Since January 2020 Elsevier has created a COVID-19 resource centre with

free information in English and Mandarin on the novel coronavirus COVID-

19. The COVID-19 resource centre is hosted on Elsevier Connect, the
company's public news and information website.

Elsevier hereby grants permission to make all its COVID-19-related

research that is available on the COVID-19 resource centre - including this
research content - immediately available in PubMed Central and other
publicly funded repositories, such as the WHO COVID database with rights
for unrestricted research re-use and analyses in any form or by any means
with acknowledgement of the original source. These permissions are
granted for free by Elsevier for as long as the COVID-19 resource centre
remains active.
 Cities 134 (2023) 104206

Contents lists available at ScienceDirect

Cities
journal homepage: www.elsevier.com/locate/cities

Factors affecting public transport performance due to the COVID-19

outbreak: A worldwide analysis
Anastasia Nikolaidou *, Aristomenis Kopsacheilis , Georgios Georgiadis , Theodoros Noutsias ,
Ioannis Politis , Ioannis Fyrogenis
Transport Engineering Laboratory, Department of Civil Engineering, Aristotle University of Thessaloniki, Thessaloniki 54124, Greece

A R T I C L E I N F O A B S T R A C T

Keywords: In this paper we investigate the public transport trip frequency variations, as well as the reasons that led to the
Public transport shift away from public transport means, due to the COVID-19 pandemic. We studied relevant data from the
Moovit Moovit platform, and we compared operational and trip frequency characteristics of public transport systems
COVID-19
before and after the outbreak of the pandemic in 87 cities worldwide. On average, waiting times at public
Income inequality
Waiting time
transport stops/stations increased while trip distances decreased, apparently due to the mobility restriction and
Trip frequency social distancing measures implemented in 2020. Most of the Moovit users who said that they abandoned public
Quality of service transport in 2020 were found in Italy and Greece. We developed linear regression analysis models to investigate
(among the 35 variables examined in the study) the relationship between public transport abandonment rates
and socioeconomic factors, quality of service characteristics, and indicators of pandemic's spread. Empirical
findings show that public transport dropout rates are positively correlated with the COVID-19 death toll figures,
the cleanliness of public transport vehicles and facilities, as well as with the income inequality (GINI) index of the
population, and thus reconfirm previous research findings. In addition, the waiting time at stops/stations and the
number of transfers required for commute trips appeared to be the most critical public transport trip segments,
which significantly determine the discontinuation of public transport use under pandemic circumstances. Our
research findings indicate specific aspects of public transport services, which require tailored adjustments in
order to recover ridership in the post-pandemic period.

1. Introduction psychosocial distress among PT users, since their feelings of fear, stress
or anxiety increased and their perception of safety, along with their
Mobility behaviour and travel mode preferences have been greatly overall travel experience, were accordingly influenced (Campisi et al.,
affected due to the COVID-19 pandemic and the corresponding lock­ 2022; Dong et al., 2021). Therefore, PT systems lost a significant share of
down and physical distancing measures. Globally, an unprecedented users and revenues, specifically during the first wave of the pandemic
mode shift occurred from public to private transport means. This when decrease rates of approximately 90 % were recorded (Fernández
happened because Public Transport (PT) modes were considered as Pozo et al., 2022; Politis, Georgiadis, Nikolaidou, Kopsacheilis, Fyr­
environments where COVID-19 transmission would be a serious threat ogenis, Sdoukopoulos, Verani, and Papadopoulos, 2021a; Vickerman,
due to the relatively small passenger cabin spaces, the often improper 2021). Even after the relaxation of lockdown measures in the fall of 2020
functioning of ventilation systems, the inability to identify infected and the vaccination campaigns, PT systems continued to suffer from
fellow passengers, the vehicles' interior surfaces (e.g., handrails, seats, ridership losses (Fernández Pozo et al., 2022; Monterde-I-bort et al.,
ticket issuing machines) that are touched by multiple users, the crowded 2022; Vickerman, 2021), at least in the short run. Recent research
conditions that favour close contact with other people on-board or at findings show that an important share of former PT passengers is not
stops/stations etc. (Huang & Li, 2022; Kłos-Adamkiewicz & Gutowski, anymore satisfied by PT cleanliness and safety and intend to perma­
2022; Kumar et al., 2021; Lucchesi et al., 2022; Monahan & Lamb, 2022; nently switch to private transport modes (Downey et al., 2022; Przy­
Tirachini & Cats, 2020). These factors were in turn associated with bylowski et al., 2021; Torbacki, 2021). Due to teleworking and increased

* Corresponding author.
E-mail addresses: nikolaid@civil.auth.gr (A. Nikolaidou), kopsacheilis@civil.auth.gr (A. Kopsacheilis), ggeorgiadis@civil.auth.gr (G. Georgiadis), pol@civil.auth.
gr (I. Politis), fyrogeni@civil.auth.gr (I. Fyrogenis).

https://doi.org/10.1016/j.cities.2023.104206
Received 14 October 2022; Received in revised form 31 December 2022; Accepted 10 January 2023
Available online 16 January 2023
0264-2751/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
A. Nikolaidou et al. Cities 134 (2023) 104206

COVID-19 safety concerns, many researchers argue that the future of PT characteristics that explain the PT modal share variations which were
is at stake and people will increasingly prefer private cars over PT researched globally.
(Christidis et al., 2021; Medlock et al., 2021; Munawar et al., 2021; In fact, a global travel preferences' survey, during the outbreak of the
Semple et al., 2022; Ton et al., 2022). At the same time, active transport- COVID-19 pandemic, indicated a significant shift from public to private
friendly initiatives have been introduced worldwide as an environ­ and non-motorized passenger transport modes, and highlighted that the
mentally friendly and social distancing mode of transport, something likelihood of using PT was comparatively higher for females, non-car
that reduced PT usage even further (Kyriakidis et al., 2023; Nikitas et al., owners, and longer trip distances (Abdullah et al., 2020). Montero-
2021). Unless these initiatives are successfully combined with PT into Lamas et al. analysed bus PT ridership and GIS data in A Coruna,
multimodal networks (Alyavina et al., 2022) or ensure a fair distribution Spain, during the COVID-19 lockdown, and they found that in areas with
of road space against all other travel modes (Creutzig et al., 2020; Nello- low-income population, high population densities and important con­
Deakin, 2019; Tzamourani et al., 2022), they might act as an additional centrations of hospitals, offices and supermarket facilities, the PT de­
barrier towards urban mobility ecosystems in which PT should operate mand was relatively higher (Montero-Lamas et al., 2022). In Lisbon,
in tandem with other transport modes. Aparicio et al. compared demand figures between 2019 and 2020 for
Nevertheless, PT continues to remain the only travel option for in­ bus, tram and metro networks and found reductions of up to 80 %, but at
dividuals who cannot afford or use private cars or active transport stations located in peripheral regions with relatively more low-income
modes (Amin & Adah, 2022; Habib & Anik, 2021). Furthermore, the role users, the decrease was lower (Aparicio et al., 2021). The fear of
of PT as backbone of sustainable urban mobility systems and catalyst for COVID-19 infection and the teleworking opportunities were among the
achieving social equity, energy efficiency and climate change overall key reasons that explained the significant reduction of PT trip frequency
goals cannot be neglected in the post-pandemic era. Therefore, the levels (up to 90 %) in Gdansk, Poland (Przybylowski et al., 2021) as well
identification of the factors that led to the abandonment or reduction in as in Warsaw, Poland, where travellers also emphasized their limited
PT use due to the pandemic is of great importance as it will define the personal security on-board due to the crowding and the non-adherence
relevant strategy and policy formulation framework in the post- to the rule of wearing masks (Kłos-Adamkiewicz & Gutowski, 2022). The
pandemic era. The clarification of these factors, in a global scale, may crowding of PT vehicles and the possibility of teleworking also explained
also lead to joint initiatives towards the improvement of quality of PT the decreased PT demand figures in Sicily, Italy, which continued to
services with a theoretically greater impact. Additionally, the recogni­ exist until the last months of the year 2020 (Basbas et al., 2021). Across
tion of both socio-economic characteristics of travellers and specific PT the twenty largest metropoles of the USA, the PT demand decreased
travel aspects, which acted as disincentives to travel by PT, may facili­ between 45 and 80 % during the first months of the COVID-19 outbreak,
tate customized planning, operational and organizational actions that and this decrease was positively correlated with the spread of the disease
will guarantee more resilient and robust PT systems in the future. (number of cases and deaths) and the educational level of the users in
Considering the above, in this paper we investigate the reasons the respective areas (Qi et al., 2021). In Chicago, USA, Hu and Chen
which led to the discontinuation of PT use after the start of the COVID- estimated an average 72 % drop of rail PT users (of the “L” train system)
19 pandemic, by considering a worldwide dataset which combines both during the first wave of the pandemic, with that decline being higher in
local specific socioeconomic characteristics as well as PT service attri­ regions with dominance of white, educated and high-income residents
butes. We synthesized PT mode choice, service characteristics and pas­ (Hu & Chen, 2021). Heavy rail PT usage also decreased in Hong Kong,
sengers' perceptions data from Moovit subscribers, along with China, both among students and adults, who reduced their rail PT trips
socioeconomic and COVID-19 related variables, to compile a cross- by up to 48 % (Zhang et al., 2021). When analyzing the ridership decline
section dataset that covers 87 cities worldwide. We study the changes of the New York subway during the spring of 2020, Sy et al. discovered
in PT service characteristics before (year 2019) and after the outbreak of that among essential workers, who are generally associated with lower
the pandemic (year 2020), the passengers' views on the desired service incomes, the mobility levels were comparatively more increased (Sy
conditions under post-pandemic circumstances, and we develop linear et al., 2021). A study in Belgium examined the hypothesis that women
regression models to identify the PT service aspects along with the are disadvantaged in the face of COVID-19 and found that females used
macro socio-economic factors that might explain the abandonment rates PT more to commute or shop (Assoumou Ella, 2021). A nationwide
on PT use, which were observed globally by the end of the year 2020. survey, after the end of the first lockdown in Canada, showed that PT
The next section reviews the impact of COVID-19 on PT use as well as demand remained decreased when compared to the pre-pandemic
the passengers' profile characteristics that had a critical role on PT mode levels, while certain mitigation measures, such as meticulous cleaning,
choice decisions during the pandemic. Section 3 describes the dataset hand sanitizing and mask-wearing would encourage people to continue
and variables we considered along with the research analysis setting we using PT modes (Labonté-Lemoyne et al., 2020). A survey in Cardiff, UK,
adopted. Our research results and empirical findings are presented and showed that PT was among the least preferable modes during the
discussed in Section 4, while conclusions are summarized in Section 5. pandemic, due to having the highest perceived risk of infection (Angell
& Potoglou, 2022). In Santiago, Chile, during the first week of the anti-
2. Literature review COVID-19 restriction measures, the PT ridership decreased by 30 %–40
% within lower income groups, but among higher income users the drop
The choice of travelling by PT during the COVID-19 pandemic is was >70 % (Tirachini & Cats, 2020). Duenas et al. studied passenger
dependent on either the impossibility of using other travel modes, due to demand data for the PT services of Bogota, Colombia, during the first
economic, social, and housing limitations, or the satisfaction from the half of the year 2020, and explained that the users with worse socio­
quality of delivered services (Lucchesi et al., 2022). In practice, though economic conditions (e.g., poverty, informal work) had comparatively
that PT systems generally experienced great ridership losses, this trend increased mobility levels after the relaxation of the first lockdown
was not always horizontal among the population groups and local measures (Dueñas et al., 2021). The greatest reduction of trip frequency
conditions considered. Previous research has indicated that people from and PT demand figures was observed in the wealthier neighbourhoods of
low-income households without a car, as well as the “essential” workers, Daejeon, South Korea, during the first wave of the pandemic (Kim et al.,
who are employed in healthcare, food supply, primary production sec­ 2021). Though that PT was rated as the most unsafe travel mode before
tors etc., and they often have comparatively lower wages, had to the first COVID-19 lockdown in India, there were no important modal
commute as usual (and by PT if that was the case) to reach their shifts from private to PT modes, possibly due to the lack of other al­
workplaces (Roberts et al., 2020; van Dorn et al., 2020; Wilbur et al., ternatives (Pawar et al., 2020). Munawar et al. analysed data from
2020). These groups of travellers create a pool of PT captive users during Moovit and other sources for Australia and found that due to the phys­
the pandemic and provide with a set of certain socioeconomic ical distancing rules and the fear of virus transmission by fellow

2
A. Nikolaidou et al. Cities 134 (2023) 104206

passengers, the PT use was reduced by 80 % in April 2020 (Munawar • Number of Transfers: The average number of transfers that people
et al., 2021). make during an average PT commute trip (one way).

3. Materials and methods Furthermore, due to the COVID-19 pandemic, Moovit collected data
on users' willingness to use PT, given 11 specific COVID-19 related in­
In this section we describe the basic elements of the methodology centives, and on the effects of COVID-19 on PT trip frequency. More
that we followed in our analysis. The data collection process and anal­ specifically, considering COVID-19, Moovit users were asked to state
ysis, as well as the use of the appropriate statistical processing tools and their preference on specific incentives that aimed to encourage PT use
techniques for modelling PT trip frequency during COVID-19 pandemic during the pandemic. These incentives (Variables 21–31 in Table 1)
period (year 2020), are analysed in detail in the following subsections. pertain to better information regarding PT services (time schedule,
vehicle occupancy, etc.) and physical and hygiene protocols (disinfec­
tion of vehicles, sanitizing equipment on vehicles, etc.). Participants in
3.1. The data the Moovit survey also stated how COVID-19 has affected their PT trip
frequency by selecting among five (5) options (“I have switched to other
The variables that were taken into consideration for this study, the modes of transportation”, “I no longer use public transportation”, “I use
source they were extracted from, their measurement units, their public transportation less frequently than before”, “It did not affect the
geographical level and their categorization are presented in Table 1. The frequency that I use public transportation”, “I use public transportation
variables were grouped into five (5) categories: (a) Socioeconomic, (b) more”). For this study the first two options were added (i.e., the per­
PT service attributes, (c) PT trip frequency rates (year 2020), (d) COVID- centage of respondents who no longer use PT and the percentage of
19-related PT incentives, and (e) COVID-19 pandemic characteristics. those who have switched to other modes of transport) to create the
All collected data referred to the year 2020 except for the data related to “Abandoned PT” variable (Table 1).
the variables “PT service attributes”, which were also collected for the The Moovit dataset was enriched with data from a wide variety of
year 2019. Due to the limited information availability in the respective other sources in order to include a dynamic range of potential factors
databases, the variables Human Development Index (HDI) and Gender that affected the PT usage, ranging from social and economic (Variables
Inequality Index (GII) also pertain to the year 2019. Table 1 also presents 1–6) to COVID-19 related ones (Variables 32–35):
basic descriptive statistics for these 35, in total, variables. In line with
our research objective, the variable: “Abandoned PT” was treated as a • The (macro) socioeconomic variables of Table 1 were selected after
dependent variable, while all the other variables were treated as inde­ considering a plethora of indicators which are stored in various re­
pendent ones. positories (Table 1). Factors concerning population characteristics
For this study, a substantial part of the data came from the publicly refer to the city's population and urban population density. Economic
available Moovit mobile phone application database (Moovit, 2021). factors include Gross Domestic Product (GDP) per capita and the Gini
Moovit is an application that allows users to view PT service charac­ Index, which measures the income distribution in a country and
teristics, such as bus arrival times, maps, and train schedules, around the ranges from 0 % to 100 %, with higher scores indicating higher in­
world. Moovit is considered an early pioneer of Mobility as a Service come inequality. More complex indicators were also included, i.e.,
(MaaS), since its launch in 2012 (Data Europa EU, 2019). Moovit covers the Human Development Index (HDI) and Gender Inequality Index
all modes of PT, local bicycle services, ride-hailing (Uber / Lyft), (GII). The HDI is a composite index that measures human achieve­
scooters, car-sharing, carpooling, etc. The Moovit app combines infor­ ment through three (3) contributing dimensions of human develop­
mation from PT operators and authorities with feedback from the user ment; “a long and healthy life, knowledge and a decent standard of
community to offer travellers a real-time picture, including the best living”, and takes values from 0 to 1, with higher scores indicating
route for their journey (Santos & Nikolaev, 2021). Moovit uses both higher human development. The GII is an indicator that is comple­
passive crowdsourced data to improve PT route schedule predictions (e. mentary to the HDI and measures the disparities in the HDI by
g., collecting information about actual arrival times of PT lines) and user gender. The closer the GII is to 1, the smaller the gap between women
reports, for instance, on incidents, delays, and other type of information and men.
(Heiskala et al., 2016). • The COVID-19 pandemic-related variables of Table 1 are: the num­
Moovit data are available at the city level. Moovit covers 87 major ber of COVID-19 cases and deaths per million citizens for the year
cities worldwide, spanning across four (4) continents; North America 2020, the Stringency Index and the Containment Health Index. The
(New York, Miami, San Francisco, Washington, etc.); South America Stringency Index is a composite measure of nine (9) response metrics
(Brasilia, Buenos Aires, Santiago, Bogota, etc.); Europe (London, Paris, (school closures, workplace closures, cancellation of public events,
Berlin, Rome, etc.) and Asia (Singapore, Bangkok, Jakarta, Kuala Lum­ restrictions on public gatherings, closures of PT, stay-at-home re­
pur, etc.). For the present study, we utilized five (5) PT service attribute quirements, public information campaigns, restrictions on internal
variables (Variables 7–16 in Table 1), that pertain to the years 2019 movements and international travel controls) (Hale et al., 2021). The
(pre-pandemic) and 2020 (pandemic). These five (5) variables are Containment Health Index uses the same nine (9) indicators as the
regularly estimated by Moovit and they are closely related to PT quality Stringency Index plus testing policy, the extent of contact tracing,
of service: requirements to wear face coverings and policies around vaccine
rollout (Hale et al., 2021).
• Average Commute Time: How long people usually commute one
way by PT (to or from home/work). This includes walking, waiting, All these 35 variables were integrated into a single database and
and travelling time. were filtered (observations with missing values were removed) and
• Average Waiting Time: How long people wait for PT, on average, cleaned (tested for errors and outliers). A schematic visualization of data
during a commute trip (one way). If a trip includes transfers, the management is presented in Fig. 1.
value is the sum of the waiting time at the different stops/stations.
• Average Walking Distance: How far people walk on average during 3.2. Analysis setting
a PT commute trip (one way). This includes walking distance from
the start to end of the trip, and during transfers. Based on our research objective, we chose to examine the relation­
• Average Trip Distance: The average distance a person travels dur­ ship between the “Abandoned PT” variable and the variables which are
ing an average PT commute trip (one way). related to the socioeconomic, PT service attributes, COVID-19 related PT

3
A. Nikolaidou et al. Cities 134 (2023) 104206

Table 1
Descriptive statistics and data collection sources for the variables of our study.
# Variable Source Units Spatial Category Max Min Mean Standard
Level Deviation

1 City Population World Population Review ( People City Socioeconomic 22,043,028 176,418 3,643,996 4,049,500
(2020) Worldpopulationreview.
com, 2021)
2 Urban Population Demographia World Urban People/ City Socioeconomic 16,240.000 620.000 4688.276 3099.132
Density (2020) Areas (Demographia, 2021) km2
3 Human United Nations (United – Country Socioeconomic 0.947 0.718 0.857 0.064
Development Index Nations Development
(HDI) (2019) Programme, 2020)
4 Gender Inequality United Nations (United – Country Socioeconomic 0.480 0.049 0.207 0.140
Index (GII) (2019) Nations Development
Programme, 2020)
5 Gini Index (2020) World Bank (World Bank, – Country Socioeconomic 54.200 26.000 39.747 6.477
2021b)
6 Gross Domestic World Bank (World Bank, $ Country Socioeconomic 59,939.000 3837.000 26,513.980 17,437.240
Product (GDP) per 2021a)
capita (2020)
7 Average Commute Moovit min City PT service 72.000 20.000 46.046 11.354
Time (2019) (Moovit, 2021) attributes (year
2019)
8 Average Commute Moovit min City PT service 71.000 22.000 46.115 11.143
Time (2020) (Moovit, 2021) attributes (year
2020)
9 Average Waiting Moovit min City PT service 25.620 7.840 13.704 4.402
Time (2019) (Moovit, 2021) attributes (year
2019)
10 Average Waiting Moovit min City PT service 31.000 8.000 14.713 5.456
Time (2020) (Moovit, 2021) attributes (year
2020)
11 Average Walking Moovit min City PT service 1330.000 518.750 817.931 167.622
Distance (2019) (Moovit, 2021) attributes (year
2019)
12 Average Walking Moovit min City PT service 1364.000 415.000 667.540 195.686
Distance (2020) (Moovit, 2021) attributes (year
2020)
13 Average Trip Moovit km City PT service 20.880 2.670 8.259 3.128
Distance (2019) (Moovit, 2021) attributes (year
2019)
14 Average Trip Moovit km City PT service 12.410 2.210 7.095 2.276
Distance (2020) (Moovit, 2021) attributes (year
2020)
15 Number of transfers Moovit Transfers City PT service 2.018 1.190 1.628 0.197
(2019) (Moovit, 2021) attributes (year
2019)
16 Number of transfers Moovit Transfers City PT service 2.068 0.895 1.622 0.227
(2020) (Moovit, 2021) attributes (year
2020)
17 I use PT less Moovit % City PT trip 54.600 30.200 42.269 5.908
frequently than (Moovit, 2021) frequency rates
before (year 2020)
18 It did not affect the Moovit % City PT trip 57.500 14.800 36.407 8.147
frequency that I use (Moovit, 2021) frequency rates
PT (year 2020)
19 I use PT more Moovit % City PT trip 22.500 0.000 7.667 4.061
(Moovit, 2021) frequency rates
(year 2020)
20 Abandoned PT Moovit % City PT trip 38.700 0.000 13.663 5.555
(Moovit, 2021) frequency rates
(year 2020)
21 Touchless payment Moovit % City COVID-19 57.400 0.000 21.386 11.048
systems and/or (Moovit, 2021) related
mobile ticketing Incentives
22 Respecting social Moovit % City COVID-19 57.600 8.800 40.568 13.616
distancing (Moovit, 2021) related
regulations on Incentives
vehicles and at
stops/stations
23 Real-time arrival Moovit % City COVID-19 60.900 6.800 39.914 14.119
information so I (Moovit, 2021) related
don't need to wait at Incentives
crowded bus stops
for long
24 More accurate Moovit % City 43.500 3.100 27.311 11.125
information on (Moovit, 2021)
(continued on next page)

4
A. Nikolaidou et al. Cities 134 (2023) 104206

Table 1 (continued )
# Variable Source Units Spatial Category Max Min Mean Standard
Level Deviation

which lines are COVID-19

working related
Incentives
25 Increase the Moovit % City COVID-19 72.300 21.400 50.048 12.378
frequency of (Moovit, 2021) related
vehicles so that each Incentives
vehicle is less packed
26 Hand sanitizer Moovit % City COVID-19 54.800 3.800 36.276 13.318
dispensers on (Moovit, 2021) related
vehicles and/or at Incentives
stops/stations
27 Flexible work Moovit % City COVID-19 41.500 0.000 22.608 11.682
schedules to (Moovit, 2021) related
eliminate peak hours Incentives
caused by traditional
9–5 office working
schedules
28 COVID-19 Moovit % City COVID-19 62.200 0.000 38.400 15.217
disinfection of (Moovit, 2021) related
vehicles, stations Incentives
and stops
29 Being notified via Moovit % City COVID-19 45.000 1.800 22.895 10.109
smartphone (Moovit, 2021) related
notifications to Incentives
know which vehicles
were recently
disinfected
30 Being able to pre- Moovit % City COVID-19 30.700 1.000 13.549 6.827
book rides on public (Moovit, 2021) related
transit vehicles with Incentives
limited capacity
31 Being able to know Moovit % City COVID-19 54.400 0.000 33.479 13.577
in advance which (Moovit, 2021) related
vehicles are crowded Incentives
32 Number of cases per Ourworldindata (Hannah Cases Country COVID-19 60,650.100 98.412 33,249.760 14,936.360
million 2020 Ritchie Edouard Mathieu pandemic
and Roser, 2020) characteristics
33 Number of deaths Ourworldindata (Hannah Deaths Country COVID-19 2789.917 0.872 846.793 431.518
per million 2020 Ritchie Edouard Mathieu pandemic
and Roser, 2020) characteristics
34 Stringency Index Ourworldindata (Hannah – Country COVID-19 82.010 50.040 64.523 6.033
Ritchie Edouard Mathieu pandemic
and Roser, 2020) characteristics
35 Containment Health Ourworldindata (Hannah – Country COVID-19 72.520 47.660 56.910 4.933
Index Ritchie Edouard Mathieu pandemic
and Roser, 2020) characteristics

incentives and COVID-19 pandemic characteristics (Table 1). We In order to identify which factors affect PT trip frequency during the
considered the existing research literature (Section 2) to construct our COVID-19 pandemic, we fit a Multiple Linear Regression model that can
initial hypotheses: adequately predict the “Abandoned PT” variable, which was presented
in the previous sub-chapter. Multiple linear regression refers to a sta­
• Regarding the PT use during the pandemic and the socioeconomic tistical technique which is used to predict the outcome of a variable
characteristics of the respective populations, we expect that cities based on the value of two or more variables. It is sometimes known
with a higher share of low-income population (and consequently simply as multiple regression, and it is an extension of linear regression.
more limited access to private car) would feature comparatively The variable that we aim to predict is known as the dependent variable,
increased PT ridership figures during the pandemic. while the variables we use to predict the value of the dependent variable
• Concerning PT service characteristics, we hypothesize that longer PT are known as independent or explanatory variables. The form of the
total travelling time (e.g., higher commute time, higher waiting model is given by the following Eq. (1) (Jobson, 1991):
times, etc.) could deter users from choosing PT, due to increased
yi = β0 + β1 xi1 + β2 xi2 + … + βp xip + ui i = 1, 2, …, n (1)
exposure to the virus.
• Previous literature findings underline that in cities that were affected
Where, in our analysis setting:
more by COVID-19 (more cases/deaths), users chose more private
modes of transport instead of PT.
• yi is the percentage of users that abandoned PT during the pandemic
• Regarding the relationship between the COVID-19 mitigation mea­
• xi1, xi2, …, xip are the explanatory variables
sures and the PT trip frequency rates, we anticipate that actions to­
• The unknown parameters β0, β1, β2, …, βp are constant coefficients
wards ensuring the proper cleaning and disinfecting of PT vehicles
• ui are the model's residuals
will have a positive impact on PT use. Additionally, increasing PT
service frequency (thus reducing on-board crowding levels), could
The statistical tests we carried out in order to evaluate the goodness
also encourage riders to choose PT, as social distancing regulations
of fit and statistical significance of our model, were:
will be better achieved.

5
A. Nikolaidou et al. Cities 134 (2023) 104206

Fig. 1. Methodological framework.

• The coefficient of multiple determination R2. This coefficient takes 4. Results and discussion
values from 0 to 1 and explains the degree to which the explanatory
variables account for the variation of the model's output. Higher 4.1. COVID-19 impact on PT travel habits
values of R2 mean that more of the model's output variation is
explained. According to the Moovit data in Table 1, regarding the PT travel
• The adjusted coefficient of multiple determination R2. This also takes frequency habits of the people in 87 cities across 21 countries, >40 % of
values from 0 to 1, but contrary to R2, penalizes the result when more citizens stated that due to the pandemic they use PT services less often,
variables are added to the model. the 36 % of them said that PT use was not affected and almost the 8 % of
• The Residual Standard Error, which is an estimate of the standard them claimed that after COVID-19 they started using PT more often.
deviation of the model's residuals. However, in this particular question what is more interesting is the
• The F-statistic, which shows whether at least one of the independent percentage of those who abandoned PT. This percentage includes both
variables has a significant correlation to y those who stopped using PT and those who switched to other means of
transport. Whether they will ever get back on board is impossible to
Finally, a Multiple Linear Regression model needs to satisfy a list of predict, but lockdowns, the rise of remote working, and stay-at-home
certain assumptions (Roback & Legler, 2021): orders have all meant to a higher than a 13 % reduction, on average,
in the volume of people commuting by PT worldwide. Generally, over
• There needs to be a linear relationship between the independent 50 % of riders have at least reduced the use or given up PT altogether.
variables X and the dependent variable Y. The maximum PT abandonment rate was observed in Thessaloniki,
• The model's residuals should be normally distributed. Greece, where the 34.3 % of Moovit subscribers no longer use PT
• The residuals at every point of the model should have a constant because of the pandemic, and where the only available PT mode is bus.
variance, which is also called homoscedasticity. Based on local data, PT trips in Thessaloniki were severely limited
• The errors of the observations should be independent. during the pandemic period (modal shares of 22 % during the pre-
• The independent variables, which are used in the model, should not pandemic period compared to 0.2 % during the pandemic period)
be highly correlated with each other, i.e., should not suffer from (Politis, Georgiadis, Papadopoulos, Fyrogenis, Nikolaidou, Kopsacheilis,
multicollinearity. Sdoukopoulos, and Verani, 2021b). Other cities, where >20 % of users
said they stopped using PT are San Francisco, USA, Athens, Greece, and
Palermo, Italy, while the top two countries where the usage has declined

6
A. Nikolaidou et al. Cities 134 (2023) 104206

the most are Greece and Italy. Additionally, 22.5 % of people in Miami, in the city of Santiago, Chile, while Antalya, Turkey presented the
USA said their PT trip frequency rates have increased. Similarly, other highest increase. The city of Venezia, Italy featured the highest decrease
cities where PT trip frequency has increased during COVID-19 are Kuala in terms of average waiting time and number of transfers. On the con­
Lumpur, Malaysia, Los Angeles, USA and Queretaro, Mexico. In the USA trary, regarding the same attributes, the highest average increase was
and Europe, the average percentage of people who stopped using PT was observed in the cities of Rosario, Argentina and Kuala Lumpur,
quite similar, while in Asian cities it was comparatively lower. Most of Malaysia, respectively. Tenerife, Spain was the city with the highest
the cities where PT use increased are located in the USA. decrease in terms of average walking distance, while on the other hand
When asked what would make them more likely to use PT again, Curitiba, Brazil featured the highest increase in the same attribute.
commuters worldwide cited a number of concerns. The top issue, cited Finally, the highest reduction regarding average trip distance was
by almost 50 % of respondents, was the need for more buses on the road observed in Dubai, United Arab Emirates, while the highest increase was
to lessen the chances of vehicles being uncomfortably full. Not far observed in Rosario, Argentina. The increase in average waiting time in
behind, 40.5 % of them requested the respecting of social distancing 2020 could possibly be appointed to the reduced service frequency,
regulations and disinfection of vehicles and stops/stations. A very while average walking distance and trip distance have been decreased
important incentive for the return of users to PT is the provision of probably due to shorter and mainly necessary trips (i.e., more commute
timely information regarding both the arrival of a bus in order to reduce trips compared to recreational trips).
the waiting time at a stop and therefore the possible exposure time, as
well as the provision of information regarding whether there is an 4.2. Discontinuance of PT use
overcrowded bus.
Fig. 2 presents the percentage change of PT service characteristics The multiple linear regression model that was best fitted, with the
between the years 2019 and 2020 in the form of box plots. City outliers percentage of users that abandoned PT in each city as a dependent
are marked with a label while continents are shown in different color. variable, indicated five (5) variables as statistically important pre­
Based on the results, the average commute time and number of transfers dictors: the Gini Index, the Average Waiting Time in 2019, the Number
have not changed significantly in 2020, in contrast to the average of Transfers in 2019, the Number of deaths per million in 2020 and the
waiting time, walking distance and trip distance. Average waiting time percentage of users that would use again PT with better COVID-19
has been increased in 2020 by one (1) minute (13.7 min in 2019
disinfection of vehicles, stations and stops.
compared to 14.7 min in 2020) while average walking distance and trip This model also fulfilled the linear regression assumptions, listed in
distance in 2020 have been decreased by 18 % and 14 % respectively. In the “Analysis Setting” section. Particularly, we ran specific tests, whose
terms of the average commute time, the highest reduction was observed results are presented in Figs. 3 – 5. The scatter plots (Fig. 3) demonstrate

Fig. 2. Change (before (year 2019)-during COVID-19 (year 2020)) on PT service characteristics per city and continent.

7
A. Nikolaidou et al. Cities 134 (2023) 104206

Fig. 3. Scatter plots for the dependent and independent variables.

Fig. 4. Normal Predictive Probability (P–P) and Quantile-Quantile (Q-Q) plots for the residuals of the linear regression model.

a linear relationship between the dependent and all the independent conclusion that the fourth assumption is not violated.
variables. Additionally, the results of the P–P and Q-Q plots presented As Table 2 and Table 3 show, our model does not suffer from mul­
in Fig. 4 suggest that the residuals follow a normal distribution and thus ticollinearity issues. The predictors used in our model are in some cases
the second assumption is fulfilled. Regarding the third assumption, we moderately correlated, but since the correlation coefficients are no >0.8,
assessed the results of the residuals plot in Fig. 5, which highlight a there is little evidence to indicate potential multicollinearity (Franke,
constant variance of the residuals as the dependent value increases. 2010). However, since Pearson correlation values are merely an indi­
Additionally, the distribution of the points on the scatter plot of Fig. 5 cation of multicollinearity, we further examined the issue through the
indicates that there is no apparent relationship between the residuals assessment of Variance Inflation Factors (VIF) and Variance Proportions
and the values of the dependent variable, which leads us to the (Table 3). Based on the results, although the VIF values suggest the

8
A. Nikolaidou et al. Cities 134 (2023) 104206

Fig. 5. Scatter plot of the residuals of the linear regression model.

Table 2
Pearson's correlation table (value, (sig value)) of the factors used in the regression model.
Gini Average Waiting Time Number of Transfers Number of deaths per COVID-19 disinfection of vehicles,
(2019) (2019) million (2020) stations and stops

0.514 − 0.093 − 0.007 0.559

Gini 1
(0.000a) (0.392) (0.947) (0.000a)
0.514
0.089 − 0.091 0.389
Average Waiting Time (2019) (0.000 1
a (0.410) (0.401) (0.000a)
)
− 0.093 0.089 − 0.143 − 0.022
Number of Transfers (2019) 1
(0.392) (0.410) (0.187) (0.837)
− 0.007 − 0.091 − 0.143 − 0.186
Number of deaths per million (2020) 1
(0.947) (0.401) (0.187) (0.084)
COVID-19 disinfection of vehicles, 0.559 0.389 − 0.022 − 0.186
1
stations and stops (0.000a) (0.000a) (0.837) (0.084)
a
Significance at 1 %.

Table 3
Variable Inflation Factors and correlation diagnostics of the factors used in the regression model.
Gini Average Waiting Time Number of transfers Number of deaths per COVID-19 disinfection of vehicles,
(2019) (2019) million (2020) stations and stops

Dimension Eigenvalue Condition Variance Proportions

Index

1 4.640 1.000 0.00 0.00 0.00 0.01 0.00

2 0.225 4.546 0.00 0.02 0.00 0.67 0.07
3 0.072 8.017 0.00 0.22 0.07 0.14 0.68
4 0.048 9.828 0.02 0.64 0.40 0.12 0.00
5 0.015 17.546 0.98 0.12 0.53 0.06 0.24
Variance Inflation Factors (VIF) 49.576 15.259 24.431 4.892 11.526

existence of multicollinearity on four out of our five predictors, the 1. In order for the intercept not to be significantly different from a value
Condition Index values are below 15 on four of the five dimensions of zero and thus to be excluded from the regression model, we
examined. Especially on the 5th dimension, which is associated with a evaluated the outcomes of the Student's t-test, which according to the
Condition Index value of 17.546, the Variance Proportions for the pre­ relevant literature (Eisenhauer, 2003; Raposo, 2016) should have a
dictors do not prove the existence of multicollinearity, since there are no Sig value <0.050, i.e. not significant at the 95 % level of confidence.
pairs of variables with values above or equal to 0.90 (Belsley, 1991). As a The Sig value of 0.145 fulfils the criterion of the test and thus the
result, we may assume that the independent variables are not correlated value of the intercept can be considered as being not different from
and thus they can be included in our multiple linear regression model. zero.
For the purpose of selecting the best possible model we evaluated its 2. The confidence interval of the intercept, and more importantly the
performance with and without the inclusion of the intercept. In order to range of its values and whether it contains the value of zero, can be
assess whether the intercept can be omitted from the model, we per­ considered as a second indication regarding the ability to omit the
formed a series of statistical tests using the IBM SPSS platform (IBM intercept (Hahn, 1977; Raposo, 2016). In our tests, the confidence
Corp, 2019): interval for the intercept ranges between − 2.783 and 18.602. As a

9
A. Nikolaidou et al. Cities 134 (2023) 104206

result, the inclusion of the value of zero in the confidence interval Time or the Number of Transfers and the PT abandonment rate during
range, fulfils the second criterion and therefore it is possible for the the COVID-19 crisis.
intercept to be excluded from the model. Beyond the predictors that were found statistically important and
were included in the model, an equal meaningful observation to make is
Considering all the above, we decided to proceed with a regression which predictors were not included in the model. As far as economic
model without an intercept. The model is presented in Table 4. It has a factors go, while the Gini Index was included in our model, the GDP per
very good goodness of fit, with an R2 of 0.911, an adjusted R2 of 0.906, a capita was not, meaning that the equal distribution of resources plays a
Residual Standard Error of 4.511 and an F-statistic of 169.3. Regarding more important role in PT usage than flat economic growth indicators.
the choice of PT service attributes that were inserted in our model, we None of the social characteristics that were under consideration were
selected the corresponding attributes for the year 2019, since PT user included in the model. This means that the abandonment rate of PT was
choices during the year 2020 will be most likely affected by the PT found to be independent of the countries' average social characteristics,
operation conditions occurring during a typical year. such as population number and urban population density. Furthermore,
Our model indicates that the PT abandonment rate in the 87 cities we indicators of social prosperity and equality, such as the HDI and the GII,
examined, is positively correlated with certain PT service characteristics were not found to be statistically important showing that evidence of
(average waiting time at stops/stations and number of transfers required human development affecting the percentage of users that abandon PT
for a commute trip) and quality of service aspects (disinfection of ve­ due to a pandemic was absent in this study. These facts come in
hicles and facilities), as well as with the severity of the pandemic contradiction to existing research, since there is evidence supporting
(number of COVID-19 deaths per million residents). On the other hand, that the PT demand was affected by urban density (Montero-Lamas
higher scores of income inequality (Gini index) were associated with et al., 2022), and women, being PT captive users on a greater degree
comparatively lower PT abandonment rates during the pandemic. than men, were affected more strongly by COVID-19 (Assoumou Ella,
More in detail, the Gini Index was found to have a coefficient of 2021). Additionally, regarding the variables describing the pandemic
− 0.276, meaning that for each increased percentage of equality, 0.28 % characteristics, while the number of COVID-19 deaths did affect the
fewer PT users would have abandoned the mode during the pandemic. percentage of users who abandoned PT in our model, the number of the
The Average Waiting Time in 2019 has a coefficient of 0.339, meaning COVID-19 cases as well as the variables describing the strictness of
that for each additional minute of average waiting time, 0.34 % more PT measures taken to halt the pandemic's advance, were not found to
users would have abandoned PT during the pandemic. The Number of importantly affect it. Finally, the variables of “Average Trip Distance”
Transfers in 2019 has a coefficient of 4.346, meaning that for each and “Average Walking Distance” were not found to be statistically
additional transfer, 4.35 % of additional users would have abandoned important, in contrast to “Number of Transfers” and “Average Waiting
PT during the pandemic. The number of COVID-19 deaths per million in Time”. This could imply that the variables which maximize the exposure
2020 has a coefficient of 0.0044, meaning that 1000 additional deaths of users in stops, stations and crowded places have a heavier effect on
would increase the percentage of PT users that abandoned it due to the the likelihood to stop using PT during a pandemic.
pandemic by 4.4 %. The percentage of users that would use PT again
given better COVID-19 disinfection of vehicles, stations and stops has a 5. Conclusions
coefficient of 0.237. This means that each additional percentage of users
who stated that they would use again PT under more improved disin­ COVID-19 had a major impact on the operation of PT systems
fection conditions, increases the percentage of users that would have worldwide. The spread of the virus along with the effect of the coun­
abandoned it by 0.237 %. Although, all “COVID-19-related PT in­ termeasures which were imposed, resulted in reduced PT ridership. Our
centives” variables indicated a strong correlation with the dependent analysis attempts to shed light on the factors that averted PT users
variable, we chose the one that had the strongest correlation (r = 0.469, worldwide, from using PT altogether during the first year of the
n = 87, p = 0.000), since all the incentive variables were strongly pandemic (year of 2020). To that end, we utilized data deriving from a
intercorrelated. survey on users of the Moovit platform concerning 87 cities worldwide.
The comparison of our results with the existing literature indicates Data included elements regarding PT service attributes in 2019 (base
an agreement regarding the relationship between PT abandonment rate year) and in 2020 (first year of the pandemic), PT trip frequency rates for
and the spread of COVID-19 (Qi et al., 2021). Additionally, other studies 2020 and incentives that could bring users back to choosing PT. The
also recognize the hygiene conditions in vehicles as promoting factors of dataset was augmented with information regarding socioeconomic as­
PT usage (Labonté-Lemoyne et al., 2020). Regarding the association of pects of the examined cities as well as with data regarding the severity of
economic factors with mobility change, present research demonstrates COVID-19.
income as a determining factor for mobility change (Hu & Chen, 2021; Descriptive statistics results show a distinction in user behaviour
Sy et al., 2021), a fact that is in agreement with the results of our model, between cities in different continents. Specifically, cities in the USA and
which indicates Gini index as a contributing factor to the PT abandon­ in Europe appear to have similar percentage of users that stopped using
ment rate. To the best of our knowledge, we did not find any supporting PT during the pandemic. On the other hand, Asian cities seemed to have
or contradicting literature on the relationship between Average Waiting lower losses in ridership. Among the cities where PT use was increased,
the majority of them were located in the USA. At the same time, the
highest percentage of users that chose to abandon PT use, was observed
Table 4
in Greece and Italy. Regarding the change of PT trip characteristics
Results of the Multiple Linear Regression Model.
between 2019 and 2020, figures indicate a clear reduction in average
Variable Estimate Std. t value Pr(>|t|)
trip and walking distance. On the contrary, average commute time and
Error
the number of transfers feature marginal changes, while average waiting
Gini Index − 0.276 0.085 − 3.248 0.002** time appears to be increased in 2020, possibly due to lower PT service
Average Waiting Time 2019 0.339 0.131 2.596 0.011*
Number of Transfers 2019 4.346 1.461 2.974 0.004**
frequencies.
Number of deaths per million 2020 0.004 0.001 3.893 0.000*** Through a multiple linear regression model, we correlated the share
COVID-19 disinfection of vehicles, of users who abandoned PT during 2020, with income equality vari­
0.237 0.040 5.969 0.000***
stations and stops ables, PT service characteristics, quality of service aspects and pandemic
Residual standard error: 4.511 on 82 degrees of freedom. severity data. Empirical findings revealed that the presence of income
Multiple R-squared: 0.911, Adjusted R-squared: 0.906. inequalities resulted in lower abandonment rate, which could mirror the
F-statistic: 169.3 on 5 and 82 DF, p-value: < 2.2e-16. behaviour of captive users. Simultaneously, regarding PT trip

10
A. Nikolaidou et al. Cities 134 (2023) 104206

characteristics, results imply that PT vehicle transfers and waiting time “smaRt ECosystem foR improvEment of public trAnsporT pErformance”
are the two most critical trip segments as per user opinion, for (Project code: ΚМР6-0284565) under the framework of the Action
continuing or not using PT modes. Additionally, results confirm our “Investment Plans of Innovation” of the Operational Program “Central
initial hypotheses that: (a) in cities with higher pandemic impact (as per Macedonia 2014-2020”, that is co-funded by the European Regional
number of deaths) users were deterred to use PT and (b) improved Development Fund and Greece.
disinfection in vehicles could contribute to sustaining ridership.
Based on the empirical findings of the regression model analysis, References
specific policy implications can be derived. The methodology of this
paper aims at identifying the key factors that influenced travel behav­ Abdullah, M., Dias, C., Muley, D., & Shahin, M. (2020). Exploring the impacts of COVID-
19 on travel behaviour and mode preferences. Transportation Research
iour, in terms of abandonment of PT, during the COVID-19 era. These
Interdisciplinary Perspectives, 8(November), Article 100255. https://doi.org/10.1016/
factors may highlight suitable measures, which can be followed by PT j.trip.2020.100255
and governmental agencies, towards maintaining PT ridership during Alyavina, E., Nikitas, A., & Njoya, E. T. (2022). Mobility as a service (MaaS): A thematic
pandemic and post-pandemic circumstances. At first, the proper and map of challenges and opportunities. Research in Transportation Business &
Management, 43, Article 100783. https://doi.org/10.1016/j.rtbm.2022.100783
frequent disinfection of vehicles could enhance the perception of safety Amin, S., & Adah, J. U. (2022). COVID-19 influence on commuters’ attitude towards
to PT users and convince them to select PT. Likewise, changes in the PT riding public buses for essential trips. Cities, 131, Article 103890. https://doi.org/
network, towards the direction of providing more direct trips, can 10.1016/j.cities.2022.103890
Angell, C., & Potoglou, D. (2022). An insight into the impacts of COVID-19 on work-
contribute to limited interaction between passengers and thus minimize related travel behaviours in the Cardiff capital region and following the UK’s first
the possibility of infection. In the same sense, PT operators should national lockdown. Cities, 124. https://doi.org/10.1016/j.cities.2022.103602
ensure that waiting facilities (bus stops, metro platforms, etc.) adhere to Aparicio, J. T., Arsenio, E., & Henriques, R. (2021). Understanding the impacts of the
COVID-19 pandemic on public transportation travel patterns in the city of Lisbon.
the health and safety protocols (clean surfaces, proper ventilation, etc.), Sustainability (Switzerland), 13(15). https://doi.org/10.3390/su13158342
in order to build confidence among users that PT is not a hotbed of Assoumou Ella, G. (2021). Gender, mobility, and Covid-19: The case of Belgium. Feminist
COVID-19 contraction. Furthermore, in regions with higher income in­ Economics, 27(1–2), 66–80. https://doi.org/10.1080/13545701.2020.1832240
Basbas, S., Campisi, T., Georgiadis, G., Al-Rashid, M. A., & Tesoriere, G. (2021). COVID-
equalities, operators should expect a comparatively lower decrease in PT
19 and public transport demand trends in Sicily: Analyzing external factors and
ridership and thus, the application of PT preferential treatment mea­ governmental recommendations. European Transport - Trasporti Europei, 83. https://
sures (dedicated bus lanes, traffic signal priority, etc.) (Pulichino & doi.org/10.48295/ET.2021.83.9
Belsley, D. A. (1991). A guide to using the collinearity diagnostics. Computer Science in
Coughlin, 2005) would increase commercial PT speeds and ultimately
Economics and Management, 4(1). https://doi.org/10.1007/BF00426854
improve PT service frequencies so as to address crowding events. Campisi, T., Basbas, S., Al-Rashid, M. A., Tesoriere, G., & Georgiadis, G. (2022). A region-
Among the limitations of the present research, we mention the lack wide survey on emotional and psychological impacts of COVID-19 on public
of variables at the city level that could better capture localities. Addi­ transport choices in Sicily, Italy. Transactions on Transport Sciences, 12(3), 34–43.
https://doi.org/10.5507/tots.2021.010
tionally, more detailed information on the socioeconomic background of Christidis, P., Christodoulou, A., Navajas-Cawood, E., & Ciuffo, B. (2021). The post-
the PT users could assist on improving the robustness of our analysis. pandemic recovery of transport activity: Emerging mobility patterns and
Future research could focus on capturing any shift in mode choice due to repercussions on future evolution. Sustainability (Switzerland), 13(11). https://doi.
org/10.3390/su13116359
vaccination campaigns, whose beginnings unfortunately coincided with Creutzig, F., Javaid, A., Soomauroo, Z., Lohrey, S., Milojevic-Dupont, N.,
the end of the survey presented in our paper. Ramakrishnan, A., Sethi, M., Liu, L., Niamir, L., Bren d’Amour, C., Weddige, U.,
Lenzi, D., Kowarsch, M., Arndt, L., Baumann, L., Betzien, J., Fonkwa, L., Huber, B.,
Mendez, E.Zausch, J. M., … (2020). Fair street space allocation: Ethical principles
CRediT authorship contribution statement and empirical insights. Transport Reviews, 40(6), 711–733. https://doi.org/10.1080/
01441647.2020.1762795
Anastasia Nikolaidou: Conceptualization, Methodology, Investi­ Data Europa EU. (2019). Moovit: A global transport application | data.europa.eu. http
s://data.europa.eu/en/news-events/news/moovit-global-transport-application.
gation, Writing – original draft, Visualization, Formal analysis, Writing – Demographia. (2021). Demographia world urban areas (built up urban areas or world
review & editing. Aristomenis Kopsacheilis: Conceptualization, aggglomerations).
Methodology, Investigation, Writing – original draft, Visualization, Dong, H., Ma, S., Jia, N., & Tian, J. (2021). Understanding public transport satisfaction in
post COVID-19 pandemic. Transport Policy, 101, 81–88. https://doi.org/10.1016/j.
Formal analysis, Writing – review & editing. Georgios Georgiadis:
tranpol.2020.12.004
Conceptualization, Methodology, Formal analysis, Investigation, Project Downey, L., Fonzone, A., Fountas, G., & Semple, T. (2022). The impact of COVID-19 on
administration, Supervision, Validation, Writing – original draft, future public transport use in Scotland. Transportation Research Part A: Policy and
Writing – review & editing. Theodoros Noutsias: Conceptualization, Practice, 163, 338–352. https://doi.org/10.1016/j.tra.2022.06.005
Dueñas, M., Campi, M., & Olmos, L. E. (2021). Changes in mobility and socioeconomic
Investigation. Ioannis Politis: Conceptualization, Funding acquisition, conditions during the COVID-19 outbreak. Humanities and Social Sciences
Methodology, Project administration, Supervision, Validation, Writing – Communications, 8(1). https://doi.org/10.1057/s41599-021-00775-0
review & editing. Ioannis Fyrogenis: Conceptualization, Writing – Eisenhauer, J. G. (2003). Regression through the origin. Teaching Statistics, 25(3), 76–80.
https://doi.org/10.1111/1467-9639.00136
original draft, Investigation. Fernández Pozo, R., Wilby, M. R., Vinagre Díaz, J. J., & Rodríguez González, A. B.
(2022). Data-driven analysis of the impact of COVID-19 on Madrid’s public transport
Declaration of competing interest during each phase of the pandemic. Cities, 127. https://doi.org/10.1016/j.
cities.2022.103723
Franke, G. R. (2010). Multicollinearity. In Wiley international encyclopedia of marketing.
The authors declare that they have no known competing financial John Wiley & Sons, Ltd. https://doi.org/10.1002/9781444316568.wiem02066.
interests or personal relationships that could have appeared to influence Habib, M. A., & Anik, M. A. H. (2021). Impacts of COVID-19 on transport modes and
mobility behaviour: Analysis of public discourse in twitter. Transportation Research
the work reported in this paper. Record: Journal of the Transportation Research Board, 036119812110299. https://doi.
org/10.1177/03611981211029926
Data availability Hahn, G. J. (1977). Fitting regression models with no intercept term. Journal of Quality
Technology, 9(2), 56–61. https://doi.org/10.1080/00224065.1977.11980770
Hale, T., Angrist, N., Goldszmidt, R., Kira, B., Petherick, A., Phillips, T., Webster, S.,
The authors would like to acknowledge Moovit for making the data Cameron-Blake, E., Hallas, L., Majumdar, S., & Tatlow, H. (2021). A global panel
of their mobility survey, upon which the present research was based on, database of pandemic policies (Oxford COVID-19 government response Tracker).
Nature Human Behaviour, 5(4), 529–538. https://doi.org/10.1038/s41562-021-
available. The data were used with respect to the Creative Commons
01079-8
Attribution 4.0 International License. Mathieu, E., Ritchie, H., Rodés-Guirao, L., Appel, C., Giattino, C., Hasell, J.,
Macdonald, B., Dattani, S., Beltekian, D., Ortiz-Ospina, E., & Roser, M. (2020).
Acknowledgements Coronavirus Pandemic (COVID-19). Published online at OurWorldInData.org.
Retrieved from: ’https://ourworldindata.org/coronavirus’.
Heiskala, M., Jokinen, J.-P., & Tinnilä, M. (2016). Crowdsensing-based transportation
Part of this study was implemented as part of the RECREATE project: services — An analysis from business model and sustainability viewpoints. Research

11
A. Nikolaidou et al. Cities 134 (2023) 104206

in Transportation Business & Management, 18, 38–48. https://doi.org/10.1016/j. Politis, I., Georgiadis, G., Papadopoulos, E., Fyrogenis, I., Nikolaidou, A.,
rtbm.2016.03.006 Kopsacheilis, A., Sdoukopoulos, A., & Verani, E. (2021). COVID-19 lockdown
Hu, S., & Chen, P. (2021). Who left riding transit? Examining socioeconomic disparities measures and travel behaviour: The case of Thessaloniki, Greece. Transportation
in the impact of COVID-19 on ridership. Transportation Research Part D: Transport and Research Interdisciplinary Perspectives, 10(January), Article 100345. https://doi.org/
Environment, 90. https://doi.org/10.1016/j.trd.2020.102654 10.1016/j.trip.2021.100345
Huang, Y., & Li, R. (2022). The lockdown, mobility, and spatial health disparities in Przybylowski, A., Stelmak, S., & Suchanek, M. (2021). Mobility behaviour in view of the
COVID-19 pandemic: A case study of New York City. Cities, 122, Article 103549. impact of the COVID-19 pandemic-public transport users in gdansk case study.
https://doi.org/10.1016/j.cities.2021.103549 Sustainability (Switzerland), 13(1), 1–12. https://doi.org/10.3390/su13010364
IBM Corp. (2019). IBM SPSS statistics for windows, version 26.0. In 2019. Pulichino, M., & Coughlin, J. F. (2005). Introducing transit preferential treatment: Is a
Jobson, J. D. (1991). Multiple linear regression. In Applied multivariate data analysis: political maverick necessary for public transportation to Innovate? Journal of Urban
Regression and experimental design (pp. 219–398). New York: Springer. https://doi. Planning and Development, 131(2), 79–86. https://doi.org/10.1061/(ASCE)0733-
org/10.1007/978-1-4612-0955-3_4. 9488(2005)131:2(79)
Kim, S., Lee, S., Ko, E., Jang, K., & Yeo, J. (2021). Changes in car and bus usage amid the Qi, Y., Liu, J., Tao, T., & Zhao, Q. (2021). Impacts of COVID-19 on public transit
COVID-19 pandemic: Relationship with land use and land price. Journal of Transport ridership. International Journal of Transportation Science and Technology. https://doi.
Geography, 96. https://doi.org/10.1016/j.jtrangeo.2021.103168 org/10.1016/j.ijtst.2021.11.003
Kłos-Adamkiewicz, Z., & Gutowski, P. (2022). The outbreak of COVID-19 pandemic in Raposo, F. (2016). Evaluation of analytical calibration based on least-squares linear
relation to sense of safety and mobility changes in public transport using the example regression for instrumental techniques: A tutorial review. TrAC Trends in Analytical
of Warsaw. Sustainability (Switzerland), 14(3). https://doi.org/10.3390/su14031780 Chemistry, 77, 167–185. https://doi.org/10.1016/j.trac.2015.12.006
Kumar, P., Khani, A., Lind, E., & Levin, J. (2021). Estimation and mitigation of epidemic Roback, P., & Legler, J. (2021). Beyond multiple linear regression. Beyond multiple linear
risk on a public transit route using automatic passenger count data. Transportation regression. https://doi.org/10.1201/9780429066665
Research Record, 2675(5), 94–106. https://doi.org/10.1177/0361198120985133 Roberts, J. D., Dickinson, K. L., Koebele, E., Neuberger, L., Banacos, N., Blanch-
Kyriakidis, C., Chatziioannou, I., Iliadis, F., Nikitas, A., & Bakogiannis, E. (2023). Hartigan, D., Welton-Mitchell, C., & Birkland, T. A. (2020). Clinicians, cooks, and
Evaluating the public acceptance of sustainable mobility interventions responding to cashiers: Examining health equity and the COVID-19 risks to essential workers.
Covid-19: The case of the great walk of Athens and the importance of citizen Toxicology and Industrial Health, 36(9), 689–702. https://doi.org/10.1177/
engagement. Cities, 132, Article 103966. https://doi.org/10.1016/j. 0748233720970439
cities.2022.103966 Santos, G., & Nikolaev, N. (2021). Mobility as a service and public transport: A rapid
Labonté-Lemoyne, É., Chen, S. L., Coursaris, C. K., Sénécal, S., & Léger, P. M. (2020). The literature review and the case of moovit. Sustainability, 13(7), 3666. https://doi.org/
unintended consequences of covid-19 mitigation measures on mass transit and car 10.3390/su13073666
use. Sustainability (Switzerland), 12(23), 1–13. https://doi.org/10.3390/su12239892 Semple, T., Fonzone, A., & Fountas, G. (2022). The impact of COVID-19, climate change
Lucchesi, S. T., Tavares, V. B., Rocha, M. K., & Larranaga, A. M. (2022). Public transport and working from home on travel choices.
COVID-19-safe: New barriers and policies to implement effective countermeasures Sy, K. T. L., Martinez, M. E., Rader, B., & White, L. F. (2021). Socioeconomic disparities
under User’s safety perspective. Sustainability (Switzerland), 14(5). https://doi.org/ in Subway use and COVID-19 outcomes in New York City. American Journal of
10.3390/su14052945 Epidemiology, 190(7), 1234–1242. https://doi.org/10.1093/aje/kwaa277
Medlock, K. B., Temzelides, T., & Hung, S. Y.(. E.). (2021). COVID-19 and the value of Tirachini, A., & Cats, O. (2020). COVID-19 and public transportation: Current
safe transport in the United States. Scientific Reports, 11(1). https://doi.org/10.1038/ assessment, prospects, and research needs. Journal of Public Transportation, 22(1),
s41598-021-01202-9 1–34. https://doi.org/10.5038/2375-0901.22.1.1
Monahan, T., & Lamb, C. G. (2022). Transit’s downward spiral: Assessing the social- Ton, D., Arendsen, K., de Bruyn, M., Severens, V., van Hagen, M., van Oort, N., &
justice implications of ride-hailing platforms and COVID-19 for public transportation Duives, D. (2022). Teleworking during COVID-19 in the Netherlands: Understanding
in the US. Cities, 120, Article 103438. https://doi.org/10.1016/j.cities.2021.103438 behaviour, attitudes, and future intentions of train travellers. Transportation Research
Monterde-I-bort, H., Sucha, M., Risser, R., & Kochetova, T. (2022). Mobility patterns and Part A: Policy and Practice, 159, 55–73. https://doi.org/10.1016/j.tra.2022.03.019
mode choice preferences during the COVID-19 situation. Sustainability (Switzerland), Torbacki, W. (2021). Achieving sustainable mobility in the Szczecin metropolitan area in
14(2). https://doi.org/10.3390/su14020768 the post-COVID-19 era: The DEMATEL and PROMETHEE II approach. Sustainability
Montero-Lamas, Y., Orro, A., Novales, M., & Varela-García, F. A. (2022). Analysis of the (Switzerland), 13(22). https://doi.org/10.3390/su132212672
relationship between the characteristics of the areas of influence of bus stops and the Tzamourani, E., Tzouras, P. G., Tsigdinos, S., Kosmidis, I., & Kepaptsoglou, K. (2022).
decrease in ridership during COVID-19 lockdowns. Sustainability (Switzerland), 14 Exploring the social acceptance of transforming urban arterials to multimodal
(7). https://doi.org/10.3390/su14074248 corridors. The case of panepistimiou avenue in Athens. International Journal of
Moovit. (2021). Moovit unveils its 2020 global public transport report. https://moovit.co Sustainable Transportation, 1–15. https://doi.org/10.1080/15568318.2022.2037793
m/press-releases/2020-global-public-transport-report/. United Nations Development Programme. (2020). The next frontier: Human
Munawar, H. S., Khan, S. I., Qadir, Z., Kouzani, A. Z., & Mahmud, M. A. P. (2021). Insight development and the anthropocene. Human Development Report, 2020, 130–158.
into the impact of COVID-19 on Australian transportation sector: An economic and van Dorn, A., Cooney, R. E., & Sabin, M. L. (2020). COVID-19 exacerbating inequalities in
community-based perspective. Sustainability (Switzerland), 13(3), 1–24. https://doi. the US. The Lancet, 395(10232), 1243–1244. https://doi.org/10.1016/s0140-6736
org/10.3390/su13031276 (20)30893-x
Nello-Deakin, S. (2019). Is there such a thing as a ‘fair’ distribution of road space? Vickerman, R. (2021). Will Covid-19 put the public back in public transport? A UK
Journal of Urban Design, 24(5), 698–714. https://doi.org/10.1080/ perspective. Transport Policy, 103, 95–102. https://doi.org/10.1016/j.
13574809.2019.1592664 tranpol.2021.01.005
Nikitas, A., Tsigdinos, S., Karolemeas, C., Kourmpa, E., & Bakogiannis, E. (2021). Cycling Wilbur, M., Ayman, A., Ouyang, A., Poon, V., Kabir, R., Vadali, A., Pugliese, P.,
in the era of COVID-19: Lessons learnt and best practice policy recommendations for Freudberg, D., Laszka, A., & Dubey, A. (2020). Impact of COVID-19 on public transit
a more bike-centric future. Sustainability, 13(9), 4620. https://doi.org/10.3390/ accessibility and ridership.
su13094620 World Bank. (2021a). GDP per capita. https://data.worldbank.org/indicator/NY.GDP.
Pawar, D. S., Yadav, A. K., Akolekar, N., & Velaga, N. R. (2020). Impact of physical PCAP.CD.
distancing due to novel coronavirus (SARS-CoV-2) on daily travel for work during World Bank. (2021b). Gini index. https://data.worldbank.org/indicator/SI.POV.GINI.
transition to lockdown. Transportation Research Interdisciplinary Perspectives, 7, Worldpopulationreview.com. (2021). World city populations 2021. https://worldpopu
Article 100203. https://doi.org/10.1016/j.trip.2020.100203 lationreview.com/world-cities.
Politis, I., Georgiadis, G., Nikolaidou, A., Kopsacheilis, A., Fyrogenis, I., Zhang, N., Jia, W., Wang, P., Dung, C.-H., Zhao, P., Leung, K., Su, B., Cheng, R., & Li, Y.
Sdoukopoulos, A., Verani, E., & Papadopoulos, E. (2021). Mapping travel behaviour (2021). Changes in local travel behaviour before and during the COVID-19 pandemic
changes during the COVID-19 lock-down: A socioeconomic analysis in Greece. in Hong Kong. Cities, 112, Article 103139. https://doi.org/10.1016/j.
European Transport Research Review, 13(1). https://doi.org/10.1186/s12544-021- cities.2021.103139
00481-7

12