Impacts of COVID-19 on public transport ridership in Sweden: Analysis of ticket validations, sales and passenger counts

Abstract

The paper analyses the impacts of COVID-19 on daily public transport ridership in the three most populated regions of Sweden (Stockholm, Västra Götaland and Skåne) during spring 2020. The analysis breaks down the overall ridership with respect to ticket types, youths and seniors, and transport modes based on ticket validations, sales and passenger counts data. By utilizing disaggregate ticket validation data with consistent card ids we further investigate to what extent fewer people travelled, or each person travelled less, during the pandemic. The decrease in public transport ridership (40%–60% across regions) was severe compared with other transport modes. Ridership was not restricted by service levels as supply generally remained unchanged throughout the period. The ridership reduction stems primarily from a lower number of active public transport travellers. Travellers switched from monthly period tickets to single tickets and travel funds, while the use and the sales of short period tickets, used predominantly by tourists, dropped to almost zero. One-year period tickets and school tickets increased from mid-April, which could indicate that the travellers using these tickets are particularly captive to the public transport system. Collaborative effort is required to put the results in the international context.

Highlights

• •Public transport ridership has been hit hard by COVID-19 compared with other modes
• •Decrease in ridership largest in Stockholm (ca 60%) and smallest in Västra Götaland (ca. 40%)
• •Reduction stems primarily from fewer active public transport travellers
• •Travellers switched from 30-day period tickets to single tickets and travel funds
• •Sales of short period tickets dropped to almost zero

1. Introduction

The COVID-19 pandemic has prompted governments and authorities around the globe to impose restrictions on transport and mobility at an unprecedented scale and magnitude. As of autumn 2020, the development is in an uncertain stage, where some regions and countries have started or are planning to lighten restrictions on mobility, while many areas are still suffering severely from the pandemic. In any case, an assessment of the first months of the pandemic is important to guide policy during its continuation as well as potential future pandemics and other crises.

Mobility service providers have published data based on geographical location data (e.g., Google COVID-19 Community Mobility Reports1 ), travel planner queries (e.g., Apple Mobility Trends2 ) or app usage (e.g., Moovit Public Transit Index3 ). Surveys of mobility patterns have been conducted in many places, e.g., Switzerland (Molloy et al., 2020), Chile (Tirachini et al., 2020) and Sweden (WSP, 2020a). While geographical settings and data vary, a consistent pattern emerges that public transport has been hit particularly hard compared to private cars and other modes.

The decline of public transport ridership is likely due to both authorities' restrictions and travellers' own choices. Public transport stations and vehicles are recognized as high-risk environments for the transmission of COVID-19 due to the limited physical space available, the abundance of surfaces that help spread the virus, and the limited testing of crew and passengers who use the system (Musselwhite et al., 2020; UITP, 2020). Evidence from Sweden shows that bus and tram drivers were among the group of professions with the highest risk of being infected (Public Health Agency of Sweden, 2020b).

Unlike many countries, Sweden opted for a strategy relying mainly on recommendations rather than mandatory enforcements to limit human interaction (Sabat et al., 2020). From the middle of March to at least summer 2020, Swedish citizens were advised to stay at home if feeling sick in any way and work from home if possible. Meetings involving more than 50 people were banned, and high schools, colleges and universities are closed for students. From early April 2020, people were advised to travel with public transport only if necessary. Public transport services were generally operated at or near nominal levels to restrict transmission risks. The distinct approach towards the pandemic makes it interesting from an international perspective to study the mobility impacts in Sweden.

The research literature on the impacts of COVID-19 on public transport use is, as of autumn 2020, still limited. Tirachini and Cats (2020) outline a number of urgent issues and important directions for research in order to cope with the ongoing crisis while continuing the development towards a sustainable transport system. Aloi et al. (2020) combine data from traffic counters, public transport GPS and ticket sales data, and pedestrian flows from traffic cameras to provide an overall assessment for Santander, Spain. Dzisi Jr. and Dei (2020) study the adherence to social distancing and mask wearing regulations in Kumasi, Ghana based on roadside observations. Tan and Ma (2020) investigate the propensity to choose rail public transport as commuting mode during COVID-19 based on questionnaire responses. The results show that occupation, pre-COVID-19 commuting mode choice, walking time from home to the nearest metro station, and indicators of infection risk in private car and public transport have significant influence on the choice. Teixeira and Lopes (2020) study the ridership in the subway and bike sharing system in New York City during the outbreak. The bike sharing system suffered a smaller ridership drop (71% vs 90%) and an increase in average trip duration (from 13 min to 19 min). The results also suggest that some subway users changed mode to the bike sharing system. Wilbur et al. (2020) analyse bus ridership in Nashville and Chattanooga, TN, USA, and find the largest drops during the morning and evening commutes, with large differences between the highest-income areas and lowest-income areas in Nashville (77% vs 58% drops). Almlöf et al. (2020) analyse the propensity to stop travelling by public transport during COVID-19 for the individual holders of 1.8 million smart cards in Stockholm, Sweden, combined with demographic data at the zonal level. The results show that education level, income, age as well as workplace type are strong predictors.

While existing work has highlighted some dimensions of the pandemic's effects on public transport many aspects are still largely unknown, including the responses for different public transport modes (which may affect different operators and travellers), different traveller groups (adults, school children, youths and seniors) and travel habits (long and short period tickets, single tickets and travel funds). From a behavioural perspective, another important question is to what extent the reduction in ridership is extensive (i.e., caused by a reduction in daily active travellers) versus intensive (i.e., caused by a reduction in the number of trips per day by each active traveller). This analysis requires detailed information from ticket validations and sales.

The aim of this paper is to address the identified research gap by analysing the impacts of COVID-19 on daily public transport ridership in the three most populated regions of Sweden (Stockholm, Västra Götaland and Skåne) during spring 2020. The Stockholm data set contains the full set of individual ticket validations. The Västra Götaland data set contains the daily number of boardings on public transport vehicles based on automatic passenger counting (APC) sensors in the vehicles, as well as daily ticket sales per ticket category. The data are available for different modes (bus, tram and train). The Skåne data, finally, contain the daily number of ticket validations for buses. A major advantage of these data compared to location-based data, such as proximity of a person to a public transport hub, or travel planner queries is that they represent actual public transport trips. Further, unlike studies based on travel surveys they represent the full set of travellers and trips in the system rather than a sample. Based on the combined data sources, the paper addresses the following questions:

• •How was public transport affected compared to other urban travel modes?
• •How did the impacts vary between regions?
• •How did the change in ridership vary across public transport modes?
• •Were fewer people travelling, or was each person travelling less?
• •Which type of tickets were travellers using?
• •How was the ridership affected for school children, youths and seniors?

The rest of the paper is organized as follows. Section 2 describes the data and the methodology used to assess the impacts, Section 3 presents the results, and Section 4 discusses the findings and concludes the paper.

2. Data and methodology

The study uses data from the three largest regional public transport authorities in Sweden: Stockholm, Västra Götaland (including the second largest city Gothenburg) and Skåne (including the third largest city Malmö). Of the three regions, Stockholm was hit the hardest by COVID-19 between March and May 2020 in terms of both absolute number of cases and cases per capita, followed by Västra Götaland (see Fig. 1 ).

Fig. 1.

Daily number of reported cases (bars) and cumulative number of reported cases per 100,000 inhabitants (lines) of COVID-19 for Stockholm, Västra Götaland and Skåne Regions.

Source: Public Health Agency of Sweden (2020a).

The major urban areas of Sweden have well-developed public transport systems with typically high mode shares. Each region has its own systems for fare validation and data collection. Furthermore, the regions differ in terms of available transport modes and ticket products. Further, the data from the three regions differ in terms of coverage and collection method. Hence, the comparison between regions must be done with some care. On the other hand, the data sets complement each other and allow for a broader set of questions to be addressed. This section describes the data available for each region and the methods used to extract the relevant information.

2.1. Stockholm

Stockholm Region has the largest public transport system in Sweden. Before COVID-19, on average 900 thousand people used the public transport system to make around 2 million trips per day. The system consists of four main transport modes: metro (44% of all trips), buses (39%), commuter trains (11%) and LRT/trams (6%). In general, different private operators are responsible for the different transport modes.

The calculation of daily public transport ridership is based on ticket validation data from the digital ticket system SL Access. Tickets are loaded on contactless cards or on smartphones. The system is tap-in only, meaning that tickets are validated at the entrances of stations or vehicles but not at the exits. Tickets are generally validated at the station gates in the metro and train system, at the front door near the driver on buses, and at the platform or manually on board in the LRT and tram system. Cards are not personal and can be shared, e.g., within a household.

Every registered tap-in generates a record containing several attributes that are used in this study, in particular the transport mode (metro, commuter train, tram/LRT or bus) and ticket product (various period cards, single ticket, travel funds, full or discounted rate, etc.). Furthermore, each record contains a unique id number for the card or mobile device carrying the ticket. This number has been anonymized by the Region Stockholm Transport Administration before we carry out our analysis.

We define a “trip” as the passenger movement corresponding to one tap-in. In general, a transfer between different public transport modes will generate a new trip, although exceptions exist at some major transfer hubs. A transfer between buses will also register a new trip, but a transfer within the metro or the commuter train system will typically not require a new tap-in. We split the trip counts into different transport modes and into different ticket products. SL Access includes a wide range of ticket products, which we group into four major categories:

• 1.Long period tickets (30 days and longer),
• 2.Short period tickets (1–7 days),
• 3.Single tickets and travel funds,
• 4.Special school and youth tickets.

The first three categories are further split into full fare and reduced fare tickets, while the fourth category is reduced fare by design. Reduced fare tickets are available to seniors (age above 65), youths (age under 20) and students.

The card id numbers allow us to count the number of active cards each day (i.e., smartcards that have registered at least one ticket validation that day), which in turn is used to compute the daily average number of trips per active cards. This separation is used to assess to what extent the overall decrease in ridership stems from a reduction of active travellers or from a fewer trips per traveller. Since cards are not personal, equating one card with one traveller is only approximate, but should in general be a reasonable interpretation.

Daily trip counts are calculated from 1 February 2020 to 31 May 2020. As reference, trip counts are also computed for the same time period the previous year, i.e., from 1 February 2019 to 31 May 2019. We calculate relative changes in ridership to the previous year by computing the ratio between the daily trip count in 2020 and the trip count of nearest day with matching weekday or holiday status in 2019. For example, Sunday 1 February 2020 is compared to Sunday 3 February 2019. This approach allows us to assess the relative ridership during Covid-19 considering the normal seasonal variations.

Since 17 March 2020, boarding on buses in Stockholm must take place through the rear doors to reduce the exposure of the drivers. Since ticket validation machines are generally installed at the front door, this means that ticket validations on buses have dropped to almost zero. It is therefore not possible to assess the effects of Covid-19 on bus ridership, and we generally exclude buses in the analysis.

2.2. Västra Götaland

Public transport services in Västra Götaland Region are provided by Västtrafik. Services are provided through buses (ca 48% of all trips in 2019), trams (46%), trains (6%) and ferries. Before COVID-19 around 450 thousand travellers used the system and around 950 thousand trips were made per day. The data used in this study comes from two different sources: The daily number of trips per transport mode is based on automatic passenger counting (APC) sensors at the vehicle doors. These sensors are available for a subset of the vehicle fleet and the number has been scaled up to be representative of the entire fleet. These data contain all transport modes except the ferries. As for Stockholm, data are available from 1 February 2020 to 31 May 2020 and the same time period the previous year. Unlike Stockholm, passenger load data for buses are available also during COVID-19.

Data are also available on the daily number of sold tickets for different ticket categories and, for some categories, the total daily sales revenue. It should be noted that there is a difference between the tickets used a particular day, which the Stockholm data contain, and the tickets bought that day. The tickets bought represent the travellers' current assessment of the suitability of different ticket types. Especially for long period tickets (30 days or more), meanwhile, the tickets used include a lag from past ticket purchases.

2.3. Skåne

Skånetrafiken provides public transport services with buses and trains in Skåne Region, the southernmost part of Sweden. Around 465,000 trips were made each day before COVID-19. Tickets can be bought and validated in a number of ways: through an app, a contactless card, paper tickets bought in ticket machines, or directly by credit card. For this study only data from the buses are available. The data contain the daily number of trips in total and for different ticket types (including school and seniors tickets). Data are available from 1 February 2020 to 31 May 2020 and the same time period the previous year.

3. Results

3.1. Total ridership per region

The relative change in the total daily number of trips for the three regions Stockholm, Västra Götaland and Skåne is shown in Fig. 2 . For Stockholm bus trips have been excluded. In all regions the ridership started dropping in the beginning of March, but at different rates. This coincides in time with the Swedish Public Health Agency increasing the risk level of COVID-19 spreading in Sweden from “low” to “moderate” on 2 March. A drastic drop in ridership occurred around 10 March, when the Public Health Agency increased the risk level to “very high”. Stockholm and Skåne experienced dramatic drops in ridership the first days and reached reductions of around 60%. Västra Götaland had a slower decline and reached a maximum loss of ca. 40%.

Fig. 2.

All regions, daily number of trips, relative change 2020 from the baseline 2019.

Fig. 3 shows the realized daily number of departures for Stockholm metro (left) and buses (right) based on automatic vehicle location data. Except for a temporary reduction in bus service from late March to early May, supply remained mostly unchanged during the period. The reduction in ridership was thus in general not caused by reduced service levels but represent a behavioural change from the travellers' side. On the contrary, the reduction in supply was initiated in response to the ridership drop, but was later reverted due to concerns over high transmission risks.

Fig. 3.

Stockholm, realized daily number of departures for the metro (left) and buses (right).

As comparison to the results above, Fig. 4 shows indicators of public transport use as reported by Google and Apple based on mobile devices located near public transport stations and travel planner use respectively. Both data sources are proxies for the actual public transport ridership. In the latter case, data are available at the city level and not the regional level. Hence, we show data for the largest city in each region: Stockholm, Gothenburg and Malmö, respectively. The proxies agree with the ridership data in some aspects: the rapid decline in mid-March and the subsequent recovery are captured, and the fact that Stockholm suffers the largest reduction. However, both proxy indicators appear to underestimate the reduction in ridership, in particular for Stockholm. Furthermore, both proxies indicate a stronger recovery in ridership during April and May than what can be observed from actual ticket validations and boarding counts.

Fig. 4.

Left: Mobility activity at public transport stations. Source: Google COVID-19 Community Activity Reports (2020). Right: Public transport travel planner search activity. Source: Apple Mobility Trends (2020).

3.2. Comparison with other modes

The public transport ridership may also be compared to other modes of transport. Fig. 5 shows the evolution of bike flows, pedestrian flows and motorized road traffic in Stockholm city based on stationary sensors and the congestion charging system. Bike and pedestrian flows are available per week for both 2019 and 2020 and are divided into the inner city and the outer city, and into weekdays and weekends. Motorized road traffic flows are available for the congestion charging cordon around the inner city and for the motorway Essingeleden passing through the city. Here, baseline values from 2019 were not available.

Fig. 5.

Stockholm city, weekly relative change in bike flows (top left) and pedestrian flows (top left), daily number of motorized road vehicles (bottom).

Source: City of Stockholm (2020).

Bike flows show no clear decline during COVID-19 compared to the previous year. In fact, biking increased in the outer city on both weekdays and weekends. However, this trend appears to have started already before the pandemic outbreak. It should be noted that bike flows are dependent on temperature and other weather conditions. Pedestrian flows remained stable in the outer city, but a significant drop occurred in the inner city. The magnitude of the decline reached 60% at most, which is on par with the public transport ridership. Road traffic flows, finally, dropped somewhat at the onset of the closedown but have since recovered to the same levels as before.

3.3. Ridership per transport mode

We break down the ridership impacts into the different public transport modes. The ridership relative to the baseline level represented by the previous year is shown in Fig. 6 for Stockholm (left) and Västra Götaland (right). In Stockholm, ridership for both metro and commuter trains fell around 60% mid-March and has remained at a similar level until end of May. The smaller decrease for commuter trains compared to metro may reflect that long-distance travellers, which dominate on the trains, have fewer alternatives and are more captive to public transport. The large decline in tram and LRT trips can be explained in part by less active ticket validation, which to some extent is done manually, to reduce exposure of the crew. Until 17 March, when ticket validation data ceased, bus ridership followed a curve close to those for metro and trains.

Fig. 6.

Daily relative change in number of trips from 2019 to 2020. Stockholm (left) and Västra Götaland (right).

In Västra Götaland, ridership on trains decreased around 60%, similar to Stockholm. The decrease for trams was somewhat smaller, around 40–50%. The smallest decrease, around 30%, occurred on the buses.

3.4. Active travellers and trips per traveller

The disaggregate ticket validation data from Stockholm allows us to investigate to what extent the reduction in ridership is caused by a reduction in daily active travellers, versus a reduction in the number of trips per day by each active traveller. Fig. 7 shows the daily total number of active smartcards and the daily average number of trips per active card (bus trips excluded). The daily number of active cards dropped from the normal level of ca. 650,000 on weekdays pre-COVID to ca. 200,000 post-COVID, ca. 60% compared to the baseline. The average daily number of trips per active card on weekdays dropped from 2.25 to 2.1, around 5% compared to the baseline. Thus, the major part of the reduced ridership comes from fewer active travellers. This is intuitive since public transport is typically used to travel both to and from a location (in particular, the daily commute). Hence, it is difficult to reduce the average daily number of trips per traveller below 2.

Fig. 7.

Stockholm, daily number of active cards and daily average number of trips per active card, in absolute numbers (left) and relative to the baseline (right). Bus trips are excluded.

3.5. Ridership per ticket type

Fig. 8 shows the daily number of trips per ticket type in Stockholm: 30-day, 90-day, and 1-year period cards, short period cars (7 days and less), and single trips and travel funds. Only full fare tickets are included. The majority of trips are carried out with 30-day period cards, which are common among commuters and other daily public transport users. Relative to the baseline from the previous year, single tickets and travel funds increased substantially between mid-March and end of May. The 1-year ticket cards also increased during this period. Meanwhile, the relative change for 30-day cards remained more or less constant while the 90-day period cards continued to decrease. This indicates that the moderate increase in overall ridership that occurred during this period can be attributed to passengers travelling with single tickets or travel funds, and with yearly cards. The former ticket category may be the option for occasional travellers, while the latter may be an option for captive public transport users. Travel with short period cards has dropped to almost zero. This may be an effect of the national and international travel restrictions that more or less stopped travel for tourism and business.

Fig. 8.

Stockholm, daily number of trips per ticket type in absolute values (left) and relative to the baseline (right), full fare tickets only. Bus trips are excluded.

Fig. 9 shows the daily ticket sales amount in Västra Götaland divided into three categories: single tickets and travel funds, short period tickets (from 90 min to 14 days) and long period tickets (from 30 days to one year). The patterns are similar to Stockholm: single trip tickets and travel funds increased steadily between mid-March and end of May. Sales of short period tickets all but stopped but have also seen a moderate return. A weekly recurring pattern can be observed for both short and long period tickets, and sales of long period cards display a large variability between days. However, no clear increase or decrease can be observed from mid-March and onward.

Fig. 9.

Västra Götaland, daily ticket sales amount in absolute numbers (left) and relative to the baseline (right), full fare tickets only.

3.6. Ridership among school children, youths and seniors

In Stockholm, reduced fare tickets can be purchased by seniors, students or youths. Youths also have the opportunity to purchase extra subsidized youth and school tickets. Fig. 10 shows the daily number of trips in Stockholm with discounted fare tickets of different types: single trip tickets and travel funds, short period cards (7 days and shorter), long period cards (30 days and longer), and youth and school tickets. The patterns for the first three categories are similar to the full fare counterparts (compare Fig. 8): Single tickets and funds decreased around 60% but have since bounced back to some extent. Long period cards dropped more than 70% and remained at that level until end of May, while short period tickets dropped more than 80%. Youth and school tickets dropped by at most around 60% but had recovered to around 50% of the baseline by end of May.

Fig. 10.

Stockholm, daily number of trips per ticket type in absolute values (left) and relative to the baseline (right), reduced fare tickets only. Bus trips are excluded.

4. Discussion and conclusions

This paper has examined the effects of COVID-19 on public transport ridership in the three largest regions of Sweden based on ticket validation, ticket sales and passenger counting data. Of the three regions, Stockholm was hit the hardest by COVID-19 cases between March and May 2020. The analysis shows that it was also in Stockholm where the decrease in public ridership was the largest (ca 60%), while the smallest decrease occurred in Västra Götaland (ca. 40%). From mid-April, ridership slowly increased but was still substantially lower than the previous year.

Comparison with other transport modes in Stockholm shows that public transport ridership has been hit particularly severely; only pedestrian flows in the inner city reached similar low levels compared to previous year. The reduction in ridership stems primarily from a reduction in the number of active public transport travellers, while the daily average number of trips per active traveller stayed relatively stable. Regarding ticket types, travellers switched from 30-day period tickets to single tickets and travel funds, while the use and the sales of short period tickets, used predominantly by tourists and other short-term visitors, dropped to almost zero. One-year period tickets and school tickets, which may represent groups that are particularly captive to the public transport system, increased from mid-April.

The results are inconclusive when it comes to the impact for different public transport modes. A deeper analysis is required to assess whether the differences between regions can be attributed to demographic differences among the public transport travellers using different modes. In any case the differences have important financial implications as the modes are generally managed by separate operators.

In summary, many travellers drastically changed their mobility patterns by abandoning public transport. Of those who remained, many switched to more flexible ticket types. Part of the moderate recovery in ridership the latter half of the period was due to returning captive public transport travellers. The outflow from public transport to private cars and to some extent bikes is in line with existing evidence (e.g., Molloy et al., 2020; WSP, 2020a). Whether this pattern is transient or signifies a shift in the long-term equilibrium is an important question for continued monitoring.

International comparisons of COVID-19 mobility impacts have so far been based on data from mobility service providers such as Google, Apple and Moovit (e.g., Tirachini and Cats, 2020). However, our analysis suggests that indicators based on human proximity to public transport stations and travel planner queries have overestimated the recovery of public transport ridership during the months following the outbreak. Hence, collaborative effort is required to put the results here, which are based on actual ridership data, in the international context.

In contrast to many other countries, Swedish authorities chose a path of strongly recommended measures rather than mandatory curfew or lockdowns. Citizens were advised to stay at home if feeling sick, work from home if possible and use public transport only if necessary, while public transport services were kept at or near nominal levels to lower crowding and transmission risks. Even so, the shift in people's activity and mobility patterns was drastic, quick and, as of yet, quite persistent. In an international perspective, Sweden exhibited high levels of transmission during spring 2020 but low levels during summer and early autumn. While the role of public transport in the transmission of COVID-19 is still not settled, our data do not suggest a strong correlation between ridership and transmission rates at an aggregated level. It may be noted that the number of reported COVID-19 cases in Västra Götaland, which had the smallest ridership drop during spring, exceeded the number of cases per capita in Stockholm by the end of June. Whether these facts are connected is beyond the ability of this study to address. In any case, higher ridership correlates strongly with more social interactions and may not itself be the driver of transmission.

As the pandemic develops, a crucial issue is how high levels of service, which are required to maintain attractiveness and limit transmission risks, can be balanced with lower ticket revenues. In Sweden, the equivalent of 300 million USD have been allocated to support public transportation in the short term. Private rail operators receive no specific support but can apply for non-sector specific governmental support (WSP, 2020b). The analysis of how people adjust their mobility behaviour must continue in order to understand both the trends at the societal level and the mechanisms at the individual level. Furthermore, development is required in terms of policy, infrastructure, technology, service planning, operations, real-time control etc. in order to rebuild public transport ridership. The resilience of energy and space efficient and sustainable urban transport should be a central theme for future research.

CRediT authorship contribution statement

Erik Jenelius: Conceptualization, Methodology, Investigation, Writing - original draft, Visualization. Matej Cebecauer: Software, Resources, Data curation, Writing - review & editing.