Impact of SARS-CoV-2 on the mobility behaviour in Germany

In Germany, COVID-19 arrived mid-February 2020, when returnees from Wuhan, China, were isolated. By the end of February, the Federal Ministry of Health and the Federal Ministry of the Interior, Building and Community set up a crisis response team. Cases of infection were confirmed in two federal states. Two days later, the crisis response team recommended the cancellation of major events. Border traffic was monitored, and regional and long-distance travellers with symptoms had to be reported to the health authorities. At the end of March, schools, day care facilities, playgrounds, non-essential shops, zoos, botanical gardens and hairdressers were ordered to stay closed. When shopping for essentials, Germans were required to keep a minimum distance of 1.5 m to others and use a face mask inside shops. In addition, six of the sixteen federal states introduced a temporary lockdown that lasted from four up to 7 weeks [25]. The lockdown meant a prohibition to leave the house without sound reason [25]. Among those sound reasons were the commute to work, grocery shopping, doctor’s appointments, important appointments (e.g. exams), and individual outdoor sports activities (e.g., cycling) or walks (within a certain radius of the home, depending on the federal state) [9].

As a consequence of the COVID-19 outbreak, flight passenger numbers decreased by up to 63% in March 2020 compared to March 2019 [23]. The number of rail passengers went down by roughly 40% for the same period. The number of people using public transport (short distance) decreased by 11% for the first quarter of 2020 (data not collected on monthly basis) [24]. According to Google [13], visitation of retail and recreation facilities went down by as much as 77% at the end of March compared to a reference day from earlier in 2020 (median of the same weekday from January 3rd to February 6th). Similar reductions were reported for transit stations with a decrease up to 68%. Workplace attendance fell by as much as 47% in early April. Analyses conducted by the Project Group Computational Epidemiology [21], who used mobile phone data to identify movements, showed that, at the end of March, weekday movements were reduced by as much as 38%. Similar results were reported by infas [10], who analysed tracking data of 1000 volunteers. The data also showed an increased proportion of walking and cycling, while the modal share of public transport decreased. Indeed, a survey conducted in early April [7] found that a considerable portion of the respondents would feel uncomfortable using public transport at time of questioning (57–63%, dependent on specific mode of transport). Six percent of those who did not own a car reported to consider buying one as result of the COVID-19 outbreak.

Aim of the analyses presented in this paper was to provide a more detailed understanding of the changes in individual mobility behaviour following the COVID-19 outbreak in Germany. Using survey data collected in the early stages of the crisis, the intent was to address questions of modal choice, trip purposes, and changes thereof. A specific focus was put on potential differences that might have arisen from the different lockdown regimes in the German federal states.¹

As the basis for our analyses, we used data that we acquired through a large-scale online survey on mobility behaviour. The survey was set up in direct response to the COVID-19 outbreak in Germany. It was launched on March 20th 2020 and distributed through multiple channels such as social media, newsletters and mailing lists. Overall, the survey consisted of three parts. In the first part, participants were required to provide information on their mobility behaviour before the outbreak of the virus. We asked them to report, e.g., their typical means of transport, their most frequent trip destinations and their trip purpose. The second part contained comparable questions about their current mobility behaviour, i.e. under the changed conditions since the beginning of the virus outbreak and the associated restrictions. The third part asked participants for a prediction of their mobility behaviour after the end of the pandemic. An overview of the three parts of the survey, the items it contained and their origin is shown in Table 1. By May 15th 2020, 6126 participants had completed the survey. A follow-up data collection with the same participants is scheduled for the same period in 2021 (21.03.-15.05.).

For the analyses presented in this paper, only responses that were recorded from March 21st to April 19th were included. This was necessary to ensure that the data only covered the period of the lockdown in those six federal states that had instated one (the Free State of Saxony ended its lockdown on April 20th, others upheld this restriction up until May 9th). To be included in the analysis, participants had to be over the age of 14 (required age to be allowed to respond to such a survey without consent of a parent or legal guardian in Germany), live in Germany, and provide information on which federal state they live in. These criteria resulted in a usable sample of 4157 completed surveys. One thousand seven hundred sixty-nine of the participants in this sample were female, 1804 male (19 other, 565 missing values), with a mean age of 40.2 years (SD = 13.9). The proportion of subscribers to public transport tickets was rather high, which might be attributed to the fact that more than 60% of all participants lived in cities with more than 100,000 inhabitants that provide a good public transport system. Additional information on the sample can be found in Table 2.

Table 3 shows how often respondents made certain types of trips before the outbreak. Nearly 70% of the respondents reported to have travelled to work at least 4–5 times a week. More than 90% went out to make purchases at least once a week. Trips to leisure destinations were frequent as well. In contrast, only a small proportion of respondents made regular trips to accompany others (e.g. for day care) or had to be mobile for business-related errands.

Asked about changes in their trips since the outbreak, nearly all respondents reported that certain trip purposes played no or only a reduced role compared to the pre-outbreak situation.² The most significant reduction was reported for recreational trips (excursions, visits, etc.), which affected the majority (87.2%) of respondents. More than two-thirds reported the elimination of trips as a result of working from home (68.8%) and the cancellation of appointments (67.8%).Half of the participants (51.1%) stated that the use of video conferencing and phone calls resulted in a reduction of trips. There were no differences between federal states with and without lockdown in the types of trips that were reduced (recreational trips: 86.5% vs. 87.7%, working from home: 69.6% vs. 68.1%, cancellation of appointments: 67.6% vs. 68.0%, video conferencing/phone calls: 50.8% vs. 51.3%).

Figure 1 shows the reported modal split before the COVID-19 outbreak. As can be seen, walking, cycling and short-distance public transport (bus/ tram) account for the largest shares (together 76%). Compared to the modal split in Germany as collected through representative surveys [14], the use of bicycles and public transport in our sample is rather high, while car usage is low. When looking at the modal split after the outbreak (Fig. 2), it becomes apparent that the proportion of trips made with public transport shrunk substantially, while the relative importance of walking, cycling and driving increased.

The applied chi-square test also showed that there is a significant difference between the modal split distribution before and during the COVID-19 outbreak, χ²(5) = 423.61, p < .000, V = 0.26.

In order to examine differences between lockdown and no lockdown with respect to the modal split, the change of modal split was considered as difference of the shares during and before COVID-19 per means of transport for each group. As can be seen in Fig. 3, a similar pattern emerged for both groups: increases for walking, cycling and driving and decreases for public transport usage.

Comparing the two groups, it is noticeable that in the lockdown group there is a larger increase in walking (U = 602,095.50, Z = − 5.85, p < .001, d = 0.24) and a larger decrease in the use of public transport (U = 621,497.50, Z = − 4.71, p < .001, d = 0.19). Without lockdown a greater increase in car use was reported than with lockdown (U = 643,187.00, Z = − 3.43, p = .001, d = 0.14). For cycling, there was a larger increase among respondents who were not in lockdown. However, this difference is not statistically significant, U = 679,207.00, Z = − 1.23, p = .218, d = 0.05. A similar picture is found for long-distance public transport (train), where a larger decrease in the no lockdown group is shown. This difference is also not statistically verifiable, U = 670,822.50, Z = − 1.90, p = .058, d = 0.07.

More than half (58.8%) of the respondents confirmed that their use of different modes of transport had changed since the outbreak of the pandemic.³ Figure 4 gives an overview of how the use of different means of transport had changed. More than half of the participants reported to use short-distance public transport (tram, suburban and underground railway, bus) slightly or significantly less. Similar, although slightly less extreme than the latter, were the changes in the use of long-distance public transport (train rides). It should also be noted that the modal share of long-distance public transport was only 5% before the crisis, which might explain a considerable portion of the “no change” responses (i.e. not used before and not used during the crisis).

Walking and cycling showed changes in the opposite direction. About one third of those surveyed reported that they had increased walking since the outbreak of the virus, while about a quarter reported an increase in cycling. For driving (as a driver), the picture was different, with about three quarters of the respondents stating that their car use had not changed, while the remaining respondents were more or less evenly split into those who reduced and those who increased their driving.

When split into federal states with and without lockdown, we found slight differences in the reported change in transport use, whereby the effects are small. For each mode of transport, a comparison was made as to whether the response frequencies (in the categories: little or significantly more, unchanged and little or significantly less) differ between respondents with and without lockdown. Table 4 shows how respondents from states with and without lockdown adjusted their use of different modes of transport during COVID-19.

For walking, 10% more respondents without lockdown indicated no change compared to the group with lockdown. Also 10% more participants in lockdown reported walking a little or significantly more than participants without lockdown, χ²(2) = 44.10, p < .001, V = 0.10. The proportions of people who reported walking a little or significantly less did not differ between lockdown and no lockdown. With regard to cycling, there were differences in the answer categories between the groups, χ²(2) = 16.06, p < .001, V = 0.06. Respondents in the lockdown stated more often that they now used the bicycle slightly or significantly more than respondents without lockdown. The proportion of respondents who indicated that they used the bicycle slightly or significantly less is about the same between the groups. Participants without lockdown stated more often that their bicycle use remained unchanged than respondents in lockdown. For the comparison of the answer categories for car use, the groups also differed statistically significant, χ²(2) = 28.28, p < .001, V = 0.08. People in lockdown stated more often than people without lockdown that they had not changed their car use. Respondents without lockdown stated more often that they used the car a little or significantly less than people in lockdown. At the same time, respondents without lockdown also stated more often that they used the car more than people in lockdown. Differences between the groups could also be found for short-distance public transport, χ²(2) = 49.54, p < .001, V = 0.11. When using public transport, less use was more frequently reported by respondents in lockdown than by respondents without lockdown. In Addition, respondents in lockdown stated less frequently that their use of short-distance public transport had not changed than respondents without lockdown. The percentage of respondents who indicated an increase in use was at a similarly low level for both groups. The same picture emerges for long-distance public transport. Again, there were differences between the groups for the three answer categories, χ²(2) = 23.10, p < .001, V = 0.08. Respondents in lockdown more often stated that they used long-distance public transport slightly or significantly less than those without lockdown. Again, fewer respondents in lockdown stated that their use of long-distance public transport was unchanged compared to the group without lockdown. The frequencies in the answer category “I use slightly or significantly more” did not differ between groups for long-distance public transport.

To identify transitions from one mode to another, we split the survey participants in groups dependent on what type of transport usage they most identified with.⁴ For each of the identified subgroups, we then looked at their main mode of transport⁵ during the outbreak (see Fig. 5). Again, it becomes obvious that public transport was highly influenced by the crisis. Even those that identify primarily as users of public transport refrained from using this mode, with many of them walking, cycling, or, to a lesser degree, using the car instead. In contrast, among those that considered themselves primarily users of cars, pedestrians or cyclists, most stuck with their typical mode of choice.

Asked for an explanation of why a change in modal choice occurred with a predefined set of response options,⁶ many respondents agreed that the intention to reduce the risk of self-infection (72%) or the risk of infecting others (58%) were relevant factors. A few (23%) also agreed that their change in modal choice might have been due to the intent to strengthen their immune system. Seven percent of the participants changed their behaviour out of necessity, as their usual means of transport was not available.

As we were only able to identify minor, isolated differences between federal states with and without lockdown, we considered the possibility that other aspects, such as the degree of urbanisation, might impact on if and how people adjust their mobility behaviour. It can be suspected that highly urbanised regions, i.e. major cities, provide more opportunity to change behaviour, as the distances that would typically have to be covered are comparatively small, and the modes of transport available rather diverse. In contrast, in less urbanised / rural areas, behavioural options might be more limited. To assess this aspect, differences between respondents from highly urbanised (≥ 100.000 inhabitants) and somewhat rural (≤ 20.000 inhabitants, small towns included) areas were examined separately. Figures 6 and 7 show the modal split before and after COVID-19 for participants from urban areas. The patterns are similar to the results for the whole sample (irrespective of the size of place of residence, see section 3.2). Figures 8 and 9 show the modal split before and during COVID-19 for rural areas. As can be seen, the car usage in rural areas was comparatively high, while public transport and cycling played a smaller role than in urban areas. When looking at the differences in modal split before and during COVID-19, the pattern is somewhat similar to that for urban areas. The share of public transport decreased, whereas walking and cycling grew in importance. However, this increase in active mobility was not as extensive as in urban areas.

Subsequently, the modal split differences (during-before COVID-19) between lockdown and no lockdown (Table 6) were examined. In principle, the findings for participants from urban areas are similar to the overall results reported in sections 3.2. This is plausible, as city dwellers account for around 63% of the total sample. The only thing noticeable about urban areas is that, in contrast to the analysis of overall differences in modal split changes (see 3.2), there is a larger decrease for long-distance public transport (train) without lockdown. Noteworthy are the findings for rural areas, which differ from urban areas in that no differences between participants with and without lockdown were found (see Fig. 10), indicating that the measure of lockdown had little impact on transport use in rural areas. There are tendencies that the modal share for driving without lockdown and for cycling in lockdown has increased during COVID-19. However, the lockdown and no lockdown group are not significantly different (see Table 6).

Asked if they intended to maintain any behavioural changes⁷ after the crisis, 26% of the respondents indicated that they might do so, with 22% being undecided (n = 2108). While the majority of the public transport users surveyed who were using a different mode of transport (n = 712) stated that they intend to use bus and train again⁸ after the crisis (63%), about 20% expected to avoid public transport for a while even after the crisis is over.