The Living,Moving and Travel Behaviour of the Growing American Solo: Implications for Cities

Abstract

Between 1930 and 2010 the share of single-person, or solo, households in the US increased from 6 per cent to almost 28 per cent, whereas the share of married-couple households decreased from 79 per cent to 49 per cent. Yet solo households have received little attention in urban planning and transport research. Given the significant increase of solo households in US cities, this study identifies the distinctive dwelling, moving and travel characteristics of the American solo households, and examines the reasons for their attraction to cities. It uses historical data from census Public Use Microdata Samples and recent national data from the American Housing Survey and the National Household Travel Survey. Descriptive statistics, basic statistical tests, binary logit models and Heckman sample selection models are used to examine various relationships. Some of the transport-related and environmental implications of the findings are discussed.

Introduction

In his recent book, Going Solo: The Extraordinary Rise and Surprising Appeal to Living Alone, sociologist Eric Klinenberg (2012) brings to the fore the growing tendency among American adults to live in single-person, or solo, households. For the first time in centuries, Klinenberg mentions, the majority of American adults are now single and approximately 31 million adults, or one of every seven, live in a solo household. From Klinenberg’s contention that the US is simply following a pattern experienced for a longer duration by other advanced countries like Japan, Germany, Canada, France and Australia, it can be anticipated that the growth of solo households in the US will continue in the foreseeable future.

Given the significant growth of solo households in the US over the past decades and its potential impact on urban areas in the future, this paper seeks to provide a better understanding about the distinctive characteristics of the American solo households. It analyses the potential reasons for their growth in cities, compares their travel and moving patterns with adults from married-couple households and discusses the implications of their growth for urban America. The paper is prepared with the premise that understanding the characteristics of the solo households and the reasons for their distinctive living, travel and moving patterns is beneficial for urban planners, transport planners and policy-makers.

Understanding the growth of solo households in US cities is important for several reasons. First, given the massive loss of population in central cities over the past decades, the growth of any type of household and jobs in cities could be perceived as a positive sign. Between 1950 and 1990, the proportion of metropolitan residents in central cities decreased from 57 per cent to 37 per cent, whereas the proportion of jobs decreased from 70 per cent to 45 per cent (Mieszkowski and Mills, 1993). Therefore, the apparent attractiveness of cities for solo households, especially working adults from solo households, provides a glimmer of hope about potential recovery of US cities. Secondly, as land is more compactly developed in central cities than suburban areas, continued growth of solo households in central cities can be expected to create a more sustainable metropolitan land development pattern in the future. Thirdly, solo households exhibit travel patterns that are more conducive to environmentally sustainable cities than the travel patterns of two-earner households. As shown in this paper and other studies, solo households own fewer cars and workers from solo households commute shorter distances, spend less time commuting and use automobiles less frequently. Considering that the increase in the use of the automobile has often been attributed to the growth of two-earner households (Cervero, 1989; Downs, 1992), the replacement of two-earner households by solo households could potentially help to reduce society’s dependence on the automobile and promote the use of public transit.

This paper focuses on the US as a whole instead of specific regions or cities of the country. Hence, data from several sources are used at the national level and city-specific comparisons are avoided. All comparisons in the paper are made between adult persons from solo households and married-couple households. By definition, married couples are those with two spouses living together, whereas solo households are composed of only one adult. Single parents are not included in any comparison because few people potentially make a choice between remaining solo and getting married with the expectation of becoming a single parent. Although many elderly persons live in solo households, the paper places a greater emphasis on working adults because of their potential impact on the economic wellbeing of cities today and in the future.

Data and Methods

This study uses data from the 2009 National Household Travel Survey (NHTS), the 2009 American Housing Survey (AHS) national sample, and Public Use Microdata Sample (PUMS) from several censuses and the 2010 American Community Survey (ACS). For analytical purposes, it uses descriptive statistics, basic statistical tests, binary logit models and the Heckman sample selection model.

The following section, or the third section of the paper, uses historical PUMS data from 1930 to 2010 to demonstrate the growth of solo households and the labour force, and to compare the relative attractiveness of cities for adults from married-couple households and solo households. The characteristics of the solo households in the US are explored by using binary logit models with 2010 ACS PUMS data in the section following the literature review on the distinctive characteristics of solo households. In the subsequent section, comparisons are made between solo households and adults from married-couple households regarding their living and travel characteristics. For this purpose, descriptive statistics from the 2009 NHTS and three Heckman sample selection models with data from the 2009 AHS national sample are used.

The Heckman sample selection model (Heckman, 1976, 1979) was used because it has clear advantages over ordinary least squares and binary logit or probit models. It uses a system of equations to jointly predict the probability of selection (for example, the decision to be solo) and the outcome variable (for example, commuting trip distance or time). Since the model predicts an outcome variable on the basis of the probability of being selected into a group, the predicted outcome can be interpreted as the result of self-selection into the group. Thus the model has a causal element that is absent in ordinary least squares models. A number of studies in transport have used the Heckman model, including Cooke and Ross (1999), Deka (2002) and Vance and Iovanna (2007). Studies in other fields have used it to compare SAT scores for coached and uncoached students (Briggs, 2004), treatment duration for individuals treated by two different drugs (Leslie and Ghomrawi, 2008), profits for family-controlled and non-family firms (Maury, 2006) and child-abuse reporting cases for Black and White populations (Ards et al., 1998).

The analysis of travel characteristics of solo households is followed by a comparison of the moving behaviour of solo and married-couple households. Three binary logit models were used with 2009 AHS data to make these comparisons. Additional information about the specific models has been provided in the sections where they have been used.

The Growth of American Solo Households and the Potential Reasons for Their Growth

To verify Klinenberg’s observations about the growth of solo households, especially their growth in cities, historical census PUMS data were obtained from the Integrated Public Use Microdata Series (Ruggles et al., 2010). The changing shares of solo households as well as population and labour force from solo households, obtained from this source, are presented in Figure 1. As evident, the share of solo households increased from 6 per cent to almost 28 per cent between 1930 and 2010. During this period, the share of married-couple households decreased from 79 per cent to 49 per cent. Because of the presence of children in non-solo households, the increase in the share of population in solo households is less substantial than the increase in the share of solo households. However, the share of labour force from solo households has been growing more rapidly than population.

Figure 1.

Share of solo households, persons and labour force in the US, excluding group quarters, 1930–2010.

Consistent with Klinenberg’s assertion about the growth of solo households in cities, the figure shows that the share of solo households, population and labour force in central cities is significantly higher than the country as a whole. While the share of solo female labour force in central cities has been higher than male labour force since 1930, the share of solo male labour force in central cities has increased considerably since 1970. As of 2010, the share of solo male and female labour force in central cities was 15 per cent and 16 per cent respectively.

Klinenberg identifies economic development, the welfare state and cultural change as the primary reasons for the growth of solo households. He contends that the rising status of women, new communications technologies, mass urbanisation and increasing longevity contributed to the trend. An additional reason for the growth of solo households in the US could be the diminishing attractiveness of marriage. As Warren and Tyagi (2003) emphasised, marriage for many working adults has become a trap that leaves them with less savings for the rainy day compared with solo households.

Since women have historically contributed more to the growth of solo households in cities than men, Kleinenberg’s assertion about the rising status of women as a reason for the growth of solo households requires further discussion. Available data on women’s education and earnings seem to provide some support to the claim that women’s status has risen in recent decades. In 1960, women received 35 per cent of the bachelor’s degrees in the US, whereas they received 58 per cent of the degrees in 2004 (Buchmann and DiPrete, 2006). The 2010 ACS PUMS data for persons in the labour force 25 or older show that 51 per cent of women from solo households have a bachelor’s degree or higher level of education, whereas 47 per cent of women from married-couple households and 45 per cent of men from both types of household have that distinction. Thus women from solo households do have a certain level of advantage in education over women from married-couple households and men.

The increasing level of education for women has been accompanied by an increase in earnings. Although women from both solo and married-couple households still earn substantially less than male workers, the increase in educational attainment among women has reduced the gender gap in earnings. That earned income for women has increased in recent years is evident from Figure 2, which shows the average earnings of persons age 25 and older from solo and married-couple households. Due to the 2007–09 recession, the earnings for all persons in the labour force stagnated in 2010. Yet a comparison between 1990 and 2010 shows that the gap between male and female earnings decreased, in all places as well as in central cities, for both married-couple households and solo households. Although the earnings for the female labour force from married-couple households are rapidly increasing, consistent with their respective educational attainment, women from solo households continue to earn more.

Figure 2.

Personal earned income in constant 1999 dollars for persons in the labour force age 25 and over in the US, excluding group quarters, 1990–2010.

Taking a cue from economists who believe that marriage is an economic institution and that the decisions to marry and divorce can be explained by utility theory (for example, Becker, 1974; Becker et al., 1977), one can argue that, due to increasing earnings, women today feel less compelled to marry for security than in the past. Becker’s theory suggests that both husband and wife benefited from production efficiencies due to the division of labour in traditional households, but, as Isen and Stevenson (2010) point out, these production efficiencies have disappeared over time due to technological advances. The decrease in the production efficiencies can be expected to reduce the motivation to marry for both men and women, thereby leading to the growth of solo households.

Another rationale for an individual to go solo is the solo’s ability to move quickly whenever desirable employment opportunities arise. Their ability to move could potentially allow them to earn a higher salary in the job market. At an age of quick job turnover, solo workers certainly have an advantage.

What brings solos to cities? According to Klinenberg, educated solo men and women are increasingly flocking to large cities of the US to live independent lives. Klinenberg emphasised the liberal culture as a reason for their concentration in cities. However, as shown in Figure 2, an earnings differential also makes cities more attractive to solos than married-couple households. While men from married-couple households earn nearly the same in central cities as elsewhere, both men and women from solo households earn far more by living in central cities. Since men have continued to remain the primary earners of married-couple households, the attractiveness of central cities for these households is lower than solo households.

The Distinctive Characteristics of Solo Households in Past Studies

A number of studies that compared the residential location and travel patterns of different types of household provide evidence of the distinctive nature of solo households. In a study for the San Francisco Bay Area, Bhat and Guo (2007) found that individuals from solo households are less likely to own cars and more likely to live in areas with high street density. In a study for the Louisville, Kentucky–Indiana metropolitan area, Scott and Horner (2008) found that persons from solo households live in areas with greater access to activities, especially commercial activities within walking distance. Similarly, Brownstone and Golob (2009) found in a California study that, compared with others, solo households locate more often in high-density areas, drive fewer miles and consume less fuel.

Studies from other countries provide more extensive evidence about solo households. In a British study, Dargay and Hanly (2007) found that solo workers are less likely to commute by automobile than others. Noting that workers from solo households spend less time commuting, Vandersmissen et al. (2003) concluded in a study for Québec City, Canada, that devoid of family constraints, solo workers have a greater opportunity than others to locate close to work. A study by Howley (2009) found that solo households are less likely than others to move out of a central-city area of Dublin. Eluru et al. (2009) found in a study for Zurich, Switzerland, that solo persons are more likely to move for education and employment compared with persons from multiple-person households.

Other studies have identified additional travel and transport attributes that distinguish solo households from others. For example, Roorda et al. (2010) found that persons from solo households make more trips per person compared with persons from married-couple households in several Canadian cities. Analysis conducted by Arentze, et al. (2005) in the Netherlands showed that solo individuals make more shopping trips than others. However, as noted by Newsome et al. (1998), solo individuals often have to make more trips than individuals from multiple-person households because of their sole responsibility of household activities.

To the detriment of solo households, several studies have shown that married-couple households can save significantly because of economies of scale generated by the sharing of goods and services within the household. For example, Lazear and Michael (1980) showed that married couples have an advantage over solo households because they save substantially on the consumption of shelter, food and other goods because of sharing. Similarly, Nelson (1988) found that larger households generate more savings in the consumption of shelter, food, household furnishings and transport compared with smaller households. A number of other studies have provided additional insights about the economies of scale due to sharing of goods and services by household members (Deaton and Paxson, 1998; Gan and Vernon, 2003; Fernández-Villaverde and Krueger, 2002; Browning et al. 2006). These studies leave little doubt that solo households sacrifice the benefits of sharing housing space and household goods and services. For the growth of solo households, countervailing factors must offset these sacrifices.

Who are the Solo in America?

Although Klinenberg cites a number of potential reasons for the growth of solo households, his objective was to describe a phenomenon rather than to provide a statistical discourse. This study takes a more methodological approach. To identify the factors associated with the likelihood of living in a solo household, three binary logit models were used with 1 per cent PUMS data for the US from the 2010 ACS. The results are presented in Table 1. In model 1, the likelihood of living in a solo household is predicted by including all individuals in the dataset excluding those below age 25 and people living in group quarters. In model 2, the data are further restricted to workers in the 25–64 age group to examine if their likelihood of living in solo households differed from the likelihood for all individuals. This model was included especially in view of the large proportion of persons in the labour force from solo households in cities. Model 3 is more elegant than model 1 and model 2 in that it combines the first two models and it can be used to show the relationships between the dependent variable and the independent variables separately for workers and non-workers. In this model, all variables are interacted with employment status except for the dummy variable on employment.

Table 1.

Binary logit model results on the likelihood of living in a solo household

	Model 1 Likelihood of living solo: all persons		Model 2 Likelihood of living solo: workers only		Model 3 Likelihood of living solo: all persons (interaction model)^b
	Coefficient (β)	S.E.	Coefficient (β)	S.E.	Coefficient (β)	S.E.
Constant	−2.807*	0.015	−2.761*	0.018	−1.383*	0.003
Employed person^a					−1.350*	0.018
Less than high school	−0.043*	0.007	-0.320*	0.013	−0.247*	0.013
High school diploma	−0.088*	0.005	−0.165*	0.008	−0.135*	0.008
Some college or associate’s degree (referent)
Bachelor’s degree	0.035*	0.006	0.189*	0.008	0.153*	0.008
Post-baccalaureate education	0.013*	0.001	0.045*	0.002	0.034*	0.002
Female^a	0.241*	0.004	-0.060*	0.006	0.097*	0.006
African American^a	0.286*	0.007	0.321*	0.009	0.332*	0.009
Hispanic^a	−0.238*	0.010	-0.235*	0.013	−0.221*	0.013
Foreign-born immigrant^a	−0.334*	0.009	−0.306*	0.012	−0.279*	0.012
Language at home is English^a	0.268*	0.009	0.257*	0.013	0.248*	0.013
Central-city resident^a	0.494*	0.006	0.591*	0.008	0.577*	0.008
Age 25–34 (referent)
Age 35–44	−0.131*	0.008	−0.143*	0.009	−0.196*	0.009
Age 45–54	0.274*	0.008	0.203*	0.008	0.119*	0.009
Age 55–64	0.674*	0.007	0.566*	0.008	0.489*	0.009
Age 65–74	1.042*	0.008			0.873*	0.012
Age 75–84	1.609*	0.009			1.311*	0.022
Age 85 or higher	2.265*	0.011			1.902*	0.061
Personal earned income less than $20k	0.059*	0.011	0.208*	0.012	0.150*	0.012
Personal earned income $20–40k	0.164*	0.011	0.343*	0.012	0.277*	0.012
Personal earned income $40–60k	0.278*	0.011	0.391*	0.012	0.352*	0.012
Personal earned income $60–80k	0.261*	0.012	0.333*	0.013	0.312*	0.013
Personal earned income $80–100k	0.214*	0.015	0.253*	0.015	0.245*	0.015
Personal earned income $100k or higher (referent)
N	2,058,284^c		1,252,437^d		2,058,284^c
-2 log likelihood	1,642,350		868,564		1,715,086
Nagelkerke R²	0.103		0.042		0.045
Model chi-squared	126,595 (p<0.0001)		27,108 (p<0.0001)		53,858 (p<0.0001)

Dummy variable, coded 1 for shown category and 0 for others.

Note that all independent variables except ‘Employed person’ are interacted variables in model 3.

Includes all persons, age 25 and over, excluding group quarters.

Includes all workers, age 25–64, excluding group quarters.

Note: * significant at the 1 per cent level.

Source: estimated from the 2010 American Community Survey Public Use Micro Sample data.

Living in a solo household is the dependent variable in all three models, which was coded 1 for persons living in a solo household and 0 for others. The independent variables in the models are the demographic and socioeconomic characteristics of the individuals in the dataset. In addition, a dummy variable on central city was included to examine the extent to which central-city living is associated with being solo.

The results of the three models in Table 1 are consistent with each other in terms of the signs of the coefficients and their statistical significance. Although many other characteristics of individuals are associated with the likelihood of living in a solo household, the three models show that ageing contributes to the likelihood more than other characteristics. This is more evident in model 1 and model 3 than in model 2 because in these two models the persons older than age 65 are also included. Model 1 shows that a person who is 85 or older is almost 900 per cent more likely, and a person between age 75 and 84 is almost 400 per cent more likely, to live in a solo household compared with a 25–34-year-old person (e^2.265 = 9.63 and e^1.609 = 4.997). Model 1 also shows that being female and being African American increases the likelihood of living in a solo household, whereas being Hispanic, a foreign-born immigrant, or a non-English speaker decreases the likelihood. The latter results are consistent with the notion that immigrants are more family-oriented than non-immigrants. Consistent with Klinenberg’s assertion about the growth of solo households in cities, both models show a positive association between central-city living and being solo. Persons with personal earned income between $40,000 and $60,000 are the most likely to live in solo households, whereas persons with lower and higher earnings are less likely. Higher education increases the likelihood of living in a solo household, although the influence of higher education is higher for workers than the general population. According to model 2, a worker with a bachelor’s degree is about 21 per cent more likely to live in a solo household compared with a person with an associate degree or some college education. For the general population, being female increases the likelihood of living in a solo household by 27 per cent, but for workers aged 25–64, women have a 6 per cent lower likelihood of living in a solo household. Further examination of the data reveals that a reason for this discrepancy is that a far larger number of elderly widows live in solo households compared with widowers. In male solo households, 34 per cent are divorced, 42 per cent have never married and 14 per cent are widowed, whereas in female solo households, 29 per cent are divorced, 25 per cent have never married and an overwhelming 40 per cent are widowed (potentially due to women’s greater longevity than men). In sum, although Klinenberg’s assertions are mostly validated by the analysis, the significant likelihood of living in a solo household by the elderly suggests that these individuals are being solo because of their circumstances instead of going solo on their own volition. Yet the analysis shows that higher education, a middle-class income and central-city residence are positively associated with the likelihood of living in a solo household.

The Distinctive Travel and Living Characteristics of the American Solo

The characteristics of solo households found in the literature pertaining to other developed countries can also be observed in the US. Basic statistics from the 2009 NHTS, presented in Table 2, show that men from solo households, on average, travel a significantly shorter distance and spend less time commuting to work than men from married-couple households. Solo women commute a shorter distance than women from married-couple households, but spend an almost identical amount of time commuting. Solo persons’ higher propensity to use public transit and walk, and a lower propensity to use personal vehicles, is evident for both men and women. As expected, a far smaller proportion of solo households own vehicles compared with married-couple households and the average number of vehicles owned by solo households is also smaller than married-couple households. Solo women have a lower propensity to own vehicles than solo men as well as men and women from married-couple households. Consistent with other studies, American solo men and women make a greater proportion of trips for shopping/errands and personal business/obligations compared with men and women from married-couple households.

Table 2.

Comparison of travel and dwelling characteristics of persons aged 25–64 from solo households and married-couple households in the US

	Couple		Solo
	Male	Female	Male	Female
Mean trip distance (miles) ^a
All trip purposes	13.08	9.02	10.09	7.18
Work trips only	14.87	10.22	10.93	8.64
Mean trip duration (minutes) ^a
All trip purposes	22.14	18.17	20.57	18.05
Work trips only	27.09	21.78	22.69	21.55
Mode used for all trip purposes (percentages) ^b
Car, truck, van and SUV	87.2	87.2	77.4	80.6
Public transit (excluding taxi, ferry, school bus and inter-regional transit)	1.5	1.6	4.2	4.1
Walk	8.7	10.0	14.0	13.5
Bicycle	1.0	0.3	2.2	0.2
Other	1.6	1.0	2.2	1.6
Total	100	100	100	100
Mode used for work trip purposes (percentages) ^b
Car, truck, van and SUV	91.5	93.9	85.8	88.5
Public transit (excluding taxi, ferry, school bus and inter-regional transit)	3.1	2.8	6.1	7.0
Walk	2.3	2.4	3.9	3.7
Bicycle	1.2	0.2	1.5	0.2
Other	1.9	0.6	2.7	0.7
Total	100	100	100	100
Trip purpose (percentages) ^b
Home	33.8	33.6	32.7	31.1
Work	20.4	12.1	18.8	17.0
Shopping/errands	15.9	19.2	20.6	22.4
Social/recreational	10.1	10.3	11.6	10.0
Family personal business/obligations	2.9	3.4	3.4	4.7
Meals	7.2	6.5	7.3	6.9
Other purposes	9.6	14.9	5.6	7.9
Total	100	100	100	100
Mean number of vehicles in household^a	2.40^c	2.37^c	1.28^c	1.02^c
Percentage of households with at least one vehicle^a	97.6	97.0	86.7	85.0
Percentage living in rented home^a	24.4	26.9	53.6	45.1
Dwelling type (percentages) ^b
Detached single house	76.4	74.8	45.3	48.3
Row house or townhouse	12.2	13.5	40.5	40.3
Other housing types	11.3	11.7	14.2	11.4
Total	100	100	100	100

Solo male and female are different from married-couple men at the 5 per cent significance level on t-test.

Differences are significant at the 5 per cent level on chi-squared test.

According to 2010 ACS PUMS, these figures are 2.33, 2.28, 1.15 and 0.97 respectively.

Source: estimated from the 2009 National Household Transportation Survey.

The travel patterns of American solo households are potentially associated with their dwelling characteristics. As evident in Table 2, less than half of the solo men and women live in detached single houses, whereas around 75 per cent of married couples live in such dwellings. Almost 54 per cent of solo men and 45 per cent of the solo women live in rented dwellings, compared with about a quarter of the men and women from married-couple households. By living in rented homes, solo households can minimise their commuting distance and time, although they spend more on shelter than married-couple households on a per-earner basis. The 2009 US Consumer Expenditure Survey data analysed by the author shows that the average per-earner shelter expenditure for an adult in a married-couple household is $1394 for a quarter, whereas the expenditure for a solo male is $1660 and for a solo female is $1561 (US Bureau of Labor Statistics, 2010).

Although the data presented in Table 2 suggest that the travel characteristics of the solo are environmentally more sustainable than individuals from married-couple households, they do not show that their distinctive travel characteristics are linked to their decision to be solo. A model that takes into account this link is the Heckman sample selection model.

The Heckman model was used to compare the commuting time and distance of workers with the 2009 AHS national sample data. The dataset was restricted to persons aged 25–64 belonging to solo households or married-couple households in urbanised areas. Persons from other types of household, such as single-parent households, were omitted so that comparisons could be made directly between solo households and married-couple households. The results of the two models are presented in Table 3 (models 1 and 2). The models were estimated by using both STATA and SAS to ensure that the results are identical.

Table 3.

Heckman model results on commuting distance, time and the likelihood of using automobile for commuting

	Model 1 Commute distance		Model 2 Commute time		Model 3 Automobile use
	Coefficient (β)	t-value	Coefficient (β)	t-value	Coefficient (β)	t-value
Outcome component
Constant	5.892**	1.83	15.810*	3.35	0.981*	3.32
Age of person	0.462*	3.13	0.534*	2.46	0.007	0.52
Age squared	−0.005*	−3.21	−0.006*	−2.37	−0.0001	−0.60
Female^a	−1.392*	−3.70	−0.916**	−1.66	0.078*	2.49
African American^a	1.007**	1.82	2.572*	3.15	−0.192*	−4.00
Hispanic^a	2.472*	3.57	3.294*	3.23	0.273*	4.94
Renter^a	−1.481*	−3.09	−1.513*	−2.15	−0.032	−0.80
Number of bedrooms in dwelling	0.724*	3.20	0.151	0.46	0.096*	4.73
Number of vehicles in household	0.557*	2.22	−1.060*	−2.89	0.502*	17.67
Lives in central city^a	−3.505*	−7.99	−1.310*	−1.99	−0.567*	−16.34
Moved in the past 5 years^a	0.402	0.94	−0.840	−1.34
Distance to work in miles					0.023*	11.81
Sigma	12.703*	81.10	18.522*	85.81
Selection component
Constant	−1.094*	−26.91	−1.097*	−26.93	−1.093*	−28.65
Age 25–34	0.103*	3.56	0.103*	3.56	0.098*	3.45
Age 35–44	−0.160*	−5.76	−0.160*	−5.76	−0.135*	−5.06
Age 45–54	-0.062*	−2.36	−0.062*	−2.35	−0.064*	−2.54
Age 55–64 (referent)
African American^a	0.347*	10.91	0.348*	10.94	0.338*	10.75
Hispanic^a	−0.176*	−4.97	−0.176*	−4.97	−0.205*	−5.88
Foreign born^a	−0.354*	−11.24	−0.352*	−11.13	−0.296*	−10.99
Less than bachelor’s degree (referant)
Bachelor’s degree	0.089*	3.65	0.094*	3.88	0.048*	2.11
Post-baccalaureate education	0.147*	4.97	0.152*	5.15	0.098*	3.53
Salary income less than $20k	−0.066**	−1.66	−0.061	−1.54	−0.019	−0.52
Salary income $20–40k	0.416*	10.41	0.416*	10.34	0.399*	10.73
Salary income $40–60k	0.362*	9.20	0.361*	9.12	0.370*	10.05
Salary income $60–80k	0.219*	5.18	0.217*	5.11	0.220*	5.55
Salary income $80–100k	0.245*	4.99	0.248*	5.06	0.240*	5.23
Salary income $100k or higher (referent)
Lives in central city^a	0.404*	18.69	0.404*	18.69	0.401*	18.69
Rho	−0.180*	−3.70	−0.130*	−2.34	−0.992*	−196.01
Lambda	−2.289		−2.412
N	22,616		22,616		22,616
Log likelihood	−29,798		−31,625		−12,617
AIC	59,652		63,306		25,288
Schwarz criterion	59,877		63,530		25,505
Wald chi-squared	179	(p<0.0001)	49	(p<0.0001)	1042	(p<0.0001)

Dummy variable, coded 1 for shown category and 0 for others.

Notes: * significant at the 5 per cent level; ** significant at the 10 per cent level.

Source: estimated from the 2009 American Housing Survey national sample.

The Heckman model consists of two components: a selection component and an outcome component. In addition to the coefficients and standard errors of the variables, the model generates two critical parameter estimates: sigma (σ) and rho (ρ). While sigma is the adjusted standard error of the outcome model, rho represents the correlation between the errors of the two models and indicates whether the factors affecting selection are associated with the outcome. For example, if the selection group is solo individuals, and rho is statistically significant and negative, it would indicate that the outcome variable, commuting time or distance, for solos is shorter than individuals from married-couple households because of the selection effect.

In addition to the models on commuting distance and time, a Heckman sample selection probit model estimating the likelihood of commuting by automobile is presented in Table 2 (model 3). Similar to the typical Heckman model with a continuous outcome variable, the binary outcome variable of this model is predicted on the basis of a selection component. The outcome component of the three models, measuring commuting distance, commuting time and the likelihood of automobile use, is shown in the top part of Table 3, whereas the selection component, measuring the likelihood of being solo, is shown in the bottom part. In the selection component of all three models, the dependent variable is a dummy variable (solo = 1, else 0), whereas the dependent variables in the outcome component of model 1 and model 2 are continuous (distance and time) and binary in model 3 (automobile use = 1, else 0).

The selection component of the three models, which predicts the likelihood of being solo, is consistent with the model results in Table 1, where the likelihood of being solo was predicted with 2010 ACS PUMS data. Similar to the models in Table 1, the selection components of the models in Table 3 show that the likelihood of being solo is higher for central-city residents, African Americans, highly educated persons and persons with moderate earnings, while it is lower for Hispanic persons, immigrants and persons in mid-life. The outcome component of the three models also shows results consistent with expectation. According to the distance model (model 1), women, renters and central-city residents commute a shorter distance, whereas African Americans, Hispanic persons, persons living in large dwellings and persons from households with a larger number of vehicles commute a longer distance. These results are consistent with the results in Crane (2007), where distance was estimated with a random-effects generalised least squares model using past AHS data.

The commute time model (model 2) is similar to the distance model for all variables except the number of vehicles in the household. Although the number of vehicles has a positive association with distance and a negative association with commute time, both results are consistent with expectation because vehicles can reduce travel time. The only variable that was not statistically significant in either of the models was the variable on movers.

The negative sign of a statistically significant rho in the commuting distance and time models indicates that the selection into the solo group decreases commute distance and time. By using the parameter estimates of the two models, it can be estimated that an average solo person commutes 3.2 miles and 3.4 minutes less than individuals from married-couple households.

An additional variable, travel distance, was included in the probit model on automobile use for commuting (model 3 in Table 3) because mode choice widely varies by distance. The variable on moving was excluded because it could not be theorised why or how moving would affect mode choice. Although fewer variables are statistically significant in this model compared with the models on commuting distance and time, the parameter estimates are generally consistent with expectation, as they show that central-city residents and African Americans are less likely to commute by automobile, whereas persons from households with a large number of vehicles and persons living in larger dwellings are more likely to do so. As expected, individuals making longer trips are more likely to use an automobile for commuting. Most importantly, the highly significant negative rho indicates that persons from solo households are less likely to use an automobile than persons from married-couple households because of a selection effect. According to the model results, the automobile usage rate for commuting for an average solo person is 4 per cent lower than that for an average person from a married-couple household. In sum, the models on commuting distance, commuting time and automobile use for commuting, estimated with AHS data, are consistent with the statistics provided in Table 2 from the 2009 NHTS. The models provide further evidence that the travel patterns of solo households are environmentally more sustainable than the travel patterns of persons from married-couple households.

The structure of the Heckman model also provides some insights about the relationship between being solo, living in central cities and having more sustainable commutes. The selection component of the model in Table 3 shows that the solos are more likely to live in central cities and the outcome component shows that, by living in cities, they commute a shorter distance and use the automobile less often for commuting. The selection and outcome components together indicate that the higher propensity of the solo households to live in central cities adds to their propensity to commute a shorter distance and use an automobile less often. However, living in central cities is one of many reasons for their observed travel patterns.

The Moving behaviour of the American Solo

It can be hypothesised that solo persons display more sustainable travel characteristics because: they have a greater ability to move; and, they have a greater concern about job proximity than acquiring larger housing space compared with married-couple households. Solo households can be expected to have a greater ability and propensity to move because they have no constraints related to other persons in the household, such as spouse’s employment or children’s school. Because they do not benefit from sharing housing space, solo households can also be expected to have a lower propensity to move in order to acquire larger dwelling units. It can be further hypothesised that their lower attraction for housing space is accompanied by a greater attraction for proximity to work because they are the sole earners in household with greater time constraints than an average adult from a married-couple household. To test three specific hypotheses—that solo households move more often, move less to acquire larger housing units and move more for greater proximity to work—statistical models were used with the 2009 AHS data.

The AHS inquires whether a respondent’s household moved within the past 12 months. The responses to this question were used to examine whether solo households move more than married-couple households. Those who moved are subsequently asked by the AHS the reasons for their move. The response categories “needed a larger house or apartment” and “to be closer to work/school/other” were used to examine the attraction of solo households for housing space and proximity to work. The restriction of the dataset to age 25–64 is expected to eliminate most of those who moved for greater proximity to school. Among the movers from married couple-households, 15 per cent men and 13 per cent women aged 25–64 reported moving for a larger dwelling, but only 3 per cent men and 4 per cent women from solo households mentioned doing so. In contrast, 7 per cent men and 8 per cent women from married-couple households mentioned moving to be closer to work/school/other, compared with 10 per cent men and 9 per cent women from solo households. Since these differences may be because of other characteristics of the individuals, the relationship between household type and moving was examined by using binary logit models to control for variations in these characteristics. Sample-selection probit models, which would estimate the probability of being solo first and use the probabilities to predict moving, were considered inappropriate because moving, the outcome variable, refers to the past. Being associative, binary logit models do not involve this conceptual problem, although their results cannot necessarily be interpreted as causal.

Results from three binary logit models are presented in Table 4. In model 1, the dependent variable is a dummy variable on moving, which was coded 1 for movers and 0 for non-movers. In model 2, where the coefficients were obtained with data from movers only, the dummy dependent variable indicates whether a household moved to acquire a larger dwelling. Those who moved to acquire larger dwelling were coded 1 and the other movers were coded 0. Similarly, in model 3, the dummy dependent variable was coded 1 for those who moved for greater proximity to work/school/other, and other movers were coded 0.

Table 4.

Binary logit model results on the likelihood of moving, moving for larger dwelling unit, and moving for job proximity

	Model 1 Likelihood of moving		Model 2 Likelihood of moving to a larger dwelling		Model 3 Likelihood of moving for job proximity
	Coefficient (β)	S.E.	Coefficient (β)	S.E.	Coefficient (β)	S.E.
Constant	−1.788*	0.102	−2.368*	0.297	−3.728*	0.349
Solo male	0.262*	0.052	−1.399*	0.200	0.293*	0.146
Solo female	0.257*	0.055	−1.123*	0.199	0.242	0.158
Married-couple male and female (referent)
Salary income less than $20k	0.374*	0.073	−0.278	0.199	0.321	0.233
Salary income $20–40k	0.251*	0.077	−0.148	0.208	0.347	0.240
Salary income $40–60k	0.119	0.077	−0.017	0.208	−0.101	0.250
Salary income $60–80k	0.060	0.083	−0.070	0.223	0.115	0.260
Salary income $80–100k	0.174*	0.096	−0.012	0.253	−0.018	0.300
Salary income $100k or higher (referent)
Age 25–34	2.051*	0.055	0.344	0.184	1.207*	0.222
Age 35–44	1.059*	0.056	0.308	0.193	1.036*	0.233
Age 45–54	0.513*	0.054	0.014	0.206	0.747*	0.240
Age 55–64 (referent)
White race^a	−0.266*	0.044	0.010	0.123	−0.239*	0.121
Hispanic^a	0.200*	0.052	0.084	0.134	0.215	0.151
Less than high school	−0.550*	0.152	0.568	0.386	−0.497	0.603
High school diploma	−0.073	0.045	0.156	0.130	−0.115	0.149
Some college or associate degree (referent)
Bachelor’s degree	−0.073	0.047	0.264*	0.133	0.395*	0.139
Post-baccalaureate education	0.057	0.056	0.308	0.158	0.622*	0.160
Child under 18 present in household^a	−0.302*	0.047	0.619	0.111	−0.225	0.149
Number of household vehicles	−0.290*	0.020	0.043	0.055	0.073	0.060
N	24,706^b		4888^c		4888^c
−2 log likelihood	21,756		3010		2786
Nagelkerke R²	0.174		0.106		0.040
Model chi-squared	2822 (p<0.0001)		260 (p<0.0001)		88 (p<0.0001)

Dummy variable, coded 1 for shown category and 0 for others.

Includes all persons in age 25–64 from solo and married couple households

Includes all movers in age 25–64 from married-couple and solo households

Note: * Significant at the 5 per cent level.

Source: estimated from the 2009 American Housing Survey national sample.

As expected, model 1 shows that solo men and women are more likely to move than adult men and women from married-couple households. According to the model results, men and women from solo households are 30 per cent and 29 per cent more likely to move in a 12-month period compared with adult men and women from married-couple households (e^0.262 = 1.30 and e^0.257 = 1.29). The model also shows that younger adults and persons with lower incomes are more likely to move compared with older adults and persons with higher incomes. Level of education does not seem to influence the likelihood of moving to a great extent, except that individuals with the lowest level of education are less likely to move than others. Consistent with expectation, having children reduces the likelihood of moving. The negative association between the number of household vehicles and moving can be interpreted as a trade-off between transport cost and moving because greater mobility using vehicles can reduce the need for moving.

Fewer independent variables are statistically significant in model 2 and model 3 than in model 1, but both provide the expected results. The coefficients of model 2 show that solo men and solo women respectively, are 75 per cent and 67 per cent less likely to move to acquire a larger dwelling unit compared with adults from married-couple households (1-e^-1.399 = 0.75 and 1-e^-1.123 = 0.67). Model 3 shows that both solo men and solo women are more likely to move for greater proximity to work compared with adults from married-couple households. However, only the variable on solo men has a confidence level of 95 per cent, whereas the variable on solo women has a confidence level of 87 per cent. The model shows that men and women from solo households are 34 per cent and 27 per cent more likely respectively to move for job proximity compared with adults from married-couple households. Model 3 also shows that younger adults and individuals with a high level of education are far more likely to move for job proximity compared with others. Overall, the model results are consistent with the notion that adults from married-couple households are more interested in acquiring larger dwelling space, whereas solo households are more concerned about job proximity. The lower concern for larger housing space and a greater concern for job proximity seemingly lead to a more sustainable living and commuting behaviour by the solo.

Summary of the Findings and Implications

The growth of solo households and workers can be perceived as a sign of recovery for American central cities. A lower propensity to live in detached single homes and a greater propensity to live close to work, walk and use public transport make them more attractive to cities than two-earner households. To the benefit of themselves and their employers, they can move more freely than married couples to take advantage of employment opportunities.

This research showed that, in addition to ageing, the increase in education and earnings of women could be a reason for the growth of solo households. By living in central cities, solo workers benefit more than the traditional bread-winners from married-couple households. This research further showed that solo households are more likely to move for job proximity and less likely to move to acquire large dwelling units compared with married-couple households. Solo households have been able to bear the excess expenses on shelter and transport so far because of an earnings advantage, but if their costs increase relative to their earnings, the growth of solo households may slow down. For the continued growth of solo households in cities, there will have to be a continued increase in earnings, stabilisation of housing and transport costs, or both. Their earnings will increase if jobs increase because of higher demand for labour.

Although solo households have so far exhibited living and travel patterns that are conducive to environmentally sustainable cities, there is no guarantee that they will continue to exhibit these patterns. For example, despite their higher propensity to use public transit, according to PUMS data, the share of transit commuting trips by 25–64-year-old female solo workers decreased from 9.6 per cent to 7.8 per cent between 1990 and 2010, while the share for solo male workers decreased from 6.5 per cent to 6.2 per cent. During this period, average commuting time by transit for solo female workers increased from 40 minutes to almost 46 minutes, while the time for solo male workers increased from 39 minutes to 45 minutes. The decrease in transit usage and increase in transit commuting time were accompanied by an increase in the average number of vehicles owned by both solo females (from 0.90 to 0.97) and solo men (from 1.13 to 1.15).

To provide better incentives to solo households, planners should address local public transport needs. Although solo workers are still far more likely to commute by transit and spend less time commuting than workers from married-couple households, if the quality of transit service deteriorates, the commuting patterns of solo workers could potentially become less environmentally friendly.

Planners should also be concerned about the long-term consequences of the growth of solo households in cities. Although the living and commuting patterns of solo households are more sustainable than those of married-couple households, many older adults are solo because of their spouse’s death, divorce, or separation. In the absence of other persons in the household to take care of their travel needs, the older adults from solo households require special attention from society. Many elderly solos need special transport services. When today’s young solos become elderly in future decades, the need for these services will only increase.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

References

Ards

Chung

Myers

S. L.

Jr (1998) The effects of sample selection bias on racial differences in child abuse reporting, Child Abuse and Neglect, 23(12), pp. 103–115.

Arentze

T. A.

Oppewal

Timmermans

H. J. P.

(2005) A multipurpose shopping trip model to assess retail agglomeration effects, Journal of Marketing Research, 42, pp. 109–115.

Becker

(1974) A theory of marriage, in: Schultz

T. W.

(Ed.) Economics of the Family, pp. 293–344. Chicago, IL: University of Chicago Press.

Becker

G. S.

Landes

E. M.

Michael

R. T.

(1977) An economic analysis of marital instability, The Journal of Political Economy, 85(6), pp. 1141–1187.

Bhat

C. R.

Guo

J. Y.

(2007) A comprehensive analysis of built environment characteristics on household residential choice and auto ownership levels, Transportation Research Part B, 41, pp. 506–526.

Briggs

D. C.

(2004) Casual inference and the Heckman model, Journal of Educational and Behavioral Statistics, 29(4), pp. 397–420.

Browning

Chiappori

Lewbel

(2006) Estimating consumption economies of scale, adult equivalence scales, and household bargaining power. Discussion Paper No. 289, Department of Economics, Oxford University (www.economics.ox.ac.uk/Research/wp/pdf/paper289.pdf; accessed 11 June 2012).

Brownstone

Golob

T. F.

(2009) The impact of residential density on vehicle usage and energy consumption, Journal of Urban Economics, 65, pp. 91–98.

Buchmann

DiPrete

T. A.

(2006) The growing female advantage in college completion: the role of family background and academic achievement, American Sociological Review, 71, pp. 515–541.

10.

Cervero

(1989) Jobs–housing balancing and regional mobility, Journal of the American Planning Association, 55(2), pp. 136–150.

11.

Cooke

T. J.

Ross

S. L.

(1999) Sample selection bias in models of commuting time, Urban Studies, 36(9), pp. 1597–1611.

12.

Crane

(2007) Is there a quiet revolution in women’s travel? Revisiting the gender gap in commuting, Journal of the American Planning Association, 73(3), pp. 298–316.

13.

Dargay

Hanly

(2007) Volatility of car ownership, commuting mode and time in the UK, Transportation Research Part A, 41, pp. 934–948.

14.

Deaton

Paxson

(1998) Economies of scale, household size, and demand for food, Journal of Political Economy, 106(5), pp. 897–930.

15.

Deka

(2002) Predicting commute time of non-workers in the context of welfare reform, Journal of Urban Affairs, 24(3), pp. 333–352.

16.

Downs

(1992) Stuck in Traffic: Coping with Peak-hour Traffic Congestion. Washington, DC: The Brookings Institution Press.

17.

Eluru

Sener

I. P.

Bhat

C. R.

Pendyala

R. M.

Axhausen

K. W.

(2009) Understanding residential mobility: joint model of the reason for residential relocation and stay duration, Transportation Research Record, 2133, pp. 64–74.

18.

Fernández-Villaverde

Krueger

(2002) Consumption over the life cycle: some facts from consumer expenditure survey data. Working Paper No. 9382, National Bureau of Economic Research, Cambridge, MA (www.nber.org/papers/w9382; accessed 11 June 2012).

19.

Gan

Vernon

(2003) Testing the Barten model of economies of scale in household consumption: toward resolving a paradox of Deaton and Paxson, Journal of Political Economy, 11(6), pp. 1361–1377.

20.

Heckman

(1976) The common structure of statistical models of truncation, sample selection and limited dependent variables and a simple estimator for such models, Annals of Economic and Social Measurement, 5(4), pp. 475–492.

21.

Heckman

(1979) Sample selection bias as a specification error, Econometrica, 47(1), pp. 153–161.

22.

Howley

(2009) Attitudes towards compact city living: towards a greater understanding of residential behavior, Land Use Policy, 26, pp. 792–798.

23.

Isen

Stevenson

(2010) Women’s education and family behavior: trends in marriage, divorce and fertility. Working Paper No. 15725, National Bureau of Economic Research, Cambridge, MA (www.nber.org/papers/w15725; accessed 12 June 2012).

24.

Klinenberg

(2012) Going Solo: The Extraordinary Rise and Surprising Appeal to Living Alone. New York: The Penguin Press.

25.

Lazear

E. P.

Michael

R. T.

(1980) Family size and the distribution of real per capita income, American Economic Review, 70(1), pp. 91–107.

26.

Leslie

R. S.

Ghomrawi

(2008) The use of propensity scores and instrumental variable methods to adjust for treatment selection bias. Paper No. 366-2008, SAS Global Forum (www2.sas.com/proceedings/forum2008/366-2008.pdf; accessed 20 July 2012).

27.

Maury

(2006) Family ownership and firm performance: empirical evidence from western European corporations, Journal of Corporate Finance, 12, pp. 321–341.

28.

Mieszkowski

Mills

E. S.

(1993) The causes of metropolitan suburbanization, Journal of Economic Perspectives, 7(3), pp. 135–147.

29.

Nelson

J. A.

(1988) Household economies of scale in consumption: theory and evidence. Econometrica, 56(6), pp. 1301–1314.

30.

Newsome

T. H.

Walcott

W. A.

Smith

P. D.

(1998) Urban activity spaces: illustrations and application of a conceptual model for integrating the time and space dimensions, Transportation, 25, pp. 357–377.

31.

Roorda

M. J.

Páez

Morency

Mercado

Farber

(2010) Trip generation of vulnerable populations in three Canadian cities: a spatial ordered probit approach, Transportation, 37, pp. 525–548.

32.

Ruggles

S. J.

Alexander

Genadek

Goeken

. (2010) Integrated public use microdata series: version 5.0 [machine-readable database]. University of Minnesota, Minneapolis, MN.

33.

Scott

D. M.

Horner

M. W.

(2008) The role of urban form in shaping access to opportunities: an exploratory spatial data analysis, Journal of Transport and Land Use, 1(2), pp. 89–119.

34.

US Bureau of Labor Statistics (2010) 2009 consumer expenditure survey. CD-ROM. US Bureau of Labor Statistics, Washington, DC.

35.

Vance

Iovanna

(2007) Gender and the automobile: analysis of nonwork service trips, Transportation Research Record, 2013, pp. 54–61.

36.

Vandersmissen

Villeneuve

Thériault

(2003) Analyzing changes in urban form and commuting time, The Professional Geographer, 55(4), pp. 446–463.

37.

Warren

Tyagi

A. W.

(2003) The Two-income Trap: Why Middle-class Parents Are Going Broke. New York: Basic Books.