Public Preferences of Shared Autonomous Vehicles in Developing Countries: A Cross-National Study of Pakistan and China

Shared autonomous vehicles (SAVs) are rapidly emerging as a viable alternative form of public transportation with the potential to provide adequate and user-friendly, on-demand services without having vehicle ownership. It has been argued that SAVs could revolutionize transportation systems and our current way of life. Although SAVs are likely to be introduced in developed countries ﬁrst, there is little doubt that they would also have a signiﬁcant eﬀect and enormous market in developing nations. This study aimed to investigate the factors that inﬂuence public acceptance of SAVs, as well as the current public attitude toward SAVs, in two developing countries, namely, Pakistan and China. A stated preference survey was conducted to understand respondents’ travel patterns, preferences, and sociodemographic data. A total of 910 valid responses were gathered: 551 from Lahore, Pakistan, and 359 from Dalian, China. A multinomial logit model and a mixed multinomial logit model with panel eﬀect were used for data analysis. The results suggested that generic attributes, such as respondents’ waiting time, travel time, and travel cost were found to be signiﬁcant in both cities. The results indicate that sociodemographic characteristics, such as education, income, travel frequency in a week, and people who had driver’s licenses, are signiﬁcantly correlated with respondents’ interest in using SAV in Lahore. The results also showed that people who had a private car indicated a greater interestin SAVs in Dalian. The study provides a new perspective to understand the public preferences toward SAVs in developing countries with diﬀerent economies and cultures, as well as a benchmark for policymakers to make eﬀective policies for the future implementation of SAVs.


Introduction
In recent years, autonomous vehicles (AVs) have emerged as a new form of transportation with the potential to improve traffic safety, efficiency, and urban environment. Shared autonomous vehicles (SAVs), also called robo-taxis or shuttles, can provide flexible and door-to-door service, with users able to summon SAVs using mobile device applications rather than having to find and walk to vehicles [1]. e amalgamation of AVs and shared mobility could further revolutionize personal transportation, given that SAVs provide the same level of convenience as more traditional forms of private transportation but without the same responsibilities and cost of ownership [2]. e US National Highway Traffic Safety Administration has reported that 90% of all vehicle accidents occur due to human errors [3]. erefore AVs, by eliminating human errors, could considerably reduce the number of vehicle accidents [4,5]. Some researchers have suggested that the greenhouse gas emissions of SAVs could be as much as 90% less than those of the current passenger vehicles, and SAVs would dominate personal mobility options by 2050 [6]. SAVs could also overcome barriers associated with traditional dynamic riding services [7,8]. It has been proposed that the automatic relocation and pick-up of SAVs could enhance their availability and flexibility [9].
SAVs are likely to transform the structure of urban areas, towns, and neighborhoods [7,10]. e deployment of SAVs will reduce parking demands in urban centers, and enable the available space of parking to be used for various economic activities, infrastructure development, and other active and green transportation modes, such as walking, bicycling, and make the downtown area more attractive and sustainable [11]. e operational costs for SAVs are $0.13-0.50 per mile, and the use of SAVs could lead to a 90% reduction of parking spaces [12,13]. It is estimated that each SAV could replace 3-14 conventional vehicles, given that SAVs could provide low-cost mobility services [14]. One latest study revealed that 10% of survey respondents were willing to move farther from the city center in the presence of AVs and SAVs in the next few years, while 15% of the survey respondents indicated that they would likely move closer to the city center, which has an impact on travel behavior and land use [7]. e initiation of SAVs also offers an opportunity for employment and business growth in the charging zone and reduces the adverse impact of road congestion pricing [15,16].
Despite the overwhelming benefits of SAVs, people are more curious about security, safety, interaction with pedestrians, and system failure [4]. SAVs might increase vehicle-miles traveled (VMT) by 8% to 17% for relocation, arrival, and departure purposes [9]. e main reasons for the increase in VMT are induced traffic demand and empty trips of SAVs [17]. e increase in VMT can be a cause of congestion and therefore impedes sustainable urban development [16]. A recent study suggests that SAVs can create bottlenecks for other road users by stopping on the curbside to take and drop passengers [18].
In developing countries such as China and Pakistan, rapid urbanization has profound negative effects in major metropolitan areas. Excessive urbanization causes a rise in travel demand and thus creates transportation problems. In particular, the absence of transportation infrastructure, insufficient public transportation, inadequate traffic management, and the lack of research and development have led to severe problems [19]. For example, Lahore, Pakistan, is plagued by severe traffic congestion. is is largely attributed to the inadequacy of the existing public transportation system, which is inefficient, disjointed, and accounts for only 20% of the mode share [20,21]. Traffic congestion has several negative consequences, such as increases in energy consumption, travel cost, travel time, air pollution, and traffic accidents [22]. e cost of traffic congestion in Lahore has been estimated at $519 million per year [20]. China is also facing severe problems due to rapid urban growth and motorization. For example, in Dalian, private car ownership rose from 0.65 million in 2010 to 1.61 million in 2017. e modal share of private car travel increased from less than 15% in 2010 to 28.8% in 2016. is increase in private car travel has caused several negative effects, such as increases in congestion, accidents, fuel consumption, air pollution [23], and public transportation ridership stagnation [23,24]. However, technologies and policies are evolving very fast in China. Kuhnert et al. [25] stated that China was likely to have a similar level of technological development as the USA and Europe by 2030. Furthermore, they argued that the percentage of Chinese using traditional vehicles will decrease substantially by 2030, due to a shift toward shared modes of transportation such as robo-taxis and shared bikes.
To the best of the authors' knowledge, there are very few studies performed to compare public attitudes toward SAVs in developing countries.
is study meets the need by comparing the factors influencing public attitudes toward SAV adoption in two developing countries with different cultures and economies, namely, Pakistan and China. e objective of this study is to explore the public acceptance of SAVs in developing countries using a stated preference (SP) survey and discrete choice modeling techniques to identify the significant factors that can affect people's interest in SAVs in the presence of conventional modes of transportation. is is motivated by the fact that public acceptance is required for the rapid development and implementation of SAVs. e findings of this study provide a benchmark for policymakers tasked with making effective policies for the future implementation of SAVs in developing countries.
is study considers SAVs without pooled options, which provide the same level of privacy as a private car. On the contrary, SAVs with pooled options imply the sharing of a vehicle with strangers, which is a downside for many passengers [26]. e remainder of the paper is organized as follows. In Section 2, the existing literature on public attitudes toward SAVs is reviewed. Section 3 contains details of the data collection, questionnaire design, and model formulation. In Section 4, the results and discussion are presented. Finally, Section 5 comprises a summary of the major findings and conclusions.

Literature Review
SAVs are a relatively new concept in the field of surface transportation, which are not yet commercially available. In recent years, however, SAVs have become the subject of intense research. Numerous studies have been performed to analyze the potential effects of SAVs, as well as their operation and market penetration. Some studies have been conducted to determine people's transportation preferences in the hypothetical presence of SAVs, with these having been carried out in several different cultures and economies [4,[26][27][28][29][30][31][32][33][34][35][36]. A summary of the existing literature on public preferences toward AVs and SAVs is provided in Table 1.
Most of the studies summarized in Table 1 collect data via online surveys, of which some have used online SP surveys. Various methodological approaches, such as multinomial logit model (MNL), nested logit model (NL), logit kernel or mixed logit model, and other regression models are often used for data analysis [4,26,[29][30][31][32]. It is noted that the vast majority of these studies were conducted in developed countries and focused on autonomous vehicles, car-sharing service, and their combination with SAVs. Studies suggest that sociodemographic attributes of people (age, education, gender) and trip attributes (travel time, travel cost, waiting time) are influencing factors in the adoption of SAVs [4,26,29,32]. Similarly, safety, comfort, and SAVs' membership costs are the key predictors that influence the public travel mode preferences toward SAVs [4,30,33,34]. Likewise, other factors, such as having a driver's license and owning a private car, have also been found to influence users' preferences for AVs and SAVs [26,29]. e conclusions from previous studies are not necessarily applicable to developing countries, where urbanization and developmental levels, cultural norms, and transportation rationales may be vastly different. erefore, examining public attitudes toward SAVs in developing countries can assist in making the analysis more relevant in terms of the expected future technological landscape in this domain. In essence, an SP survey design that concentrates on a respondent's current travel selections and rationality may be able to increase our understanding of traveler preferences for SAVs. is current study intends to add valuable inputs from developing countries, which would subsequently help policymakers to make more rigorous decisions about SAVs worldwide.

Study Area.
In this research, we consider two cities: Lahore, Pakistan, and Dalian, China. Lahore is the capital of the Punjab Province, and the second megacity of Pakistan, with a population of 11.13 million and an area of 1,792 km 2 [37]. It comprises 12.7% of the total urban population of Pakistan. In 2017, it had a per capita GDP of approximately $5,855. Lahore is a hub for educational, medical, and industrial employment opportunities, and contains various other allied facilities. Together, these factors place enormous demands on the road network [22], and consequently, Lahore is at the forefront of public transportation and infrastructure development in Pakistan. e first bus rapid transit system (BRT) in Pakistan, the Lahore metro bus system, was launched in 2013. An average of 135,000 people travel on the BRT every day [21].
Dalian is located on the southern tip of the Liaodong Peninsula in northeast China, with the Bohai Sea to the west and the Yellow Sea to the east. Dalian is one of the most popular business areas in China, a tourist destination, and a distribution hub for goods and materials, with the Dalian container port linking more than 20 international oceanshipping routes [23,24]. Dalian is a heavily developed industrial area of China, with a population of 6.98 million and a total administrative area of 13,237 km 2 . In 2017, the GDP per capita in Dalian was approximately $14,816. It is a flexible and innovative city in terms of public transportation, with four subway lines, one BRT route, two trolley bus lines, and approximately 130 conventional bus lines. e summary statistics for Lahore and Dalian are provided in Table 2 [23]. e findings of previous studies have indicated that major metropolitan cities in developing countries face significant negative effects related to urbanization, such as increasing car ownership leading to traffic congestion, environmental problems, and land-use problems [21,23,[38][39][40]. Dalian and Lahore have different levels of development, different economies, and different cultures. However, they have similar transportation systems and public transportation policies. In both cities, private cars symbolize freedom and mobility, while the possibilities for expanding road infrastructure are limited. Both cities have invested heavily in promoting and optimizing public transportation systems. Moreover, neither city has car restrictions nor congestion charges. Finally, Lahore and Dalian are broadly representative of other cities in Pakistan and China, respectively, in terms of their travel patterns [21,23,40].

3.2.
Design of the SP Survey. SP surveys have been widely used in the field of transportation to analyze people's travel behavior, including their preferences for alternative forms of transportation that are not yet available [41]. In the SP survey, various hypothetical scenarios are presented to respondents, who then select their preferred choice(s) regarding each scenario from a finite set of attributes and alternatives [42][43][44]. SAVs are not yet available in the public transportation sector, and thus choice behavior and preferences toward SAVs cannot be observed. An SP survey was therefore conducted to elicit the respondents' willingness to use SAVs. e SP survey questionnaire in this study had three sections. In the first section, the respondents were familiarized with SAVs, via a general description and a pictorial representation in Figure 1, and were asked whether they possessed a private car and a driver's license. In addition, the respondents were asked questions regarding their technological interests, awareness of SAVs, and inclination to use existing dynamic riding services. In the second section, respondents were asked to select their preferred mode of transportation from a list of possible alternatives. In the third section, the respondents were asked questions relating to sociodemographic characteristics, namely, gender, age, educational attainment, employment, and monthly income.

Alternatives and Attributes of Choice Experiments.
Four alternatives (i.e., modes of transportation) and four attributes were presented in the SP survey questionnaires to respondents. For Lahore, the alternatives were personal cars, conventional taxis, minibuses, and SAVs. For Dalian, the alternatives were personal cars, conventional taxis, conventional buses, and SAVs. Our study focuses on SAVs with level 5 autonomy as defined by the Society of Automotive Engineers, which means that automated vehicles complete all driving tasks autonomously, without any human intervention [32]. For both cities, the four attributes were walking distance, waiting time, travel time, and travel cost. ese attributes were selected as they were considered the main factors that influence people's decisions on whether to travel by private car or by public transportation. To aid understanding of different scenarios, and elicit more realistic responses, pictorial versions of each alternative were presented to respondents. Each attribute was assigned three levels, and thus the total factorial was 3^1 3 . e levels of the attributes for each of the alternatives in Lahore and Dalian are provided in Tables 3 and 4, respectively.
To increase the realism of the choices, the levels of the attributes of the hypothetical alternatives were pivoted around the levels of the attributes of the revealed preferences [29]. For both cities, the levels of the attributes were based on traveling a distance of 10 km in the central business district.
is is particularly appropriate in Lahore, where the majority of the population lives within a 7-8 km radius of the city center [45]. e travel time for personal cars was based on the average speed of cars in both cities. Similarly, the travel times for conventional taxis, minibuses, and conventional buses were based on their average speeds. e travel cost of a personal car was calculated by considering the fuel, depreciation, maintenance, and miscellaneous (e.g., parking, carwash, tolls) costs.
is is somewhat different from previous studies, in which fuel cost and parking cost were considered separately, but the maintenance, depreciation, and miscellaneous costs were ignored. e travel costs of the conventional taxi, conventional bus, and minibus were calculated from existing fares. Finally, the travel time and travel cost for SAVs were taken from existing research.
A Taguchi fractional factorial experimental design was used to implement the choice experiments, using Minitab software. is design enables a subset of all possible scenarios to be created. In this case, a total of 27 scenarios were generated. An example of a scenario for Lahore is depicted in Figure 2, and another example of a scenario for Dalian is depicted in Figure 3. To achieve more realistic responses, the 27 scenarios were divided into nine blocks of three each. Each respondent thus only had to consider three scenarios.

Data Collection.
In this study, data were collected from respondents in Lahore, Pakistan, and Dalian, China, respectively, with Lahore respondents surveyed first. e questionnaire was devised and uploaded in English as a Google Form. Hyperlinks were generated for each of the nine experimental blocks. e online questionnaire was then disseminated via social media and emails to randomly selected residents in Lahore. is strategy enables rapid delivery, randomly selected respondents, easy switching between scenarios, and is inexpensive [46]. e questionnaires were distributed to the potential respondents in nine separate groups, who had a fortnight to complete the questionnaires from the date of receipt. e data were collected between April 10, 2019 and July 5, 2019, with a total of 551 valid responses being received. Each respondent considered three scenarios, and thus the total number of observations was 1,653. A summary of the sociodemographic characteristics of the respondents is provided in Table 5.
In the second phase, data were collected in Dalian. e questionnaire was devised, then translated into Chinese, and uploaded as a Wenjuanxing Form (https://www.wjx.cn). Nine hyperlinks and nine QR codes were generated for the nine experimental blocks. Fliers containing the formal greeting from the questionnaire, one of the QR codes, and an outline of the study objectives were designed. e questionnaire was then randomly distributed by sharing the fliers online via social media platforms WeChat and QQ, and printed fliers were also distributed manually within the central business district, with the help of colleagues and friends. e fliers were distributed in nine separate groups, and respondents were allowed two weeks to complete the questionnaires from the date of receipt. e data were collected between July 15, 2019 and October 30, 2019, and a total of 359 valid responses were received. Each respondent considered three scenarios, and the total number of observations was therefore 1,077. A summary of the sociodemographic characteristics of the respondents is provided in Table 5. e high percentage of students and younger respondents in both cities may be attributed to the fact that unemployed, less educated, and elderly people are usually less approachable through online surveys.
Nontrading behavior is one of the major drawbacks of online SP surveys. Nontrader implies that the respondent always selects the same alternative for each choice scenario of preference. is behavior of respondents may be attributed to several factors. First, a respondent may have an extreme preference for one alternative. Second, a respondent usually gets bored and does not pay full attention to the survey.
ird, a respondent makes strategic or political decisions. However, it is a difficult task to distinguish the three causes of nontrading behavior. Some researchers have reported that nontrader behaviors can also affect the marginal utility coefficients [47]. e current study has been conducted on traders and nontrader groups. However, for the reason of concision, we are only focused on the traders' group.
In the Lahore survey, 61 out of the 612 respondents (9.97%) represented nontrading behavior. 38% of nontrader respondents chose SAVs, 31% chose a private car, 15% chose a conventional taxi, and 16% chose the minibus. Most nontraders were highly educated with a bachelor's and a SAVs will pickup and dropoff you at home (door-to-door service) without walking to station and back from station to home.
SAVs gets rid user from cost of vehicle ownership, insurance, toll fee, fuel cost, depreciation, maintenance and parking cost. Only user have to pays for travel cost per minute or kilometer (Km).
SAVs are highly safe due to automated driving and advance technology.
SAVs are environmentlly friendly due to low emissions of CO2, and less chances of congestion due to advance sensing technology.  cars. In terms of employment, 32 out of 45 student respondents opted for private cars. 34 out of the 47 respondents whose income was less than 3,000 CNY/month always opted for private cars.
Based on the above descriptive results of nontraders, it is noted that the majority of the nontraders were employed, highly educated, and high-income people in both cities. In this study, the nontraders had an equal preference for SAVs and private cars in Lahore. However, the nontraders in Dalian had a strong preference for private cars. After excluding the nontrader respondents from the sample, the remaining 551 and 359 trader respondents from Lahore and Dalian were used for the discrete choice analysis.   the random utility maximization framework. It is assumed that individuals choose the alternative over the course of alternatives that has the maximum utility for them [10]. Various discrete choice models have been used to analyze travelers' preferences for AVs and SAVs, such as the multinomial logit model, the nested logit model, and the mixed logit model [4,26,29,32]. In this study, the multinomial logit (MNL) model and the mixed multinomial logit (MMNL) model with panel effect were used.

Multinomial Logit
Model. e MNL model is frequently used to interpret and calibrate mode-choice data [48]. In the MNL model, the utility is decomposed into two parts, namely, the systematic utility and the random error: where U in is the overall utility of the alternative i to the decision maker n, ε in is the unobserved error component of utility, and V in is the systematic utility function. e systematic utility function is further decomposed as where β alt ik are the parameters of alternative-specific attributes, β ind im are the parameters of individual-related variables, x is the explanatory variable, K is the total number of alternative-related parameters in the model, and M is the total number of individual-related parameters in the model. e probability of each individual n choosing alternative i from a set of alternatives j (j � 1, 2, . . ., J) can be estimated via

Mixed Multinomial Logit
Model. e MMNL model generalizes the MNL model by allowing the coefficients β to be random. In particular, it is now assumed that β is distributed according to β ∼ f(β), for some density f(β). In this case, conditional on β, the probability of each individual n choosing alternative i from a set of alternatives j (j � 1, 2, . . ., J) is given by where it is now made explicit that the systematic utility function V in depends on the parameter β, by writing V in � V in (β). As β is now random and not known, the (unconditional) choice probability is now the integral of the previous probability over the density of β [49,50], as given by e choice probabilities are thus considered as a mixture of the logit function evaluated over the values of parameters β, with f(β) the density of the parameters. is model does not constrain β to follow a normal distribution, unlike the probit model, and thus can be used with any other distribution (e.g., triangular, uniform, and lognormal). In this sense, it is considered a more powerful and flexible model in treating the variances and correlations of the random component.

Descriptive Statistics.
In the first section of the survey questionnaire, respondents were asked questions regarding their awareness of SAVs, technological interests, and inclination to use existing dynamic riding services. Regarding awareness of SAVs, 75% (81%) of respondents had heard of SAVs in Lahore (Dalian), as shown in Figure 4. e technological interest of the respondents in advanced technology was analyzed on a Likert scale (extremely interested, moderately interested, slightly interested, and not at all interested). Individuals' interest in advanced technology was high. Approximately 71% and 86% of respondents in Lahore and Dalian had extreme and moderate interest, respectively. Figure 5 shows that individuals interested in new technology were more willing to adopt SAVs for schooling and working trips than other trips in both cities. is suggests that the "automation" of vehicles would be more effective during peak hours of work and schooling. is finding is in line with the previous study that early adopters showed more interest in new technology. e diffusion of innovation theory by Roger Everett's classic model has been divided into five categories; innovators, early adopters, early majority, late majority, and laggards. According to the diffusion of innovation theory, early adopters embrace technology more rapidly [51].
e results of individuals interested in new technology and their preference for shopping and social trips are shown in Figure 6 and 7.
Finally, regarding the inclination of the respondents to use existing dynamic riding services, e.g., Careem or Uber in Pakistan and Didi in China, 82% (88%) had used these services in Lahore (Dalian). In addition, as shown in Figure 8, the individuals who use dynamic riding services were willing to adopt SAVs and private cars (conventional bus) for commuting to school and the workplace as compared to other trips in Lahore (Dalian). e results of individuals who used dynamic riding service and their preference for shopping and social trips are shown in Figure 9 and 10. Based on these descriptive results, respondents had a slightly higher preference for SAVs with sharing systems rather than automation in both cities, as shown in Figures 5 and 6, and Figures 8 and 9. Besides this, in Dalian, respondents had a slightly higher preference for SAVs in terms of automation for social trips as shown in Figures 7 and 10. A prior study conducted in Lahore on mobile app taxi services revealed that Careem and Uber were preferred by 78% and 75% of respondents, respectively. In addition, the respondents reported that mobile app taxis were more reliable and safer than conventional public transportation [52]. It is thus arguable that, when SAVs enter the market, the share of these dynamic riding services will significantly increase.

Generic Variables and Goodness-of-Fit of the Models'
Results. As outlined in Section 3.5, discrete choice models were used for data analysis and calibration. NLOGIT software was used for estimating the model parameters and for testing the goodness-of-fit. Analysis of the MNL model was provided first, as it was simpler and easier to interpret the model. All the alternative-specific and sociodemographic variables were considered for possible inclusion in this model, and were included in the final model if they were found to be significant at the 90% level (p < 0.10).
A summary of the results for Lahore is provided in  parameters is expected, as the utility and the choice probability decrease when the travel cost, travel time, and walking distance increase. A summary of the results for Dalian is provided in Table 7, where it can be seen that the waiting time, travel time, and travel cost all have negative coefficients. As outlined above, these observations are unsurprising: people tend to choose the mode of transportation that has the minimum travel time and travel cost. is result is also consistent with the results of previous studies [29].
e MNL model is based on the assumption of independent irrelevant alternatives (IIA) and homogeneous responses.
e MMNL model can relax these assumptions, potentially resulting in a better model. e MMNL model with panel effect was employed to capture the correlation of the repeated observations. is estimation is important for the analysis since SP survey respondents always provide multiple responses and are not independent from one another.
erefore, the panel effect should take into account the     Tables 6 and 7. e panel MMNL model was fitted using a lognormal distribution for the alternative-specific variables, after testing several different distributions for these parameters. In particular, specifying a lognormal distribution for the travel time yielded the best model output. e lognormal distribution is bounded by a positive value (0, 1). It is assumed that the opposite of travel time follows a lognormal distribution. Let the coefficient of the opposite of travel time be β, then the distribution of β mtt is: ß and σ are parameters that need to be estimated, and v mtt is a stochastic error, which follows a standard normal distribution. Since v mtt follows a standard normal distribution, if we do not consider the impact of v mtt , on average, the coefficient of travel time is β tt � − exp(β). Tables 6 and 7 show that the coefficients of travel time and travel cost are negative as expected. e standard deviations (sigma) of the travel time coefficient are statistically significant, verifying that there is heterogeneity in the population. e conventional approach to simulation estimation is to use random draws from the specified distribution. A good performance in simulation estimation usually requires large numbers of draws, which require a large amount of computation. Bhat has found that a Halton sequence with a far small number of replications is often as effective as a far larger number of random draws [54]. erefore, we used the Halton draws for the simulation of the mixed logit model. To achieve stable results for variables, 1,000 Halton draws were used. e lognormal distributions of travel time in Lahore and Dalian are presented in Figure 11. In Dalian, the standard deviations of travel time are more deviated from the mean than in Lahore.

Journal of Advanced Transportation 13
an overestimation of standard deviations. Despite its limitations, the lognormal distribution is preferred over normal distribution in the case of coefficients with a strong priori sign assumption, such as travel cost and travel time coefficients [55].

Sociodemographic Variables
Results. e model estimates for the sociodemographic variables are provided in Tables 6 and 7, and a descriptive interpretation of these estimates is summarized in Table 8. In terms of age, elderly people were more likely to use a conventional taxi than other modes of travel in Lahore. e results show counterintuitive behavior of the respondents, as the conventional taxi is an expensive mode of transportation in Lahore. is can be explained by the fact that elderly people rely on other family members and drivers in Pakistan. e conventional taxi drivers assist passengers, such as dropping passenger's luggage at the doorstep rather than other public  McFadden R 2 � 0.1222 Halton draws � 1,000 Note: * * * , * * , and * denotes statistical significance at 1%, 5%, and 10% level respectively.
transportation services. In Pakistan, the fare is often set by open negotiations between the passenger and the driver, and it is primarily dependent on the driver's mood, location, and time. In Dalian, elder people were also more likely to use the conventional taxi. Dalian's conventional taxi is slightly more expensive than other existing travel modes, and the fare is calculated on a taximeter that charges by both time and distance. e younger people, however, expressed more inclination to use private cars in Dalian.  Income + is indicates that people with higher incomes (i.e., a monthly income >50,000 PKR) were more likely to use SAVs. is observation is intuitive because the high-income public is more likely to explore new modes of transportation. Higher earners typically also care more strongly about the principal advantages of SAVs, such as the reduced risk of accidents and environmental benefits. is result is also in line with a previous study conducted in developed countries that higher-income people were more likely to use SAVs [60].
Car ownership -Insignificant for SAVs. + is indicates that people who had their private cars were more likely to use SAVs. is is likely due to the perception that SAVs could provide the same level of convenience and on-demand service but without the cost associated with car ownership. In addition, this positive behavior of respondents indicated that they would use SAVs and give up the existing private vehicles.
Driver license + is suggests that people with driver's licenses were more likely to use SAVs.
Frequency of travel in a week + is implies that people who traveled more than two days a week were more likely to use SAVs. is finding reaffirmed the preference of people who were used to conventional taxis for their trips. ey are thus likely to be attracted by SAVs, which represent an advanced form of taxis that will eliminate the walking distance and decrease the waiting of the users and provide door-to-door services.
From the perspective of gender, men were perceived as more likely to use private cars in Lahore. is observation is not surprising because family owned private cars were primarily used by men in Pakistan. On the other hand, it was found that women were more likely to use conventional taxis. is is in stark contrast with a previous study's finding that women did not hire taxis due to cultural norms and privacy constraints [56]. It can be partly explained by the fact that women in Pakistan are marginalized due to privacy constraints and have restricted transit options. Due to limited public transportation options, women usually prefer to take traditional taxis and ride-sharing services [57]. e SAVs will overcome these issues and provide an alternative form of private transportation. It has been expected that women will prefer SAVs when they are practically viable. On the other hand, no such gender disparity in transit options was found in Dalian.
In Dalian, employment was also found to have a significant positive relationship with the endogenous variable, suggesting that employed people and entrepreneurs were more likely to use a conventional taxi. is is reasonable because taxis allow people to perform other tasks while traveling, making time spent on traveling more desirable and more valuable. In contrast, in Lahore, employment was found to be insignificantly associated with any of the transportation alternatives. e association between education and the endogenous variable was positive and significant in Lahore, implying that people having bachelor's degrees and above were more likely to choose SAVs. Conversely, the education variable was found to be insignificantly associated with any of the transportation alternatives in Dalian. ere was a significant positive association between income and the endogenous variable in Lahore, but not in Dalian. In Dalian, private car ownership was found to have a significant association with a preference for SAVs. is positive behavior of respondents indicates that they would use SAVs and give up the existing private vehicles. is finding contradicts a previous US study which found that 18.6% people were likely and 7.3% were extremely likely to relinquish an existing vehicle [30].
In Lahore, people with driver's licenses were significantly more likely to use SAVs, private cars, and conventional taxis. In contrast, in Dalian, possession of a driver's license was found to be insignificantly associated with any of the transportation alternatives. In addition, people who traveled more than two days a week were more likely to use SAVs in Lahore. e people in Lahore are reluctant to take public transportation due to disintegrating land-use mismanagement and a cumbersome public transportation system. For these reasons, private cars and motorcycles are the dominant modes of transportation in Lahore. e result supports the preference of people who would use SAVs to travel in the future to get rid of recurrent public transportation issues in Lahore. On the other hand, in Dalian, the frequency of travel in a week was perceived to be insignificantly associated with any of the transportation alternatives. e value of travel time saving (VTTS) is a core concept of transportation demand modeling and appraisal. e VTTS depends on several parameters and varies from country to country, from industry to industry, and even from person to person, based on socioeconomic factors, the purpose of travel, the length of travel, etc. [58]. e VTTS is calculated as the ratio between the coefficient of travel time and the coefficient of travel cost parameters. In Lahore, the coefficient value of travel cost is β tc � −0.00725, and the travel time coefficient is β tt � −0.0429, resulting in an average VTTS of 5.917 PKR/min or 355 PKR/hr ($2.28/hr) across all modes. Likewise, in Dalian, the travel cost coefficient is β tc � −0.02304, and the travel time coefficient is β tt � −0.0156, resulting in an average VTTS of 0.677 CNY/min or 40.62 CNY/hr ($6.36/hr) across all modes. For comparison, the VTTS in the United Kingdom is 11.7 pounds/hr ($16.51/hr) across all modes with an average distance of 12.87 km [59].

Conclusions
is study aimed to examine the public attitude toward the adoption of SAVs in a cross-national context, and an SP survey was performed to capture the travel mode preferences of respondents in the cities of Lahore (Pakistan) and Dalian (China). e SP surveys contained detailed information on socio-demographic attributes of travelers and service attributes of the four travel mode choices considered, i.e., SAVs, private car, taxi, and bus. e MNL model and the MMNL model were used for model estimation. Model estimation results showed that the MMNL model was more robust than the MNL model in predicting the travel mode choice. e findings of the study revealed that generic attributes such as travel time, waiting time, and travel cost have a significant influence on transportation preferences in both cities. In addition, the results showed that highly educated people with high incomes were more likely to use SAVs in Lahore. e study also suggested that people in the city of Lahore who traveled more than two days a week and had driver's licenses were more likely to use SAVs. Based on our inference, private vehicles and motorcycles in the city could be replaced by bike-sharing, e-bikes, and car-sharing to achieve a sustainable transportation system. Surprisingly, in Dalian, only one sociodemographic variable, i.e., car ownership, was found to be significant for the preference of SAVs. is implies that people who had their private cars were more likely to use SAVs.
is positive behavior of respondents indicated that people in Dalian were more likely to give up their private vehicles when SAVs practically come into the market. It is an interesting result that in developed countries such as the USA, the majority of people still prefer private cars to travel [30]. e descriptive results also showed that respondents had slightly higher preferences for SAVs for different trip purposes due to sharing of SAVs rather than automation in both cities. is result demonstrates that the public is accustomed to shared mobility options and is wary of an autonomous vehicle. e findings of this study will provide a useful benchmark for policymakers tasked with devising effective policies for the future implementation and deployment of SAVs in cities worldwide. e study has a few limitations that must be acknowledged. First, the percentage of students and younger respondents was high, which might cause bias in analyzing the behavior of respondents. erefore, revealed preference surveys and stratified sampling should be considered in future research. Second, this study was first conducted in Pakistan and focused only on SAVs, private cars, conventional taxis, and minibuses in Lahore, ignoring other modes such as motorcycle, walking, and BRT to reduce the cognitive burden and complexity of the survey. Forthcoming studies can emphasize all modes of transportation to exhaustively capture the preferences of respondents.
Data Availability e data are available on request from the corresponding author.

Conflicts of Interest
e authors declare no conflict of interest.