Railway infrastructure breakthrough and mode choice: evidence from the Ceneri Base Tunnel in Switzerland

Abstract

Major investments in railway infrastructure and service provision are often justified by their potential to increase public transport (PT) use and induce modal shift away from private motorized transport (PMT), thereby contributing to more sustainable mobility. Yet, empirical evidence on the extent to which large-scale railway interventions affect individual mode choice remains scattered. This study evaluates the impact of the Ceneri Base Tunnel (CBT), opened in Ticino, Switzerland, in 2020, together with associated PT offer improvements, on work and study commuting mode choice. Drawing on the INDMOB travel survey, that collected data from 24,577 individuals, we analyze changes in commuting practices between 2019 and 2023 and estimate the causal effect of the CBT opening using a Difference-in-Differences (DiD) approach. We find that the CBT (and associated PT improvements) increased the probability of using PT rather than PMT as the main commuting mode by 4.2 percentage points (p.p.) among commuters along its north–south corridors, relative to the control group, after controlling for socio-demographics, mobility equipment, and PT level of service. This corresponds to a 13% relative increase compared with the baseline PT share in the treatment group. We further show that PT use is associated with individual and spatial characteristics: women and younger commuters, individuals with a PT subscription, and those living in areas with higher PT levels of service are more likely to use PT, conditional on the other covariates. Our findings contribute to the transport literature on the effects of major railway infrastructure improvements on mode choice and support rail investments as an effective component of modal shift strategies, while highlighting the importance of individual mobility resources and the broader PT service context in shaping their potential.

1 Introduction

1.1 Background and motivations

The transport sector alone is responsible of around 30% of total greenhouse gas (GHG) emissions in Europe (Transport and Environment, 2024) and nearly 40% in Switzerland (Luh et al., 2024). This substantial environmental footprint is largely driven by the persistent reliance on carbon-intensive transport modes, particularly private motorized transport (PMT), for daily mobility. Consequently, promoting a modal shift from PMT towards more sustainable alternatives, such as public transport (PT), has become a central policy strategy for transport decarbonization. PT can substantially reduce emissions per passenger-kilometer compared with private transportation, while also contributing to lower car dependency and reduced vehicle kilometers traveled (Litman, 2023); particularly in urban and metropolitan areas, where a large share of trips remains potentially transferable (Velázquez-Cortés et al., 2026). In these contexts, improving PT infrastructure and service provision is considered a key policy lever to enable sustainable modal shifts and achieve climate objectives. To the extent that they generate actual behavioral changes and induce a measurable shift in travel behavior (Anagnostopoulou et al., 2020), large-scale investments in railway systems are often justified by their potential to attract users away from carbon-intensive modes including PMT, thereby reducing transport-related GHG emissions (Baker et al., 2010; Zhang et al., 2021).

Large-scale investments in railway systems can involve major tunnel projects aimed at overcoming geographical constraints, particularly in mountainous regions. The opening of the Ceneri Base Tunnel (CBT) in 2020 completed a broader process of PT infrastructure improvement in southern Switzerland initiated with the opening of the Gotthard Base Tunnel (GBT) in 2016. By reducing travel times along key corridors in Ticino, the CBT substantially improved regional connectivity and was indeed followed by increased PT use and reduced car use among its residents, as shown by a recently published report (Repubblica e Cantone Ticino, 2025). However, descriptive evidence alone is insufficient to causally assess the impact of this infrastructure intervention on commuting behavior. In the broader context of understanding and fostering modal shift towards sustainable transportation, this paper asks whether the CBT influenced residents’ mode choice between PMT and PT. More specifically, we examine whether improvements in rail travel times, service frequency, and network integration associated with the CBT and related policy measures contributed to changes in commuting practices, by shaping them in favor of PT—particularly among commuters whose work/study trips were most directly affected by the new rail connection.

1.2 Literature review

We draw, first, on that body of scholarly literature evaluating rail transit supply shocks¹ through impact evaluation methods, including Difference-in-Differences (DiD). Studies indeed show that this kind of railway infrastructure improvements usually reconfigures mode share in favor of rail use by reducing car use of treated groups (Dai et al., 2020, 2024), although with modest effects on mode substitution from PMT to PT (Albalate and León-Gómez, 2024; Borsati and Albalate, 2020; Givoni and Dobruszkes, 2013). In fact, effects of major PT supply shocks are mostly leading to ‘induced demand²’ (Gumy et al., 2026; Sohoni et al., 2017) and are strongly context-dependent (Uddin and Rana, 2023; Wang et al., 2023, 2025). That is, rail transit supply shocks usually generate heterogeneous and polarized impacts across population groups (Wallimann et al., 2026); leaving behind, among others, more car-dependent, suburban, less educated, and older groups (Gumy et al., 2026; Rames et al., 2021). Similar conclusions are found when considering other supply measures such as fare subsidies (Conte et al., under review; Daniel et al., 2025) or the implementation of mobility applications (Goetz et al., 2024). Nonetheless, external shocks can also produce the opposite, undesired effect; inducing modal shift from PT to PMT—following, for instance, new mobility services entering the urban transport system (Ward et al., 2021) and COVID-19 social distancing policies (Das et al., 2021)—or from PT to active modes (Kraus and Koch, 2021).

We draw, second, on the literature examining the determinants of mode choice typically mediating the effect of transport supply shocks such as railway infrastructure improvements. In fact, mode choice is not determined solely by the availability of alternative modes, but by the interaction between conditions of different type that enable or constrain the adoption of more sustainable mobility practices. According to mobility scholars, these typically include individual (e.g., socio-demographic, socio-psychological) and contextual, structural (e.g., spatial, journey characteristic) factors (De Witte et al., 2013; Van Acker et al., 2010). Interacting with each other, these factors shape the probability of choosing a given mode and—importantly for our case—the extent to which infrastructure improvements translate to behavior change.

Concerning individual factors, we know that younger, more educated, female, less car equipped, and PT-satisfied households tend to use more PT (Chng et al., 2016; Ingvardson and Nielsen, 2019; Kawgan-Kagan, 2020; Roos et al., 2020). On top of socio-demographic differences, mode choice action logics³ define the mobility rationales and travel preferences of individuals (Kaufmann, 2002), influencing how transport improvements affect social perceptions of available modes—and, eventually, travel choices (Gumy et al., 2026; Kaufmann, 2026). These are, in fact, mainly shaped by commuters’ ability to move—in terms of access, competences, and appropriation (Grandvillemin et al., forthcoming; Guitton et al., 2025; Kaufmann et al., 2004)—and the integration of commuting practices into user needs, travel patterns, and everyday mobility routines (De Vos et al., 2016; Lyons et al., 2007; Meinherz and Binder, 2020).

Regarding contextual and structural factors, scholars highlight the role of, among others, travel time and pricing (Nordenholz et al., 2017), and temporal (i.e., service frequency) and spatial (i.e., capacity, accessibility) availability of transport services (Göransson and Andersson, 2023; Redman et al., 2013) in determining PT use; besides less tangible elements such as comfort, perceived (in)security, and the symbolic value of mode use (Abbasi et al., 2023; Bernier et al., 2022; Kaufmann, 2026). The impact of transport supply shocks on travel behavior is especially constrained by socio-spatial barriers, varying by spatial proximity, urban density, infrastructure quality, and regional transport competitiveness (Dai et al., 2020, 2024; Daniel et al., 2025; Goetz et al., 2024). Effective modal shift therefore requires the integration of mobility practices—for all travel purposes (Grisé et al., 2019; Pot et al., 2023; Yu and Higgins, 2024)—into the existing time-space constraints (Cass and Faulconbridge, 2016; Scheiner, 2010; Stanton et al., 2013), especially after transport infrastructure shocks (Gumy et al., 2026).

1.3 Research gap and objectives

Our first research objective is to estimate the causal effect of the CBT on commuting practices, building notably upon recent work conducted in similar geographic contexts (Wallimann et al., 2026). Although we previously contributed to a descriptive assessment of the impact of the CBT opening (Repubblica e Cantone Ticino, 2025), causal evidence on its effects on mode choice remains lacking. While recent working papers have addressed related questions (Guggisberg et al., 2026), they rely upon substantially different theoretical and methodological approaches and, most importantly, do not directly assess the impact of the CBT on the mobility practices of Ticino residents. We address this gap by providing the first (to our knowledge) quantitative impact evaluation of the opening of the CBT on mobility practices. Although our study does not directly assess modal shift, our research design allows to identify changes in mode choice minimizing concerns related to induced demand (Gumy et al., 2026; Hortelano et al., 2016; Wallimann et al., 2026; Wardman and Murphy, 2015). By incorporating the main determinants of transportation mode choice identified in the literature (De Witte et al., 2013; Van Acker et al., 2010) as covariates (i.e., control variables), our analytical framework additionally helps connect the literature on rail transit supply shocks with that one on modal shift.

Our second objective is to derive policy implications and key recommendations from our findings that may foster the transition towards more sustainable transportation. Transport policy being increasingly driven by climate mitigation goals and seeking to promote modal shifts from PMT to more sustainable modes (Intergovernmental Panel on Climate Change (IPCC), 2023; International Transport Forum, 2023), updated and robust empirical evidence on the behavioral impact of PT infrastructure investments is particularly needed. Beyond informing the evaluation of large-scale transportation investments, we hope our findings contribute to a broader understanding not only of the conditions under which railway infrastructure can support the transition towards sustainable mobility, but also of the dynamics of how broader PT measures ultimately encourage modal shift.

The paper is structured as follows. The next section presents the data and the methods adopted; with a detailed explanation of our impact evaluation approach and the models we used for our analysis. Then we present our results, namely the quantitative estimates of the causal effect of the opening of the CBT on Ticino work/study commuters’ mode choice. Subsequently, we outline the key limitations of our current methodological approach, provide suggestions for further research in the field, discuss our results vis-à-vis the relevant scholarly literature, and finally offer targeted recommendations for policy makers in the domain of sustainable transportation.

2 Materials and methods

2.1 Study context

The study area is the Canton of Ticino, Switzerland’s southernmost region. Bordering Italy to the south, Ticino is a geographically diverse area, with pre-Alpine mountainous regions, relatively small urban centers (notably, Lugano, Bellinzona, Locarno) and their suburban areas in the main valleys, and ample rural areas in between. Compared with the rest of Switzerland, Ticino is characterized by slightly lower income levels, greater gender disparities in labor-market participation, and an older population (Federal Statistical Office, 2025b). Moreover, although it is integrated into the same Federal transport system—notably regarding rail transport—, Ticino has comparatively lower levels of (regional) PT service quality. Figure 1 illustrates the PT level of service in Ticino based on PT frequency as well as the type and the accessibility of stops: railway station, bus or tram stop (Federal Office for Spatial Development, 2026). At the regional level, urban areas benefit from the best PT service, particularly Lugano, where the spatial coverage of categories A and B is the highest (Figure 1).

Figure 1

In the last decade, this region’s PT infrastructure was radically improved by two major interventions, both of which formed part of a broader project to modernize the Swiss railway network across the Alps (Federal Department of the Environment, Transport, Energy and Communications, 2025).⁴ First, the GBT, that went in operation in 2016, is the longest railway tunnel in the world and thus represented a gamechanger along the north-south axis between Zurich, Switzerland and Milan, Italy; reducing up to 25% the travel time between the two cities and leading to a modal shift of around 5% from car users (Wallimann et al., 2026). Second, the CBT, opened in 2020, together with the implementation of the ‘TP21’ offer (2021 PT system overhaul⁵), significantly improved the regional PT service and enhanced rail network connectivity between northern (Locarno, Bellinzona, Tre Valli) and southern (Mendrisio, Malcantone, Vedeggio, Lugano) Ticino (Figure 1). It indeed halved travel times along the north-south axis within the region, along the corridors Lugano-Locarno (30 min) and Lugano-Bellinzona (15 min)⁶ (Repubblica e Cantone Ticino, 2025).

Those PT improvements contributed to the development of the so-called "Città Ticino⁷", by upgrading the main north-south railway corridor in Ticino, linking Chiasso in the south to Airolo in the north and serving key urban areas such as Mendrisio, Lugano, Bellinzona, and Biasca all along the way (Ferrata, 2022). Then, it restructured regional spatial organization by reducing commuting costs, strengthening Lugano’s role as an employment center, and enabling residential decentralization around Bellinzona (Guggisberg et al., 2026). Finally, it reconfigured commuting practices and mode choice. Despite the stronger reliance on PMT in Ticino compared with the national average (Federal Statistical Office, 2024), the introduction of the CBT was followed by a +3.2% increase in PT use (+15.6% along the Bellinzona-Lugano corridor) and a −5% decline (−10% along the corridor) in car use (Repubblica e Cantone Ticino, 2025). In light of the three arguments discussed above and despite the CBT arguably having wider impacts on both passenger and freight transport at the national and continental scales, this study will focus on north-south or south-north daily travel within Ticino, where the CBT is likely to mostly involved, to test the effect of the CBT introduction on work/study commuters’ mode choice. This, in turn, prompts us to describe the data used for this purpose.

2.2 Data and variables

The research institute LINK administered the main INDMOB survey⁸ from September to December 2023 via an online questionnaire to 50,000 Ticino residents aged 19 or older⁹, randomly selected from the Canton's population register, the so-called Movpop. Besides questions on current (fall 2023) travel behavior, the survey includes a retrospective module on fall 2019 (recalled) mobility practices; since no pre-CBT travel data existed, this information was needed to evaluate the impact of CBT. Precisely, the INDMOB questionnaire covered seven domains to evaluate mobility practices changes post-CBT: (i) socio-demographics part 1 (e.g., residence location in 2019 and in 2023, work/study location in 2019 and 2023); (ii) access to transport (e.g., car availability and PT subscriptions in 2019 and 2023); (iii) professional status and mobility practices in 2023 (e.g., parking options, mode of transport options to commute); and (iv) 2019” (see Data Sheet); (v) mobility practices changes between 2019 and 2023 (e.g., reasons for shifting from car to car alternatives, or vice versa); (vi) knowledge of CBT (e.g., CBT operationalization since 2021); (vii) socio-demographics part 2 (e.g., household composition).

Of the 50,000 Ticino residents randomly selected, a total of 24,577 completed, at least in part, the main questionnaire¹⁰. The sample (n = 24,577 respondents) is regionally representative, with a concentration of middle-aged and older cohorts (over 35 years), predominantly households of two or more people, and approximately half of the full sample is employed¹¹ (Repubblica e Cantone Ticino, 2025). For the remainder of this article, PT refers to trains and buses while PMT refers to cars and motorized two-wheelers.

To assess the CBT’s impact on mode choice, we conducted a four steps methodological approach, as illustrated by Figure 2.

Figure 2

First, we restricted the sample to commuters; specifically, those respondents who provided both home and work or study locations (i.e., trip origin and destination). More precisely, we excluded from the sample the following categories of individuals: (i) non-working and non-studying respondents (e.g., unemployed, pensioners); (ii) those with missing commuting data; and (iii) active modes’ users (i.e., walking, cycling), as our focus is on shifts from PMT to PT. After data cleaning (i.e., missing values, quality control), the final sample comprises 12,683 individuals (see Research design and Figure 2).

Second, based on the theoretical framework of De Witte et al. (2013), we selected three groups of mode choice determinants: (i) socio-demographic (i.e., age, gender, education level), (ii) mobility equipment (i.e., driver license possession, PT subscription, PT digital application usage), and (iii) spatial/journey characteristics, as illustrated in Figure 1 (i.e., the PT level of service, or ARE index,¹² from the Swiss Federal Office for Spatial Development). We do not include respondents’ subjective perceptions of PT level of service—in line with the socio-psychological determinants theorized by De Witte et al. (2013)—as a variable since one-third of respondents did not provide complete answers to that question (Figure 2). Income was also excluded due to the high share of respondents selecting ‘do not know/prefer not to answer’ (20% of the sample). Accounting for those three determinants allows us to consider potential differences in individual, spatial, and mobility-related characteristics between groups, thus improving the accuracy of the estimated effect of the CBT opening on commuting mode choice.

Third, to select the most relevant variables from each of the three dimensions (i.e., socio-demographic, mobility equipment, and spatial/journey characteristics), we statistically tested the association between their corresponding variables and the commuting mode choice variable (using Cramér’s V) as well as the significance of the results (using a Chi-squared test). Table 1 establishes the explanatory power of each variable with respect to mode choice (Cramér’s V), the statistical significance of the association through the Chi-Squared test (p-value < 0.001), and the absence of any significant multicollinearity (Variance Inflation Factors (VIFs) of all variables comprised between 1 and 1.5). Fourth, we estimate the Average Treatment effect on the Treated (ATT) with Difference-in-Differences models (with and without covariates).

Table 2 enables comparison between the full sample of all Ticino commuters and the subsample of employed people commuting by PMT or PT. The demographic differences between the full survey sample and our DiD subsample reflect societal trends, for instance women’s labor market participation rate, or the fact that our oldest age class, above 60 years old, mostly consists of retired people (i.e., not-working/studying commuters). Education levels are comparable between both samples. In terms of mobility equipment, the commuters of the subsample tends to be more equipped (in terms of driver’s license possession, PT subscription, and PT application use), which is to be expected as employed people are generally more mobile than the average population.

Fourth, we run a DiD model to assess the impact of the CBT opening on work/study commuters’ mode choice between 2019 and 2023. We provide further details about our final methodological step, and research design, in the following sub-section.

2.3 Research design

Difference-in-Differences (DiD) is widely regarded as the “gold standard” of econometric methods for establishing causality (Angrist and Pischke, 2009), following seminal economists’ work which underlies the award of the 2021 Nobel Prize in Economic Sciences (Card and Krueger, 1994). The first use of this approach is historically attributed to John Snow who, in 1854, was able to prove the causal relationship between contaminated water and Cholera in London (Lechner, 2011). Since then, thousands of studies employing this method have been published in various fields, ranging from public health (Currie and Gruber, 1996) to transport policy (Gibbons et al., 2019).¹³

We adopt a DiD approach in order to estimate the causal effect of the CBT on commuters’ mode choice. Treating the tunnel opening as a quasi-natural experiment, our design and data enable a before-after evaluation of the reconfiguration of regional rail services and its effects on travel behavior of our subsample. By comparing pre- (2019) and post-intervention (2023) observations, we thus identify changes in mode choice patterns, with particular attention to PT versus PMT. Our model’s dependent variable is indeed an indicator of the main mode used in the two periods, built as a dummy variable (PT = 1, PMT = 0). For the purpose of this study, we do not consider active mobility, as we assume that the tunnel due to the geographical distances involved and the descriptive results available (Repubblica e Cantone Ticino, 2025) has no effect on the use of these modes.

Drawing upon the survey items on residence and work/study locations (see Supplementary material), we isolate two population groups. The first, serving as our treatment group, includes those commuters potentially affected by the CBT, since their home-to-work or home-to-school commutes lie along its axes: namely, those whose residence place is located in the north (Bellinzona, Locarno, Trevalli) and work/study location is in the south (Lugano, Mendrisio, Vedeggio, Malcantone) of the region—or vice versa. The second, serving as our control group, includes the rest of the sample: namely, those living and working/studying in one of the two regions (north or south), and thus not necessarily commuting along the axes of the CBT. In the final sample, our treatment group consists in 1,734 individuals (13.7% of the sample), while our control group represents 10,949 individuals (86.3% of the sample). We further provide descriptive evidence of the observable characteristics of both groups in Appendix (Table 6).

This identification strategy relies upon distinguishing between the direct effects of the CBT opening (i.e., drastically shorter travel times along its north-south axis) and the broader changes affecting the whole regional PT system (e.g., more frequent trains, new early morning and late-night trains). The CBT opening was accompanied by the introduction of a Canton-wide PT service upgrade—the "TP21" offer, including the TILO 2021 timetable—which reorganized regional services in the context of the new infrastructure. While these broader PT improvements may have affected mobility conditions in Ticino more generally, our treatment definition focuses by design on the main mechanism through which the CBT is expected to affect travel behavior: the improvement in interregional accessibility generated by the tunnel itself. Individuals commuting across the Ceneri axis are directly exposed to the reduction in travel times induced by the CBT (see Study context) as well as to improved rail connections between northern and southern Ticino. Conversely, individuals who live and work/study within the same zone are not directly exposed to this rail infrastructure shock, although they may still benefit from broader improvements in the regional PT system. Therefore, the control group should be interpreted as being potentially exposed to general PT improvements, but not to the main treatment channel associated with the CBT. Comparing individuals directly exposed to the CBT with those who are not, while controlling for observed covariates, indeed allows us to estimate the differential change in mode choice associated with the tunnel opening.

After recoding the involved variables into categorial dummies, we obtain the following model (Equation 1):

In this model, θ denotes the DiD estimator and captures the Average Treatment Effect on the Treated (ATT), namely the causal effect of the intervention on exposed units. The β coefficients respectively capture the main effects of the treatment group (treated commuters vs. control residents), time period (2023 vs. 2019) dimensions, both specified as dummy variables, as well as the covariates X_it. Here, coefficients are interpreted as percentage-point (p.p.) changes, with θ thus representing the differential change in mode share between the treatment and control groups over time; that is, the effect of the CBT opening on the shift between PMT and PT between 2019 and 2023. The dependent variable we use is the main mode of transport respondents declared to use to travel to their work/study locations.¹⁴ In other words, we do not account for intermodality; instead, we focus on the main commuting mode.

3 Results

3.1 Models and assumptions

To assess the impact of the CBT opening on commuting mode choice, we evaluate two alternative DiD model specifications (Table 3). The first, referred to as the ‘base model’ (A1), estimates the treatment effect without including any additional covariates, thereby capturing the unconditional average effect of the CBT opening on mode choice. The second, referred to as the ‘full model’ (A2), estimates the DiD effect while controlling for the set of covariates identified as statistically significant through the preliminary analysis—namely, Cramér’s V and Chi-squared association tests (see Data and variables).

These control variables account for relevant differences in socio-demographic, spatial, and mobility-related characteristics between individuals exposed (i.e., treatment group) and not exposed (i.e., control group) to the intervention. We further remove variables that did not add any explanatory power to our full model, through a process of stepwise backwards elimination. For instance, we removed education level, driver’s license possession, and PT application use from the model. Accordingly, while the base model provides the overall estimated effect of the CBT opening, the full model estimates the treatment effect conditional upon observable characteristics that may also explain differences in mode choice between individuals.

Our research design—especially the base model—relies upon the plausibility of the Parallel Trends Assumption (PTA) underlying DiD models. This implies that, in the absence of the treatment (i.e., the CBT opening), the evolution of outcomes would have been similar between the treatment and control groups. In our setting, this means that mode choice among north-south and south-north commuters (i.e., the treatment group) would have followed the same underlying trend as among residents who live and work/study within the same zone (i.e., the control group). Although we do not dispose of multiple pre-treatment observations allowing for an empirical assessment of pre-treatment trends (i.e., so-called “placebo test”), we evaluate the plausibility of the PTA through institutional and contextual arguments. First, both our treatment and control groups are located within the same regional context and are exposed to common changes affecting mobility practices in Ticino, including transport policies, economic conditions, and the COVID-19 shock. Second, the only key difference between the treatment and control groups is their potential exposure to the improvement in rail travel times for their commute journey, while both groups remain exposed to broader trends affecting PT demand. Third, aggregate mobility statistics show a long-term evolution of travel behavior in Ticino towards greater PT use, from 11% of distance traveled in 2005 to 22% in 2015, with no evidence of structural breaks prior to the CBT opening¹⁵ (Federal Office for Spatial Development, 2017). Moreover, available official statistics also suggest possibly similar future trends that are justified by the national context and European mobility patterns towards PT use. While it is methodologically impossible to demonstrate the validity of the PTA, these global trends highly support its plausibility (Roth, 2022; Roth et al., 2023).

The fully adjusted model additionally relies upon the Conditional Parallel Trends Assumption (CPTA), according to which treatment and control groups would have followed similar outcome trends in the absence of treatment (i.e., CBT opening) after conditioning on observed covariates. By controlling for baseline individual characteristics (i.e., age and gender), mobility equipment (i.e., PT subscription), and PT level of service, the full model reduces potential compositional differences between the two groups that may influence mode choice. As with the unconditional PTA, the conditional version cannot be directly tested and exclusively relies on the plausibility that the included covariates capture relevant observable sources of differential trends.

In addition, the Stable Unit Treatment Value Assumption (SUTVA) requires that the potential outcome (i.e., mode choice) for any individual does not vary with the treatment of other individuals of the study. In other words, we assume that mode choice of a survey respondent is unrelated to other respondents’ commuting situation (intra-zone or along the north-south corridors). Since respondents were sampled at random from different households, it is reasonable to assume that their mode choices are independent from each other.

Furthermore, the identification strategy relies on the no-anticipation assumption, which requires that individuals do not systematically adjust their behavior in response to the expected treatment before its actual implementation. Although this assumption is more difficult to verify empirically in a single pre-treatment period design, it is arguably supported by the behavioral inertia generally featuring mobility choices, with changes in residential location, workplace (or school place), and habitual travel patterns typically occurring gradually—and after—rather than in anticipation of infrastructure interventions (González et al., 2017). The complementary measures accompanying the CBT opening (i.e., the TILO 2021 offer) did improve PT service frequency and strengthen regional network integration. Besides partly attenuating potential differences across our sample population in exposure to pre-existing PT supply conditions, controlling for baseline PT level of service (i.e., ARE index) also responds to the no-anticipation assumption. Indeed, when comparing PT service quality between 2019 and 2023, several new A and B categories appear in 2023 (i.e., the frequency of PT use has improved at these locations, such as in Lugano, Bellinzona, Locarno). Their spatial extent is also greater, particularly in the Malcantone area, namely west of Lugano (Figure 1). This represented an overall change in transport supply that only materialized once the new rail system schedule became fully operational (late 2020—early 2021), further reducing the likelihood of immediate anticipatory adjustments in mode choice.

Finally, our design also relies on the common support (aka strong overlap) assumption, which requires sufficient overlap in the distribution of relevant attributes between the treatment and control groups. In other words, individuals exposed and not exposed to the CBT intervention should share comparable characteristics, ensuring that the estimated treatment effect is identified from meaningful comparisons rather than from systematic differences between groups. Our preliminary analysis, in fact, indicates that the distributions of socio-demographic characteristics, mobility equipment, and PT level of service (i.e., ARE index) are similar between north-to-south and south-to-north commuters (i.e., the treatment group) and individuals living and working/studying within the same zone (i.e., the control group), with an average level of 2.45 for both groups on a scale from 1–5. This evidence, together with additional descriptive information on the observable characteristics of the groups provided in Appendix (Table 6), supports the plausibility of the common support assumption.

3.2 Difference-in-differences estimates

Estimating model A1 via Ordinary Least Squares regression (OLS) yields an ATT, θ, of 0.067 (95% CI: [0.032, 0.102]). The ATT thus corresponds to a 6.7 p.p. increase, with very high statistical significance (p = 0.000). In other words, the CBT opening is associated with a statistically significant increase of 6.7 p.p. in the share of individuals choosing PT rather than PMT as main commuting mode in the treatment group, relative to the control group (Table 4).

Although our outcome variable is binary, we chose to opt for linear as opposed to logistic regression, to maximize interpretability of coefficients, as recommended by Gomila (2021). In accordance with the literature, we use heteroskedasticity-robust standard errors (MacKinnon and White, 1985). We leverage the python library Pyfixest, which is based on the R library fixest, to implement these models (Bergé et al., 2026). Table 4 summarizes the results of both model specifications (i.e., A1 and A2), showing a positive and statistically significant treatment effect of the CBT opening on commuting mode choice.

When introducing the set of significant covariates identified in the preliminary analysis (i.e., model A2), the estimated effect remains positive and statistically significant, although its magnitude slightly decreases, as expected. The conditional DiD estimator is indeed θ = 0.042 (p = 0.001; 95% CI: [0.017, 0.067]), corresponding to an increase of 4.2 p.p. in PT mode share after the CBT opening, conditional on observed socio-demographic, mobility equipment, and spatial (PT level of service) characteristics. The slight reduction in the estimated effect compared with the baseline specification suggests that part of the unconditional difference between treatment and control groups is explained by these observable characteristics. Nonetheless, the persistence of a significant positive effect indicates that the CBT opening is associated with a robust shift towards PT use among affected commuters. Figure 3 provides a graphical representation of this DiD framework, showing the common pre-treatment trend and the estimated treatment effect associated with the CBT opening.

Figure 3

Furthermore, the inclusion of covariates substantially improves overall model fit. While the baseline specification achieves a relatively low explanatory power (R² = 0.02), adding observed characteristics in model A2 increases the R² to 0.43, indicating that these variables capture a substantial share of the variance in the dependent variable. This improvement indeed reflects the significant contribution of individual and contextual characteristics included in our model (see Data and variables) in explaining mode choice and supports the use of the conditional specification for estimating the CBT effect. Table 5 summarizes the coefficients for the different covariates included in model A2.

Among them, PT subscription (β = 0.480), gender = female (β = 0.057) and PT level of service (β = 0.024) are positively associated with PT mode choice, whereas age is negatively associated with PT use (β = −0.095). However, these coefficients should be interpreted as conditional associations rather than causal effects of the covariates themselves. In other words, the covariates are included in the model primarily to improve comparability between the treatment and control groups and to estimate the CBT opening’s effect conditional on key observed characteristics.

Overall, our results provide robust evidence of a meaningful change of mode choice towards PT use induced by the CBT. This effect is consistent across model specifications and remains statistically significant after accounting for both observed and unobserved time-invariant individual heterogeneity. This suggests that the improved PT service and reduced travel times associated with the tunnel opening did contribute to a shift towards more sustainable commuting behaviors among the commuters directly exposed—by design—to the new railway infrastructure. In other words, the positive rail transit supply shock generated by the tunnel increased the attractiveness of PT compared with PMT along the north-south axis, leading to a reallocation of commuting mode shares in favor of PT. Finally, we conducted a manual DiD, which yielded similar evidence, with the details reported in Appendix (Table 7).

4 Discussion

4.1 Study contributions

Our analysis shows that the CBT induced a structural and statistically robust shift in Ticino residents’ commuting mode choices, moving a significant share of commuters from PMT to PT use. In particular, the 4.2 p.p. increase observed in the share of commuters using PT as their main mode in our treatment group—compared to the control group—is a relatively large effect in transport policy terms, especially when contextualized into the baseline levels of PT use in Ticino. Relative to the pre-treatment share of commuters using PT as their main mode in the treatment group, it corresponds to a 13% increase, from 32% to 36.2% (Table 6). This complements previous descriptive results on mobility practices change after the CBT opening (Repubblica e Cantone Ticino, 2025) by providing causal evidence of the latter fostering PT use compared to PMT. This finding also adds to (1) the studies evaluating the effects of large-scale rail infrastructure breakthroughs, including tunnel openings, on travel behavior (Gumy et al., 2026; Wallimann et al., 2026) and (2) the transport literature on the impact of major rail transit supply shocks on modal shift (Albalate and León-Gómez, 2024; Borsati and Albalate, 2020; Givoni and Dobruszkes, 2013).

Though our method cannot in principle explain the underlying decision-making process through which the CBT opening ultimately affected the commuting mode choice of Ticino residents (for a discussion on the limits of causal inference, see Conte, 2024), our findings do underline the positive impact of railway infrastructure improvements in promoting PT use (Lalive et al., 2013; Uddin and Rana, 2023; Wang et al., 2023, 2025). This resonates with the literature on rail supply shocks (Daniel et al., 2025; Goetz et al., 2024; Hortelano et al., 2016; Wardman and Murphy, 2015): the CBT, together with the complementary PT 2021 offer, improved travel times and service frequency of rail transit along the north-south axis of the region, thus increasing the attractiveness of PT relatively to PMT (Göransson and Andersson, 2023; Nordenholz et al., 2017; Redman et al., 2013). Our study further shows that this effect was especially significant for individuals who regularly commute along the corridors involved by the tunnel: Lugano-Bellinzona and Lugano-Locarno, respectively.

Concerning socio-demographics, our results confirm the evidence that rail transit supply shocks usually generate heterogeneous and polarized impacts across population groups (Gumy et al., 2026; Rames et al., 2021) and not necessarily translate to a large-scale modal shift towards PT (Sohoni et al., 2017). In our case, this reflects into women having a 5.7 p.p. higher—and older age groups a 9.5 p.p. lower— probability of choosing PT over PMT after the CBT opening, conditional on the other covariates (Table 5). This is consistent with studies showing that female and younger households typically use more PT (Chng et al., 2016; Ingvardson and Nielsen, 2019; Kawgan-Kagan, 2020; Roos et al., 2020). Our design, which focuses on work and study commuters, does not allow us to infer the CBT’s effects on travel behavior among the broader regional population. Yet, we assume that these are likely to be more limited among the population groups that are underrepresented in our treatment group, but overrepresented in Ticino relative to the national average, such as car-dependent, suburban and rural residents (Federal Statistical Office, 2024; Repubblica e Cantone Ticino, 2025).

Regarding mobility equipment, our findings show that individuals with a PT subscription have a 48.0 p.p. higher probability of choosing PT rather than PMT after the CBT opening, conditional on the other covariates. Although this is a relatively strong effect, much larger in magnitude than those associated with the other variables tested, the finding is not surprising. On the one hand, the very significant role of mobility equipment following a supply shock echoes the ‘motility’ literature, which explains how successful modal shift hinges upon commuters’ ability to move in terms of access, competences, and appropriation (Grandvillemin et al., forthcoming; Guitton et al., 2025; Kaufmann et al., 2004), among other factors. On the other hand, the positive association between PT subscription and our dependent variable is consistent with the idea that access to subsidized fares and reduced marginal cost per trip eventually nudges commuters towards higher PT use (Parry and Small, 2009; Simma and Axhausen, 2001; Strommer et al., 2023).

Finally, our results show that a one-unit increase in the PT level of service is associated with a 2.4 p.p. increase in the probability of choosing PT rather than PMT as main commuting mode, conditional on the other covariates. This is in line with literature linking mode choice and PT use with socio-spatial barriers: spatial proximity, urban density, infrastructure quality, and regional transport competitiveness (Cass and Faulconbridge, 2016; Dai et al., 2020, 2024; Daniel et al., 2025; Goetz et al., 2024; Scheiner, 2010; Stanton et al., 2013). Our covariate measuring PT level of service (i.e., ARE index) increased by 10% between 2019 and 2023 (Repubblica e Cantone Ticino, 2025), suggesting an improvement in the integration of PT network—and thus access—across the region; which possibly allowed for a successful modal shift of our treatment group towards PT. However, non-PT users still suffer from these transport barriers (e.g., accessibility, service disruptions, high fares), which remain the main obstacle to modal shift—for commuting purposes—in the region (Repubblica e Cantone Ticino, 2025). More importantly, since improved job accessibility¹⁶ (through better PT access) does not necessarily translates into better service accessibility (Grisé et al., 2019), limited levels of the latter can still encourage PMT use across different travel purposes; thus constraining the impact of policies targeting commuting behavior (Pot et al., 2023; Yu and Higgins, 2024).

4.2 Policy implications

Our study provides several implications for transport policy, especially concerning the potential effects—and limits—of rail transit improvements on mode choice and modal shift from PMT. If scholars typically provide recommendations following negative supply shocks (i.e., the impact of rail service disruptions on mobility practices) to tackle their decreasing effects on PT ridership (Adler et al., 2021; Bešinović, 2020; Liu et al., 2023), our case calls for the opposite: drawing insights from the effects of a positive supply shock—that of the CBT opening—on mode choice. From this perspective, a 4.2 p.p. increase in PT mode share among the treated (i.e., exposed) commuters is a particularly relevant policy result, indicating that large-scale rail infrastructure interventions can indeed produce measurable effects—and eventually lasting behavioral changes in mobility practices. Consistently with other studies in the Swiss region, which reached similar conclusions (Gumy et al., 2026; Wallimann et al., 2026) and comparable magnitude effects, our findings suggest that rail transit infrastructure investments can reconfigure mode choice and commuting practices in favor of PT.

However, for supply shocks to successfully translate into modal shifts, transport improvements need to account for several factors (Bernier et al., 2022; De Witte et al., 2013; Kaufmann, 2026; Van Acker et al., 2010). Firstly, demand-side characteristics, including individual attributes and dispositions, can shape individuals’ responses to changes in transport supply. Our results indicate that some population groups (i.e., women and younger individuals) are more likely to choose PT rather than PMT in response to rail supply shocks. This finding not only highlights the role of understanding the socio-demographic context in which transport policies are implemented but also suggests that targeting these interventions towards groups with a greater propensity to respond to them may ultimately help maximize their effectiveness (Velázquez-Cortés et al., 2026). This is also consistent with approaches adopted in other policies, such as fare-free PT schemes that target the population groups most in need (Conte et al., under review).

The second key factor is represented by mobility equipment. On the one hand, the positive and strong association between PT subscription and PT use suggests that the effects of rail supply shocks may be concentrated among individuals who already have some degree of engagement with PT—something that echoes the problem of induced demand (Gumy et al., 2026). This could constrain their potential to induce modal shift among car-dependent individuals and, more broadly, population groups that are less familiar with PT (Rames et al., 2021). On the other hand, mobility resources of competing modes (i.e., driver’s license and car ownership)—discarded from our final model for statistical reasons (Figure 2)—can also shape mode choice. Ticino has the highest vehicle ownership rate among Swiss Cantons, at 635 cars per 1000 people, well above the national average (535 cars per 1000 people) and those of the three largest Swiss metropolitan areas¹⁷ (Federal Statistical Office, 2025a). The relationship between car ownership and mode choice has long been documented (Train, 1980), with the former also shown to mediate the influence of built environment characteristics on the latter (Ding et al., 2017). Against this context of high car dependence, the observed shift from PMT to PT following the CBT opening is particularly noteworthy, suggesting that rail infrastructure improvements may foster PT use even where car ownership is widespread.

Third, spatial factors also shape the conditions for successful modal shift. Our results indicate that pre-existing PT level of service (i.e., the ARE index) is positively associated with the probability of shifting towards PT following a rail supply shock: work/study commuters living in areas with higher PT service quality (i.e., Lugano, Bellinzona, Locarno—the three main urban poles along the corridors affected by the CBT) have a 2.4 p.p. higher probability of shifting from PMT to PT. Besides being consistent with the descriptive evidence showing higher PT use among residents of these cities after the intervention (Repubblica e Cantone Ticino, 2025), this suggests that the effects of rail infrastructure improvements are partly conditioned by the surrounding PT level of service quality. A well-developed network may facilitate the integration of improved rail services into individuals’ existing mobility practices (Cass and Faulconbridge, 2016) as well as time-space constraints (Scheiner, 2010; Stanton et al., 2013), thereby making modal shift more likely. Although smaller than the associations of the other covariates in our model, this effect highlights the importance of pre-existing PT network conditions. Major interventions should indeed be embedded within broader transport strategies aimed at improving PT accessibility, particularly in regions like the Canton of Ticino, where PT service provision remains below the national average (Federal Office for Spatial Development, 2026). At the same time, improving PT accessibility does not necessarily translate into better service accessibility (Grisé et al., 2019), underscoring the need to address both dimensions (Grandvillemin et al., forthcoming; see also Kaufmann et al., 2004).

Despite leveraging these individual and spatial factors, PT supply shocks alone may not be sufficient to scale up modal shift towards PT and sustainable transport at the broader regional or national level (for an account of the determinants of modal shift, see Kaufmann, 2026). To fully exploit the potential of railway infrastructure breakthroughs and potentially amplify the effects identified in this and similar studies (Gumy et al., 2026; Wallimann et al., 2026), transport policies should consider complementary interventions (Winkler et al., 2023; Xiao et al., 2022). These include, among others, coercive measures (Abbasi et al., 2023; Banister, 2008), ticket fare subsidies (Conte et al., under review; Daniel et al., 2025), and behavioral interventions (Cellina et al., 2025; Franssens et al., 2021; Steffen et al., 2024).

4.3 Limitations and future research

The main limitations of our study are threefold. Firstly, our research design made us artificially neglect possibly relevant population groups (see Research design): namely, (1) those who travel for leisure or other reasons than work and study (e.g., retired people, tourists coming from outside Ticino); (2) cross-border commuters traveling from Italy to (southern more than northern) Ticino for work/study reasons; (3) people traveling through the tunnel despite not living and working/studying at each end of its axis (e.g., for professional reasons besides commuting, continuing education, longer trips to the north of the Alps); (4) those mainly commuting to the workplace (or school place) by active modes. The first group was excluded for data availability reasons, since the absence of travel destination information for non-work/study travel did not allow us to consider this broader sample. The second and third groups were excluded due to their sample size; representing a minority of the examined population (Repubblica e Cantone Ticino, 2025). Given the geographical distance involved, the fourth group similarly represents a very small portion of commuters, which further lies beyond the scope of our study (i.e., focused on PT and PMT only). Although we do not expect the main results to substantially change, further research may want to extend our study design by including some of these groups (i.e., people traveling for non-work purposes, such as leisure) to the impact evaluation analysis.

A second limitation concerns our method relying exclusively on survey data. The collected information about travel behavior remains merely grounded on declared statements, without necessarily capturing actual behavior. Although focusing on the main commuting mode rather than the declared frequency of mode use somehow mitigates this risk, the attitude-behavior gap poses fundamental issues (Lanzini and Khan, 2017; Weyer and Hoffmann, 2025). In our case, they are possibly compounded by the survey design, which relies on data collected in 2023 asking respondents to recall their travel behavior in 2019; thereby increasing the potential for recall bias. To tackle this issue, quite common in transport research, future studies could run field experiments (e.g., Randomized Controlled Trials) where the policy to be assessed already embeds a data collection phase. We thus call on transport policy institutions to plan measures that include a research design component aimed at evaluating their impact. Besides reducing the biases mentioned above, this would also increase the reliability of causal inference and enable a more precise assessment of the behavioral effects of transport policies.

Thirdly, the period covered by our study overlaps with a major disruption to everyday life, namely the COVID-19 pandemic. Constituting an external and global shock, it potentially affected travel behavior and mode choice of our population. Although this in principle does not change the interpretation of our results (i.e., everything that happened between 2019 and 2023 is by default absorbed by the time fixed effect), it is worth to ask whether COVID-19 may have caused the post-treatment trends to differ. That is, our estimates may still be biased if the pandemic had heterogeneous impacts across the two groups. One plausible mechanism in this sense is remote working and distance learning practices, which may have reshaped post-pandemic mobility patterns differently across regions and population groups (e.g., by affecting urban residents and north-south commuters disproportionately compared to suburban and rural residents and intra-zonal commuters). Having access to additional data in this sense would allow us to refute this hypothesis—which remains highly unlikely, given our population characteristics—and eventually make our results more robust, further confirming whether they still apply after this external shock¹⁸ (for a discussion of this point, see Gumy et al., 2026).

Future research may additionally want to extend our analysis by collecting data from multiple time periods (i.e., more than three years after the CBT opening). At the technical level, this would allow to explicitly test pre-treatment trends as well as the long-term dynamics of the target effect (e.g., delayed or adjustment effects). At the conceptual level, this would allow to assess the policy’s lasting effects on travel behavior and provide insights on behavioral inertia (González et al., 2017; Ingvardson et al., 2025) following infrastructure interventions and life events (e.g., change of residence, job location) in influencing mode choice (Gerber et al., 2017). Moreover, future studies may explore the heterogeneity of the effect (e.g., by PT level of service, PT subscription status) to understand where the tunnel had the greatest impact and where there is room for complementary interventions. This would further allow to consider more subtle dimensions of mode choice (e.g., individual perceptions, social norms), which often reveal key factors determining how transport improvements translate into commuting behavior (Abou-Zeid et al., 2012; Bernier et al., 2022; De Vos et al., 2016; Gumy et al., 2026; Kaufmann, 2026; Lyons et al., 2007; Meinherz and Binder, 2020).

5 Conclusion

Our study provides causal evidence that the Ceneri Base Tunnel (CBT) opening shifted mode choice towards public transport (PT) and away from private motorized transport (PMT) among work and study commuters along its north-south axis. Using a Difference-in-Differences (DiD) approach, we estimate a positive statistically significant effect of 4.2 percentage points (p.p.) on the probability of using PT as the main work/study commuting mode in Ticino between 2019 and 2023, corresponding to a 13% relative increase compared with the baseline PT share in the treatment group. Our results further show that PT use is associated with socio-spatial characteristics: women and younger commuters, individuals with a PT subscription, and those living in areas with higher PT levels of service (i.e., urban areas) are more likely to use PT, conditional on the other covariates.

Besides contributing to the transport literature on the effects of railway improvements on mode choice, our findings support major rail investments as an effective component of modal shift policies. Beyond techno-solutionist approaches focused on vehicle energy efficiency or autonomous driving, reshaping commuting practices towards PT remains key to sustainable mobility. In highly car-dependent contexts such as Ticino, where relatively low PT use and high car ownership may indicate substantial potential for modal shift, railway infrastructure breakthroughs could act as "flagship measures" that challenge established mobility patterns (Kaufmann, 2026). At the same time, the association between PT use and individual attributes, mobility resources and PT levels of service highlights the importance of considering all these factors when designing broader strategies for sustainable transport transitions.

References

• 1

  AbbasiM.MamdoohiA. R.ArndtW.-H. (2023). Modal shift behavior of car users to light rail transit, some evidence from the field. Trans. Transp. Sci.14, 24–29. doi: 10.5507/tots.2022.021

  • CrossRef
  • Google Scholar

• 2

  Abou-ZeidM.WitterR.BierlaireM.KaufmannV.Ben-AkivaM. (2012). Happiness and travel mode switching: findings from a Swiss public transportation experiment. Transp. Policy19, 93–104. doi: 10.1016/j.tranpol.2011.09.009

  • CrossRef
  • Google Scholar

• 3

  AdlerM. W.LiberiniF.RussoA.OmmerenJ. N. V. (2021). The congestion relief benefit of public transit: evidence from Rome. J. Econ. Geogr.21, 397–431. doi: 10.1093/jeg/lbaa037

  • CrossRef
  • Google Scholar

• 4

  AlbalateD.León-GómezC. R. (2024). High-speed rail: a game changer for Spanish motorway transport?. Econ. Transp.39:100360. doi: 10.1016/j.ecotra.2024.100360

  • CrossRef
  • Google Scholar

• 5

  AngristJ. D.PischkeJ. S. (2009). Mostly Harmless Econometrics: An Empiricist’s Companion. Princeton, NJ: Princeton University Press.

  • Google Scholar

• 6

  AnagnostopoulouE.UrbančičJ.BothosE.MagoutasB.BradeskoL.SchrammelJ.et al. (2020). From mobility patterns to behavioural change: leveraging travel behaviour and personality profiles to nudge for sustainable transportation. J. Intell. Inf. Syst.54, 157–178. doi: 10.1007/s10844-018-0528-1

  • CrossRef
  • Google Scholar

• 7

  BakerA.CallawayB.CunninghamS.Goodman-BaconA.Sant’AnnaP. (2026). Difference-in-differences designs: a practitioner’s guide. J. Econ. Lit.64, 498–557. doi: 10.1257/jel.20251650

  • CrossRef
  • Google Scholar

• 8

  BakerC. J.ChapmanL.QuinnA.DobneyK. (2010). Climate change and the railway industry: a review. Proc. Inst. Mech. Eng., Part F: J. Rail Rapid Transit224, 519–528. doi: 10.1243/09544062jmes1558

  • CrossRef
  • Google Scholar

• 9

  BanisterD. (2008). The sustainable mobility paradigm. Transp. Policy15, 73–80. doi: 10.1016/j.tranpol.2007.10.005

  • CrossRef
  • Google Scholar

• 10

  BergéL. R.ButtsK.McDermottG. (2026). Fast and user-friendly econometrics estimations: the R package fixest. arXiv:2601.21749. doi: 10.48550/arXiv.2601.21749

  • CrossRef
  • Google Scholar

• 11

  BernierE.KaufmannV.MasseF.GonzalezJ. (2022). Analysing the potential for modal shift based on the logic of modal choice. disP. - Plan. Rev.58, 54–67. doi: 10.1080/02513625.2022.2200659

  • CrossRef
  • Google Scholar

• 12

  BešinovićN. (2020). Resilience in railway transport systems: a literature review and research agenda. Transp. Rev.40, 457–478. doi: 10.1080/01441647.2020.1728419

  • CrossRef
  • Google Scholar

• 13

  BorsatiM.AlbalateD. (2020). On the modal shift from motorway to high-speed rail: evidence from Italy. Transp. Res. A Policy Pract.137, 145–164. doi: 10.1016/j.tra.2020.04.006

  • CrossRef
  • Google Scholar

• 14

  CardD.KruegerA. B. (1994). Minimum wages and employment: a case study of the fast-food industry in New Jersey and Pennsylvania. Am. Econ. Rev.84, 772–793.

  • Google Scholar

• 15

  CassN.FaulconbridgeJ. (2016). Commuting practices: new insights into modal shift from theories of social practice. Transp. Policy45, 1–14. doi: 10.1016/j.tranpol.2015.08.002

  • CrossRef
  • Google Scholar

• 16

  CellinaF.GerosaT.VentimigliaL.HoerlerR.Del DuceA.Klopfenstein FreiN.et al. (2025). Can we become car-free in just one month? Evidence from Switzerland. Transp. Res. Part D Transp. Environ.148:104991. doi: 10.1016/j.trd.2025.104991

  • CrossRef
  • Google Scholar

• 17

  ChngS.WhiteM.AbrahamC.SkipponS. (2016). Commuting and wellbeing in London: the roles of commute mode and local public transport connectivity. Prev. Med.88, 182–188. doi: 10.1016/j.ypmed.2016.04.014, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 18

  ConteL. (2024). Qualitative economics: integrating qualitative research methods into economic analysis. (doctoral thesis). University of Fribourg, Switzerland. Available online at: https://folia.unifr.ch/unifr/documents/329492 (accessed August 31, 2026).

  • Google Scholar

• 19

  ConteL.RamesC.BernierE.RavaletE.KaufmannV.ReckD. J. (under review). The impact of fare-free public transport (FFPT) on travel behaviour: a comprehensive review. Transp. Policy

  • Google Scholar

• 20

  CurrieJ.GruberJ. (1996). Health insurance eligibility, utilization of medical care, and child health. Q. J. Econ.111, 431–466. doi: 10.2307/2946684

  • CrossRef
  • Google Scholar

• 21

  DaiF.DiaoM.SingT. F. (2020). Effects of rail transit on individual travel mode shares: a two-dimensional propensity score matching approach. Transp. Res. Part D Transp. Environ.89:102601. doi: 10.1016/j.trd.2020.102601

  • CrossRef
  • Google Scholar

• 22

  DaiF.DiaoM.SingT. F. (2024). A two-dimensional propensity score matching approach to estimating the treatment effect of urban rail transit lines on vehicle travel. Transportation51, 853–873. doi: 10.1007/s11116-022-10360-2

  • CrossRef
  • Google Scholar

• 23

  DanielT.GailM. M.KlotzP.-A. (2025). From highway to rail? Germany’s Public Transport Ticket Experiment (SSRN Working Paper No. 5138487). SSRN.

  • Google Scholar

• 24

  DasS.BoruahA.BanerjeeA.RaoniarR.NamaS.MauryaA. K. (2021). Impact of COVID-19: a radical modal shift from public to private transport mode. Transp. Policy109, 1–11. doi: 10.1016/j.tranpol.2021.05.005, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 25

  De GottardiR. (2022). L’evoluzione della domanda di trasporto di persone in Svizzera e in Ticino (Documenti No. 14). Ufficio di statistica, Repubblica e Cantone Ticino.

  • Google Scholar

• 26

  De VosJ.MokhtarianP. L.SchwanenT.Van AckerV.WitloxF. (2016). Travel mode choice and travel satisfaction: bridging the gap between decision utility and experienced utility. Transportation43, 771–796. doi: 10.1007/s11116-015-9619-9

  • CrossRef
  • Google Scholar

• 27

  De WitteA.HollevoetJ.DobruszkesF.HubertM.MacharisC. (2013). Linking modal choice to motility: a comprehensive review. Transp. Res. A Policy Pract.49, 329–341. doi: 10.1016/j.tra.2013.01.009

  • CrossRef
  • Google Scholar

• 28

  DingC.WangD.LiuC.ZhangY.YangJ. (2017). Exploring the influence of built environment on travel mode choice considering the mediating effects of car ownership and travel distance. Transp. Res. A Policy Pract.100, 65–80. doi: 10.1016/j.tra.2017.04.008

  • CrossRef
  • Google Scholar

• 29

  Federal Department of the Environment, Transport, Energy and Communications (2025). The New Rail Link through the Alps NRLA. Available online at: https://www.uvek.admin.ch/en/nrla (accessed September 4, 2026).

  • Google Scholar

• 30

  Federal Office for Spatial Development (2017). Mobility and Transport Microcensus 2015: Population’s Mobility Behaviour in Switzerland. Available online at: https://www.are.admin.ch/en/mtmc

  • Google Scholar

• 31

  Federal Office for Spatial Development (2026). Desserte en Suisse. Available online at:https://www.are.admin.ch/fr/desserte (accessed August 2, 2026).

  • Google Scholar

• 32

  Federal Statistical Office (2024). Statistical Atlas of Switzerland: Modal Split of Commuters – Private Motor Vehicles, 2020–2022. Swiss Confederation. Available online at: https://mapexplorer.bfs.admin.ch/#c=home

  • Google Scholar

• 33

  Federal Statistical Office. (2025a). Mobility and Transport. Available online at: https://www.bfs.admin.ch/bfs/en/home/statistics/mobility-transport.html (accessed August 14, 2026).

  • Google Scholar

• 34

  Federal Statistical Office (2025b). Permanent and Non-permanent Resident Population by Canton, Citizenship (Selection), Country of Birth, Sex and Age Class, 2020–2025. Available online at: https://www.bfs.admin.ch/asset/en/px-x-0103010000_325 (accessed August 14, 2026).

  • Google Scholar

• 35

  FerrataC. (2022). Fare la Città Ticino: Una Metafora Geografica per il territorio. GEA – Associazione dei geografi. Available online at: https://www.gea-ticino.ch/wp-content/uploads/farelacitta-ticino.pdf (accessed July 10, 2026).

  • Google Scholar

• 36

  FranssensS.BotchwayE.de SwartW.DewitteS. (2021). Nudging commuters to increase public transport use: a field experiment in Rotterdam. Front. Psychol.12:633865. doi: 10.3389/fpsyg.2021.633865, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 37

  GerberP.MaT. Y.KleinO.SchiebelJ.Carpentier-PostelS. (2017). Cross-border residential mobility, quality of life and modal shift: a Luxembourg case study. Transp. Res. A Policy Pract.104, 238–254. doi: 10.1016/j.tra.2017.06.015

  • CrossRef
  • Google Scholar

• 38

  GibbonsS.LyytikäinenT.OvermanH. G.Sanchis-GuarnerR. (2019). New road infrastructure: the effects on firms. J. Urban Econ.110, 35–50. doi: 10.1016/j.jue.2019.01.002

  • CrossRef
  • Google Scholar

• 39

  GivoniM.DobruszkesF. (2013). A review of ex-post evidence for mode substitution and induced demand following the introduction of high-speed rail. Transp. Rev.33, 720–742. doi: 10.1080/01441647.2013.853707

  • CrossRef
  • Google Scholar

• 40

  GoetzA.MarinicaI.MayrH.MosettiL.SchubertR. (2024). Do mobile applications foster sustainable mobility? Evidence from a field experiment. Swiss J. Econ. Stat.160:12. doi: 10.1186/s41937-024-00129-y, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 41

  GomilaR. (2021). Logistic or linear? Estimating causal effects of experimental treatments on binary outcomes using regression analysis. J. Exp. Psychol. Gen.150, 700–709. doi: 10.1037/xge0000920, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 42

  GonzálezR. M.MarreroÁ. S.CherchiE. (2017). Testing for inertia effect when a new tram is implemented. Transp. Res. A Policy Pract.98, 150–159. doi: 10.1016/j.tra.2017.02.007

  • CrossRef
  • Google Scholar

• 43

  GöranssonJ.AnderssonH. (2023). Factors that make public transport systems attractive: a review of travel preferences and travel mode choices. Eur. Transp. Res. Rev.15:32. doi: 10.1186/s12544-023-00609-x

  • CrossRef
  • Google Scholar

• 44

  GrandvilleminJ.MasseF.Carpentier-PostelS.KaufmannV. (forthcoming). Evaluating sustainable mobility potential through geographical accessibility and individual access, skills and appropriation. Environ. Plann. B. Plann. Des. (Special Issue: Novel Approaches to the 15-Minute City).

  • Google Scholar

• 45

  GriséE.BoisjolyG.MaguireM.El-GeneidyA. (2019). Elevating access: comparing accessibility to jobs by public transport for individuals with and without a physical disability. Transp. Res. A Policy Pract.125, 280–293. doi: 10.1016/j.tra.2018.02.017

  • CrossRef
  • Google Scholar

• 46

  GuggisbergD.HörcherD.HeinenE.de LapparentM. (2026). "Towards a quantitative spatial model for Switzerland: prototype implementation and lessons learned from Ticino,” in Paper presented at the 26th Swiss Transport Research Conference (STRC 2026), Ascona.

  • Google Scholar

• 47

  GuittonE.EisenmanL.GuerinC.PotelM.SomatA. (2025). Defining motility: the uses, operationalisations and limits of a concept. Mobilities20, 969–985. doi: 10.1080/17450101.2024.2449515

  • CrossRef
  • Google Scholar

• 48

  GumyA.KaufmannV.VaroneF.Masood DechevrensM. (2026). The partial adaptation of mobility practices to public transport shock effects: evidence from the Greater Geneva cross-border region. Transp. Res. Interdiscip. Perspect.39:102219. doi: 10.1016/j.trip.2026.102219

  • CrossRef
  • Google Scholar

• 49

  HensherD. A.RoseJ. M. (2007). Development of commuter and non-commuter mode choice models for induced travel demand estimation. Transp. Res. A Policy Pract.41, 241–253. doi: 10.1016/j.tra.2006.09.006

  • CrossRef
  • Google Scholar

• 50

  HortelanoA. O.GuzmanA. F.PrestonJ.VassalloJ. M. (2016). Price elasticity of demand on the high-speed rail lines of Spain: impact of the new pricing scheme. Transp. Res. Rec.2597, 90–98. doi: 10.3141/2597-12

  • CrossRef
  • Google Scholar

• 51

  IngvardsonJ. B.NielsenO. A. (2019). The relationship between norms, satisfaction and public transport use: a comparison across six European cities using structural equation modelling. Transp. Res. A Policy Pract.126, 37–57. doi: 10.1016/j.tra.2019.05.016

  • CrossRef
  • Google Scholar

• 52

  IngvardsonJ. B.RaveauS.Soza-ParraJ. (2025). Inertia and shock effects in public transport: the case of metro line 6 in Santiago using smart card data. Transp. Res. A Policy Pract.192:104352. doi: 10.1016/j.tra.2024.104352

  • CrossRef
  • Google Scholar

• 53

  Intergovernmental Panel on Climate Change (IPCC) (2023) Climate Change 2022 - Mitigation of Climate Change: Working Group III Contribution to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change.Cambridge: Cambridge University Press.

  • Google Scholar

• 54

  International Transport Forum (2023). ITF Transport Outlook 2023. Paris: OECD Publishing. Available online at: https://www.oecd.org/en/publications/itf-transport-outlook-2023_b6cc9ad5-en.html (accessed August 13, 2026).

  • Google Scholar

• 55

  KaufmannV. (2002). Re-thinking Mobility: Contemporary Sociology. Aldershot: Ashgate.

  • Google Scholar

• 56

  KaufmannV. (2026). Le report modal: du désir à la pratique. EPFL Press.

  • Google Scholar

• 57

  KaufmannV.BergmanM. M.JoyeD. (2004). Motility: mobility as capital. Int. J. Urban Reg. Res.28, 745–756. doi: 10.1111/j.0309-1317.2004.00549.x

  • CrossRef
  • Google Scholar

• 58

  Kawgan-KaganI. (2020). Are women greener than men? A preference analysis of women and men from major German cities over sustainable urban mobility. Transp. Res. Interdiscip. Perspect.8:100236. doi: 10.1016/j.trip.2020.100236

  • CrossRef
  • Google Scholar

• 59

  KrausS.KochN. (2021). Provisional COVID-19 infrastructure induces large, rapid increases in cycling. Proc. Natl. Acad. Sci.118:e2024399118. doi: 10.1073/pnas.2024399118, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 60

  LaliveR.LuechingerS.SchmutzlerA. (2013). Does Supporting Passenger Railways Reduce Road Traffic Externalities? (CEPR Discussion Paper No. 9335), Centre for Economic Policy Research.

  • Google Scholar

• 61

  LanziniP.KhanS. A. (2017). Shedding light on the psychological and behavioral determinants of travel mode choice: a meta-analysis. Transp. Res. Part F Psychol. Behav.48, 13–27. doi: 10.1016/j.trf.2017.04.020

  • CrossRef
  • Google Scholar

• 62

  LechnerM. (2011). The estimation of causal effects by difference-in-difference methods. Found. Trends Econom.4, 165–224. doi: 10.1561/0800000014

  • CrossRef
  • Google Scholar

• 63

  LitmanT. (2023). Evaluating Public Transportation Health Benefits. Victoria Transport Policy Institute. Available online at: https://www.vtpi.org/tran_health.pdf (accessed June 2, 2026).

  • Google Scholar

• 64

  LiuY.OsorioJ.OuyangY. (2023). How long it took transit ridership to recover from disruptive events: a review into the recent history. J. Public Transp.25:100051. doi: 10.1016/j.jpubtr.2023.100051

  • CrossRef
  • Google Scholar

• 65

  LuhS.KannanR.McKennaR.SchmidtT. J.KoberT. (2024). Quantifying the impact of travel time duration and valuation on modal shift in Swiss passenger transportation. Appl. Energy356:Article 122412. doi: 10.1016/j.apenergy.2023.122412

  • CrossRef
  • Google Scholar

• 66

  LyonsG.JainJ.HolleyD. (2007). The use of travel time by rail passengers in Great Britain. Transp. Res. A Policy Pract.41, 107–120. doi: 10.1016/j.tra.2006.05.012

  • CrossRef
  • Google Scholar

• 67

  MacKinnonJ. G.WhiteH. (1985). Some heteroskedasticity-consistent covariance matrix estimators with improved finite sample properties. J. Econ.29, 305–325. doi: 10.1016/0304-4076(85)90158-7

  • CrossRef
  • Google Scholar

• 68

  MeinherzF.BinderC. R. (2020). The dynamics of modal shifts in (sub)urban commuting: an empirical analysis based on practice theories. J. Transp. Geogr.86:102763. doi: 10.1016/j.jtrangeo.2020.102763

  • CrossRef
  • Google Scholar

• 69

  NordenholzF.WinklerC.KnörrW. (2017). Analysing the modal shift to rail potential within the long-distance passenger travel market in Germany. Transp. Res. Procedia26, 81–91. doi: 10.1016/j.trpro.2017.07.010

  • CrossRef
  • Google Scholar

• 70

  ParryI. W. H.KennethA. S. (2009). Should urban transit subsidies be reduced?. Am. Econ. Rev.99, 700–724. doi: 10.1257/aer.99.3.700

  • CrossRef
  • Google Scholar

• 71

  PotF. J.KosterS.TillemaT. (2023). Perceived accessibility and residential self-selection in the Netherlands. J. Transp. Geogr.108:103555. doi: 10.1016/j.jtrangeo.2023.103555

  • CrossRef
  • Google Scholar

• 72

  RamesC.WilsonA. M.Zimny-SchmittD.NeriC.SperlingJ.Romero-LankaoP. (2021). A data-driven mobility–energy typology framework for New York State. Environ. Plann. B: Urban Anal. City Sci.48, 2254–2271. doi: 10.1177/2399808320974032

  • CrossRef
  • Google Scholar

• 73

  RedmanL.FrimanM.GärlingT.HartigT. (2013). Quality attributes of public transport that attract car users: a research review. Transp. Policy25, 119–127. doi: 10.1016/j.tranpol.2012.11.005

  • CrossRef
  • Google Scholar

• 74

  Repubblica e Cantone Ticino (2025). Indagine sui cambiamenti delle abitudini di mobilità nel Canton Ticino seguito all’apertura della galleria di base del Ceneri e all’introduzione del trasporto pubblico 2021. Dipartimento del territorio, Sezione della mobilità. Available online at: https://www.ti.ch/fileadmin/dt/mobilita_ticino/File/20250210_INDMOB_Rapporto_completo.pdf (accessed March 30, 2026).

  • Google Scholar

• 75

  RoosJ. M.SpreiF.HolmbergU. (2020). Sociodemography, geography, and personality as determinants of car driving and use of public transportation. Behav. Sci.10:93. doi: 10.3390/bs10060093

  • CrossRef
  • Google Scholar

• 76

  RothJ. (2022). Pretest with caution: event-study estimates after testing for parallel trends. Am. Econ. Rev.: Insights4, 305–322. doi: 10.1257/aeri.20210236

  • CrossRef
  • Google Scholar

• 77

  RothJ.Sant’AnnaP. H. C.BilinskiA.PoeJ. (2023). What’s trending in difference-in-differences? A synthesis of the recent econometrics literature. J. Econ.235, 2218–2244. doi: 10.1016/j.jeconom.2023.03.008

  • CrossRef
  • Google Scholar

• 78

  ScheinerJ. (2010). Interrelations between travel mode choice and trip distance: trends in Germany 1976–2002. J. Transp. Geogr.18, 75–84. doi: 10.1016/j.jtrangeo.2009.01.001

  • CrossRef
  • Google Scholar

• 79

  SimmaA.AxhausenK. W. (2001). Structures of commitment in mode use: a comparison of Switzerland, Germany and Great Britain. Transp. Policy8, 279–288. doi: 10.1016/S0967-070X(01)00023-3

  • CrossRef
  • Google Scholar

• 80

  SohoniA. V.ThomasM.RaoK. V. K. (2017). Mode shift behavior of commuters due to the introduction of new rail transit mode. Transp. Res. Procedia25, 2603–2618. doi: 10.1016/j.trpro.2017.05.311

  • CrossRef
  • Google Scholar

• 81

  StantonN. A.McIlroyR. C.HarveyC.BlaineyS.HickfordA.PrestonJ. M.et al. (2013). Following the cognitive work analysis train of thought: exploring the constraints of modal shift to rail transport. Ergonomics56, 522–540. doi: 10.1080/00140139.2012.718366, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 82

  SteffenJ.HookH.WitloxF. (2024). Improving interest in public, active, and shared travel modes through nudging interventions. Transp. Res. Part F Psychol. Behav.103, 353–367. doi: 10.1016/j.trf.2024.04.020

  • CrossRef
  • Google Scholar

• 83

  StrommerT.HörcherD.MunkácsyA. (2023). Crowding externalities and optimal subsidies in public transport: revisiting the Parry–Small model. Res. Transp. Econ.100:101324. doi: 10.1016/j.retrec.2023.101324

  • CrossRef
  • Google Scholar

• 84

  TrainK. (1980). A structured logit model of auto ownership and mode choice. Rev. Econ. Stud.47, 357–370. doi: 10.2307/2296997

  • CrossRef
  • Google Scholar

• 85

  Transport and Environment (2024). The State of European Transport 2024: An Overview of the EU’s Largest Climate Problem. Available online at: https://www.transportenvironment.org/uploads/files/TE_SoT_2024_report_2024-04-29-155117_davx.pdf (Accessed April 15, 2026).

  • Google Scholar

• 86

  UddinS. Z.RanaM. (2023). Systematic review on the impact of large-scale railway infrastructure on regional connectivity and resilience in the U.S. Int. J. Sci. Interdiscip. Res.4, 177–223. doi: 10.63125/p06cv674

  • CrossRef
  • Google Scholar

• 87

  Van AckerV.Van WeeB.WitloxF. (2010). When transport geography meets social psychology: toward a conceptual model of travel behaviour. Transp. Rev.30, 219–240. doi: 10.1080/01441640902943453

  • CrossRef
  • Google Scholar

• 88

  Velázquez-CortésD.Rojas-RuedaD.TonneC.MuellerN.NieuwenhuijsenM. J. (2026). Substituting car trips with public and active transport in European urban areas: quantifying potential modal shift across the EU27. J. Transp. Health50:102346. doi: 10.1016/j.jth.2026.102346

  • CrossRef
  • Google Scholar

• 89

  WallimannH.Von ArxW.HesseA. (2026). Rail infrastructure and road use: causal evidence from the Gotthard base tunnel. Transp. Res. A Policy Pract.206:104922. doi: 10.1016/j.tra.2026.104922

  • CrossRef
  • Google Scholar

• 90

  WangJ.ChengW.LuY.WangD. (2023). Effect of rail transit on travel behavior: a systematic review and meta-analysis. Transp. Res. Part D Transp. Environ.122:103882. doi: 10.1016/j.trd.2023.103882

  • CrossRef
  • Google Scholar

• 91

  WangJ.DuH.LinK. (2025). Substitution effects of high-speed railway on carbon mitigation: from theory to empirics. Energy Econ.142:108179. doi: 10.1016/j.eneco.2025.108179

  • CrossRef
  • Google Scholar

• 92

  WardJ. W.MichalekJ. J.SamarasC.AzevedoI. L.HenaoA.RamesC.et al. (2021). The impact of Uber and Lyft on vehicle ownership, fuel economy, and transit across U.S. cities. Science24:101933. doi: 10.1016/j.isci.2020.101933, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 93

  WardmanM.MurphyP. (2015). Passengers’ valuations of train seating layout, position and occupancy. Transp. Res. A Policy Pract.74, 222–238. doi: 10.1016/j.tra.2015.01.007

  • CrossRef
  • Google Scholar

• 94

  WeyerJ.HoffmannS. (2025). Bridging the attitude-behaviour gap: an explanation of travel mode choice using analytical sociology. PLoS One20:e0330073. doi: 10.1371/journal.pone.0330073, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 95

  WinklerL.PearceD.NelsonJ.BabacanO. (2023). The effect of sustainable mobility transition policies on cumulative urban transport emissions and energy demand. Nat. Commun.14:2357. doi: 10.1038/s41467-023-37728-x, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 96

  XiaoC.Van SluijsE.OgilvieD.PattersonR.PanterJ. (2022). Shifting towards healthier transport: carrots or sticks? Systematic review and meta-analysis of population-level interventions. Lancet Planet. Health6, e858–e869. doi: 10.1016/S2542-5196(22)00220-0, .

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 97

  YuA.HigginsC. D. (2024). Travel behaviour and the 15-min city: access intensity, sufficiency, and non-work car use in Toronto. Travel Behav. Soc.36:100786. doi: 10.1016/j.tbs.2024.100786

  • CrossRef
  • Google Scholar

• 98

  ZhangY.WangD.ChenX.LiQ. (2021). Impacts of high-speed rail projects on CO₂ emissions due to modal interactions: a review. Transp. Res. Part D Transp. Environ.100:103081. doi: 10.1016/j.trd.2021.103081

  • CrossRef
  • Google Scholar