How Much of Which Mode? Using Revealed Preference Data to Design Mobility As a Service Plans

Abstract

Mobility as a service (MaaS) seeks to integrate emerging shared mobility modes with existing public transportation (PT). Decisive to its uptake will be attractive subscription plans that cater for heterogeneous mobility needs. Research on willingness to pay for such plans has commenced, yet remains divided on a central question: how much to include of which mode, and how? Complementing previous research building on stated preference data, in this study revealed preference data is used to analyze the viability of different subscription plan components (PT, car-sharing, bike-sharing, taxi), modes of inclusion (budgets in minutes and season tickets) and subscription cycles (weekly, monthly). PT season tickets are found to be viable for 83% of all respondents. Interestingly, the viability of minute budgets of car- and bike-sharing depends on subscription cycle length. Using a monthly subscription cycle, car-/bike-sharing appears viable to include in a bundle for 35%/31% of all respondents, respectively. Using a weekly subscription cycle, these figures drop to 1.4%/0.4%, respectively, as weekly variation in demand is much higher than monthly variation. In contrast to many current MaaS pilots, taxi use remains too infrequent to include as recurring credit in MaaS plans. Rather, pay-as-you-go is the economically more sensible option for consumers. This research therefore challenges the idea of all-inclusive mobility flat rates and suggests a more modular design.

Shared and electricity-powered transportation modes (e.g., [e-]scooters, [e-]bikes, ride-hailing) have started to appear in many cities worldwide. Their integration with existing public transportation (PT), often referred to as Mobility as a Service (MaaS), could facilitate their joint use and thereby present a new alternative to the private car. A corresponding shift to shared modes could help reduce congestion and greenhouse gas emissions at times when urgently needed (1 –3).

At its core, MaaS consists of a platform integrating payment and routing across multiple transportation providers, allowing customers to plan and pay for intermodal trips using a single app (“one-stop shop”) and subscribe to recurring, multimodal transportation plans. Current MaaS pilots (e.g., UbiGo, WHIM, zengo) exhibit a wide variety of plans, typically including a sole “pay-as-you-go” option, an “all-inclusive” mobility flat rate and a package in between consisting of (discounted) budgets for specific modes (e.g., car-sharing, bike-sharing, taxi). While it is too early to assess the success of the different pilots, it is clear that plans need to cater for the heterogeneous mobility needs of potential customers.

Research on MaaS plans is still in its infancy (4,5). It has begun with stated preference experiments on willingness to pay (WTP) and adoption rates of different components (4,6 –8), but remains divided on a central question: how much to include of which mode, and how? While Matyas and Kamargianni (9) found that respondents in London generally do not prefer shared modes (car-sharing, bike-sharing, taxi) in their plans, Guidon et al. (8) found a higher WTP for PT and car-sharing and a lower WTP for (e-)bike-sharing and taxi in a bundle when compared with stand-alone WTP in Zürich. This suggests that customers prefer certain shared modes (car-sharing) in their plans and that preferences for modes in general may differ from preferences for modes in bundles. Ho et al. (6) also found a general preference for car-sharing in bundles, noting, however, that preferences for shared modes vary significantly across the respondents, with discounts for taxi and ride-hailing only appealing to regular taxi/Uber users.

Despite evidence that travel behavior is influenced by habit (10) and that MaaS customers may consequently seek to purchase plans that best cater for their current mobility needs (6,7,9), revealed preference data has to date not been used for research on the design of MaaS plans. Yet, it offers insight into the amount and variability of transportation demand on the individual level and its substitutability with shared modes over time.

This paper reports a study that used the mobility traces of 555 students over one semester (13 weeks) in Copenhagen, Denmark, to construct a MaaS scenario where private car trips are substituted by shared modes based on generalized costs. The viability of different subscription plan components (PT, car-sharing, bike-sharing, taxi), modes of inclusion (budgets in minutes and season tickets) and subscription cycle durations (weekly, monthly) for each student are analyzed. The results offer evidence that a PT season ticket represents a viable core of MaaS plans for the majority. In contrast to current MaaS pilots, car-, bike-sharing and taxi use remain too infrequent in this sample to include as weekly recurring credit. When considering monthly subscription cycles, the picture changes for car-sharing and bike-sharing. The results thus suggest a more modular approach to designing MaaS bundles than all-encompassing mobility flat rates.

This paper is structured as follows. We first review the literature on MaaS. We then introduce our data and methods before reporting on the results. We conclude with a summary and a discussion of the implications for transportation research, business and policy.

Literature Review

The recent emergence and plentiful appearance of new shared transportation modes (e.g., shared e-scooters, ride-hailing, [e-]bike-sharing) has motivated research on their use and resulting changes in travel behavior. While suppliers often frame their impact in sustainable terms (i.e., more active travel, less carbon footprint, less private automobile usage, less congestion) and emphasize complementarity with PT as first/last mile solutions, these claims have recently been questioned and partially refuted. Academic peer-reviewed empirical studies of ride-hailing (for a recent review, see Tirachini [11]) tend to show increases in congestion, vehicle kilometers travelled and resulting carbon footprint (12 –16) and lean toward a prevalence of substitution effects of PT (15,17,18). In relation to changes in travel behavior, most ride-hailing trips seem to substitute the taxi or PT and a small share seem to be additional travel that would not have happened if the option was not available (induced demand) (14,16,17,19). Bike-sharing studies (for reviews, see Fishman et al. [20] and Ricci [21]) reveal that bike-sharing is predominantly used instead of walking and PT (20,22 –26) while in general the frequency of using a bicycle seems to increase as a result of bike-sharing system prevalence (20,21). For shared e-scooters, the emerging (yet scarce) evidence suggests that many trips are short and substitute active modes (walking, cycling) while the link between PT and shared e-scooters used as first/last mile access modes is generally weak (27 –29).

It is in this context of increasing evidence that emerging transportation modes might substitute PT more than complement it that the potential of MaaS becomes clearest. At its core, MaaS aims to integrate emerging transportation modes with PT to facilitate seamless (intermodal) planning, booking and payment, thereby addressing barriers to their joint use as a (more sustainable) alternative to the private car. MaaS has been conceptualized using levels of integration (30 –32). These typically range from no to full integration across operational, informational and transactional dimensions; for an overview, see Reck et al. (33). Bundling mobility services into multimodal transportation plans that users can subscribe to is typically seen as the step preceding full integration of the previous dimensions, although this sequence is not necessarily followed in practice (WHIM in Helsinki, Finland, and swa Augsburg in Augsburg, Germany, are examples).

Matyas and Kamargianni (4,9) were among the first to highlight and analyze the research gap on multimodal transportation plans. They conducted a stated choice survey in Greater London, England, asking respondents to choose from different MaaS plans. They estimated mixed multinomial logit models on plan choices and found that coefficients for shared modes besides PT (car-sharing, bike-sharing, taxi) are negative, implying that respondents do not prefer any of these shared modes in their MaaS plans. Interestingly, ∼38% of all respondents would still consider buying a MaaS plan, however, the authors noted that this could be because of the hypothetical nature of the experiment. In their successive work, Matyas and Kamargianni (34) found a strong correlation between currently used modes and stated preferences and hypothesized the reason to be habit persistence.

Guidon et al. (8) conducted a stated choice survey in Zürich, Switzerland, to analyze the valuation of components in MaaS bundles versus stand-alone. They find a higher WTP for PT and car-sharing and a lower WTP for (e-)bike-sharing and taxi in a bundle when compared with stand-alone, suggesting that customers prefer certain shared modes (car-sharing) in their plans. This contrast to the findings of Matyas and Kamargianni (9,34) may be because of regional or methodological differences in eliciting and analyzing customer preferences in a hypothetical experiment.

Ho et al. (6) conducted a stated choice survey in Sydney, Australia, to analyze the potential uptake of MaaS plans and the WTP for its components. While also finding a general preference for car-sharing in bundles, they note that preferences for shared modes vary significantly across respondents. This finding is reinforced by a subsequent study of the authors using the same experimental design in Tyneside, UK (7). This could explain the differences between Guidon et al. (8) and Matyas and Kamargianni (9) on car-sharing preferences. Ho et al. (6,7) also found that uptake levels correlate with current usage of mobility tools (e.g., discounts for taxi and ride-hailing only appealed to regular taxi/Uber users). This corresponds with Matyas and Kamargianni’s (34) finding of a strong correlation between currently used modes and stated preferences and is rational to expect given the effect of habit on travel choices (10).

Despite the shared finding that MaaS customers may seek to purchase plans that best cater for their current mobility needs, which corresponds with earlier evidence on habitual travel behavior, revealed preference data has to date not been used for research on the design of MaaS plans. Yet, it offers insight into the amount and variability of multimodal transportation demand on the individual level over time and its substitutability with shared modes over time.

Data

The mobility trajectories of participants were used in a 24-month longitudinal experiment, the Copenhagen Networks Study (CNS); for a more detailed description of the study, see Stopczynski et al. (35). The CNS was conducted between September 2013 and September 2015. All prospective undergraduate students at the Technical University of Denmark were invited to participate in the study by pamphlets sent with official university acceptance letters during spring/summer 2013. Free smartphones (LG Nexus 4), to be distributed in the first weeks on campus, were promised to students participating in the 24-month study. This way, ∼500 students were recruited up to September 2013. As the participation rate reached a plateau, further undergraduate students from former years were also invited to participate, of which an additional ∼300 signed up. In total, more than 800 students participated in the study.

Registration before the distribution of the smartphones included a questionnaire, written consent to data collection, a symbolic cash deposit for each phone and finalization of the pre-installed data collection app (“SensibleDTU”). Data was subsequently collected via the smartphones in various ways including questionnaires, online social networks and smartphone sensor data (GPS, Bluetooth and WiFi signals). The sensor data was collected passively and uploaded to the server automatically every 2 h.

Naturally, passively collected location data is not perfectly precise. The estimation error for locations in this study was below 50 m in 95% of all cases (36). Subsequently, locations were separated into stop-locations (dwell time above 5 min) and travel stages within the Danish regions of Hovedstaden and Sjælland (Figure 1). For each stage, the mode of travel was inferred with an overall accuracy of 90% (37).

The total number of 634,269 identified stages for 814 students was filtered to contain only semester times (4 × 13-week periods), as it was assumed that students’ mobility routines would be more stable during these periods, thus more suitable for subscription-based mobility plans. Widely available semester PT cards support this choice. From the remaining dataset with 365,911 stages for 801 students, 29 students with a time span of less than 4 weeks between the first observation and the last were excluded to ensure sufficiently long mobility trajectories. The remaining 772 students participated with varying reliability in the experiment. While the number of smartphones with GPS and WiFi traces varied over time with its peak in the second semester, some traces show gaps of several days or weeks. In order to ensure a sufficiently dense and large dataset, the analyses were thus focused on semester two (of four) and further excluded students with less than one stage per day averaged over all days. The final dataset included 131,338 stages for 555 students.

Study Context

The main campus of the Technical University of Denmark (DTU) is located in Kongens-Lyngby, 12 km north of Copenhagen (Figure 1). It is connected to the city by several bus lines and has a substantial number of parking places. Students can live on campus or off campus in Kogens-Lyngby and Copenhagen center.

Copenhagen is the capital of Denmark and its most populous city. At the time of the study, the city of Copenhagen had ∼649,000 inhabitants and Copenhagen metropolitan region ∼1.92 million (38). The city is known for its extensive bicycle infrastructure and high share of active modes. The trip-based modal split during the time of study was 22% PT, 38% active modes (bicycle, walking) and 40% private car (38). Around 25% of the inhabitants of the metropolitan region lived less than 500 m away from the nearest PT station and ∼55% less than 1 km (38).

Several shared modes additional to PT were offered or introduced to the city during the study period. Bike-sharing provider Bycyklen has a long history of operating in Copenhagen. Car-sharing providers car2go and DriveNow began operating in September 2014 and August 2015, respectively. Uber did not yet operate in Copenhagen during the study period.

Methodology

Complementing the dataset (131,338 stages across all modes), we constructed a MaaS scenario by substituting car stages with the best alternative shared mode (PT, car-sharing, bike-sharing, taxi) in relation to generalized travel costs. This allows a subsequent analysis of the viability of different MaaS plan components.

Consistent values of travel time and cost components for all alternative modes and stages were obtained by mapping the data to the 2015 Danish National Transport Model (DNTM) zones. The DNTM includes submodels for population synthesis, household synthesis, car ownership and assignment models. For a detailed introduction to each submodel, see Rich and Hansen (39). Demand estimation is based on the Danish National Travel Survey, which has been conducted for almost 20 years. The DNTM calculates (and partly simulates) demand for the entire Danish population (5.4 million individuals, 2.4 million households) and iterates demand with supply until convergence is obtained. Assignment takes into account congestion and trips as a function of time (pseudo-dynamic assignment as a day is separated into 10 time periods). The choice process is based on the random utility framework and includes long-term mobility tool ownership decisions as well as short-term travel decisions (trip frequency, destination choice, mode choice). The zone system has a total of 3,670 zones and each zone has an average number of 612 addresses in Sjælland. Five modes are implemented in the model: walk, bike, car (driver), car (passenger) and PT.

For each of the 131,338 stages (a stage is denoted s in the following), travel time and costs were calculated according to the following formulas, that have recently been applied by Rich and Vandet (40) to similar data using the DNTM:

tim e_{P T} (s) = ivt (s) + γ_{1} * noc (s) + γ_{2} * wwt (s)

(1)

e x p_{P T} (s) = fare (s)

(2)

Travel time for PT is composed of in-vehicle time ivt(s), the number of changes noc(s) multiplied by a transfer penalty γ₁ of 6 min, and additional waiting and walking times wwt(s) multiplied by a penalty factor γ₂ of 1.5; see Rich (41) for detailed documentation of the DNTM and associated γ-factors. Travel costs are based on 2015 single ticket fares (discounts for season tickets are applied later).

Travel time for car-sharing and taxi stages are assumed to be similar to car stages and are composed of free-flow times and additional delays caused by congestion multiplied by a penalty factor of 1.35 (40). In sum, time spent finding a parking space, and access and egress time is assumed to be similar for the private car and car-sharing as many dedicated car-sharing parking spots in Copenhagen can be assumed to shorten the time needed to find a parking space, while access time can sometimes be longer. For taxis, there is usually no egress or access time nor is time required to find a parking spot. On the other hand, there is additional waiting time. As waiting time for the taxi is often shorter than the total time needed to find a parking space, access and egress time, 5 min of the total travel time was deducted for taxi stages. Travel costs are based on 2015 prices of car-sharing provider DriveNow and taxi company Taxa DK (non-shared taxi prices). For bike-sharing, travel times are included in the DNTM, and travel costs are based on the 2015 prices of bike-sharing provider Bycyklen. Car-sharing and bike-sharing are free-floating and dockless, respectively, and only available in the center of Copenhagen. For the service area, an approximation of the current DriveNow zones was used.

Consistent with the DNTM and previous Danish studies on the value of travel time savings (VTTS) (39,40), generalized travel costs for each alternative mode m and stage s are then computed using an income-specific VTTS:

gencos t_{m} (s) = tim e_{m} (s) + \frac{e x p_{m} (s)}{VTTS}

(3)

As the income of the student population can be assumed to be fairly homogeneous, a VTTS value of 59 DKK/h (∼9 USD/h) was used that was computed for a sample consisting mainly of DTU students (42). Variations across trip purposes and modes were found to be small in previous Danish VTTS studies (43) and thus not further considered.

Finally, weather data for each trip was added and the availability of the bike-sharing alternative constrained to trips with a travel time of less than 30 min and times when local precipitation was zero.

Results

Alternative Shared Modes

Table 1 shows the observed modal split and the modal split in the MaaS scenario, where all stages previously conducted with a car have been substituted with the best alternative available in terms of generalized costs. Overall, PT sees by far the greatest overall gain (+16.41 pp.), while car-sharing (+3.29 pp.) and bike-sharing (+2.64 pp.) account for the remaining stages. Within the car-/bike-sharing zones, which cover a large part of the inner city of Copenhagen, car-sharing sees the greatest gain (+6.95 pp.), followed closely by bike-sharing (+5.57 pp.). These results are plausible as car/bike-sharing presents a viable alternative to PT on shorter inner-city distances with a median of 3.3 km (∼2 mi). Interestingly, taxi trips, although included in existing MaaS packages, are always among the least preferable alternatives in terms of generalized costs (because of high fares), thus no car stages are substituted with the taxi.

Table 1.

Alternative Modes Substituting Car Stages in MaaS Scenario

Mode	All zones		Car-/bike-sharing zones (∼Copenhagen city)
Mode	Observed	MaaS scenario	Observed	MaaS scenario
PT	34.94%	+16.41 pp.	34.04%	+0.94 pp.
Car	22.34%	na	13.46%	na
Car-sharing	na	+3.29 pp.	na	+6.95 pp.
Taxi	na	na	na	na
Bike-sharing	na	+2.64 pp.	na	+5.57 pp.
Walk/bike	42.72%	na	52.50%	na

Note: pp. = percentage points; na = not applicable.

Viability of MaaS Plan Components

In the following, we test the viability of different MaaS plan components (PT season tickets, weekly and monthly budgets for car-/bike-sharing in minutes) for each student. It was assumed that it is financially sensible to purchase a PT season ticket if a student spends more than the weekly/monthly cost of a season ticket on PT during the entire timespan (13 weeks/3 months) on average. It was further assumed that a budget of X min for car-/bike-sharing is financially sensible to purchase if a student uses at least X min of the respective mode every week/month. This follows the rationale that car-/bike-sharing is usually charged by the minute, thus a student would be better off not to purchase a certain budget if it were underused.

Figure 2 illustrates the approach. It shows the weekly demand for bike-sharing, car-sharing and PT for one student participating in this study. It can be seen from Figure 2 that this student used bike-sharing irregularly, with a weekly demand varying between 0 min and ∼100 min. The student also used car-sharing irregularly (between 0 min and ∼30 min), however a more stable usage is visible, with an average ∼13 min per week. The student used PT for most trips, at an average weekly cost of ∼500 DKK (∼75 USD). Following the assumptions above, it would not seem sensible for this student to purchase weekly budgets for car-/bike-sharing, as the minimum weekly demand is zero (solid blue line). A PT season ticket, however, would seem sensible to purchase, as the average weekly cost of PT (∼500 DKK/solid blue line) surpasses its cost per week at the discounted student rate (∼145 DKK/dashed black line) for Hovedstaden, which includes both DTU and Copenhagen municipality.

This analysis was conducted for all 555 students, finding that for 459 students (82.7%) a PT season ticket (region Hovedstaden) would seem sensible to purchase based on a weekly subscription cycle. Just eight students (1.4%) showed a weekly car-sharing demand greater than zero, suggesting a low viability of a weekly recurring car-sharing budget. For bike-sharing, this number is even lower (two students, 0.4%). This is because of the high variability of weekly demand for both modes: on average, students show demand for car-/bike-sharing only in 28%/25% of all weeks, respectively. As intuition suggests, variability (in relation to zero demand periods for car-/bike-sharing) decreases when the length of the subscription cycle is changed from weekly to monthly (Figure 2 versus Figure 3), increasing the viability of multimodal bundles. This effect is most pronounced for car-sharing, where 196 students (+188 when compared with a weekly bundle) show a monthly demand greater than zero. The effect for bike-sharing is also strong with 170 students (+168) showing a monthly demand greater than zero. PT season ticket viability remains similarly high (470 students, +11). Absolute demand among these students for car-/bike-sharing is low, though. On average, the monthly demand for bike-sharing was 4.1 h (SD: 3.7 h) and 2.6 h (SD: 7.59 h) for car-sharing.

Discussion

In the introductory section of this paper it was stated that the uptake of MaaS will depend on how well multimodal plans cater for the heterogeneous mobility needs of potential customers. Research on MaaS plans has begun with stated preference experiments, but remains divided on a central question: how much to include of which mode, and how? While the partially contradictory findings of the four main contributions so far (4,6 –8) may stem from regional and methodological differences, all authors agree that there is a strong correlation between current travel patterns and preferences for MaaS plans, suggesting that potential customers will seek to reproduce their current mobility tool usage in a future MaaS plan. Despite this agreement, revealed preference data has to date not been used for research into the design of MaaS plans. Yet, it offers insight into the amount and variability of transportation demand on the individual level and its substitutability with shared modes over time.

In this study, the mobility traces of 555 students over one semester (13 weeks) were used to construct a MaaS scenario where shared modes replaced car trips based on generalized costs. The viability of different MaaS plan components (PT, car-sharing, bike-sharing, taxi), different modes of inclusion (budgets in minutes and season tickets) and subscription cycles (weekly, monthly) were analyzed. For weekly subscription cycles, it was found that a PT season ticket forms a viable core of MaaS plans for the majority of the sample (83%). In contrast to many current MaaS pilots, car-/bike-sharing and taxi use remain too infrequent to include as recurring credit for most students. Rather, pay-as-you-go appears to be the economically more sensible option for those modes. For monthly subscription cycles, the picture changes. Now, car- and bike-sharing appear viable to include in bundles for 35.3% and 30.6% of the students, respectively, with mean absolute levels of demand of 4.1 h for bike-sharing and 2.6 h for car-sharing. This finding suggests that the duration of the subscription cycle is an important aspect in MaaS bundle design.

Demand is not always motivated economically, however. There is evidence of the “flat rate bias” (44,45) in PT season ticket purchasing decisions (46,47) and first indications show similar patterns for all-inclusive mobility flat rates (e.g., SBB Green Class, WHIM Unlimited), too (48). Yet all-inclusive mobility flat rates seem to appeal only to a very specific socioeconomic group (48). The results of this study suggest a more modular design to appeal to larger parts of society.

These findings are consistent with Matyas and Kamargianni (9) in that, for the majority of the sample, recurring credit for car-/bike-sharing and taxi would not be economically sensible to purchase as part of a MaaS plan. They are also consistent with Ho et al. (6,7) in that preferences for MaaS plans are likely to vary substantially across potential customers as their individual mobility needs diverge.

Payment in a MaaS context can be ex post (i.e., pay-as-you-go) or ex ante (i.e., subscription). This study identified demand variability as a challenge when designing multimodal subscriptions (at least for this sample of students). A question that arises is whether subscriptions need to be tied to specific modes and budgets (e.g., a PT season ticket, X min of car-sharing) or whether they can be conceptualized as flexible “mobility wallets.” Similar to pre-paid phone cards, pre-paid credit could be used for travel until depleted/expired and users could choose between a manual or a recurring top-up option. One question that arises is how high should recurring top-ups be? This question recalls the literature on variability/stability in monetary travel budgets (49 –52). How stable are monetary travel budgets? Mokhtarian and Chen (50) conducted an extensive review of more than 25 years of evidence on the topic and concluded that stability in travel budgets (time and monetary) has only been shown on the most aggregate level, while the initial claim of stability cannot be supported at the disaggregate (i.e., person-) level. Here, travel budgets vary with socioeconomic characteristics (e.g., income, car ownership, location of residence), transportation system supply and urban structure (50 –52). The amount of the recurring top-up for a mobility wallet could be flexible, though, thus accounting for inter-person differences. The question in the present case is rather: Can we assume intra-person monetary travel budgets to be constant over time? The literature on habitual travel behavior (10) and the existence of PT season tickets supports this idea. The question is whether stability can also be assumed for shared modes (i.e., car-sharing, bike-sharing, taxi). The results of this study suggest otherwise (at least for this sample of students). This can be explained with the type of trips these shared modes are used for. Previous studies suggest that a large share of trips conducted with these shared modes are of recreational and social nature and that only a small share of trips serve the regular commute (53). This clearly is an early hypothesis and further research is needed to investigate the stability of intra-person travel budgets (money and time) for different modes. Returning to the idea of mobility wallets, it can thus be stated that a recurring top-up mechanism would require (a decent degree of) stability in monetary travel budgets, which is questionable (at least for users with higher modal shares of car-sharing, bike-sharing, taxi and ride-hailing).

It remains an open question whether MaaS as a concept and bundles as one of its layers have the potential to influence and change travel behavior. This has been argued for theoretically (3,31), as MaaS essentially integrates new mobility modes with public transport, decreasing the cognitive user effort to plan, book and pay for multimodal trips on operational, informational and transactional dimensions. However, existing empirical evidence of behavioral change is scarce (54) and indicative at best. After all, travel behavior is influenced by many more aspects than the level of integration of different modes (e.g., income, household structure, mobility tool ownership, attitudes) and PT systems need to have the capacity to satisfy potential modal shifts. The authors are involved in several large-scale empirical studies to evaluate the impact of MaaS bundles on travel behavior in Switzerland, Germany (Augsburg) and Australia (Sydney) and will report on first results in the following months.

Finally, while presenting a new approach to designing MaaS plans, this study has several limitations that call for future work. First and foremost, as the sample is composed of students, the results are not necessarily generalizable to other parts of the population. Students show high variation of daily schedules (even in term times) and are limited both in their current mode choices and their economic conditions. We thus suggest conducting similar analyses with other longitudinal datasets, ideally with revealed preference data from actual MaaS trials to circumvent hypothetical scenarios. Second, generalized costs were calculated with a single VTTS for all students. Estimating individualized VTTS would clearly be preferable, however, it was not possible because of data limitations in this case. Future studies should consider this as a methodological suggestion. Third, additional relevant factors influencing mode choice (such as attitudes, income, mobility tool ownership, the need to travel with others or carry heavy objects) could be included as additional boundary conditions when assessing the availability of alternative modes and their generalized costs to make findings more robust. Fourth, this study focuses on existing modes of transportation as the DNTM does not yet account for future modes such as autonomous taxis. Future studies could include yet-to-be-established modes by building on transportation demand models that already incorporate them (such as MATSim, for example). This would provide a more holistic perspective of the viability of future multimodal mobility bundles.

Footnotes

Acknowledgments

The authors thank Sune Lehmann and Otto Anker Nielsen for providing access to the CNS dataset and the Danish National Transport Model data, respectively. Thanks also to the many helpful researchers at DTU and KU for their comments and suggestions (Andreas Bjerre-Nielsen, Kelton Ray Minor, Andrea Vanesa Papu Carrone and Daniele Romanini) and MOVIA (Sonia Eriksen, Carsten Jensen and Peter Andreas Rosbak Juhl) for making available their zone and pricing data. Four anonymous reviewers have enabled us to improve this paper substantially with their valuable feedback.

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: D. J. Reck and K. W. Axhausen; data collection: D. J. Reck; analysis and interpretation of results: D. J. Reck; draft manuscript preparation: D. J. Reck; manuscript review and editing: K. W. Axhausen. Both authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

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

Funding

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

References

Kamargianni

, W. Li, M. Matyas, and A. Schäfer. A Critical Review of New Mobility Services for Urban Transport. Transportation Research Procedia , Vol. 14, 2016, pp. 3294–3303.

Jittrapirom

, V. Caiati, A. M. Feneri, S. Ebrahimigharehbaghi, M. J. Alonso González, and J. Narayan. Mobility As a Service: A Critical Review of Definitions, Assessments of Schemes, and Key Challenges. Urban Planning , Vol. 2, No. 2, 2017, pp. 13–25.

Mulley

Mobility As a Services (MaaS) – Does it Have Critical Mass? Transport Reviews , Vol. 37, No. 3, 2017, pp. 247–251.

Matyas

, and M. Kamargianni.

Survey Design for Exploring Demand for Mobility As a Service Plans.

Transportation , Vol. 46, 2019, pp. 1525–1558.

Esztergár-Kiss

, and T. Kerényi. Creation of Mobility Packages Based on the MaaS Concept. Travel Behaviour and Society , 2019. https://doi.org/10.1016/j.tbs.2019.05.007

C. Q.

, D. A. Hensher, C. Mulley, and Y. Z. Wong. Potential Uptake and Willingness-to-Pay for Mobility As a Service (MaaS): A Stated Choice Study. Transportation Research Part A: Policy and Practice , Vol. 117, 2018, pp. 302–318.

C. Q.

, C. Mulley, and D. A. Hensher. Public Preferences for Mobility As a Service: Insights from Stated Preference Surveys. Transportation Research Part A: Policy and Practice , Vol. 131, 2020, pp. 70–90.

Guidon

, M. Wicki, T. Bernauer, and K. W. Axhausen. Transportation Service Bundling – for Whose Benefit? Consumer Valuation of Pure Bundling in the Passenger Transportation Market. Transportation Research Part A: Policy and Practice , Vol. 131, 2020, pp. 91–106.

Matyas

, and M. Kamargianni. The Potential of Mobility As a Service Bundles As a Mobility Management Tool. Transportation , Vol. 46, 2019, pp. 1951–1968.

10.

Gärling

, and K. W. Axhausen.

Introduction: Habitual Travel Choice.

Transportation , Vol. 30, No. 1, 2003, pp. 1–11.

11.

Tirachini

Ride‐Hailing, Travel Behaviour and Sustainable Mobility: An International Review.

Transportation , 2019, pp. 1–37. https://doi.org/10.1007/s11116-019-10070-2.

12.

Erhardt

G. D.

, S. Roy, D. Cooper, B. Sana, M. Chen, and J. Castiglione. Do Transportation Network Companies Decrease or Increase Congestion? Science Advances , Vol. 5, No. 5, 2019, p. eaau2670.

13.

Lewis

E. O. C.

, and D. MacKenzie. UberHOP in Seattle: Who, Why, and How?. Transportation Research Record: Journal of the Transportation Research Board , 2017. 2650: 101–111.

14.

Rayle

, D. Dai, N. Chan, R. Cervero, and S. Shaheen. Just a Better Taxi? A Survey-Based Comparison of Taxis, Transit, and Ridesourcing Services in San Francisco. Transport Policy , Vol. 45, 2016, pp. 168–178.

15.

Tirachini

, and M. del Río. Ride-Hailing in Santiago De Chile: Users’ Characterisation and Effects on Travel Behaviour. Transport Policy , Vol. 82, 2019, pp. 46–57.

16.

Tirachini

, and A. Gomez-Lobo. Does Ride-Hailing Increase or Decrease Vehicle Kilometers Traveled (VKT)? A Simulation Approach for Santiago De Chile. International Journal of Sustainable Transportation , Vol. 14, No. 3, 2020, pp. 187–204.

17.

Alemi

, G. Circella, S. Handy, and P. Mokhtarian. What Influences Travelers to Use Uber? Exploring the Factors Affecting the Adoption of On-Demand Ride Services in California. Travel Behaviour and Society , Vol. 13, 2018, pp. 88–104.

18.

Graehler

, R. A. Mucci, and G. D. Erhardt. Understanding the Recent Transit Ridership Decline in Major US Cities: Service Cuts or Emerging Modes? Presented at 98th Annual Meeting of the Transportation Research Board, Washington, D.C., 2019.

19.

Tang

B. J.

, X. Y. Li, B. Yu, and Y. M. Wei. How App-Based Ride-Hailing Services Influence Travel Behavior: An Empirical Study from China. International Journal of Sustainable Transportation , 2019, pp.1–15. https://doi.org/10.1080/15568318.2019.1584932

20.

Fishman

, S. Washington, and N. Haworth. Bike Share: A Synthesis of the Literature. Transport Reviews , Vol. 33, No. 2: 148–165.

21.

Ricci

Bike Sharing: A Review of Evidence on Impacts and Processes of Implementation and Operation. Research in Transportation Business & Management , Vol. 15, 2015, pp. 28–38.

22.

Bachand-Marleau

, B. H. Y. Lee, and A. M. El-Geneidy. Better Understanding of Factors Influencing Likelihood of using Shared Bicycle Systems and Frequency of Use. Transportation Research Record: Journal of the Transportation Research Board , 2012. 2314: 66–71.

23.

Campbell

K. B.

, and C. Brakewood Sharing Riders: How Bikesharing Impacts Bus Ridership in New York City. Transportation Research Part A: Policy and Practice , Vol. 100, 2017, pp. 264–282.

24.

Fishman

, S. Washington, and N. Haworth. Bike Share’s Impact on Car Use: Evidence from the United States, Great Britain, and Australia. Transportation Research Part D: Transport and Environment , Vol. 31, 2014, pp. 13–20.

25.

Murphy

, and J. Usher. The Role of Bicycle-Sharing in the City: Analysis of the Irish Experience. International Journal of Sustainable Transportation , Vol. 9, No. 2, 2015, pp. 116–125.

26.

Shaheen

, H. Zhang, E. Martin, and S. Guzman. Hangzhou Public Bicycle: Understanding Early Adoption and Behavioral Response to Bikesharing. Transportation Research Record: Journal of the Transportation Research Board, 2011. 2247: 33–41.

27.

Caspi

, M. J. Smart, and R. B. Noland. Spatial Associations in Dockless Shared E-Scooter Usage. Presented at 99th Annual Meeting of the Transportation Research Board, Washington, D.C., 2020.

28.

Reck

D. J.

, S. Guidon, and K. W. Axhausen. Modelling Shared E-Scooters in Louisville, Kentucky: A Spatial Regression Approach. Presented at 99th Annual Meeting of the Transportation Research Board, Washington, D.C., 2020.

29.

Zuniga-Garcia

, and R. Machemehl. Dockless Electric Scooters and Transit Use in an Urban/University Environment. Presented at 99th Annual Meeting of the Transportation Research Board, Washington, D.C., 2020.

30.

Hensher

D. A.

, C. Mulley, C. Q. Ho, J. Nelson, G. Smith, and Y. Wong. Understanding MaaS: Past, Present and Future. Proc., 2nd International Conference on Mobility as a Service, Tampere, 2019.

31.

Lyons

, P. Hammond, and K. Mackay. The Importance of User Perspective in the Evolution of MaaS. Transportation Research Part A: Policy and Practice , Vol. 121, 2019, pp. 22–36.

32.

Sochor

, H. Arby, I. M. Karlsson, and S. Sarasini. A Topological Approach to Mobility As a Service: A Proposed Tool for Understanding Requirements and Effects, and for Aiding the Integration of Societal Goals. Research in Transportation Business & Management , Vol. 27, 2018, pp. 3–14.

33.

Reck

D. J.

, D. A. Hensher, and C. Q. Ho. MaaS Bundle Design. ITLS Working Paper, No. ITLS-WP-20-06, 2020. Submitted to Transportation Research Part A: Policy and Practice. Available at: https://ses.library.usyd.edu.au/bitstream/handle/2123/21987/ITLS-WP-20-06.pdf?sequence=1&isAllowed=y

34.

Matyas

, and M. Kamargianni. Using Mixed Methods to Examine User Preferences for Mobility As a Service Subscription Plans. Presented at 98th Annual Meeting of the Transportation Research Board, Washington, D.C., 2019.

35.

Stopczynski

, V. Sekara, P. Sapiezynski, A. Cuttone, M. M. Madsen, F. E. Larsen, and S. Lehmann.

Measuring Large-Scale Social Networks with High Resolution.

PLoS One , Vol. 9, No. 4, 2014, pp. e95978.

36.

Alessandretti

, S. Lehmann, and A. Baronchelli. Understanding the Interplay between Social and Spatial Behavior. EPJ Data Science , Vol. 7, No. 1, 2018, pp. 36.

37.

Bjerre-Nielsen

, K. Minor, P. Sapiezynski, S. Lehmann, and D. D. Lassen. Wi-Finding: Urban Transportation Sensing using Crowdsourced Wi-Fi. Pervasive and Mobile Computing , Forthcoming.

38.

Floater

, P. Rode, and D. Zenghelis. Copenhagen: Green Economy Leader Report . London School of Economics and Political Science, London, 2014.

39.

Rich

, and C. O. Hansen. The Danish National Passenger Model – Model Specification and Results. European Journal of Transport and Infrastructure Research , Vol. 16, No. 4, 2016, pp. 573–599.

40.

Rich

, and C. A. Vandet. Is the Value of Travel Time Savings Increasing? Analysis throughout a Financial Crisis. Transportation Research Part A: Policy and Practice , Vol. 124, 2019, pp. 145–168.

41.

Rich

The Weeday Demand Model in LTM – Model for Generation, Destination and Mode Choice . Doc. no: 35243-004. Technical University of Denmark, Lyngby, 2017.

42.

Papu Carrone

, V. M. Hoening, A. F. Jensen, S. E. Mabit, and J. Rich. Car Sharing Preferences and Mode Substitution: A Stated Choice Experiment. Proc., World Conference on Transport Research, Mumbai, 2019.

43.

Fosgerau

, K. Hjorth, and L. J. S. Vincent. The Danish Value of Time Study. Danish Transport Research Institute, Technical University of Denmark, Lyngby, 2007. http://www.transport.dtu.dk/-/media/Institutter/Transport/forskning/publikationer/publikationer-dtf/2007/the_danish_value_of_time_study_250208.ashx. Accessed July 29, 2019.

44.

Train

K. E.

, D. L. McFadden, and M. Ben-Akiva. The Demand for Local Telephone Service: A Fully Discrete Model of Residential Calling Patterns and Service Choices. Rand Journal of Economics , Vol. 18, No. 1, 1987, pp. 109–123.

45.

Lambrecht

, and B. Skiera. Paying Too Much and Being Happy about It: Existence, Causes, and Consequences of Tariff-Choice Biases. Journal of Marketing Research , Vol. 43, No. 2, 2006, pp. 212–223.

46.

Axhausen

K. W.

, H. Köll, and M. Bader. Public Transport Usage Intensity of Season Ticket Holders in the City of Innsbruck. Report to the Innsbrucker Verkehrsbetriebe GmbH, Innsbruck and Ampass, 1998.

47.

Wirtz

, P. Vortisch, and B. Chlond. Flat Rate Bias in Public Transportation – Magnitude and Reasoning. Presented at 94th Annual Meeting of the Transportation Research Board, Washington, D.C., 2015.

48.

Martin

, H. Becker, D. Bucher, D. Jonietz, M. Raubal, and K. W. Axhausen. Begleitstudie SBB Green Class – Abschlussbericht. Working Paper No. 1439. Institute for Transport Planning and Systems, ETH Zürich, Zürich, 2019.

49.

Gunn

H. F.

Travel Budgets—A Review of Evidence and Modelling Implications. Transportation Research Part A: General , Vol. 15, No. 1, 1981, pp. 7–23.

50.

Mokhtarian

P. L.

, and C. Chen. TTB or not TTB, That Is the Question: A Review and Analysis of the Empirical Literature on Travel Time (and Money) Budgets. Transportation Research Part A: Policy and Practice , Vol. 38, No. 9–10, 2004, pp. 643–675.

51.

Tanner

J. C.

Expenditure of Time and Money on Travel. Transportation Research Part A: General , Vol. 15, No. 1, 1981, pp. 25–38.

52.

Zahavi

, and A. Talvitie. Regularities in Travel Time and Money Expenditures. Transportation Research Record: Journal of the Transportation Research Board , 1980. 750: 13–19.

53.

Young

, and S. Farber. The Who, Why, and When of Uber and Other Ride-Hailing Trips: An Examination of a Large Sample Household Travel Survey. Transportation Research Part A: Policy and Practice , Vol. 119, 2019, pp. 383–392.

54.

Strömberg

, I. M. Karlsson, and J. Sochor. Inviting Travelers to the Smorgasbord of Sustainable Urban Transport: Evidence from a MaaS Field Trial. Transportation , Vol. 45, No. 6, 2018, pp. 1655–1670.