Content uploaded by Isabella Waldorf
Author content
All content in this area was uploaded by Isabella Waldorf on Oct 17, 2023
Content may be subject to copyright.
Germany’s 9-Euro-Ticket: Impacts on1
Disadvantaged Groups Using a Causal Inference2
Approach3
Isabella Waldorf*4
Chair of Traffic Engineering and Control, Technical University of Munich, Germany5
Email: isabella.waldorf@tum.de6
Allister Loder7
Chair of Traffic Engineering and Control, Technical University of Munich, Germany8
Email: allister.loder@tum.de9
Stefan Wurster10
Assistant Professorship of Policy Analysis, Technical University of Munich, Germany11
Email: stefan.wurster@tum.de12
Klaus Bogenberger13
Chair of Traffic Engineering and Control, Technical University of Munich, Germany14
Email: klaus.bogenberger@tum.de15
* Corresponding author16
Word count: 4819 words + 7 table(s) ×250 + 750 words for references = 7319 words17
18
Submitted: August 1, 202319
20
Paper submitted for presentation at the 103rd Annual Meeting Transportation Research Board,21
Washington D.C., January 202422
ABSTRACT1
Free-fare policies have been proposed as a means to reduce emissions in the transport sector2
and promote equitable mobility. However, their potential distributional impacts on disadvantaged3
groups remain uncertain. Using data from Germany’s 9-Euro-Ticket, we analyze the effects of4
nearly free transit on individuals with a low economic status, women and individuals with a dis-5
ability. To offer a comprehensive evaluation we include the effects on activity participation, use of6
public transport and financial relief in our analysis. Relying on observational data where users self-7
select the treatment rather than being randomly assigned, we utilize a quasi-experimental method,8
Propensity Score Matching, combined with weighted regression models. This doubly robust ap-9
proach enables us to identify causal effects. Our findings indicate that the 9-Euro-Ticket increased10
public transport use across all groups and improved activity participation, particularly among eco-11
nomically disadvantaged individuals. However, the program did not seem to offer targeted finan-12
cial relief for economically marginalized individuals, and its benefits were less pronounced for13
women and people with disabilities. These results underscore the positive impact of low-fare pub-14
lic transport on economically marginalized individuals but also highlight its limited effectiveness15
in addressing barriers faced by other disadvantaged groups. The findings have important impli-16
cations for policymakers and transport planners seeking to make public transport more accessible17
and equitable.18
Keywords: Transport Equity, Propensity Score Matching, Public Transit, Fares, Policy Evaluation,19
9-Euro-Ticket20
Waldorf et al. 2
INTRODUCTION1
Free fare policies have been discussed as an instrument for sustainable development, seeking to2
promote public transit, increase ridership, reduce the negative externalities of car traffic and im-3
prove mobility for all (1). The 9-Euro-Ticket that was introduced by the German government4
between June and August 2022 allows to evaluate the effects of nearly fare-free transit for all. The5
ticket cost nine euros (around 10 USD) per calendar month and was valid on all local and regional6
buses and trains throughout Germany. It was part of a federal relief package in reaction to the war7
in Ukraine that negatively affected the German economy, aiming to provide financial relief. Next to8
the economic aspect the 9-Euro-Ticket was also meant as an incentive to switch to climate-friendly9
public transport and save fuels (2). The total of the funds for three months, amounting to 2.5 bil-10
lion euros (around 2.8 billion USD), was derived from the forecast of the lost ticket revenue of the11
federal states. Overall, fifty-two million tickets were sold from June to August 2022. Additionally,12
around 10 million subscribers received the ticket automatically for the period of its availability (3).13
While economists generally agree that optimal pricing of public transport should equal14
its marginal costs, this principle is only applicable if all other transport prices are also based on15
marginal cost pricing (4). Transit subsidies are furthermore typically justified on three grounds:16
to guarantee the provision of a public service that is often unprofitable for the operators, to secure17
the positive externalities of public transport and to redistribute income to specific groups (5). Ac-18
cording to German Law, public transport is considered a public service that the state must provide19
to ensure its citizens’ mobility. To fulfill this obligation, the state secures the necessary funding20
(6). In contrast to targeted fare subsidies, the 9-Euro-Ticket was available to purchase for all travel21
users reducing overall administrative costs for the government. However, as the ticket was a blunt22
policy instrument, it remains unclear whether all social groups could benefit. Decisions regarding23
the transportation system greatly influence people’s lives by decreasing or creating access to a wide24
range of opportunities, impacting individuals’ life chances and agency (7). How transport policies25
affect social groups with different transportation abilities and needs is thus also a question of equity26
(8). Transportation equity has been conceptualized as improving accessibility to social and eco-27
nomic opportunities, especially for marginalized groups (9). Vertical equity is used as a concept28
in transport planning to promote the mobility of disadvantaged groups (10). As most transporta-29
tion interventions cause costs and benefits, it is crucial to analyze the differential social impacts30
of the 9-Euro-Ticket (11). A subsidized fare mainly addresses the price of public transit and not31
additional barriers faced by women, namely fear of harassment (12–14) and hate crimes and in-32
accessibility affecting individuals with a disability (15–20). Therefore, we expect those who are33
disadvantaged due to their gender or a disability to gain less mobility compared to economically34
marginalized persons for whom the cost of transportation is the main barrier (21–24).35
One challenge in quantifying the social impacts of nearly fare-free transit is that few stan-36
dardized methods exist, and insecurity remains which social impacts to include (25). In the past,37
most studies focused on analyzing social welfare benefits of transit subsidies (26,27). While some38
studies found low-income households to benefit most (28,29), others found most forms of subsi-39
dies catering more to higher income individuals (30,31). Most previous research takes into account40
the taxation source of the subsidy and the distribution of public resources for different modes of41
transit that are used by different income groups (32). Different transit pricing strategies, such as42
flat fares, distance-based fares or mode-dependent fares are also found to cater to different income43
groups (33–35).44
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 3
However, analyzing the effect of fare subsidies from a welfare perspective often only fo-1
cuses on direct monetary benefits. Furthermore, aggregated metrics can conceal relevant infor-2
mation (36). In our paper we want to expand the scope to include other benefits, such as activity3
participation and public transport usage, taking into account the literature on transportation equity4
and mobility justice (37–42). Furthermore, we do not only analyze the distribution of benefits5
according to economic status but also to gender and ability. When evaluating policies, conduct-6
ing randomized controlled trials to determine causal impacts is often infeasible because of ethical7
concerns (43). Randomized treatment assignment would imply in the case of the 9-Euro-Ticket,8
that only a randomly chosen group received the ticket. Relying on observational data, treatments9
are selected rather than assigned (44). In terms of the 9-Euro-Ticket, participants chose to buy the10
ticket or not. The participants will generally buy the ticket if the expected benefit is higher than11
the associated costs. Treatment and control groups can thus not be directly compared because we12
might assume that the two groups differ fundamentally in their baseline characteristics (45). Even13
in the absence of the 9-Euro-Ticket, the two groups would potentially have differing outcomes.14
In order to estimate causal effects in the presence of non-random treatment assignment we use a15
quasi-experimental approach, Propensity Score Matching. Propensity Score Matching has been16
used in other transportation policy contexts (46).17
This paper presents three contributions. Firstly, it extends beyond considering economic18
status and addresses other factors in the context of transport-deprivation such as gender and abil-19
ity (22). Secondly, instead of focusing solely on net monetary benefits, this study explores the20
effectiveness of subsidized fares in benefiting disadvantaged groups in a broader sense. The bene-21
fits evaluated in this study cover activity participation, public transport usage, and financial relief.22
Lastly, this research seeks to identify the causal effects of almost fare-free public transport using23
a causal inference method. The effects of the 9-Euro-Ticket are estimated using Propensity Score24
Matching, and weighted regression models. This doubly robust approach reduces the bias of the25
estimates compared with traditional models, such as Ordinary Least Squares (44).26
The paper is organized as follows. We first describe our data and methodology. Then we27
present the results from the weighted regression models. Finally, we discuss our findings and give28
an outlook on future research. The research design is displayed in Figure 1.29
DATA AND METHODOLOGY30
The data used is part of the Mobilität.Leben study with a total of 2,569 participants. More informa-31
tion on the study design can be found in earlier publications (47–49). 1650 participants (64.2%) are32
part of a study focusing on the Munich Metropolitan Region, the rest of the sample (919, 35,8%)33
were recruited nationally. The non-Munich sample is representative. The Munich Metropolitan34
Region comprises different types of spatial structure, covering both rural and metropolitan areas to35
represent different mobility behavior, furthermore all genders and ages are represented (47). Until36
July 2023 there have been six survey waves. We will draw on data from the first three waves,37
the timing of the distribution is depicted in Figure 2. The first survey was completed by 2,14138
participants and the second survey by 1,733 participants. Completion rates were higher in the na-39
tional sample than in the Munich sample. 117 observations were discarded as unreliable because40
of implausible completion times. Table 1 describes the variables used in the analysis.41
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 4
TABLE 1 : Variable Description
Variable Role Type Survey Question Answer
Person with a Disability Covariate Binary Are you limited by a health problem in activities of daily living? Yes, very limited/ Yes, somewhat lim-
ited
Ticket: Bad Idea Covariate Binary Please indicate your agreement with the statement: "The 9-Euro-
Ticket is a good idea!"
Completely disagree/ disagree
Ticket: Neutral Covariate Binary Please indicate your agreement with the statement: "The 9-Euro-
Ticket is a good idea!"
Neither agree nor disagree
Ticket: Good Idea Covariate Binary Please indicate your agreement with the statement: "The 9-Euro-
Ticket is a good idea!"
Completely agree/ agree
Access to Public Transport Covariate Continuous What public transport options are available within a 5-minute walk
of your home?
Number of selected options (bus, S-
Bahn, tram, U-Bahn)
Activity Level Covariate Continuous On average, how many days do you travel to different locations in
a week?
Sum of answers for work, leisure, and
errands
Mode of Transport: Often (May) Covariate Binary How often do you use the following modes of transport in a week? Daily/ 4-5 days a week
Mode of Transport: Sometimes (May) Covariate Binary How often do you use the following modes of transport in a week? 2-3 days a week/ Once a week/ Less
than once a week
Mode of Transport: Never (May) Covariate Binary How often do you use the following modes of transport in a week? Never
Economically Marginalized Person Covariate Binary How does your household cope with price increases? How much
does the following statement apply to you? "Because of the in-
creased prices, I have to forgo many things in my life." Which state-
ment about saving applies to your household in a typical month?
At least one of the following: Very bad,
Completely true, The household must
draw on savings or borrow money
Savings Outcome Binary How strongly do you agree with the following statement? "I can
spend the money saved from the 9-Euro-Ticket on more useful
things."
Completely agree/ Agree
Activity Level: Leisure (Jul/Sep) Outcome Continuous On average, how many days do you travel to different locations in
a week?
Answer for "leisure"
Activity Level: Errand (Jul/Sep) Outcome Continuous On average, how many days do you travel to different locations in
a week?
Answer for "errand"
Public Transport: Often (Jul/Sep) Outcome Binary How often do you use public transport in a week? Daily/ 4-5 days a week
Public Transport: Sometimes (Jul/Sep) Outcome Binary How often do you use public transport in a week? 2-3 days a week/ Once a week/ Less
than once a week
Public Transport: Never (Jul/Sep) Outcome Binary How often do you use public transport in a week? Never
9-Euro-Ticket Treatment Binary Did you purchase the 9-Euro-Ticket for the month of June/July? At least once "yes"
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 5
9-Euro-Ticket Users in
June and/or July 2022
(Treatment Group)
Propensity Score Matching
9-Euro-Ticket Non-Users in
June and July 2022
(Control Group)
Subgroup Analyses Based on Economic Status, Gender and Ability
Weig hted Regression Models
Outcome 1:
Activity Participation
Outcome 2:
PT Usage
Outcome 3:
Financial Relief
Users Without a Public Transport Pass in May 2022
FIGURE 1 : Research Design
Research Design1
To estimate the effects of the 9-Euro-Ticket, the treatment and the relevant population must be2
defined. The results of the second survey wave that was distributed at the end of July are used to3
compute the effects. Therefore, the treatment is defined as purchasing the 9-Euro-Ticket at least4
once in June and/or July. Participants who only bought the 9-Euro-Ticket in August were therefore5
not included in the treatment group. Furthermore, it must be considered that 543 participants (33.76
%) had already owned a transport subscription in May. While the 9-Euro-Ticket will lead to a7
cost reduction for this group, their habitual travel behavior will plausibly remain unaffected (50).8
Nearly all participants with a previous public transport subscription (96.3 %) received or bought9
the 9-Euro-Ticket. In terms of experimental designs, this group is pre-treated as they had already10
been subject to the treatment of having a public transport subscription. This is also in line with11
the policy objective of the 9-Euro-Ticket to cater primarily to new public transport users. Hence,12
the treatment effects are estimated for those without a public transport pass in May 2023 (N =13
1,067). All further statements will apply to this subset of the data. We conducted moderation14
analyses using subgroups based on economic status, gender, and ability. Economic status was15
determined by identifying individuals experiencing economic pressure, with the specific survey16
questions provided in table 1. Disability was defined broadly as limitations in activities of daily17
living due to health issues, in line with previous research (51). One participant with a diverse18
gender was included in the analysis and categorized under the female participants.19
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 6
Propensity Score Matching1
When estimating the effects of an intervention, one is interested in the different outcomes of an2
individual under each treatment state. However, researchers can only ever observe one outcome3
per individual, the other (potential) outcome remains counterfactual and exists only in theory.4
Therefore, causal effects cannot be estimated for individual units (44). Thus, experimental research5
designs are used to estimate treatment effects: Participants are randomly assigned to the treatment6
or the control group. As policies are usually directed at certain parts of the population, the Average7
Treatment Effect on the Treated (ATT) is of special interest.8
However, randomized controlled trials are often infeasible for policy evaluation because of9
ethical concerns (43). In the case of the 9-Euro-Ticket, it was not randomly assigned but partici-10
pants decided whether to purchase it taking into account the expected benefits and the associated11
costs. Therefore, treatment and control groups can not be directly compared because we might12
assume that the two groups differ fundamentally in their baseline characteristics (45). Even in13
the absence of the 9-Euro-Ticket, the two groups would potentially have differing outcomes. One14
possible approach for observational data is conditioning the sample on a set of variables (X) that15
predict treatment assignment. This implies that the potential outcomes of treatment and control16
group are independent of the treatment assignment (D) given their observed characteristics (X)17
(52).18
Y0,Y1⊥⊥ D|X(1)
If this assumption holds, a robust ATT estimate can be calculated in the presence of a non-randomly19
assigned treatment.20
E[δ|D=1] = E[Y1|D=1,X]−E[Y0|D=1,X](2)
Conditional on X, there are no systematic differences between treatment and control group. X21
is thus a straightforward balancing score b(x) that is specified so that the conditional distribution22
of X given b(x) does not differ between the treatment and the control group (45). However, due23
to the "curse of dimensionality", it is oftentimes not feasible to match units from treatment and24
control group on all covariates contained in X. Rosenbaum showed that the propensity score can25
be used as a balancing score (45). The propensity score is the estimated probability of taking the26
treatment, modeled as a function of covariates predicting the treatment assignment. The true form27
of the propensity score is unknown when working with observational data. Therefore, propensity28
score estimations are used (44).29
The Propensity Score Model30
When selecting the covariates to be included in the matching process, the goal is to satisfy the31
assumption of "strong ignorability", which means that conditional on the observed covariates, there32
are no unobserved differences between the control and treatment group. There is little cost in33
including unnecessary variables, i.e. variables that are unrelated to the treatment assignment. They34
may slightly increase the variance of the model (53). In contrast, omitting relevant confounding35
variables will significantly increase bias. It is therefore advisable to include all variables that could36
influence treatment assignment and/or outcome (53). This is also true for the 9-Euro-Ticket, where37
several factors influence the decision to purchase the 9-Euro-Ticket and the outcomes. Table 138
shows the relevant dependent variables (outcome and treatment) when choosing which variables39
to include in the propensity score model.40
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 7
Vincent Kaufmann’s concept of motility provides a framework for considering the space in1
which mobility decisions are situated (54). The categories provided in the motility framework are2
taken into account when selecting the variables for matching. Table 2 presents an overview of the3
categories and their corresponding variables in the propensity score model. It should be noted that4
the categories are interdependent and therefore variables could potentially fit into different cate-5
gories. For example, the socioeconomic variables included may affect all three categories. Because6
the data can only be matched on observed characteristics, some variables such as "knowledge of7
PT" or "preference" are proxied by past behavior.8
TABLE 2 : Motility and Variables used for PSM
Motility Categories Variables Included in the Propensity
Score
Variable Name in the Model
Access
• Options (Transportation, Ser-
vices)
• Conditions (Costs, Logistics,
Constraints)
• Public Transport Access, Re-
giostar Classification (German
Classification of Regional Type)
• Other Socio-Economic Variables
(Age, Gender, Economic Status,
Employment, Household Size,
Children)
• PTAccess, RegiostarClassifica-
tion
• Age, Gender, EconomicStatus,
Employment, HouseholdSize,
Children
Competence
• Physical Ability
• Acquired Skills
• Organizational Skills
• Disability
• Driving License
• Knowledge about PT (e.g., About
Schedules; Proxied by Experi-
ence Through Past Use)
• Disability
• License
• ModeUsage
Appropriation
• Needs
• Plans
• Aspirations
• Understandings
• Attitude towards Climate Change,
Political Attitudes, Attitude to-
wards the 9-Euro-Ticket
• Preference (Proxied by Previous
Activity Participation and Mode
Use)
• AttitudeClimateChange, Politica-
lAttitude, AttitudeTicket
• ActivityLevel
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 8
The propensity score b(x) is specified in the following way:
b(x) = Pr[D=1|X] = (3)
β0+β1PTAccessi+β2RegiostarClassificationi+β3Agei+β4Genderi+β5EconomicStatusi
+β6Employmenti+β7HouseholdSizei+β8Childreni+β9Disabilityi+β10Licensei
+β11ModeUsagei+β12AttitudeClimateChangei+β13PoliticalAttitudei+β14AttitudeTicketi
+β15ActivityLeveli+εi
Figure 2 shows the propensity score distribution depending on ticket purchase, a logit1
model was used. The "common support" assumption implies that the treatment and control groups2
overlap substantially in their propensity score distribution. The density of the distribution may dif-3
fer (53). As shown in Figure 2, the range of propensity scores is similar between the two groups.4
In the group without the ticket, lower propensity scores are estimated; in the group with the 9-5
Euro-Ticket, more propensity score estimates are closer to one. No estimates are exactly 0 or 1,6
allowing the propensity score to be used for matching.7
No ticket
Ticket in June and/or July
0.25 0.50 0.75 1.00 0.25 0.50 0.75 1.00
0
25
50
75
b(x)
count
FIGURE 2 : Distribution of the Propensity Score b(x)
Matching8
Testing various matching methods, a combined propensity score with exact matching on economic9
status, ability and gender (55) using optimal full matching resulted in the most balanced matched10
sample. The R matchIt package was used for matching (56). Figure 3 shows the effectiveness of11
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 9
PSM in reducing covariate imbalance between the control and treatment group. The commonly1
used threshold for the standardized mean difference of 0.1 is displayed. For almost all variables,2
matching reduced the standardized mean difference between treatment and control group, although3
perfect balance was not achieved. The sample size after matching is 567 in the treatment group4
(ESS: 69.66) and 260 in the control group. Since full matching was used no observations were5
discarded. As logistic regression requires complete observations, the sample available for matching6
decreases in size compared to the original sample.7
Weighted Regression Models8
Propensity Score Matching does not estimate effects by itself, but must be combined with other9
models such as linear regression (53). Regression after matching can further reduce bias due10
to remaining imbalances in the matched data (44). Weighted regression models provide less bi-11
ased estimates by accounting for individual-level heterogeneity between the treatment and control12
group. The weights used are propensity score estimates (44). Because the weighted regression es-13
timates condition on the covariates twice (both in the matching process and in the regression), the14
results are said to be doubly robust (44). All treatment effects were estimated using g-computation15
and cluster-robust standard errors with the marginaleffects R package (57). The following model16
specifications were used to estimate the ATT in weighted regression models using linear models.17
Outcome =
β0+β1Treatment +β2PropensityScore +β3Covariates +β4(Treatment ×Covariates)+
β5(Treatment ×PropensityScore) + εi
(4)
The propensity score was added to the regression model to increase robustness (58). In addition,18
to minimize the impact of any remaining imbalances in the matched data, all covariates used for19
matching were included as controls in the model, including the interaction effects with the treat-20
ment variable. Subgroup effects were estimated based on gender, economic status, and ability.21
Some survey questions were only addressed to those who purchased a ticket. For these outcome22
variables, logistic regression models were fitted to the subset of the treatment group, including all23
covariates used for matching as controls.24
RESULTS25
In this section, we will present our results using weighted regression models to estimate the ATT26
and logistic regression models for outcomes concerning only the ticket users.27
Activity Participation28
Two models were estimated with the average number of days per week participants participated in29
leisure activities and ran errands as outcomes. The results are presented in Table 3. The results30
indicate a significant positive effect of the 9-Euro-Ticket on participation in leisure and errand ac-31
tivities for the entire sample. The moderation analysis shows that the effect varied across subgroups32
depending on the activity. The 9-Euro-Ticket had a significant positive effect on leisure activities33
for men, those not economically marginalized, and those without disabilities with effect sizes up34
to 0.54. This indicates, that the 9-Euro-Ticket led some groups to participate in leisure activities35
on one additional day every two weeks. No significant effects were observed for the other groups.36
Women, the economically marginalized and those without disabilities experienced an increase in37
the number of days running errands. The results suggest no increase in activity participation for38
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 10
Bike: Sometimes
Carsharing: Often
Children
Regiostar 74
Household size: 2
Household size: more than two
Household size: 1
Regiostar 73
Studying
Working
Bikesharing: Often
Activity Level²
Activity Level
Male
Unemployed
Studying and Working
Regiostar 75
Regiostar 72
Driving License
Economically Marginalized
Regiostar 76
Ticket: Bad Idea
Person with Disability
Age²
Walk: Never
Walk: sometimes
Age
Ticket: Neutral
Car: Sometimes
PT: Often
Regiostar 77
Bike: Never
Politically Right Leaning²
Concerned about Climate Change
Bike: Often
Walk: Often
Concerned about Climate Change²
Politically Right Leaning
Ticket: Good Idea
Bikesharing: Sometimes
Bikesharing: Never
Car: Often
Access PT
Carsharing: Never
PT: Sometimes
Car: Never
Carsharing: Sometimes
Access PT²
Regiostar 71
PT: Never
0 0.1 0.25 0.5 0.75 1
Absolute Standardized Mean
Differences
Sample Original Optimal Full Matching Combined with Exact Matching
Covariate Balance
FIGURE 3 : Covariate Balance in the Original Sample and the Matched Sample.
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 11
people with disabilities in any category.1
TABLE 3 : Effect of the 9-Euro-Ticket on Activity Participation
ATT: Avg. Days with Leisure Activities ATT: Avg. Days with Errand Activities
All 0.34∗∗ 0.30∗
(0.14) (0.16)
Female 0.20 0.84∗∗∗
(0.25) (0.27)
Male 0.44∗∗∗ −0.03
(0.17) (0.20)
Economically Marginalized −0.35 0.99∗∗∗
(0.40) (0.33)
Not Economically Marginalized 0.52∗∗∗ 0.13
(0.15) (0.18)
Person with a Disability −0.31 −0.53
(0.34) (0.32)
Person without a Disability 0.54∗∗∗ 0.56∗∗∗
(0.15) (0.18)
Standard errors in parentheses
∗p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01
The difference in leisure activity participation is not significant between the genders.
The difference in leisure activity participation is significant between (non) economically marginalized persons on the 0.05 level.
The difference in leisure activity participation is significant between persons with a disability and those without on the 0.05 level.
The difference in errand activity participation is significant between the genders on the 0.01 level.
The difference in errand activity participation between (non) economically marginalized persons is significant on the 0.05 level.
The difference in errand activity participation is significant between persons with a disability and those without on the 0.01 level.
In addition, the survey asked participants whether they participated in more activities be-2
cause of the 9-Euro-Ticket. Since this question is only relevant for the ticket holders, a logistic3
regression model was fitted to the subset of the treatment group. Table 4 shows the significant re-4
sults of the analysis. Economic status was the only identity marker influencing the binary outcome,5
with economically marginalized individuals more likely to report increased activity participation.6
Use of Public Transport7
The results displayed in table 5 suggest that the 9-Euro-Ticket had a positive effect on the proba-8
bility of using public transport often and led to a decrease in the probability of never using public9
transport, both for the entire sample and for all subgroups. The ticket seemed to be most effective10
in reducing the probability of never using public transport, with the treatment effect being a reduc-11
tion of 49 percentage points for the whole treated sample. However, the effect size varied across12
subgroups, with people with disabilities experiencing a significantly smaller reduction than those13
without disabilities. Furthermore, the 9-Euro-Ticket impacted the use of public transport after its14
validity period for certain subgroups. Men, economically marginalized people and people without15
disabilities were more likely to use public transport often in September if they had bought the 9-16
Euro-Ticket. However, the effect size was small and no effects were found for the other groups.17
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 12
TABLE 4 : Logistic Regression: More Activities Because of the 9-Euro-Ticket (Excerpt)
Dependent variable:
More Activities
Working −0.47∗
(0.27)
Ticket: Good Idea 0.91∗∗
(0.36)
Male 0.02
(0.19)
Economically Marginalized 0.66∗∗
(0.31)
Person with a Disability −0.14
(0.29)
Constant −1.09
(2.32)
Observations 566
Log Likelihood −359.83
Akaike Inf. Crit. 791.65
McFadden’s Pseudo-R20.06
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
The 9-Euro-Ticket also reduced the probability of never using public transport after its validity1
period for all groups except the economically marginalized. Overall, the results indicate that the2
9-Euro-Ticket had a limited effect on public transport usage after its validity period.3
Financial Relief4
The effect of the 9-Euro-Ticket on financial relief was also studied using two models. The first5
model estimated the treatment effect on participants’ agreement with the statement "I can spend6
the money saved by the 9-Euro-Ticket on more useful things". This statement was presented7
alongside other hypothetical statements for participants to indicate their agreement. Table 6 shows8
the results, which indicate that the ticket had a positive effect on the agreement rates for most of the9
treated subgroups. However, no significant effect was observed among economically marginalized10
individuals who purchased the ticket, significantly differing from the estimate for individuals with11
a higher economic status.12
The survey also included a question asking participants whether they could benefit finan-13
cially from the 9-Euro-Ticket. To analyze this question, a logistic model was fitted to the treatment14
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 13
TABLE 5 : Effect of the 9-Euro-Ticket on Public Transport Usage
ATT: Public Transport Often (During) ATT: Public Transport Never (During) ATT: Public Transport Often (After) ATT: Public Transport Never (After)
All 0.10∗ ∗ ∗ −0.49∗∗∗ 0.02∗∗ ∗ −0.25∗∗∗
(0.02) (0.05) (0.01) (0.04)
Female 0.08∗ ∗ ∗ −0.56∗∗∗ 0.01 −0.25∗∗∗
(0.03) (0.06) (0.02) (0.06)
Male 0.11∗ ∗ ∗ −0.44∗∗∗ 0.04∗∗ ∗ −0.24∗∗∗
(0.03) (0.06) (0.01) (0.06)
Economically Marginalized 0.15∗∗ ∗ −0.30∗∗ 0.08∗ ∗ ∗ −0.04
(0.04) (0.12) (0.02) (0.10)
Not Economically Marginalized 0.09∗ ∗ ∗ −0.54∗∗∗ 0.01 −0.30∗∗∗
(0.02) (0.05) (0.01) (0.05)
Person with a Disability 0.10∗ ∗ ∗ −0.31∗∗∗ 0.00 −0.35∗∗∗
(0.04) (0.11) (0.03) (0.08)
Person without a Disability 0.10∗ ∗ ∗ −0.55∗∗∗ 0.03∗∗ ∗ −0.21∗∗∗
(0.02) (0.06) (0.01) (0.05)
Standard errors in parentheses
*p<0.10, ** p<0.05, *** p<0.01
In the first column, there are no significant differences in the subgroup effects.
In the second column, the difference in effect estimates is significant between persons with a disability and those without on the 0.10 level; the difference in effect estimates between economic status is significant on the 0.10 level.
In the third column, the difference in effect estimates is significant between (not) economically marginalized persons on the 0.01 level; the difference in effect estimates between the genders is significant on the 0.05 level.
In the fourth column, the difference in effect estimates is significant between (not) economically marginalized persons on the 0.05 level.
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 14
TABLE 6 : Effect of the 9-Euro-Ticket on Agreement to the Statement: "I can spend the money
saved by the 9-Euro-Ticket on more useful things"
ATT: Agreement to the Statement
All 0.13∗∗∗
(0.05)
Female 0.18∗∗
(0.07)
Male 0.10∗
(0.06)
Economically Marginalized −0.16
(0.10)
Not Economically Marginalized 0.20∗∗∗
(0.05)
Person with a Disability 0.09
(0.12)
Person without a Disability 0.14∗∗∗
(0.05)
Standard errors in parentheses
∗p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01
The difference in effect estimates between (not) economically marginalized persons is significant on the 0.01 level.
For the other groups, there are no significant differences.
group subset, including all covariates used for matching. Table 7 summarizes the regression re-1
sults, displaying the significant variables and the subgroup characteristics. The results suggest that2
gender, economic status, and disability did not significantly influence the outcome.3
DISCUSSION4
Overall, our results suggest that the effect of the 9-Euro-Ticket varied across indicators and sub-5
groups. Our main findings include:6
7
• increased public transport use for all groups during the ticket’s validity period8
• an increase in activity participation especially for economically marginalized individuals9
• no targeted financial relief for economically marginalized individuals10
• less benefits for women and individuals with a disability11
Economic Status12
A logistic regression showed that the economic status is a significant factor in explaining whether13
9-Euro-Ticket users could participate in more activities because of the 9-Euro-Ticket. This finding14
supports that the cost of mobility presents a major barrier for economically marginalized individ-15
uals and that a discounted price can improve their mobility. The effects on transport use after the16
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 15
TABLE 7 : Logistic Regression: Financial Benefit Because of the 9-Euro-Ticket (Excerpt)
Dependent variable:
Financial Benefit
Age −0.02∗
(0.01)
Regiostar 72 −0.70∗
(0.43)
Driving License 1.19∗∗
(0.51)
Concerned about Climate Change 0.12∗
(0.06)
Car: Sometimes 0.77∗∗∗
(0.29)
Car: Never 1.18∗∗∗
(0.46)
Male −0.11
(0.23)
Economically Marginalized −0.46
(0.36)
Person with a Disability −0.19
(0.33)
Constant 13.32
(613.79)
Observations 566
Log Likelihood −271.59
Akaike Inf. Crit. 615.17
McFadden’s Pseudo-R20.10
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 16
validity period support this interpretation. While all other groups reduced the probability of never1
using public transport after the intervention, there was no significant effect for the economically2
marginalized. At the same time, the probability of using public transport often after the 9-Euro-3
Ticket increased. This suggests that some economically marginalized people may have continued4
to use public transport because it has allowed them to participate in activities they could not be-5
fore. However, a certain proportion of this group could only afford to use public transport at a6
reduced fare and has become restricted again in their mobility. The data suggest that economically7
marginalized individuals used the ticket primarily for errands. For this group, the new mobility8
offered by the 9-Euro-Ticket may have been used mainly for essential daily tasks before increasing9
leisure activities (59). Leisure activities may also be less accessible for economically marginalized10
individuals due to the additional costs associated with them, such as entrance or participation fees.11
The evidence on whether economically marginalized individuals also benefited financially12
is mixed. The treatment effect of the 9-Euro-Ticket on agreement with the statement "I can spend13
the money saved by the 9-Euro-Ticket on more useful things" was not significant for this group.14
This may be because economically marginalized individuals had used public transport less fre-15
quently before the 9-Euro-Ticket. The ticket allowed them to be more mobile but did not lead to16
savings. Also, economic status was insignificant in a logistic regression predicting the financial17
benefit of the 9-Euro-Ticket. Collectively, these findings indicate that individuals facing economic18
marginalization did not receive focused financial relief.19
Disability20
In contrast to economically marginalized individuals, people with a disability had a significantly21
smaller reduction in the probability of never using public transport than people without a disability.22
This may suggest that additional barriers related to accessibility or fear of victimization prevented23
people with a disability from switching to public transport. Also, individuals with disabilities did24
not experience an increase in activity participation across all activity categories. Given the concept25
of the "accessible trip chain" (17), barriers may be associated with both transportation and the26
activity itself. Since the data show that people with disabilities were more likely to use public27
transportation with the 9-Euro-Ticket, this finding suggests that the challenges to participating in28
the activity may be due to the accessibility of the activity or other limitations. This would imply29
that people with disabilities used the 9-Euro-Ticket to reach their usual destinations, but did not30
increase their overall level of activity.31
Gender32
The effect of the ticket on women was also mixed. In contrast to men, they were not more likely33
to use public transport more often after the intervention. This finding suggests that while they may34
have tried to use public transport during the ticket’s validity period, they were not convinced to35
make a more permanent switch, possibly due to other barriers they encountered when using public36
transport. In terms of activity participation, women used the ticket primarily for errands, while37
men used it more for leisure. Traditional gender roles may explain this pattern. Women typically38
spend more time on housework than men (60).39
Limitations40
There are several limitations of this study. The (effective) sample size is relatively small and41
conclusive survey weights have not yet been calculated. Therefore, it is uncertain whether the42
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 17
results are representative of the entire German population. Due to the sample size, the moderation1
analysis was also limited to gender, ability, and economic status as monolithic categories. It is2
plausible that the effect might further differ, for example, between women with a disability and men3
with a disability, given the intersectional nature of identity categories (61). More subgroups based4
on different identity markers could be created and analyzed in a larger sample. In addition, because5
PSM relies on observed variables, it may be subject to omitted variable bias. Another limitation is6
that the dependent variables were based on self-reported behavior, which may be subject to bias.7
Participants may have over-reported their use of public transportation during the validity period of8
the 9-Euro-Ticket because they wanted to demonstrate a desired behavior or a psychological wish9
to justify their investment. In addition, activity participation was measured as the number of days10
per week that participants participated in a specific activity, which may not capture multi-purpose11
trips. This could particularly affect the treatment effect for women, as they are reported to use trip12
chaining (62). It is possible that the 9-Euro-Ticket led to increased participation in activities, but13
these activities were bundled rather than spread over several days. Trip chaining would then lead14
to an underestimation of the increase in activity due to the 9-Euro-Ticket.15
CONCLUSION16
The 9-Euro-Ticket was a German nationwide policy initiative that introduced an almost fare-free17
public transport system for a period of three months. Using Propensity Score Matching and18
weighted regression models on study participants without previous public transport subscriptions,19
we showed that economically marginalized people benefited most from the almost fare-free transit.20
In contrast, the effects on women and persons with disabilities were mixed. The results suggest21
that there was no targeted financial relief for the economically marginalized. Overall, therefore,22
the policy seemed to have been effective in addressing the most important barrier for economically23
marginalized persons: the cost of public transport. It also motivated some individuals to continue24
using public transport after the intervention. These results can be valuable for other countries25
that seek to increase public transport ridership and improve access for individuals with a low eco-26
nomic status, especially in the presence of an otherwise accessible public transport service. Despite27
these impressive effects, the 9-Euro-Ticket was not a cure-all. Structural barriers for marginalized28
groups, such as victimization or lack of accessibility, still exist. To fulfill every citizen’s basic29
mobility needs is in line with the understanding of public transit as a public service and normative30
conceptions of transport equity and mobility justice. This justifies government subsidies to make31
public transport more accessible. Additional policy instruments are needed to implement a public32
transport system that enables public transport use and activity participation for all while providing33
targeted financial relief to economically marginalized individuals. While this research explored34
the effectiveness of the ticket, in the next steps we will include the costs in a comprehensive policy35
evaluation in the form of a cost-benefit-analysis or cost-effectiveness-analysis. Furthermore, we36
plan to analyze the effects of the recently introduced 49-Euro-Ticket to explore the impact of a less37
subsidized fare with the data we collected as part of our study.38
AUTHOR CONTRIBUTIONS39
The authors confirm contribution to the paper as follows: study conception and design: Isabella40
Waldorf, Allister Loder, Stefan Wurster, Klaus Bogenberger; data collection: Allister Loder; anal-41
ysis and interpretation of results: Isabella Waldorf, Allister Loder; draft manuscript preparation:42
Isabella Waldorf. All authors reviewed the results and approved the final version of the manuscript.43
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 18
ACKNOWLEDGEMENTS1
The authors would like to thank the TUM Think Tank at the Munich School of Politics and Public2
Policy for its support. Allister Loder acknowledges funding by the Bavarian State Ministry of3
Science and the Arts in the framework of the bidt Graduate Center for Postdocs. Isabella Waldorf4
would like to thank Tobias Rommel, David Dúran and Philipp Mennig for their helpful discussions.5
GPT-3 assisted in debugging the R code, creating LaTeX tables, and making language edits.6
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 19
References1
1. Štraub, D. and V. Jaroš, Free fare policy as a tool for sustainable development of public trans-2
port services. Human Geographies, Vol. 13, No. 1, 2019, pp. 45–59.3
2. Die Bundesregierung, Fragen und Antworten zum 9-Euro-Ticket ab Juni 2022, 2022.4
3. VDV - Die Verkehrsunternehmen, Bilanz 9-Euro-Ticket, 2022.5
4. Rouwendal, J. and E. T. Verhoef, Basic economic principles of road pricing: From theory to6
applications. Transport Policy, Vol. 13, No. 2, 2006, pp. 106–114.7
5. Ubbels, B., P. Nijkamp, E. Verhoef, S. Potter, and M. Enoch, Alternative Ways of Funding8
Public Transport: A Case Study Assessment. European Journal of Transport and Infrastruc-9
ture Research, Vol. 1, No. 1, 2001, p. 73–89.10
6. Regionalisierungsgesetz § 1 Öffentliche Aufgabe, Zuständigkeit: RegG, 2022.11
7. Wellman, G. C., Transportation exploitation, mobility and social justice. In Mobilities, Mobil-12
ity Justice and Social Justice (N. Cook and D. Butz, eds.), Routledge, London and New York,13
2020, pp. 67–80.14
8. Wee, B. v., B. v. Wee, and K. Geurs, Discussing Equity and Social Exclusion in Accessibility15
Evaluations. European Journal of Transport and Infrastructure Research, Vol. 11, No. 4, 2011.16
9. Pereira, R. H. M., T. Schwanen, and D. Banister, Distributive justice and equity in transporta-17
tion. Transport Reviews, Vol. 37, No. 2, 2017, pp. 170–191.18
10. Camporeale, R., L. Caggiani, A. Fonzone, and M. Ottomanelli, Quantifying the impacts of19
horizontal and vertical equity in transit route planning. Transportation Planning and Technol-20
ogy, Vol. 40, No. 1, 2017, pp. 28–44.21
11. Haas, T., The political economy of mobility justice. Experiences from Germany. Mobilities,22
Vol. 17, No. 6, 2021, pp. 1–15.23
12. Chowdhury, S. and B. van Wee, Examining women’s perception of safety during waiting times24
at public transport terminals. Transport Policy, Vol. 94, 2020, pp. 102–108.25
13. Dunckel Graglia, A., Finding mobility: women negotiating fear and violence in Mexico City’s26
public transit system. Gender, Place & Culture, Vol. 23, No. 5, 2016, pp. 624–640.27
14. Loukaitou-Sideris, A., Fear and safety in transit environments from the women’s perspective.28
Security Journal, Vol. 27, No. 2, 2014, pp. 242–256.29
15. Imrie, R., Disability and Discourses of Mobility and Movement. Environment and Planning30
A: Economy and Space, Vol. 32, No. 9, 2000, pp. 1641–1656.31
16. Unsworth, C., M. H. So, J. Chua, P. Gudimetla, and A. Naweed, A systematic review of public32
transport accessibility for people using mobility devices. Disability and rehabilitation, Vol. 43,33
No. 16, 2021, pp. 2253–2267.34
17. Park, J. and S. Chowdhury, Investigating the barriers in a typical journey by public transport35
users with disabilities: Manuscript. Journal of Transport & Health, Vol. 10, 2018, pp. 1–12.36
18. Gallagher, B. A. M., P. M. Hart, C. O’Brien, M. R. Stevenson, and A. J. Jackson, Mobility and37
access to transport issues as experienced by people with vision impairment living in urban and38
rural Ireland. Disability and rehabilitation, Vol. 33, No. 12, 2011, pp. 979–988.39
19. Wilkin, D., Disability Hate Crime: Experiences of Everyday Hostility on Public Transport.40
Palgrave Hate Studies, Springer International Publishing AG, Cham, 2020.41
20. B˛eczkowska, S. A. and Z. Zysk, Safety of People with Special Needs in Public Transport.42
Sustainability, Vol. 13, No. 19, 2021, p. 10733.43
21. Aberle, C., S. Daubitz, C. Gertz, and O. Schweder, Mobilitätsbezogene soziale Exklusion in44
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 20
Großstädten: Empirische Bestandsaufnahme und Strategieentwicklung für Berlin und Ham-1
burg. Journal für Mobilität und Verkehr, , No. 14, 2022, pp. 18–27.2
22. Clifton, K. and K. Lucas, Examining the empirical evidence of transport inequality in the US3
and UK. In Running on empty (K. Lucas, ed.), Policy Press, Bristol, UK, 2004, pp. 15–36.4
23. Olvera, L. D., D. Mignot, and C. Paulo, Daily Mobility and Inequality: The Situation of the5
Poor. Built Environment, Vol. 30, No. 2, 2004, pp. 153–160.6
24. Purwanto, J., Does a Rise in Income Inequality Lead to Rises in Transportation Inequality and7
Mobility Practice Inequality? Social Inclusion, Vol. 4, No. 3, 2016, pp. 110–132.8
25. Cadena, P. C. B., J. M. Vassallo, I. Herraiz, and M. Loro, Social and Distributional Effects of9
Public Transport Fares and Subsidy Policies: Case of Madrid, Spain. Transportation Research10
Record: Journal of the Transportation Research Board, Vol. 2544, No. 1, 2016, pp. 47–54.11
26. Savage, I. and A. Schupp, Evaluating Transit Subsidies in Chicago. Journal of Public Trans-12
portation, Vol. 1, No. 2, 1997, pp. 93–118.13
27. Parry, I. W. H. and K. A. Small, Should Urban Transit Subsidies Be Reduced? American14
Economic Review, Vol. 99, No. 3, 2009, pp. 700–724.15
28. Guria, J. C. and A. E. A. Gollin, Net tax incidence for urban public transit subsidies in New16
Zealand. Transportation, Vol. 13, No. 4, 1986, pp. 319–328.17
29. Basso, L. J. and H. E. Silva, Efficiency and Substitutability of Transit Subsidies and Other18
Urban Transport Policies. American Economic Journal: Economic Policy, Vol. 6, No. 4, 2014,19
pp. 1–33.20
30. Frankena, M., Income Distributional Effects of Urban Transit Subsidies. Journal of Transport21
Economics and Policy, Vol. 7, No. 3, 1973, pp. 215–230.22
31. Cropper, M. and S. Bhattacharya, Public Transport Subsidies and Affordability in Mumbai,23
India. Urban Studies Research, Vol. 2012, 2012, pp. 1–7.24
32. Pucher, J., Equity in Transit Finance: Distribution of Transit Subsidy Benefits and Costs25
Among Income Classes. Journal of the American Planning Association, Vol. 47, No. 4, 1981,26
pp. 387–407.27
33. Brown, A. E., Fair fares? How flat and variable fares affect transit equity in Los Angeles. Case28
Studies on Transport Policy, Vol. 6, No. 4, 2018, pp. 765–773.29
34. Nuworsoo, C., A. Golub, and E. Deakin, Analyzing equity impacts of transit fare changes:30
Case study of Alameda-Contra Costa Transit, California. Evaluation and program planning,31
Vol. 32, No. 4, 2009, pp. 360–368.32
35. Zhao, P. and Y. Zhang, The effects of metro fare increase on transport equity: New evidence33
from Beijing. Transport Policy, Vol. 74, 2019, pp. 73–83.34
36. Carleton, P. R. and J. D. Porter, A comparative analysis of the challenges in measuring transit35
equity: definitions, interpretations, and limitations. Journal of Transport Geography, Vol. 72,36
2018, pp. 64–75.37
37. Martens, K., Transport Justice: Designing fair transportation systems. Routledge, New York,38
N.Y., 2017.39
38. Elliott, A. and J. Urry, Mobile lives: Self, excess and nature. International Library of Sociol-40
ogy, Routledge, New York, NY, 2010.41
39. Soja, E. W., Seeking Spatial Justice, Vol. 16 of Globalization and Community series. Univer-42
sity of Minnesota Press, Minneapolis, 2010.43
40. Allen, J. and S. Farber, Planning transport for social inclusion: An accessibility-activity par-44
ticipation approach: Manuscript Preprint Version. Transportation Research Part D: Transport45
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 21
and Environment, Vol. 78, 2020, pp. 1–29.1
41. Cook, N. and D. Butz, Moving toward mobility justice. In Mobilities, mobility justice and2
social justice (N. Cook and D. Butz, eds.), Routledge, London and New York, 2020, pp. 3–22.3
42. Sheller, M., Mobility justice: The politics of movement in the age of extremes. Verso, London4
and Brooklyn, NY, 2018.5
43. Mennig, P. and J. Sauer, The impact of agri-environment schemes on farm productivity: a6
DID-matching approach. European Review of Agricultural Economics, Vol. 47, No. 3, 2020,7
pp. 1045–1093.8
44. Morgan, S. L. and C. Winship, Counterfactuals and causal inference: Methods and principles9
for social research. Analytical methods for social research, Cambridge University Press, New10
York, NY, USA, second edition, reprinted with corrections ed., 2015.11
45. Rosenbaum, P. R. and D. B. Rubin, The central role of the propensity score in observational12
studies for causal effects. Biometrika, Vol. 70, No. 1, 1983, pp. 41–55.13
46. Zhang, N., D. J. Graham, D. Hörcher, and P. Bansal, A causal inference approach to measure14
the vulnerability of urban metro systems. Transportation, Vol. 48, No. 6, 2021, pp. 3269–3300.15
47. Loder, A., F. Cantner, L. Adenaw, M. Siewert, S. Goerg, M. Lienkamp, and K. Bogenberger,16
A nation-wide experiment: fuel tax cuts and almost free public transport for three months in17
Germany – Report 1 Study design, recruiting and participation, 2022.18
48. Cantner, F., N. Nachtigall, L. S. Hamm, A. Cadavid, L. Adenaw, A. Loder, M. B. Siewert,19
S. Goerg, M. Lienkamp, and K. Bogenberger, A nation-wide experiment: fuel tax cuts and20
almost free public transport for three months in Germany – Report 2 First wave results, 2022.21
49. Loder, A., F. Cantner, A. Cadavid, M. B. Siewert, S. Wurster, S. Goerg, and K. Bogenberger,22
A nation-wide experiment: fuel tax cuts and almost free public transport for three months in23
Germany – Report 3 Second wave results, 2022.24
50. Gaus, D., N. Murray, and H. Link, 9-Euro-Ticket: Niedrigere Preise allein stärken Alltagsmo-25
bilität mit öffentlichen Verkehrsmitteln nicht, 2023.26
51. Flamm, M. and V. Kaufmann, Operationalising the Concept of Motility: A Qualitative Study.27
Mobilities, Vol. 1, No. 2, 2006, pp. 167–189.28
52. Rubin, D. B., Assignment to Treatment Group on the Basis of a Covariate. Journal of Educa-29
tional Statistics, Vol. 2, No. 1, 1977, pp. 1–26.30
53. Stuart, E. A., Matching methods for causal inference: A review and a look forward. Statistical31
science : a review journal of the Institute of Mathematical Statistics, Vol. 25, No. 1, 2010, pp.32
1–21.33
54. Kaufmann, V., M. M. Bergman, and D. Joye, Motility: mobility as capital. International Jour-34
nal of Urban and Regional Research, Vol. 28, No. 4, 2004, pp. 745–756.35
55. Green, K. M. and E. A. Stuart, Examining moderation analyses in propensity score methods:36
application to depression and substance use: Author’s manuscript. Journal of consulting and37
clinical psychology, Vol. 82, No. 5, 2014, pp. 1–22.38
56. Greifer, N., Matching Methods, 2022.39
57. Arel-Bundock, V., marginaleffects: Predictions, Comparisons, Slopes, Marginal Means, and40
Hypothesis Tests, 2023, r package version 0.12.0.9017.41
58. Austin, P. C., Double propensity-score adjustment: A solution to design bias or bias due to42
incomplete matching. Statistical methods in medical research, Vol. 26, No. 1, 2017, pp. 201–43
222.44
59. Hille, C. and M. Gather, Ergebnisbericht: „Das 9-Euro-Ticket hat mir gezeigt, dass man nicht45
TRB Annual Meeting 2024 Initial Paper Submittal
Waldorf et al. 22
alleine sein muss.“ – Mit dem 9-Euro-Ticket zu mehr sozialer Teilhabe? Ergebnisse einer1
Befragung von einkommensschwachen Haushalten zur Wirkung des 9-Euro-Tickets auf das2
Mobilitätsverhalten und ausgewählte Dimensionen der sozialen Teilhabe, 2022.3
60. Lachance-Grzela, M. and G. Bouchard, Why Do Women Do the Lion’s Share of Housework?4
A Decade of Research. Sex Roles, Vol. 63, No. 11-12, 2010, pp. 767–780.5
61. Crenshaw, K., On intersectionality: Essential writings. The New Press, New York, 2014.6
62. Levy, C., Travel choice reframed: “deep distribution” and gender in urban transport. Environ-7
ment and Urbanization, Vol. 25, No. 1, 2013, pp. 47–63.8
TRB Annual Meeting 2024 Initial Paper Submittal
