MOIA’s 9-euro ticket dilemma

MOIA Engineering
10 min readAug 26, 2022


by: Tim Sadler, Christof Pfundstein and Marc Dziakowski

At MOIA, we strongly believe that the future of mobility will be multimodal and independent of private cars. That is why we are working toward a world where people can reach their destination using a variety of alternatives, such as their own or rented bicycles, scooters, trams, trains, buses and, if a car is needed, car sharing, taxis and, of course our own service, ride-pooling. Together, these services are working to challenge the supremacy of the single-owner car.

When in spring 2022 the federal government of Germany introduced a ticket that would allow unlimited use of local and regional public transport for a flat fee of 9 Euro per month, we at MOIA were in favor of this government support for public transport. The public also welcomed the ticket widely, as it meant a drastic reduction in public transport fees and a refund for season ticket holders. Additionally, the ticket’s validity is not limited to a region, but it is valid all over Germany in local and regional public transport.

A study conducted by HVV, Hamburg’s public transportation provider, from beginning of August revealed that about half of the respondents have been in possession of a 9-euro ticket. Given the assumption that there is a high overlap between HVV customers and MOIA customers, we can assume that also a high number of MOIA customers are in possession of the ticket.

The ticket will end at the end of August. While there are discussion whether a similar offer should continue in the future, the government has not announced official plans. Besides these discussions and others, regarding crowded trains or the intervention into market dynamics, the ticket presented a serious issue to private mobility companies like us: What will the ticket do to our demand and therefore indirectly our ability to pay wages and invest into our product. In other words: will we lose money due to the 9-euro ticket?

While we are happy about any policy stimulating public transport, we were slightly worried that the ticket will eat up a lot of our demand, and thus, revenue. This, in short, we call the ‘MOIA 9-euro ticket dilemma’.

How to solve this dilemma

As a knowledge driven company, we always strive to understand the factors influencing our service. Therefore, we started viewing the ticket as an experiment that allows us to gain knowledge about the impact of cheap and easy to use public transportation tickets on our service. We were therefore asking ourselves:

“How does the 9-euro ticket affect demand for MOIA?”

In this case demand can be simplified as the number of trips taken with MOIA.

To analytically derive the impact of the 9-euro ticket on the average trips per customer, one would normally want to conduct a controlled experiment. Of course, here a controlled experiment was not possible, as it would have required us to set up two comparable environments (i.e., two Germanys), one with the ticket issued and the other one without. Even worse, we did not even know who had a 9-euro ticket.

Fortunately, our mobility analysts were in the process of conducting a survey for a different purpose, and we were allowed to ask additional questions. In this survey, 1,000 MOIA customers with at least two trips since December 2021 were asked about their mobility behavior. We added the following questions:

  1. Do you own a 9-euro ticket?
  2. When did you purchase your 9-euro ticket?
    (Including the option: “I got it automatically”)
  3. What do you think, how does the 9-euro ticket influence your future MOIA behavior?

Based on the second question, we identified three user groups that likely differ in their behavior, depicted as the horizontal categories in the following figure.

Additionally, we can differentiate across time, in the control and the active periods on the vertical axis, meaning the period before the ticket was activated or thereafter.

For Group A, the ticket was added automatically, meaning that these customers held a public transportation ticket before that was upgraded to a 9-euro ticket on June 1st 2022. For these customers, the control period is the time before June 1st, the active period begins on June 1st.

For Group P, we know that they proactively purchased the ticket. Therefore, their control period is the time before they purchased the 9-euro ticket or before it was activated (any time before June 1st or their actual date of purchase).

The Control Group neither holds a 9-euro ticket nor a public transportation ticket.

Due to the inherent differences between Groups A & P in the way they obtained their 9-euro ticket, we need to apply different methodologies to grasp the ticket’s impact: a Natural Experiment vs. Propensity Score Matching.

The following figure illustrated the two approaches and the respective periods used in them.

Natural Experiment: Remember, we cannot conduct a controlled experiment here. However, for the group of customers who automatically obtained their 9-euro ticket as an upgrade, Group A, we can apply a Natural Experiment approach.

Propensity Score Matching: For the group who decided to purchase their 9-euro ticket, Group P, we cannot pursue this, due to selection bias. Selection bias arises from the fact that the customer themself decides to purchase the ticket (i.e., select themself into Group P). Therefore, we cannot just compare Group P to the Control Group, as it might have an entirely different structure of mobility needs, leading to purchasing the ticket in the first place. Here we will apply a Propensity Score Matching approach, i.e., we will look for close empirical twins between Group P and the Control Group.

Group A: A Natural Experiment

For Group A (automatically obtaining the ticket), we will first look at their control and active period in comparison with the Control Group. Note, that the following plot displays “model-free” evidence, meaning that differences between groups are not controlled for (which we will do later in a statistical model).

The graph below shows the average weekly MOIA trips per customer since December 2021 for the Control Group (grey line) and Group A (golden line). The dashed line indicates the start date of the 9-euro ticket.

Group A and the Control Group appear to be close, with Group A being somewhat higher, suggesting that public transport subscription ticket holders might have a higher frequency of travel with MOIA. This is in line with our expectations. This behavior seems to continue after the introduction of the 9-euro ticket. While Group A has increased their trips near the dotted line, this started already before the introduction of the 9-euro ticket and the grey line overtakes the golden line shortly after. Therefore, we could expect a slight difference between the grey and golden line in the control period and possibly no difference in the active period, when controlling for differences in a statistical model.

Again, the above chart displays model-free evidence. This means that it merely plots the mean weekly trips per customer across both groups, without taking into account that we have some information about differences between those groups. From the model-free evidence, we already suspect differences after the introduction of the 9-euro ticket to be quite small. Next, we will apply a statistical model in which we include further control variables to isolate the “real” effect of the ticket.

For Group A, the 9-euro introduction can be thought of as an experimental stimulus. Both Group A and the Control Group were exposed to the same influence of time-varying external factors. The differences between the groups can therefore be controlled by the behavior in the control period and the answers in the survey (careful: this does not mean that the effects are necessarily causal).

For this experiment, we considered the probability of taking a trip on a given day as the dependent variable and the period from May to the end of June as the Estimation Period. We consider has_nine_euro_ticket (= 1, if customer is in Group A, 0 if they are in Control Group), before_nine_euro_ticket (= 1, if the date is in the Control Period, 0 if it is in the Active Period), as well as the Interaction Effect between the two variables, as the independent variables and include various Control Variables from the survey and the Control Period behavior.

The effects these three variables have on our dependent variable, the probability of taking a trip on a given day, are displayed in the following table. The coefficient displays the effect of the mentioned variable, while the stars indicate whether this effect has happened randomly. If no stars are present, we cannot rule out statistically that it has happened randomly. For sake of readability, we omitted the large set of control variables, which are only included in the model to “clean-up” the effect of our key independent variables.

In this model, no significant change in behavior was found for any of the groups after the 9-euro ticket, and there was no significant difference between Group A and the Control Group either before or during the 9-euro ticket period.

In other words: There is no evidence that the group of customers holding a public transport subscription, and therefore automatically obtaining the 9-euro ticket (Group A), is any different to the group of non-ticket holders (Control Group). Also, neither of the groups appear to change their behavior as a result of the 9-euro ticket.

Group P: Propensity Score Matching

For Group P, we decided to use a Propensity Score Matching approach. The idea behind this approach is to find empirical twins between Group P and the Control Group. These twins should be as similar as possible except for the 9-euro ticket ownership. To do this, we first identify independent variables that capture the variance between the customers in the time before they bought the ticket. The accuracy of the matching depends on the information we provide as independent variables. We then regress this set of “matching variables” on the binary variable of whether the customer bought a 9-euro ticket. The resulting logits from the logistic regression can be used to identify “twins” by identifying one customer with the 9-euro ticket and then the customer without it with the closest logit score.

Looking at the model-free evidence by comparing the members of Group P (blue line) with their empirical twins without the 9-euro ticket in the Control Group (grey line), we see a slight indication that the Control Group has a higher MOIA use especially between March and May. However, once the 9-euro ticket was introduced, there appears to be no difference between the two groups. In contrast to Group A, not everyone in Group P bought the 9-euro ticket in advance, which leads to a different start of the 9-euro ticket. Therefore, no dashed line is indicated.

In a second step, to identify the significance of the impact of the 9-euro ticket, we again apply a statistical model. We define the dependent variable of the Estimation Period as the number of trips per day after activating the ticket, until the 28th of June (which is the end of our observation period). This can be a linear regression, a count model or again a logistic regression, depending on the problem — we tried several and used the count model as it provided the best fit.

In another statistical model, we then regress the binary variable (1, if the customer made a trip, 0 else) on our new dependent variable, strictly including only the set of twin pairs. We additionally use some of the variables from the matching step as control variables in this model. The effect of the has_nine_euro_ticket variable (= 1, if customer is in Group P, 0 if they are in the Control Group) in the second model now shows the structural differences between Group P and the Control Group. The variable has_nine_euro_ticket_at_date (= 1, if the customer has an active ticket at the respective date, 0 else) indicates the effect of the 9-euro ticket after it has been activated.

As with Group A and the Control Group, we do not see a significant difference between Group P and the Control Group over the entire period (as indicated by the absence of stars, nor do we see significant evidence for a behavioral change once the 9-euro ticket was available. Thus, there is no evidence that the financial nudging towards public transportation led to a reduction of MOIA usage in the observed groups.

What we’ve learned

The study conducted by our colleagues at HVV (Hamburg’s public transportation provider) revealed that weekly public transport usage by 9-euro ticket owners increased by 7 % and about half of the respondents report to have used their car less often, resulting in about 12 % reduction of car rides. These are customers open for multi-model transportation, likely fostered by the 9-euro ticket.

We can thus conclude, for us as a privately owned mobility company, on the one hand, the ticket did not pose a threat; the reduction in the price of public transport did not lead to a decrease in ridership among our existing customers. On the other hand, the ticket even increased the multi-modality of its holders. Those are the customers moving within the city, open to alternative mobility solutions. If we make sure we provide comfort, usability and high quality service, a government intervention that causes public transportation to be used more often can even increase usage of other modes of shared mobility.