Discrete/Continuous Model for Motorcycle Ownership and Utilization Across Income Groups in a Vietnamese Medium-Sized Motorcycle-Dependent City

Abstract

Understanding vehicle choice and utilization is increasingly considered in addressing motorcycle-related challenges by policymakers. However, the linkage between the discrete-continuous decisions has not been examined across income groups. This study measures the interdependence between vehicle choice and utilization for income groups in a Vietnamese medium-sized city. Model estimation results show that contributing factors to households’ motorcycle ownership include individual and household characteristics and some aspects of the built environment. However, differences exist in motorcycle ownership and utilization between low- and medium-income groups. Results from this study may suggest that households may be more likely to buy an extra motorcycle as their incomes increase and this may generate more motorcycle-related issues in the future. On the positive side, we found that the non-motorized mobility resources available in the low-income group played a key role in reducing motorcycle ownership and usage. Therefore, we suggest that medium-sized cities should develop cycling infrastructure to facilitate the mobility of low-income groups as well as promote well-established travel behavior from cyclists; then, they may be less willing to switch to motorized vehicles as their income increases.

Keywords

planning and analysis mode choices pattern (behavior choices etc.)

The motorcycle is the most popular mode in Vietnam, and the number of registered motorcycles has exploded over the last three decades with over 65 million motorcycles in 2020 compared to 1.2 million in 1990 ( 1 ). In Vietnamese cities, where the major of households live in small alleys of less than 5 m, a motorcycle is considered to be a more flexible, more convenient, and faster transportation mode for door-to-door travel compared to a bike and transit, and a much cheaper mode compared to a car. As a result, many Vietnamese cities have substandard public transit systems, and major urban travel patterns are now dominated by motorcycles. Motorcycle ownership rates are around 600–700 motorcycles per 1000 inhabitants in so-called motorcycle-dependent cities ( 2 ).

Recently, the sharp rise in motorcycle ownership and usage has attracted the attention of policymakers and researchers because of its contribution to the unique transportation challenges in many Asian countries, including Vietnam. These challenges include the sharp decline in non-motorized transport (NMT) and public transport usage, particularly for medium-sized cities; high involvement of motorcycles in traffic accidents and fatalities; high contribution of motorcycles to congestion and pollution; and high frequency of motorcycle robberies ( 3 – 6 ). These challenges put high pressure on policymakers and push them to make more effort to reduce motorcycle-related accidents and congestion as well as avoid motorcycle snatch theft. Many studies in Vietnam claim there are several solutions to reduce motorcycle usage, such as the promotion of public transit ( 7 – 9 ), increasing travel time by motorcycles compared to public transit ( 10 ), increasing parking fees ( 10 , 11 ), and promoting active travel modes ( 12 ). However, one has shown that these solutions cannot significantly decrease motorcycle usage because motorcyclists’ travel behavior is well-established and they are unwilling to switch to other modes ( 13 ). This means that motorcycles continue to play a significant role in residents’ lives for improving mobility and combating poverty in the future. Therefore, before addressing further motorcycle-related challenges, it is necessary to understand residents’ decisions of ownership and utilization in the current context.

Previous research on motorcycle ownership and utilization, while plentiful, has not addressed these issues because it has not adequately accounted for the dynamic aspects of motorcycle ownership. In addition, such research failed to consider the relationship between household decisions with respect to the number of vehicles to own and the extent to which vehicles are used.

The purpose of this paper is to analyze the determinants of household motorcycle ownership and utilization across income groups to help understand the deeper aspects of travel behavior. In particular, the following questions need to be addressed: (1) How responsive, with respect to motorcycle ownership, would households be to changes in household characteristics and the built environment? (2) To what extent are household members' preferences toward motorcycle usage? Given the inherent challenges of motorcycles, an understanding of the linkage between motorcycle ownership and utilization is crucial in supporting policymakers in Vietnam.

We chose to define the two income groups in this study as low- and medium-income rather than high-income groups because of the possibility that they may be more sensitive to shift to environmental-friendly modes. The income levels are defined based on the World Bank’s Lower-Middle Income Class poverty line (i.e., USD 1.9/day 2011 purchasing power parity, PPP, for extreme poverty or USD 3.2/day 2011 PPP for lower-middle income class). Adapted to the context of Can Tho city, the low-income threshold is defined in this study as the regional minimum wage in Can Tho. The medium-income threshold is defined as from 30% to 100% of the increased value compared to the regional minimum wage. Therefore, the number of low-income households is taken as the sum of all household members belonging to or below the low-income threshold.

For the analysis, household travel survey data are used. The data were collected by a consulting company through paper surveys in Can Tho, a medium-sized city in Vietnam. For model estimation, six discrete-continuous models were considered: (1) the low-income household’s decision to own one motorcycle (LHH-one) and the corresponding utilization of the vehicle (i.e., kilometers per day); (2) the medium-income household’s decision to own one motorcycle (MHH-one) and the corresponding utilization of the vehicle; (3) the low-income household’s decision to own two motorcycles (LHH-two) and the corresponding utilization of the vehicles; (4) the medium-income household’s decision to own two motorcycles (MHH-two) and the corresponding utilization of the vehicles; (5) the low-income household’s decision to own three motorcycles (LHH-three) or more and the corresponding utilization of the vehicles; (6) the medium-income household’s decision to own three motorcycles or more (MHH-three) and the corresponding utilization of the vehicles. The discrete component of the choices (i.e., the decision to own a motorcycle) is modeled using the binary logit structure. The continuous components of the choices (i.e., the utilization of the vehicle) are each modeled using a linear regression structure with the natural logarithm of the corresponding utilization as the choice variable.

The paper is structured as follows. In the second section, we review the previous studies on factors influencing vehicle ownership and mode choice. In the methodology in the third section, we present the models of household motorcycle ownership and utilization. The data is described in the fourth section. In the fifth section we report and analyze estimation results for single-vehicle and two-vehicle households. Finally, the sixth section contains the conclusions.

Literature

The literature includes a large volume of empirical research on the effect of income on households’ decisions in holding and using vehicles. Evidence from the existing literature shows that income is a widely used indicator of a household’s ownership and mode share ( 14 – 19 ).

Apart from income, determinants of mode choice have been identified by demographic characteristics, trip characteristics, and the built environment. More specifically, demographic characteristics include gender, age, occupation, and household size ( 16 , 20 –22). Trip characteristics involve trip length, travel time, travel cost, and trip purpose ( 23 – 25 ). The built environment contains land use, density, urban design, accessibility, and proximity to a public transit station ( 26 – 32 ).

For model estimation, a wide range of modeling methods has been applied to predict vehicle ownership and utilization. Because of the discrete nature of vehicle selection outcomes, multinomial logit models have been widely applied to investigate the effects of significant factors in vehicle choice ( 32 – 36 ). Alternatively, nested logit models have also been employed to partially address the endogenous correlations among different mode choice outcomes ( 37 , 38 ). Moreover, considering the intuitive ordering of mode choice outcomes, ordered logit and probit models are used ( 15 ). Recently, unobserved heterogeneity has received growing concern. Random parameter approaches and their variants, such as latent class models ( 39 , 40 ) and mixed logit models with heterogeneity in means, variances, or both ( 41 ), have been applied to cope with the unobserved heterogeneity in travel behavior analysis.

In the specific context of motorcycles, motorcycle ownership and usage are explored through both aggregate and disaggregate models. While disaggregate models seek to understand the individual and household characteristics that affect motorcycles’ ownership and usage, and they found that income, gender, age, and spatial trip characteristics have a significant association with motorcycle ownership and usage ( 42 – 44 ), some aggregate approaches seek to determine the trends in motorcycle ownership and usage in a city or country ( 17 , 45 ). Some studies have compared motorcycle and car ownership characteristics (e.g., Law et al. [ 17 ], Sanko et al. [ 46 ], Wong [ 47 ], and Lai et al. [ 48 ]) and concluded that car and motorcycle ownership are not independent. More specifically, travel mode choice in Vietnamese cities has been traditionally modeled with a discrete choice framework using multinomial logit models ( 7 , 13 ), probit models ( 49 ), and latent class models ( 9 ). These studies provide a good understanding of contributing factors associated with mode choice outcomes in Vietnamese cities. However, these studies have not considered the linkage between households’ decisions, such as whether households’ decide to purchase motorcycles and how they utilize them. The lack of understanding of the linkage between households’ decisions in motorcycle-dependent cities is the motivation of this study.

The advantage of joint models over single models has also been examined in several previous studies. For example, Ermagun et al. ( 50 , 51 ) found that joint models perform better as compared to single models when they were applied to explore activity type decisions. Similarly, Spissu et al. ( 52 ) highlighted that the mobility sources available in the household and usage were significantly interrelated. Therefore, the joint model will provide an improvement over the mode choice model.

The connection between vehicle ownership and utilization has received extensive attention as researchers seek to confirm the linkages between them. Efforts to examine the interdependencies between a household’s vehicle ownership and travel-related decisions have been accommodated explicitly ( 31 , 37 , 52 –56). As a result, they contribute to accurately forecasting travel patterns and evaluating the impacts of policy actions. However, these studies do not allow for understanding the linkages in household decisions across the population subgroups defined by income groups. Therefore, it strongly motivates follow-up studies to conduct more in-depth investigations surrounding the relationship between vehicle ownership and mode choice across income groups in emerging cities.

Methods

Model Selection

This study applied a reduced-form model structure to consider the household’s choice to own a motorcycle (discrete) and the corresponding utilization of the vehicle. Here we expect the utility of owning a motorcycle is a function of household characteristics, the built environment, and a disturbance term. Since we use a reduced-form structure, the utility will also be dependent on the continuous variable, utilization. Let i represent the index of four discrete-continuous choices, which can be one of the following: (1) LHH own one motorcycle; (2) LHH own two motorcycles; (3) LHH own three or more motorcycles; (4) MHH own one motorcycle; (5) MHH own two or more motorcycles; and (6) MHH own three or more motorcycles. The discrete choice of motorcycle ownership is as follows:

U_{i} = β_{i} Z_{i} + ϕ y_{i} + ε_{i}

(1)

where U_i is the utility that household type I derives from owning a motorcycle, Z_i is the vector of household and built environment characteristics, y_i is the utilization of a motorcycle with the corresponding household’s motorcycle ownership (i.e., kilometers per day), ε_i is a disturbance term accounting for unobserved effects, β_i is a vector of estimable parameters, φ is an estimable parameter, and the continuous choice of utilization is as follows:

y_{i} = α_{i} X_{i} + η_{i}

(2)

where X_i is a vector of household, individual, travel, and built environment characteristics, η_i is the unobserved household, individual, travel, and built environment characteristics, and α_i is a vector of estimable parameters.

The substitution of Equation 2 into Equation 1 giving the “reduced-form” utility ( 57 ), in which the overall utility is developed as the joint utility of the discrete-continuous choice:

U_{i} = β_{i} Z_{i} + ϕ α X_{i} + ϕ η_{i} + ε_{i}

(3)

Mannering and Hensher ( 57 ) and Train ( 58 ) showed a joint discrete-continuous cause of the estimation bias between error terms (e.g., ε_i in Equation 1 and u_i in Equation 2). The correction of the estimation bias could take in two major forms: (1) the indirect method, which regresses the endogenous variable in Equation 2 with a series of exogenous variables, and the value of the endogenous variable predicted by the resulting regression equation would then be used in the estimation of Equation 2 ( 58 – 60 ); (2) the direct method, which provides for an explicit interaction between the discrete and continuous components through the bias correction term, expected value, and full information ( 57 ).

This study applies the direct method with the bias correction approach developed by Heckman ( 61 ). This approach seeks to develop a single term that can be used to correct the selectivity bias in continuous equations. To demonstrate, the form of the Heckman two-step method is presented below:

y_{j} = α_{j} X_{j} + η_{j}

(4)

with

V_{j} = β_{j} Z_{j} + ε_{j}

(5)

where V_j is a binary variable (select or not), with y_i only observed when V_j = 1. Here, η_j and ε_j are error terms that follow a bivariate normal distribution:

[\begin{matrix} η_{j} \\ ε_{j} \end{matrix}] ~ N [\begin{matrix} σ^{2} & ρ σ \\ ρ σ & 1 \end{matrix}]

where σ is the scale parameter and ρ is the correlation coefficient.

Equation 4 is generally referred to as the response equation, with y_j the variable of interest. Equation 5 is termed the selection equation and determines whether y_j is observed or not:

E (y_{j} | V_{j} = 1) = α_{j} X_{j} + ρ σ λ_{j} (β_{j} Z_{j})

(6)

where λ(X) = ϕ(X)/φ(X) is the inverse Mills ratio ( 62 ) and ϕ and φ are the standard normal density and cumulative distribution function, respectively. Then we may specify a regression model:

y_{j} = α_{j} X_{j} + ρ σ λ_{j} (β_{j} Z_{j}) + η_{j}

(7)

The two-step method proceeds by first estimating a probit regression for Equation 5 to obtain an estimate of β^*_j, from which λ_j(β^*_j Z_j) may be calculated. A least-squares regression of y_j on $α_{j} X_{j}$ and β^*_j:

y_{j} = α_{j} X_{j} + ρ σ λ_{j}^{*} + η_{j}

(8)

is then computed, yielding consistent estimates of α and θ = ρσ.

The estimator of the coefficient covariance matrix is given by the following:

Φ = {\hat{σ}}^{2} {(X^{*^{'}} X^{*})}^{- 1} (X^{*^{'}} (I - {\hat{ρ}}^{2} \hat{Δ}) X^{*} + Q) {(X^{*^{'}} X^{*})}^{- 1}

(9)

where $X_{j}^{*} = (X_{j}, \underset{j}{\hat{λ}})^{'}$ , $\hat{Δ}$ is a diagonal matrix with $\underset{j}{\hat{δ}} = \underset{j}{\hat{λ}} (\underset{j}{\hat{λ}} - β_{j} Z_{j})$ on the diagonals, I is an identity matrix, $Q = \overset{2}{\hat{ρ}} (X^{*^{'}} \hat{Δ} Z) V ((X^{*^{'}} \hat{Δ} Z)$ , and V is the coefficient covariance matrix from the probit estimation of Equation 5.

Data

This study was undertaken based on a mobility dataset extracted from Household Survey acquired from the Sustainable Urban Development Joint Stock Company. The data were collected from December 2019 to February 2020 in Can Tho, a medium-sized city in the Mekong River delta region, Vietnam, to identify the household’s motorcycle ownership and use as well as household members’ travel patterns. The survey was conducted through a random sampling method, in which several households from each zone are randomly selected to respond to the questionnaire. A total of 2000 households participated in the survey, a sample of about 1% of the total households in urban areas. The survey was led by the Sustainable Urban Development Joint Stock Company in the framework of the project “Can Tho Development and Resilience.” The questionnaire was approved by and under the control of the Can Tho Department of Transport. The field survey needed strong support from the leaders of the administrative organization in the neighborhood, who are respected and trusted people in the neighborhood groups. These leaders provided a list of households under their management and, based on that, supervisors and surveyors randomly select the households. Then the leaders informed and persuaded households to participate in the survey. Therefore, the response rate was 100%. Finally, there were 5060 household members, who are 6 years old and above, participating in this survey.

Can Tho City is a central-level city in the Mekong River delta of Vietnam with over 1.24 million people (the urban population accounts for 70%) living under this city’s jurisdiction in 2020 ( 63 ). At the time of the survey, public transport by bus remains in this city but its mode share is limited with 0.1% ( 3 ). In 2020, the average gross regional domestic product (GRDP) per capita was 71.9 million VND (about USD 3084) and the mode share of the motorcycle mode was around 85.9% ( 3 ).

The number of registered motorized vehicles rapidly increased, with high annual growth rates of 8.3% for motorcycles and 13.9% for passenger cars (as estimated from Table 1 between 2014 and 2020). However, the number of cars was still modest compared to the number of motorcycles.

Table 1.

Vehicle Growth in Can Tho

Year	Population (000)	Private car	Urban bus	Others	MC	Growth rate (%)		MC per 1000 inhabitants
Year	Population (000)	Private car	Urban bus	Others	MC	MC	Car	MC per 1000 inhabitants
2014	1238	21,881	55	1720	542,561	na	na	438
2015	1208	25,743	65	1970	638,309	17.6	17.7	528
2016	1215	29,735	63	2357	679,806	6.5	15.5	560
2017	1222	33,873	60	3126	728,170	7.1	13.9	596
2018	1228	37,687	54	4190	774,005	6.3	11.3	630
2019	1236	42,763	55	5760	827,079	6.9	13.5	669
2020	1262	47,683	49	6433	873,752	5.6	11.5	692

Note: MC = motorcycle; na = not applicable.

Source: Can Tho Department of Transportation ( 64 ).

The survey was designed in four sections: (i) household sociodemographic characteristics; (ii) individual sociodemographic characteristics; (iii) workday travel diary; and (iv) built environment. The definition and sample statistics of the explanatory variables are presented in Table 2. The average number of members in a household is 3.22, mostly of working age. The majority of households have a medium income; therefore, they cannot afford to buy a car, which is costly at a price from USD 20,000, but their income is enough to support them to own a motorcycle, which costs about USD 1000. Individual factors covered in this survey include gender (49.87% female), population group from 25 to 60 years old (71.92%), who have permanent jobs, such as officers (45.58%) and workers (16.15%), that support them to gain a medium income (63.78%). The average density in Can Tho is 5348 people/km² with high variance; the minimum density value assigned is 640 persons/km² and the maximum is 15,774 persons/km². It is highlighted that public transport is not attractive in Can Tho in relation to network coverage; this indicator is measured through the proximity to bus stops, which is over 1.5 km on average, three times higher than the recommended 500-m threshold for walking distance. Therefore, it is not surprising that Can Tho households have little choice except bikes and motorcycles for their daily activities. The linkage between ownership and usage is because they are the outcomes of common causes, such as preference for motorcycles and dependence on motorcycles.

Table 2.

Summary Statistics of Independent Variables

Variable	Definition	Mean or percentage	SD	Min.	Max.
Household characteristics
HH_size	Household (HH) size (persons)	3.22	0.94	1	12
HH_18	Number of members 18 years old and above	3.17	1.55	1	12
HH_kids	Number of members 12 years old and below	.01	.16	0	3
HH_bike	Number of bikes owned by household	.44	.65	0	4
Trip_rate	Number of trips per day	3.78	1.70	1	15
HH_lowincome	Dummy: 1 if HH belongs to low-income groups	24.14%	na	na	na
HH_medium	Dummy: 1 if HH belongs to medium-income groups	62.70%	na	na	na
HH_highincome	Dummy: 1 if HH belongs to high-income groups	10.16%	na	na	na
Individual characteristics
Female	Dummy: 1 if HH member is female	49.87%	na	na	na
Age_18_24	Dummy: 1 if age is 18–24 years old	9.03%	na	na	na
Age_25_40	Dummy: 1 if age is 25–40 years old	32.35%	na	na	na
Age_41_60	Dummy: 1 if age is 41–60 years old	39.57%	na	na	na
Age_above60	Dummy: 1 if age is above 60 years old	13.29%	na	na	na
Servant	Dummy: 1 if HH member is official staff or civil servant	45.58%	na	na	na
Worker	Dummy: 1 if HH member is worker	16.15%	na	na	na
Student	Dummy: 1 if HH member is student	11.10%	na	na	na
Part_time	Dummy: 1 if HH member works part-time	8.34%	na	na	na
Full_time	Dummy: 1 if HH member works full-time	88.86%	na	na	na
Travel characteristics
Work	Number of trips by HH member to work	.86	.91	0	9
School	Number of trips by HH member to school	.098	.40	0	4
Individual	Number of trips by HH member with private purpose	.91	.98	0	10
Built environment
Home_proximity	Home transit proximity (km)	1.62	.64	0.6	3.6
Work_proximity	Workplace/school transit proximity (km)	1.49	.60	0.32	3.6
Pop_density	Population density (000 persons/km²)	5.35	5.06	0.64	15.77

Note: SD = standard deviation; Min. = minimum; Max. = maximum; na = not applicable.

Descriptive statistics of the dependent variables are shown in Table 3. The average number of motorcycles owned in a household is 2.66. The variance in motorcycle ownership is high, with some households having a total of 11 motorcycles and some none. The average daily distance driven by each motorcycle is 15.33 km, about 2.5 times higher than the daily distance of bikes. The utilization is also marked with high variance. The combination of owning motorcycles and bikes in Table 4 shows that the majority of households hold two or more motorcycles and zero bikes (1734 observations), followed by households holding two or more motorcycles and one bike (443 observations), while the combination least observed is zero motorcycles and two or more bikes (three observations).

Table 3.

Summary Statistics of Dependent Variables

Definition	Mean	SD	Min.	Max.
Number of motorcycles owned by the household	2.66	1.09	0	11
Average daily distance driven by motorcycle (km)	15.33	12.58	0.5	122

Note: SD = standard deviation; Min. = minimum; Max. = maximum.

Table 4.

Tabulation of Vehicle Counts

	Number of motorcycles
Number of bikes	0	1	2	3 or more	Total
0	4	104	519	663	1290
1	7	59	226	217	509
2 or more	3	6	36	56	101
Total	14	169	781	936	1900

Results and Discussion

The model estimation is conducted using STATA 16 software. For ease of presentation, the discrete components (motorcycle ownership) are discussed first, followed by a discussion of the corresponding continuous components (utilization). Finally, we discuss the correction bias.

Ownership Model Estimation

The detailed estimation results of the ownership model are shown in Table 5, including the coefficients, standard errors, and the p-value. Given the availability of data, the model explored here exhibits the utility of motorcycle ownership as a function of (1) household characteristics, (2) individual characteristics, and (3) the built environment.

Table 5.

Results of Ownership Model of Motorcycle

	One MC, LHH			Two MCs, LHH			Three or more MCs, LHH			One MC, MHH			Two MCs, MHH			Three or more MCs, MHH
	Coef.	SE	p-value	Coef.	SE	p-value	Coef.	SE	p-value	Coef.	SE	p-value	Coef.	SE	p-value	Coef.	SE	p-value
Variables with household characteristics
HH_size	−.021	.065	0.747	−.137	.065	0.035	.149	.059	0.013	−.174	.092	0.058	−.270	.039	0.000	.299	.039	0.000
HH_18	−.373	.050	0.000	−.160	.037	0.000	.450	.041	0.000	−.645	.087	0.000	−.474	.027	0.000	.554	.028	0.000
HH_kids	.197	.230	0.394	−.035	.226	0.877	−.130	.256	0.612	−6.495	237	1.000	−.021	.159	0.894	.087	.161	0.587
Trip_rate	−.071	.032	0.024	.004	.024	0.862	.054	.025	0.034	−.058	.037	0.116	.041	.016	0.010	.035	.016	0.028
HH_bike	.397	.077	0.000	.091	.062	0.145	−.490	.070	0.000	.335	.087	0.000	.302	.042	0.000	−.365	.042	0.000
Variables with individual characteristics
Female	.144	.094	0.127	−.103	.074	0.162	−.023	.079	0.766	.025	.117	0.827	.000	.052	0.996	−.013	.053	0.809
Age_18_24	.573	.307	0.062	.042	.234	0.858	.696	.254	0.006	1.577	.436	0.000	1.262	.179	0.000	1.464	.181	0.000
Age_25_40	.472	.363	0.194	.224	.283	0.429	.801	.305	0.009	1.247	.486	0.010	1.079	.214	0.000	1.256	.215	0.000
Age_41_60	.566	.362	0.118	.132	.284	0.642	.751	.306	0.014	1.365	.484	0.005	.968	.213	0.000	1.159	.214	0.000
Age_over60	.439	.380	0.248	.076	.297	0.799	−.642	.319	0.044	1.438	.5155	0.005	1.059	.224	0.000	1.262	.226	0.000
Servants	−.278	.118	0.018	.179	.095	0.060	.043	.102	0.675	.171	.150	0.256	−.151	.065	0.021	.124	.066	0.059
Worker	−.052	.145	0.720	.062	.116	0.592	−.015	.125	0.902	.004	.211	0.983	−.363	.092	0.000	.336	.093	0.000
Student	−.533	.315	0.091	−.217	.235	0.354	.587	.246	0.017	−.269	.393	0.493	−.456	.167	0.006	.493	.167	0.003
Part_time	.051	.286	0.858	−.326	.229	0.156	.295	.246	0.229	.330	.400	0.409	.114	.189	0.547	−.148	.190	0.437
Full_time	−.101	.254	0.691	−.197	.202	0.330	.277	.216	0.199	−.142	.371	0.701	−.009	.168	0.953	.045	.169	0.790
Variables with built environment
Home_proximity	−.832	.234	0.000	−.713	.179	0.000	1.456	.195	0.000	−.694	.279	0.013	−.679	.115	0.000	.816	.116	0.000
Work_proximity	.337	.179	0.059	−.303	.138	0.028	.201	.146	0.169	.445	.208	0.032	.133	.097	0.169	−.201	.098	0.040
Pop_density	.025	.012	0.050	.034	.011	0.001	.023	.011	0.041	.007	.017	0.673	.018	.008	0.024	−.018	.008	0.023

Note: MC = motorcycle; LHH = low-income household; MHH = medium-income household; Coef. = coefficient; SE = standard error; Bold denotes that explanatory variables are significant with p < 0.05.

Household Characteristics

Several household characteristics were found to affect motorcycle ownership across income groups. The presence of extra people in the household decreased the propensity of owning two motorcycles across income households, while increasing the propensity of households to pursue owning three or more motorcycles. These findings reveal that as the number of adults in the household increased, the more likely households were to own extra motorcycles to allow people to drive independently. These results are consistent with previous studies ( 7 ) and provide some insights as to why the number of motorcycles increases over the years.

Trip rate also played a significant role in motorcycle ownership. Trips were more likely to be taken by households holding two motorcycles or more. These findings may be attributable to the possibility that households increase the propensity of buying new motorcycles as their members undertake independent activities outside the home.

The mobility resources available in the household played a key role as well in the motorcycle ownership probabilities for the income groups. Low-income households having bikes were more likely to own one motorcycle, while medium-income households having bikes were more likely to own one or two motorcycles. These findings may suggest that improvement of the bike infrastructure structure or policy may affect bike consumption and thus reduce motorcycle ownership. Among groups, bike ownership was associated with motorcycle ownership in the low-income group better than in the medium-income group.

Individual Characteristics

The age and occupation of individuals influenced motorcycle ownership probabilities in different ways across income groups. We found that people between the age of 18 and 24 years were more likely to own motorcycles than older people for medium-income households. These findings prove that motorcycles play a more significant role in medium-income youths’ lives than in those of low-income people. Also, medium-income groups in Can Tho tend to become “captive” motorcycle users rather than “choice” motorcycle users compared to low-income groups—which can be considered as a discouraging factor for biking or public transport among medium-income people.

Occupation is another individual characteristic that was found to affect motorcycle ownership outcome probabilities. The presence of workers in the household increases the propensity of households to own three motorcycles or more. Similarly, student-related households were likely to own three motorcycles or more. These findings may be attributable to the possibility that medium-income households give priority to purchasing motorcycles for members who are workers or students. Further, we found that the influence of student-related medium-income households on motorcycle ownership was stronger than that of other job statuses.

Transit accessibility also plays a significant role in motorcycle ownership. For example, households located far from bus stops were more likely to own three motorcycles or more. Further, we found that the influence of home transit proximity was greater than for workplace transit proximity. These findings may contribute to a requirement for improvements of the bus network coverage or improvement of accessibility that can be considered as an encouraging factor for public transit among motorcyclists.

Population densities were significantly positive to motorcycle ownership in the LHH-two and MHH-two models but significantly negative for the MHH-three model. When density increased, households were less likely to hold three or more motorcycles. This finding may be attributed to the possibility that households are more likely to select their residential location close to their household members’ place of employment to minimize travel time, distance, and cost of travel.

Utilization Model Estimation

The results for the continuous components (utilization) are presented in Table 6. The explanatory variables are classified into four categories and the impacts of each of these categories are discussed below.

Table 6.

Results of Utilization Model of Motorcycles

	One MC, LHH			Two MCs, LHH			Three or more MCs, LHH			One MC, MHH			Two MCs, MHH			Three or more MCs, MHH
	Coef.	SE	p-value	Coef.	SE	p-value	Coef.	SE	p-value	Coef.	SE	p-value	Coef.	SE	p-value	Coef.	SE	p-value
Variables with household characteristics
HH_size	.400	1.477	0.786	.796	.933	0.394	1.578	1.365	0.248	−9.016	5.445	0.098	−.911	.672	0.175	−.089	.314	0.775
HH_18	−.513	1.228	0.676	−.251	.629	0.690	.638	3.178	0.841	−2.713	6.466	0.675	−.259	.752	0.731	.328	.315	0.298
HH_kids	−.049	3.388	0.988	−.520	2.647	0.844	−1.129	4.731	0.811	−.394	2.905	0.892	4.886	2.3003	0.034	.371	1.652	0.822
HH_bike	−3.712	1.696	0.029	−3.337	.751	0.000	−4.210	3.830	0.272	−2.368	3.636	0.515	.082	.736	0.910	−.217	.437	0.620
Variables with individual characteristics
Female	−4.603	1.717	0.007	−3.187	.879	0.000	−.828	1.254	0.509	−2.368	3.636	0.515	−1.378	.746	0.065	−2.074	.499	0.000
Age_18_24	2.279	4.896	0.642	5.206	2.975	0.080	10.055	6.129	0.101	9.625	21.190	0.650	10.572	3.128	0.001	8.669	1.783	0.000
Age_25_40	.818	5.354	0.878	4.422	3.447	0.200	12.262	7.268	0.092	13.439	20.641	0.515	10.831	3.419	0.002	8.533	2.042	0.000
Age_41_60	−.070	5.411	0.990	3.597	3.4751	0.301	10.652	6.999	0.128	17.892	21.108	0.397	9.572	3.366	0.004	6.582	2.035	0.001
Age_over60	−.478	5.709	0.933	−1.718	3.630	0.636	9.649	6.691	0.149	13.572	22.140	0.540	9.888	3.593	0.006	4.657	2.144	0.030
Servants	3.623	2.128	0.089	2.552	1.229	0.038	.376	1.756	0.830	−1.216	6.048	0.841	1.370	.986	0.165	2.174	.685	0.002
Worker	1.949	2.404	0.417	3.639	1.446	0.012	4.13	2.097	0.049	−10.308	8.185	0.208	5.566	1.453	0.000	4.293	.882	0.000
Student	−3.736	5.812	0.520	1.963	3.054	0.520	4.870	5.514	0.377	−4.492	14.370	0.755	5.255	2.603	0.044	.775	1.623	2.731
Part_time	−.486	4.823	0.920	3.996	2.718	0.142	11.074	4.316	0.010	14.508	14.515	0.318	−3.634	2.752	0.187	−2.894	1.794	0.107
Full_time	2.917	4.349	0.502	3.275	2.329	0.160	7.226	3.928	0.066	18.597	12.860	0.148	−1.388	2.492	0.578	−1.161	1.570	0.459
Variables with travel characteristics
Work	.046	1.807	0.979	1.673	.973	0.085	1.765	1.364	0.196	5.665	5.519	0.305	.0479	.699	0.945	.435	.578	0.452
School	2.926	3.589	0.415	.238	1.539	0.877	4.963	1.841	0.007	2.473	8.094	0.760	3.364	1.463	0.022	1.938	.935	0.038
Individual	−1.553	1.560	0.320	−1.102	.893	0.217	−1.315	1.185	0.267	1.675	5.320	0.753	−1.859	.689	0.007	−.948	.521	0.069
Variables with built environment
Home_proximity	−2.473	4.815	0.607	1.871	2.503	0.455	5.289	11.202	0.637	−24.178	11.147	0.030	−4.169	1.909	0.029	.253	1.145	0.825
Work_proximity	3.030	3.158	0.337	1.801	1.728	0.298	9.054	2.632	0.001	16.153	8.365	0.054	1.228	1.366	0.369	2.194	.922	0.017
Pop_density	−.217	.231	0.347	−.255	.142	0.072	−.374	.269	0.166	−1.091	.565	0.054	−.618	.121	0.000	−.401	.074	0.000
Goodness-of-fit assessment
Bias correction factor
Mills lambda λ	−0.400	0.789	0.000	−0.185	0.358	0.000	0.120	0.114	0.029	−0.921	0.469	0.000	−0.114	0.110	0.029	−0.087	0.285	0.003
Rho	−0.681	na	na	−0.710	na	na	0.808	na	na	−0.612	na	na	−0.794	na	na	−0.828	na	na
Sigma	0.535	na	na	0.619	na	na	0.698	na	na	0.539	na	na	0.652	na	na	0.702	na	na
Wald chi-square	220	na	na	155	na	na	84	na	na	252	na	na	152	na	na	61	na	na

Note: MC = motorcycle; LHH = low-income household; MHH = medium-income household; Coef. = coefficient; SE = standard error; na =not applicable.Bold denotes that explanatory variables are significant with p < 0.05.

Household Characteristics

The household characteristics affecting motorcycle utilization were the number of children in the household and the number of bikes. MHH-two with more children increased the trip length by motorcycles. This finding likely captures the trip chaining behavior of parents who often link their daily trips with children’s trips, for example, parents in Vietnamese cities drop off children at school on the way to work or pick the children up on the way home. The number of bikes in low-income groups is also related to less driving by motorcycles.

Individual Characteristics

Gender, age, and occupation were the individual characteristics affecting motorcycle utilization decisions. Females were more likely to travel short distances regardless of whether households have more or fewer motorcycles. This finding may likely capture the gendered division of work in households, where women often have multiple tasks and activities in the home. This trend was found to be similar in many cities, such as Auckland, Dublin, Hanoi, Helsinki, Jakarta, Kuala Lumpur, Lisbon, and Manila ( 65 ). Age groups related strongly positively to the likelihood of driving in MHH-two or MHH-three, but did not show a significant relationship with motorcycle utilization in low-income households. In addition, adults in MHH-two were more likely to travel than those in MHH-three. This finding may be attributable to the possibility that motorcyclists in MHH-two linked their daily trips together to save time and to seek for convenience when the mobility resources in the household are limited. Further, we found that young members (age of elder household adult between 18 and 40 years) are found to be more likely to travel a longer distance. Finally, employed adults were also found to be more likely to increase their motorcycle kilometers traveled, but in a different way. Civil servants and workers were more likely to increase their travel distance. Households with part-time job members had a significant positive association with LHH-three, but its association with motorcycle ownership in other models was insignificant.

Travel Characteristics

The next set of variables captures the effects of trip purpose on motorcycle utilization. Trips for work purposes were found not to be associated with an increase of motorcycle utilization. Increase of number of school trips was positively associated with motorcycle utilization in LHH-three, MHH-two, and MHH-three. This highlighted again the linkage between parent’s activities with children’s activities in the household. These findings may be attributed to the possibility that the combination of trip chaining and the presence of children in the household prevents its members from shifting to public transport.

Built Environment

Transit accessibility continues to play a significant role in motorcycle utilization in income groups. For example, households who are far from bus transit were less likely to travel if they own only one motorcycle. Further, we found that the negative impact of proximity in MHH-one on motorcycle utilization was much stronger than that of MHH-two. Longer distance from bus transit to workplace/school was strongly associated with higher probabilities of riding motorcycles in LHH-three and MHH-three. Finally, when the density increased, the frequency of using motorcycles decreased.

Correction Bias

To test for bias, we examine the relationship between the residual for the two models (discrete model and continuous model). If the unobservables in the discrete model are correlated with the unobservables in the continuous model, we have biased estimates without correction, which implies that unobservables in the motorcycle ownership selection are also affecting the utilization model. As indicated in Table 6, the error terms for ρ changed from −0.612 to 0.808 depending on the model, meaning that the unobserved factors that cause motorcycle ownership and utilization are positively and negatively correlated with one another, respectively. The error terms across the income groups have a large standard deviation, which implies that the heterogeneity is captured with respect to ownership levels and income. The Wald test reveals that the joint models are preferred to the independent probit and linear regression models, with values varying from 61 to 252. This indicates that the models reject the independence of the equation at the 95% confidence level. The selectivity bias correction factor, λ, is statistically significant and negative. The implication of this finding is that unobserved factors that influenced motorcycle ownership also influenced motorcycle utilization. Specifically, a negative sign on the estimated correlation coefficient, as obtained, indicates that unobserved factors that increase the propensity to hold a motorcycle decrease the mileage.

Conclusions

This paper presented an analysis of households’ decisions in holding and using motorcycles. A discrete-continuous structure was developed that captures the linkage between motorcycle ownership and utilization and recognizes the impacts of common unobserved factors on households’ decisions. The model was estimated using data from the 2019–2020 Household Travel Survey. Overall, the empirical findings from this research highlight the importance of explicitly incorporating interdependencies in households’ decisions across income. For example, we found that some household characteristics, such as household size and the number of adults, were more likely to affect motorcycle ownership in the medium-income groups than in the low-income groups, while another household indicator such as the number of bikes as a mobility resource available in the household was less likely to affect motorcycle ownership in the medium-income groups than in the low-income groups. These findings may suggest that households in Can Tho may be more likely to buy an extra motorcycle as their incomes increase, potentially generating more motorcycle-related issues in the future. On the positive side, we found that the non-motorized mobility resources available in the low-income group played a key role in motorcycle ownership and usage; the more bikes in the household, the more likely it is that low-income households own only one motorcycle, and the fewer motorcycle trips were taken. Therefore, we suggest that medium-sized cities should develop cycling infrastructure to facilitate the mobility of low-income groups and promote well-established travel patterns among cyclists, and they will then be less willing to switch to motorized vehicles as their income increases. This suggestion is particularly relevant in the context of significant policy interests and financial incentives over the last decade that have been aimed at enhancing public transport ridership. Indeed, bikes will play an important role as a convenient and inexpensive solution for the “first and last mile” to/from public transit. In addition, Can Tho is a city with many famous destinations, such as Ninh Kieu Wharf, Cai Rang Floating Market, Phong Dien Floating Market, and so forth, which attract approximately 3 million domestic and foreign tourists annually. Therefore, the provision of walking space and a public bicycle service is perfectly suitable for tourists. Currently, bicycles are circulating with motorized vehicles and there are no regulations on space and operating areas for non-motorized vehicles in Can Tho City. As road infrastructure plays an important role in securing and encouraging the use of non-motorized vehicles, if bicycles operate alongside motorized traffic flow, it can be dangerous and polluting for bicycle users. The creation of a bicycle habit must be associated with improving infrastructure, especially designing exclusive bicycle lanes (either under the roadway or on wide sidewalks). Besides the use of bicycles for different purposes, bicycles should also be encouraged as a means of connecting to public transport by bus. This method should be implemented in a way that provides parking lots near bus stops/terminals to make bike trips become viable for last mile trips to/from public transit. The second way to serve residents and tourists is to develop public bicycles. In addition, we found that those aged 18–60 years in medium-income groups were positively associated with the probability of owning two or more motorcycles and riding them (here we exclude those aged over 60 years because their behaviors were heterogeneous). These findings imply that adults in Can Tho tend to become motorcycle-dependent users rather than motorcycle-selective users, which can be considered a discouraging factor for biking or using public transport among adults. Recently, policymakers in Vietnam have started introducing a motorcycle ban policy that prohibits motorcycle usage within five cities, including Can Tho. These cities plan to ban all motorcycle usage by 2030. With the specific context in Can Tho, and possibly in other medium-sized cities, we strongly assert the negative impacts of the motorcycle ban policy on residents’ lives with respect to ensuring mobility and combating poverty. Given that the people of Can Tho depend heavily on motorcycles, we would rather see the prospect of a travel demand picture with a mix of public transport and motorcycles together instead of advocating the motorcycle ban policy. These effects cannot be captured by travel models, which model individuals independently. On the other hand, the importance of motorcycles in the overall transport mix needs to be highlighted in Can Tho, as well as in other Vietnamese medium-sized cities, and perhaps in other cities with high motorcycle rates—they are likely to be a major part of the solution and not necessarily part of the problem.

This study had some methodological limitations. By applying a binary discrete-continuous choice model to model the binary choice of whether a household owns a given number of vehicles or not (binary choice) and the mileage accrued by the household (continuous choice), it may create bias in model estimation and inconsistencies in prediction. Although bias was taken into account in the binary discrete/continuous model presented in this study, considering the multinomial discrete-continuous choice model (as a future study) may further provide an improvement over the binary joint model.

Footnotes

Acknowledgements

The author would like to thank colleagues at the Urban Sustainable Development Joint Stock Company for their support in data collection.

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: A.M. Ngoc, H. Nishiuchi; data collection: A.M. Ngoc; analysis and interpretation of results: A.M. Ngoc, H. Nishiuchi; draft manuscript preparation: A.M. Ngoc. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

An Minh Ngoc

References

National Traffic Safety Committee (NTSC). Motorcycle and Car Statistics, 1990-2020. Printing Report. 2021.

Hung

K. V.

Traffic Management in Motorcycle-Dependent Cities. PhD dissertation. 2006. https://tuprints.ulb.tu-darmstadt.de/767/.

Ngoc

A. M.

Nishiuchi

Truong

N. V.

Huyen

L. T.

A Comparative Study on Travel Mode Share, Emission, and Safety in Five Vietnamese Cities. International Journal of Intelligent Transportation Systems Research, Vol. 20, No. 2022, pp. 157–169.

Ngoc

A. M.

Nishiuchi

Nhu

N. T.

Huyen

L. T.

Ensuring Traffic Safety of Cargo Motorcycle Drivers in Last-Mile Delivery Services in Major Vietnamese Cities. Case Studies on Transport Policy, Vol. 10, 2022, pp. 1735–1742.

Ngoc

A. M.

Minh

C. C.

Nhu

N. T.

Nishiuchi

Huynh

Influence of the Human Development Index, Motorcycle Growth and Policy Intervention on Road Traffic Fatalities – A Case Study of Vietnam. International Journal of Transportation Science and Technology, 2022. https://doi.org/10.1016/j.ijtst.2022.09.004.

N. T.

Binh

D. T.

Huyen

L. T.

Ngoc

A. M.

Impact of Congestion Charging on Traffic and Environment in Hanoi: Results from Simulation Analysis. In CIGOS 2021, Emerging Technologies and Applications for Green Infrastructure ( C.

Ha-Minh

Tang

A. M.

Bui

T. Q.

X. H.

Huynh

D. V. K.

, eds.). Lecture Notes in Civil Engineering, Vol. 203. Springer, Singapore, 2022, pp. 1643–1652. https://doi.org/10.1007/978-981-16-7160-9_166.

Tuan

V. A.

Mode Choice Behavior and Modal Shift to Public Transport in Developing Countries – The Case of Hanoi City. Journal of the Eastern Asia Society for Transportation Studies, Vol. 11, 2015, pp. 473–487.

Ngoc

A. M.

Hung

K. V.

Tuan

V. A.

Towards the Development of Quality Standards for Public Transport Service in Developing Countries: Analysis of Public Transport Users’ Behavior. Transsportation Research Procedia, Vol. 25, 2017, pp. 4564–4583.

Nguyen

N. T.

Miwa

Morikawa

Response to the Planned Public Transport System in Ho Chi Minh City: Analysis of Latent Classes. Clean Technologies and Environmental Policy, Vol. 21, 2019, pp. 1925–1935.

10.

Chu

C. C.

Nguyen

L. X.

Ton

T. T.

Huynh

Assessment of Motorcycle Ownership, Use, and Potential Changes due to Transportation Policies in Hochiminh City, Vietnam. Journal of Transport Engineering, Part A: Systems, Vol. 145, No. 12, 2019, p. 05019007.

11.

Truong

T. M. T.

Ngoc

A. M.

Parking Behavior and the Possible Impacts on Travel Alternatives in Motorcycle-Dominated Cities. Transport Research Procedia, Vol. 48, 2019, pp. 3469–3485.

12.

Ngoc

A. M.

Nishiuchi

Minh

C. C.

Key Factor Associated with Traffic Crashes and the Role of Crash Experiences in Mode Choice for School Trips – a Case Study of Can Tho, Vietnam. Travel Behaviour and Sociey, Vol. 30, 2023, pp. 240–248.

13.

Bray

D. J.

Holyoak

Motorcycles in Developing Asian Cities: A Case Study of Hanoi. Proc., 37th Australasian Transport Research Forum, Sydney, 2015.

14.

Creemers

Cools

Tormans

Lateur

P. J.

Janssens

Wets

Identifying the Determinants of Light Rail Mode Choice for Medium- and Long-Distance Trips: Results from a Stated Preference Study. Transportation Research Record: Journal of the Transportation Research Board, 2012. 2275: 30–38.

15.

Hess

D. B.

Ong

P. M.

Traditional Neighborhoods and Automobile Ownership. Transportation Research Record: Journal of the Transportation Research Board, 2002. 1805: 35–44.

16.

Kim

Wang

Empirical Examination of Neighborhood Context of Individual Travel Behaviors. Applied Geography, Vol. 60, 2015, pp. 230–239.

17.

Law

T. H.

Hamid

Goh

C. N.

The Motorcycle to Passenger Car Ownership Ratio and Economic Growth: A Cross-Country Analysis. Journal of Transport Geography, Vol. 46, 2015, pp. 122–128. https://doi.org/10.1016/j.jtrangeo.2015.06.007.

18.

Mercado

Paez

Farber

Roorda

Morency

Explaining Transport Mode Use of Low-Income Persons for Journey to Work in Urban Areas: A Case Study of Ontario and Quebec. Transportmetrics, Vol. 8, 2012, pp. 157–179.

19.

Nolan

A Dynamic Analysis of Household Car Ownership. Transportation Research Part A: Policy Practice, Vol. 44, No. 6, 2010, pp. 446–455.

20.

Cirillo

Axhausen

Comparing Urban Activity Travel Behaviour. Presented at 81th Annual Meeting of the Transportation Research Board, Washington, D.C., January 13–17, 2002, 27p.

21.

Hensher

Rose

Development of Commuter and Non-Commuter Mode Choice Models for the Assessment of New Public Transport Infrastructure Projects: A Case Study. Transportation Research Part A: Policy and Practice, Vol. 41, 2007, pp. 428–443.

22.

Limtanakool

Dijst

Schwanen

The Influence of Socioeconomic Characteristics, Land Use and Travel Time Considerations on Mode Choice for Medium- and Longer-Distance Trips. Journal of Transport Geography, Vol. 14, 2006, pp. 327–341.

23.

de Witte

Hollevoet

Dobruszkes

Hubert

Macharis

Linking Modal Choice to Mobility: A Comprehensive Review. Transportation Research Part A: Policy and Practice, Vol. 49, 2013, pp. 329–341.

24.

Kajita

Toi

Chishaki

Matsuoka

Structural Mechanism of Modal Choice Based on the Linked Structure of Trip Purpose and Transportation Choice. Memoirs of the Faculty of Engineering, Kyushu University, Vol. 64, No. 1, 2004, pp. 17–33.

25.

Rietveld

The Accessibility of Railway Stations: The Role of the Bicycle in The Netherlands. Transport Research Part D: Transport and Environment, Vol. 5, 2000, pp. 71–75.

26.

Cao

Mokhtarian

Handy

Examining the Impacts of Residential Self-Selection on Travel Behavior: A Focus on Empirical Findings. Transport Reviews, Vol. 29, 2009, pp. 359–395.

27.

Chakaborty

Mishra

Land Use and Transit Ridership Connections: Implications for State-Level Planning Agencies. Land Use Policy, Vol. 30, No. 1, 2013, pp. 458–469.

28.

Chen

Gong

Paaswell

Role of the Built Environment on Mode Choice Decisions: Additional Evidence on the Impact of Density. Transportation, Vol. 35, 2007, pp. 285–299.

29.

Chow

Zhao

Liu

Ubaka

Transit Ridership Model Based on Geographically Weighted Regression. Transport Research Record: Journal of the Transportation Research Board, 2006. 1972: 105–114.

30.

Cui

Boisjoly

Miranda-Moreno

El-Geneidy

Accessibility Matters: Exploring the Determinants of Public Transport Mode Share Across Income Groups in Canadian Cities. Transportation Research Part D: Transport and Environment, Vol. 80, 2020, p. 102276.

31.

Dinh

Wang

Tang

Mishra

Liu

Joint Analysis of the Spatial Impacts of Built Environment on Car Ownership and Travel Mode Choice. Transportation Research Part D: Transport and Environment, Vol. 60, 2018, pp. 28–40.

32.

Ding

Cao

Wang

Synergistic Effects of the Built Environment and Commuting Programs on Commute Mode Choice. Transportation Research Part A: Policy and Practice, Vol. 118, 2018, pp. 104–118.

33.

Bhat

C. R.

Incorporating Observed and Unobserved Heterogeneity in Urban Work Travel Mode Choice Modeling. Transport Science, Vol. 34, No. 2, 2000, pp. 228–238.

34.

Fan

Ermagun

Cao

Wang

Das

Public Bicycle as a Feeder Mode to Rail Transit in China: The Role of Gender, Age, Income, Trip Purpose, and Bicycle Theft Experience. International Journal of Sustainable Transportation, Vol. 11, No. 4, 2017, pp. 308–317.

35.

Yang

Jin

Gao

Exploring Spatially Varying Influences on Metro-Bike-Share Transfer: A Geographically Weighted Poisson Regression Approach. Sustainability, Vol. 10, No. 5, 2018, p. 1526.

36.

Qiao

Saeed

T. U.

Chen

Nateghi

Labi

Acquiring Insights into Infrastructure Repair Policy Using Discrete Choice Models. Transportation Research Part A: Policy and Practice, Vol. 113, 2018, pp. 491–508.

37.

Shen

Chen

Pan

Factors Affecting Car Ownership and Mode Choice in Rail Transit-Supported Suburbs of a Large Chinese City. Transportation Research Part A: Policy and Practice, Vol. 94, 2016, pp. 31–44.

38.

Vega

Reynolds

A Methodological Framework for the Study of Residential Location and Travel-to-Work Mode Choice Under Central and Suburban Employment Destination Patterns. Transportation Research Part A: Policy and Practice, Vol. 43, No. 4, 2009, pp. 401–419. https://doi.org/10.1016/j.tra.2008.11.011.

39.

Bolduc

Boucher

Daziano

R. A.

Hybrid Choice Modeling of New Technologies for Car Choice in Canada. Transportation Research Record: Journal of the Transportation Research Board, 2008. 2082: 63–71.

40.

Greene

W. H.

Hensher

D. A.

A Latent Class Model for Discrete Choice Analysis: Contrasts with Mixed Logit. Transportation Research Part B: Methodological, Vol. 37, No. 8, 2003, pp. 681–698.

41.

Guo

Wang

Peeta

Anastasopoulos

P. C.

Personal and Societal Impacts of Motorcycle Ban Policy on Motorcyclists’ Home-to-Work Morning Commute in China. Travel Behavior and Society, Vol. 19, 2020, pp. 137–150.

42.

Kepaptsoglou

Milioti

Christoforou

Karlaftis

M. G.

An Empirical Investigation of Motorcycle Ownership and Usage. Proc., 1st Congress of Transportation and Development Institute (TDI), ASCE, Reston, VA, 2011, pp. 1075–1084.

43.

Senbil

Zhang

Fujiwara

Motorization in Asia: 14 Countries and Three Metropolitan Areas. IATSS Research, Vol. 31, No. 1, 2007, pp. 46–58.

44.

Wen

C. H.

Chiou

Y. C.

Huang

W. L.

A Dynamic Analysis of Motorcycle Ownership and Usage: A Panel Data Modeling Approach. Accident Analysis & Prevention, Vol. 49, 2012, pp. 193–202. https://doi.org/10.1016/j.aap.2011.03.006.

45.

Velagapudi

S. P.

Ray

G. G.

A Study on Motorcycle Usage and Comfort in Urban India. Proc., 19th Triennial Congress of the IEA, International Ergonomics Association, Melbourne, Australia, 2015.

46.

Sanko

Dissanayake

D. D.

Kurauchi

Maesoba

Yamamoto

Morikawa

Inter-Temporal Analysis of Household Car and Motorcycle Ownership Behaviors: The Case in the Nagoya Metropolitan area of Japan, 1981–2001. IATSS Research, Vol. 33, No. 2, 2009, pp. 39–53. https://doi.org/10.1016/S0386-1112(14)60243-0.

47.

Wong

K. L.

An Analysis of Car and Motorcycle Ownership in Macao. International Journal of Sustainable Transportation, Vol. 7, No. 3, 2013, pp. 204–225. https://doi.org/10.1080/15568318.2013.710057.

48.

Lai

W. T.

J. L.

Chiang

Y. S.

The Study of Multi-Vehicle Household Mix Demand Model on Auto/Motorcycle Ownership and Use in Taiwan Area. Transportation Planning Journal, Vol. 35, No. 3, 2006, pp. 309–336.

49.

Nguyen

N. T.

Miwa

Morikawa

Demand Forecast of Public Transport Considering Positive and Negative Mass Effects. Transportation Research Part D: Transport and Environment, Vol. 85, 2020, p. 102466.

50.

Ermagun

Hossein Rashidi

Samimi

A Joint Model for Mode Choice and Escort Decisions of School Trips. Transportmetrica A: Transport Science, Vol. 11, No. 3, 2015, pp. 270–289.

51.

Ermagun

Hajivosough

Samimi

Rashidi

T. H.

A Joint Model for Trip Purpose and Escorting Patterns of the Disabled. Travel Behaviour and Society, Vol. 3, 2016, pp. 51–58.

52.

Spissu

Pinjari

A. R.

Pendyala

R. M.

Bhat

C. R.

A Copula-Based Joint Multinomial Discrete-Continuous Model of Vehicle Type Choice and Miles of Travel. Transportation, Vol. 36, 2009, pp. 403–422.

53.

Bhat

C. R.

Sen

Household Vehicle Type Holdings and Usage: An Application of the Multiple Discrete-Continuous Extreme Value (MDCEV) Model. Transportation Research Part B: Methodological, Vol. 40, No. 1, 2006, pp. 35–53.

54.

Fang

H. A.

A Discrete-Continuous Model of Households’ Vehicle Choice and Usage, with an Application to the Effects of Residential Density. Transportation Research Part B: Methodological, Vol. 42, 2008, pp.736–758.

55.

Hensher

D. A.

Milthorpe

F. W.

Selectivity Correction in Discrete-Continuous Choice Analysis. Regional Science and Urban Economics, Vol.17, 1987, pp. 123–150.

56.

Mannering

Winston

A Dynamic Empirical Analysis of Household Vehicle Ownership and Utilization. The Rand Journal of Economics, Vol. 16, No. 2, 1985, pp. 215–236.

57.

Mannering

Hensher

D. A.

Discrete/continuous econometric models and their application to transport analysis. Transport Reviews, Vol. 7, No. 3, 1987, pp. 227–244.

58.

Train

Qualitative Choice Analysis: Theory, Econometrics and an Application to Automobile Demand. MIT Press, Cambridge, 1986.

59.

Dubin

McFadden

An Econometric Analysis of Residential Electric Appliance and Holding and Consumption. Econometrica, Vol. 51, 1984, pp. 345–362.

60.

Train

Lohrer

Vehicle Ownership and Usage: An Integrated System of Disaggregate Demand Models. Cambridge Systematics, Berkeley, CA, 1982.

61.

Heckman

Sample Selection Bias as a Specification Error. Econometrica, Vol. 47, 1979, pp. 153–162.

62.

Greene

W. H.

Econometric Analysis, 6th ed. Pearson Prentice Hall, Upper Saddle River, NJ, 2008.

63.

General Statistics Office (GSO). 2021 n.d. https://www.gso.gov.vn/en/homepage/

64.

Can Tho DOT. Infrastructure and Vehicle Statistic Data. Unpublished Report. Can Tho, Vietnam, 2021.

65.

W.-S.

Acker

Understanding Urban Travel Behaviour by Gender for Efficient and Equitable Transport Policies. International Transport Forum Discussion Papers 2018/01. OECD Publishing. https://www.itf-oecd.org/sites/default/files/docs/urban-travel-behaviour-gender.pdf