increase if two-earner households own a house. We find gender gap exists regardless of residential location (urban, suburban, and rural areas). As expected, both men and women living in urban neighborhoods have shorter commute distances as compared to those who living in rural neighborhoods.

The differences in travel distances may be attributed to the different roles/responsibilities of men and women, and their trip purposes. Therefore, we examine various trip purposes, their distances, and conduct t-tests to examine the gender differences across these categories (Table 4). These average distances represent the average daily miles traveled per person, and missing values are discarded when calculating an average. Table 4 also shows the percentage of trips per purpose and gender.

Table 4. T-test results of travel distances/the percentage of trips per purpose.

		Average Travel Distance (miles)
		2001		2009		2017		% Change^a
		Men	Women	Men	Women	Men	Women	Men	Women
Work trips	Distance	14.0	10.5	15.6	11.9	15.2	12.3	8.55	17.73
	% of trips^b	100.0	100.0	100.0	100.0	100.0	100.0
School/Daycare/Religious	Distance	8.16	7.89	8.12	8.04	9.65	8.85	18.34	12.21
	% of trips	5.7	8.0	6.2	8.3	5.4	6.8
Medical/Dental services	Distance	10.38	10.14	11.74	11.16	11.10	10.37	7.02	2.25
	% of trips	3.1	5.3	3.5	5.7	2.6	4.1
Shopping/Family Errands	Distance	15.54	15.23	14.49	15.03	13.73	13.89	−11.68	−8.75
	% of trips^b	63.3	71.1	39.5	46.9	54.4	58.7
Social/Recreational activity	Distance	14.76	12.62	13.32	11.69	12.37	11.87	−16.15	−5.92
	% of trips	28.8	30.0	54.0	57.7	25.2	26.3
Something else	Distance	13.67	10.89	22.62	20.57	14.30	12.54	4.62	15.23
	% of trips	24.6	20.2	29.1	29.1	9.5	8.0

Note: Mean differences by gender that are statistically significant at 95% level are presented in bold at each year.

^a Percentage change from 2001 to 2017.

^b The percentage of trips per purpose were calculated at a rate of the total number of individuals to the number of individuals who conduct the trip. For example, all individuals have work trips, thus, the percentage of work trips is 100%. 54.4% of men had shopping/Family Errands trips in 2017.

NHTS dataset classifies trip purposes into various categories: work trips, school/daycare/religious activity related trips, medical/dental services, shopping/family errands, social/recreational activity related trips, and others. We find gender gap exists in all survey years for work trip distances. We observe gender gaps do not exist in school/daycare/religious activity related trips, medical/dental services, and others in all survey years. In 2009, we find while women travel longer distances for shopping and family errands, men travel longer for social and recreational activity.

The first research question posed in this study was whether gender differences in commute distances are growing or shrinking. Based on our descriptive analyses and tests, we find gender differences still exist in general. However, while commute distances have increased for both men and women, the increase have been on the larger side for women, and therefore gender gaps have narrowed.

3.1. Seemingly unrelated regression (SUR) models

The second research question of this study is how dual-earner households compromise on their commute decisions. As noted above, for two-earner households, commute decisions potentially are not determined by separate individuals, but rather are made jointly at the household level. Factors associated with commute distances overall may affect women and men in different ways, and therefore either increase or decrease the resulting gender differences. In order to understand the factors associated with household commute distances, and their contributions to gender differences, we estimate SUR models.

The SUR model has several regression equations ($m$ regression equations) which have its own dependent variable and potentially different sets of independent variables (Zellner, 1962). Although the equations seem unrelated, they are related to one another through the correlation in the error. The error terms are assumed to be correlated across the equations. In general, the SUR model equation can be written as $$y_{\mathcal{i}}={\beta }_{\mathcal{i}}{X}_{\mathcal{i}}+{\epsilon }_{\mathcal{i}}\mathcal{ }(\mathcal{i}=1,\dots ,m)$$ where $\mathcal{i}$ represents the equation number, each equation has a set of explanatory variables $X_{\mathcal{i}},$ a vector of parameters ${\beta }_{\mathcal{i}},$ and an error term ${\epsilon }_{\mathcal{i}}.$

With $m$ samples of such relationship stacked, we can stack the $m$ equations into a SUR model as follows: $$\left[\begin{array}{c}{y}_1\\ y_2\\ .\\ .\\ .\\ y_m\end{array}\right]=\left[\begin{array}{c}\begin{array}{ccc}{x}_1& 0···& 0\end{array}\\ \begin{array}{ccc}⋮& \ddots & ⋮\end{array}\\ \begin{array}{ccc}0& ···& x_n\end{array}\end{array}\right]\left[\begin{array}{c}{\beta }_1\\ {\beta }_2\\ .\\ .\\ .\\ {\beta }_m\end{array}\right]+\left[\begin{array}{c}{\epsilon }_1\\ {\epsilon }_2\\ .\\ .\\ .\\ {\epsilon }_m\end{array}\right]=\beta X+\epsilon$$

The SUR models allow for estimation of multiple models simultaneously while accounting for the correlated errors. With SUR models, numerous studies have estimated total cost and derived input cost share functions in areas such as bus transit, energy consumption, and VMT (Akar & Guldmann, 2012). For instance, Akar and Guldmann (2012) examine the determinants of total VMT in two-vehicle households and how this VMT is distributed among the two vehicles (new vehicle and older vehicle) using a SUR model with two equations. In their study, the first equation is to estimate total VMT, and the second equation is the share equation, explaining the percentage use of the first vehicle.

In our study, the second research question focuses on the determinants of total commute distances in two-earner households, and how this total distance is distributed between men and women. An SUR model with two equations is estimated. The first equation estimates total household commute distance, and the second equation is the share equation, explaining the percentage of women’s share in total household commute distance. The first equation is as follows: $$\log \left(\mathit{Total}\mathrm{ }\mathit{Distance}\right(\mathit{man}+\mathit{woman}\left)\right)=f\left\{\begin{array}{c}h\mathit{ousehold} \mathit{characteristics}, \\ a\mathit{ttributes}\mathit{\text{ of }}\mathit{residential} \mathit{location}\end{array}\right\}$$

The second equation is the share equation, explaining the percentage of women’s share in total household commute distance. Since women’s share in total household commute distance is lower than that of men, increases in this share indicate shrinking gender gap. If Dist_woman is the total commute distance of women, then that share is defined as follows: $${\mathrm{Woman}}_{\mathrm{d}}=100\mathrm{*}{\mathrm{Dist}}_{\mathrm{woman}}/{\mathrm{Dist}}_{\mathrm{total}}$$

The commute distance share of men, ${\mathrm{Man}}_{\mathrm{d}},$ is computed as follows: $${\mathrm{Man}}_{\mathrm{d}}=100-{\mathrm{Woman}}_{\mathrm{d}}$$

The share of men is simply calculated by subtracting the share women from the total (100%), there is no need for estimating a share equation for men.

Therefore, the second equation of the model is then $$\log \left({\mathit{Woman}}_d\right)=f\left\{\begin{array}{c}c\mathit{haracteristics}\mathit{\text{ of }}\mathit{men}, \mathit{characteristics}\mathit{\text{ of }}\mathit{women},\\ h\mathit{ousehold} \mathit{characteristics}, \\ a\mathit{ttributes}\mathit{\text{ of }}\mathit{residentail} \mathit{location}\end{array}\right\}$$

4. Model results and discussion

In order to analyze the determinants of total commute distances in two-earner households, and how this total distance is distributed between men and women, we employ Seemingly Unrelated Regression (SUR) models on pooled data (covering all three survey years).

In this study, to determine which independent variables contribute significantly to explaining the variability in the dependent variables, we include variables based on theory and previous studies. The resulting sample includes 50,176 households.

An often-used specification test for the SUR model is the Breusch-Pagan test of independent error. The Breusch-Pagan test is used to test the assumption that the errors across equations are contemporaneously correlated. The null hypothesis is no contemporaneous correlation. In this study, we find that the correlation of the residual in the total commute distance and share of women’s distance equations is −0.155 and $x^2$ is 1200.07 (p < 0.001), thus we reject the hypothesis that this correlation is zero and conclude these two equations are correlated. We report the SUR model results in Table 5.

Table 5. SUR results for total distance and share of women’s distance.

	Log (Total commute distance)		Log (Share of women’s distance)
	Coeff.	t-value	Coeff.	t-value
Household Race (reference: Others)
Non-Hispanic White	−0.133	−10.67***	−0.007	−0.68
African American	0.131	5.82***	0.039	1.97**
Number of adults	−0.018	−2.35**	−0.019	−2.83***
Number of children (age ≤ 16)	0.022	6.26***	−0.025	−7.15***
Number of HH vehicles	0.041	10.44***	−0.015	−4.12***
Log (Household income)	0.266	32.67***	0.001	0.16
Home Ownership (reference: Rent)
Own	0.028	2.17**	0.010	0.84
Log (Employment density)	−0.095	−27.62***	−0.018	−6.00***
Residential Location (reference: Suburban)
Urban	−0.109	−7.20***	0.029	2.13**
Rural	0.101	8.77***	−0.024	−2.36**
MSA (reference: Others)
MSA of 1 million or more, with rail	0.366	33.28***	−0.093	−9.47***
Survey Year (reference: Y2001)
Y2009	0.124	11.76***	0.028	2.98***
Y2017	0.167	15.85***	0.035	3.73***
Man’s Age			−0.005	−1.43
Man’s Age^2			0.000	2.05**
Man’s Occupation (reference: Manufacturing or Farming)
Sales or service			0.046	4.58***
Professional			0.002	0.26
Clerical or administrative			0.057	2.86***
Man’s Work Status (reference: Full-time)
Part-time			0.096	6.84***
Woman’s Age			−0.013	−3.47***
Woman’s Age^2			0.000	2.37**
Woman’s Occupation (reference: Manufacturing or Farming)
Sales or service			−0.151	−8.29***
Professional			−0.010	−0.55
Clerical or administrative			−0.065	−3.60***
Woman’s Work Status (reference: Full-time)
Part-time			−0.138	−17.02***
Man’s Commute Mode (reference: Auto)
Walk or Bicycle			0.632	27.05***
Public transit			−0.287	−11.88***
Woman’s Commute Mode (reference: Auto)
Walk or Bicycle			−1.515	−56.87***
Public transit			0.342	13.73***
Constant	0.281	2.99	−0.249	−2.53
N	50,176		50,176

Note:

* p < 0.10.

** p < 0.05.

*** p < 0.01. System-weighted R²=.097, R²=.084, respectively.

The first model explains the factors associated with total commute distances. All variables related to the household and residential location characteristics reveal significant relationships with total commute distances of two-earner households. While being Non-Hispanic White decreases total commute distances, being African American has the opposite effect, as compared to other race category (reference group). As expected, increases in number of children, household vehicles, and household incomes are likely to lead to longer total commute distances. In particular, a unit increase in the number of children leads to a 2.2 percent increase in total commute distance.

The residential location-related variables include employment density, neighborhood categories (urban, suburban, and rural), and metropolitan statistical area (MSA) designation. We find that high employment density in residential locations, and urban neighborhoods are negatively related to total commute distances. For instance, a unit percent increase in employment density leads to a 0.095 percent decrease in total commute distance, on average. Two-earner households living in urban neighborhoods are likely to have shorter commute distances than those living in suburban neighborhoods (reference group). These results indicate, having employment centers in close proximity to residential locations shorten commute distances. In contrast, two-earner households living in rural neighborhoods are likely to have longer commute distances than those living in suburban neighborhoods (reference group). In addition, total commute distances are more likely to increase if two-earner households live in metropolitan areas of 1 million or more with rail systems, than those living in other areas.

Since we use pooled data (data from 2001, 2009 and 2017 U.S. NHTS), we include year dummies to account for their effects on total commute distances and gender differences. All else being equal, two-earner household commute distances are 13.2 percent longer in 2009 and 18.2 percent longer in 2017 as compared to 2001 (reference group). We have checked for the presence of interaction effects between the years and other variables, however we did not find any statistically significant associations.

The second model focuses on the share of women’s commute distances at the household level. Since women tend to have shorter commute distances in general, those that increase this share contribute to shrinking the gender gap. We include characteristics of men and women, such as age, occupations, work status, and commute mode, as well as household characteristics and attributes of residential location. We find that most characteristics related to men and women have significant effects on the gender gap in commute distances. The share of women’s commute distances are likely to increase if men work in the sales/service industries or hole clerical/administrative jobs or part-time positions. On the other hand, the share of women’s commute distances are likely to decrease if women work in the sales/service industries or hold clerical/administrative jobs or part-time positions. This may be due to the fact that sales or sales/service or clerical/administrative jobs are more evenly distributed across the area (Assaad & Arntz, 2005).

In general, those who walk, or bicycle cover shorter distances (Shaw et al., 2020). The share of women’s commute distance is likely to increase if men walk or ride a bicycle when they go to work. On the other hand, the share of women’s commute distance is likely to decrease if men use public transit. The results are similar when we look at women’s commute modes. The share of women’s commute distance is likely to decrease if they walk or ride a bicycle, and vice versa.

The variables related to the characteristics of a household reveal that while increases in the number of children are likely to lead to longer commute distances for two-earner households in general, increases in the number of adults and number of children are likely to decrease the share of women’s commute distance. This finding is consistent with the previous studies that support the Household Responsibility Hypothesis (i.e., employed women carry greater household and childcare responsibilities than men and single women, and therefore have shorter commutes) (Gimenez‐Nadal & Molina, 2016; Sermons & Koppelman, 2001).

The variables related to the attributes of the residential location are statistically significant when it comes to women’s share in commute distances. We find that higher employment densities or rural neighborhoods (as compared to suburban neighborhoods) and living in MSAs with 1 million or more with rail systems decrease women’s share in total commute distances.

We find that the coefficients of the year dummies for 2009 and 2017 are positive and statistically significant. All else being equal, the shares of women’s distances are 2.8 percent more in 2009 and 3.6 percent more in 2017 as compared to 2001 (reference group).

5. Conclusions

Researchers in the transportation, urban planning and geography fields studied magnitude of, and reasons for gender differences in commute outcomes. However, the evidence of gender differences in commute behavior has been inconclusive, and the existing literature has not paid much attention to understanding the extent of these gaps and their determinants over time (Chidambaram & Scheiner, 2020; Gimenez‐Nadal & Molina, 2016). While women’s mobility patterns and their underlying factors have changed (e.g., increases in commute distances, having full-time and professional occupations), our findings reveal gender gaps persist, although they have narrowed over the years.

This study focused on gender differences in commute distances and durations with a focus on two-earner households. Using data from the 2001, 2009 and 2017 U.S. National Household Travel Survey (NHTS), we compared commute distances across men and women to understand whether gender differences are growing, shrinking, or persistent over the years. We also analyzed the determinants of these commute distances and gender differences considering the interrelationship between men and women using Seemingly Unrelated Regression (SUR) models.

The descriptive analyses show that the gender gaps in terms of commute distances have narrowed over the years; however, the magnitude of change is small. We also observe that some of the underlying factors have changed. We find that the percentages of women holding professional jobs and full-time position have increased. Our empirical models reveal the varying effects of commute related variables on men and women.

Key findings from our empirical analyses are as follows: 1) Women continue to have shorter commute distances; 2) There are gender gaps if both men and women use automobile, or if one uses automobile and the other uses public transit. We did not observe gender gaps if both use non-motorized modes or public transit; 3) The gender gap is likely to increase if the couple has a child between the ages of 6 to 15; 4) High employment density in close proximity leads to shorter commutes for both genders; and 5) Individual’s (both men and women) occupation related characteristics (sector and full-/part-time status) have statistically significant effects on the share of women’s commute distances.

Our empirical results show that women still have shorter commute distances especially in households with children between the ages of 6 to 15. As MacDonald (1999) argues, women’s shorter work trips can be explained by the women’s dual duties as mothers and earners. Even though women contribute to household incomes just as men do, women still hold more household and child-care responsibilities. Disproportionate burden of household responsibilities on women undoubtedly makes it difficult for them to work further away from home, and this may in return limit their access to a wider range of occupational opportunities (Crane, 2007; Rosenbloom, 2004; Wheatley, 2013). Therefore, having employment opportunities within close proximity may be crucial for increasing and sustaining women’s participation in different sectors. As argued by Crane (2007), understanding the gender differences in commute lengths may help to predict residential and workplace location preferences. Our study proves relevant to this issue.

Our findings can be useful for urban planners and policymakers as they look into the determinants of commute distances, and search for strategies to reduce these distances. The models we estimate control for detailed household and individual characteristics, while revealing significant land-use effects. Not surprising, we find that increased employment density at one’s residential location leads to shorter commute distances for both men and women. That is, a mixture of residential and nonresidential land uses makes it possible to increase employment access, possibly benefiting both spouses in dual-earner households. These can be linked to the recent concept of “the 15-minute city” which is defined by its ability to provide access to all human needs by walking or bicycling for a quarter hour or less (Pozoukidou & Chatziyiannaki, 2021).

While gender gaps in commuting durations and distances are generally attributed to equity issues (inequality in access to opportunities) (Hanson, 2010), there are different points of view as well. Recently some studies have pointed out women’s shorter commutes are more sustainable since shorter commutes have lower carbon emission and lead to less commuting stress (Chatterjee et al., 2020; Lancée et al., 2017; Miralles-Guasch et al., 2016; Roberts et al., 2011). For instance, Wang and Yang (2019) research examines the relationship between long commutes and maternal stress during pregnancy and find that a 10 miles increase in commute distance is associated with increases in the probabilities of low birth weight and intrauterine growth restriction by 0.9 and 0.6 percentage points, respectively.

As a limitation, we rely on national-level data (i.e., U.S. NHTS), without detailed residential or employment location information. Therefore, we cannot capture neighborhood effects such as accessibility, amenities, and land-use characteristics. Future research could use micro-level data to achieve a better understanding on how neighborhood factors affect gender differences in commute behavior. In this study, we use pooled data (data from 2001, 2009 and 2017 U.S. NHTS), but NHTS does not offer individual’s income in all survey years and individual’s race except for 2017. The commute lengths may vary between groups of women, for instance, between higher-income women and lower-income women or between White and African American women. In terms of access to opportunities, it is very important to discuss these differences between diverse population subgroups. In addition, our findings reveal women continue to have shorter commute distances. However, we cannot determine whether shorter work trips represent preferences or other constraints (e.g., household responsibilities and spatial configurations), because U.S. NHTS datasets do not include such detailed information. This can raise specific questions for transportation planning and policy in the future.

Notes

1 The percentages of same-sex married couples are very small in U.S. NHTS datasets (0.19% in 2001, 0.31% in 2009, and 2.35% in 2017).