Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

Contents lists available at ScienceDirect

Transportation Research Interdisciplinary Perspectives

journal homepage: www.sciencedirect.com/journal/transportation-
research-interdisciplinary-perspectives

Effect of socio-economic and demographic factors on ride-sourcing services

in Dhaka City, Bangladesh
Md. Hishamur Rahman a, Soumik Nafis Sadeek a, *, Asif Ahmed a, Shakil Mohammad Rifaat b,
Masnun Abrar b
a
Department of Civil Engineering, IUBAT – International University of Business Agriculture and Technology, Uttara, Dhaka, Bangladesh
b
Department of Civil and Environmental Engineering, Islamic University of Technology (IUT), Board Bazar, Gazipur, Bangladesh

A R T I C L E I N F O A B S T R A C T

Keywords: Ride-sourcing services have emerged in the entire world through advancement of information technology. This
Ride-sourcing service has also penetrated in the market of the developing countries similar to Bangladesh with immense po­
Socio-economic tential. People have adopted this new trip concept with a hope to ease their daily lives by hassle free movement.
Demographics
In developed countries, making trips using ride-sourcing services have been greatly influenced by users’ socio-
Bivariate-ordered probit
economic and demographic backgrounds. But, it is hitherto unclear how this ride-service concept is influ­
Marginal effects
Emerging mobility enced by the social and economic characteristics of the users from developing countries. Hence, this research
attempts to unveil the inherent relationships between users’ usage of motorcycle and car ride-sourcing services
and their socio-economic and demographic backgrounds. The sample size contains responses of 1003 re­
spondents. Sixteen questions are asked indicating different socio-economic and demographic characteristics of
the respondents towards their choice between motorbike and car ride-sourcing: two most popular ride-sourcing
services in Bangladesh. A Bivariate-ordered Probit Model has been used to model jointly their choices towards
ride-sourcing services. The outcomes indicate that people with certain types of socio-demographic background
prefer to use motorcycles over cars for ride-sourcing and vice-versa. The result also shows that gender, household
density, income, educational qualification, and smart-phone usage have a propensity to increase ride-sourcing
choice towards both motorbikes and cars. The outcomes are expected to provide useful insights to the ride-
sourcing service providers, transport agencies and policy makers of Bangladesh in understanding the choice
behaviors from the ride-sourcing user perspectives.

1. Introduction Chuxing, Ola, Lyft, etc. Uber alone is delivering its services in 700
metropolitan areas in 65 countries (DMR, 2019) including few devel­
Ride-sourcing, aka ride-hailing service is a system where freelance oping countries.
drivers provide on-demand transport service at their flexible working Uber established its business in Bangladesh in 2016. Along with
hours. The system responds to its user in real-time which has made this Uber, two other domestic companies: Pathao and Amar Ride also
service an efficient, quick and convenient one, resulting in increasing emerged in the same yard at nearly the same time (IDLC, 2018). At this
popularity. Generally, ride-sourcing system providers are known as moment, Bangladesh Road Transport Authority (BRTA) has approved
Transport Network Companies (TNCs). The service providers run the the licenses of twelve ride-sharing and ride-sourcing companies in
entire system through an internet-based platform. Thus, real-time Bangladesh (BRTA, 2020). According to the IDLC (2018) report, around
communication can take place between service providers and passen­ 5,00,000 commuters used ride-hailing service through cars and bikes
gers by dint of mobile devices (Wang and Yang, 2019). Therefore, mo­ during 2017–2018. It is important to note that after the introduction of
bile technologies equipped with internet facilities have enhanced the these three companies, especially Uber, the number of registered mo­
ride-sourcing market. Around the world, several companies have come torbikes and cars had increased significantly. According to the BRTA-
forward to deliver services through ride-sourcing, namely: Uber, Didi 2018 report, during 2016–2018, the number of registered motorbikes

* Corresponding author.
E-mail addresses: hishamur@iubat.edu (Md.H. Rahman), soumik.sadeek@iubat.edu (S.N. Sadeek), asif.ahmed@iubat.edu (A. Ahmed), smrifaat@iut-dhaka.edu
(S.M. Rifaat), masnunabrar@iut-dhaka.edu (M. Abrar).

https://doi.org/10.1016/j.trip.2021.100492
Received 7 April 2021; Received in revised form 26 July 2021; Accepted 27 October 2021
Available online 2 November 2021
2590-1982/© 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Md.H. Rahman et al. Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

and cars increased by 35.86% and 8.1%, respectively in Bangladesh. In sourcing users by combining geographical areas and demographic
Dhaka alone, this number was 40.1% for motor-bikes and 8.6% for cars. characteristics. This study provided a framework to correlate ride-
Also, in a recent study, Wadud (2020) investigated that motorcycle sourcing demand and waiting time with age, income, and education.
ownership in Dhaka had increased significantly after the introduction of They expected that the cities can utilize the results of this research to
motorcycle based e-hailing app services. As a result, it is evident from deal with the issues of vehicle deadheading, vehicle occupancy, mode
the data that cars and motor-bikes are the most preferable ride-sourcing replacement, efficiency, and VMT in more detail and realize what they
modes in Bangladesh. People with all social status are using these might actually gain or lose.
technology-enabled services in Bangladesh. Deterioration in public Gilibert et al. (2017) found that young people are more inclined
transport, steady growth of middle class people, fast communication, towards ride-sourcing services. Alemi et al. (2018) found a greater
and hike in motorbike sale have made both of these ride-sourcing ser­ propensity of using ride-sourcing services among highly educated and
vices popular in Bangladesh (IDLC, 2018). older people, long-distance business trip travelers, frequent users of
However, it is unclear how people from various social, economic, mobile apps, and in areas with land-use mix and accessibility by car. The
and demographic backgrounds are using both motorcycle and car as findings are expected to meet the expanding need to coordinate policy
ride-sourcing services. Majority of the previous studies on behavioral making and incentives in order to achieve the potential benefits of these
choice modeling of ride-sourcing services focused only on either of the services while reducing the negative effects. They suggested that the
two modes—car or motorcycle. Therefore, it is important to conduct a highest public benefits would be achieved by advertising the pooling
research on how frequently the users choose various ride-sourcing ser­ version of these services. Lavieri et al. (2018) used publicly available
vices (i.e., car and motorcycle) together and how their choices are data to investigate trip generation patterns from a specific zone for both
affected by their socio-economic and demographic characteristics in the weekdays and weekends. They aimed to identify characteristics of those
context of a megacity, Dhaka - the capital city of Bangladesh. Moreover, zones that attract ride-sourcing services using a multivariate count
the research also attempts to find whether there exists any pattern be­ model. The findings were analogous to the expectations. For instance,
tween bike and car mode choices based on different socio-demographic zones with more male population are intended to use more ride-sourcing
background of the people and any similarities or differences exist be­ trips on weekdays and less on weekends. Also, there exists a positive
tween cities from developed and developing countries from users relation between population density of a zone and work-to-home trips.
perspective. Hence, a joint modeling would benefit in inferring the Furthermore, vehicle ownership is found to be more likely to reduce the
factors, for instance - behaviour, demography, lifestyle, etc that might ride-sourcing trips on weekdays. The trip distribution model indicated
have an affect on both of the usage of car and motor-bike together as ride that, airport is a major ridesourcing trip attraction. A better under­
sourcing services. It is important to note that there exists very scarce standing of this finding is expected to help future parking planning at
evidence of this kind of research based on the data in the context of airports. Furthermore, their findings can be used to determine better
developing worlds. strategies to allocate drivers to rides or to estimate optimal fleet size
The benefits of this research are two-fold. Firstly, this study tries to when entering a new market.
find socio-demographic characteristics of the group of people who How ride-hailing services are experienced at an individual level was
intend to use motor-bikes and cars as a ride-sourcing service and to unclear before the investigation made by Lavieri and Bhat (2019). This
unveil which one is preferable while comparing them in-between. Sec­ study examined ride-hailing experience and related trip characteristics
ondly, this research helps to formulate policy for ride-sourcing service by observing the influence of the unobserved lifestyle of individuals,
providers based on users’ socio-economic and demographic back­ transportation-related choices, and socio-demographic factors. Based on
grounds which will eventually guide them to expand their services in an their findings they urged from both transportation as well as physical
efficient way. Expansion of ride-sourcing services in an optimized way perspective to discourage the substitution of short-distance “walkable”
would certainly reduce the number of personalized vehicles in a con­ trips by ride-hailing, i.e., through a pricing scheme that more heavily
gested city of a developing country. prices the first mile would be extremely effective.
The remaining paper is structured as follows: Section 2 conducts a Using a random parameters logit model, Barbour et al. (2019) made
critical review regarding the current state-of-the-art of incorporating an effort to identify the factors influencing bike-sharing, and to assess
socio-economic and demography as important factors in ride-sourcing whether bike-sharing services displace auto-trip or not and if yes then
services in the perspective of both developed and developing coun­ how. Travel behaviors, health-related factors, and socio-demographic
tries. Section 3 presents a detailed description of the study area, data factors were considered as explanatory variables. Age, income, trip
collection procedure and related descriptive statistics of each of the types, vehicle ownership were found to have a significant role in ride
categories of the factors considered. Section 4 describes the methodol­ hailing and mode substitution. The results of this paper are expected to
ogy selected for this study with mathematical expression. Section 5 il­ support developing a new understanding of how bike sharing decisions
lustrates the discussion of the outcomes with the suggestions of ride- were highly correlated with people characteristics, which could guide in
sourcing policy implication. Section 6 concludes the paper along with the further development of policies to improve the performance of
limitations and future scopes. bikesharing systems and to make them a more viable transportation
option.
2. Literature review Alemi et al. (2019) tried to understand the frequency of ride-sourcing
service usage by applying ordered probit and zero-inflated ordered
Several research has been carried out in developed worlds to unveil probit model with correlated error terms. This study investigated
the complex relationship between ride-sourcing usage and socio- different socio-economic and demographic factors those were later
economic characteristics. Therefore, at the outset of this paper, some identified as the most important variables in explaining ride-sourcing
recent studies and their attempt to link socio-economic and de­ trip frequency. By providing policymakers a better understanding of
mographic variables with ride-sourcing services have been reviewed. who might use ridesourcing services, potential plus points are expected
For example, Dias et al (2017) made an effort to understand the influ­ to come from ridesourcing, e.g., reduced traffic congestion, road infra­
ence of socio-economic and demographic variables on the use of ride- structure costs, greenhouse gas emissions, and parking demand, if these
sourcing and car-sharing services simultaneously using a bivariate or­ services serve as solutions for sustainable urban mobility by reducing
dered probit model. It concluded that educated young individuals with the use of single occupant ride hailing. Also, Dias et al. (2019) proposed
high income residing in high-density areas are the probable dominant an interesting approach of data fusion to gain a large source of better-
users of these disruptive services. Looking into some other studies, quality data of ride-sourcing services. A publicly available data was
Henao and Marshall (2019) pointed on equity issues among ride- fused with census and parcel data to infer trip purpose, OD information,

2
Md.H. Rahman et al. Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

and demographic characteristics. Then a multivariate ordered probit accessibility had an impact on ride-sourcing demand. A data-driven
model was deployed. The result indicated socio-demographic variables model was used to understand the relationship controlling socio-
as an important factor influencing trip frequency. An activity-based economic factors. The results showed that population density, employ­
pattern was extracted from the analysis. For example, lower-income ment, transit-stop density, and land-use mix were positively correlated
people were tending to make more ride-sourcing trips for utilitarian with ride-sourcing demand. Based on previously derived knowledge of
purposes while people with high incomes were prone to use service-trips how the built environment affects traditional modes of travel, they
more for the airport and recreational purposes. This study found heavy found that their results could be put together by both city planners, and
usage of ride hailing services on nights and weekends and suggested that TNCs to develop holistic policy initiatives, regulations, and planning
this finding may call for the development of special night or weekend strategies that would be able to utilize the possible benefits of ride-
travel forecasting models so that bottlenecks and congestion points can sourcing and to minimize its negative effects.
be better identified. Yu and Peng (2019) focused on the impacts of the On the other hand, Shokoohyar et al. (2020) focused on the acces­
built environment in understanding ride-service travel behavior. By sibility of Uber in the city. Population density, age, gender, education,
applying Geographically Weighted Poisson Regression (GWPR), the income, etc along with transportation infrastructure had been used as
spatial relationships were established between the built environment variables. K-means clustering was applied to develop a spatial lag model
and ride-sourcing demand. Modeling results presented strong relation­ that showed ride-hailing accessibility depends on the characteristics of
ships between ride-sourcing demand and built environment variables (i. the neighborhoods. This study indicated areas with dense population,
e., density, land use, infrastructure, and transit accessibility). higher number of male people, less access to amenities in the walkable
Haddad et al. (2019) attempted to estimate the socio-economic im­ distance, high crime rate produced more ride-sourcing trips. This study
pacts of e-hailing emergence on commuters’ travel time and accessibility pointed out some serious imbalance in availability and popularity in a
in São Paulo, Brazil. The results indicated that 83% of current e-hailing certain area and expected that the outcome of this study could be key
trips derived from trips by traditional motorized private modes. They factors for better expansion of ride hailing services. Yan et al. (2020)
also found that e-hailing supply had negligible effects on travel times attempted to predict ride-sourcing demand using publicly available data
and congestion. Moreover, Tarabay and Abou-Zeid (2019) developed a in Chicago. The explanatory variables they had used were related to the
hybrid choice model to predict how students of Lebanon switched their built environment, demographics, transit-supply data, etc. With an aim
trip choice from traditional modes to ride-sourcing services for social to predict demand, Random forest – a machine learning technique was
purposes. This behavioral shift was affected by travel time, cost, income, used. The results came out with higher predictive accuracy which out­
age, vehicle ownership, accessibility of transit, trip purpose, destination, performed the traditional multiplicative models. It is important to note
etc. that socio-economic and demographic variables contributed half of the
Additionally, Irawan et al. (2020) applied a structural equation predictive power of the model.
model to investigate the relationships among motorcycle-based ride- There are some review papers generated based on the ride-sourcing
sourcing (MBRS), motorcycle taxis, and public transport in the Jakarta system. For instance, Ricci (2015) provided a state-of-the-art on bike-
Metropolitan Area. The result showed that motorcycle-based ride- sharing with an aim to investigate their popularity, impact, operation,
sourcing was complementary to bus and commuter trains, but a and business implementation strategies. In the end, they concluded that
competitor of the motorcycle taxis. Since they found that the MBRS deploying bikes as a ride-service improved health and comfort, reduced
could support the existence of public transport, they suggested the travel time and cost, thus resulting in increased popularity among users.
transportation authority’s focus their attention more on improving the However, the authors also mentioned that there existed an unequal
maintenance of the public transport. Chen (2015) concluded that social distribution among young males from the user’s perspective. The service
and recreation trips, shorter trip length, and vehicle occupancy rates seems to facilitate those who are economically stable and have accept­
make dominance in choosing ride-sourcing service. able social status. Wang and Yang (2019) conducted a comprehensive
By deploying a structural equation model, Yu and Peng (2020) tried review on ride-sourcing demand, supply, pricing, incentives, operations,
to understand the effect of the built environment on ride-sourcing competitions, impacts, and regulations where an impact of socio-
mobility. The outcome affirmed a strong impact of the built environ­ economic, and demographic factors had been predominant for the pas­
ment. The model demonstrated that greater population, more jobs, good senger demand ride-sourcing market.
quality pavement, road and land use mix, accessibility tend to increase Apart from developed countries, a few studies have been carried out
ride-sourcing demand. Barbour et al. (2020) developed a multinomial in Bangladesh focused on ride-sourcing services as well. For instance,
logit model considering heterogeneity to understand usage rates of ride- Islam et al (2019) and Sakib and Mia (2019) – both studies carried out a
sourcing options. Along with the traditional socio-demographic vari­ descriptive analysis based on questionnaires to shade light on current
ables, travel and health-related factors were also investigated in this states of ride-sourcing business and services in Bangladesh. They
study. The outcomes postulated a probable way to reduce inequity. concluded that with different demographic backgrounds, a significant
Encouragement of sharing ride-sourcing services with the people of respondents had preferred to use this alternative transport service (i.e.,
lower-income, old age, and small children were considered beneficial for ride-sourcing). The potential reasons of this popularity were mainly
the entire transport ride-sourcing systems. Based on their findings, they flexibility, time-saving advantage, convenience, etc. The popularity was
stressed on encouraging sharing of single occupancy rides with others found among the young generation. Moreover, Abedin et al (2020)
for enhancing the efficiency of a vehicle and breaking the single occu­ recently conducted a study with an attempt to design a ride-sourcing
pancy mindset. service plan for the corporate workers (i.e., small and medium enter­
Moreover, Sabouri et al. (2020a) examined the connection between prise employees) of Dhaka City. They used cost, time, distance, and
ride-sourcing services and vehicle ownership. The study used the Na­ transport modes as their primary factors. To understand the travel be­
tional Travel Survey Data. To investigate the relationship, the authors haviors, this study also used demographic information of the users, e.g.,
used household socio-economic characteristics and built environments age, gender, occupation, affiliation, etc. They expected that their find­
as explanatory variables. To capture the non-linear relationship, a ings could be used by TNC services for designing a work commuter
multilevel Poisson model ensembled with random forest had been “pickup and drop-off” service. Moreover, in this study, a combined
deployed as the modeling method. The outcome exhibited a negative transportation measures were also suggested for those who were inter­
correlation between ride-sourcing service and vehicle ownership. They ested about giving a reliable work commute for their staff.
pointed out that their study would aid the planners in balancing the The understanding of how people use ride-sourcing services in a
parking supply and demand. Sabouri et al. (2020b) also investigated social network setting is required to assist policy makers and transport
how built environments such as density, diversity, design, distance, and planners in developing a more efficient ride-sourcing system. Such

3
Md.H. Rahman et al. Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

aspects were addressed by Sadri (2016) for joint trips to study the effects Table 1
of personal network attributes on the number of shared trips for different Descriptive statistics of response and explanatory variables.
activities by utilizing zero-inflated Poisson models. Furthermore, online Description of variables Count Percentage (%)
social media have become an indispensable part of our social life,
Gender
especially during the era of Covid-19. Identifying user communities Male 670 66.8
based on their social media communications could identify unobserved Female 333 33.2
factors affecting the ride-sourcing services and its users. To that end, Age
Mojumder et al. (2021) tried to understand the ride-sourcing risk 18–24 400 39.9
25–34 424 42.3
communication patterns by applying numerous natural language pro­ 35–44 116 11.6
cessing techniques on large-scale crowdsourced data collected from 45–54 35 3.5
social media to understand risk perception of the different stakeholders 55–64 14 1.4
associated with ride-sourcing services. Their findings indicated that the ≥ 65 14 1.4
Marital Status
ride-sourcing stakeholders are had been concerned about the various
Single 607 60.5
negative effects of restricted mobility due to pandemic. Married 396 39.5
From the discussion above, it is clear that a significant attempt has One Person Household
been made to understand how socio-economic and demographic factors Yes 440 43.9
affect the ride-sourcing choice and its frequent usage. But, social, eco­ No 563 56.1
Residential Density (per square km)
nomic, and demographic functions vary from one country to another 0 – 20 k 201 20
country. How people from developing countries react to ride-sourcing 20 k – 40 k 381 38
systems, what influences them in choosing ride-sourcing for their day- 40 k – 60 k 187 18.6
to-day trips, how their social status, economic condition, and de­ 60 k – 80 k 202 20.1
80 k+ 34 3.4
mographic characteristics impact their decision in making ride-sourcing
Education
trips are necessary to understand. So, in this research paper, the main Less than High School 17 1.7
objective is to develop a choice model to understand the socio- O-level or equivalent 18 1.8
demographic effects on ride-sourcing choices by the users of A-level of equivalent 65 6.5
Bangladesh and to make an inference about their choices between two Diploma 52 5.2
Graduate (Bachelor) 368 36.7
most preferable modes: motorbikes and cars. Findings will unearth the
Post-graduate or more 483 48.2
market penetration factors related to ride sourcing services. Addition­ Employment Status
ally, the results potentially could be used in travel demand forecasting Full Time 453 45.2
models to predict the demand service. Part-Time 68 6.8
Self Employed 132 13.2
Unemployed 350 34.8
3. Study area and data collection Occupation Sector
Government 42 4.2
This study aims to unveil the effect of socio-economic and de­ Private 464 46.3
mographic factors on ride-sourcing demand by applying an ordered Business 114 11.4
Other Services 383 38.2
choice method. Hence, to explore peoples’ (users’) opinions regarding Monthly Income (BDT)
their choice and to extract their socio-economic and demographic data, Less than 25 k 513 51.1
a questionnaire was prepared to conduct a survey over more than one 25,000 – 49,999 232 23.1
thousand people who use ride-sourcing services for their trips and are 50,000 – 74,999 132 23.1
75000–99999 53 13.2
living in Dhaka city-the capital of Bangladesh. According to BRTA report
100 k or more 73 5.3
2021 (February), Dhaka city alone accounts for 26% of the total regis­ Smartphone Availability
tered motorcycles and 85% of the total registered private cars of the Yes 961 95.8
country. Until now, BRTA approved 12 transport network companies No 42 4.2
and ride-sharing companies to operate. Two mostly used ride-sourcing Private Car Availability
Yes 245 24.4
companies (i.e., Uber and Pathao) have only extended their services to No 758 75.6
Chattogram, Sylhet and Rajshahi – the other three cities of Bangladesh. Motorcycle Availability
Other than ride-sourcing services, mostly courier and food delivery Yes 779 77.7
services are provided in those three areas. So it is imperative that it is No 224 22.3
Driving License
Dhaka, the capital where the highest numbers of TNCs are operating and
Yes 270 26.9
most of the people are using this as a service. The authors collected No 733 73.1
sample data using a random sampling method. It should be noted that Ride-sourcing with Motorcycle
respondents less than 18 years were removed from the dataset to create a Never shared 425 42.4
model that represents only adult users. Therefore, the final dataset Not in the past 30 days 178 17.7
1 to 3 rides per month 189 18.8
contains the response of 1003 adult ride-sourcing users. The question­ One ride per week 31 3.1
naire comprises 16 questions indicating different socio-demographic Twice or more rides per week 180 17.9
characteristics of the respondents. Ride-sourcing with Cars
Table 1 represents the description of explanatory variables selected Never shared 256 25.5
Not in the past 30 days 201 20
for the analysis. The variable-selection has been done with three steps.
1 to 3 rides per month 372 37.1
Firstly, previous studies where socio-economic and demographic factors One ride per week 102 10.2
were present were searched in Google Scholar, ProQuest, ResearchGate, Twice or more rides per week 72 7.2
Web of Science, Scopus, and society journal databases. Secondly, a pilot
study in the local context of ride-sourcing was carried out to sync it with
the variables used in the literature. Lastly, a questionnaire was prepared respondents, the dominating age range was 25 to 34 years (42.3%) and
following the preceding steps. second highest users were from 18 to 24 years old (39.9%). The per­
Among the respondents, the number of male users of ride-sourcing centage of married people (39.5%) was less than unmarried (60.5%)
services (66.8%) were more than that of female users. Among the while using ride-sourcing services.

4
Md.H. Rahman et al. Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

Also, households having multiple persons used ride-sourcing service tackle two response variables simultaneously. It can also account for the
(56.1%) more compared to those of one-person households. Surveyed correlation among the response variables (Lord et al., 2021). It indicates
data indicates that among the respondents, the majority of them were that if two or more dependent variables tend to show inter relationship,
graduates (84.9%), full-time employees (45.2%), had private jobs and share unobserved characteristics, then multivariate models (i.e.,
(46.3%) and had monthly income of less than 25 k (51.1%). Neverthe­ here bivariate probit model) can explain this correlation well and pro­
less, proportions of respondents were also significant who are unem­ vide less biased estimation of the parameters (Washington et al., 2020).
ployed (34.8%) and engaged in other services, for example, giving In Fig. 1, a basic framework of the entire study is given for easy go
tuitions in-home and coaching centers (38.2%). Also , 95.8% and 77.7% through with the flow of the paper. Model development process is
of the respondents had owned smartphones and motorcycles, respec­ mentioned as well.
tively, 75.6% of them did not have private cars, and 73.1% of the re­ In this research, authors have ordered and coded the response of
spondents did not have driving licenses. ride-sourcing service usage into five categories i.e., never shared = 0;
To come up with the trip frequency of the ride-sourcing users, the not in the past 30 days = 1; one to three ride(s) per month = 2; at least
authors have ordered the responses from 0 to 4 as individual categories. one ride per week = 3, and twice or more rides per week = 4 to un­
Each of these ordered-categories represents how frequently the derstand the frequency of using motorbikes and cars as ride-sourcing
respondent has made their trips using either motor-bike or car or both. services. As these choice data follow ordered structure from least to
The options of each of the level are: most severe service users, ordered probit models present a useful
analytical framework.
“Never shared = 0; not in the past 30 days = 1; one to three ride(s) In this research, a joint model comprising motorbike and car ride-
per month = 2; at least one ride per week = 3, and twice or more sourcing services is analyzed using a bivariate ordered probit (BOP)
rides per week = 4” model. It is derived by defining the ordered data, y for each of the ob­
servations. For instance, while analyzing two outcomes i.e., motor-bike
From Table 1, we can observe that among other categories, 37.1% and car ride-sourcing, according to Greene and Hensher (2010) and
and 18.8% of the respondents used cars and motorbikes one to three Greene (2012), the outcomes could be formulated as:
times per month, respectively as a ride-sourcing service. Interestingly,
frequent ride-sourcing users, i.e., twice or more per week, preferred yi1 = β1 Xi1 + εi1 , yi1 = k if ak− 1 < yi1 < ak where k = 0 to K1
motorcycles (17.9%) than cars (7.2%). It should be noted that the
collected data has a significant percentage of people who have never yi2 = β2 Xi2 + εi2 , yi2 = k if bk− 1 < yi2 < bk where k = 0 to K2 (1)
used motorbikes (42.4%) and cars (25.5%) as ride-sourcing services. where,
From Table 1, it can also be deduced that the use of cars as ride-source is yi1 = ordering of motor-bike choice as a ride-sourcing service and yi2
a little higher than that of motorbikes in this sample. The reason is easily = ordering of car choice as a ride-sourcing service; β1 and β2 = vectors of
understandable as it was stated in the previous section that the number the slope of the parameter estimates for motor-bike and car respectively;
of registered cars is more in Dhaka city compared to motorcycles ac­ Xi1 and Xi2 = vectors of explanatory variables that affect users’ frequent
cording to the BRTA report 2021 (February). choice of motor-bike and car ride-sourcing, respectively; a, b = threshold
parameters that define yi1 and yi2 ; k = order level (0 to 4); εi1 and εi2 =
4. Methodology error terms for motor-bike and car ride-sourcing service, respectively.
(This term is assumed to be normally distributed)
An extension of probit model that can deal with two dependent The correlated error terms which are important to derive whether
response variables at a time, namely bivariate ordered probit probit the joint modeling is justified or not could be estimated using the
model has been adopted for this current study. Probit models require following matrix, i.e.,
normal distribution as an assumption to model unobserved components. ( ) [( ) ]
In the literatures this assumption as well as the model are being used εi1
∼ Normal
0
, (1 ρ ρ 1) (2)
overwhelmingly. The normal distribution is motivated due to its’ appeal εi2 0
to central limit theorem, and also the distribution itself influences
where, ρ = correlation coefficient of the error terms.
human behavior modeling (Hess and Daly, 2014). Moreover, in most
The bivariate ordered probit model with ordered selection joint
real-life cases, normal distribution is proved to provide enough vignette
probability for motor-bike, yi1 = j and car, yi2 = q can then be defined as
of the unobserved random components (Train, 2012). Therefore, the ( )
probit model could be applied to estimate the model. From the earlier prob yi1 = j, yi2 = q given Xi1 , Xi2 =
studies, it has been evident that the ordered probit model is quite pop­ ( ( ) ) ( ( ) )
φ[(ak − β1 Xi1 ), bq − β2 Xi2 , ρ] φ[(ak − β1 Xi1 ), bq− 1 − β2 Xi2 , ρ]
ular in investigating the effect of socio-economic and demographic ( ) − ( )
− φ[(ak− 1 − β1 Xi1 ), bq − β2 Xi2 , ρ] − φ[(ak− 1 − β1 Xi1 ), bq − β2 Xi2 , ρ]
characteristics on ride-sourcing services. For instance, Dias et al. (2017)
used a bivariate ordered probit model where joint modeling was con­ (3)
ducted using ride-sourcing and car-sharing simultaneously. Dias et al.
(2019) also used a multivariate ordered probit model in predicting ride- where, φ = standard normal cumulative distribution function.
sourcing demand. Moreover, Alemi et al. (2019) deployed ordered Once the parameters β1 and β2 are estimated, the corresponding
probit, and zero-inflated ordered probit models for examining the effect values are interpreted depending on its plus-minus sign. For instance, if
of socio-demographic characteristics. However, very few studies have parameters exhibit positive sign, then it indicates an increase in the
considered two contemporary preferred modes of ride-sourcing services probability of the most severe order i.e., twice or more rides per week
together in the context of developing countries like Bangladesh. (=4) and decrease in the least severe outcome i.e., never shared (=0). If
Therefore, there is a need to study how various social and economic the sign indicates negative, then the converse is true.
backgrounds of the users make an impact on the frequency of using ride- Using these parameters of the BOP, only the most and least order
sourcing services. Moreover, according to Dias et al (2017), joint outcomes are interpretable. To observe the effect of intermediate out­
modeling would be an additional benefit to infer common unobserved comes i.e., not in the past 30 days = 1; one to three ride(s) per month =
factors e.g., behaviour, demography, lifestyle etc that may affect both 2; at least one ride per week = 3; marginal effects need to be computed at
the usage of car and motor-bike as ride sourcing services. As a result, the mean of the each of the categories according to Green and Hensher
bivariate ordered probit has been adopted as a modeling method. The (2010), Green (2012) and Washington et al. (2020). The marginal effect
advantage of bivariate ordered probit model is primarily its strength to could be formulated as:

5
Md.H. Rahman et al. Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

Fig. 1. Research framework of the current study.

prob(y = k) 5. Results and Discussion

= [Φ(ωk− 1 − βx) − Φ(ωk − βx)]β (4)
∂x
As mentioned above, a Bivariate-ordered Probit Model has been
where, prob (y = k) = probability of the ride-sourcing services users selected as a modeling method to explore the effect of socio-economic
experience kth - order of ride-trips number; ω = thresholds; k = order and demographic factors of Bangladesh on ride-sourcing usage. Here,
level (0 to 4); Φ() = probability mass function of the standard normal as response variables, two candidates have been selected i.e., motor-
distribution. cycle and car-ride sourcing. As in this research, a joint modeling

Table 2
Parameter Estimation Results for Bivariate-ordered Probit Model for Motorcycle and Car Ride-sourcing Response affected by Socio-economic Factors.
Description of the Variables Motorcycle Ride-sourcing Car Ride-sourcing

β SE z-value p-value β SE z-value p-value

Gender: (Base: Female)

Male 0.86 0.08 10.18 0.000*** − 0.19 0.08 − 2.24 0.024*
Age: (Base: 18–34)
45 – 54 − 0.49 0.11 − 4.32 0.000***
≥ 65 − 1.22 0.29 − 4.21 0.000***
Household Status (Base: No) − 0.16 0.07 − 2.14 0.032*
Household Density (Base: 0 – 20,000)
40,000 to 60,000 0.36 0.10 3.64 0.000*** 0.59 0.10 6.12 0.000***
≥ 60,000 0.27 0.09 3.02 0.002** 0.39 0.09 4.40 0.000***
Academic Background (Base: Less than High Scholl)
Post-graduate 0.120 0.09 2.23 0.026* 0.41 0.09 4.67 0.000***
Occupation Sector (Base: Govt.)
Private 0.24 0.09 2.64 0.008** 0.30 0.09 3.32 0.000***
Monthly Income (Base: 0 – 49,999 BDT)
50,000 BDT to 74,999 BDT 0.19 0.10 1.83 0.066.
≥75,000 BDT 0.33 0.13 2.65 0.008**
Smart Phone (Base: No)
Yes 0.69 0.29 2.40 0.016* 0.52 0.18 2.83 0.004**
Private Car (Base: No)
Yes 0.29 0.09 3.05 0.002**
Motorcycle (Base: No)
Yes − 0.19 0.11 − 1.69 0.091.
Driving License (Base: No)
Yes 0.19 0.11 1.74 0.082.
Threshold values
µ1 and Ω1 1.10 0.24 − 0.04 0.199
µ2 and Ω2 1.63 0.25 0.57 0.198
µ3 and Ω3 2.25 0.25 1.73 0.203
µ4 and Ω4 2.37 0.25 1.94 0.205
Sample Size 1003
Log-likelihood − 2623.4991
Wald-chi2 219.02

6
Md.H. Rahman et al. Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

approach has been decided to deploy, so a proper justification is respectively. The results seem to be analogous to the one of the studies
necessary whether the formulation of the model could be able to conducted in Austin, Texas by Yu and Peng (2019) where they indicated
recognize the effect or not. From Table 3, it could be observed that the that if population densities increased, ride-sourcing demand also
error correlation between motor-cycle (Bikes) and car-ride sourcing is increased and Sabouri et al. (2020b) also showed that ride-sourcing
positive (ρ = 0.27) and statistically significant at less than 1% signifi­ demand were positively correlated with population density in almost
cance level. This result indicates that the potential unobserved factors 24 regions in United States. This finding may have two important policy
could have a strong chance to affect positively in using both motor-cycle implications. Firstly, car ride-sourcing companies may transfer their
and car-ride sourcing. It implies that people of Dhaka take both of the focus on high-density areas. It is obvious that high-density areas tend to
rides as their daily-life mobility services frequently. So, it is evident from generate more trips. Thus, it may increase the exposure of a particular
this result that a joint modeling approach could be beneficial in deter­ company to promote their services. Secondly, as alone in Dhaka, almost
mining the impacts of the socio-economic and demographic factors on 12 ride-sourcing companies are being operated and therefore, high
choice as well as use of motor-bikes and cars. market competition exists here, so service expansion towards high
Table 2 represents the final outcomes of the model consisting of all density areas may trigger huge returns for the company. Also, as mo­
statistically significant variables with less than 0.1%, 1%, 5%, and 10% torbikes tend to be less popular in high-density areas, therefore, services
significance level, respectively. Initially, sixteen variables associated like Pathao may propose new schemes and proposals to popularize their
with socio-economic and demography had been taken. After conducting services in those areas.
several trials, eleven variables at different categories have been found People having completed post-graduation are likely to use both
statistically significant. sources of rides for their trip compared to those who did not even pro­
From the results given in Table 2, it could be inferred that males are mote to high schools. As expected perhaps more academic qualification
likely to exhibit their preference more in motorbike ride-sourcing (β = implies economic affluence because of having handsome salary in
0.86). The result seems similar to the study of Ricci (2015) where males Bangladesh. However, while comparing between these two modes, the
were more likely to be engaged in bike-sharing, whereas females are preference is more for car ride-sourcing (β = 0.41) relative to the
more likely to use car ride-sourcing (β = -0.19) compared to males. Both motorcycle ride-sourcing (β = 0.12). The result is somewhat similar to
of the results are somewhat expected in the social context of Bangladesh. the study of Henao and Marshall (2019) where it has been stated that car
Male-users are more likely to save the ride-cost and shorten journey time ridesourcing users have high educational qualification. Highly educated
and prefer easy accessibility during boarding and alighting which is persons may perceive well the disadvantage of using public transit as
getable through the usage of motor-bikes. These might skew their choice well as safety issues of using motorcycles as a ride-sourcing service in
towards motorcycle over car ride-sourcing. On the contrary, females are Dhaka city. This result indicates to users’ unsatisfactory level towards
more likely to choose a ride that delivers both comfort and safety. This public transit and motorbikes. Also, highly educated people may think of
counterintuitive finding might be crucial in terms of safety. If male users earning more by registering their personal cars as ride-sourcing services.
emphasize on saving time, then they are likely to compromise their People working in the private sector are more likely to use ride-
safety during the journey. Therefore, infrastructures with more safety sourcing services compared to those who have government jobs. Here,
and robust regulations as well as social awareness regarding safe the magnitude of the coefficient is slightly greater for car ride-sourcing
mobility in motorbikes could be emphasized through policy (β = 0.30) in comparison to motor-cycle (β = 0.24). It indicates that
implementation. people having private jobs tend to use car ride-sourcing slightly more.
People aged between 45 and 54 years and greater or equal to 65 High wage, strict time management, social-status might be the probable
years are found less likely to choose motor-bike ride compared to other reasons for this selection. Moreover, ride-sourcing services, especially
age groups. However, no statistically significant variable has been found via car, may assure greater flexibility and easy accessibility towards the
from different age categories in case of car ride-sourcing. The findings destinations. These findings seem to provide a cluster of working people
demonstrate that aged people may not be willing-to-accept motor-bikes which indicates that the primary focus of a ride-sourcing company may
as a ride-sourcing service if they prioritize safety and comfort more. start providing services for the employees of private sectors. Also, it is
Also, aged people may not feel comfortable in using the newly intro­ equally important to find a way to promote ride-sourcing services
duced technology-based system. On the contrary, people aged 18 to 34 among government servants. In this case, one possible solution may be
years are likely to exhibit an adventurous nature in selecting mode decreasing fare and advertising affordable schemes for both of the ride-
choice. They might prefer taking more risks to reach their destinations. sourcing modes for the government service holders.
They perhaps prioritize travel-time and cost over comfort and safety. In Income exhibits interesting results by observing that people having
this case, policy makers may think of alternative marketing strategies to higher incomes tend to use motorcycle ride-sourcing. For the monthly
promote and popularize ride-sourcing services among elderly people to income category 50,000 to 74,999 BDT (85BDT ≈1 USD) and more than
increase their acceptance level. Moreover, these findings may trigger to 75,000 BDT, the coefficient values are 0.19 and 0.33 respectively. These
find out other alternative groups who may have similar sort of concerns results indicate that people with higher income are likely to use
about ride-sourcing services. motorcycle ride-sourcing services more frequently. But surprisingly, no
People belonging to household densities of 40,000 to 60,000/ km2 significant results have been found in case of car ride-sourcing service.
and more than 60,000/ km2 tend to use both of the ride-sourcing The result of selecting motor-bikes as a ride-sourcing service is similar to
frequently compared to 0 to 20,000/ km2 of density. If magnitude of β the study of Dias et al. (2017) where young people with high income
is observed, car ride-sourcing is likely to be preferable for both cate­ preferred ride-sourcing. Similarly, Barbour et al. (2019) found a positive
gories i.e., 40,000 to 60,000/ km2 (β = 0.59) and more than 60,000/ correlation between bike-sharing and high income in the United States.
km2 (β = 0.39) compared to motorcycle where the magnitudes show A probable reason may be greater flexibility to call up motorcycles using
0.36 and 0.27 for the density 40,000 to 60,000/ km2 and 60,000/ km2, app services while planning for a trip, as well as low cost may lead to
choosing motorbikes more compared to cars. However, it is expected
that income should have positive association with education though this
Table 3 relationship between these two does not exist in the finding of this study.
Error Correlation Table. This study has also found that highly educated people prefer to choose
Error ρ SE z-value p-value cars more, whereas the completely opposite scenario is being observed
while examining the effect of income. Thus, this contrary outcome may
Motorcycle and car ride-sourcing 0.27 0.037 7.07 0.000*** need further explorations. Our result suggests that if a user is highly
(Note: Significant values: 0 ‘***’, 0.001 ‘**’, 0.01 ‘*’, 0.05 ‘.’, 0.1 ‘ ’). qualified as well as poses high income, he/she may choose motorbikes to

7
 Md.H. Rahman et al. Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

enjoy ride-sourcing service, however, the magnitude of choosing mo­

Pseudo-elasticity torbikes in terms of education is found less compared to cars. Hence, an
extended study needs to be executed to find the relationship between
− 6.183 income and education level while considering their effects of choosing
different ride-sharing services . This result resembles that of the study of
− 1.26
− 5.41
− 4.55

− 2.52

− 3.09
21.56
13.31

13.16

13.21
9.59

5.54

9.68

6.11
Irawan et al (2020).
As expected, people who have smart phones are more likely to use
ride-sourcing services. The magnitude of the parameter is more for
0.131***
0.078***

0.074***

0.054***
− 0.035*

− 0.027*

0.056**
− 0.007
− 0.029

− 0.014

− 0.017
motorcycles (β = 0.69) than that of cars (β = 0.52). This result is ex­

0.035.
0.032

0.066
pected as it is more likely that the prevalence of smart phones equipped
5

with internet facilities promotes ride-sourcing services. These results

match with the outcomes of a study conducted in California by Alemi
0.030***
0.020***

0.021***
− 0.010*

− 0.008*

0.015**

0.023**

0.015**
− 0.002
− 0.009

− 0.004

− 0.005
et al. (2018) where they stated that people who use transportation-

0.009.
0.008
related apps tend to avail and adapt more with on-demand ride ser­
4

vices compared with those who never used them or just heard the names
of them only.
0.061***
0.052***

0.067***

0.049***

0.041***
− 0.030*

− 0.027*
− 0.007
− 0.035

− 0.014

− 0.018
0.111* People who have cars of their own (β = 0.29) and those who have
0.025.

0.029.
Car Ride-sharing(Ordered-category)

driving licenses (β = 0.19) as well are liable to avail car ride-sourcing

3

services. However, these two variables have not shown any significant
result for motorcycle ride-sourcing in joint modeling. These findings
− 0.070***
− 0.042***

− 0.038***

− 0.029***

− 0.030***
− 0.022**

may be more insightful if it is integrated with education and occupation

0.019*

0.015*

− 0.019.
0.004
0.014

0.007
− 0.017

0.009

level in this study. People with higher education and private jobs tend to
select cars as a ride-sourcing service. So, this cluster of socio-
2

demographic people might be the target group for the ride-sourcing

companies to expand their businesses. A different policy initiative
− 0.152***
− 0.108***

0.123***

0.090***

− 0.082***
0.055*

0.048*

0.179*

− 0.054.

could be adopted for these cohorts of people compared to those who do

0.012
0.059

0.025
0.047

0.031
Marginal Effect of Bivariate-ordered Probit Model for Socio-economic Factors affecting on Motorcycle and Car Ride-sourcing Services.

not belong to these categories.

1

Moreover, people who have personal motorcycles tend to use less (β

= − 0.19) motorcycles as a ride-sourcing service. The results echoed the
Pseudo-elasticity

findings of the study by Lavieri et al. (2018), Barbour et al. (2019), and
Sabouri et al. (2020a) where these studies found less association of using
ride-sourcing service with people having their own cars and motorcycles
− 1.41

7.16

8.96

6.79

0.64

− 6.64

− 3.15
27.96

− 15.42
− 24.97

14.48
10.61

13.55

18.34

in the USA. The result is not surprising as if people have their own ve­
hicles (i.e., motorbikes) at home, it reduces the tendency to seek ride
sourcing services.
To investigate the effect of intermediate ordered categories other
0.167***

− 0.091***
− 0.137***

0.091***

0.105***
0.066**

0.055**
0.044*

0.085*
0.042.

− 0.040.

than the highest and the lowest ride-sourcing choices, marginal effect
− 0.009

0.004

− 0.019

analysis is performed. In Table 4, the marginal effect of the two ride-

5

sourcing services are presented with several significant levels. For

each socio-economic and demographic variable, the values of the mar­
0.024***

− 0.014***
− 0.026***

0.010**
0.008**

0.010**

0.018**

ginal effect are calculated for each ordered category. These marginal
0.0005
0.005*

0.007*

0.005.
− 0.001

− 0.006

− 0.003

effects represent the mean change in probability of specific choice that

corresponds to a one-unit change of an explanatory variable (i.e., socio-
4

economic variables). For example, when people are using smart phones,
Motorcycle Ride-sharing(Ordered-category)

their use of motor-bikes and cars twice or more in a week will increase
0.111***

− 0.066***
− 0.153***

0.038***
0.030***

0.035***
0.029**

0.093**
0.024*

(Note: Significant values: 0 ‘***’, 0.001 ‘**’, 0.01 ‘*’, 0.05 ‘.’, 0.1 ‘ ’).
0.021.

by 10.5% and 6.6%, respectively, than the people who are not using
− 0.005

0.002

− 0.024

− 0.011

smartphones. The other variables and their corresponding categories

3

could be explained in a similar manner.

The marginal effects in Table 4 are utilized to estimate pseudo-
0.032***

0.107**

elasticities of the independent variables on the expected total number

0.0005

0.0008

0.0003

0.0001
0.021*

0.043.
0.004

0.002

0.002

0.004

0.004

0.001

of trips per year for both motorcycle and car ride-sourcing services by
assigning cardinal values for each ordinal level. This calculation is done
2

Academic Background (Base: Less than High Scholl)

−
−
−

−
−

based on the study conducted by Dias et al. (2017). We assigned the

following cardinal values for the ordinal levels in the model: (1) Never
− 0.33***

− 0.14***
− 0.10**
0.19***
0.42***

− 0.006
0.09**

0.12**

0.26**

0.074.

shared = 0 (0 trips per month), (2) Not in the past 30 days = 3 (0.25 trips
0.034
0.08*

0.07.
0.02

Monthly Income (Base: 0 – 49,999 BDT)

per month), (3) 1–3 times per month = 12 (2 trips per month), (4) 1 ride
Household Density (Base: 0 – 20,000)
1

per week = 48 (4 trips per month), (5) Two or more rides per week =
Occupation Sector (Base: Govt.)

144 (12 trips per month). These values are multiplied with the corre­
Description of the Variables

Household Status(Base: No)

sponding marginal effects and added up separately for motorcycle and

50,000 BDT to 74,999 BDT

Driving License (Base: No)

Smart Phone (Base: No)

car ride-sourcing services to estimate the pseudo-elasticities of the in­

Gender: (Base: Female)

Private Car (Base: No)

Motorcycle (Base: No)

dependent variables in Table 4. The difference between the pseudo-

Age: (Base: 18–34)

40,000 to 60,000

elasticities for the two services i.e., car and motorcycle, provide the
Post-graduate

≥75,000 BDT

relative preference of one service over the other. For example, in com­
≥ 60,000

parison to government sector employees, a private sector employee

45 – 54

Private
Table 4

takes, on average, 8.96 more motorcycle ride-sourcing trips per year and
≥ 65
Male

Yes

Yes

Yes

Yes

9.59 more car ride-sourcing trips per year than the government sector

8
Md.H. Rahman et al. Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

employee and their difference indicates that a private sector employee Also, consideration of orthogonal study design could be a potential
takes more car ride-sourcing trips than motorcycle ride-sourcing trips. approach to address the similar issue to adjust for confounding factors
and intermediate variables. Therefore, incorporating causal inference
6. Conclusion could have a chance to unveil effect of latent, immeasurable and con­
founding factors with counterfactual analysis.
The study aims at assessing the impact of social, economic and de­
mographic characteristics of the users on usage frequency of car and CRediT authorship contribution statement
motorcycle ride-hailing services in Dhaka – a megacity of Bangladesh.
To achieve this goal, a joint model on usage of motor-bikes and cars as Md. Hishamur Rahman: Conceptualization, Methodology, Soft­
ride-sourcing services was developed by deploying Bivariate-ordered ware, Formal analysis, Investigation. Soumik Nafis Sadeek: Method­
Probit Model. The model was found effective in estimating the impact ology, Software, Resources, Data curation, Formal analysis,
of socio-economic and demographic variables on both travelers’ choices Investigation. Asif Ahmed: Validation, Investigation, Writing – review
in selecting ride-sourcing services as well as the frequency of their usage. & editing. Shakil Mohammad Rifaat: Supervision, Writing – review &
The results show that males, young travelers, earning higher degrees, editing. Masnun Abrar: Data curation, Writing – review & editing.
having private jobs, earning higher income, availing smart phones in­
crease the frequency of usage of both ride-sourcing services in the Declaration of Competing Interest
context of Dhaka – capital of Bangladesh, a developing country. The
study finds that male travelers are more prone to use bike ride-sourcing The authors declare that they have no known competing financial
whereas women prefer more to use car for riding. TNCs and ride- interests or personal relationships that could have appeared to influence
sourcing companies may enhance the number of female scooty riders the work reported in this paper.
so that more females can consider themselves to use bike-services too.
Also, estimated parameters indicate that if the household density is more Acknowledgements
in an area, people having higher educational qualification and engaging
in private jobs, then propensity to select car increases as ride-sourcing Authors would like to acknowledge the efforts and help of two un­
compared to motorcycle. Policy makers of ride-sourcing service com­ dergraduate students of IUBAT – International University of Business
panies can use this information by concentrating more on dense loca­ Agriculture and Technology – Md. Jahidul Islam and Alamin Sumon,
tions with higher number of educational institutions and private during the data collection for this study.
organizations to increase car riding service. Additionally, model finds
that people who have smartphones are more likely to choose motorcycle
Funding
as their preferable mode. The outcomes are found similar to those of the
earlier studies carried out in the developed countries. Especially, gender,
The authors did not receive any grants from any funding bodies
household density in a area, academic qualifications, income, technol­
related to public, private, commercial, or non-profit sectors.
ogy usage, car ownerships are found to have similar impact on peoples’
choice for selecting between motorcycle and car as a service of ride-
References
sourcing in both developing and developed countries. However, other
explanatory variables e.g., age, job types might not show similar effect in Abedin, N., Rahman, M.M., Hossain, M.I., Hisazumi, K., Ahmed, A., 2020. Travel
both developed and developing countries perhaps due to the variation of behavior of SME employees in their work commute in emerging cities: A case study
individual choices. in Dhaka City, Bangladesh. Sustainability 12 (24), 10337. https://doi.org/10.3390/
su122410337.
Ongoing popularity and increasing usage of car and motorcycle ride- Alemi, F., Circella, G., Handy, S., Mokhtarian, P., 2018. What influences travelers to use
sourcing in developing countries, TNCs might adapt various policies to Uber? Exploring the factors affecting the adoption of on-demand ride services in
expand their services more. Valuable information while finding the california. Travel Behav. Soc. 13, 88–104.
Alemi, F., Circella, G., Mokhtarian, P., Handy, S., 2019. What drives the use of
impact of socio-economic and demographic factors on choosing car and ridehailing in California? Ordered probit models of the usage frequency of Uber and
bike ride sourcing service in this study would certainly help policy Lyft. Transport. Res. Part C: Emerg. Technol. 102, 233–248. https://doi.org/
makers to expand more market penetration. Moreover, findings in this 10.1016/j.trc.2018.12.016.
Barbour, N., Zhang, Y., Mannering, F., 2019. A statistical analysis of bike sharing usage
study can be used in travel demand forecasting models which eventually
and its potential as an auto-trip substitute. J. Transport Health 12 (January),
would help the users to choose preferable ride-sourcing services bearing 253–262. https://doi.org/10.1016/j.jth.2019.02.004.
in mind the traffic congestion on roads. Barbour, N., Zhang, Y., Mannering, F., 2020. An exploratory analysis of the role of socio-
This study has some limitations as well. Considering the large sample demographic and health-related factors in ridesourcing behavior. J. Transport
Health 16 (February), 100832. https://doi.org/10.1016/j.jth.2020.100832.
size, this study was limited to socio-economic and demographic vari­ BRTA Report, 2018. Number Of Registered Motor Vehicles In Bangladesh (Yearwise).
ables only which might have created omitted variable bias. Travel URL: https://brta.portal.gov.bd/sites/default/files/files/brta.portal.gov.bd/
characteristics such as preferred ride-sourcing mode based on time-of- monthly_report/d4d56177_644f_44f8_99c4_3417b3d7b0f4/MV_statistics-
bangladesh-march-18.pdf.
day (i.e., peak/off-peak) were not considered in this study and this BRTA Report, 2021 (February). Number Of Registered Motor Vehicles In Bangladesh and
may provide more interesting findings. Furthermore, the study did not Dhaka (Yearwise).
consider interaction effects among the explanatory variables that might BRTA Report, 2020. Number of registered ride-sharing companies in Bangladesh. URL:
http://www.brta.gov.bd/sites/default/files/files/brta.portal.gov.bd/page/
have unveiled more insights on the relation between various socio- 1758a94f_5ca7_40bc_aa1e_a00bc47dbc6b/List%20of%20approved%20Ridesharing
economic and demographic factors and ride-sourcing services. More­ %20Companies.pdf.
over, data balancing methods were not considered in this study that Chen, Z., 2015. Impact of ride-sourcing services on travel habits and transportation
planning. M.Sc Thesis. University of Pittsburg.
could have removed model biases towards one category to another Dias, F.F., Lavieri, P.S., Garikapati, V.M., Astroza, S., Pendyala, R.M., Bhat, C.R., 2017.
categories. Therefore, these limitations could be addressed in the future A behavioral choice model of the use of car-sharing and ride-sourcing services.
extension of this kind of research. Moreover, more complex discrete Transportation 44 (6), 1307–1323. https://doi.org/10.1007/s11116-017-9797-8.
Dias, F.F., Lavieri, P.S., Kim, T., Bhat, C.R., Pendyala, R.M., 2019. Fusing Multiple
choice analysis e.g., latent class ordered choice model, random-
Sources of Data to Understand Ride-Hailing Use. Transp. Res. Rec. 2673 (6),
parameter multinomial logit model could be tested in modeling indi­ 214–224. https://doi.org/10.1177/0361198119841031.
vidual users’ mode preferences of availing ride-sourcing services. DMR, 2019. 110 Amazing Uber stats and facts (2019) by the numbers. Accessed on April
Additionally, in this study, only the effect of statistical significant vari­ 20, 2020. URL: https://expandedramblings.com/index.php/uber-statistics/.
Gilibert, M., Ribas, I., & Rodriguez-Donaire, S. (2017). Analysis of mobility patterns and
ables have been addressed. However, insignificant variables may also be intended use of shared mobility services in the Barcelona region. In: European
considered to measure the indirect effects as intermediate variables. Transport Conference 2017: Strands, pp. 1–20.

9
Md.H. Rahman et al. Transportation Research Interdisciplinary Perspectives 12 (2021) 100492

Greene, W.H., 2012. Econometric Analysis. Prentice Hall, Pearson. Sabouri, S., Brewer, S., Ewing, R., 2020b. Exploring the relationship between ride-
Greene, W.H., Hensher, D.A., 2010. Modeling Ordered Choices: A Primer. Cambridge sourcing services and vehicle ownership, using both inferential and machine
University Press. learning approaches. Landscape Urban Plann. 198 (March), 103797 https://doi.org/
Haddad, E.A., Vieira, R.S., Jacob, M.S., Guerrini, A.W., Germani, E., Barreto, F., 10.1016/j.landurbplan.2020.103797.
Bucalem, M.L., Sayon, P.L., 2019. A socioeconomic analysis of ride-hailing Sadri, A.M. (2016). Social Network Influence on Ridesharing, Disaster Communications,
emergence and expansion in São Paulo, Brazil. Transportation Research and Community Interactions.“ ProQuest Dissertations and Theses (January): 284.
Interdisciplinary Perspectives 1, 100016. https://doi.org/10.1016/j. https://docs.lib.purdue.edu/open_access_dissertations/1230.
trip.2019.100016. Sakib, M.N., Mia, M.H., 2019. The ride-sharing services in Bangladesh: Current status,
Henao, A., Marshall, W.E., 2019. The impact of ride-hailing on vehicle miles traveled. porspects, and challenges. Europ. J. Busin. Manage. 11 (31), 41–52.
Transportation 46 (6), 2173–2194. https://doi.org/10.1007/s11116-018-9923-2. Shokoohyar, S., Sobhani, A., & Ramezanpour Nargesi, S. R. (2020). On the determinants
Hess, S., Daly, A., 2014. Handbook of Choice Modelling. Elgar original reference. Edgar of Uber accessibility and its spatial distribution: Evidence from Uber in Philadelphia.
Elgar Publishing. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, (October 2019),
IDLC monthly: Business Review (2018). Ride-sharing in Bangladesh: Disrupting the way 1–15. Doi: 10.1002/widm.1362.
we commute. IDLC Finance Limited. URL: https://idlc.com/public/documents/mbr/ Tarabay, R., & Abou-Zeid, M. (2019). Modeling the choice to switch from traditional
18/Monthly%20Business%20Review%20-%20June%202018.pdf. modes to ridesourcing services for social/recreational trips in Lebanon.
Irawan, M.Z., Belgiawan, P.F., Tarigan, A.K.M., Wijanarko, F., 2020. To compete or not Transportation. Springer US. Doi: 10.1007/s11116-019-09973-x.
compete: exploring the relationships between motorcycle-based ride-sourcing, Train, K.E. (2012). Discrete Choice Methods with Simulation (2nd Edition), Cambridge
motorcycle taxis, and public transport in the Jakarta metropolitan area. University Press. DOI: 10.1017/CBO9780511805271.005.
Transportation 47 (5), 2367–2389. https://doi.org/10.1007/s11116-019-10019-5. Wadud, Z., 2020. The effects of e-ridehailing on motorcycle ownership in an emerging-
Islam, S., Huda, E., Nasrin, F., 2019. Ride-sharing service in Bangladesh: Contemporary country mehacity. Transportat. Res. Part A: Policy Pract. 137, 301–312.
states and prospects. Int. J. Business Manage. 14 (9), 65–75. Wang, H., Yang, H., 2019. Ridesourcing systems: A framework and review. Transport.
Lavieri, P.S., Dias, F.F., Juri, N.R., Kuhr, J., Bhat, C.R., 2018. A Model of Ridesourcing Res. Part B Methodol. 129, 122–155. https://doi.org/10.1016/j.trb.2019.07.009.
Demand Generation and Distribution. Transp. Res. Rec. 2672 (46), 31–40. https:// Washington, S., Karlaftis, M.G., Mannering, F., Anastasopulos, P., 2020. Statistical and
doi.org/10.1177/0361198118756628. Econometric Methods for Transportation Data Analysis, 3rd Edition. Taylor and
Lavieri, P.S., Bhat, C.R., 2019. Investigating objective and subjective factors influencing Franxcis Group, Routledge.
the adoption, frequency, and characteristics of ride-hailing trips. Transport. Res. Part Yan, X., Liu, X., Zhao, X., 2020. Using machine learning for direct demand modeling of
C: Emerg. Technol. 105 (May), 100–125. https://doi.org/10.1016/j. ridesourcing services in Chicago. J. Transp. Geogr. 83 (February), 102661 https://
trc.2019.05.037. doi.org/10.1016/j.jtrangeo.2020.102661.
Lord, D., Qin, X., Geedipally, S., 2021. Highway Safety Analytics and Modeling. Elsevier. Yu, H., Peng, Z.-R., 2019. Exploring the spatial variation of ridesourcing demand and its
Mojumder, M.N.H., Ahmed, M.A.A., Sadri, A.M., 2021. Identifying Ridesharing Risk, relationship to built environment and socioeconomic factors with the geographically
Response and Challenges in the Emergence of Novel Coronavirus (COVID-19) Using weighted Poisson regression. J. Transp. Geogr. 75, 147–163. https://doi.org/
Interactions in Uber Drivers Forum. Front. Built Environ. 7, 4. 10.1016/j.jtrangeo.2019.01.004.
Ricci, M., 2015. Bike sharing: a review of evidence on impacts and processes of Yu, H., Peng, Z.R., 2020. The impacts of built environment on ridesourcing demand: A
implementation and operation. Res. Transport. Busin. Manage. 15, 28–38. https:// neighbourhood level analysis in Austin, Texas. Urban Studies 57 (1), 152–175.
doi.org/10.1016/j.rtbm.2015.03.003. https://doi.org/10.1177/0042098019828180.
Sabouri, S., Park, K., Smith, A., Tian, G., Ewing, R., 2020a. Exploring the influence of
built environment on Uber demand. Transport. Res. Part D: Transp. Environ. 81,
102296. https://doi.org/10.1016/j.trd.2020.102296.

10