CC BY-NC-ND
0ARTICLE IN PRESS
IATSSR-00129; No of Pages 8
IATSS Research xxx (2016) xxx-xxx
Contents lists available at ScienceDirect
IATSS Research
Determinants of long-duration commuting and long-duration commuters' perceptions and attitudes toward commuting time: Evidence from Kunming, China
Mingwei He *, Shengchuan Zhao
School of Transportation & Logistics, Dalian University of Technology, Liaoning 116024, China
ARTICLE INFO ABSTRACT
Understanding the commuting patterns of long-duration commuters and the possible changes in these patterns can help policymakers adopt the more reasonable land use and transportation policies. With Kunming in China as a case study, the determinants of long-duration commuting trips were identified based on logistic regression model. The results indicated that age, education level, number of workers, presence of retirees, and residential location have a significant impact on the occurrence of long-duration commuting trips. The ideal commuting times and tolerance thresholds of commuting time of long-duration commuters were also investigated. The statistical results revealed the distributions of ideal commuting times and tolerance thresholds of commuting time of both short- and long-duration commuters. The average tolerance threshold of commuting time and the average ideal commuting time of long-duration commuters were greater than those of short-duration commuters. For 97.2% of the long-duration commuters, their actual commuting time was longer than the ideal commuting time; this finding indicates that most long-duration commuters are dissatisfied with their commuting time. The actual commuting time of 40.1% long-duration commuters exceeded their tolerance thresholds; these commuters are eager to reduce their commuting time.
© 2016 International Association of Traffic and Safety Sciences. Production and hosting by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Available online xxxx
Keywords: Commuting time Long-duration commuting Ideal commuting time Tolerance threshold China
1. Introduction
Chinese cities have been experiencing profound social, economic, institutional, and spatial transformations since the 1980s. These transformations have changed the jobs-housing relationship and commuting behaviors. As a result, the commuting time and distance for Chinese urban residents have also experienced significant changes. Long-duration commuting accounts for a large percentage of commuting trips and plays an increasingly important role in transportation demand, especially during morning-evening rush hours. From the perspective of the entire society, long-duration commuting trips are suboptimal because they may lead to increased pressure on infrastructure [1] and increased air pollution and energy consumption [2]. From the perspective of individuals, an increase in commuting time may decrease the time for participating in other activities, such as personal entertainment activities [2]. In addition, with the increase in commuting time, the individual will be subjected to a higher commuting cost. Most
* Corresponding author. E-mail addresses: hmwei01@163.com (M. He), szhao@dlut.edu.cn (S. Zhao). Peer review under responsibility of International Association of Traffic and Safety Sciences.
commuters try to achieve the expected commuting time by changing residential or job location or changing travel mode. Therefore, studies on the determinants of long-duration commuting trips and on the perceptions and attitudes of long-duration commuters with regard to commuting time are of great significance to residential location choice, jobs-housing relationship, commuting pattern, and land use. Such studies can help policymakers better understand commuting patterns and possible changes in these patterns so that more reasonable land use and transportation policies can be adopted.
Many studies have investigated various factors that influence commuting time. Some empirical studies have reported that jobs-housing relationship has a significant impact on commuting time [2-6]. Levinson [5] found that commuters living in job-rich areas and working in housing-rich areas have shorter commuting time. With Beijing as a case study, Zhao et al. [2] found that a statistically significant correlation exists between jobs-housing balance and a worker's commuting time when the factors of transport accessibility, population density, and socioeconomic characteristics are controlled; the higher the jobs-housing balance is, the shorter the worker's commuting time is. However, several researchers have questioned the significant impact of jobs-housing relationship on commuting time. Giuliano and Small [7] found that in five counties of the Los Angeles region, only extreme jobs-housing imbalance has a significant impact on average commuting
http://dx.doi.org/10.1016/j.iatssr.2016.08.001
0386-1112/© 2016 International Association of Traffic and Safety Sciences. Production and hosting by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
ARTICLE IN PRESS
M. He, S. Zhao / ¡ATSS Research xxx (2016) xxx-xxx
time. Ewing [8] found that in Palm Beach County in the US, only a weak relationship exists between job-housing balance and average commuting time when socio-demographic variables are controlled.
Apart from the jobs-housing relationship, other variables also influence commuting time. These variables include travel mode and socio-economic and demographic characteristics. Residential location, workplace location, and travel mode are all exogenous variables; they are directly or indirectly influenced by the socio-economic and demographic characteristics of commuters. Existing studies paid more attention to the effects of the socio-economic and demographic variables of commuters on commuting time. In a study on Korean commuters, Lee and McDonald [9] found that commuting distances and times are longer for male workers, full-time salaried workers, workers with more education, and homeowners. In a study on commuting time in Los Angeles, Giuliano [10] found that longer commutes are associated with males, higher income, home-ownership, age (between 25 and 50), use of mass transit, and work status. A number of empirical studies have focused on gender differences in commuting, and one overwhelming conclusion is that females have shorter commuting trips than males in western countries [11-14]. One of the major explanations for the differences in commuting between males and females is that females undertake more household responsibilities than males [15]. However, different from these conclusions, Wang and Chai's [16] conclusion is that the average commuting time of females in Beijing (34.1 min) is longer than that of males (31.9 min). He et al. [17] investigated gender differences in commuting in Kunming and found that the commuting distance of females is obviously shorter than that of males; meanwhile, the commuting time of females is longer than that of males. The main reason for this finding is that the mobility of females is far lower than that of males. Many studies have paid attention to the effects of the presence of children on commuting time, and some controversy surrounds such effects. Preston et al. [18] found that the presence and ages of children significantly reduced women's commuting times, although the effects of parenthood were muted for minority women. Singell and Lillydahl [19] found that the presence of children has a negative effect on commuting time for both male and female workers. Johnston-Anumonwo [20] reported that the presence of children is not responsible for women's shorter work trips. Kim [21] found that the presence of children has no significant impact on the commuting time of males or females.
As shown above, previous studies have identified various factors that influence commuting time. However, only a few studies have focused on long-duration commuting trips. To the best of the authors' knowledge, no detailed investigations have been conducted on long-duration commuters' perceptions and attitudes towards commuting time. In the present study, with Kunming in China as a case study, we focused on long-duration commuters. Long-duration commuters are defined as people who have 40 min or more one-way commuting time.
First, the effects of socio-economic and demographic variables on long-duration commuting trips were studied. Differences exist between China and western countries in terms of economic and social settings, particularly in family structure and social roles of females. The employment rate of females in China is over 80%, which is far higher than that in western countries [22]. When a female participates in the labor market, she becomes a full-time worker because finding part-time jobs is difficult in China [23,24]. At the same time, female mobility and travel speed are also far lower than those of males. Furthermore, China has many big families, in which parents or parents-in-law live together with the young couple or unmarried adult children. The legal retirement age in China is 60 for men, 55 for women in managerial functions, and 50 for all other women compared with 60 or 65 in Western countries [24,25]. The older generation have the ability and willingness to provide substantial help in taking care of children and other housework. Therefore, identifying whether gender, the presence of children, and the presence of retirees have a significant impact on long-duration commuting choice would be interesting.
Second, the perceptions and attitudes of long-duration commuters toward commuting time were investigated. Several interesting but unexplored questions need to be answered. What are long-duration commuters' ideal commuting times? What are their tolerance thresholds for commuting time? What is the difference between actual commuting time and ideal commuting time? The answers to these questions would help determine the possible changes in the commuting behaviors and residential and job location choices of long-duration commuters.
2. Data
Kunming, which is the capital of Yunnan Province in southwest China, was adopted as the case study. Kunming's population exceeded 6.0 million, and the per capita gross domestic product was 56,236 CNY/year (about 9070 US dollars/year) in 2014 [26]. Kunming's urban area is about 305 km2 (see Fig. 1). About 3.2 million people live in the urban area of Kunming. As a second tier city, Kunming is representative for a multitude of Chinese cities with a comparative size.
Typically, many Chinese cities have a hierarchical center structure with a main center, a few sub-centers and a number of local centers and community centers [27]. Kunming is a representative of this kind of city. The main center is developed based historical urban center and has the highest density of workplaces, commercial and service facilities. Population density is higher in the main center than in the outer part of this region. The main center has an unchallenged status as the center area of the urban area. About 1.2 million people live in the center area of the urban area of Kunming (within the second ring road). The urban area supplies 1.7 million jobs, and about 0.9 million jobs are provided within the second ring road [28].
A questionnaire survey was conducted in Kunming in August 2014. Two sampling strategies were applied: stratified and simple random sampling. Stratified random sampling divides the entire sampling frame into several strata and then draws out samples from each stratum. In this study, different scale enterprises, public institutions, government agencies, supermarkets, and retail stores in different regions were regarded as different strata. Stratified random sampling involved 13 organizations, and 692 samples were obtained. Meanwhile, 103 samples were obtained through simple random sampling in the streets. A total of 795 commuters completed the questionnaires, and 42 questionnaires were eliminated because of incomplete responses. Overall, the database used in this study includes 753 valid samples. Table 1 presents the profile of the samples.
3. Determinants of long-duration commuting
A commuting time of 40 min has emerged as the threshold of long commuting trip in the previous studies [29]. The distribution of actual commuting time is presented in Table 2. In our samples, 23.6% of commuters have a commuting time longer than or equal to 40 min. Based on the previous studies and the distribution of commuting time, we use 40 min as the threshold of long commuting trip. The dependent variable used in the models is LT40, which is equal to 1 if a worker spends at least 40 min on commuting to work. The independent variables used in the models were selected based on the findings ofprevious studies and data availability. The definitions and values of the variables are shown in Table 3.
The dataset was split randomly into two mutually exclusive subsets: the training set containing 590 cases and the test set containing 163 cases. The training data was used for fitting a model and the test data was used to measure the predictive accuracy of the model. Two logistic regression models were employed to identify the determinants of long-duration commuting trips. The statistical package SPSS 19.0 was utilized to estimate the models. Model 1 provides estimation results after all variables were entered. The prediction accuracy of Model 1 for the training data is 64.2%. Based on the test data, Model 1 performs a prediction
ARTICLE IN PRESS
Fig. 1. The map of the urban area of Kunming.
accuracy of 63.8% with a sensitivity of 55.6% and a specificity of 66.1%. Model 2 excludes non-significant variables by a stepwise method (sig. <0.1). The prediction accuracy of Model 2 for the training data is 63.4%. Based on the test data, Model 2 performs a prediction accuracy of 63.2% with a sensitivity of 61.1% and a specificity of 63.8%. The estimation results of the models are shown in Table 4. The significance
Table 1
Profile of the sample.
Variables Classification Cases Percentage
Commuting time 40 min and more 177 23.5%
Less than 40 min 576 76.5%
Gender Male 417 55.4%
Female 336 44.6%
Age 30 and younger than 30 320 42.5%
More than 30 433 57.5%
Education level High school or below 301 40.0%
College or above 452 60.0%
Occupation Public institution staff or civil 109 14.5%
servant
Others 644 85.5%
Annual household income Less than 50,000 413 54.8%
(CNY) 50,000 and more 340 45.2%
Number of workers in a Single worker 196 26.0%
household Two or more workers 557 74.0%
Presence of children With children aged < 13 years 215 28.6%
Without children aged <13 538 71.4%
Presence of retirees With retirees 206 27.4%
Without retirees 547 72.6%
Residential location Center area 331 44.0%
Peripheral area 422 56.0%
test indicates that age, education level, number of workers, presence of retirees, and residential location have a significant impact on the occurrence of long-duration commuting trips.
In terms of individual attributes, commuters younger than 30 have a lower possibility of experiencing commuting time longer than or equal to 40 min, which is only 0.674 time that of commuters older than 30. One of the major reasons for this finding is that many young Chinese fail to own their own houses because of high house prices; most of them merely rent houses. In 2014, the average house price in Kunming was 8206 CNY/km2 [30], while our sample shows that average monthly personal income of young people was 3455 CNY. The gap between house price and income is huge. As a result, many young people can't afford to buy a house. Thus, they consider commuting costs and tend to rent houses near their workplaces. In addition, many young people move into the city from rural area. These rural migrant workers are employed in low-skill and low-income jobs and can be replaced easily.
Table 2
Distribution of actual commuting time.
Time interval (min) Cases Percentage Cumulative percentage
<10 34 4.5% 4.5%
[10, 20) 149 19.8% 24.3%
[20, 30) 190 25.2% 49.5%
[30, 40) 203 27.0% 76.5%
[40, 50) 113 15.0% 91.5%
[50, 60) 29 3.9% 95.4%
[60, 70) 23 3.1% 98.4%
[70, 80) 3 0.4% 98.8%
[80, 90) 1 0.1% 98.9%
>90 8 1.1% 100.0%
ARTICLE IN PRESS
M. He, S. Zhao / IATSS Research xxx (2016) xxx-xxx
Table 3
Definitions of the variables used in the logistic regression models. Variables Definitions
LT40 1 if the commuting time is more than or equal to 40 min; 0 otherwise
Gender 1 if the person is a male; 0 otherwise Age 1 if the person is 30 or younger; 0 otherwise
Education 1 if the person has a college education or above; 0 otherwise Occupation 1 if the person is a public institution staff or civil servant; 0 otherwise H_income 1 if the annual household income is less than 50,000; 0 otherwise Nworker 1 if the household has one worker; 0 otherwise Children 1 if the household has at least one child younger than 13 years; 0 otherwise
Retiree 1 if the household has at least one retiree; 0 otherwise R_Location 1 if the person lives in the urban central area (within the Second Ring Road); 0 otherwise
Most of these workers opt to rent houses near their workplaces or merely look for jobs near their houses. Education level also has a significant impact on whether people experience long-duration commuting trips. Commuters with college education or above have a higher possibility of experiencing commuting time longer than or equal to 40 min, which is 2.157 times that of commuters with high school education or below. High education level often leads to better-paid jobs, which can compensate for commute costs [31]. In this study, the coefficient of the variable gender in Model 1 is negative, with sig. = 0.232 (>0.1), and excluded in Model 2 by the significance test. This result indicates that no statistically significant difference exists between females and males.
In terms of household characteristics, commuters in one-worker households have a lower possibility of experiencing commuting time longer than or equal to 40 min, which is 0.534 time that of commuters in households with two or more workers. The reason is that it is difficult for a household with two or more workers to select residential locations close to both workplaces. Commuters in a household with retirees have a higher possibility of experiencing commuting time longer than or equal to 40 min, which is 1.705 times that of commuters in households without retirees. This result could be due to the fact that retirees share household responsibilities so that workers can spend more time on commuting. In this study, the coefficient of the variable Children in Model 1 is negative, with sig. = 0.274 (>0.1), and excluded in Model 2 by the significance test. This result indicates that although commuters in a household with children have a lower possibility of experiencing long-duration commuting trips, no statistically significant difference exists between commuters in a household with children and those in a household without children.
The results also show that residential location has a significant impact on the occurrence of long-duration commuting trips. Commuters living in the urban central area have a lower possibility of experiencing commuting time longer than or equal to 40 min. Compare with urban peripheral area, urban central area has higher density land use pattern
Table 4
Estimation results of the logistic regression models.
Variables Model 1 Model 2
B Wald Sig. Exp (B) B Wald Sig. Exp (B)
ASC - 0.934 8.823 0.003 0.393 -1.311 39.755 0.000 0.269
Gender - 0.248 1.430 0.232 0.780 - - - -
Age - 0.464 3.419 0.064 0.629 - 0.394 3.267 0.071 0.674
Education 0.718 6.340 0.012 2.050 0.768 10.167 0.001 2.157
Occupation - 0.027 0.009 0.926 0.973 - - - -
H_income - 0.255 1.177 0.278 0.775 - - - -
Nworker - 0.578 4.400 0.036 0.561 - 0.628 5.466 0.019 0.534
Children - 0.279 1.198 0.274 0.756 - - - -
Retiree 0.630 7.150 0.007 1.877 0.534 6.049 0.014 1.705
R Location - 0.525 6.177 0.013 0.592 - 0.518 6.111 0.013 0.596
Cox & Snell R2 0.066 0.061
Nagelkerke R2 0.099 0.092
and higher concentration of workplaces, commercial and service facilities. As mentioned above, the urban area of Kunming supplies 1.7 million jobs, and about 0.9 million jobs are provided within the central area. Higher job density and more employment opportunities in the urban central area result in shorter commuting times for the residents living in this area. A previous study also confirmed that commuters living in job-rich areas have shorter commuting time [5].
4. Long-duration commuters' perceptions and attitudes towards commuting time
4.1. Two indicators
The concept of constant travel time budget originated from Zahavi [32] and has been supported by much previous literature [33,34]. However, the concept that travel time budget keeps a constant also is subject to some controversy [35-37]. Regardless of whether stability exists in the daily travel time budget or not, for commuters, the increase in commuting time inevitably means using other travel time budgets and further affects participation in other activities, such as leisure, shopping, and family activities. Commuting cost is composed of Commuting time and derived travel expenses. With the increase in commuting time, travelers will be subjected to higher commuting costs. The length of commuting time also has a significant effect on commuters' satisfaction, security risk, and efficiency during office hours. Thus, for individuals, a tolerance threshold of commuting time must exist at cognitive and mental levels. The tolerance threshold of commuting time represents the maximum amount of one-way commuting time that an individual can tolerate.
Meanwhile, commuters also perceive an ideal commuting time at cognitive and mental levels. Under allowable conditions, commuters are likely to pursue their ideal commuting time by changing their residential location, workplace, or travel mode [38]. In modeling travel behavior, travel is commonly assumed to be a source of disutility; hence, travelers attempt to minimize their travel time and cost [39]. However, some researchers noted that travel can provide positive utility in its own right [40-43]. Specifically, with respect to commuting, such positive utility may mainly manifest in the transition between home and work and obtaining a better dwelling condition and more available job opportunities by spending more commuting time [44-46]. Commuting time is not unequivocally a source of disutility to be minimized, but rather offers some benefits; most commuters have a non-zero ideal commuting time [39]. Commuters may seek to optimize commuting time rather than minimize it.
The ideal commuting time and the tolerance threshold of commuting time are of great significance for urban and land use planning. With the rapid urbanization of China, significant changes in urban land use and spatial layout arise. For example, some cities gradually change from "single center + ring roads" to a multi-center pattern; there are increasing trends of jobs-housing mismatch and suburbanization. Thus, several important issues require in-depth studies, such as land use patterns, the radiation scope of a central city, space-time relationship between main center and new towns. Studies on these issues should consider the ideal commuting time and tolerance threshold of commuting time. In addition, the ideal commuting time and tolerance threshold of commuting time can serve as a reference in the study of travel mode choice, residential location choice, job accessibility, and travel subjective well-being.
4.2. Statistical analysis
In the questionnaire, the tolerance threshold of commuting time was set to nine intervals. The tolerance threshold of commuting time in each interval was as follows: ifit falls within 20 min, it is noted as 15 min; ifit falls in the interval of 90 min or more, it is noted as 90 min; and ifit falls in other intervals, it is noted as the mid-point of the interval. For
ARTICLE IN PRESS
M. He, S. Zhao / IATSS Research xxx (2016) xxx-xxx 5
Statistics of ACT, ICT, and TTCT1.
ACT TTCT ICT
Mean value (min) Standard deviation Mean value (min) Standard deviation Mean value (min) Standard deviation
Long-duration commuters 48.8 14.2 47.1 14.8 24.4 10.0
Short-duration commutersb 22.6 8.2 34.5 13.0 16.9 8.1
All 28.7 14.9 37.4 14.4 18.6 9.1
a Actual commuting time is abbreviated as ACT, ideal commuting time is abbreviated as ICT, and tolerance threshold of commuting time is abbreviated as TTCT. The same applies to the next table.
b Short-duration commuters are defined as people who have less than 40 min one-way commuting time.
example, if it falls within the interval [20, 30), it is noted as 25 min. As shown in Table 5, for all commuters, the average tolerance threshold of commuting time is 37.4 min. The average tolerance threshold of commuting time of long-duration commuters is greater than that of short-duration commuters (47.1 min vs. 34.5 min). As shown in Fig. 2, the tolerance threshold of commuting time of long-duration commuters peaks at the interval of [50, 60), accounting for 37.3%. About half (49.2%) of long-duration commuters have a tolerance threshold of commuting time of 50 min or more. Generally, a commuting time of 60 min on a daily basis was considered a maximum for most Chinese people. Our results also show that only 5.3% of all commuters and 12.5% of long-duration commuters have a tolerance threshold of commuting time of 60 min or more. The majority of the long-duration commuters with a tolerance threshold of commuting time of 50 min or more fall in the interval of [50, 60). The tolerance threshold of commuting time of short-duration commuters peaks at the interval of [30, 40), accounting for 30.6%. Short-duration commuters with a tolerance threshold of commuting time that is less than 50 min account for 90.5% of all short-duration commuters.
In the questionnaire, ideal commuting time was set to seven intervals. The ideal commuting time in each interval was as follows: if it falls within 10 min, it is noted as 5 min; if it falls in the interval of 60 min or more, it is noted as 60 min; and if it falls in other intervals, it is noted as the mid-point of the interval. For example, if it falls within the interval [10, 20), it is noted as 15 min. For all commuters, the average ideal commuting time is 18.6 min. The average ideal commuting time of long-duration commuters is greater than that of short-duration commuters (24.4 min vs. 16.9 min). As shown in Fig. 3, the ideal commuting time of long-duration commuters peaks at the interval of [20, 30), accounting for 40.7%. The ideal commuting time of short-duration commuters peaks at the interval of [10, 20), accounting for 47.7%. Long-duration commuters with an ideal commuting time of less than 30 min account for 73.5% of all long-duration commuters;
short-duration commuters with an ideal commuting time of less than 30 min account for 95.6% of all short-duration commuters.
Overall, a correlation exists among actual commuting time, the tolerance threshold of commuting time, and ideal commuting time (see Table 6). The tolerance threshold of commuting time and ideal commuting time of long-duration commuters are greater than those of short-duration commuters. Meanwhile, ideal commuting time does not always tend to be 0, which supports the previous findings that commuting can offer positive utility [38,39].
The scatter plot of actual versus ideal commuting times is shown in Fig. 4. As it is, an overwhelming majority of long-duration commuters fall under the 45 degree line, which indicates that most long-duration commuters have shorter ideal commuting times than actual commuting times. The scatter plot of tolerance thresholds versus actual commuting times are shown in Fig. 5. As it is, the scatter of points is disperse, with a majority of them falling above the 45 degree line. Compared with long-duration commuters, more short-duration commuters fall above the 45 degree line. Further, the comparison of actual and ideal commuting time, and tolerance threshold of commuting time is presented in Table 7. If the ideal commuting time or tolerance threshold of commuting time and the actual commuting time belong to the same interval, then the ideal commuting time or tolerance threshold of commuting time is approximately equal to actual commuting time. As shown in Table 5, for the long-duration commuters, the actual commuting time of 97.2% long-duration commuters is greater than the ideal commuting time. This result indicates that most long-duration commuters are dissatisfied with their commuting time and expect to shorten it. The actual commuting time of 40.1% long-duration commuters exceeds the tolerance threshold; these commuters are eager to reduce their commuting time. To achieve a reasonable travel time, Travelers may decrease travel distances and/or increase travel speed [47,48]. Therefore, commuters have three means of reducing commuting time, namely, changing residential location, working place, or travel mode. They will be more likely
40% 35% 30% 25% 20% 15% 10% 5% 0%
37.' % • Long »-duration c ommuter
28 3% 30.6% -■-Shor t-duration c ommuter
18.1% 17.5%
13.C %
11.1% 6.4% 7.3% 9
1.7 % 2.1% 2.3% 2.3% C 0.6%
0.2% 0.3% 0.5%
<20 [20,30) [30,40) [40,50) [50,60) [60,70) [70,80) [80,90) >90
Time interval (min)
Fig. 2. Distribution of the tolerance threshold of commuting Time.
ARTICLE IN PRESS
M. He, S. Zhao / IATSS Research xxx (2016) xxx-xxx
47.7% ■Long-duratic n commuter
40.7% ■Short-duratk >n commuter
19.3% 26.6% i 22.6% i
6.2% 3.6% 2.3% 1.1% 0.6%
0.0% 0.0%
<10 [10,20) [20,30) [30,40) [40,50) [50,60) >60
Time interval (min)
Fig. 3. Distribution of ideal commuting time.
Table 6
Correlation matrix.
ACT ICT TTCT
ACT 1.000 0.400** 0.515**
ICT 1.000 0.524**
TTCT 1.000
** Correlation is significant at the 0.01 level (2-tailed).
to choose faster travel mode. [f there is no fast and reliable public transport, such as rail transit, it is very possible to promote them to choose private cars. Urban land use and transport planning should consider the willingness of commuters and these possible changes in order to adopt reasonable policies.
5. Conclusion
Understanding the commuting patterns of long-duration commuters and the possible changes can help policymakers adopt the more reasonable land use and transportation policies. With Kunming in China as an example, the determinants of long-duration commuting trips were identified based on logistic regression model. The research results indicated that age, education level, number of workers in a household, presence of retirees, and residential location have a significant impact on the occurrence of long-duration commuting trips. [f a commuter is younger than 30, belongs to one-worker household, or
lives in the urban central area, he/she has a lower possibility of experiencing commuting time longer than or equal to 40 min. Commuters with an education level of college or above have a higher possibility of experiencing commuting time longer than or equal to 40 min.
As mentioned above, differences in family structure and social roles of females exist between China and western countries. This paper identified whether gender, presence of children, and presence of retirees have a significant impact on long-duration commuting choice. The results indicated that gender and presence of children have no significant impact on the occurrence of long-duration commuting trips. Commuters in a household with retirees have a higher possibility of experiencing commuting time longer than or equal to 40 min.
The ideal commuting time and tolerance thresholds of commuting time of long-duration commuters were also investigated. The average tolerance threshold of commuting time and the average ideal commuting time of long-duration commuters are greater than those of short-duration commuters. [deal commuting time does not always tend to be 0, which supports the previous finding that commuting can offer positive utility. The actual commuting time of 97.2% long-duration commuters is greater than the ideal commuting time, which indicates that most long-duration commuters are dissatisfied with their commuting time and expect to shorten it. The actual commuting time of 40.1% long-duration commuters exceeds the tolerance threshold; these commuters are eager to reduce their commuting time. Urban land use and transportation planning should consider the willingness of commuters and these possible changes in the commuting behaviors and residential and job location choices.
ACT (min)
Fig. 4. Scatter plot oflCT versus ACT (n = 753).
ARTICLE IN PRESS
M. He, S. Zhao / IATSS Research xxx (2016) xxx-xxx 7
ACT (min)
Fig. 5. Scatter plot ofICT versus TTCT (n = 753).
Comparison of ACT, ICT, and TTCT.
ACT < ICT ACT к ICT ACT > TTCT ACT к TTCT
Case Percentage Case Percentage Case Percentage Case Percentage
Long-duration commuters 2 1.13% 3 1.69% 71 40.1% 45 25.4%
Short-duration commuters 61 10.6% 150 26.0% 44 7.6% 173 30.0%
All 63 8.4% 153 20.3% 115 15.3% 218 29.0%
This study is a preliminary investigation on the long-duration [5 commuting behaviors of China's urban residents. The focus was on the effects of the socio-economic characteristics of commuters on long- [6 duration commuting trips and the descriptive statistics of ideal [7 commuting time and tolerance thresholds of commuting time of long-duration commuters. Further study should be conducted to identify [8 the heterogeneity of ideal commuting time and tolerance thresholds [9 of commuting time. Meanwhile, future studies should take different urban sizes and urban structures as case studies so as to acquire a thorough understanding of long-duration commuting behaviors. Furthermore, to achieve the expected commuting time, long-duration commuters could make changes in their commuting behaviors and job-housing locations. The effects of these changes on transportation systems and land use should be further studied to obtain valuable policy implications for their improvements.
Acknowledgments
This research is supported by the National Natural Science Foundation of China (NO. 51278087).
References
[1] M. Van Ham, P. Hooimeijer, Regional differences in spatial flexibility: long commutes and job related migration intentions in the Netherlands, Appl. Spat. Anal. Policy 2 (2) (2009) 129-146.
[2] P. Zhao, B. Lu, G. de Roo, Impact of the jobs-housing balance on urban commuting in Beijing in the transformation era, J. Transp. Geogr. 19(2011)59-69.
[3] R Cervero, Jobs-housing balancing and regional mobility, J. Am. Plan. Assoc. 55 (1989) 136-150.
[4] A. Downs, Stuck in Traffic: Coping With Peak-hour Traffic Congestion, The Brookings Institution, Washington, 1992.
D.M. Levinson, Accessibility and the journey to work J. Transp. Geogr. 6 (1) (1998) 11-21.
W.H. Horner, Extensions to the concept of excess commuting, Environ. Plan. A 34 (3) (2002) 543-566.
G. Giuliano, K. Small, Is the journey to work explained by urban structure? Urban Stud. 30 (9) (1993) 1485-1500.
R Ewing, Beyond density, mode choice, and single trips, Transp. Q. 49 (4) (1995) 15-24.
B.S. Lee, J.F. McDonald, Determinants of commuting time and distance for Seoul residents : the impact of family status on the commuting of women, Urban Stud. 40 (7) (2003) 1283-1302.
G. Giuliano, Information technology, work patterns and intrametropolitan location: a case study, Urban Stud. 35 (1998) 1077-1095.
J.F. Madden, Why women work closer to home, Urban Stud. 18 (1981) 181-194.
0. Blumen, Gender differences in the journey to work, Urban Geogr. 16 (1994) 223-225.
M.W. Sermons, F.S. Koppelman, Representing the differences between female and male commute behavior in residential location choice models, J. Transp. Geogr. 9 (2) (2001) 101-110.
J. Prashker, Y. Shiftan, P. Hershkovitch-Sarusi, Residential choice location, gender and the commute trip to work in Tel Aviv, J. Transp. Geogr. 16 (2008) 332-341. T. Turner, D. Niemeier, Travel to work and household responsibility: new evidence, Transportation 24 (1997) 397-419.
D. Wang, Y. Chai, The jobs-housing relationship and commuting in Beijing, China: the legacy of danwei, J. Transp. Geogr. 17 (2009) 30-38.
M.W. He, S.C. Zhao, M. He, Determinants of Commute Time and Distance for Urban Residents: A Case Study in Kunming, China, CICTP2015, © ASCE, 2015 3663-3673.
V. Preston, S. MCLafferty, E. Hamilton, The impact of family status on black, white, and Hispanic women's commuting, Urban Geogr. 14 (3) (1993) 228-250. L.D. Singell, J.H. Lillydahl, An empirical analysis of the commute to work patterns of males and females in two-earner households, Urban Stud. 23 (2) (1986) 119-129.
1. Johnston-Anumonwo, The influence of household type on gender differences in work trip distance, Prof. Geogr. 44 (2) (1992) 161-169.
S. Kim, Excess commuting for 2-worker households in the Los Angeles metropolitan area, J. Urban Econ. 38 (2) (1995) 166-182.
World Bank, Labor Participation Rate, Female of the World, World Bank Washington, 2013.
C.C. Fan, J. Regulska, Gender and the labor market in China and Poland, in: J.R. Logan (Ed.), Urban China in Transition, Blackwell Publishing, Oxford 2008, pp. 89-111.
ARTICLE IN PRESS
8 M. He, S. Zhao / IATSS R
[24] J. Feng, M. Dijst, B. Wissink, J. Prillwitz, The impacts of household structure on the travel behavior of seniors and young parents in China, J. Transp. Geogr. 30 (2013) 117-126.
[25] SCPRC, Regulations for Retirement, The State Council of the People's Republic of China, Beijing, 1978.
[26] Kunming Statistics Bureau, 2014 Kunming National Economic and Social Development Statistics Bulletin, Kunming Statistics Bureau, Kunming, 2014.
[27] P. Nsss, Residential location, travel, and energy use in the Hangzhou metropolitan area, J. Transp. Land Use 23 (3) (2010) 27-59.
[28] Kunming university of science and technology, Passenger Flow Forecast of the 2st Phase Project of Kunming Metro Line 2, Kunming university of science and technology, Kunming, 2014.
[29] S.H. Cassel, Z. Macuchova, N. Rudholm, A. Rydell, Willingness to commute long distance among job seekers in Dalarna, Sweden, J. Transp. Geogr. 28 (2013) 49-55.
[30] Sohu Focus, Kunming annual market report: data analysis of housing market, http:// news.focus.cn/km/2015-01-13/5958011_1.html. 2015 (accessed 16. 06.25).
[31] E. Sandow, Commuting behavior in sparsely populated areas: evidence from Northern Sweden, J. Transp. Geogr. 16 (2008) 14-27.
[32] Y. Zahavi, UMOT Project, Report D0T-RSPA-DPB-20-79-3, U.S. Department of Transportation and Ministry of Transport, Federal Republic of Germany, 1979.
[33] P. Gordon, H. Richardson, M. Jun, The commuting paradox: evidence from the top twenty, J. Am. Plan. Assoc. 57 (1991) 416-420.
[34] A. Schafer, Regularities in travel demand: an international perspective, J. Transp. Stat. 3 (2000) 1-31.
[35] C.L. Purvis, [. Pas, Changes in regional travel characteristics and travel expenditures in San Francisco Bay area: 1960-1990, Transp. Res. Rec. 1466 (1994) 99-110.
[36] A. Kumar, D. Levinson, Temporal variations on allocation of time, Transp. Res. Rec. 1493 (1995) 118-127.
ch xxx (2016) xxx-xxx
[37] L. Toole-Holt, S.E. Polzin, R.M. Pendyala, Two minutes per person per day each year: an exploration of the growth in travel time expenditures, Transp. Res. Rec. 1917 (2005) 45-53.
[38] M.W. He, S.C. Zhao, M. He, [deal commute time analysis based on traveler's cognition, J. Transp. Syst. Eng. Inf. Technol. 15 (4) (2015) 161-165.
[39] L.S. Redmond, P.L. Mokhtarian, The positive utility of the commute: modeling ideal commute time and relative desired commute amount, Transportation 28 (2001) 179-205.
[40] [. Salomon, P.L. Mokhtarian, What happens when mobility-inclined market segments face accessibility-enhancing policies? Transp. Res. D 3 (3) (1998) 129-140.
[41 ] L. Steg, Car use: lust and must: instrumental, symbolic and affective motives for car use, Transp. Res. A 39 (2005) 147-162.
[42] P.L Mokhtarian, C. Chen, TTB or not TTB, that is the question: a review and analysis of the empirical literature on travel time (and money) budgets, Transp. Res. A 38 (2004) 643-675.
[43] C. Jakobsson Bergstad, A. Gamble, O. Hagman, M. Polk, T. Garling, L.E. Olsson, Affective-symbolic and instrumental-independence psychological motives mediating effects of socio-demographic variables on daily car use, J. Transp. Geogr. 19 (2011) 33-38.
[44] J.Jain, G. Lyons, The gift of travel time, J. Transp. Geogr. 16 (2008) 81-89.
[45] J. Van Ommeren, P. Rietveld, P. Nijkamp, Commuting: in search of jobs and residences, J. Urban Econ. 42 (1997) 402-421.
[46] P.L. Mokhtarian, [. Salomon, How derived is the demand for travel? Some conceptual and measurement considerations, Transp. Res. A 35 (2001) 695-719.
[47] D. Banister, The trilogy of distance, speed and time, J. Transp. Geogr. 19 (2011) 950-959.
[48] D.S. Vale, Does commuting time tolerance impede sustainable urban mobility? analyzing the impacts on commuting behavior as a result of workplace relocation to a mixed-use centre in Lisbon, J. Transp. Geogr. 32 (2013) 38-48.
