Impact of COVID-19 pandemic on the people’s choice of urban public transportation modes and mobility in Addis Ababa and Hawassa city, Ethiopia

ABSTRACT

The aim of this study is to investigate the impacts of COVID-19 pandemic on public transportation mode choices in Ethiopia. Protection motivation theory and multinomial regression model were applied to estimate the mode choice behavior during the pandemic, compared to the pre-COVID period. Results indicated that public transport modes are the dominant modes before and during the pandemic but the choice of these modes declined during the pandemic by about 33 percent. Respondents placed more priority on pandemic-related risk, cleanliness, social-physical distancing, good airflow, and travel cost when deciding mode choice during the pandemic, compared to the pre-COVID period. Pandemic-related factors, gender, income, distance of residences, and travel cost are the significant predictors of mode choice. Passengers are also likely to choose a lesser amount of public transportation whereas more choice for alternative modes. The reasons are the passengers’ perception of travel risks, greater exposure to infections; shortage of infection prevention services; unaffordable service fees; disruption of services, being made out of work and working from home. Overall, the study could inform that COVID-19 pandemic is really impacting the mode choice behaviors of urban people. It will imply policymaking to prepare pandemic-sensitive and adaptive transport systems during & post-pandemic period.

KEYWORDS:

• COVID-19
• public transportation
• mode choice behavior
• protection motivation theory
• urban mobility
• transport policy
• urban planning

1. Introduction

The basic types of mobility in urban areas include public transportation such as trains, city-buses, and mini-buses plus 3-wheel rickshaw or Bajaj; private transportation; freight transportation, and non-motorized transport like walking (Jean-Paul et al., 2017). Sustainable transportation is widely supported through a shift from private transportation to public transportation (PT) mainly for environmental, economic, and social sustainability goals (Berechman et al., 2006; J. Zhang, 2020; Litman, 2016).

Enormous efforts have been invested to encourage the use of PT modes such as trains, city buses, mini-buses, and multimodal transport. However, Ozbilen et al. (2021) confirmed that the global spread of COVID-19 and related measures such as travel restrictions have exerted negative pressures on public transport systems in urban areas to date. Currently, in many urban areas of the world transport and travel plans have been strongly disrupted (Brianne, 2020; De Vos, 2020; Jakkie et al., 2020).

There are several studies such as Hotle et al. (2020) and Parady et al. (2020) that show the use of overall transportation, specifically public transportation, highly reduced during the pandemic eras compared to private and non-motorized transportations. From a theoretical perspective, it is also obvious that people perceive a similar reduction in mode choices and travel due to COVID-19 in developing countries.

However, to the authors’ knowledge, even though these all impacts mentioned in PMT are at least theoretically acceptable, only a small number of transport studies have practically explored the potential for travelers’ psychological factors (i.e. fear of pandemic and travel risks) and consequent motivation to adopt protective measures or behavioral intentions (i.e. inclination to avoid travel, continue or reduce transit choice) to alter their transit choice behavior during the pandemic.

In the field of transportation, particularly in developing countries, critical questions remain unanswered, such as: which particular PT mode choice is reduced now and in the near future? How are the perceived COVID-19 impacts and PT choice reductions actually correlated during and in the post-COVID scenarios? Which kinds of travellers are more affected, i.e. based on their demographic and economic characteristics? And how public transport choice is impacted from the context of protection motivation theory?

Besides, in the contemporary world, research is highly required to focus on the issues and questions such as: What are the determinants or factors affecting the public choice of transport during the pandemic period? To what extent COVID-19 pandemic affect future travel behavior and transport mode choice? Accordingly, these critical questions and surprising things motivated this study to address and inform based on the protection motivation theory (PMT). The study aimed to address this research gap, particularly in developing countries using PMT. Considering the application of this theory in previous studies and bearing in mind the given potential for COVID-19 pandemic-related perceptions of risk, fear, and protective measures to affect people’s choices, PMT model has a good fit for this study. Additionally, PMT and multinomial regression models were estimated considering attitudes towards the pandemic and public transportation choices under different travel contexts such as pre- and during-COVID scenarios and in situations where personal and government safety measures have been implemented or not as responses against the pandemic. Based on this, the study models will shed light on the fitness of PMT to explore the implications of the pandemic and implementation of safety policies on public transportation choices during the pandemic.

As a result, the evidences will contribute to policymakers better understand the specific PT mode choices for improved policy formulations for the post COVID-19 world considering the socio-economic and demographic characteristics of travellers as well as the unique constraints of the developing world. This research recognized that the countries especially in the sub-Saharan Africa have unique socio-economic and social constraints and require special transport-related policies to address mobility issues during a pandemic situation. Thus, the motivation of the research is to investigate the impacts from the perspective of developing countries and fill the gap in the literature by providing insights into how public transport mode choices are impacted and changed due to the COVID-19 pandemic in developing countries like Ethiopia.

Thus, this study aimed at understanding the impacts of COVID-19 on the public transport mode choice decision and travel behavior during the COVID-19 pandemic relative to the pre-COVID-19 pandemic period in the major urban areas of Ethiopia. To this end, it used new and most up-to-date theory and variables including perception of travel risk of COVID-19 infection as well as the degree of sensitivity to various PT attributes which together affect the decision whether to continue using the public transportation modes in comparison with the pre-COVID period. It investigated the determinants and impacts of COVID-19 on the public transport mode choice mainly for workplace destinations in two big cities of Ethiopia such as Addis Ababa and Hawassa.

The modeling framework for this contemporary topic of investigation allows the study to use recent theoretical foundations and advanced analysis models. The modeling framework is generally composed of three steps. Initially, the research applied the protection motivation theory as a theoretical foundation and conceptual framework to identify the new and COVID-19-related transport mode choice determinant factors together with the conventional ones. Second, the study collected primary data of 763 observations from passengers through a survey approach as input for multi-criteria impact analysis. The Wilcoxon signed-rank test was then used to analyze the underlying factors significantly affecting mode choice before and during the pandemic. Third, a multinomial logistic regression model was applied to model the data sample and the impact of the predictor variables on transport mode choice.

Following the PMT, the present study advances a deeper understanding of the relationships between the perceived COVID-19 impacts and PT choice reductions to understand the travellers’ perceptions of travel during and in the post-COVID-19 pandemic and offers practical policy implications to revive the PT choices and usage in the market. A sound understanding of travel-mode choice determinants is needed to design interventions to slow down and prevent the spread of the COVID-19 pandemic with sustainable mobility through consistent or adaptable strategies. Therefore, the current study also tries to bridge the research gap in this emerging topic and contributes to protection motivation theory and existing literature.

As the current study focused on an emerging issue, i.e. impacts of COVID-19 pandemic and transport mode choice based on protective motivation theory in Ethiopia, the empirical findings can be informative for charting out strategic interventions for urban transport to prevent the impacts of the spread of the COVID-19 pandemic. It can also inform the recent developments, trends, new challenges, and opportunities for policymakers in designing innovative pathways to meet the transport needs of people.

Therefore, the motivation of this study is to find pieces of evidence and answer the following state-of-the-art research questions:

1. To what extent COVID-19 pandemic changed the travel behavior of public transport users in Addis Ababa and Hawassa City

2. How were the public transport mode choice decisions affected and changed during the pandemic period?

3. How COVID-19 pandemic affect the anticipated transport mode choices and shifts in the future post-COVID period (in 18–24 months)?

• Are respondents likely to choose or shift public transport modes in the future?

• What are the anticipated levels of mode choice for the public transport modes compared to the alternative modes under the status quo of current policy?

• What are the justifications for the anticipated choice and use of public transport modes?

The rest of the article is organized into various sections such as literature review, theoretical and conceptual framework; material and methods; results, discussions; conclusion, and policy implications.

2. Literature review

2.1. Theoretical framework

Globally, governments have implemented new policies such as travel restrictions, temporarily closed businesses and schools, and instituted social distancing as responses to the threat posed by COVID-19 (Yezli & Khan, 2020). Even if the intention and implementation of these policies were to control the pandemic and protect public health, they also have the potential to affect people’s perceptions of the pandemic-related risks (Zheng et al., 2021). Literatures have shown that the pandemic and related policies, latent attitudinal factors and perceptions can affect transport mode choice of people, mainly public transit choices and uses (Nikolaidou et al., 2023; Oum & Wang, 2020).

To identify the major determinants of public choice of transport mode during the pandemic, an array of new variables or pandemic-related variables (example, cleanliness, infection concern, and social physical distance) were used. These are perceptions of COVID-19 pandemic effects, fear and perception of travel risks which affect the decision of whether to continue using public transportation modes, together with the widely known conventional variables. For this purpose, the protection motivation theory (PMT) was chosen as a theoretical framework and foundation for conceptual framework of this study. Thus, the new variables and indicators that may affect the public’s choice of transport mode are developed from the principles of this up-to-date theory. As PMT is nowadays gaining wider support due to its better relevance to the studies focusing on the impacts of pandemics and risks such as COVID-19, it is found to be more suitable for the evaluation of COVID-19 pandemic impacts particularly on public transport mode choice in the urban transport sector.

According to Cox et al. (2004) and Rogers (1983) protection motivation theory is a major social psychological model initially designed by Ronald Rogers in 1970s and improved later in 1983. It explains how individuals cognitively perceive or assess any risk that threatens their lives and how they adopt self-protective behaviors or measures. The three components of the PMT model are attitude change, cognitive mediating processes, and fear appeal. In order to influence people to change their behavior, the fear appeal the harmful implications of a certain issue. Risk appraisal and risk coping appraisal are the two different cognitive processes that make up the cognitive mediating processes. People assess their risk of suffering from the threat and the threat’s seriousness during the risk appraisal process. It states that how seriously people perceive the threat affects their cognition and conduct. In the risk coping appraisal, individuals take preventive strategies that eliminate or reduce the recurrence of threats and develop confidence in their ability to adhere to preventive measures through three cognitive processes, including response efficacy, self-efficacy, and response-cost. Therefore, risk and coping appraisals as mediators could influence the relationship between risk information and protective behavior (Norman et al., 2005; Rogers, 1975).

Risk assessment indicates that people have a higher possibility of contracting COVID-19 during the pandemic. In the risk appraisal, people judge the seriousness of the travel risk and the likelihood of contracting COVID-19. Therefore, people engaged in three cognitive processes to choose preventative actions that lower the risk of contracting COVID-19 during the risk coping evaluation. People first perceive avoiding health and safety concerns through reaction efficacy. Second, people’s perception that they can successfully prevent health hazards is based on their sense of self-efficacy. Third, response-cost describes the additional resources needed to ward off COVID-19. Therefore, the individuals’ perceptions of risk appraisal and risk coping measures lead to their public transport choice reduction, commute disruptions and avoidance of travel (Rogers, 1983).

Furthermore, Rosenstock (1990) showed that the perceived risk, considered as a predictor of intentions and behavioral change, is a foundation of the Health Belief Model, which explains the likelihood of engagement in health-promoting behavior in response to stimuli or cues to action. In general, this theory accounts for how individuals change perceptions, attitudes, and behavioral styles when interpreting a threat appraisal and reacting to a coping appraisal to fear appeals and stressful stimuli such as COVID-19 pandemic on the public transport mode choice in this study.

There are studies that applied protection motivation theory. Wang et al. (2021) and Zheng et al. (2021) in their tourism-related research showed that in response to a public health threat, individuals may develop perceptions of the risks and turn to adaptive coping mechanisms that aim to protect against threats based on PMT. Besides, there are some studies that applied PMT in transport fields such as Mashrur et al. (2022) to analyze the impact of the Pandemic on the anticipated transit usage in Post-pandemic period particularly in Canada and by Harbeck et al. (2018) to investigate young drivers’ perceived risk and risky driving. A. Chen and Lu (2021) also applied it to explore the mechanisms of passengers’ protective behavior in ride-sharing. Lu and Wei (2019) widely applied this theory to evaluate individuals’ perception of risks such as traffic accident.

However, this theory was not primarily applied in studies of public transportation choice and from the perspective of developing countries. Thus, the current study was highly motivated to apply it to this transport topic, particularly to determine the factors related to the threats of the pandemic that could affect public transport choice decisions during the pandemic era in developing nations like Ethiopia, considering the given potential for COVID-19 pandemic-related perceptions of risk, fear, and protective measures to affect the choice of public transit.

Since the theory has been successfully applied to several health promotion activities and healthy lifestyles (Cox et al., 2004; Rivera, 2004; Rogers, 1983), for this health and transport-related study, the modified PMT model was thus applied as framework to assess transport mode choice behavior and intention with regard to COVID-19 pandemic-related attributes (Figure 3).

The application of PMT was to quantify, estimate, and understand the impact of COVID-19-related factors (e.g. the perceived infection fear, travel risks, and safety) on the choice of passengers for transit modes during and in future post-pandemic periods. The major role of PMT in this study was to serve as framework to explore the relationships among pandemic fear, travel risks, protective responses, and changes in transit choice. Thus, it was applied to examine the behavioral consequences of pandemic fear, travel risks, and protective measures taken by passengers on their public transit choice decisions.

In addition to the current impacts, this study applied the PMT to predict the impacts of COVID-19 and related government measures on the PT mode choice intentions in the short-term of the future scenario. The basic concepts of PMT, an individual’s risk assessment model, applied in this study to assess the links and impacts of COVID-19 pandemic on the public transport mode choice using the conceptual framework indicated in Figure 1 and mixed research design.

Figure 1. Public transportation network of Addis Ababa city mainly city-buses and minibus taxis.

Source: Adapted from Addis Ababa city structure and master plan (2017–2032).

2.2. Impacts of COVID-19 and subsequent measures on public transit choices

With each passing day, reports on rising total confirmed cases of COVID-19 and deaths continue to dominate the global conscience, and the virus is now present on every continent. As thousands of people have perished as the effects of COVID-19 touch us all, the resulting fear is more pervasive. Concerning the pathogens’ ability to travel, buses and trains are, of course, excellent ways of spreading infections-the study on the infection of 9 passengers on a long-distance bus in Huwan in January became famous in this respect; the study has been denied meanwhile without giving reasons. In many Asian cities, such as Wuhan, Huanggang, or Delhi, public transport was suspended to contain the virus. Even though the total shut down of public transport systems is not a measure taken by all cities which are affected by the epidemic, it is important to systematically identify areas of action to minimize the risks for public transport staff and passengers (Rudy, 2020; World Health Organization [WHO], 2020).

The response by governments, companies, and communities to the pandemic has suddenly impacted our way of life and our local, regional, and global transportation systems (Ibold et al., 2020).

We can see the impacts of the pandemic and the consequences of measures to stop or slow the spread of the virus worldwide. Impacts are felt in various fields such as urban transport, the economy, social behavior, climate, and urban environments (International Organization for Migration [IOM], 2020; Timothy, 2020). As a response to COVID-19, millions of people around the world were quarantined in their homes, while infrastructure and transportation systems that bonded us globally, nationally, and locally are being used more thinly these days, mainly in urban areas (Abdullah et al., 2020; Rudy, 2020).

A study by Basbas et al. (2021), Tsavdari et al. (2022) and Roger (2020) show that transport usage is and will change after partial and complete lockdown measures due to the COVID-19 pandemic. This study, for example, shows that the number of people using public transport in Britain’s cities could be 20 percent lower than normal after the end of coronavirus lockdown. In London, commuters using buses and tubes could go down by as much as 40% from pre-lockdown levels. According to a survey conducted by transport consultants SYSTRA, rail use could drop by 27%. There could also be a boom in walking and cycling in a population that may be more interested in health messages. Such results are bad news for the government, which wants more people to use public transport to cut carbon emissions that are fuelling climate heating or global warming.

Baruch (2020) and Moslem et al. (2020) showed that COVID-19 has caused the intra-city mobility landscape and most transit agencies to enter crisis mode, mainly in the US. Ridership has dropped by 50 to 95 percent in most major cities during the shutdown. Though transit ridership has been declining since 2014, long before COVID-19, every major transit agency in the country lost ridership between 2018 and 2019. Per capita transit ridership has dropped even faster, dropping from 287 trips per urban resident to 38 trips per urban resident in recent times. Despite the 87 percent drop in ridership, New York’s Metropolitan Transportation Authority (MTA) has cut service by only 25–35 percent. The Chicago Transit Authority recently saw its ridership drop by 82 percent, but it has no plans to reduce service. The Metropolitan Atlanta Rapid Transportation Authority (MARTA) cut bus service by 20% after ridership dropped by 55%. Among large transit agencies, only the Washington Metropolitan Area Transit Authority (WMATA) has made drastic reductions. It is now operating on its modified Sunday schedule on weekdays and operating only 27 ‘lifeline’ bus routes on weekends (Okunlola et al., 2020).

Besides, transportation agencies are also spending more resources sanitizing buses and trains. Many transit operators are boarding bus passengers through the back door to protect drivers, foregoing the fare box revenue that is typically collected at the front door. As transit managers are focused on short-term day-to-day operations as opposed to long-term strategies, many public transit agencies of the 21st century cling to 20th-century worlds and operations. And their short-term approach focuses on operating or provision of as much service as possible, sometimes at the expense of their employees’ health and agencies’ bottom lines (Abdullah et al., 2020; De Vos, 2020; Mogaji, 2020).

As movement becomes restricted, public transit services are asked to reduce their carrying capacity and daily service time. Public transit use and ridership are greatly reduced, and then urban shared mobility is impacted for social distancing reasons. As cities are limiting access to the public transit system, residents are going back to basics. They are also pointing out to their city leaders that they have too narrow sidewalks and cannot maintain the minimum social distance when out for walking. Some city leaders, like in Bogota, Colombia, New York City, and others, are using this opportunity to create emergency lanes by re-appropriating what are now empty roads and making space so that walking and cycling are the preferred ways to get around (Jain & Tiwari, 2019; Koehl, 2020).

It is reported that transit ridership dropped by as much as 90% because of pandemic-related factors, including the perceived risk and fear of getting infected while using transit and government response measures such as travel restrictions (Cohen, 2020; Pakpour & Griffiths, 2020).

Most of the studies conducted on COVID-19 and its impact on passenger travel demand such as Beck et al. (2020) and Wang, Liu, et al. (2021) have focused on the alteration of travel behaviors caused by the pandemic and confirmed that the pandemic had changed the behaviors of travelers worldwide.

It is important to keep in mind that the pandemic fear, travel health and safety risks as well as response measures enacted now aiming to mitigate the pandemic impacted the mode choice decisions and commuters differently depending on their gender, job type and income. The consequences are not distributed evenly: transportation-disadvantaged groups such as urban poor or lower income, women, the elderly, and people with disabilities may be disproportionately affected (Brianne, 2020; Ibold et al., 2020; TUMI Initiative, 2020).

Many travel-mode choice studies such as Adeel et al. (2016) exist in Pakistan, but those studies were before the COVID-19 outbreak or under normal conditions, not during a pandemic. These studies disclosed that gender has a relationship with travel-mode choice. As compared to females, males were most likely to travel to workplaces and markets by 2 and 3 wheelers, followed by active modes, and least likely to travel by car. Unlike in the pre-COVID-19 study, men were more likely to walk and use public transportation (Abdullah et al., 2021).

Rith et al. (2019) showed that males had a higher baseline preference than females to travel to work by motorcycle during the COVID-19 outbreak in Metro Manila, the Philippines. Females were more likely than males to stay at home rather than go to workplaces and markets. A similar finding was reported for a random day pre-COVID-19 that females were more likely to stay at home than travel, probably due to societal norms and cultural values. Also, women tended to participate less frequently in out-of-home activities (Adeel et al., 2016). Gender had no significant relationship with the travel-mode choice for the hospital destination, whereas the workplace and market destinations did.

A study by Shah et al. (2021) revealed that the proportion of public transport mode was minor for the three key destinations such as workplace, market, and hospital, in Islamabad, Pakistan, during the COVID-19 outbreak. Large cities in Pakistan have many travel-mode choices, and travelers prefer more comfortable modes at a higher travel cost, i.e. Light Rail Transit. It showed that male passengers were most likely to travel by 3 wheelers and least likely to travel by car to the workplace.

Those who were likely to stay at home include females, unemployed persons, and students. Self-employed and government employed people were found to be most likely to go by car. Therefore, students were likely to study online at home during school closures (Rith & Piantanakulchai, 2020; Shah et al., 2021).

Adeel et al. (2014) and Rith et al. (2019) revealed that respondents who belong to families with a household income of more than 500 USD/month were more likely than respondents with a lower household income to travel to hospitals by car. People with high household incomes had a higher baseline of car ownership. The reason that few people living in towns or rural areas use public transport is the limited availability of public transport. Furthermore, the ban on intra-city and inter-city movement of passengers by public transport (except for the metro bus) is another major reason.

In addition to direct and current demand reductions for particular PT modes, findings also show that people often adopt new and other transport practices (such as motorbikes, Bajaj, walking, and auto) during and immediately following a pandemic, which can become permanent, depending on a range of factors.

Thus, it will have an indifferent or even negative effect on sustainable mobility later unless measures are also targeted specifically. Otherwise, the legacy of the COVID-19 shock will damage economies and mobility systems. The pandemic offers us a unique once-in-a-lifetime opportunity to re-imagine the transportation system and move it towards a more resilient, equitable, and seamless experience (IOM, 2020; Timothy, 2020).

3. Material and methods

A pragmatic research paradigm and mixed research approach are ideal for this study, providing both quantitative and qualitative evidence. Particularly, a sequential mixed approach was more suitable to obtain different but complementary data on the topic than using a single research approach.

3.1. Description of the study areas

3.1.1. Addis Ababa city and its public transport network

Ethiopia is located in the Horn of Africa; the country has around 105 million inhabitants and is characterized as being predominantly rural, with about 20% of its population living in urban areas. However, the country’s urban population will triple in absolute numbers between 2010 and 2040 because of the rapidly accelerating urbanization, which is mainly driven by continued migration to urban areas. The capital Addis Ababa accounts for a large part of this urbanization trend, but there are another seven urban growth poles with a population of over 150,000 people, such as Hawassa city (UN-Habitat, 2020).

According to Wubneh (2013), Addis Ababa is the national capital city of Ethiopia. It holds about 527 square kilometers of area and is currently subdivided into eleven administrative sub-cities. It is one of the most rapidly urbanizing cities in Africa, mainly due to internal migration from rural areas. For example, the population in 2021 was 5,006,000, a 4.42% increase from 2020. Though a recent phenomenon, this internal migration is expected to increase significantly during the years to come due to the fact that Addis Ababa is not only a financial and political center but also a diplomatic capital. Accordingly, the city is burdened by about four basic transport-related multi-dimensional problems and challenges. Among these problems, the huge gap between transport demand and supply is the primary challenge. The other challenges include problems with road traffic safety, traffic congestion, and air pollution. Regarding the demand-supply gaps, the daily transport trip demand of the city is about 4.5 million, but only about 3.2 million can be supplied so far.

Besides, according to the World Bank (2015) 54% of the population in the city still relies on walking as a major means of transportation. Whereas, the overall share of the public transportation system is about 31%, and the remaining 15% goes to autos and taxis. Currently, the public transportation network is composed of formal and publicly owned modes, including city buses (such as Sheger and Anbessa buses), the public service employees transport service enterprise (PSETSE), and the light rail transit service, as well as informal and privately owned modes such as midibuses (with a loading capacity of about 24–30 persons), minibus taxis (with a loading capacity of 11–16 persons), and Bajajs, a kind of three-wheeled motorized rickshaw vehicle with a loading capacity of 3–6 persons.

Particularly, Addis Ababa light rail transit (AA-LRT) is another new and electrified public transport mode operating on a total of 34 km of two lines using 39 stations, of which five are common stations between these two lines. Currently, this LRT network is providing passenger services for over 125,000 people per day on the two corridors, namely the north-south and east-west corridors.

According to the Addis Ababa City Transport Authority (AATA) report of 2020, the composition of the public transportation network covering all the eleven administrative sub-cities of the city and the surrounding suburbs shows that minibus taxis account for the largest share with 86%, followed by Anbesa citybuses with 5%, both minibusses, and PSETSE have 3% each, AALRT has 2%, and the remaining 1% is covered by sheer citybus (Addis Ababa Transport Bureau, 2019).

In the city’s public transportation network indicated in Figure 1, about 8,911 minibus taxis are operating on 1265 routes in all sub-cities, mostly with blue and white vehicles, and the routes of their network are permitted by the Addis Ababa City Transport Authority. About 687 Anbessa city buses are operating on 124 routes in the city. Based on data from 2018–2019, the service carries an average of 309,888 passenger trips per day. Given the extensive route network operated by the Anbesa buses, headways are sometimes as long as 90 minutes.

Moreover, Sheger city buses are also operating, using about 217 vehicles on 48 corridors and serving about 198,000 passengers per day. Even though Sheger buses operate along many of the same corridors served by Anbessa buses, the routes and network are predominantly concentrated in the inner parts of the city.

3.1.2. Hawassa city and its public transport network

Hawassa City, also called Awasa, is located in the Great Rift Valley of Ethiopia, 273 km south of Addis Ababa, 1,125 km north of Nairobi, Kenya, and on the railway connecting Addis Ababa to Kenya. Currently, the city serves as the capital for one of the regional states in Ethiopia, namely Sidama regional state. It has a total area of 157.2 square kilometers and is divided into eight sub-cities. Hawassa City is one of the fastest-growing cities in terms of building infrastructure and population size (Central Statistical Agency of Ethiopia [CSA], 2019).

According to the Ethiopian Central Statistical Agency or CSA, the total population of Hawassa city administration was estimated at 315,267 in 2016, but these days it is growing to about 436,992 with a growth rate of 4.02% in 2020. Out of the total number, 292,525 people live in an urban area with a 4.8% growth rate, while the remaining 144,467 people live in rural and peri-urban areas within the administrative boundary. A large part of this rapid population growth was due to high immigration levels from other areas, especially rural areas and smaller towns. Thus, in Hawassa City, there are massive numbers of people who use public transport to make home-based and non-home-based trips and achieve their daily activities or mobility needs.

Based on the data from UN-Habitat (2020), there are three types of public transportation services: mass transport via the city bus and midibus lines (run by the municipality or city administration), automobile Lada rickshaw taxis and Bajaj as well as minibus taxis (both are privately managed). These public transport services were used in some heavily trafficked corridors of the city. Besides, like other regional towns of Ethiopia, the transportation in Hawassa city was mainly based on walking, horse-driven carts, bicycles, and motorbikes. The recent development of a very flexible system of vehicles has witnessed the shift of the transportation system into Bajaj taxi transport.

Regarding the proportion of modes of transport used in Hawassa, the auto lada rickshaw taxis, Bajaj and minibus taxis account for about 55%, midibuses and city buses (19.5%), private cars (8%), walking (2.5%), motorbikes (10.5%), and others (4.5%). Thus, in most parts of Hawassa city, public transportation accounts for the dominant modal share compared to private modes of transport. However, in the newly established and outskirt areas of the city, such as Tula, walking and motorbike usage are widely seen because of the absence of a public transport system.

Nowadays, even though there are about 2500 Bajaj taxis registered and organized in about eleven associations, each with its own starting point that can overlap with that of another agency, there are also about 500 unregistered vehicles that illegally provide services. There are six Bajaj taxi main lines that are spread throughout the city, sometimes with more than one line covering the same area.

The automobile lada rickshaw taxis and Bajaj and minibus taxi service systems connect the most attended spots by the people of the city, mainly in the central areas, with some overlap, for example, at the Gabriel stop that links three different lines. This is because they can drop off and connect passengers from areas that the official bus lines do not reach, enabling them to provide an efficient mobility system (UN-Habitat, 2020). Hawassa city has a promising road grid with large avenues and 2–3 lines per direction that compose the main public transportation axes.

As indicated in Figure 2, in this city, the public transportation network is composed of six routes composed of arterial, collector, and local streets. These are:

Figure 2. Public transport network of Hawassa city.

Source: Adapted from (UN-Habitat, 2020).

1. From Tikur wuha to Monopol, which is 11.91 km long, the axis uses five buses: one articulated bus, one normal (12 m) bus and three midi-buses.

2. From Piaza to Debub Condominium (5.18 km) is served by four buses: two normal (12 m) and two midi-buses.

3. From Blue Nile to Referral Hospital (3.85 km), the axis uses four buses: two normal and two midi-buses.

4. Membo (TTC) to New Bus Station (3.18 km)

5. From Membo to Nigist Fura to Membo (3.53 km). This axis uses four buses: two normal (12-meter) and two midi buses.

6. From Tula to the New Bus Station (7.93 km), this only uses one articulated bus.

The extension of these routes in the network varies from approximately 3 km to 12 km. The existing bus routes cover mostly the southern area of the city, but the rest of the built-up areas are unserviced. In addition, in each bus route, most paths are not doing a closed journey, are too short, or might make the most efficient route. Though there is no clear database system for public transport, considering its fleet and operation system, the urban bus service organization’s provisions report showed that it has a potential capacity to transport 39,200 people per day. However, due to operational limitations, each city bus is currently making seven trips per day and could make at least ten round trips per day on average. This means the organization is transporting about 20,400 people per day, which is 52% of its capacity. Moreover, there is no clear delimitation of public transport stops, which is negatively impacting the service and the overall use of the streets (UN-Habitat, 2020).

The growth of public transport users is fast, and this was combined with the somewhat limited development of public transport facilities and services within the city. In spite of the lack of an overall policy on transport, Hawassa does have clear regulations for public transport operation bodies. The role of each operator and stakeholder involved in public transport services is clearly defined in the regulation. However, the city’s capacity for public transport provision, operation, monitoring, and control is very limited. Currently, the public transport service within the city is organized as per the directive formulated by the regional road and transport bureau, and the enterprise is working with a service subsidy. The regional capital, Hawassa, one of the most rapidly urbanizing cities in the country, is also experiencing similar problems. It was observed that there was overcrowding and congestion of the public transport demands every year in the city’s urban area (Shiferaw et al., 2015 cited in Gebre, 2021).

According to the World Bank and the transport policy of the country (2015), achieving modal shifts into public transportation, mainly in the major urban areas, is the priority goal to tackle the multi-dimensional transport problems and challenges. Even though these cities have a diverse range of public transportation options and the government’s investment in them is growing, there are still significant limitations in the provision of public transport services.

Insufficient supply of public transport services is a significant challenge. For example, the interview with the head of the Addis Ababa Transport Bureau in February 2023 revealed that, including the minibus taxis, the total number of public transportation modes is about 2000 in Addis Ababa. Of these, the number of city buses is not more than 1250, and the most widely used buses are named Anbesa and sheer city buses, including the articulated ones. This means the amount of demand-supply gaps in terms of the city buses is about 3000–3500 buses, which is a huge gap. The long waits, queues of passengers, and overcrowded vehicles being seen every day, mainly during peak hours in many parts of the city, are manifestations of shortages of public transportation services.

Even the new AA-LRT system is characterized by very long intervals of 10–15 minutes between trains. In addition to the limited network coverage, particularly in outlying areas, the quality and comfort of public transportation vehicles are poor (Planning & Implementing Sustainable Transport in Addis Ababa, n.d; World Bank, 2015).

The COVID-19 pandemic has disrupted economic activities and livelihoods all around the world. Despite increasing levels of vaccination, the coronavirus continues to spread, having an unprecedented impact on people’s lives. Ethiopia is no exception, and the first case of COVID-19 in Ethiopia was reported on 13 March 2020 (FMOH, 2020; Kopsidas et al., 2021). Followed by early preventative measures such as mandatory quarantine for travelers, mask mandates, and communication efforts, the government of Ethiopia declared a national state of emergency on 8 April 2020. In addition, the government decided on restrictions on movement and limited access to transportation, the banning of public meetings, and school closures mainly in urban areas. The Ethiopian Federal Ministry of Health swiftly implemented a series of national COVID-19 response policies, including national guidelines to sustain essential health services (WHO, 2020; Zikargae, 2020).

APTA (2020) showed that at global level there are about four phases of COVID-19 pandemic that basically require unique types and specifically targted measures. According to the COVID-19 Map-Johns Hopkins Coronavirus Resource Center (2021), the situation and corresponding stage of the pandemic in Ethiopia have evolved rapidly in the last about 18 months. While the initial wave in the first half of 2020 progressed more slowly all over Africa relative to other continents, like other African countries, Ethiopia was smashed much harder by the second wave (mainly February – June 2021). At this stage, the highest number of cases and frequency were recorded, and since then there has been no sign of slowing down, especially given the arrival of the Delta variant.

Ethiopia entered the third wave mainly at the time of the survey collection for this study (i.e. October to December of 2021), and the negative impacts of the pandemic were being further aggravated by other crises, the outbreak of desert locusts, and armed conflict in several regions. At the time when the surveys were conducted, about 365,776 confirmed cases and 6,486 deaths from COVID-19 were reported in the country. However, with a total number of about 500,014 cases and 7,572 deaths up to March 2023, Ethiopia currently ranks the fourth highest in Africa. In spite of the slight increase seen recently, a lower fatality rate is recorded in Ethiopia compared to the regional average throughout the pandemic. Next to the second stage, the infection rate reached its highest level during the third stage. This is due to the fact that the actual case and death count is estimated to be much higher than the reported ones (Ethiopia: WHO Coronavirus Disease (COVID-19) Dashboard with Vaccination Data, n.d.).

Despite the gradual re-opening of schools, the return of children to education, and the overall recovery in employment during the time of this survey in 2021, the economic and social costs of the pandemic continue to be substantial. The COVID-19 situation continues to be precarious and has varying impacts to date across different regions of the country, with notable disparities between urban and rural areas and different sectors and socioeconomic groups.

In general, urban areas and public transportation all over the world are both the sources and the primary victims of the multidimensional impacts of the COVID-19 pandemic. That is why the study selected these two rapidly urbanizing cities in Ethiopia to investigate the implications and impacts of the pandemic on the mobility and transport modal choice behaviors of people in urban areas of developing countries.

3.2. Impact evaluation approaches

This study was undertaken based on key elements and assumptions of a protection motivation theory that can be used as evidence and a basis for COVID-19 impact evaluation. Through a mixed research design, evidence was drawn from different methodologies and sources. Much of the evidence came from econometric studies, case studies and particularly pre- and post-evaluation model using multiple impact indicators.

Accordingly, this study used comprehensive impact evaluation models that could bring complete and adequate evidence to public transport choice decisions. This impact approach could also provide new evidence about the effects on the public’s choice of transport mode. It could test the basic assumptions about the effects of the new pandemic-related and conventional choice-determinant variables on the mobility behavior and mode choice decisions of urban people.

Thus, it was used as evidence to explain the impacts of COVID-19 on the travel behavior and mode choice decisions of people, as well as the poor, outer-city residents, and the elderly in particular. These pieces of evidence represent the change or impact across time and targeted passengers based on the defined impact indicators relative to situations in the pre-COVID period. These percentage changes in low-income, outer-city, gender-specific, and elderly passengers were also evaluated and compared to a corresponding percentage change in their counterparts.

3.3. Comparative approaches

In the transport market it is really important to make consistent evaluations about the effects of various factors on travel behavior and mode choice decisions of service users. Impacts of COVID-19 pandemic on mode choice decisions of passengers are best evaluated through comparative approaches and the quantification of outcome variables and impact indicators. Thus, a comparative study design was used to effectively describe and explain the phenomena in the current study. This design could help to establish whether there are statistically significant differences between dependent or independent groups concerning some key outcome variables. The dependent groups are used for pre/post temporal comparison using responses and observations on the determinants of mode choice before and after pandemic scenario from the same subjects. Whereas, independent groups are used for differential impact analysis among income, gender, age social groups as well as for a spatial comparison among transit sites/corridors and inner-city vs outer-city situations.

3.3.1. Pre-post analysis

Evidence from Klatt and Taylor-Powell (2005) and Toepoel and Schonlau (2017) showed that this type of pre/post analysis model specifically a retrospective pre-test evaluation design is better for such situations that include measuring change over a very short period (i.e. a 4-hour course), capturing factual or routine information, attempting to gauge perceptions of change as a result of project implementation, trying to diminish response-shift bias, or trying to evaluate change without having collected baseline data before the start of project efforts (Howard, 1980; Sudman et al., 1996).

Thus, a pre-post analysis approach was used based on before-after comparisons that assume that all impacts and changes in mode choice over time are due to the COVID-19 pandemic and not due to other trends or factors. This temporal comparison and impact evaluation methodology were predominantly based on before and after-COVID analyses. The impacts of COVID-19 and responses were compared with a scenario that would have existed had the corvid-19 and response measures not been undertaken. To this end, the temporal comparison technique was applied using pre and post-COVIDoutbreak scenarios (i.e. before and after March 2020).

As the report of the first confirmed case of COVID-19 and response measures started in March 2020 in the country, this date was used as a reference point for the pre/post-COVID scenario of impact evaluation and temporal comparison. The situation of 12 months after March 2020 was compared with the equivalent 12 months before March 2020.

This temporal comparison study design was used on the same intra-city transit users by making two observations or measurements for the pre and post-COVID outbreak scenarios on the defined outcome variables. Since there are no similar ‘before’ studies and the length of the reference period is shorter to easily memorize the past situations, the historical review for data about the pre-COVID period is based on the memories of respondents. This means that assumptions about how the public transport mode choice decision of transit users in the pre-COVID period are often not directly tested. Heath et al. (2020) showed that pre/post-program survey participants report more easily and significantly when the reference period is shorter because it requires shorter time memories.

3.4. Methods to collect data and materials

Given the pace of events surrounding the COVID-19 pandemic and the urgency of the situation, compounded by potential long-term negative impacts, public transportation assessment methods must adapt and overcome barriers to maintain a steady and reliable flow of information. So, the authors used possible research methods considering the gaps in the literature and information as well as the influences of prevailing circumstances on data collection and research work.

This study employed multiple approaches for various types of data from various sources through both quantitative and qualitative methods. Accordingly, close-ended questionnaire, key informants interview (KII), field observations, document and literature reviews, and informal expert consultations were employed to collect evidence materials and data.

The passengers or users of public transportation are the study’s target population in both Addis Ababa and Hawasa, based on the study’s purpose. Unlike household surveys, the transport service users included in most transport surveys are normally characterized as having no fixed address and no well-organized document about their exact numbers.

Accordingly, to estimate the sample sizes, the formula of Cochran’s (1977) for the infinite or unknown study population was separately applied for each city using a 95% confidence interval. The computed sample size was 384 from each city, and this number was adjusted to 763 considering the proportion of passengers and the sizes of both cities. Since the sample size is determined by a scientific formula and based on the desired accuracy of a 95% width of the confidence interval, its sufficiency for such a quantitative survey is justified. The collection of qualitative data from interviewees was also carried out through the attainment of saturation.

Although probability sampling is more rigorous, sometimes it is only theoretically sensible, especially in such transport service surveys where it is not feasible and practical because of the portable nature of transport service users and the lack of a convenient sampling frame. After the proportional assignment of a quota of passengers was made for the purposefully chosen study sites (terminals, stations, corridors, transit modes, and residential areas in both cities), sample respondents were thus selected using segmentation analysis and accidental or convenience sampling techniques.

Even though the passengers who were simply available on board and at stations during the time of the surveys were chosen as samples, data collectors and survey moderators made much effort to consider the composition and proportion of samples from various demographic backgrounds such as age, gender, seat, and residential locations. This could help to reduce the selection bias and errors mostly associated with non-probability sampling techniques such as accidental or convenience sampling techniques.

The actual practice of the onboard survey in transport research was difficult and time-consuming, mainly considering the challenges of data collection from this large sample size, mainly during the COVID-19 pandemic, and resource limitations. Consequently, some household surveys were also conducted on those households and individuals believed to be typical users of public transportation. The administration of these household socio-economic and travel surveys was carried out along with and to supplement the on-board survey and achieve the representativeness of samples.

Overall, the surveys were also conducted from October 1 to 20 December 2021, during peak and off-peak hours, weekdays and weekends, to increase the representativeness and validity of the data. The surveys were also administered by the trained data enumerators and with the collaboration of public transport operators and staff, mainly at stations, terminals, and onboard.

The questionnaire tool was composed of three sections including socio-economic and demographic backgrounds, travel behaviors, and pandemic-related and conventional factors affecting PT mode choice before and during the COVID-19 pandemic using 5-point Likert scale items. Therefore, the conventional variables and new pandemic-related variables were identified and included in the questionnaire for the respondents to place a priority on each factor when choosing a transport mode (indicated in Table 1).

Table 1. Operationalization of variables used to analyze the factors of mode choice.

Variables/Determinants of mode choice	Type
How high you place a priority on “total travel time savings” while choosing a mode?	Conventional
How high you place a priority on “travel cost affordability” while choosing a mode?	Conventional
How high you place a priority on “service availability” while choosing a mode?	Conventional
How high you place a priority on “accessibility of service” while choosing a mode?	Conventional
How high you place a priority on “comfort” while choosing a mode?	Conventional
How high you place a priority on “predictability of arrivals” while choosing a mode?	Conventional
How high you place a priority on “environment concern” while choosing a mode?	Conventional
How high you place a priority on “cleanliness” while choosing a mode?	Conventional
How high you place a priority on “COVID infection risk” while choosing a mode?	Pandemic-related
How high you place priority on “social distancing” while choosing a mode?	Pandemic-related
How high you place a priority on “Good air flow while traveling” when choosing?	Pandemic-related

Source: Adapted from Norman et al. (2005), Rogers (1975) and Rosenstock (1990).

Following the questionnaire survey, a total of 15 semi-structured interviews were conducted based on the data saturation principle with purposively chosen key interview informants (KII) from passengers, experts, and authorities in both cities to gather meanings and opinions for qualitative data. Considering the volume of transport services and passengers, ten of the KII were made in Addis Ababa and the remaining five in Hawassa.

The nine key interview informants, particularly passengers, were purposefully chosen considering their lived experiences of public transportation before and after the pandemic period and the more relevant information and knowledge they possess on the topic. Even though the use of transport services is a common characteristic, they were made to be composed of various backgrounds using certain selection criteria, including sex, age, city, location, the typical mode they use, and the length of service usage, to increase the representativeness and validity of the data. Additional six key informants from the other group of experts and authorities were particularly chosen based on certain selection criteria, including their qualifications and responsibilities relevant to the study, such as transport, urban planning, and technical expertise, as well as the political decision-making power they have to manage the sector.

In addition to the in-person interview, the study tool on the phone was fitted with checks to ensure data completeness and accuracy. These data were used to complement and explain the quantitative evidence regarding the impacts of the pandemic on mode choices and mobility behavior.

3.5. Conceptual framework

In this study, the conceptual framework (indicated in Figure 3) for the investigation of casual relationship and impacts of COVID-19 pandemic related attributes on public transportation mode choice is designed based on PMT and relevant literatures. Brewer and Fazekas (2007) and Li et al. (2021) revealed that people’s perception of risk is based on their cognitive assessment of its seriousness, whereas severity of indicators of risk reflect their awareness of serious circumstances.

Figure 3. Conceptual Framework: Causal Links between COVID-19 pandemic, PT Mode Choices and Travel Risks as Mediating Variables.

Source: Adapted from a protection motivation theory (Rogers, 1983; Norman et al., 2005).

Studies such as Cahyanto et al. (2016) confirm that perceptions of travel risk have a big impact on people not choosing for transport modes and not traveling in risky situations. Travel risk perceptions are the main factor in deterring mode choice and travel decisions (Kozak et al., 2007). The travelers’ and tourists’ perceptions on the travel risks influenced their destination selections, and their opinions of the location’s image influenced their travel choices (Wang et al., 2021).

Recent studies on the COVID-19 pandemic acknowledged that it had a detrimental impact on people’s travel intentions as well as their travel choices (Liu et al., 2020; Teeroovengadum et al., 2020). Studies have also shown that a person’s fear and perception of risk affects how they act to protect themselves (Dryhurst et al., 2020).

The role of PMT is to serve as a basis for this framework of transport-related study to indicate how travelers perceive COVID-19 pandemic and the severity of the contagious disease, and how they perceive travel risk during the pandemic. It could also help to understand the response measures and triggers that alert individuals make to potential threats. These triggers include fear messages that encourage individuals to take protective measures or refrain from activities that might harm themselves or others. Consequently, how the fear, perception changes and protective measures correlated with and impacted public transportation choice changes.

Based on the PMT and preceding literature, the conceptual framework assume that passengers have assessed the pandemic infection risks including travel risk (TRP) of public transport modes as well as the health and safety risks (HSP) that are associated with mode vulnerability, vehicles cleanliness, implementation level of intervention policies. Consequently, they have taken protective measures to avoid the severe effects on their health and lives.

The role of this PMT-based framework was generally to guide the pipeline of the study data and methods of impact analysis. Its role in this transport study is that of an analysis tool that provides a systematic framework for representing how travel behavior and public transport choice changed in response to different input assumptions such as COVID-19 infection risk perceptions (severity and vulnerability), travel fear, and protective measures of passengers. It could thus help to develop and test the following hypotheses:

• Impacts of COVID-19-related factors are related with TRP & HSP, and positively correlated with public transportation choice reductions. TRP and HSP mediate between the COVID-19-related factors and reductions in the choices of public transportation mode.

• The larger COVID-19 infection risk perceptions (severity and vulnerability) and travel fear of passengers will produce larger motivation to take protective measures.

• The larger the motivation to take protective measures, the larger the tendency to avoid or reduce travel by modes of transportation.

• The lower the likelihood of more frequent public transportation choice during the pandemic, compared to the pre-COVID period.

3.6. Operationalization of constructs and variables

In the theoretical and conceptual framework, it is discussed that the role of the modified PMT in the present research was to quantify and examine the perceived impacts and types of correlations that exist between the pandemic-related attributes and the PT mode choice change intentions and behavior. Appropriate constructs for the variables of interest in the PMT such as perception of pandemic severity and fear, vulnerability of travel risks, and protective measures were identified, revised, and quantified by items based on PMT and prior studies such as Cox et al. (2004) and Rogers (1983) on protection behavior to ensure their validity.

Accordingly, three models were used to test the hypotheses and estimate the COVID-19-related factors and their impacts on changes in public transport mode choice during the pandemic. To this end, survey participants were first asked to rate their level of concern about the number of new and total COVID-19 positive cases, infection rates, and death rates every day in the area as indicators of their ‘perception of pandemic infection risk and fear’. These indicators were measured and rated by participants using a 5-point Likert scale (i.e. 5 = very high and 1 = very low) response format. Secondly, in this model ‘vulnerability of travel risks’ is defined in terms of vulnerability level of PT service (measured by considering their cleanliness) as well as the implementation level of safety intervention measures and health strategies (measured by considering Social physical distancing, Good airflow and information) in the transport system to control the pandemic. These were rated by participants using similar question and response format.

Finally, the construct ‘protective measures’ was also quantified and measured using the frequency of scores given by participants’ responses to the selected indicators. These observed indicators are ‘the likelihood to continue the choice for PT modes’, the likelihood to reduce PT choice and shift to others, and ‘avoid travel or choice to any mode’ during the pandemic, compared to the pre-COVID scenario. Out of these three protective measures, the most frequently selected one is considered to be the impact of pandemic-related aspects on PT mode choice changes.

As a result of the first two construct variables, protective measures of travelers were estimated for two travel environments before and after the pandemic period, taking into account their public transportation choice attitudes and decisions. The implications of pandemic-oriented government restrictions and intervention safety measures and their level of implementation are also considered to justify their PT mode choice decisions. These indicators of protective measures refer to how safe passengers felt and how willing they were to choose PT modes in the given environments.

Since the passengers’ cognitive perceptions of risk differ with their economic, social, and demographic characteristics, the current study intends to test the differential impacts and statistical significance of differences between various groups of passengers using multi-group comparative analysis. Consequently, among those socio-economic and demographic factors, gender, monthly income, distance from residential location, and travel cost were found to be significant predictors.

Thus, they were used to categorize the respondents into various groups and analyze the variability of mode choices among the groups or categories. The analysis of socio-economic and demographic factors and travel characteristics of the households and individuals obtained from the sample clearly explains that there are significant variations between employment, income, and education levels, as well as vehicle ownership in the study areas. T-tests and ANOVA can be used to compare the effects of males and females, lower, middle, and upper income levels, employed and unemployed groups, and so on.

A sound understanding of travel-mode choice determinants is needed to design interventions to slow down and prevent the spread of the COVID-19 pandemic. The implication of the pandemic on travel behavior and choice of public transport modes of urban transit users were well contextualized and subdivided into various dimensions and variables. Accordingly, the framework or model for the evaluation of perceived risks of COVID-19 infection in this transport study is performed revolving around the protection motivation theory (PMT). The new variables and indicators that may affect the choice of transport mode are developed from the principles of this theory.

As indicated in Table 1, the recent and pandemic-related variables included as one of the possible determinants during the pandemic are infection concern, social physical distance, disinfection services and cleanliness, and good airflow. They are the perception of risk of COVID-19 infection which affects the decision whether to continue using the public transportation modes. To this end, the COVID-19 pandemic related attributes were used as an independent or predictor variable together with the conventional variables that entail the socio-economic and demographic characteristics and travel behaviors such as travel time savings, travel cost, availability, comfort, and reliability.

On the other hand, the dependent variable was the impacts and changes on the choice of public transport modes during the pandemic. The dependent or outcome variables were represented and measured using more statistically significant variables or impact indicators, for example, the priority placed by passengers on each variable while making public transportation mode choice decisions.

3.7. Data analysis

All tools of the study were pretested in a similar setting and adjusted accordingly before data collection to ensure they capitulate the information required. Meetings were conducted at the end of each day to check for consistency, and completeness, and to ensure proper data collection through questionnaires. Accordingly, the primary dataset of 763 observations were stored and analyzed using IBM SPSS Statistics (2016, Version 24). Presentation of data and results was done using tables, figures, graphs, plates, maps, and textual narratives. The comparative and comprehensive impact evaluation methodologies were supported by relevant quantitative and qualitative data analysis approaches. Data were quantitatively analyzed using inferential statistical models to determine the statistical significance level of differences among various groups in their transport mode choice. Data were also analyzed using descriptive statistics and other qualitative tools.

Concerning the first and second research questions, Wilcoxon signed-rank test was used to analyze and compare median differences between before and after-COVID outbreak observations on the same subjects (i.e. transit users) for the selected outcome variable or mode choice. This test was also run to identify the significant variables to be used as input for predictor variables in the Multinomial logistic regression to model the impact of the pandemic on public transport mode choice.

3.7.1. Model specification

PT mode choice behavior was modeled using Multinomial (MNL) logistic regression models, a discrete choice model. It was used to model the probability of choosing a particular public transport mode for daily trips during the pandemic. This was due to the fact that the outcome variable (i.e. PT mode choice) was nominal and comprised three categories: city buses, mini buses, and Bajajs, with city buses serving as the reference category. According to Chen and Li, (2017) and Train (2003), this model helps choice makers opt only one alternative from a choice set. All possible alternatives are included in the choice set and the number of alternatives is finite. It was also chosen for its robustness in modeling disaggregated data, simplicity in the mathematical framework, and adaptability to any sample size and data type. Unlike the probit model, this model does not require a large sample size. Besides, for the MNL model, the unobserved term of one alternative is assumed to be unrelated to the unobserved term of another alternative. The intercept coefficients are included to capture the average unobserved effects, and they have no interpretable meaning.

This model is more appropriate and robust for this study because it provides the goodness of model fitting values and results in more meaningful interpretations. In addition, the likelihood ratio test and multinomial logit estimate of this model are widely acknowledged for their robustness to indicate the significance of the improvement produced by the developed model on the intercept-only model. They are also capable of observing and modeling how and to what extent the sampled respondents of each predictor variable are likely to choose each outcome variable when compared to their counterparts, given the other independent variables in the model are held constant. This means it could compare the likelihood of choice of males to females, lower-income to middle and upper-income people, etc., as well as the likelihood of choosing each mode, such as city-buses to mini-buses and bajajs.

This model was used to predict the probability of choosing a particular public transportation mode and the probability of an alternative being chosen by a choice maker during the pandemic using the following equation form:

$\mathrm{p}=\frac{\mathrm{e}\mathrm{x}\mathrm{p}\left(\mathrm{a}+\mathrm{b}1\times 1+\mathrm{b}2\times 2+\mathrm{b}3\times 3+\dots \right)}{1+\mathrm{e}\mathrm{x}\mathrm{p}\left(\mathrm{a}+\mathrm{b}1\times 1+\mathrm{b}2\times 2+\mathrm{b}3\times 3+\dots \right)}$

Where:

p = log (odds) = logit (P): the probability that a case is in a particular category. This logistic transformation of the odds (referred to as logit) serves as the dependent variable.

exp = the exponential (approx. 2.72),

X_1, X_2, X_{3 … .}are the independent/predictor variables and

a = the constant of the equation and,

b_1, b_2, b_{3 … .} = the coefficient of the predictor or independent variables.

The predictor variables used as input in the MNL regression were composed of the conventional mode choice determinants, the underlying new pandemic related factors, demographic and socio-economic characteristics of respondents. Out of the total ten variables (i.e. the conventional and new pandemic-related mode choice determinants) indicated in Table 1 those which were found to be statistically significant in the Wilcoxon signed-rank test were primarily chosen as input variable. These are the pandemic infection risk, cleanliness, social physical distancing, good airflow, and travel cost.

Moreover, the other input variables were gender, age, monthly income and employment status and residence distance from central business district (CBD). As indicated in Table 4, the actual data contained in these input variables were a mix of binary, nominal and ordinal data through the use of dummy variables. For example, for gender, it was binary data such as male and female, for pandemic-related factors it was nominal, and for age, monthly income, residence location distance from city-center and travel cost were ordinal data.

Table 2. Perception of pandemic severity, travel risk vulnerability and protective measures.

S/N	Perception of pandemic severity, travel risk vulnerability and protective measures	Response (Frequency)
		Very high	High	Moderate	Low	Very low	Total
1	Size of new & total COVID positive cases every day in the area	292	338	23	11	6	670
2	Number of people recovered from COVID infection every day in the area	31	26	20	274	320	671
3	Pandemic-related death rates and numbers every day in the area	289	332	25	13	8	667
4	Vulnerability of travel risks while using PT, relative to pre-pandemic scenario	294	340	13	15	9	671
5	Implementation levels of intervention measures & safety strategies	37	32	19	268	313	669
6	Likelihood of continuing to choose PT as frequently as in the pre-pandemic period	101	n.a.	n.a.	n.a.	n.a.	669
7	Likelihood of reducing the choice of PT and shift to others, relative to pre-COVID period	499	n.a.	n.a.	n.a.	n.a.
8	Likelihood of avoiding travel or choosing any modes, relative to pre-COVID period	69	n.a.	n.a.	n.a.	n.a.

Very high = 5, High = 4, Moderate = 3, Low = 2, Very low = 1 and n.a.= not applicable.

Source: Field survey (2021).

Table 3. Mode choices of Passengers for daily trips before and during the pandemic.

Scenario	City	Modal Choice/share (Frequency)
		City-Bus	Midi-Bus	Mini-Bus	Bajaj	AA-LRT	Other	Total	Percent
Before COVID	Addis Ababa	104	63	81	25	75	n.a.	348	51.9
	Hawassa	58	20	111	134	0	n.a.	323	48.1
	Total	162	83	192	159	75	n.a.	671	100
During COVID	Addis Ababa	84	54	66	24	62	58	348	51.9
	Hawassa	49	16	105	129	0	24	323	48.1
	Total	133	70	171	153	62	82	671	100

Source: Field survey (2021).

Table 4. Factors determining PT mode choice before and during COVID-19 pandemic.

S/N	Wilcoxon Signed-Rank test on each factor/variable	Ranks		Test statistics
		Negative ranks	Positive ranks	Z	Sig. (2-tailed)
1	Travel time savings After COVID- Travel time savings before COVID	348	317	−10.452	0.05
2	Travel cost affordability After COVID- Travel cost affordability before COVID	101	562	−14.412	0.001
3	Service availability After COVID- Service availability before COVID	360	303	−0.2101	0.06
4	Accessibility of service After COVID- Accessibility of service before COVID	355	306	−0.0240	0.06
5	Comfort After COVID- Comfort before COVID	351	312	−0.0320	0.07
6	Predictability of arrivals After COVID- Predictability of arrivals before COVID	362	297	−0.1220	0.07
7	Cleanliness After COVID- Cleanliness before COVID	79	582	−12.153	0.001
8	Pandemic infection risk After COVID- Pandemic infection risk before COVID	71	592	−17.108	0.03
9	Social physical distancing After COVID- Social physical distancing before COVID	84	575	−16.108	0.001
10	Good airflow while traveling After COVID- Good airflow while traveling before COVID	87	570	−11.960	0.001

Source: Computed using survey data (2021).

Re-coding these categorical variables into a new set of dummy variables with values of 1 and 0 was done to facilitate greater interpretation of the intercept in the model. For the dummy coded variables, one group was treated as a reference category, which is a baseline against which all other categories are compared.

In this study, the reference categories are: male for gender, good airflow for pandemic-related factors, 48 years and above for age, monthly Income-3 (i.e. over 7800 ETB or 163 USD) for monthly income, location distance-2 (or 10–20 km) for residential distance from city center, and trip cost-3 (i.e.≥30 ETB or 0.6 USD) for average travel cost. All the descriptive and inferential statistics were achieved at a 95% confidence interval. Finally, to assess the changes and effects faced and felt by transport users, qualitative data analysis mainly thematic narrative analysis was utilized. Accordingly, selected key informants mainly passengers and experts could freely express the choice of the public transport modes before, during, and after the COVID-19 outbreak. The qualitative data were transcribed and analyzed based on the relevant themes such as perception of the pandemic, travel risks, mode choice and travel behaviour changes, and why and how public transportation mode choice changed. These qualitative data analysis could explain the findings of questionnaire surveys.

4. Results

Out of the planned 763 survey respondents, valid data were collected from a total of 734 samples using a survey questionnaire with a response rate of 96.1 percent. What this section presented is the analysis of data and findings for each research question.

4.1. The changes in travel behaviors of public transport users during the COVID-19 pandemic, relative to the pre-COVID period

To assess the changes brought by the COVID-19 pandemic on the daily travel behavior of people, data analyses were made on the dominant public transport mode choice as well as the change in the frequency of public transport usage for daily trips in both before and during the pandemic scenarios. Thus, the key question is, what is the actual situation regarding public transit mode choice right now.

4.1.1. The mode choice behavior for daily trips before and during the COVID pandemic

As indicated in Table 2, respondents were asked to report their perceptions and attitudes toward COVID-19 pandemic risk severity and fear based on their information about the daily number of new and total positive cases, people recovered and deaths in the area.

According to the summary of the responses and findings of PMT model presented in Table 2, out of 670 respondents, about 291 (or 44%) and 338 (or 50%) of them perceived that the daily number of new and total positive cases was very high and high, respectively. With regard to the daily pandemic-related deaths, 289 (or 43%) and 339 (or 51%) of the respondents also reported very high and high deaths, respectively. On the other hand, 274 (or 41%) and 320 (or 48%) of the respondents indicated that the daily new and total number of people recovered from COVID-19 infection are low and very low, respectively, as opposed to 8% who reported high and very high levels to this question. These findings showed that most respondents were highly concerned and feared about the severity of the pandemic infection risk.

Concerning the second construct, attitudes toward vulnerability of travel risks while using public transportation considering their cleanliness, 294 (or 44%) and 340 (or 51%) of the total 671 respondents indicated very high and high vulnerability levels, respectively, during the pandemic than during the pre-COVID period. On the contrary, some of the respondents (about 4%) described low and very low levels of vulnerability to travel risk.

In addition, for the level of implementation of safety intervention measures and strategies in the transportation system, 268 (or 40%) and 313 (or 47%) of the respondents further indicated low and very low levels, respectively, as opposed to 10% who reported high and very high levels to this question. These evidences revealed that most respondents perceived themselves to be more vulnerable to risks and were uninterested in the cleanliness of and traveling by PT modes. The implementation of COVID-19 response policies and practices (such as social physical distancing, good airflow, and information) in place were unsatisfactory.

Finally, regarding the passengers’ protective measures, out of the 669 respondents, 499 (or 75%) of them reported they were more likely to reduce their choice of PT and shift to other transport options during the pandemic than during the pre-COVID period. In contrast, 101 (or 15%) of respondents indicated that they would continue to use PT as frequently as they did before the pandemic, as opposed to 69 (or 10%) who indicated that they would avoid travel or choose any mode of transportation in comparison to before the pandemic. The findings confirmed that most of the respondents developed an intention and behavior to significantly reduce their choice of public transportation and shift to other options during the pandemic as a response to pandemic risks and fear.

Based on these findings of the PMT model, this study identified the most significant variables including pandemic infection risk, cleanliness, social physical distancing, and good airflow to be named ‘pandemic-related factors’ and used as input for further analysis in the Wilcoxon Rank tests and MLN regression model.

As indicated by Table 3 the usual and existing public transport modal choice or use in Addis Ababa city were City-buses, Mini-buses, AA-LRT (Addis Ababa light rail transit), Midi-Buses, and Bajajs respectively for both before and during COVID-19 period. Findings also showed that the choice of Bajaj in the city is lower in both periods, compared to Hawassa city. The amount of modal choice was reduced for each mode during the COVID-19 period.

Regarding Hawassa city findings showed the usual and existing public transport mode choice in both periods were public transit including Bajaj, Mini-buses, and city-bus, respectively, and other options of public transit. But, in this city, the amount of modal choice was reduced during the COVID-19 period for each mode. There is also a low mode choice for Midi-buses and no provision of light rail service.

For Addis Ababa, the percentage share of overall PT modes is found to be 51.9 percent similarly in both before and during COVID-19 period, whereas for Hawassa it is 48.1 percent for both periods. However, the percentage share of choice for each mode showed a reduction during the pandemic in each city due to the new choice for alternative transport options that were absent before the pandemic.

Figure 4 shows the total distribution of responses for the dominant mode choice and other options of public transit for daily trips before and during the COVID-19 pandemic period in both cities. Before the COVID-19 period, the dominant mode of public transportation for the majority (or 29%) of passengers was the Mini-bus in both cities followed by city-buses and Bajaj, Midi-buses, and AA-LRT with 24%, 12%, and 11%, respectively. Whereas, during the COVID-19 period the dominant modal choice or modal share for the majority of passengers were Mini-buses, Bajajs, and City-buses with 25.5%, 23%, and 20%, respectively.

Figure 4. Overall percentage share of public transport mode choice in both cities before and during the pandemic.

Source: Field survey (2021).

As expected, the use of public transport modes reduced during the COVID-19 period compared to the situation before the COVID-19 period. Among the public transits in both cities, Mini-bus is found to be the dominant mode choice or share in both periods, but its share reduced by 3.5% during the COVID-19 period. Pieces of evidence also confirm that the modal choice or share of Bajaj, city-buses, Midi-buses, and AA-LRT reduced by 1%, 4%, 1.5%, and 2%.

Findings indicate that for the majority of the respondents in both cities the most dominant mode choices are city-buses, Mini-buses, and Bajaj in both periods. In contrast, for a small percentage of respondents, the public transport mode choices are Midi-buses and AA-LRT in both periods. One reason for this may be the availability of AA-LRT in Addis Ababa city only but not in Hawassa city.

Regarding other transport options beyond PT modes in both periods, findings in Figure 4 and Table 3 show that there was almost no use of private transport and non-motorized transportation before the COVID-19 period. Nonetheless, the use of private transport, non-motorized transportation (such as walking and bicycling) and motorcycles increased (from almost zero to 12%) during the COVID period compared to the pre-COVID era. This means the pandemic has contributed to the rise in the choice of private transport and non-motorized transportation as alternative options in both cities.

Of the total number of passengers who were using public transportation modes before the COVID-19 period, about 12% shifted to other modes of transport such as walking, cars, or others during the COVID-19 period. Qualitative findings revealed that this shift from PT is mainly due to the tremendously lower probability of infection risk from pandemics. Although the proportion of these transportation options increased during the pandemic, it is a small percentage compared to PT options. Thus, the COVID-19 pandemic is responsible for the reduction of the choice of public transit, but for the increased choice of other transport options such as private cars and non-motorized modes.

We approached individuals and households that travel frequently before and during the pandemic using PT modes and other options to make intermodal comparisons and multi-modal analysis on mode choices. However, due to the small proportion of choice and frequency of usage of these private auto and non-motorized modes during the pandemic, they were of course not included in the regression model analysis as an outcome variable. This does not mean the majority of people very likely have very few options other than public transit. The options other than public transit, such as private transport, walking, cycling and motorcycling, are in place, but the actual level of choice is lower.

4.1.2. Frequency of public transport usage before & during the COVID-19 pandemic

Concerning the mode choice decision, respondents were asked how often they used Public transportation (PT) modes per week before and during the COVID-19 pandemic period. Figure 5 shows that the majority of respondents, i.e. 52 and 29 percent replied ‘most of the time’ and ‘all of the time’, respectively, for their weekly use of public transportation modes before the COVID-19 period.

Figure 5. Total public transport choice frequency per week before and after COVID-19 in both cities.

Source: Computed using survey data (2021).

Besides, 12% and 7% of them replied that they used public transportation modes ‘some of the time’ and ‘a little of time’ per week, respectively, before the COVID-19 period. During the pandemic period 35, 26 and 24 percent of respondents used public transportation modes ‘most of the time’, ‘some of the time’, and a ‘little of time’ per week, respectively.

In addition, 10 and 5 percent of them used public transportation modes ‘None of the time’ and ‘All of the time’, respectively, per week during the COVID-19 period. Even though the majority of the respondents used public transportation modes ‘most of the time’ per week in both periods, this weekly usage frequency showed a reduction of about 33 percent (from 52 percent to 35 percent) during the COVID period compared to the period before-COVID pandemic.

4.2. Determinants of public transport mode choice during COVID-19 pandemic

4.2.1. Determinants of public transport mode choice before and during COVID-19

Regarding the actual situation of public transit mode choice right now, the dominant types of factors or determinants of public transport mode choice decisions were analyzed in both cities using the same kinds of determinant factors. The variations in the significance level of the dominant factors were also compared between the two periods based on their median differences using the Wilcoxon Signed-Rank test. A mix of both the conventional factors and the new pandemic-related factors was used to analyze and compare the dominant factors used by the public for mode choice decisions in both periods.

Out of the total ten determinant variables indicated in Table 4, the last four variables are new and pandemic-related determinants of mode choice decisions while the rest six are the conventional ones. Out of the selected total determinants, using this proportion (about 40%) of the pandemic-related determinants is sufficient to measure their correlation and effects on mode choice. Even though the effect of government pandemic control measures are partially considered in these pandemic-related determinants (i.e. with social physical distancing good air flow practices), it is well analyzed through the discussion with key informants.

Concerning Pre and during COVID-19 scenario of factors of mode choice comparison, in Table 4, the Wilcoxon Signed-Rank test indicated that median differences between before and during the COVID-19 period concerning the determinant factors used for mode choice decisions have statistical significance.

More importantly, the determinant factors that show a statistically significant difference are the Pandemic infection risk, Cleanliness, Social physical distancing, Good airflow, and Travel cost affordability. As expected, Wilcoxon Rank tests indicated that most of the respondents placed more priority on pandemic-related factors when making transport mode choice decisions during the pandemic period, compared to what they did in the past or before the COVID-19 period. For example, 592, 582, 575, 570, and 562 respondents put more priority on pandemic-related determinant factors particularly; Pandemic infection risk, Cleanliness, Social physical distancing, Good airflow, and average double trip cost affordability, respectively, during the COVID-19 pandemic period than Pre-COVID period. This means, except for travel cost affordability which is a conventional determinant variable, all of the significant determinants of mode choice decisions are new and pandemic-related variables. Accordingly, these new pandemic-related variables and an average travel cost (as travel behavior) were used as input for the Multinomial logistic regression model together with other independent variables to model mode choice.

However, the median differences between transport mode choices before and after the COVID-19 period were not statistically significant for the remaining factors like Predictability of arrivals, Comfort, Accessibility of service, service availability, and Travel time savings. Most of the respondents do not place more priority on these factors when making mode choice decisions in the Post-COVID pandemic period, compared to the Pre-COVID period. Because of this reason, these variables were not used as input for the Multinomial logistic (MNL) regression model.

Furthermore, results of interview discussions with most key informants (n = 14) also showed that in addition to the perception of pandemic-related risks and factors, the measures taken by government to contain the pandemic could exert pressure and effect on the reduction of public transport mode choice of people. These are restrictions on intra and inter-city travel, availability and carrying capacity of vehicles and increase of travel charge tariff.

4.2.2. Modeling the impact of the COVID-19 pandemic on public mode choice

In Table 5, Multinomial logistic regression model was applied to estimate the regression coefficients and model the relative influence of independent variables on the existing public mode choice during the COVID-19 pandemic period. As indicated by preceding findings in Table 3 and Figure 2 among the five public transport modes in both periods (before and during the COVID-19 pandemic) the most dominant mode choices were City-buses, Mini-buses, and Bajaj, relative to others such as Midi-buses and AA-LRT. Thus, the nominal outcome variable input in this model was mode choice which consisted of three categories namely, City-buses, Mini-buses, and Bajaj.

Table 5. Mode choice modeling during the COVID-19 period.

Independent variables	Parameter estimates
	Bajaj			Mini-bus
	B	Exp (B)	Sig.	B	Exp (B)	Sig.
Intercept	1.31	.657	0.000	−0.01	.365	0.004
Pandemic-related factor: infection risk fear	1.40	1.03	0.000	−0.01	1.46	0.004
Pandemic-related factor: cleanliness	0.90	0.723	0.021	0.12	0.615	0.011
Pandemic-related factor: social distancing	−0.12	0.24	0.072	−0.93	0.87	0.020
Sex-female	−0.70	0.54	0.036	0.42	1.32	0.000
Age 18–32 years	1.13	1.38	0.021	0.08	1.10	0.012
Age 33–47 years	0.40	.091	0.064	0.20	.791	0.106
Monthly Income −1	−0.15	0.43	0.000	−1.04	1.03	0.211
Monthly Income-2	0.24	0.09	0.013	0.50	0.86	0.000
Location distance-1	0.64	1.98	0.010	1.27	2.53	0.021
Location distance-3	0.63	0.98	0.100	−1.31	0.95	0.012
Trip cost-1	n.a.	n.a.	n.a.	−0.45	1.06	0.000
Trip cost-2	n.a.	n.a.	n.a.	−0.57	1.27	0.041

Source: Computed using survey data (2021).

Where,

Monthly Income-1 refers to below 3000 ETB (or 63 USD)

Monthly Income-2 refers to 3001–7800 ETB (or 63-163USD)

Monthly Income-3 refers to Over 7800 ETB (or 163 USD)

Location distance-1 refers to Inner-city (≤10 km)

Location distance-2 refers to Intermediate (10–20 km)

Location distance-3 refers to Outer-city (≥20 km)

Trip cost-1 refers to the Average double trip cost of ≤15ETB (or 0.3 USD)

Trip cost-2 refers to the Average double trip cost of 15-30ETB (or 0.3 USD-0.6 USD)

Trip cost-3 refers to the Average double trip cost of ≥30ETB (or 0.6 USD)

In addition to the demographic and socio-economic characteristics, pandemic-related factor and travel cost were analyzed as predictors because both variables were found to be significant in determining transport mode choice before and during the pandemic (Table 4 the Wilcoxon Signed-Rank test). A total of 562 questionnaire responses given about the choice for the three dominant modes were considered in this model analysis excluding the responses given for the others, i.e. Midi-buses and AA-LRT mode choices.

The statistically significant level (p-value) and regression estimation coefficients of the variables obtained from the analysis are also revealed in Table 5. In the analysis, the fitness of the selected model was checked to the set of data using likelihood ratio chi-square statistics. The result of the model fitting information proves that there is a goodness-of-fit of the model and significant relationship between the outcome variable and the predictor variables in the final model at [LR x² (18) = 70.34, p < .001].

According to the Pseudo R-Square statistic, the model explained about 41% (Nagelkerke R Square) of the variance in mode choice is associated with the predictor variables. The likelihood ratio test was found to be significant indicating that the developed model is a significant improvement over the intercept-only and final model.

The regression coefficient of each statistically significant independent variable is estimated by measuring the change in the dependent variable or logit for a one-unit change in each independent variable, ceteris paribus (i.e. other independent variables remaining constant in the model). Here, the positive signs of regression coefficients indicate the positive correlation and effect of predictor variable on the corresponding outcome variable. This means a one unit increase in the predictor variable is associated with a certain unit increase in the outcome variable, but a vice versa for the negative sign.

Those independent variables which are correlated to the dependent variables and have a statistically significant p-value less than 0.05 alpha value are presented below based on the model output interpretation.

• The findings of the model reveal that the pandemic-related underlying factor (such as infection risk fear and cleanliness), gender, monthly income, distance of residential location, and travel cost were significant predictors of mode choice for work-related daily trips during the pandemic.

• The first pandemic-related factor, ‘infection risk fear’, is significant and positive for Bajaj (b = 1.40, p < 0.001) but significantly negative for Mini-bus (b = −0.01, p = 0.004). The log-odds of choosing the ‘Bajaj’ category (relative to the ‘city bus’ category, coded 0) are predicted to be 1.40 points higher than for respondents who prioritized ‘good airflow’ (coded 0 and reference group). The odds ratio indicates that respondents are more inclined to choose Bajajs and less inclined to choose city buses. Regarding the choice of minibuses, respondents who placed a priority on infection risk fear are at a lower inclination than those who placed a higher priority on good airflow.

• The ‘cleanliness’ predictor is positive and significant for Bajaj (b = 0.90, p = 0.021) and Mini-bus (b = 0.12, p = 0.011). Respondents who placed more priority on ‘cleanliness’ had a higher inclination to choose Bajaj and minibuses than city buses.

• The ‘social distancing’ predictor is negative and significant for Bajaj (b = −0.12, p = 0.072) and Mini-bus (b = −0.93, p = 0.020). Compared to respondents who placed a priority on ‘good airflow’, respondents giving a priority on ‘social distancing’ are at a lower inclination to choose Bajaj and minibusses, respectively, than city buses.

• The ‘Sex-Female’ is significant and negative for Bajaj (b = −0.70, p = 0.036), but positive and significant for Mini-bus (b = 0.42, p = .001). Females (coded 1) are 0.54 points less likely than males (coded 0) to prefer Bajaj relative to City-bus during the pandemic period. Whereas, for the choice of the minibus, their odds ratio (1.32) is found to be greater than 1. This means females have a greater tendency to prefer a ‘minibus’ (relative to the ‘city bus’), compared to males.

• The age variable was treated as categorical and each category represented a dummy variable comprising a younger (18–29), middle-aged (30–47), and an older group (above 47). Only the first dummy variable was significant and positive for Bajaj and minibus. Younger respondents had a greater tendency to choose Bajaj as well as minibuses over citybuses.

• Monthly Income-1 (below 3000 ETB or 63 USD), Monthly Income-2 (3001-7800 ETB/63-163 USD), and Monthly Income-3 (over 7800 ETB or 163 USD) have a significant and negative correlation with Monthly Income-1 predictor (b = −0.15, p.001). This indicates that persons in the lower income group are predicted to have a 0.43-point lower tendency of selecting Bajaj than citybuses, while those with an upper income have higher tendency to choose Bajajs and minibus over a citybuses.

• The location distance-1 predictor (inner-city with ≤10 km) was found to be significant and positive for Bajaj and Minibus, while the location distance-3 predictor was significant and negative for the minibus only. People living in the inner city had a 1.98 and 2.53 point higher inclination to choose Bajaj over citybuses than people living in the intermediate location distance (10–20 km).

• Regarding the average double-trip travel-cost variable, trip cost-3 (i.e.≥30ETB or 0.6 USD), was coded 0 and reference category. Predictors for choosing Minibus and citybus were found to be significant and negative, respectively, with people in the trip-cost 1 having 1.06-point lower inclination to choose a minibus than a citybus.

4.3. The anticipated choice of PT in the future COVID period (in 18–24 months)

In addition to the impacts of the pandemic on public transport mode choice and transport behavior of passengers now during the pandemic period, its impacts on the future mode choice and modal shift are also analyzed as follows:

4.3.1. The likelihood of respondents to opt PT modes in the future COVID period

As indicated in Table 6, out of the total 671 respondents about 222 and 192 of them confirm that they are not likely to choose and use public transportation services in the near future (18–24 months) considering the implementation of COVID-19 prevention measures and perception of travel risks. Whereas, 99 and 75 of respondents are likely to choose public transportation modes such as city-buses, mini-buses, midi-buses, Bajajs, and light rail transit services in the upcoming 18–24 months.

Table 6. Probability of modal shift (from PT) in the future post-COVID (in 18–24 months).

S/N	Agreement/Disagreement level about the modal choice in the future (during & post-COVID period)	Response (Frequency)
		Mostly disagree	Disagree	Moderate	Agree	Mostly agree	Total
1	You are likely to use PT service again and again during the pandemic	192	222	83	99	75	671
2	You are likely to recommend PT service to friends, colleagues…	183	219	78	97	89	666
3	You are likely to shift (from PT) into the use of alternative modes of transport	78	99	85	218	189	669

Source: Computed using survey data (2021).

About 219 and 183 of the respondents reveal that they are not likely to recommend friends, colleagues, families, etc. to choose public transportation services. While 97 and 89 of respondents are likely to recommend public transportation modes to friends, colleagues, families, etc. in the near future.

Besides, 218 and 189 of the respondents disclose that they are likely to shift from the use of public transportation modes to other alternative modes of transport in the near future. However, 99 and 78 of them are not likely to shift into the use of alternative modes of transport.

In addition, out of the 15 key interview informants who possess better experience of public transport modes in both periods, majority (n = 11) are not liable to choose again and again the public transport modes in the 18–24 months of the future COVID period, compared to the likelihood they had in the pre-COVID period. They are not also likely to recommend friends, families, and close relatives to opt the services of public transport modes including city-buses, mini-buses, midi-buses, Bajajs, and light rail, etc.

As a result, these key informants are more likely to make a modal shift into other alternative modes considering the weak measures being taken to prevent and control the pandemic as well as their perception of higher travel-related health and safety risks.

4.3.2. The anticipated public transport mode choice levels compared to alternative modes

As shown in Table 7, respondents were asked to anticipate their future mode choice in the future COVID period (in 18–24 months) considering the implementation of COVID-19 prevention measures and the perception of travel risks. Accordingly, 54%, 44%, 39%, and 41% of respondents anticipated, in general, less choices for public transport modes such as City-bus, Midi-bus, Mini-bus, and light rail transit, respectively, in the city of Addis Ababa and Hawassa, Ethiopia. However, for these four public transport modes 25%, 30%, 25%, and 29% of respondents anticipated in general a more mode choice, respectively. On the other hand, 51%, 56%, and 55% of respondents anticipated a more choice for other transport alternatives namely 2 and 3 wheel vehicles, Walking, and Bicycle, respectively. Nonetheless, for these three alternative transport options a less choice is anticipated by 23%, 32%, and 22% of respondents, respectively. Moreover, more choice is generally expected for the ‘None/wouldn’t make travel’ option by 61% of respondents. For the ‘Auto/car use’ option 51% and 36% of respondents predicted a less and more choice, respectively.

Table 7. The anticipated modal choice in the future or post-COVID (in 18–24 months).

The mode anticipated to be chosen in the future	Measures of the anticipated modal choice (%)					Total
	Much less	Some what less	About the same	Some what more	Much more
City-buses	22	32	21	19	6	201
Midi-buses	29	15	26	18	12	126
Mini-buses	9	30	36	14	11	398
Bajajs	10	13	26	20	31	256
Light rail	19	22	30	15	14	105
Walking	14	18	12	40	16	793
Bicycle	11	11	23	22	33	421
Auto/car use	33	18	13	21	15	41
None/wouldn’t make travel	10	13	16	25	36	539
Others	13	15	33	18	21	54
Total	n.a.	n.a.	n.a.	n.a.	n.a.	2934

The sum is beyond the total sample size because of the multiple responses of each respondent.

Source: Computed using survey data (2021).

According to the text analysis of qualitative data, the anticipated mode choices labeled as ‘Others’ in Table 7 are for example ‘Lada taxi’ use or RIDE-sharing, Motorcycle and free transport services offered by public and private employer organizations to their employees. In terms of the largest number of total responses, the leading anticipated mode choices in the near future are walking, ‘None/wouldn’t make travel’, Bicycle and Motorcycle, respectively. Whereas, those anticipated mode choices with the lowest number of total responses are Auto/car use, ride-sharing, light rail, and Midi-buses, respectively.

Therefore, from the evidence, it is possible to understand that majority of respondents anticipated choosing and using a lesser amount of public transportation modes such as City-bus, Midi-bus, Mini-bus, and light rail transit in the future of COVID period (in 18–24 months) due to poor implementation of COVID-19 prevention measures, greater travel risks and other various reasons. In contrast, the majority of respondents anticipated choosing Bajaj, Walking and Bicycle more. Furthermore, the majority of respondents anticipated choosing none or would not make travel because of various reasons.

Even though almost all of the key interview informants (n = 14) are currently using public transport modes, they are not agreeable to choosing public transport modes more in the near future, compared to the choices they made in the current and pre-COVID period. Instead, they expected to make a modal shift into other alternatives such as walking and no travel at all considering the unsatisfying measures being taken to control the pandemic and their perception of travel risks.

4.3.3. Justifications for the anticipated less choice for public transport modes and shifting to other alternative options

As it is indicated well in the preceding findings, most of the existing public transport modes specifically the City-bus, Midi-bus, Mini-bus, and light rail transit are anticipated to be chosen less in the future COVID period by the respondents of Addis Ababa and Hawassa city. Whereas they anticipated choosing alternative transport options such as walking, no travel at all, bicycle and motorcycle more in the future, compared to their choice levels during and in the pre-COVID pandemic period.

Consequently, as indicated in Table 8 further qualitative and quantitative analyses were made to find out the significant and priority rationale that impacts the choice for the public transport modes and other alternative modes in the near future.

Table 8. Justifications for the anticipated less use of and shift from public transport in the future.

Justification for why the shift from PT usage?	Frequency	Ranking
Greater exposure to and risk of COVID infections	376	1
Shortage of COVID infection prevention services on board	329	2
Unaffordable PT service fee	275	3
Out of necessity- disruption of service provision & availability	252	4
Commuters have been laid off	204	5
PT services become irregular and infrequent	187	6
Commuters have been working from home	146	7
By choice i.e. lower service quality	120	8
Commuters retirement	72	9
Inability to walk to/from stops or get on/off vehicles	69	10
Possibility of location change	63	11
PT services start too late and end too early	61	12
Possibility of using other modes (auto, bicycle, bike…)	47	13
Unwillingness to travel with strangers	45	14
PT services are not environmentally friendly	36	15
Others	26	16
Total	2308	n.a.

The sum is beyond the total sample size because of the multiple responses of each respondent.

Source: Computed using survey data (2021).

For this purpose, respondents were requested to select multiple responses from the lists of variables of interest considerably matching their reasons for the anticipated less and more mode choices. The lists of variables of interest are composed of quantitative and qualitative factors considering the economic, social, and environmental aspects of transport. They are also made to be a mix of the conventional factors used in the normal periods and the new pandemic-related variables that are widely considered in the theoretical framework and conceptual framework of this study. Then the justifications were categorically analyzed and ranked in ascending order based on their total frequency of responses to indicate their relative level of significance.

Therefore, findings show that the first five (first to the fifth rank) most significant reasons or justifications that affect the respondents’ PT mode choices during and in the future COVID period are: Greater exposure to COVID infections and travel risks; Shortage of COVID infection prevention services on board; Unaffordable PT service fee; Out of necessity – disruption of service provision and availability; and Commuters have been laid off from their employment, respectively, in this period of COVID-19 pandemic.

More to the point, the next five (sixth to the tenth rank) more significant reasons are PT services becoming irregular and infrequent; Commuters have been working from home; By choice-lower service quality; Commuters’ retirement; and Physical inability to walk to/from stops or get on/off vehicles, respectively.

In contrast, the last five (eleventh to the fifteenth rank) or least significant reasons are the possibility of residential or work location change; PT services start too late and end too early; Possibility of using other transport options (such as auto, bicycle, RIDE sharing, etc.); Unwillingness to travel with strangers; and transport services are not environmental friendly, respectively.

The findings of all key informant interviews and group discussions (n = 15) also confirm that the most significant reasons are directly and indirectly related to the impacts of COVID-19 pandemic; passengers’ perception of travel risks; and safety and health risks. These justifications are also related to the measures taken by governments, employers, service providers and stakeholders to prevent and contain the pandemic such as restrictions on travel, provision of services, service durations, carrying capacity of vehicles and other rules. In addition, most of the first ten significant factors that reduced the choice for public transportation modes are new and pandemic-related variables except few conventional factors such as lower service quality, commuters’ retirement, and physical inability to walk to/from stops or get on/off vehicles.

The results of qualitative data analysis also confirm the significance of the aforementioned reasons that affect the mode choice decisions during and in the near future of the COVID-19 period, particularly among outer-city residents, women, unemployed and low-income people. It also showed that location change for residential and workplace mainly because of the pandemic-related effects is among the justifications labeled as ‘Others’.

5. Discussions

5.1. The changes in travel choice behaviors of passengers during the pandemic, relative to the pre-COVID period

5.1.1. Perception of pandemic severity, travel risk vulnerability & protective measures

This study adapted the modified PMT model as a theoretical foundation and assessed the effectiveness of the model to measure and assesses passengers’ public transport choice intentions and behaviors with regard to the impacts of COVID-19 pandemic-related factors. The results of this study have demonstrated that the modified PMT is a good model for estimating and predicting passengers’ intention and public transportation choice behavior during a pandemic.

This is because evidence revealed that most passengers were anxious and concerned about the severity of the pandemic infection risks as a result of the higher number of daily positive cases and death rates reported in both cities. They believed that there were more vulnerability risks associated with cleanliness problems and traveling during the pandemic, mainly using public transportation. Additionally, these passengers also displayed lower implementation of social physical distancing, good airflow and other safety intervention strategies.

Consequently, it is possible to clearly understand that passengers could develop higher pandemic fear and tendency to take protective measures. Accordingly, they had a higher inclination to reduce choice of public transit and shift to other transport options. The tendency to significantly reduce public transportation choice intentions and behaviors was viewed as a protective measure against the highest level of travel vulnerability and the severity of pandemic threats. Based on these findings of the PMT model, this study could identify variables including pandemic infection risk, cleanliness, social physical distancing, and good airflow as the major pandemic-related factors influencing public transport choice reductions during the pandemic.

Hence, the model results were consistent with the hypothesized causal relationships and protection motivation theory, stating that higher pandemic fear persuaded the respondents to take more protective measures. This means, the larger the perception of the severity of COVID-19 infection, travel fear, and vulnerability of travel risks, the larger the motivation to take protective measures. The larger the motivation to take protective measures, the larger the tendency to reduce travel by modes of transportation mainly transits. These findings also aligned with studies such as those by Floyd et al. (2000) that confirmed the PMT model provides an understanding of why attitudes and behaviors change when people are confronted with threats.

Findings were also consistent with past studies including Ozbilen et al. (2021) and Zheng et al. (2021) that revealed people’s pandemic travel fear and risk perception of public transit were greater than that of private modes during the pandemic, potentially affecting their transit choice and usage behavior.

5.1.2. Transport mode choice changes during the pandemic

A study by Shah et al. (2021) exposed that the share of public transport mode choice was small during the pandemic mainly for the workplace destinations. Roger (2020) shows that transport usage is and will change after partial and complete lockdown measures.

Studies that specifically studied transit demand, for example, Hotle et al. (2020) and Parady et al. (2020), found that passengers perceiving a higher risk of influenza infection in public transit were more likely to avoid transit trips but shift to other options such as private transport. Similarly, evidences of current study showed that public transport mode usage is the dominant one for both before and during the pandemic period scenarios. However, the level of public transport choice significantly declined during the pandemic period mainly because of the direct and indirect impacts of the COVID-19 pandemic, travel risks and related government prevention measures in both cities.

Findings showed that the usual public transport modal choice in Addis Ababa and Hawassa city were City-buses, Mini-buses, AA-LRT (Addis Ababa light rail transit), Midi-Buses, and Bajajs for both before and during COVID-19 period. To mention some of the variation between the two cities, in the city of Addis Ababa the choice and use of Bajaj are lower in both periods whereas in Hawassa city there is low mode choice for Midi-buses and no provision of light rail service. However, the amount of modal choice is generally reduced for each mode during the COVID period in both cities, compared to the pre-COVOD period. Besides, before the COVID-19 period, the dominant mode of public transportation for the majority of passengers was the Mini-bus in Addis Ababa and Hawassa city followed by city-buses, Bajaj, Midi-buses, and AA-LRT.

Whereas, during the COVID-19 period the dominant modal choice or share was Mini-buses, Bajaj, and City-buses. Regarding the frequency of public transport choice, majority of passengers choose the modes most of the time and all of the time for their weekly trips before the COVID-19 period. On the other hand, during the pandemic period majority of passengers choose public transportation modes most of the time, some of the time, and a little time per week. This means the frequency of public transport mode choice (particularly for ‘most of the time’ parameter) was highly reduced by about 33 percent now during the pandemic period in both cities, compared to the pre-COVID pandemic period. Besides, evidences also revealed that the weekly public transport usage frequency particularly for ‘All of the time’ parameter reduced by six-fold on average now during the pandemic compared to the frequency in the pre-COVID-19 scenario.

Off course data were collected during the pandemic period by carefully targeting the passengers who were the frequent public transport service users in the pre-COVID period. Most of these passengers were fortunately frequent public transport service users during the pandemic as well. This could help the passengers to compare their mode choice behaviours between the two periods. Unlike the mode choice behaviours in pre-COVID period, few travellers (about 10%) could adopt new and non-public transport options for their weekly trips during the pandemic. This means the predominant mode choice of all passengers was public transport modes before the pandemic. Except for 10 percent of the passengers, the dominant choice of most of the passengers during the pandemic is public transport. About the 10 percent of the passengers showed modal shift from public transport into other alternative options.

Even though the choice and frequency of usage of PT modes mainly for work-related trips substantially declined during the pandemic, particularly the essential workers of public sector and private institution still need to make travels. There are also people who still demand transportation services mainly for primary and compelling purposes such as health and shopping reasons. For many people who have no permanent salaries or income sources and depend on temporary business and employment opportunities to sustain lives of families, there are demands for travel and transportation services. That is why PT modes are still the dominantly chosen ones even if it is below the situation in pre-pandemic period.

5.2. Determinants of public transport mode choice decisions during the pandemic

Previous studies confirmed that variables such as sex, car ownership, employment status, and travel cost were the most significant predictors of mode choice during the pandemic (Abdullah et al., 2020, 2021).

Overall, considering the shreds of evidence of the MNL logistic regression model in the current study, the pandemic-related underlying factors, socio-economic and travel characteristics of passengers possess significant relationships with and impacts on their public transport mode choice. However, significant variations are seen among the predictor variables. The most significant predictors of public transport (i.e. Mini-bus, City-bus and Bajaj) choice decisions during the pandemic are the pandemic-related underlying factors, gender, monthly income, distance of residential location, and average travel cost. More specifically, the most significant pandemic-related determinant factors are the perception of pandemic infection risk fear, cleanliness, social physical distancing, and good airflow.

Thus, except for travel cost affordability, which is a conventional determinant variable, all of the significant determinants of mode choice are new and pandemic-related variables. In addition, these findings are consistent and aligned with the preceding results of PMT model. Most of the passengers placed more priority on pandemic-related factors and risks when making mode choice decisions during the pandemic period, compared to what they did in the past or before the pandemic. Thus, these findings of the MNL logistic regression model could justify the fitness and good performance of the PMT as framework of this transport and health-related study. This implies that COVID-19 is significantly affecting the public transport mode choice decisions in major urban areas now during the pandemic.

Studies such as by Nikolaidou et al. (2023) and Tirachini and Cats (2020) showed that reduced public transport use since COVID-19 has been aggravated by the opinion that considers public transportation is riskier than personal transport modes. This was mainly due to the possible closer physical contact with other people while sharing the same vehicles and seats.

According to Dryhurst et al. (2020), Liu et al. (2020) and Teeroovengadum et al. (2020) COVID-19 pandemic and person’s perception of risk had a detrimental impact on people’s travel intentions and travel mode choices.

People’s perception of risk is on the basis of their cognitive evaluation of its seriousness, while severity indicators of risk indicate their awareness of serious circumstances (Brewer & Fazekas, 2007; Li et al., 2021).

Just like the conclusions made by such empirical studies, the impact of the COVID-19 pandemic and related prevention strategies on the public transportation mode choices is most significant in the current study, relative to the conventional determinants widely used in the past or before the COVID-19 period. Accordingly, there was a decrease in the choice of public transport modes including Mini-buses, City-buses, and Bajajs during the pandemic because of the COVID-related new factors and infection risks. Besides, the impact of the COVID-19 pandemic and related prevention strategies become more significant, relative to the conventional determinant factors widely used in the normal periods before the COVID-19 period. Evidences have also shown that a passenger’s perception of travel, health and safety risks affect how they act to protect themselves by changing the public transportation mode choice decisions. In this study, how passengers perceive the pandemic, severity of the disease risks and the existing pandemic preventive measures affects how they perceive travel risk of transport service usage, which is strongly correlated with public transportation choice reductions.

Evidences show that the underlying assumptions of the protective motivation theory are applicable and the hypotheses designed based on the conceptual framework are verified in this study. The impact of COVID-19 outbreak is positively correlated with passengers’ reduction of public transportation choice and associated modal shift to other alternatives now during the pandemic. These public transportation choice reductions are also the effects of travelers’ greater perception of health and safety risks, lower cleanliness, and increased vulnerability of transit services. Passengers’ perceptions of travel risk and safety, based mainly on a judgment that considers public transit modes riskier because of closer physical contact with crowds of people, led to a decrease in public transportation choices in Ethiopia.

The other reason that reduced the choice of public transport is the unsatisfactory government response measures, such as dramatic limitations in service supply, social distancing, disinfection, and good airflow.

Furthermore, the majority of interview informants (n = 13) showed that people are less likely to travel by public transportation due to restrictions on transit service provision, increased tariffs, decreased frequency, and protection of passengers’ safety. Thus, government response measures to the COVID-19 pandemic could have unintended negative impacts on travel behavior and public transportation mode choices. This evidence suggests that revisiting and adjustments to control measures are needed to address these negative effects.

Brianne (2020), Ibold et al. (2020) and TUMI Initiative (2020) confirmed that the negative impacts of the pandemic disproportionately affected the mode choice of people mainly the urban residents, lower-income, women, the elderly, and people with disabilities. Similarly, in the current study, the regression model unveiled that in addition to the pandemic-related underlying factor the demographic backgrounds and economic factors such as gender, monthly income, distance of residential location, and travel cost significantly impacted on and resulted in the variation of the choice for each public transport mode during the pandemic.

With regard to demographic factors, Adeel et al. (2016) and Rith et al. (2019) illustrate that males had a higher baseline preference than females to travel by motorcycle to work and females were more likely than males to stay at home during the COVID-19 outbreak. Conversely, a study by Shah et al. (2021) confirmed that male passengers were most likely to travel by 3 wheelers or Bajajs.

Although gender had a significant relationship with the choice for public transportation modes (particularly City-buses, Mini-buses and Bajajs), a dissimilar finding is reported in the current study. Females tended to prefer mini-buses more than males, while males and lower-age (i.e. 18–33 years old) passengers were most likely to prefer Bajajs to other public transit modes mainly city bus. This is probably due to their positive perception of travel risks on these modes. Just like mini-bus for females, for males and lower-aged passengers Bajajs were considered as less risky even if the service fees were relatively larger than city buses. Evidences also show that even for females, city-buses are the second best choices. Hence, it is vital to bear in mind that even though public transit choices are overall reduced, Mini-buses and Bajajs are the priority choices, considering the impacts of the pandemic-related risks, gender and age.

According to Adeel et al. (2014) and Rith et al. (2019) families with a household monthly income of above 500 USD were more likely to travel by car than others with a lower household income.

Likely the current study revealed a corresponding finding with past studies regarding the relationship between monthly income levels and public transport modal choice. Passengers with lower household monthly income of below 3000 ETB or 63 USD and those residing in outer-city locations were generally inclined to choose city-buses, mainly due to relatively lower service charges during the pandemic. In spite of the more expensive service fees, people with an income between 3001 and 7800 ETB (or 63 and 163 USD), spending below 30ETB or 0.6 USD for a double trip and those living in inner-city areas gave more priority to pandemic impacts and travel risks when deciding to choose mini-buses and Bajaj that were also more available than city-buses.

This implies that for the majority and lower-income and outer-city resident passengers city-buses were the priority choices of all public transport modes during the pandemic, primarily considering their affordable services though there was limited service availability and pandemic control measures. Even though the overall choices for public transportation modes have greatly reduced due to the pandemic-related impacts as pointed out in the protection motivation theory, specialized attention is essential to specific mode types considering the demographic and socio-economic backgrounds such as gender, age, income, residential location distances purchasing capacity, pandemic-related perceptions, health and safety of people in urban areas.

5.3. The anticipated choice of public transport modes in the future COVID period

According to IOM (2020) and Timothy (2020) impacts of the pandemic and the consequences of government measures are felt in the fields of urban transport, the economy, and social behaviour. Basbas et al. (2021), Tsavdari et al. (2022) and Roger (2020) also verified that transport usage is and will change due to the COVID-19 pandemic. Concerning the impacts on the anticipated future transport mode choices and travel-related behaviour, pieces of evidence in this study consistently affirm that majority of respondents (61.69 percent) are unlikely to choose the services of public transportation modes in the next about 18–24 months. They are also not likely even to recommend their friends, colleagues, families, etc. to choose public transportation services in the future.

Roger (2020) shows that the number of people using public transport such as buses and rail use in Britain’s cities could go down by about 40% and 27%, respectively, from pre-lockdown levels. There could also be a boom in walking and cycling in a population that may be more interested in health messages.

In the current study findings are similar and public transport mode choice is expected to reduce and people will make modal choice shifts into other alternative modes during and in the future COVID period, compared to what they did in the past or before COVID-19 period. More specifically City-bus, Midi-bus, Mini-bus, and light rail transit will be chosen less by passengers in the city of Addis Ababa and Hawassa. Whereas the anticipated choice for alternative transports options including as Bajaj, Walking, Bicycle and motorcycle is more in the future COVID period due to passengers’ perception of lower travel risks. Furthermore, considering the existing pandemic prevention and control measures and travel risks there are also many passengers anticipating an avoidance of travel due to various reasons.

5.3.1. Justifications for the anticipated less choice and use of public transport modes and shifting to other alternative modes

Abdullah et al. (2020) and Tirachini and Cats (2020) confirmed that as a result of the possible and inevitable closer bodily contact with a large number of passengers while traveling using public transport modes there is a decline in the choice of these modes. COVID-19 has been aggravated by the opinion that considers public transport is riskier than other option modes.

The passengers’ perceptions of travel risk appraisal and risk coping measures lead to their public transport choice reduction, commute disruptions and avoidance of travel (Lu & Wei, 2019; Norman et al., 2005; Rogers, 1983).

Cahyanto et al. (2016), Kozak et al. (2007) and Wang et al. (2021) confirm that perceptions of travel risk are the main factor in deterring mode choice and have a big impact on people not choosing for transport modes and not traveling. Rosenstock (1990) illustrated the perceived risk is a predictor of intentions and behavioral change, which explains the likelihood of engagement in certain behavior in response to stimuli.

Similar to the findings indicated by the previous studies, the evidence of the current study verified that the implication and impacts of the COVID-19 pandemic and related prevention and control measures significantly affected not just the existing public transport mode choices but also the anticipated choices to public transport modes in the near future in urban areas such as Addis Ababa and Hawassa city of Ethiopia because of several factors.

For most of the passengers the underlying reasons for the reduction of current and anticipated choice for public transport modes are greater exposure to COVID infections and travel risks; shortage of pandemic prevention services on board; unaffordable service fees; the disruption of service provisions and availability; and commuters have been laid off from their employment or made out of work by their employers.

Studies by Ibold et al. (2020) indicated that governments, companies, and communities’ responses to control the pandemic have abruptly impacted people’s way of life and transportation systems. Similarly, following the inception of the COVID-19 pandemic the passengers in Hawassa and Addis Ababa city experience dissatisfaction in their everyday life, access to employment and work productivity that resulted in overall income reductions. There was also a large increase in the number of people working from home, made out of work, retired from work, and using online shopping. There are several passengers expecting to change their current work and residential locations. Passengers also shifted into active and non-motorized transport options such as cycling and walking due to the contribution to social development in terms of mobility and increased physical and out-of-home activities for physical fitness and sports.

In addition to direct demand reductions for particular PT modes, findings also show that few travelers adopt new and other transport practices such as motorcycles and auto during and in post-pandemic periods, which can become permanent, depending on a range of factors such as low-quality transit services, privacy concerns and increase of incomes or living standards. However, expensive ride-sharing and carpooling options were chosen by some passengers for more essential, urgent and safer travels.

These all factors are the significant justifications for the current and future reduction of public transport mode choice. This means that COVID-19 pandemic and perception of travel risks are the major cause and factor both directly and indirectly. Beyond the impact of the conventional factors of the normal COVID-free periods, the recent and pandemic-related impacts are more responsible for the reduction of public transport mode choice now during and in the future pandemic period. As the impacts of COVID-19 strongly related to the reductions in public transportation choice, it is possible to understand that the principles of protection motivation theory could inform and imply the findings of the current study. Accordingly, passengers could have assessed the travel risk as well as the health and safety risks of public transport modes and have taken protective measures such as reducing their choice for these transport modes to avoid the impacts of the pandemic.

The passengers’ perceptions of the travel risks influenced their mode selections, and opinions of the vehicle’s image and safety that result in public transport choice reductions.

Brianne (2020), Ibold et al. (2020) and TUMI Initiative (2020) confirmed that the negative impacts of the pandemic are unevenly distributed and urban lower income, women, the elderly, and people with disabilities may be disproportionately affected. Particularly for outer-city residents, women, and low-income passengers the reductions of public transport mode choices were more significant in the current study. Since its declaration as a world pandemic, COVID-19 has brought significant impacts and changes to the nature of public transport mode choices, travel behavior, transport demand, and supply.

Though outdoor travels and public transport mode choices reduced mainly for work and education purposes during the pandemic, people in urban areas still need to travel for various primary and essential reasons such as health, shopping and other compelling reasons. Based on the implications of the pandemic on the urban public transport and mobility, the following specific suggestions are forwarded to address the declining choices for public transport modes in the cities of Ethiopia including Addis Ababa and Hawassa.

• Since public transport service provisions such as city-buses, Mini-buses, and light rail transit were reduced during the pandemic, regional and city governments should encourage public transport priority strategies, public–private partnership to restore the pre-COVID public transport mode choice levels, and accommodate the growing travel demand mainly in rapidly growing cities.

• Understanding the primary concerns of residents who are typically public transport service dependents during, in the future and post-COVID period is crucial to provide social, economic and transport-related support to address the transport mobility demands and bring back the public transport mode choices.

• The city governments should have provided incentives and subsidies to public transport service providers to expand service provisions and frequencies. Such subsidy policy needs to focus on transport disadvantaged groups such as the low incomes, unemployed, outer-city residents to help them afford their travel costs.

• In addition to public transport, city governments and concerned stakeholders should selectively promote new and modern transport solutions such as ride-sharing and carpooling options for more urgent and safer trips mainly for those who can afford. Active or non-motorized transport modes such as walking and bicycling need to be enhanced in selected routes for their contribution to increased mobility and out-of-home activities such as physical fitness and sports. Multimodal transport need also be considered to accommodate the varied travel mode choices.

• The private and government-owned public transport service providers should have pandemic-sensitive transport strategies, increased the availability, frequency and daily length of services together with pandemic prevention services and quality of services of public transport modes are vital to reduce travel, health and safety risks.

• Having considered the protection motivation theory, transport and urban policymakers need to dwell on the overall negative impacts and challenges of the pandemics and devise innovative and adaptive solutions that can accommodate the transport, health, and socio-economic aspects of urban life during the pandemic. It includes reducing the perceived travel risk of public transportation during and in the future of the pandemic period to enhance mobility.

• Further studies should focus on the driving factors that encourage people to reduce public transport mode choices, to stay at home and their socio-economic & environmental effects during and in the future of pandemic period using advanced transport modeling, and more indicators. Besides, further studies need to done in the country-levels to discover new impacts and innovative interventions.

6. Summary

The aim of the current study was to investigate the impacts of the COVID-19 pandemic and related response measures on the mode choice and mobility of people using an up-to-date model called protection motivation theory. For this purpose, the study focused on the widely used public transportation services in the rapidly growing cities of Addis Ababa and Hawassa, Ethiopia.

However, the main constraints and limitations of the study were the exclusion of impacts from the perspective of public transport companies or operators; only the context of passengers in Addis Ababa and Hawassa city was focused. In the pre/post-impact analysis and comparison, there was a dependency on the memory of respondents about the measurements of situations in the pre-COVID period due to a lack of past study, well-organized data, and observation. Besides, the socio-economic and environmental effects of the pandemic were also not considered in this study.

Findings and shards of evidence show that the choice of public transportation was significantly reduced during the pandemic compared to the pre-COVID outbreak periods in both cities. In line with the assumptions of the protection motivation theory, passengers’ perceptions of the pandemic and travel risks were the significant factors affecting public transportation mode choice during the pandemic. Overall, COVID-19-related factors, including passengers’ perceptions of danger, greater likelihood of infection risks, and inconvenience in public transportation services, are the major causes for the reduction of public transportation choices during and after COVID. The shift in mode choice during and in the post-COVID period could aid in long-term demand planning and bring the choice for public transportation back at least to the pre-COVID levels relatively quickly.

Although the long-term implications and impacts of COVID-19 on the public transportation system and general mobility behavior cannot be fully assessed at the moment, the current study made clear that all possible efforts need to be made to ensure that measures taken by governmental agencies, public transport, and shared mobility companies in order to ensure the safety of passengers as well as a further spread of COVID-19 shall be based on comprehensive impact assessments. Unless other measures are also targeted specifically, the legacy of the COVID-19 pandemic will damage mobility systems.

7. Conclusions

This study investigated the impacts of COVID-19 pandemic and related response measures on the public transportation mode choice and mobility of people using protection motivation theory and MNL regression to model the data sample. Willingness to choose public transportation modes is significantly dependent on infection, safety and health risks of COVID-19 and related preventive measures.

The COVID-19 pandemic and response measures such as travel restrictions played a more significant role in affecting the change of the public transport mode choice of people than the conventional mode choice determinants. Moreover, the perception of travel risks in public transportation services is characterized by the physical, social distancing, free air flow and sanitization measures; significantly determine the willingness of people towards the choice of public transportation systems. Due to the impacts of the pandemic, the choice of public transportation strongly reduced during the pandemic. Evidences have shown that the pandemic-related factors as well as gender, monthly income and travel cost were found to have relationship with the reduction of choice for public transportation particularly City-bus, Midi-bus, Mini-bus, and light rail transit. These crises could even result in long-lasting future reductions in public transportation choice and shifts towards other alternative options especially when crucial transport policies are not considered.

Thus, the revival of choice for PT to pre-COVID level requires innovative actions and adequate considerations. This will depend on a range of factors including COVID-19 related factors (such as passengers’ perception of travel risk, likelihood of infection and effectiveness of preventive measures) and when supported by other factors such as travel cost, convenience, and the availability. Even if travels and public transport mode choices reduced mainly for work purposes during the pandemic, there are essential workers and people who still need to travel for other primary and compelling reasons such as health and shopping. Thus, it is crucial to understand the public transport mode choice and travel behavior during and in the future of the pandemic period.

The study concludes with a focus on the need for policies that could be implemented for safe and sustainable public transportation and mobility during and after the COVID period. These are pandemic-sensitive, public transportation-priority policies that include subsidies, high service levels, a focus on passengers’ safety, and the promotion of innovation, digital solutions, and e-payment. The Avoid-Shift-Improve approach needs to be applied and prioritized differently depending on the severity of the four phases of the pandemic. Policies should also consider not only the dramatic one-sided reduction of transport supply but also the demand for mobility, mainly for essential travel. Further studies should focus on the socio-economic and environmental effects of the reduction in public transport mode choices and consider the impacts on public transport companies or operators with advanced models and a broader scope, mainly at the country level.

Acknowledgments

The authors would like to acknowledge Professor Samson Kassahun regarding the guides about to publish articles and make professional contributions. Besides, the support of Ethiopian Civil Service University through a partial finance as staff development program for data collection is gratefully acknowledged.

Disclosure statement

No potential conflict of interest was reported by the authors.