Impacts of shared e-scooters on urban transportation systems

Universidade do Minho Escola de Engenharia Gabriel José Cabral Dias Impacts of Shared E-scooters on Urban Transportation Systems outubro de 2024 UMinho | 2024 Gabriel José Cabral Dias Impacts of Shared E-scooters on Urban Transportation Systems Gabriel José Cabral Dias Impacts of Shared E-scooters on Urban Transportation Systems Tese de Doutoramento Programa Doutoral em Engenharia Civil Trabalho efetuado sob a orientação do Professor Doutor Paulo Jorge Gomes Ribeiro e da Doutora Engenheira Elisabete Maria Mourinho Arsénio Guterres de Almeida Universidade do Minho Escola de Engenharia outubro de 2024 Impacts of Shared E-scooters on Urban Transportation Systems ii DIREITOS DE AUTOR E CONDIÇÕES DE UTILIZAÇÃO DO TRABALHO POR TERCEIROS Este é um trabalho académico que pode ser utilizado por terceiros desde que respeitadas as regras e boas práticas internacionalmente aceites, no que concerne aos direitos de autor e direitos conexos. Assim, o presente trabalho pode ser utilizado nos termos previstos na licença abaixo indicada. Caso o utilizador necessite de permissão para poder fazer um uso do trabalho em condições não previstas no licenciamento indicado, deverá contactar o autor, através do RepositóriUM da Universidade do Minho. Atribuição-NãoComercial-CompartilhaIgual CC BY-NC-SA https://creativecommons.org/licenses/by-nc-sa/4.0/ Impacts of Shared E-scooters on Urban Transportation Systems iii ACKNOWLEDGMENTS I want to thank my supervisors for all their contributions to this thesis. To Professor Paulo Ribeiro, thank you for all the support, for being available for worthwhile discussions, and for constantly pushing me to broaden my academic and professional boundaries. To Dr. Elisabete Arsenio, thank you for all the support and fruitful discussions and for inciting my interest in Discrete Choice Models. Together, you provided me with the necessary tools to accomplish this work. Thanks also to the host institutions, the Centre for Territory, Environment and Construction (CTAC) at the University of Minho and the National Laboratory for Civil Engineering (LNEC), for providing the necessary means to accomplish the research tasks. I would like to thank the Municipality of Braga for being available and providing data on the usage of shared e-scooters in the city. Muito obrigado aos meus pais, José e Maristela, pelo esforço incansável durante todos estes anos para que eu pudesse chegar até aqui. Este trabalho não seria possível sem vocês. Meu muito obrigado. Agradeço também à minha irmã, Nathália, por sempre me apoiar em todas as decisões, e por ser um exemplo de profissional em que me espelho. Thank you, João, for your support and partnership these past few years. Finally, I would like to thank the Portuguese Foundation for Science and Technology (FCT) for the grant (https://doi.org/10.54499/2020.05041.BD) that made this research possible. Impacts of Shared E-scooters on Urban Transportation Systems iv STATEMENT OF INTEGRITY I hereby declare having conducted this academic work with integrity. I confirm that I have not used plagiarism or any form of undue use of information or falsification of results along the process leading to its elaboration. I further declare that I have fully acknowledged the Code of Ethical Conduct of the University of Minho. Impacts of Shared E-scooters on Urban Transportation Systems v Impactos de Trotinetes Elétricas Partilhadas no Sistema de Transporte Urbano Resumo O recente aumento significativo no uso de trotinetes elétricas partilhadas nas cidades levantou preocupações entre cidadãos, políticos e planeadores urbanos, nomeadamente no que se refere ao impacto na segurança e no funcionamento dos sistemas de transporte e mobilidade urbana, assim como na utilização dos espaços e infraestruturas viárias urbanas. Neste contexto, este estudo tem como principal objetivo compreender e avaliar os impactos no sistema de transportes urbanos, com vista ao desenvolvimento de ferramentas de planeamento e regulação mais eficazes para tornar o uso deste modo de transporte mais seguro e sustentável. Para este efeito, a investigação avalia os principais motivadores, determinantes e barreiras associados ao uso das trotinetes partilhadas. Para atingir esses objetivos, foram realizados dois inquéritos na cidade de Braga, no Norte de Portugal. O primeiro consistiu num inquérito de preferência revelada, destinado a determinar os principais fatores que influenciam o uso de trotinetes partilhadas, traçar o perfil sociodemográfico dos seus utilizadores e identificar os principais atributos que podem incentivar o uso deste modo de transporte. Numa segunda fase, foram conduzidos dois inquéritos de preferência declarada para avaliar o contributo deste modo para a sustentabilidade do sistema de mobilidade, nomeadamente como uma alternativa mais sustentável ao automóvel em viagens pendulares e de lazer. Os resultados indicam que o perfil dos utilizadores é maioritariamente composto por jovens adultos do sexo masculino e com um salário semelhante à média para a cidade. As vias dedicadas, o estacionamento e as zonas exclusivas para circulação de trotinetes surgem como os principais fatores determinantes para a sua utilização, especialmente entre mulheres. No caso das viagens pendulares, as trotinetes partilhadas não conseguem promover uma transferência modal dos automóveis para este modo, sendo o transporte público frequentemente uma opção mais eficiente. Já em relação às viagens de lazer, as trotinetes podem ser vistas como uma alternativa potencial ao carro, abrangendo um perfil mais diversificado de utilizadores, especialmente se houver infraestrutura dedicada e se forem oferecidos incentivos financeiros (como vouchers) aos utilizadores, particularmente os de rendimentos mais baixos. Palavras-chave: impacto das trotinetes partilhadas; micromobilidade partilhada; mobilidade urbana sustentável; planeamento para trotinetes partilhadas; trotinete partilhada. Impacts of Shared E-scooters on Urban Transportation Systems vi Impacts of Shared E-scooters on Urban Transportation Systems Abstract The recent rise in shared e-scooter usage has raised concerns among citizens, policymakers, and city planners regarding its impact on the safety and the synergy of urban transportation systems, as well as the usage of urban space and the available road infrastructure. In this context, this research aims to understand and evaluate the impacts of shared e-scooters on urban transportation systems, with a focus on developing tools, planning, and regulation strategies to harness the potential benefits of shared e-scooter usage. By making it more efficient, safe, and sustainable, shared e-scooters could significantly improve urban transportation. Thus, this research evaluates the main drivers/motivations and barriers associated with the usage of this mode of transportation. To achieve these goals, two survey methods were deployed in Braga, North of Portugal. Firstly, a revealed preference survey is deployed to delineate the primary factors that affect shared e-scooter usage, identify the sociodemographic profile of the users, and identify the primary attributes that can lead to using shared e-scooters. Then, two stated choice experiments are deployed to assess how shared e-scooters can contribute to the sustainability of urban mobility, namely as a more sustainable option than cars for commuting and leisure trips. The results indicate that the shared e-scooter user profile is mainly composed of young male adults with a salary similar to the city average. Dedicated lanes, parking spots, and exclusive areas for e-scooters appear to be the main determining factors for their use, especially among women. In the case of commuting, shared scooters cannot promote a modal shift from cars, with public transportation often being a more efficient option. However, in relation to leisure travel, shared e-scooters can be seen as a potential alternative to cars, reaching a more diverse user profile, especially if there is dedicated infrastructure and if financial incentives, such as vouchers, are offered to users and potential users, particularly those with lower incomes. This suggests that financial incentives could be a key solution to increasing e-scooter usage. Keywords: shared micromobility; shared e-scooters; shared e-scooter impacts; shared e-scooter planning; sustainable urban mobility. Impacts of Shared E-scooters on Urban Transportation Systems vii TABLE OF CONTENT CHAPTER I. INTRODUCTION 1 I.1. Context 1 I.2. Motivation 4 I.3. Objectives 5 I.4. Hypothesis and research questions 6 I.5. Structure of the thesis 9 CHAPTER II. STATE OF THE ART 11 II.1. Background 11 II.1.1. The concept of micromobility, shared micromobility, and active transportation 11 II.1.2. Bike sharing 13 II.1.3. E-scooter sharing 16 II.2. Worldwide use of shared e-scooters 19 II.3. Usage and users’ profile of shared e-scooters 25 II.4. Problems regarding shared e-scooter usage 27 II.5. Shared e-scooter regulation 29 II.6. Impacts of shared e-scooters on sustainability 30 II.7. Impacts of the COVID-19 pandemic on urban mobility and the role of shared e-scooters 35 II.8. Synthesis of main considerations 37 CHAPTER III. RESEARCH METHODOLOGY 39 III.1. Revealed preference (RP) 40 III.2. Stated choice (SC) 42 III.2.1. Preferences and the nature of economic valuation 42 III.2.2. Stated choice application in transportation 44 III.2.3. Processes to prepare a stated choice experiment 45 III.2.4. Building stated choice experiments from revealed preference data 50 III.3. Theoretical framework 51 III.4. Logit model 52 III.4.1. Binary logit model 53 III.4.2. Multinomial logit model (MNL) 54 III.5. Synthesis of main considerations 57 CHAPTER IV. CASE STUDY: CITY OF BRAGA 59 Impacts of Shared E-scooters on Urban Transportation Systems xiv To my nephews Davi José and Arthur Custódio Impacts of Shared E-scooters on Urban Transportation Systems 1 Chapter I. Introduction This chapter provides a general overview of the thesis, and it is divided into five main subsections. Section I.1. presents the theoretical framework for developing the thesis. Section I.2. presents the motivation for selecting the research field of shared micromobility in cities. Section I.3 introduces the objectives of the thesis. Section I.4. presents the hypothesis and research questions to be answered by the thesis research work. Section I.5. introduces the structure of the thesis. I.1. Context Sustainable urban mobility planning is believed to be a critical factor in achieving a cleaner, greener, and more innovative future for all. The prioritization of public transportation, walking, cycling, and zeroemission solutions for urban fleets, as well as new digital solutions and services, will put the transportation sector on track to cutting its emissions by 90% by 2050 (European Commission, 2021a). Thus, to make transportation more sustainable, appropriate conditions need to be created for a higher uptake of greener alternatives that are safe, competitive, and affordable, including the subsequent mobility patterns created in recent years, such as shared solutions facilitated by the creation of the “sharing economy” and digital advancements (European Commission, 2020c). The concept of the “sharing economy,” which enables the sharing of a diverse range of goods and services through the availability of various online information systems (Hamari et al., 2015), came to life to fulfill environmental, economic, and social needs to reduce the ownership of goods and consequently reducing the need for their production and the production of waste (Cohen & Shaheen, 2016a). This specific new Chapter I – Introduction 2 form of sharing is present in many sectors, including lodging (e.g., Airbnb), labor (e.g., TaskRabbit, Handy), equipment (e.g., EquipmentShare), food (e.g., Nommery), and transportation (e.g., Lyft, Uber, Gira) (S. Shaheen et al., 2020). One of the motivations for the sharing economy or collaborative consumption is that it provides economic benefits to users, such as travel cost savings, facilitating access to resources, and free riding (Hamari et al., 2015). On the other hand, participation in the sharing economy is usually expected to be highly ecologically sustainable (Prothero et al., 2011). The sharing economy is suggested to be used to foster a sustainable marketplace (Phipps et al., 2013) that “optimizes the environmental, social, and economic consequences of consumption to meet the needs of both current and future generations” (Luchs et al., 2011, p. 2). According to the United Kingdom Department for Transportation (2019), while public transportation remains a fundamental form of shared mobility, new models based on shared use or vehicle ownership are rising. Digital platforms increasingly enable these models and align with shifts toward a sharing economy in other sectors. Shared mobility includes various service models and modes of transportation, such as car sharing, ridesharing, bike-sharing, scooter sharing, ride-sourcing, and alternative public transportation services that have transformative effects on urban mobility and local planning (Castellanos et al., 2022; Cohen & Shaheen, 2016a). The evolution in technology and the change in consumption patterns allowed the introduction of some new ways of displacement in cities that generated positive outcomes. One example is the bike-sharing programs worldwide, which bring several benefits, such as travel time saving, connections with public transportation systems, improved personal health, better air quality, and a decrease in noise pollution (S. A. Shaheen et al., 2013; Todd et al., 2021). The shared mobility service has a transformative impact on many major cities worldwide by enhancing transportation accessibility while reducing personal vehicle ownership and driving among users (FHWA, 2018). This impact has several independent synergies that affect the work of planners and civic leaders, such as transportation and circulation, zoning, land use, growth management, urban design, housing, economic development and environmental policy, conservation, and climate action (Cohen & Shaheen, 2016a). Impacts of Shared E-scooters on Urban Transportation Systems 3 These impacts in urban areas started with the expansion of bicycle infrastructure that accompanied the first wave of micromobility and led to opportunities for the current flux of dockless bicycles and e-scooters (National League of Cities, 2019). E-scooters themselves are not new, as they were an active part of transportation in the 1910s-40s; what is new is the shared ability to access e-scooters using personal smartphone devices and the evolution of electric batteries that now power them (Six, 2019). In the United States, shared micromobility tools began to arrive in cities in 2010 (NACTO, 2019). In Europe, Paris pioneered one of the first city bike schemes, the “Vélib”, and projected it onto the global stage. The system was successful and inspired similar global schemes, such as Milan in 2008, London in 2010, and even New York City in 2013 (World Economic Forum, 2019a). Bike-sharing services, especially dockless or free-floating ones, where users can park their bicycles wherever the ride ends, have seen an abrupt increase in popularity (Richter, 2018). Data from the World Economic Forum (2019b) shows that in 2015, there were 700,000 public-use bicycles in the world, while in 2016, the number reached 2,294,600. China has the highest number of public-use bicycle programs, followed by France, Italy, the United States, Germany, and Spain. E-scooter sharing was first introduced in the United States. These self-service, free-floating rental services started to proliferate in Europe after the arrival of the first providers in Paris in the summer of 2018 (6tbureau de recherche, 2019). The number of shared e-scooters in Europe doubled between 2021 and 2022. By the end of 2022, there were 700,560 e-scooters in Europe compared to the 285,000 e-scooters available in 2021 (Pinheiro, 2023). However, one major problem in expanding shared e-scooter services is the lack of regulation surrounding this “new” mode of transportation. Recently, cities worldwide have been working on shared e-scooter trials and pilot programs to regulate their usage better. In the United States of America, Chicago performed a shared e-scooter pilot in 2019 to understand how the service could be better used when implemented to boost mobility and reduce inequality among users (City of Chicago, 2020). In the European context, France conducted a nationwide e-scooter survey to understand better how, where, and when people ride shared e-scooters (6t-bureau de recherche, 2019a); however, the service was shut down for popular demand in September of 2023. London established e-scooter trials to support Chapter I – Introduction 4 a green restart of local travel and help mitigate public transportation capacity due to COVID-19 pandemic restrictions (Department for Transportation & Transportation for London, 2020). Even if pilot programs have been developed, more is still needed to know how shared e-scooters could affect the urban transportation system. This approach to acknowledging the effects of shared e-scooters in cities can lead to the adjustment of the service to increase its positive impacts on the urban transportation systems and the sustainability of the urban environment. However, further studies must be conducted on this matter to allow the creation and establishment of a sound and systemic regulation for shared e-scooters, which is still a gap in the deployment and massive usage of this mode of transportation in cities. I.2. Motivation One of the main concerns in urban planning is how to manage the transportation systems to accommodate the best mobility options that meet the population’s needs, decrease traffic congestion and air pollution, and allow low-income communities access to safe, comfortable, and efficient public transportation. With the introduction of new technologies (electric-powered vehicles) and new consumption patterns (sharing economy), cities have been looking forward to integrating new modes of transportation to increase the versatility and efficiency of urban mobility. Considering the relevance of the above, shared e-scooter systems were chosen as the research theme to be studied and understood in urban environments. The goal is to determine how this mode could contribute to a more sustainable mobility option in urban centers. Thus, after a preliminary evaluation of the bibliography available, it was possible to identify the need for further studies to assess the impacts of shared e-scooter systems in cities once the addition of this new mode of transportation has been causing some dilemmas associated with how they should be managed to bring the best solutions to help mobility, the environment in urban centers. Cities all around the world, such as Chicago, San Francisco, São Paulo, Lisbon, and Braga, have seen the quick rise of shared e-scooter use in their streets, as it comes with several problems that can be overcome by proper planning and policymaking (Petzhold & Pasqual, 2019; San Francisco Municipal Transportation Agency - SFMTA, 2020). The issue is that there is no consensus on managing e-scootersharing systems in cities. Often, cities allow the implementation of these services without fully Impacts of Shared E-scooters on Urban Transportation Systems 5 understanding the impacts they will cause on the transportation system and public space, which leads to the banning of the service right after they are established. Thus, the primary motivation of the thesis is to complement and further study the many current possibilities available to establish shared e-scooter services in cities. The focus will be on understanding users' and potential users' perceptions and how the sociodemographic characteristics of the population can affect the choice to use shared e-scooters. I.3. Objectives This thesis aims to identify the impacts of shared e-scooters in urban transportation systems, considering the determinants of their usage and how different attributes influence their choice of daily rides. It also compares and contrasts the use of shared bicycles and shared e-scooters in cities to understand what features unite and separate them as modes of transportation in cities. A case study of the town of Braga was used for this. The primary challenge is to develop a methodology to treat the available information to determine the characteristics of shared e-scooter use and how the implementation of this service is managed in urban planning. The second challenge is finding the optimal interface to analyze, compare, and contrast the shared e-scooters and shared bicycles to determine whether one is better for public use and a sustainable city strategy. The specific objectives of this research work are thus listed as follows: ● This research examines how implementing shared e-scooter services in cities can contribute to diversifying the mobility options available to the population. Using available data, it will be possible to determine people's main points of interest in using shared e-scooters, where they feel more vulnerable, where they usually ride their e-scooters, how users evaluate the service, and identify its best and worst features; ● This research investigates the immediate problems caused by shared e-scooters in urban mobility and public space; ● This research investigates the potential to change the modal share in cities, i.e., the probability of shifting trips from other modes, such as individual vehicles, to the usage of shared e-scooters Chapter I – Introduction 6 and other sustainable modes of transportation. It also evaluates its contribution to reducing traffic congestion and decarbonizing the cities; ● This research evaluates the environmental and social benefits of shifting modes of transportation to shared e-scooters in cities; ● This research assesses whether shared bicycles and shared e-scooters can complement each other in cities or if the availability of both services can cause competition. I.4. Hypothesis and research questions Since there is still yet to be a consensus on the key factors influencing shared e-scooter planning and implementation in cities, more studies are needed to prove that this can be a safe and sustainable transportation option. Therefore, this research author has been conducting research on this matter since 2020, culminating in publishing six scientific papers in international journals as the main author, with the contribution and oversight of the supervisors of this thesis. Thus, this thesis comprises a miscellaneous work already published and further studies submitted for peer review. In addition, due to the period when the research was conducted (between 2020 and 2024), amidst the mobility restrictions of the COVID-19 pandemic, this subject was also identified as a possible influence on the usage of shared e-scooters. Thus, it was included in the research questions presented below. Considering the context of the shared e-scooter implementation in cities, as well as the lack of consensus on its impacts in cities, the following hypothesis and research questions were formulated: The research hypothesis proposed for the thesis is: “A shared e-scooter system can be a main driver of sustainable mobility for short city trips if considering the proper planning and infrastructure needed to promote a safe and comfortable ride, which can mitigate its negative impacts on urban transportation systems.” The main research questions of this research work are:  RQ1 – How did the COVID-19 pandemic influence the usage of shared e-scooters? Impacts of Shared E-scooters on Urban Transportation Systems 7 The scope of this research question is to understand the differences in shared e-scooter usage in the preand post-pandemic context, the usage of shared e-scooters during the pandemic in some specific contexts, and what happened in Braga during this period. This research question is further investigated in Chapter II of this research. The following published paper helps to answer RQ1: Dias, G., Arsenio, E., & Ribeiro, P. (2021). The Role of Shared E-scooter Systems in Urban Sustainability and Resilience during the COVID-19 Mobility Restrictions. Sustainability (Switzerland), 13(13). https://doi.org/10.3390/su13137084  RQ2 – What are the main drivers/determinants of shared e-scooter usage for commute and leisure trips? This research question aims to understand the main motivations and drawbacks of shared e-scooter usage. In addition, it targets the assessment of how different attributes affect the choice of shared escooters for daily trips. Research question two is further investigated in Chapter IV of this research. The following published papers help to answer RQ2: Dias, G., Ribeiro, P., & Arsenio, E. (2023). Determinants of shared e-scooters usage in the city of Braga: Results from a mobility survey and trip data analysis. Transportation Research Procedia, 72, 4002–4009. https://doi.org/10.1016/j.trpro.2023.11.379 Dias, G., Ribeiro, P., & Arsenio, E. (2024). Perceptions of shared e-scooters service among university students in Braga, Portugal. Transportation Engineering, 16(April 2023), 100231. https://doi.org/10.1016/j.treng.2024.100231  RQ3 – How do transportation planning and traffic management influence the safe usage of shared e-scooters? This research question intends to understand that shared e-scooters need specific planning measures to be widely used so the population feels safe and comfortable using them. The results presented in Chapter IV of this research are used to answer this research question. Chapter I – Introduction 8 The following published paper helps to answer RQ3: Dias, G., Ribeiro, P., & Arsenio, E. (2024). Determinants of shared e-scooter usage and their policy implications. Findings from a survey in Braga, Portugal. European Transportation Research Review, 16(1). https://doi.org/10.1186/s12544-024-00642-4  RQ4 – What are the impacts of shared e-scooter usage on the sustainability of the transportation system? The target of RQ4 is to understand the modal shift shared e-scooters can promote and the possible environmental benefits of replacing vehicle trips with shared e-scooters. Chapter IV of this research investigates the answer to this research question. The following published papers help to answer RQ4: Dias, G., Arsenio, E., & Ribeiro, P. (2021). The Role of Shared E-scooter Systems in Urban Sustainability and Resilience during the COVID-19 Mobility Restrictions. Sustainability (Switzerland), 13(13). https://doi.org/10.3390/su13137084 Dias, G., Ribeiro, P., & Arsenio, E. (2023). Shared E-Scooters and the Promotion of Equity across Urban Public Spaces – A Case Study in Braga, Portugal. Applied Sciences, 13(3653). https://doi.org/10.3390/app13063653  RQ5 – Are shared e-scooters more competitive than shared e-bikes for short and mediumdistance trips? RQ5 intends to understand the choice of shared e-scooters and e-bikes in Braga and compare how each mode of transportation is chosen. The interaction of these two modes is presented in Chapter IV of this research. The following paper helps to answer RQ5: Impacts of Shared E-scooters on Urban Transportation Systems 9 Dias, G., Ribeiro, P., & Arsenio, E. (2024). Are shared e-bikes disruptive of established shared escooter services? A case study of Braga, Portugal. Lecture Notes in Mobility (Proceedings of TRA2024), Springer: in process of publication. I.5. Structure of the thesis The thesis is composed of five main chapters and is organized as follows: Chapter I is the introduction of the thesis, offering a comprehensive research outline. It presents the theoretical framework and main motivations for selecting shared micromobility as the theme for this research. Then, the objectives of the thesis are outlined, as well as the hypothesis and the research questions. Additionally, it presents the structure of the thesis document. Chapter II focuses on the literature review and the theoretical support for the theme explored. This chapter presents the concepts of micromobility and shared micromobility and the cities’ take-up on shared escooter services worldwide. Then, the profile of shared e-scooter riders and the main problems caused in the towns by exploiting this mode of transportation without any prior planning will be presented. The chapter concludes by extending the review to the usage of shared e-scooters for improving sustainability in cities, the impacts of the COVID-19 pandemic on urban mobility, and the role of shared e-scooters in the towns during this period. The information presented in this chapter helps to answer RQ1 regarding the influence of the pandemic on shared e-scooter usage. Chapter III introduces the methodology used for the thesis. It provides an overview of the two survey models used to collect data on the determinants of shared e-scooter usage in Braga, as well as how the choice of modes of transportation is affected by different attributes. Therefore, a review of Revealed Preference (RP) surveys and the tools used to process the data collected in this part of the research are presented. In addition, the stated choice (SC) method and the steps to formulate a stated choice experiment are presented. Then, Multinomial Logit (MNL) models will be introduced and used to model the data retrieved from the SC experiment. Chapter IV presents the results of the RP and SC surveys collected in Braga's case study. It starts characterizing the case study site and explaining the outline of the surveys deployed in the city. The RP survey collected data on the main determinants of shared e-scooter usage in Braga and the primary needs Chapter II – State of the Art 16 More recently, a bicycle-sharing system was implemented in the “Comunidade Intermunicipal do Médio Tejo,” where sixty-eight bicycle stations are available in eleven municipalities, including Alcanena, Constância, Entroncamento, Ferreira do Zêzere, Mação, Ourém, Sertã, Tomar, Torres Novas, Vila de Rei, and Vila Nova de Barquinha. In these towns, 250 electric bicycles are available, and only in the first month of availability have the systems already reached more than 1,500 people who subscribed to the app (CIM Médio Tejo, 2024). Another essential feature of bike-sharing programs is the possibility of connection to public transportation services, which helps in the first and last mile of the journey. As a means to promote the integration of bicycles and public transportation, Wang & Chen (2020) state that there is a statistically significant and positive increase in shared bicycle usage when their availability near bus stops and subway entrances is promoted. II.1.3. E-scooter sharing The e-scooter-sharing service was first introduced in the United States in 2017 and has increased worldwide as a mobility trend. In Europe, an example of the spread of this system is Paris, where the first provider of self-service, free-floating e-scooter rental service was introduced in 2018 (6t-bureau de recherche, 2019c). As of early June 2019, the French capital hosted 12 free-floating scooter services. However, Parisians voted in 2023 to ban shared e-scooters from the city starting in September of the same year (Giuffrida, 2023) due to the high number of accidents involving this mode of transportation. Yet, the dockless e-scooter service model has taken off and witnessed aggressive growth in its early years (Figure 3). Several US-based scooter-sharing companies have reached unicorn status at lightning speeds as big investors pour millions of dollars into this business (CB Insights, 2019). Moreover, about 70% of Americans living in major urban areas and current shared e-scooter users worldwide view e-scooters positively. The positive perception is primarily due to their ability to expand transportation options, enable a car-free lifestyle, provide a convenient replacement for sort car trips, and complement public transportation (Populus, 2018a; Pourfalatoun et al., 2023). Impacts of Shared E-scooters on Urban Transportation Systems 17 Figure 3: Shared e-scooters in Europe (own author) In Asia, shared e-scooters have been gaining traction as a viable solution for short-distance public transportation trips because of their excessive indirectness, multiple transfers, and long access walking distances, mainly in urban areas. In Singapore, e-scooter-sharing services have been identified as an effective last-mile solution that complements public transportation services for longer journeys. In the Asian market, factors such as fare and walking distances significantly influence the choice between shared e-scooters and traditional public transportation modes (Cao et al., 2021a) However, some cities still consider shared e-scooters dangerous, with the case of Paris, which banned shared e-scooters from its streets in 2023. While many feared that other towns would follow this example, it has not happened, except for Madrid, which is banning shared e-scooters in October of 2024. Different cities have opted for cuts to fleet sizes rather than bans; this is the case of Berlin, which cut up to 6,000 shared e-scooters from the fleet, and Brussels, which opted to cut 12,000 e-scooters from the city in 2023 (Maxwell, Krieg, & Corbett, 2023). The opposite happened in Eastern Europe, where fleets have grown 33% in the same period. Guaquelin (2019) and Cao et al. (2021a) say that in France, like the U.S.A. and Singapore, the primary users of e-scooters are males with high income, with an average revenue of EUR 2,500/month in the Chapter II – State of the Art 18 U.S.A. versus an average revenue of EUR 2,202/month in Paris. However, a couple of differences can be found, as in Paris, 42% of the users were tourists, which lowers the importance of the shared e-scooter as an alternative for commuting. Students are over-represented despite the high price of the service. Companies usually charge a EUR 1.0 fee to unlock the e-scooter and more than EUR 0.15 per minute traveled. According to 6t-bureau de recherche (2019c), 58% of e-scooter users in France are wealthy local executives and students who use e-scooters in their city. The following 33% are foreign visitors, and the last 9% are French visitors who use e-scooters outside their town. The main motivations for using the services are related to leisure and the fun aspects of the e-scooters that make a ride pleasant, the time people can save in traffic by using this mode of transportation, and the possibility of taking door-to-door trips (Laa & Leth, 2020a). On the other hand, the main obstacles are price, feeling unsafe, and exposure to poor weather (Hardt & Bogenberger, 2017a). Research from NACTO (2019b) shows that people use e-scooters in U.S.A. cities mainly for recreation and exercise, to go to and from work, to connect to public transportation, and for social reasons. In addition, a survey conducted in Germany (Hardt & Bogenberger, 2019) shows that e-scooters were used mainly for commuting, leisure, and business trips within the city. These trip purposes represent a chance to establish e-scooters as an alternative to cars since commuting and leisure trips are still dominated by car usage. In Singapore, e-scooters are primarily used to replace traditional public transportation modes that do not meet people’s needs (Cao et al., 2021a) However, the modal shift in trips can also represent changes in people’s walking and bicycling patterns. Chang, Miranda-Moreno, Clewlow, & Lijun Sun (2019) show that in a survey deployed in the U.S.A., more than half of the respondents said that e-scooter sharing essentially replaces walking and cycling. Moreover, 46% of respondents stated that they would have walked or cycled if a shared e-scooter had not been available for their last trip (A. Chang et al., 2019). In France, a survey showed that 44% of local users would have walked to take their last trip instead of using a free-floating e-scooter. 30% would have used public transportation, but only 6% have taken fewer urban transportation options since they started using e-scooters. Moreover, 23% of these trips are Impacts of Shared E-scooters on Urban Transportation Systems 19 intermodal, which combines the e-scooter with another mode of transportation, such as public transportation for 66% and walking for 19% (Polis, 2019). Trips made by e-scooters have their limitations. This mode of transportation could perform better in hilly areas or brick-line streets; it needs to be improved for lousy weather, and riders need somewhere to stow groceries or other belongings (Schellong et al., 2019a). E-scooter use can also cause difficulties for pedestrians on sidewalks. Several studies have reported that e-scooters disrupted pedestrian travel and blocked pedestrian right-of-way (Fang et al., 2018; Radavoi & Potter, 2019). II.2. Worldwide use of shared e-scooters The advent of shared e-scooter schemes in 2017 changed how people commute in many cities worldwide, such as Madrid, San Diego, Santa Monica, and Portland. To exemplify, in 2012 and 2013, 2% of the population over the age of 18 in metropolitan areas were members of carsharing services (Populus, 2018b), while in less than one year of availability in some cities, e-scooter sharing has experienced an average adoption rate of 3.6% as measures by the percentage of people who have used these services (Populus, 2018c). Nearly 90% of the micromobility investments were going to shared e-scooter companies in the core of Asia, Europe, and North America, representing approximately more than EUR 8 billion from 2028 to 2022 (Heineke et al., 2022). In the United States of America, by August 2020, 145 e-scooter-sharing schemes were serving around 40 cities across the country (U.S. Department of Transportation, 2020a). The number of e-scooter schemes increased through 2019, then declined by almost 40% from 2019 to 2020 due to the Covid-19 pandemic restrictions. In 2018, there were 82 e-scooter systems throughout the USA, while in 2019, 115 systems were seen, and in 2020, only around 40 remained (U.S. Department of Transportation, 2020a). After the pandemic of COVID-19, in 2022, dockless e-scooter trips dropped by 10% due to the end of operations of some companies in the North American market, which made people take 58.5 million trips on dockless shared e-scooters in 2022, compared to 65 million in 2021 (Nacto, 2023). Some cities that provide e-scooter sharing services in the U.S.A. and the companies running them can be found in Table 1. Chapter II – State of the Art 20 Table 1: Shared e-scooter services in the U.S.A. by 2020 (U.S. Department of Transportation, 2020b) City Companies Operating (by 2020) San Francisco, CA Spin, Scoot San Diego, CA Bird, Spin, Lyft, Wheels Los Angeles, CA Bird, Jump, Lyft, Spin, Sherpa, Wheels Long Beach, CA Bird, Lime, Gruv, Razor Santa Monica, CA Bird, Lyft Freemont, CA Hopr Portland, OR Spin, Razor, Lime, Bird Boise, ID Lime, Bird, Spin Bozeman, MT Blink Rides Salt Lake City, UT Bird, Lime, Spin, Razor Ogden, UT Lime Farmington, UT Spin Denver, CO Lyft, Razor, Spin, Bird, Lime Tucson, AZ Razor (Pilot) Tempe, AZ Razor Austin, TX Bird, Lime, Spin, Wheels Corpus Christ, TX Blue Duck San Antonio, TX Bird, Razor Dallas, TX Lime, Bird, Wheels Plano, TX Boaz Bikes, Razor Oklahoma City, OK Lime Stillwater, OK Bird, Lime Kansas City, KS RideKC scooter, Bird, Spin Wichita, KS Spin, Voeride Minneapolis, MN Bird, Lyft St. Paul, MN Lime Detroit, MI Bird, Spin Ann Arbor, MI Spin Lansing, MI Gotcha Nashville, TN Lime Atlanta, GA Bird, Spin, Helbiz, Veoride Tampa, FL Spin, Bird, Lime Fort Pierce, FL Zagster Orlando, FL Wheels Charlotte, NC Bird, Lime, Spin Blacksburg, VA Spin Virginia Beach, VA Bird, Lime, Veoride, Spin Washington, D.C. Jump, Lyft, Skip, Spin Baltimore, MD Lime, Spin Providence, RI Veoride, Spin While in 2020, only around 40 cities in the U.S.A. provided shared e-scooter services, by 2022, there were 286 shared e-scooter services across the U.S.A. However, this also represents a decrease from the Impacts of Shared E-scooters on Urban Transportation Systems 21 peak number of 311 systems in 2021. Figure 4 shows the sizes of shared e-scooter services in North America by 2022. Figure 4: Shared e-scooter systems size in North America by 2022 (NACTO, 2022) Despite the quick rise of the service in many North American cities and the growth in shared e-scooter usage in the years preceding the COVID-19 pandemic, the current state of availability shows a decrease in the number of vehicles present in cities, which has affected the number of trips generated. Following this pattern, European cities face the same trend regarding shared e-scooter availability and usage in their towns. In several European cities, many e-scooters and free-floating bicycle schemes have been rolled out, and 2019 can be considered the year of the shared e-scooter (van Wijnendaele, 2019). The early days of shared e-scooters in Europe show that in Brussels, for example, there were around 3000 free-floating escooters and bicycles by 2019, and the e-scooters are mainly used for 2 km to 3 km trips (van Wijnendaele, 2019). In Paris, more than 20,000 shared e-scooters (Crellin, 2019) used to be rented daily for commuting during the weekday for a 19-minute trip (6t-bureau de recherche, 2019b). This massive adoption of shared e-scooters is also promoted by enlarging internet access, allowing people to rent the e-scooters through mobile apps and share their experience through user-generated content (UGC) (Owusu et al., 2016). Chapter II – State of the Art 22 A survey in Europe showed that among countries with shared e-scooter systems (Austria, Belgium, Czech Republic, Denmark, Germany, Greece, Finland, France, Hungary, Italy, Netherlands, Norway, Poland, Portugal, Serbia, Spain, Sweden, and Switzerland) there is a pattern in the number of shared e-scooters per resident in different countries, such as Denmark, Norway, Sweden and Germany (Table 2) (Kamphuis & van Schagen, 2020). Even if the global number of shared e-scooters differs in distinct contexts, it usually represents a similar ratio of vehicles per inhabitant. Table 2: Availability of shared e-scooters in different European countries (Kamphuis & van Schagen, 2020) Location Number of shared escooters Shared e-scooters per resident Denmark 7,000 1/1000 Norway 11,000 2/1000 Sweden 17,000 2/1000 Germany 50,000 0.6/1000 In other countries, only the estimate of shared e-scooters in a town or city was possible, such as Thessaloniki city (Greece) with more than 100 shared e-scooters (e-scooter per resident: 1/1000), Lisbon (Portugal) with around 4,000 shared e-scooters (e-scooter per resident: 8/1000), and Stockholm (Sweden) with around 9,000 shared e-scooters (e-scooter per resident: 9/1000) (Kamphuis & van Schagen, 2020). Some companies that dominate the market also influence the spread of shared e-scooter services in Europe. Table 3 shows data from three companies that provide shared e-scooter services in Europe and the cities where the offer is present by 2021. Table 3: Shared e-scooter services in Europe (Bird, 2021; Bolt, 2021; Lime, 2021) E-scooter company Country Cities Bolt Czech Republic Boskovice, Brno, Frydek-Mistek, Mladá Boleslav, Olomouc, Pardubice, Prague Estonia Parnu, Talin, Tartu Bolt France Bordeaux Croatia Osijek, Rijeka Lithuania Kaunas, Klaipeda, Panevezys, Vilnius Latvia Daugavpils, Jelgava, Liepaja, Riga, Valmiera Malta Malta Norway Bergen, Fredrikstad, Lillestrom, Oslo Impacts of Shared E-scooters on Urban Transportation Systems 23 E-scooter company Country Cities Bolt Poland Bialystok, Gdansk, Gdynia, Katowice, Krakow, Lublin, Poznan, Silesia, Sopot, Trojmiasto, Warsaw, Wroclaw Portugal Braga, Lisbon, Romania Bucharest, Cluj-Napoca, Constanta, Galati, Sweden Gothenburg, Stockholm Lime Austria Vienna Belgium Brussels Bulgaria Sofia Czech Republic Brno, Pilsen, Prague Denmark Copenhagen Finland Helsinki France Lyon, Toulouse Germany Berlin, Bonn, Brunswick, Cologne, Dortmund, Dresden, Dusseldorf, Essen, Frankfurt, Hamburg, Hannover, Karlsruhe, Munich, Nuremberg, Ruhrpott, Stuttgart Greece Chania, Rethymno, Thessaloniki Hungary Balaton, Budapest Italy Milan, Rimini, Rome, Turin, Verona Norway Oslo Poland Krakow, Poznan, Tricity, Warsaw, Wroclaw Portugal Lisbon Romania Bucharest, Cluj-Napoca Spain Barcelona, Madrid, Malaga, Seville Sweden Gothenburg, Malmo, Stockholm United Kingdom London, Milton Keynes Bird Austria Vienna Belgium Antwerp, Brussels France Bordeaux, Bretigny-sur-Orge, Marseille, Orange, Villemonble, Viry-Chatillon Germany Berlin, Chemnitz, Cologne, Darmstadt, Dortmund, Dusseldorf, Erfurt, Frankfurt, Gelsenkirchen, Gottingen, Hamburg, Hannover, Heidelberg, Heilbronn, Karlsruhe, Kassel, Ludwigshafen, Munich, Neckarsulm, Neu-Ulm, Oldenburg, Pforzheim, Regensburg, Reutlingen, Rostock, Troisdorf, Ulm, Wurzburg Bird Italy Aprilia, Firenze, Milano, Palermo, Pesaro, Rimini, Rome, Torino, Verona, Viareggio Norway Bergen, Oslo Portugal Braga, Lisbon, Porto Spain Madrid, Malaga, Terragona, Zaragoza Sweden Gothenburg, Stockholm Switzerland Basel, Winterthur, Zurich UK Canterbury, Redditch Chapter II – State of the Art 24 The data from Table 3 shows that more than 18 European countries offer a shared e-scooter service by 2021, reaching big and small cities in different countries. By 2023, 91 cities in Europe offered shared escooter services (less four than in 2022), with a fleet size of approximately 514,000 e-scooters, representing 64% of the shared vehicles fleet in Europe (Maxwell, Krieg, Venkatesh, et al., 2023). Even with the high number of shared e-scooters in the urban fleet, their presence has shrunk since 2022, the same as in the U.S.A. Figure 5 shows Europe's shared e-scooter fleet difference between 2022 and 2023. Figure 5: Shared e-scooter fleet in Europe in comparison to 2022 (Maxwell, Krieg, & Corbett, 2023) In one year, Eastern Europe experienced a high rise in the shared e-scooter fleet. At the same time, southern Europe was affected by a 33% decrease in the same year, followed by France and Benelux, primarily due to the ban on shared e-scooter usage in Paris. The differences in the fleets of shared escooters in the U.S.A. and Europe show that after an exponential rise in usage and fleet during the first years of operation, shared e-scooters have experienced a decrease in availability and use worldwide. The decline can be attributed to companies struggling to cope with market competition and some bans that occurred in some cities, such as Paris and Madrid. The Asian-Pacific market for shared e-scooters has been growing in recent years; however, data on usage and information on the patterns of geographic vehicle distribution are still scarce. This is the world's third biggest shared e-scooter market, only behind Europe and North America, representing a 10% market share. The number of users of shared e-scooters in Asia is expected to continue growing in the following years, reaching 19.25 million by 2029. The projected penetration of 0.3% in 2024 is expected to rise to 0.4% by 2029 (Statista, 2023). Even if the market's growth is projected for the following years, shared e- Impacts of Shared E-scooters on Urban Transportation Systems 25 scooters are still banned in most parts of China, for example. In 2024, a trial was proposed in Suzhou, China to introduce 10,000 shared e-scooters and e-bikes by the end of the year, aiming to demonstrate the safety and potential of shared micromobility through the controlling context of a trial (Musa, 2024). II.3. Usage and users’ profile of shared e-scooters Temporal, sociodemographic, and built-environment variables have influenced the usage of shared escooters in cities. However, the profile of the population exposed to shared e-scooter services highly influences the usage of this mode of transportation. Hence, understanding the relationship between the usage of shared e-scooters and the sociodemographic characteristics of the population can be crucial to establishing e-scooters as a mode of transportation (Nikiforiadis et al., 2021). Therefore, some researchers have indicated that shared e-scooters are more used in areas with more employment rates between residents and where bicycle infrastructure is higher (Mehzabin Tuli et al., 2021). Mixed-use regions have the most minor portion of non-recreational trips. In contrast, downtown areas have the highest usage of shared e-scooters for non-recreational trips and institutional-oriented mixed-type regions (Liu et al., 2020). Shared e-scooters also reach their peak usage in the middle of the day or the evenings for trips focused on recreation and educational usage (Dibaj et al., 2021). Besides, the presence of commercial regions, along with an encouraging environment for active transportation (e.g., walking and cycling), and the proximity to public transportation stations and hubs were also correlated with high usage of shared e-scooters, which can suggest that this mode of transportation can be associated positively with sustainable development (Hosseinzadeh et al., 2021a; Latinopoulos et al., 2021). In addition, the presence of shared e-scooters near and on university campuses, combined with the ability to appreciate the ease of use of this “new” mode of transportation, the availability of smartphones, and online payments, make university students the main adopting group of shared e-scooter services around the world (Fistola et al., 2022). Students and other specific groups of users have been reported to be constantly using shared e-scooters. In Paris, for example, shared e-scooter users used to be 18 to 29-year-old males with a high educational level. They can be students or executives significantly wealthier than the general French population (6t- Chapter II – State of the Art 32 transportation, some strategies need to be taken, such as increasing the lifespan of e-scooters, the reduction of collection and redistribution of e-scooters by using more efficient vehicles, as well as making less frequent battery charges (Bortoli, 2020; Hollingsworth et al., 2019; Reis et al., 2023b). In the case where a shared e-scooter lifespan could increase to 6,500 km, a reduction of about 28% in GHG emissions could be achieved, while a lifespan increase to 10,000 km could lead to an emission reduction of 46% (EU Urban Mobility Observatory, 2023). As a result of investments to develop shared-scooter service and lifespan, in Munich, Germany, shared escooters are seen as a benefit to the environment, as they have low energy consumption, produce less noise and fewer air pollutants, as well as help overcome congested, exhausted streets and over-crowded parking spaces by substituting cars. The data from the pilot program in Munich shows that shared escooters could cover up to 60% of daily trips by replacing cars, especially the ones for commuting and leisure purposes. Also, up to 50% of all trips in Munich are suitable for an e-scooter (Hardt & Bogenberger, 2017b). Despite the possible complaints from dwellers about pedestrian safety and the impact of the emerging shared e-scooters on European urban public spaces. They are believed to help cities ease problems with traffic, emissions, and parking (Hardt & Bogenberger, 2019; Zagorskas & Burinskien˙, 2019). A survey conducted in Brussels showed that 25% of the respondents said that one of the reasons for them to use shared e-scooters was to lower the air pollution in the city (Moreau et al., 2020), which can translate into a positive effect on human footprint, especially if shared e-scooter trips replace private car trips (C. C. Chang et al., 2016). Besides, using micromobility in urban logistics to replace traditional combustion engine vehicles has been demonstrated to have environmental and social effectiveness (Nocerino et al., 2016). If considering the social benefits of shared e-scooter usage, it was deployed in cities as a beacon of hope to reduce the gender gap in transportation use, as women would feel safer on e-scooters because they are smaller in comparison to other modes of transportation (International Transportation Forum - ITF, 2020) and are more easily ridden on sidewalks; along with the possibility of wearing skirts and dresses that make it easier to stand on an e-scooter than on a bicycle. Women are also more distance-sensitive and are less likely to bike long distances, and both e-bikes and e-scooters enable everyone to travel greater distances easily (CB Insights, 2019). However, recent literature has shown that women feel unsafe on Impacts of Shared E-scooters on Urban Transportation Systems 33 shared e-scooters and constantly avoid this mode of transportation (Campisi et al., 2021; Gauquelin, 2019). Furthermore, new micromobility solutions such as shared e-scooters can help cities develop policies to incentivize the equitable spatial deployment of micro-vehicles (Caspi et al., 2020; Clewlow et al., 2018). One possibility could be creating programs and policies to distribute shared e-scooters in cities considering undeserved developments and focusing on portions of the city with the highest social impact (Hosseinzadeh et al., 2021b; Olabi et al., 2023). Lee et al. (2021) also point out that low-income people prefer shared e-scooter services for first-mile and last-mile trips. Still, governments should take particular action to promote a policy reducing the fares of shared e-scooters and fare integration systems, allowing for multiple modes of public transportation. Thus, disadvantaged groups could be integrated to the group of shared e-scooter users and gain in accessibility compared to the rest of the population (Abouelela et al., 2024). For economic purposes, shared scooters primary motivation for usage can be seen as travel time savings followed by playfulness and money savings (Christoforou et al., 2021a), as they are cheaper than hailing a ride-share vehicle (Schellong et al., 2019b). For urban drivers on short-distance trips, shared e-scooters would be a more cost-effective alternative to car ownership (Smith & Schwieterman, 2018). However, even if shared e-scooters can be more affordable than car ownership and ride-hailing services, one of the main drawbacks for its usage is still the price of the trip. Thus, some cities have required that shared escooter companies address equity issues by supplying a specific number of e-scooters in underserved areas and offering a low-income fare for the resident population (McQueen et al., 2021). Also, the business model for shared e-scooter deployment and their economic sustainability emphasizes functionality over ownership. It incorporates practices such as increasing e-scooter lifecycles, recycling, and using renewable energy to enhance sustainability. However, shared e-scooter services still face challenges in economic sustainability, requiring careful consideration of business operations and policies to mitigate adverse impacts (Sundqvist-Andberg et al., 2021). The number of trips generated, which translates into profit for e-scooter companies, and the span of the vehicle's lifecycle can lead the business to profitability, as well as environmental advancements, through Chapter II – State of the Art 34 the longer usage of e-scooters. These are some of the aspects that contribute to making shared e-scooters economically sustainable. Table 7 shows a collection of references that approach the three aspects of sustainability in shared escooter usage to shed light on how this service can be used in urban environments to achieve more sustainable mobility. Table 7: Sustainability in shared e-scooter usage Reference Sustainability Aspect Methodology Environmental Social Economic Chang et al., 2016 X LCA Clewlow et al., 2018 X Geospatial analysis Populus, 2018 X Geospatial analysis Zagorskas and Burinskien˙, 2019 X Review Smith and Schwieterman, 2019 X X Geospatial analysis Hollingsworth et al., 2019 X LCA Schellong et al., 2019 X Review Hardt and Bogenberger, 2019 X Pilot study Fong and Mcdermott, 2019 X Review Moreau et al., 2020 X LCA Gössling, 2020 X X Review Caspi et al., 2020 X Geospatial analysis Holm Moller and Simlett, 2020 X Review Bortoli, 2020 X LCA/Survey Christoforou et al., 2021 X X Survey Hosseinzadeh et al., 2021 X Review Lee et al., 2021 X Survey Almannaa et al., 2021 X X Survey The environmental effects of shared e-scooter usage are still the main research subject in recent literature. In contrast, its social and economic effects are still understudied. On the other hand, geospatial analysis, Life Cycle Assessment (LCA), and discussions/revisions are the main methods used to describe the impact of shared e-scooters on cities. Impact assessment and attitudinal studies are needed to deepen the studies on this matter. Impacts of Shared E-scooters on Urban Transportation Systems 35 II.7. Impacts of the COVID-19 pandemic on urban mobility and the role of shared escooters In the first quarter of 2020, a virus started spreading worldwide, resulting in lockdowns, reduced displacements, and social distancing. At that time, shared e-scooters had been deployed in cities for nearly two years. Therefore, it was unknown what would happen to this mode of transportation amid the changes in travel behavior on all continents. Hence, the assessment of the usage or not usage of shared e-scooters during the pandemic became a subject of interest for this research since, in times of uncertainty, this “new” mode of transportation could not have returned to cities after the mobility restrictions were lifted. Before the COVID-19 pandemic started, 52% of Europeans used a car as their primary mode of transportation for daily trips, while only 1% used a shared bicycle or e-scooter (European Commission, 2020a). This pandemic induced several behavioral changes that disrupted urban mobility, including avoiding crowded spaces, such as public transportation (Kakimoto et al., 2020). Therefore, it was of utmost importance to understand the role of shared e-scooters in urban mobility during the pandemic and how urban mobility could change after the restrictions were lifted. To stop the spread of COVID-19, people started avoiding crowded public transportation in many cities worldwide (De Vos, 2020), culminating in historic low passenger demand for public transportation use (Governors Highway Safety Association, 2020). To bypass the shift from public transportation to individual vehicles, cities worldwide joined forces to invest in infrastructure to increase micromobility usage, primarily to address daily mobility needs among commuters. In order to increase micromobility usage, many countries adopted specific policies, including financial support, for the relaunch of transportation through more sustainable modes, which included an investment of EUR 100 million in Italy to incentivize the transition to sustainable solutions, such as electric vehicles, e-bikes, and e-scooters, as well as the possibility for e-scooters to become part of the public transportation supply (Lozzi et al., 2020). To boost micromobility usage during the pandemic of COVID-19, Rotterdam, Netherlands, partnered with micromobility providers, such as Donkey Republic, Felyx, Check, Go Sharing, and Rotterdam Ahoy, to Chapter II – State of the Art 36 ensure that 1,500 shared bicycles and more than 1,500 shared e-scooters were available in more than 25 public transit hubs throughout the city (RET, n.d.). Milan, Italy, invested in the Lazzaretto and Isola pilot projects to create more protected and accessible roads for everyone. The municipality offers new public spaces for adults and children. It encourages travel by foot/bicycle/e-scooter for urban displacements through a diversified, complementary, alternative offer to public transportation and private cars. The municipality aims to free up public space by temporarily pedestrianizing some roads, widening sidewalks and connections with existing cycle paths, implementing new tactical urban planning interventions, and installing outdoor areas (Comune di Milano, 2020). In Rome, drastic measures were taken by the capital municipality to reduce dependence on personal automobiles. E-scooter operators such as Bird supported this effort by promoting micromobility as an everyday alternative to cars, allowing for social distancing (Bird, 2020). The e-scooter provider Lyft, with its program Lyftup, started offering free 30-minute rides for first respondents, the transit workforce, and healthcare providers in Denver, Colorado; Los Angeles, California; the Washington, D. C. area; San Diego, California; and Santa Monica, California. This program is in addition to bicycle share programs offering free membership for critical workers on networks that Lyft operates, including City Bike (New York), Divvy (Chicago), Bluebikes (Boston area), and Bay Wheels (SF Bay Area) (Lyft, 2020). In Europe, the shared escooter company Bird equipped Red Cross volunteers with a donated fleet of Bird e-scooters to enable more accessible transportation around urban neighborhoods (Bird, n.d.) so they could provide healthcare assistance to those in need. Other than that, to evaluate and study the resilience of alternative mobility systems such as public transportation, bicycles, and e-scooters during the COVID-19 pandemic, shared micromobility companies such as Bird, Spin, Biketown Spin, and TriMet are sharing their survey data on riders’ habits, preferences, and attitudes towards various shared modes of transportation with researchers (Goddard, 2020). Data from this study shows that almost 90% of respondents said they do not avoid taking shared e-scooter trips because of COVID-19; also, nearly 20% of respondents said they replaced public transportation trips with e-scooter trips during the pandemic (Cherry et al., 2020). Furthermore, this mode of transportation could promote well-being and sustainable and safe transportation for people who need it the most when the pandemic started. This leads to the advantages Impacts of Shared E-scooters on Urban Transportation Systems 37 of having shared systems in cities to help decrease the need for “traditional” modes of transportation (Hamad Almannaa et al., 2021), such as vehicles and public transportation, which became an unsafe option during the COVID-19 pandemic. In many cities, shared e-scooter services also played an essential role in not letting people rely only on cars to avoid public transportation, consequently increasing urban mobility and environmental problems related to the massive usage of private vehicles. However, some cities worldwide opted for shutting down the shared e-scooter services in order to discourage urban trips. Braga, Portugal was one of these cities, where shared e-scooters were collected from the streets and riders could not use this vehicle to displace, which resulted in an increase in car usage to avoid public transportation. Therefore, more information on the impacts of the COVID-19 pandemic on shared e-scooter usage in Braga will be presented in the case study section (Chapter IV). II.8. Synthesis of main considerations Micromobility and its shared form are seen as an opportunity to increase the quality of life in cities by achieving environmental goals and decreasing traffic congestion and noise. Shared bicycle services have been around for a long time in cities, contributing to the eminent benefits brought by its usage and the knowledge on how it should be used and ridden to promote safe and sustainable mobility. Shared e-scooters, on the other hand, were first deployed in cities in 2017. At first, this new mode of transportation was revealed to be an opportunity for cities to invest in a clean, fast, economically viable, equitable, and fun way of displacement. However, the shortcoming of shared e-scooters was the increase in the number of accidents involving not only riders but also pedestrians, the expansion of e-scooters “flooding” sidewalks and disrupting pedestrians’ right-of-way, as well as the unknown benefits for the environment in its current business model. The unplanned deployment of shared e-scooters displayed the differences between what was considered this mode of transportation and what it is in cities. Wealthy, well-educated men are the primary users of this mode of transportation in downtown areas, near business districts, and university campuses. However, cities' commitment to improving mobility showed that pilot programs and trials could disrupt the negative impacts caused by shared e-scooters, by testing procedures and measuring how they can affect the safe and sustainable usage of this mode of transportation. Chapter II – State of the Art 38 The primary outcomes from the pilots and trials show that shared e-scooter companies and authorities need to work together to offer a sustainable, safe, equitable, and economically viable service for the population. Firstly, the rules for where shared e-scooters must be ridden must be clear, even if they differ depending on the location. Secondly, the recharging of batteries and the collection of e-scooters need to be done by non-polluting modes of transport, and the available fleet must be offered not only downtown but also in underprivileged communities with special discounts that allow people access to the service. The positive outcomes from shared e-scooter deployment in cities can also be felt in disruptive events like the COVID-19 pandemic. Subsection II.7 brings information to answer RQ 1 concerning the usage of this mode of transportation to promote safe and inclusive mobility to respond to the rejection for crowded public transportation options worldwide. When the pandemic spread and essential workers needed to make their journeys to their job sites every day, shared e-scooters were used as an alternative to public transport and the massive usage of private vehicles. If no other sustainable option were available for the population, the choice of the car would have been more prominent. Shared e-scooters were essential in providing a sustainable transportation option when they were available for essential workers in many cities. The discounts offered, the dedicated infrastructure that was available through pop-up cycle lanes that became effective after the mobility restrictions were revoked, and the programs to boost this mode of transportation through the cooperation of companies and governments made it possible to take people from public transportation where the virus could spread quickly, to a better option than the car. However, this was only the case in some cities. In Braga, the case study of this thesis, the shared escooter service was completely shut down during the pandemic to prevent people from making their daily journeys. The city also did not benefit from special programs for first respondents and essential workers to use e-scooters, and it lost the opportunity to create pop-up cycle lanes that could have been transformed into permanent infrastructure for micromobility. This decision to ban e-scooters during the pandemic could have caused an increase in car usage since the mobility restrictions were lifted in the country once people got used to using cars more often and declined public transportation usage. Impacts of Shared E-scooters on Urban Transportation Systems 39 Chapter III. Research Methodology This chapter presents the methodology used to conduct the research developed in this thesis. This research comprises a literature review presented in Chapter II and is used to complement the information presented in the case study (Chapter IV) to create two survey methods. These survey methods were deployed to assess the main determinants of shared e-scooter usage in the case study of Braga and how different attributes influence the choice of sustainable modes of transportation for commutes and leisure trips. Therefore, the methodology used to deploy the first survey is presented in section III.1, where the Revealed Preference (RP) survey is disclosed, and the statistical approach used to analyze the data retrieved from this survey is shown. Next, section III.2 presents the stated choice (SC) methodology related to the second set of surveys deployed to assess the likelihood of replacing a car with a sustainable mode of transportation in urban trips. The collection of the results from the case study, added to the information from the literature review, is used to draw policy recommendations for implementing shared e-scooters in cities. Figure 7 shows an illustration of the methodology used. The methodological approach starts with deploying a Revealed Preference (RP) survey to gather information on the main determinants of shared e-scooter usage, its main drivers, and the primary needs to boost it. The latter is used to compose the stated choice experiments for commute and leisure trips that are used to assess the impacts of shared e-scooters on sustainability, the influence of transportation planning on shared e-scooter usage, the main determinants for it, and the competitiveness of shared escooters and e-bikes. Chapter III – Research Methodology 40 Figure 7: Methodology used for the research III.1. Revealed preference (RP) Revealed preference (RP) has long been a fundamental approach in mobility planning, providing a robust framework for understanding and predicting consumer behavior. This approach posits that individuals' true preferences can be inferred from their observable choices rather than relying solely on self-reported preferences or stated intentions (Fischhoff et al., 1978). Revealed preference techniques have gained traction as a valuable tool for researchers and policymakers, allowing them to delve deeper into the complex drivers of consumer decision-making. Therefore, Revealed Preference (RP) surveys are used to report the actual behavior of travelers; thus, the data retrieved does not suffer from hypothetical biases (Rudloff & Straub, 2021). The revealed preference theory developed by Samuelson (1948) suggests that an individual’s true preferences can be best indicated by their choices or actions. Therefore, RP choices are partly governed by different agents, such as beliefs and desires (Holmes, 2022). One of the primary advantages of revealed preference surveys is their ability to capture actual behavior, thereby reducing the hypothetical bias often associated with stated choice surveys. Hypothetical bias occurs when there is a discrepancy between what people say they will do and what they do. A study on blood donation preferences found that stated intentions significantly overestimated actual donation Impacts of Shared E-scooters on Urban Transportation Systems 41 frequencies. Researchers could adjust their models by incorporating revealed preference data and reducing this overestimation, leading to more accurate predictions (de Corte et al., 2021). Furthermore, combining revealed and stated choice data can provide a more comprehensive understanding of consumer behavior. This approach leverages the strengths of both methods: the realworld accuracy of revealed preferences and the flexibility of stated choice to explore hypothetical scenarios. For example, in travel behavior, revealed preference data from actual travel patterns can be combined with stated choice data from surveys about hypothetical travel scenarios, such as the introduction of autonomous vehicles. This combined approach allows for more robust modeling and forecasting of future travel behaviors (Yu & Jayakrishnan, 2018). One of the critical applications of revealed preference surveys is the study of the mode of transportation choice. Researchers have used this approach to understand how individuals select their mode of transportation, such as car, bus, or train, based on their observed travel behavior (D. Basu et al., 2018). For example, a Mumbai, India study used a revealed preference survey to capture information on individuals' travel patterns, socio-economic characteristics, and attitudes. It then employed structural equation modeling to analyze the factors influencing their decision to share cab rides (Shah et al., 2020). Besides, RP surveys are often used in transportation to identify attributes for stated choice (SC) experiments. Bottero, Bravi, Caprioli, & Dell (2023) describe the usage of RP surveys to collect information from respondents on the features they value the most on a subject; then, this data is used to design the alternatives included in the SC experiment. Interestingly, recent research has explored the potential limitations of revealed preference techniques, highlighting the need for a more nuanced understanding of their underlying assumptions and methodological considerations. For example, studies have shown that consumers may employ different cognitive strategies when evaluating new products, which standard preference measurement techniques must better capture (Scheibehenne et al., 2015). The steps for creating the RP survey for this research, as well as the statistical treatment for the data collected are presented and explainded in details in Chapter IV, when the case study and the survey structure is presented. Chapter III – Research Methodology 48 and is asked to select one of the alternatives. Full factorial designs guarantee that all attribute effects of interest are genuinely independent and that the main interaction effects are considered (Louviere et al., 2000). However, the number of situations generated (i.e., scenarios) for the respondents to choose from can become large as the total number of alternatives, attributes, and attribute levels increase. The number of choice situations in the full factorial design can be calculated using Equation 1 (ChoiceMetrics, 2021a). 𝑆𝑓𝑓=∏∏𝑙𝑗𝑘 𝐾𝑗 𝑘=1 𝐽 𝑗=1 (1) Where:  𝑆𝑓𝑓= total number of choice situations;  J = alternatives;  K = attributes of the alternatives;  𝑙 = levels within the attributes. Because each decision maker is presented with each choice situation in a full factorial design, the workload for the respondent becomes excessively large in all but the smallest experiments (Hensher et al., 2005a). In addition, as the number of choice situations presented to a respondent increases, the amount of effort the respondent puts into analyzing the alternatives and selecting the most favorable alternative has been found to decrease (Louviere et al., 2000). One possibility for reducing the number of choice situations presented to each respondent is to divide the choice situations among the respondents instead of assigning all situations to all respondents. However, this can lead to biased outcomes (ChoiceMetrics, 2021a). Another option for limiting the number of treatment situations presented to each respondent is to systematically select the most critical situations for producing the desired statistical results, and one of the options is the fractional factorial or orthogonal designs. III.2.3.3. Fractional factorial/orthogonal design Researchers commonly employ orthogonal coding to label attribute levels when constructing an orthogonal fractional factorial choice experiment. This coding approach simplifies the creation of the Impacts of Shared E-scooters on Urban Transportation Systems 49 experimental design. Orthogonal coding ensures that the sum of each column of attribute levels equals zero. For instance, conventionally assigned values would be 1 and -1 when dealing with an attribute having two levels. Similarly, the labels would be 1, 0, and -1 for an attribute with three levels. Notably, it exclusively utilizes odd numbers, avoiding using the number 5 because orthogonal coding relies on balanced and orthogonal arrays designed to ensure that all levels of each factor are equally represented and that the factors are uncorrelated. Creating an orthogonal fractional factorial design through orthogonal coding can be laborious (ChoiceMetrics, 2021a). A minimum of six choice situations must be included to meet the requirements of attribute balance and degrees of freedom. However, the minimum number of choice situations may vary significantly based on the number of alternatives, attributes, and attribute levels incorporated into the experimental design (Louviere et al., 2000). When the number of choice situations is high, a technique called blocking is employed to reduce the survey burden on respondents. This method splits the design into smaller, non-orthogonal designs while ensuring that the sum of these designs maintains orthogonality (Hensher et al., 2005a). Recent advancements in stated choice experimental design theory, particularly the introduction and refinement of efficient stated choice designs, have brought attention to several limitations of the orthogonal fractional factorial method. Maintaining orthogonality within the dataset when creating choice models is atypical for several reasons (ChoiceMetrics, 2021a). Firstly, orthogonality is compromised when data is missing for any choice situations. If blocking is employed and a respondent fails to complete their choice task, the entire dataset’s orthogonality is jeopardized (Louviere et al., 2000). Secondly, the transition from design codes to orthogonal codes can result in a loss of orthogonality (ChoiceMetrics, 2021a). However, ChoiceMetrics (2021a) acknowledges that orthogonal design is ease of construction or acquisition, whether through software packages or academic literature, which facilitates their widespread use. Also, the historical impetus behind the prevalence of orthogonal designs in studies related to stated choice (SC) stems from the experimental design literature’s historical focus on linear models, particularly linear regression models. In such models, orthogonality serves two critical purposes: (a) preventing multicollinearity and (b) minimizing parameter estimate variances, which are derived from the variancecovariance (VC) matrix of the model (Equation 2). Chapter III – Research Methodology 50 𝑉𝐶=𝜎2[𝑋′𝑋]−1 (2) Where:  𝜎2 = the model variance;  𝑋 = the matrix of attribute levels in the design or in the data to be used in the estimation. Controlling for the model variance, which serves as a mere scaling factor, the variance-covariance (VC) matrix elements in linear regression models are minimized when the design matrix (X matrix) is orthogonal. Opting for a design that minimizes the VC matrix elements offers two distinct advantages. Firstly, such a design yields the minor possible standard errors (i.e., square roots of the variances), thereby maximizing the t-ratios derived from the model. Secondly, an orthogonal design (or dataset) ensures zero-off diagonals in the VC matrix, effectively disentangling parameter estimates and mitigating multicollinearity. Orthogonal designs, particularly in the context of linear models, satisfy the two essential criteria for effective experimental design outlined in the introduction. Firstly, they facilitate the independent assessment of each attribute’s impact on the dependent variable. Secondly, they enhance the design’s statistical power to detect significant relationships (as reflected in t-ratios) at any given sample size. This means that the different attributes used are combined in the design so that the attribute’s effects can be estimated independently of the others, and that the attributes are uncorrelated, allowing for a clear and unbiased estimation of each attribute’s impact on the choice made by respondents. III.2.4. Building stated choice experiments from revealed preference data Recent literature has highlighted the potential enhancement of choice experiment realism by constructing alternatives based on existing situations (Starmer, 2000). Using a reference alternative may yield more meaningful disaggregated data, as participants can contextualize the choice task using their existing memory (Rose & Bliemer, 2004). The recommended approach involves using both data sources, namely revealed preference and stated choice data, which allows one to exploit their advantages and overcome their limitations (Ben-Akiva & Morikawa, 1990b; Louviere et al., 2000). A methodology adopted by Cherchi & Ortúzar (2002) focuses on retrieving revealed preference in the early stages of the stated choice experiment's development to Impacts of Shared E-scooters on Urban Transportation Systems 51 gather better information on people’s attitudes towards a particular mode of transportation. These authors conducted an RP survey to obtain data on actual choices and select a reference trip using a customized SC design. This approach offers the advantage of presenting attributes and attribute levels that align more closely with respondents’ expectations, thereby creating more realistic choice scenarios (Hensher et al., 2005a). When designing a pivoted stated choice experiment, careful consideration of the attribute levels used for pivoting is essential, as they significantly impact the situations/scenarios presented to respondents and the usability of the resulting data. According to Rose & Bliemer (2004), a promising approach for constructing a stated choice experiment is to base its design on revealed preference data. In this method, respondents participate in an initial survey phase that collects revealed preference (RP) data. Subsequently, this RP data is utilized to create a stated choice experiment in the second phase. From a statistical perspective, this two-phase approach is commendable because it generates a data-specific design by minimizing values within the attributevalue-cost (AVC) matrix using the same dataset employed for model estimation (Rose & Bliemer, 2004). III.3. Theoretical framework According to Ortúzar et al. (2008), the random utility theory is the most common theoretical framework for generating discrete choice models. This theory assumes that individuals belong to a homogeneous population, act rationally, and possess perfect information. This means that they always choose to be in a situation that maximizes their net personal utility subject to legal, social, physical, and budgetary (in time and money) constraints. In this theory, a given individual is supplied with attributes to face different choice sets according to the available alternatives and the vectors (levels) of attributes. It is then assumed that one’s choice set is predetermined, implying that the effect of constraints has already been taken care of and does not affect the selection process among the available alternatives. Each respondent's option is associated with a net utility for the individual. The researcher does not have complete information about all the elements the respondents (choice makers) considered. Therefore, the utility is assumed to be represented by two components: Chapter III – Research Methodology 52  A measurable, systematic, or representative part 𝑉𝑗𝑞, which is a function of the measured attributes X;  A random part 𝜀𝑗𝑞 reflects each individual's peculiarities and particular tastes, together with any measurement or observational errors made by the researcher. Therefore, the random utility theory can be formulated according to equation 3. 𝑈𝑗𝑞=𝑉𝑗𝑞+𝜖𝑗𝑞 (3) The random utility theory formulation allows two apparent illogicalities to be explained: that two individuals with the same attributes and facing the same choice set may select different options, and that some individuals may not always select the best alternative III.4. Logit model Logit models are a fundamental tool in statistical analysis and are convenient for modeling binary and multinomial outcome variables. These models estimate the probability of a given event based on one or more predictor variables. Multinomial and Mixed logit models represent an advanced extension of the basic logit model, allowing for random heterogeneity in the population. This means the model can account for variations in preferences or behaviors not captured by observed variables. These models are instrumental in discrete choice modeling, where they can incorporate both revealed preference and stated choice data (Hensher & Greene, 2003). Another variant, nested logit models, is designed to handle situations where choices can be grouped into a hierarchical structure. These models are advantageous in scenarios where alternatives can be naturally grouped based on attributes, such as flight itineraries grouped by departure time and number of stops. Despite their utility, nested logit models can be computationally intensive, especially when dealing with large datasets (Carrasco & De Dios Ortuzar, 2002). Recent advancements have focused on developing more efficient estimation methods for large-scale applications. Impacts of Shared E-scooters on Urban Transportation Systems 53 In addition to these specific models, various specification tests are available for logit models. These tests can be used to check for omitted variables and heteroskedasticity, ensuring that the model is correctly specified. Lagrange Multiplier tests, for example, can be computed inexpensively and help identify potential issues in the model (Davidson & Mackinnon, 1984). Logit models and their extensions offer powerful tools for modeling binary outcomes and discrete choices. This model was chosen because the coefficients in a logit model can be interpreted as the change in the odds of the dependent variable for a one-unit change in the predictor variables; they are computationally simpler and faster to estimate, more widely used and understood in different fields, and often more robust to deviations from the underlying assumptions compared to other models. III.4.1. Binary logit model Binary logit models are a fundamental tool in statistical analysis, particularly useful for modeling binary response variables. These models are a type of regression model where the dependent variable is binary, taking on values of 0 or 1. The primary goal of binary logit models is to estimate the probability that a given observation falls into one of the two categories based on one or more predictor variables. The basic form of a binary logit model involves the logistic function, which ensures that the predicted probabilities lie between 0 and 1. The logistic function is defined as equation 4 (Ben-Akiva & Lerman, 2000): 𝑃𝑛(1)=1 1+𝑒−𝜇𝑉𝑛 (4) One of the critical advantages of binary logit models is their ability to handle non-linear relationships between the dependent and independent variables. This is particularly useful when the relationship between variables is often complex and not strictly linear (Cramer, 1998). However, interpreting the coefficients in a logit model can be more complex than linear models. The coefficients represent the change in the log odds of the dependent variable for a one-unit change in the predictor variable. This can be transformed into odds ratios, which are easier to interpret but should be used cautiously, especially when comparing different studies or models (Norton & Dowd, 2018). Chapter III – Research Methodology 54 Extensions of the binary logit model include the bivariate binary logit (BBL) model, which handles two correlated binary responses. This model is advantageous when dealing with data that can be represented in a 2x2 contingency table and follows a multinomial distribution. The BBL model uses maximum likelihood estimation and provides robust hypothesis testing methods (Purhadi & Fathurahman, 2021). Another extension is the dynamic logit model, which incorporates time-dependent covariates and can handle repeated measurements of binary responses. This model is helpful for longitudinal data analysis, where the same subjects are observed under different conditions over time (Zheng & Sutradhar, 2018). Despite their widespread use, binary logit models have limitations. For instance, they can be sensitive to unbalanced samples, where the frequency of one outcome is much lower than the other. This imbalance can lead to biased prediction probabilities and affect the model's performance. Partial remedies, such as adjusting the model fit or using alternative diagnostics, can somewhat mitigate these issues (Cramer, 1998). Moreover, binary logit models are a versatile and powerful tool for analyzing binary response data. They offer flexibility in modeling complex relationships and can be extended to handle correlated responses and repeated measurements. However, careful consideration must be given to interpreting coefficients and handling unbalanced samples to ensure accurate and meaningful results. III.4.2. Multinomial logit model (MNL) Multinomial logit models (MNLs) are powerful statistical tools for predicting outcomes where the dependent variable consists of multiple non-ordinary categories. These models are instrumental in marketing, health economics, industrial organization, and transportation, where researchers need to analyze choices among numerous alternatives. The MNL uses a maximum likelihood estimator to handle situations with several categories, making it a versatile method for various applications. The formulation of this method considers a universal set of alternatives composed of multiple modes in an individual subset, which defines a restricted set of options for a particular decision-maker . The general expression for MNL models is presented in equation 5 (Ben-Akiva & Lerman, 2000): 𝑃(𝑖|𝐶𝑛)=𝑒𝜇𝑉𝑖𝑛 ∑𝑒𝜇𝑉𝑗𝑛 𝑗∈𝐶𝑛 (5) Impacts of Shared E-scooters on Urban Transportation Systems 55 Where:  The utility for the decision-maker in a mode is 𝑈𝑖𝑛=𝑉𝑖𝑛+𝜀𝑖𝑛;  𝑃(𝑖|𝐶𝑛) : the probability of the decision maker choosing a mode of transportation;  The numerator is the utility of mode for the decision-maker , and the denominator is the summation of the utilities for all alternative modes for a decision-maker;  The disturbances  are distributed in independent and identical forms. In this case, the model satisfies the axiom of independence of irrelevant alternatives. Therefore, where any two alternatives have a non-zero probability of being chosen, the ratio of one probability over the other is unaffected by the presence or absence of any additional alternative in the choice set (Luce & Suppes, 1965; Ortúzar & Willumsen, 2008) One of the key advantages of the multinomial logit model is its ability to incorporate continuous and discrete individual heterogeneity, depending on the model specification. This is achieved by allowing the parameters to vary randomly over individuals according to a chosen distribution, which can be constant, discrete, or a mixture of both. For instance, the Biogeme software (Bierlaire, 2023) supports the estimation of various multinomial logit models, including the mixed multinomial logit, the scale heterogeneity multinomial logit, and the generalized multinomial logit. These models can be estimated using either the maximum likelihood estimator or the maximum simulated likelihood estimator, providing flexibility and robustness in handling different types of data (Sarrias & Daziano, 2017). However, the application of multinomial logit models is not without challenges. One significant issue is the interpretation of results. Researchers often face difficulties interpreting the model coefficients due to the independence of irrelevant alternative assumptions. This assumption implies that the relative odds of choosing between any two alternatives are unaffected by the presence of other alternatives. To address this, guidelines and practical steps have been developed to aid in analyzing and interpreting MNL results. These include intuitive graphical representations of predicted probabilities and marginal effects, which are crucial for interpreting and communicating results (Wulff, 2015). Moreover, multinomial logit models are versatile and powerful tools for analyzing choices among multiple alternatives. They can handle continuous and discrete individual heterogeneity, provide intuitive interpretations of results, and be extended to account for scale and coefficient heterogeneity. Despite the Chapter III – Research Methodology 56 challenges in high-dimensional settings, advancements in estimation techniques and variable selection methods continue to enhance the applicability and robustness of these models in various fields. III.4.2.1. Maximum likelihood The maximum likelihood estimation (MLE) method is a cornerstone in applying multinomial logit (MNL) models, which are widely used for modeling categorical response variables. MLE is particularly advantageous because it provides a consistent and efficient means of estimating the parameters of these models, even in complex settings. Equation 6 gives the mathematical representation for the calculation of the maximum likelihood in the Biogeme software: 𝐿(𝛽)=𝑙𝑛𝐿∗(𝛽)=∑𝑙𝑛𝑃(𝑦𝑛|𝑋𝑛,𝛽) 𝑁 𝑛=1 =∑(∑𝑦𝑖𝑛 𝑖𝜖𝐶𝑛𝑙𝑛𝑃(𝑖|𝐶𝑛)) 𝑁 𝑛=1 (6) Where:  𝑦𝑖𝑛=1 if n chose alternative I, 0 otherwise. In Biogeme, the maximum likelihood estimation (MLE) is used to estimate the parameters of discrete choice models. The Log-Likelihood Function measures how well the model with a given set of parameters explains the observed data (Bierlaire, 2023). III.4.2.2. Rho-square To determine the optimal model design, assessing the overall goodness of fit of the estimated model in relation to the collected data is essential. A common approach for evaluating how well a discrete choice model aligns with the data involves using a statistical metric known as the likelihood ratio index, also called rho-square. This index allows the comparison of the performance of the model with estimated parameters against two reference scenarios: one with all null parameters (equivalent to having no model, where all modes share the same distribution) and another with only the specific constants of the alternatives (comparable to the model based on market distributions from the sample used). The rhosquare for the null model is calculated using equation 7, and the rho-square for the initial model is calculated using equation 8: Impacts of Shared E-scooters on Urban Transportation Systems 57 𝜌2=1−ℒ∗ ℒ0 (7) 𝜌2=1−ℒ∗ ℒ𝑖 (8) In economics or stated choice experiments, the variability of the model can be explained by using different rho-squared values. However, as practical modeling guidance, a value of 0.2 can be considered excellent for the models. III.5. Synthesis of main considerations Two survey methods were chosen to gather information on the determinants of shared e-scooter usage and the effects of attributes on the choice of sustainable modes of transportation in Braga. Revealed preference and stated choice surveys have been used in the transportation sector in order to assess people’s willingness to shift modes of transportation, as well as understand how different attributes affect mode choice. These two survey methods were selected for this thesis to complement its information and retrieve the current state of data, as well as the choice for hypothetical scenarios. Firstly, the revealed preference survey was deployed in Braga to collect data on the main motivations and drawbacks of using or not using shared e-scooters in Braga, as well as the perspectives of users and non-users about this mode of transport. This information was used to create two stated choice experiments. The first experiment comprises choice sets with multiple transportation options, which explains the usage of the MNL model to design the utility of the choices presented to the respondents. The second experiment comprises a binary choice between transportation modes, which corroborates the choice of a binary logit model to delineate the choices according to the attributes presented. The choice of the RP and SC methods for this thesis is based on the best practices for surveying users of different modes of transportation in cities to acknowledge the main observed features that define the usage or not usage of a specific transportation option. Since the choice is often dependent of different Chapter IV – Case Study: City of Braga 64 IV.1.3. Urban mobility in Braga IV.1.3.1. Daily trips and modal share In Braga, 113,902 commuting trips are generated daily, 61.5% within the municipality. 78,321 trips are made mainly by working individuals (68.8%), followed by trips made by students, representing 31.8% of the displacements (Instituto Nacional de Estatística, 2022a). The most common modes of transportation used in these trips are cars (70%), followed by walking (16%) and bus (10%). This tendency to use individual vehicles has been growing in Braga compared to data from the 2011 Census, and a decrease in trips by foot and public transportation has also been noticed (Figure 12). Figure 12: Modal share in Braga in 2011 and 2021 The distribution of the commutes within the municipality of Braga shows that the parishes with higher commute rates are located in the central area of the city and are São Vítor (Northeast of Braga), the parish union of Nogueira, Fraião, and Lamaçães (Northeast of Braga), and the parish union of Real, Dume, and Semelhe (Northwest of Braga). These three parishes have a daily commuter share of 16.8%, 8.5%, and 7.6%, respectively. On the other hand, the parishes with a smaller number of daily commuters are Lamas (0.5%) (South of Braga), Espinho (0.5%) (East of Braga), and parish union of Santa Lucrécia de Algeriz and Navarra (0.5%) (North of Braga), located in the periphery of Braga. 010000 20000 30000 40000 50000 60000 70000 80000 90000 Car Walk Bus Bicycle Others 2021 2011 Impacts of Shared E-scooters on Urban Transportation Systems 65 IV.1.3.2. Time spent on daily trips According to data from Instituto Nacional de Estatística (2022a), Braga's average commute time is 17.6 minutes, lower than the national average for 2021 (19.9 min). Table 8 shows the percentage of commuter trips by the time spent in some cities in Braga. Table 8: Commuter trips by duration Town Duration of commute trips Up to 15 min 16 – 30 min 31 – 60 min 61 – 90 min More than 90 min Braga 66,197 34,366 10,848 1,496 995 Most commuter trips in Braga last up to 15 minutes (58.1%), which means that a large portion of those trips would be suitable for active modes of transportation (e.g., walking, bicycle). The change in modal share for short trips would represent a decrease in car usage and a chance for Braga to shift its current massive car usage in town. IV.1.3.3. Presence of active modes of transportation The city of Braga has been working in the last few years to promote walkability in the city by expanding its pedestrian zone (Figure 13). These pedestrian zones were created in 1995, with the pedestrianization of the city center's main streets, including Republic Square and Souto Street (Silva, 2017). It has more than 20,000 m² of area exclusively for pedestrian use (Souza & Dias, 2020). Figure 14 and Figure 15 show the changes in the urban built environment since the introduction of pedestrian zones in Braga, which shows that the city of Braga, throughout time, has been working to promote better and greener ways of transportation for its population. This is also reflected in better usage of active modes in Braga compared to other municipalities, such as Porto, Coimbra, Aveiro, and Faro (MPT, 2018a). Chapter IV – Case Study: City of Braga 66 Figure 13: Pedestrian zone in Braga (own author) Figure 14: Republic Square in Braga in 1940 (left) and 2018 (right) Figure 15: Liberty Avenue in Braga in 2000 (left) and 2018 (right) Impacts of Shared E-scooters on Urban Transportation Systems 67 However, it is also important to note that the suitable conditions for pedestrian use in Braga are only constant in some of its territory. In Braga’s city center, for example, the pavements, their color, and the continuity of the walkways are well maintained. At the same time, the connected pathways suffer from discontinuation of the width and length of walkways (Rahaman et al., 2010). Studies from MPT (2018a) show that in other areas of the city, such as the city’s traditional consolidated area and areas of linear shaping, pedestrian mobility is classified as low because it shows a lack of elements and needs improvements in the quality of the built environment. Bicycle infrastructure in Braga is also a recurring theme among urban mobility discussions. According to Meireles & Ribeiro (2020), the local government has been assuming the need to offer citizens adequate infrastructure to enable active mobility to achieve the sustainable mobility goals from the National Strategy for Active Mobility, which culminated in the creation of 22km of cycling routes along with parking facilities to promote multimodality. The bicycle network in Braga is presented in different ways. There is an infrastructure shared with pedestrians along with the extension of the Este River, called eco via. Cycle paths are present in distinct areas of the city, and there are signed shared roads. Figure 16 shows the current cycle network in the city of Braga. Figure 16: Bicycle infrastructure network (MPT, 2019) Chapter IV – Case Study: City of Braga 68 The Municipality of Braga states in its Municipal Master Plan that the city needs to reach a modal share of 10% for bicycle use by improving cyclists' safety, developing and keeping the cycle network connected and attractive, providing parking facilities for bicycles, and identifying partners to create educational plans for bicycle use (Município de Braga, 2015). Yet another measure that promotes the increase in safety for the most vulnerable in Braga's urban mobility (bicyclists and pedestrians) is the creation of 30km/h zones. According to data from Portugal 2020 (2019), the constraint on speed must be implemented in residential areas with high economic activity or in the surrounding areas of schools. The project includes the creation of crosswalks with tactile floors, traffic calming measures, installations of access ramps, and widening sidewalks. Figure 17 shows the location of Braga's existing and future 30km/h zones. Figure 17: Areas where 30km/h zones will be implemented (MPT, 2019) In addition to walking and cycling in Braga, the city has also worked to establish a shared e-scooter service as a new mode of transportation to help ease traffic congestion, pollution, and noise problems. The Municipality holds the shared e-scooters as a promise to improve mobility in Braga, even if some constraints are reported in this mode of transport. Impacts of Shared E-scooters on Urban Transportation Systems 69 IV.1.4. Shared e-scooters in Braga The e-scooter-sharing service started operating in Braga on August 19th, 2019, with only one company providing the service. This company started its operation with about 80 e-scooters distributed around the city center, with expected expansion in the short term. Later, two other companies began to provide the same service in the following months of the same year. However, only two companies remain offering the service in the city and currently make available 379 shared e-scooters for the population. The e-scooters were available in 25 different parking spaces allocated within the city center area (City of Braga, 2019) at the beginning. Still, currently, there are more than 50 parking spaces available near malls and educational institutions to stimulate the usage of the service by students (Figure 18). These parking spots were converted from car parking spaces to accommodate bicycles and e-scooters (Figure 19). Figure 18: Location of shared e-scooter parking spots (own author) Chapter IV – Case Study: City of Braga 70 Figure 19: Micromobility parking spots (own author) Although riders don't need to park their shared e-scooters in the parking spots, decreasing the number of e-scooters parked on sidewalks blocking the walkways is a great asset. Besides the parking spots, the City of Braga also created a “red zone” –seen in the companies’ app - where the maximum speed of the e-scooter must be reduced, and the wheels are blocked when the rider reaches a “no-circulation zone”. These areas were created to increase the safety of the e-scooter users and the pedestrians circulating there. This “no-circulation” zone covers some critical streets of the city, such as Dom Diogo de Sousa Street, Souto Street, Dom Paio Mendes Street, Dom Gonçalo Pereira Street, Rossio da Sé, Misericórdia Street, Eça de Queirós Street, Dr. Justino Cuz street, Francisco Sanches Street, S. Marcos Street, Liberdade Avenue, Dr. Gonçalo Sampaio Street. The map in Figure 20 covers the city of Braga with the “red zones” printed in “red”, and it is essential to point out that besides the streets, a large part of the city center of Braga is covered with these constraints. Impacts of Shared E-scooters on Urban Transportation Systems 71 Figure 20: Red zone for shared e-scooter usage (own author) Moreover, some rules were established by the Municipality (City of Braga, 2019) to be followed by the companies that provide the service once it can help increase the users' safety. The rules are as follows:  The company needs to make e-scooters available in the designated locations (dedicated parking sports) under the authorizations and monitoring of the Municipality of Braga;  Safeguard that the equipment is maintained and safe for users;  Provide users with information about operations, road safety, and usage rules. At registration, all information on good practices for use and circulation on public roads and parking rules will be given;  Ensure that the equipment does not pose a danger to other road users. The company must remove or relocate shared e-scooters parked in inappropriate places on its initiative and whenever requested;  Inform users about authorized stopping places;  Distribute e-scooters in all locations and regularly;  Apply mandatory checkout mechanisms in the app when riders reach the “red zones”, where vulnerable users recognize more sensitivity. Chapter IV – Case Study: City of Braga 72 IV.1.4.1. Pre-pandemic usage of shared e-scooters in Braga The shared e-scooter service was offered in Braga before the COVID-19 pandemic started. Data from August until December 2019 shows that, in this period, 174 e-scooters were operating in the city by the end of October. These e-scooters made an average of 4,000 trips per month, and according to the data retrieved, from August to October, e-scooters made more daily trips. In November and December of the same year, the average number of trips per e-scooter decreased to 0.8, which can be justified as in these months, the temperatures decrease, the days are shorter, and it starts raining more often in Braga. In addition, the weekends and hours between 5 p.m. and 7 p.m. were when e-scooters were more solicited, representing that e-scooters were more used for leisure activities. Table 9 shows the metrics for shared e-scooter trips in the second half of 2019 in Braga. Table 9: E-scooter usage in Braga in the second half of 2029 Metric Month of 2019 August September October November December Number of e-scooters 46 110 174 103 98 Number of trips 1746 7,159 6,843 2,439 2,549 Number of trips per e-scooter 1.26 2.17 1.31 0.8 0.8 Number of trips (average/day) 58 239 228 81 82 Number of km traveled 4,624 14,924 11,770 4,004 4,076 Most solicited weekday Sunday Sunday Saturday Saturday Tuesday Most solicited hour of the day 6 p.m. 5 p.m. 7 p.m. 6 p.m. 5 p.m. At the end of the five months after the implementation of the service, shared e-scooters have made more than 20,000 trips and traveled more than 40,000 km. According to data from the company, more than 95% of the users in this period were locals, and they took trips mainly during the evening, between 5 p.m. and 6 p.m. The geographic sprawl of trips changed during these months. When the service started operations in October, people would start and finish their trips in the central area of Braga (i.e., the historic center), while in the last months of the year, people would take further trips to different areas of the city, namely near Braga Parque mall and the University of Minho campus. Moreover, the only pre-pandemic data available for Braga comprises the second half of 2019, and there is no consistent data from the beginning of 2020. Impacts of Shared E-scooters on Urban Transportation Systems 73 IV.1.4.2. Impacts of the COVID-19 pandemic on the shared e-scooters in Braga With the spread of COVID-19 in Europe, Portugal started a lockdown on 18 March 2020. The government took several measures, including allowing residents to circulate on the streets only for a few reasons, including grocery shopping, health care, assisting vulnerable people, and exercising (Republica Portuguesa, 2020). In Braga, with the state of lockdown, the shared e-scooter service was suspended until people could walk freely on the streets, which happened at the beginning of 2021. Since then, three different companies have resumed their services to the population. Starting in April 2021, the city offered about 500 shared e-scooters. For the service to be resumed, some measures were taken to try to reduce the negative impact of this mode of transportation in the city, such as the control of the speed, as e-scooters cannot run at speeds higher than 25 km/h in order not to disturb and to harm people’s accessibility to public spaces. In addition, e-scooters must be allocated in specific parking spots distributed within the city, where companies must ensure their correct parking (Agência Lusa, 2021). Besides, companies started offering price reductions to increase the use of this mode of transportation in the city. One of the companies provided the service with a free e-scooter unlock and only EUR 0.05 per minute traveled. While the other started offering 15-minute trips for only EUR 1.0 and some trip packs to be used for one month or 90 days with a price reduction. These measures increased the usage of shared e-scooters in Braga in the following months. The companies that deployed their e-scooters on the streets of Braga saw the success of this mode of transportation, as reported in the increased number of trips presented in the following section. IV.1.4.3. The current shared e-scooter service in Braga Currently, the two companies that offer the service in Braga have applications that allow users to unlock and ride their e-scooters throughout the city. The apps also provide the location of the specific parking spots for e-scooters and the exact location of the e-scooters for rent. Figure 21 shows the parishes where shared e-scooters are available. Chapter IV – Case Study: City of Braga 80 Figure 26: Number of trips in shared e-bikes Figure 27: Distance traveled and trip duration of shared e-bikes Shared e-bike riders spent an average of 10 minutes traveling an average of 1.0 km per trip in Braga. Figure 25 shows that the highest duration of trips occurred in June 2023 since this month also presented the highest distance traveled. In the other months, the duration of the trips remained almost constant, with a considerable drop in March 2023, when the average duration of trips was 7 minutes. It is essential to mention that the distance traveled, the duration of trips, and even the number of trips show a slight increase in the months when the number of shared e-bikes increases. 0 100 200 300 400 500 600 700 Sep-22 Oct-22 Nov-22 Dec-22 Jan-23 Feb-23 Mar-23 Apr-23 May-23 Jun-23 Number of trips 0 2 4 6 8 10 12 14 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Sep-22 Oct-22 Nov-22 Dec-22 Jan-23 Feb-23 Mar-23 Apr-23 May-23 Jun-23 Average trip duration (minutes) Average distance traveled (meters) distance_origin Duration Impacts of Shared E-scooters on Urban Transportation Systems 81 Following the same pattern as shared e-scooters, the trips made by shared e-bikes had similar origins and destinations. Most trips started and ended in São Victor parish, the city's most populated and dense area. Shared e-bikes were most solicited on Fridays and the weekend. IV.1.6. Interaction between shared e-scooters and e-bikes To check the interaction between shared e-scooters and shared e-bikes, 273 days from September 2022 to May 2023 are evaluated since both services ran in the city this period. Within this period, 185,355 trips were reported by shared e-scooters and shared e-bikes in the city. The highest number of trips were made by e-scooters (182,850), while only 2,505 trips were made by e-bikes. The number of trips shows an increase in e-scooter and e-bike rides during the months of higher solar exposure and less rainfall, which are September, October, April, and May. However, during the colder and rainy months of November until mid-March, these modes of transportation face a decline in the number of rides. Even if the usage pattern throughout this timeframe seems similar for e-bikes and e-scooters, the number of trips made by each mode is different. E-scooters represent over 97% of shared micromobility trips during the ten months. These patterns can represent that e-scooters have more acceptance among users of shared micromobility in Braga or that residents are more used to this mode once it has been available to the population since 2019. Another difference in e-scooter and e-bike usage in Braga is related to the characteristics of the trips regarding the distance traveled by each mode and the duration of these trips. Table 11 shows the descriptive statistics for the number of trips made by shared e-scooters and shared e-bikes during the ten months. The data used to compose the table comprises all trips generated during this period. Table 11: Descriptive statistics of shared e-bike and e-scooter trips Shared e-bike (N=2,505) Shared e-scooter (N=182,850) Distance traveled (m) Trip duration (min) Distance traveled (m) Trip duration (min) Mean 972.5 9.8 1508.1 8.6 Std. Dev 502.9 4.1 94.4 1.0 Min. 118.5 1.7 1302.9 6.7 Max. 3269.83 25.7 1822.1 11.6 Chapter IV – Case Study: City of Braga 82 The average distance traveled by e-scooters is higher than the ones made by e-bikes. However, e-bikes take longer to reach their destination. Even though shared e-scooters are used more often than e-bikes, the trips made by e-bikes usually reach further distances in longer journeys. Figure 28 shows the percentage of trips made per day of the week for shared e-scooters and shared e-bikes. These trips are made mainly on the weekends, with the most used days being Friday, Saturday, and Sunday, with a slight peak also on Tuesday. The e-scooter usage presents an almost even utilization rate during the week, reaching the peak of trip numbers on Friday. Figure 28: Percentage of shared e-bike and e-scooter trips during the week When comparing the main areas of usage for shared micromobility in the city, it is possible to identify that e-bikes and e-scooters are used relatively in the same regions of the city, with particular attention to the darker colors in Figure 29, which represent the most populated parishes that comprise the most significant number of services for the population, such as the shopping mall, school districts, the University of Minho campus and city facilities. The darkest color represents São Victor parish, the most populated and dense in the city. 0% 5% 10% 15% 20% E-bike usage during the week Mon Tue Wed Thu Fri Sat Sun 0% 5% 10% 15% 20% E-scooter usage during the week Mon Tue Wed Thu Fri Sat Sun Impacts of Shared E-scooters on Urban Transportation Systems 83 Figure 29: Shared e-bikes and e-scooters main area of usage (own author) In Braga, the implementation of shared e-scooters occurred before the implementation of the shared ebike service, which is not common, since bicycle sharing services have been around for longer. The early rise of the former contributed to familiarizing the general population with this mode. The trip data shows an excessive disadvantage of e-bikes when considering the number of trips studied. This can be due to the early familiarization of e-scooters. Another reason that can cause an advantage of e-scooter usage over e-bikes is the trip's price. While e-scooters can reach the lower cost of only EUR 0.21 per minute traveled, e-bikes used to cost EUR 1.0 to be unlocked, plus EUR 0.18 per minute traveled. This also confirms the role of deploying mode choice experiments in the city. The need for dedicated infrastructure can be another crucial factor in determining the lack of competitiveness between e-bikes and e-scooters. This is because in the area of the city where most micromobility trips are made, cycle lanes are almost non-existent, and the streets are narrow and paved with cobblestones. Therefore, e-scooters would have priority usage since they are easier to ride on the sidewalk, even though it is legally forbidden in Portugal. Chapter IV – Case Study: City of Braga 84 In Braga, both modes are present and used in the same area of the city, representing that they are accessed in the same locations. When comparing the usage throughout the week, patterns remain similar since both micromobility modes present higher ridership rates on Fridays. Furthermore, shared e-scooters and e-bikes in Braga can coexist and provide mobility options for shortdistance trips. In Braga, to balance the usage of both micromobility modes, more efforts needed to be made to minimize price differences among them and to provide equal opportunities for people to ride safely (i.e., provision of dedicated infrastructure) so the population could be familiarized with e-scooters as well as e-bikes and use them according to their mobility needs. Impacts of Shared E-scooters on Urban Transportation Systems 85 IV.2. Revealed preference (RP) survey This sub-section displays the objectives of the administered survey, the methodology used, and the data collection. In addition, the results are presented in two blocks: 1) results from the survey administered with the population sample and 2) results from the survey administered with the university students sample. IV.2.1. Objectives of the RP survey The main objective of the Revealed Preference survey deployed in Braga is to identify the main attributes (i.e., motivations and barriers) for shared e-scooters in Braga that can be used to create the SP experiment. Also, the RP survey is used to identify the users’ and non-users perceptions of shared escooters, the profile of the users and potential users of the service in Braga, the main determinants of shared e-scooter usage in Braga, and the possible planning strategy that needs to be taken into consideration to promote a safe usage of this mode of transport. Therefore, the results of the RP survey are closely related to the following research questions:  RQ2: What are the main drivers/determinants of shared e-scooter usage for commute and leisure trips?  RQ3: How do transportation planning and traffic management influence the safe usage of shared e-scooters?  RQ4: What are the impacts of shared e-scooter usage on the sustainability of the transportation system? IV.2.2. Methodology of the RP survey The RP survey was developed to extract information from users and non-users of Braga's shared e-scooter service. The online Google Forms survey (Appendix A) was disseminated from 24 January to 10 July 2022 in downtown Braga, in the vicinity of shared e-scooter stations that comprise the main traffic generators in the city, near school districts, as well as the University of Minho campus (i.e., students and staff). The survey administration was divided into two approaches: collecting data from a sample from the general population of the city and a sample from the university student body. The choice to collect data specifically for the latter group is due to previous research showing that shared e-scooters are used mainly by younger Chapter IV – Case Study: City of Braga 86 people (Y. Guo & Zhang, 2021b; McKenzie, 2019), and heavy e-scooter traffic is expected to occur in downtown and university campus areas (Caspi et al., 2020; Huo et al., 2021). As people were still highly concerned about the contamination from COVID-19 in the first half of 2022, the survey was disseminated with focus on the two needed samples. The survey was disseminated through e-mail, institutional e-mail and on campus through a QR code for the university students. For the general population of Braga, the survey QR code near the city center was propagated near shopping malls and school districts targeting students and parents, so people could respond to the survey on their cellphones and avoid close contact with the researcher (Figure 30). Figure 30: QR Code flyer used to disseminate the RP survey A stratified random sampling method was employed to target current users of shared e-scooters and potential service users. The survey incorporated multiple question types, including single-choice, multiplechoice, and Likert-type scales. These questions explored vital aspects, such as the main factors not to use shared e-scooters in Braga (for non-users), the main motivations to use this mode of transportation (for current users), and what they would like to see improved in the provision of the service to increase its usage. Impacts of Shared E-scooters on Urban Transportation Systems 87 IV.2.2.1. Structure of the RP survey and sampling The same survey structure was used to collect data from the general and university student populations. Firstly, the respondents had to answer whether or not they use the shared e-scooter service in Braga. Then, if the answer was negative (potential users of shared e-scooters in the future), the respondents were forwarded to a selection of questions regarding the main factors for them not to use the service and if they were willing to start using shared e-scooters if the current perceived negative factors were solved. If the respondents use the service, they were asked questions regarding their primary motivations for using this mode of transportation and how they use the micro vehicle and the service across the municipality (e.g., origin-destination, frequency). In the end, all respondents were asked sociodemographic questions. Thus, to better identify the profile of the population and their point of view on the shared e-scooter service, the structure of the survey is as shown in Figure 31. Figure 31: Structure of the RP survey For the general population sample, the data collection resulted in 541 answers for the survey, but 108 had to be discarded due to inconsistent answers (e.g., surveys that were left partially blank or when single- Chapter IV – Case Study: City of Braga 88 answer questions received more than one answer), which resulted in 433 (N = 433) valid answers to estimate the results. For the university student sample, 376 valid answers were collected. The sample size for both respondent categories (i.e., general population and university students) was calculated using valid approaches in designing and conducting survey research (Ayaz et al., 2021; Cochran, 1977) (equation 8). To statistically representative of the population of Braga at a 95% confidence level (margin of error of 5%), a population of 193,324 people who live in the city (Instituto Nacional de Estatística, 2022b) was considered, a z-score of 1.96 (Kelley, 2007; Krejcie & Morgan, 1970), and 40% male population. It is essential to mention that from the 433 valid answers, 78 represent current users of shared e-scooters in Braga. In contrast, 355 represent potential users (i.e., people who do not use shared e-scooters but could start using the service if some improvements were made). To statistically represent the university student population at the University of Minho at a 95% confidence level and 5% margin of error, a population of 15,350 students in the University of Minho campus in Braga was used, considering a z-score of 1.96, and 40% male population. 𝑆𝑎𝑚𝑝𝑙𝑒 𝑠𝑖𝑧𝑒= ( 𝑧2∗𝑝(1−𝑝) 𝑒2 1+(𝑧2∗𝑝(1−𝑝) 𝑒2𝑁)) (8) Where:  e – desired level of precision, the margin of error;  p – the fraction of the population (as percentage);  z – the z-score. IV.2.2.2. Statistical treatment of results To determine whether the use or non-use of shared e-scooters in Braga and the sociodemographic characteristics of the sample are independent (unrelated), Pearson’s chi-square tests (equation 9) were performed to correlate the data for the general population and university student sample. This test is used because the collected data is categorical, meaning it is divided into categories and does not fall within a parametric or continuous data group. Impacts of Shared E-scooters on Urban Transportation Systems 89 𝑋2=∑(𝑂−𝐸)2 𝐸 (9) Where:  X2 – is the chi-square test statistic;   is the summation operator;  O is the observed frequency;  E is the expected frequency. In addition, since university students constitute a large portion of shared e-scooter users (Nikiforiadis et al., 2023a), and the younger population is more open to shifting modes of transportation due to their sustainability features, only the university student sample data was used to assess the correlation between the perception of sustainability of shared e-scooters and their sociodemographic characteristics. The Mann-Whitney test (equation 10) and the Kruskal-Wallis H test (equation 11) were performed (Pestana & Gageiro, 2014). 𝑈=𝑛1𝑛2+𝑛1(𝑛1+1) 2−𝑅1 (10) Where:  n1 – sample size one;  n2 – sample size two;  R1 – rank of the sample size. 𝐻= 12 𝑛(𝑛+1)∑𝑅𝑗2 𝑛𝑗 𝑘𝑗−3(𝑛+1) (11) Where:  n – total number of observations;  k – number of groups that are being compared;  Rj – is the sum of ranks for group j;  nj – number of observations in each group. Chapter IV – Case Study: City of Braga 96 Figure 35: Intervention for users to continue using shared e-scooters As it was expected, the offer of riding lessons was the least chosen option by shared e-scooter users. However, even if they are familiar with riding shared e-scooters, the provision of safety infrastructures is still a concern since the most chosen options are related to the increase in the protection and safety of an e-scooter rider (e.g., cycle lanes, dedicated parking, zoning). Shared e-scooter users would also appreciate the availability of this mode of transportation where public transportation is scarce. In Braga, the bus service (the only public transportation option available) is reduced in some parishes distant from downtown. This could contribute to the spread of other sustainable transportation options for the population. Users also appreciate the creation of low-speed zones. However, there is a distinction between implementing a 30km/h zone and a 20km/h zone. In most cases, users prefer that the speed be reduced to 30km/h, and this can be due to their already experience riding an e-scooter and consequently having more confidence in the traffic. IV.2.3.6. The main factors for not using shared e-scooters The non-users of shared e-scooters could also report the main reasons for not using them in Braga. In a multiple-choice question, they could choose among infrastructure and weather-related options. Yet, the 62,80% 46,20% 43,60% 32,10% 30,80% 17,90% 17,90% 14,10% 12,80% 12,80% 0,00% 10,00% 20,00% 30,00% 40,00% 50,00% 60,00% 70,00% Cycle lanes Dedicated parking Zoning Availability where PT is scarse 30km/h zones Integration with PT Regulation Diverse payment options 20km/h zones Riding classes Impacts of Shared E-scooters on Urban Transportation Systems 97 most chosen factor for not using shared e-scooters was the lack of safety (34.6%) and the preference for using other modes of transport, such as the car (33.2%) and the bus (29.3%). Figure 36 shows the choices for the factors for not using shared e-scooters. Figure 36: Main factors for not using shared e-scooters Another critical issue that needs to be addressed regarding the factor that negatively affects the usage of shared e-scooters in Braga is that people need to learn how to use the service or ride an e-scooter. In a car-centric country, such as Portugal, where people have access to learn how to drive a car when they are eighteen years old, but they do not have the same opportunity to learn how to ride a micro vehicle, such as a bicycle and an e-scooter, it is improbable that they even consider these modes as a valid transportation option for their daily trips. Therefore, this could also cause a low preference for bicycle use (11.0%). The current pavement conditions in Braga also disrupt the usage of shared e-scooters. The pavement's poor condition increases the trip's discomfort since it is a vehicle with little or sometimes no suspension system that allows a smooth ride. In addition, the high prices of the service can be a drawback from the usage since the usual cost of EUR 1.0 to unlock the vehicle plus EUR 0.15 per kilometer traveled can make the trip more expensive than a single bus ticket (EUR 1.55 in Braga) or even a ride healing (e.g., Uber, Bolt). 34,60% 33,20% 29,30% 29,30% 23,70% 23,70% 21,70% 17,20% 15,20% 11,30% 11,00% 8,70% 5,90% 0,00% 5,00% 10,00% 15,00% 20,00% 25,00% 30,00% 35,00% 40,00% Lack of safety Prefer to use the car Prefer to use PT I do not know how to use it Bad pavement conditions High price Bad weather conditions I prefer to walk High speed limit Lack of dedicated parking I prefer to use a bicycle Small geofence area Uncomfortable Chapter IV – Case Study: City of Braga 98 Another aspect highly considered by non-users is the weather conditions. Braga is known for experiencing heavy rain during late fall and winter, which contributes to the unwillingness to use a mode of transportation that provides no weather protection for the rider. The city of Braga experiences more than 114 days of rain during the year (p>1mm) and more than 48 days during the year with precipitation of more than 10mm (IPMA, 2010). Thus, usage of shared e-scooters during rainy months can be decreased. It is also reported that women suffer more than men due to the lack of road safety in riding this mode of transport. In addition, with the advancement of the age of respondents, one feels more unsafe being on the road on a shared e-scooter. Despite this micro vehicle, the preference for a car is also more prominent for women than men. In addition, the increase in family income is proportional to the preference for using a car instead of a shared e-scooter. The contrary is found with the preference for public transportation over e-scooters since the preference for PT decreases as the family income increases. Women are also more affected by the lack of knowledge about how to use a shared e-scooter than men, and as the educational level of the respondents increases, the knowledge of how to ride e-scooters increases. Price affects usage on all academic levels almost equally, affecting more men than women. IV.2.3.7. The primary interventions to start using shared e-scooter In addition to the main factor preventing shared e-scooter usage, respondents were also asked about the interventions they find relevant to allow shared e-scooter usage. The primary choice is related to providing infrastructure that allows a safe ride, which is the implementation of cycle lanes since shared e-scooters would have a dedicated channel to be ridden on. This would reduce the interaction with cars and pedestrians when they are erroneously ridden on sidewalks. Figure 37 shows the selected interventions. Impacts of Shared E-scooters on Urban Transportation Systems 99 Figure 37: Main interventions for non-users to start using shared e-scooters Non-users of shared e-scooters are also seeking to improve safety in their trips. The three most selected options were the implementation of cycle lanes (54.9%), provision of zoning (49.6%), and dedicated parking (34.1%). This aligns with the answers provided by service users, as well. While the availability of diverse payment options is an intervention needed for non-users, this was one of the least chosen options by users of the service. Riding lessons is another contrast between users and non-users; while this was the least chosen option by users, riding lessons still play an essential role for non-users. This situation can be induced by the fact that one of the main reasons people do not use shared e-scooters is that they do not know how to ride them. The integration of shared e-scooters with public transportation has almost the same percentage of choice between users (17.9%) and non-users (20.8%), indicating that the two groups value almost the same degree the possibility of using shared e-scooters in the first/last mile traveled, which could increase the catchment area of the bus stops. IV.2.3.8. Main determinants of shared e-scooter usage After analyzing the primary motivations for using shared e-scooters, their main drawbacks, and the sample profile, it was possible to audit the main determinants for using this service. The determinants are related 54,90% 49,60% 34,10% 33,80% 29,00% 22,30% 21,40% 20,80% 16,10% 10,10% 0,00% 10,00% 20,00% 30,00% 40,00% 50,00% 60,00% Cycle lanes Zoning Dedicated parking Diverse payment options Regulation Availability where PT is scarse Riding lessons Integration with PT 30km/h zones 20km/h zones Chapter IV – Case Study: City of Braga 100 to the sociodemographic characteristics of the population, such as gender, age range, educational stage, employment status, family size, family monthly income, parish of origin of most of the trips, parish of destination of most of the trips, and the mode of transportation used daily. The results from the statistical correlation between the use or not use of shared e-scooters (yes or no) and the sociodemographic characteristics can be seen in Table 13. Table 13: Pearson’s chi-square for the determinants of shared e-scooter usage Utilization determinant Use of shared e-scooters (yes or no) Pearson chisquare Df Fisher’s exact test  Gender 15.698 1 <0.001 <0.001 Age 7.367 9 0.751 0.602 Educational stage 1.401 3 0.702 0.707 Employment status 1.725 3 0.722 0.668 Family size 2.113 4 0.677 0.718 Family monthly income 3.353 8 0.919 0.919 Parish of origin 26.020 1 <0.001 <0.001 Parish of destination 1.581 1 0.271 0.271 Mode of transportation 37.979 3 <0.001 <0.001 There is evidence of a significant statistical association between the utilization of shared e-scooters and the respondents' gender. This is because there is a lack of women's participation in shared e-scooter usage in Braga (critical z = -4.0), while men are overrepresented (critical z = 4.0), meaning that the adjusted residuals for women are lower than expected and higher than expected for men. Thus, gender is likely to influence the willingness to use the service because it is reported that women suffer more than men due to the lack of road safety when riding this mode of transport. The parish of origin of the trips also represents a significant statistical association with the use or not use of shared e-scooters. Thus, if a person lives in the central area of the city, where shared e-scooter services are available, the probability of using this mode of transportation for daily trips is higher than people who live in peripheral areas. The critical z for this correlation is -5.1, which is related to the uneven distribution of the e-scooter service in Braga since this mode of transportation is only available downtown and in its immediate surroundings. Respondents' primary mode of transportation also influences their use or non-use of shared e-scooters. People tend not to use the service if they have access to a car (critical z = 2.5, meaning that the adjusted Impacts of Shared E-scooters on Urban Transportation Systems 101 residual for this mode of transportation is higher than expected for non-users of shared e-scooters) or public transportation (critical z = 3.9, meaning that the adjusted residual for this mode of transportation is higher than expected for non-users of shared e-scooters). On the other hand, people tend to use shared e-scooters if they need to walk (critical z = 4.7, meaning that the adjusted residual for this mode of transportation is higher than expected for users of shared e-scooters). However, shared e-scooters can act as a mode of transportation that attends to the needs of people from multiple family sizes, incomes, employment statuses, and educational stages since there is no evidence of a significant statistical association between using or not using e-scooters and these determinants. Respondents tend to use shared e-scooters regardless of their family’s monthly income. This can be influenced by the fact that most respondents are students who depend on their parents’ incomes to pay for transportation options. Also, most respondents earn more than EUR 1,501, which is higher than the average income in Braga, which is EUR 1,146 (Instituto Nacional de Estatística, 2022a). In addition, the employment status of the population shows no significant statistical association with the use or not use of shared e-scooters in Braga, which shows that students, employees, unemployed people, and self-employed people have the same willingness to use this mode of transport, which is also in line with the work of Mitra & Hess (2021). IV.2.3.9. Representativeness of different sociodemographic groups in shared e-scooter usage This section correlates the sociodemographic profiles of users and potential users (i.e., non-users who would start using shared e-scooters if the service were improved) with the sociodemographic profile of the population in Braga (Instituto Nacional de Estatística, 2022b). For this, the chi-square goodness of fit is used to determine whether the sociodemographic characteristics of users and potential users are similar to the sociodemographic characteristics of the population of Braga. The aim is to assess whether the general population is represented in the service’s usage. Among current users (N = 78), statistical observations show that shared e-scooters do not provide equity services for the diverse population of Braga (Table 14). Regarding gender differences, women are underrepresented compared to men since the data distribution from shared e-scooter users is not consistent with the distribution of the data from the 2021 census. In this case, a higher representation of Chapter IV – Case Study: City of Braga 102 women was expected among service users, as the number of observations from the survey for women (N = 23) was lower than expected to represent the population of Braga (N = 39.8). Table 14: Chi-square goodness of fit test for shared e-scooter users (N = 78) Sociodemographic characteristics Chi-square Goodness of Fit Df  Gender 14.932 1 <0.05 Age 98.288 3 <0.05 Educational level 18.016 2 <0.05 Employment status 122.855 1 <0.05 The correlation of education level shows that people with higher education are overrepresented, while people with lower educational levels are underrepresented. In this case, the number of observations of respondents who completed their undergraduate studies (N = 27) was higher than expected to represent the population of Braga (N = 14). In comparison, the number of observations for people who have completed their high school degree (N = 42) is lower than the number of observations expected for the population of Braga (N = 58). The professional status of the population is also another aspect of inequity among shared e-scooter users. Students are the most prominent group of people served (observed N = 58, expected N = 17.4), while unemployed people (i.e., people with reduced economic means) are left behind. When potential users are added to the total sample in Braga (N = 433), the representation of the general population in the usage of shared e-scooters somewhat changes, but this mode of transportation is still deficient in enabling equitable service across the population (Table 15). Interestingly, if potential users start riding shared e-scooters in Braga, gender differences will decrease as more women begin riding this micro vehicle. Considering the improvement in gender representativeness, men and women would have shares of almost 50% in usage, as the number of observations for men is N = 207, and the expected number of observations to represent the population of Braga would be N = 200.4. The number of observations for women is N = 214, and the expected number of observations for this gender in Braga would be N = 220.4. Table 15: Chi-square goodness of fit test for shared e-scooter users and potential users (N = 433) Sociodemographic characteristics Chi-square Goodness of Fit Df  Gender 0.415 1 0.519 Impacts of Shared E-scooters on Urban Transportation Systems 103 Sociodemographic characteristics Chi-square Goodness of Fit Df  Age 431.799 3 <0.05 Educational level 70.225 2 <0.05 Employment status 639.346 3 <0.05 However, shared e-scooters would continue providing an inequitable service across all other socioeconomic groups surveyed. Regarding the age of users, people from 20 to 34 would still be overrepresented since the number of observations for this group is N = 228, and the expected number of observations in Braga would be N = 77.6. People 35 years old and above are underrepresented because the expected number of observations for Braga would be N = 100.1, and the actual number of observations was N = 54. In this scenario, users' educational levels continue to over represent people with high academic levels, mainly people who graduated from the university and pursued a master’s or doctoral degree. Respondents who concluded their undergraduate studies have a number of observations of N = 130, and people with master’s or doctoral degrees have a number of observations of N = 61, while the expected number of observations for Braga would be N = 76.8 and N = 37.4, respectively. Shared e-scooters still underserve people at an economic disadvantage or unemployment, while students are the most prominent users. In this case, the observed number of unemployed people is N = 6, and the expected number of observations for the population of Braga would be N = 18.2. It is essential to mention that the increase in potential users is due to some expected changes in providing shared e-scooters in Braga. Nevertheless, more efforts need to be made to increase equity issues through other measures, such as the ones mentioned by respondents of the survey, which includes the implementation of more dedicated infrastructure (e.g., cycle lanes), which could influence road safety perception, as well as the creation of specific zones where shared e-scooters, and micromobility in general, have priority over cars, and implement other payment methods other than credit cards. IV.2.4. Results from the RP survey for the university student sample To better understand how young users perceive shared e-scooter services, the same survey was deployed in Braga in the spring of 2022 among university students. The survey's main objectives are to assess the Chapter IV – Case Study: City of Braga 104 perception of sustainability of shared e-scooter services among university students and how regulation could affect this type of user's usage of this service. IV.2.4.1. Profile of the respondents The total number of answers retrieved from the survey is 376, and the prevailing age range of the respondents is from 20 to 24 years of age. Of all enquired students, 58% are female, 39% are male, and 3% preferred not to say their gender. Their average family monthly income ranges between EUR 1,000 to EUR 1,500, corresponding to the average monthly income of families in Portugal (Instituto Nacional de Estatística, 2022a). The respondents' main modes of transportation are private cars (36%), public transportation (35%), and walking (27%). The profile of the respondents and the student body of the University of Minho in Braga can be seen in Table 16. Table 16: Sample description for university students Category Subcategory N Sample % Sample % University Gender Female 218 57.8% 60.1% Male 146 38.7% 39.9% Other 2 0.5% - Prefer not to say 11 2.9% - Age Up to 19 years old 148 39% 26% 20-24 years old 181 48% 45% 25-29 years old 21 5.6% 12% 30-34 years old 12 3.2% 5% 35-39 years old 6 1.6% 4% 40-44 years old 4 1.1% 3% 45-49 years old 1 0.1% 2% 50-54 years old 3 0.8% 1% 55-59 years old 0 0% 1% Above 60 years old 1 0.3% 0% Household monthly income Up to EUR 665 17 4.5% - EUR 666 – EUR 1000 38 10.1% - EUR 1001 – EUR 1500 66 17.5% - EUR 1501 – EUR 2000 36 9.5% - EUR 2001 – EUR 2500 28 7.4% - EUR 2501 – EUR 3000 15 4.0% - EUR 3001 – EUR 3500 7 1.9% - EUR 3501 – EUR 4000 3 0.8% - Above 4000 6 1.6% - The main mode of transportation Walking 103 27.3% - Private car 135 35.8% - Public transport 130 34.5% - Impacts of Shared E-scooters on Urban Transportation Systems 105 Category Subcategory N Sample % Sample % University The main mode of transportation Micromobility (bicycle or escooter) 4 1.1% - The sample of 376 students statistically represents the wider University population (N = 15,350). However, data on household income and the main mode of transportation used by students were not available. Moreover, the sample used in this research work is aligned to represent the gender and age distribution of the University's student body evenly. IV.2.4.2. University students’ perception on shared e-scooter usage and sustainability Of all survey respondents in Braga, 86% do not use the service, while only 14% use shared e-scooters regularly. The main reasons for using or not using shared e-scooters in Braga (Figure 38) show that students prefer to displace in other modes of transport, such as public transportation and private cars. Also, the need for knowledge on using a shared e-scooter (digital unlock system, online payment option, how to ride the vehicle) and insecurity on the road lead young students to avoid this shared micro vehicle. Figure 38: Main reason for university students to not use shared e-scooters Moreover, Pearson’s chi-square tests were performed to acknowledge how the main reasons for not using shared e-scooters are affected by different socioeconomic groups. The results presented in Table 17 show a statistical correlation between the gender of the non-user and the feeling of lack of road safety, as well 38,70% 38,10% 37,50% 34,40% 24,10% 23,50% 22,00% 11,50% 9,00% 8,40% 5,90% 0,00% 5,00% 10,00% 15,00% 20,00% 25,00% 30,00% 35,00% 40,00% 45,00% Prefer to use PT Prefer to use a car I do not know how to use it Lack of safety Bad weather conditions High price Bad pavement conditions High speed limit Lack of dedicated parking Small geofence area Uncomfortable Chapter IV – Case Study: City of Braga 112 variables that influence modal choice on trips respondents make to and from the University by bicycle. In this experiment, respondents had seven different hypothetical choices that comprised the following attributes: car travel time, bicycle travel time, cost, and presence of dedicated infrastructure. In Belgium, Adnan, Altaf, Bellemans, Yasar, & Shakshuki (2019) used stated choice to identify the willingness to use a shared bicycle for the last mile of journeys through the rail. The experiment had nine different hypothetical choices with attributes regarding the cost of the trip, travel time, trip distance, presence of segregated bike lanes, rainfall, availability of other modes, temperature, and bicycle parking at the destination. In order to better understand mode choice when micromobility services are available, namely shared bicycles, Li & Kamargianni (2019), Masoumi (2019), and Politis, Fyrogenis, Papadopoulos, Nikolaidou, & Verani (2020) deployed SC experiments in China, Greece, and the Middle East, respectively. The attributes used in the experiment were travel time, cost, walking time, parking availability and cost, bus availability, and mobile app usage. To test the choice of a micromobility mode if there is a public transportation strike in London, UK, Manca, Sivakumar, & Polak (2019) deployed an SC experiment to compare shared taxi services, docked bikesharing services, and dockless bike-sharing services. Civil engineering students from two different universities were given twelve hypothetical choices, each with attributes such as in-vehicle time or cycling time, cost, walking time, and weather conditions. Since 2021, shared e-scooters have also been introduced in stated choice experiments regarding the usage of micromobility in cities. Studies from Baek, Lee, Chung, & Kim (2021), (Awad-Núñez, Julio, Gomez, Moya-Gómez, & Sastre González (2021), van Kuijk, de Almeida Correia, van Oort, & van Arem (2022), McQueen & Clifton (2022), Nikiforiadis et al. (2023b) and Yan, Zhao, Broaddus, Johnson, & Srinivasan (2023) surveyed students and public transportation users in the USA, Europe, and Asia to identify their willingness to use of this mode of transportation for last-mile trips connected to public transport. The attributes in the choice experiments focused on the travel time, cost, waiting time for public transport, public transportation frequency, walking time, and the extra cost promoted by using a new mode of transportation for the last mile traveled. In addition, Hong, Jang, & Lee (2023) used stated choice Impacts of Shared E-scooters on Urban Transportation Systems 113 to identify a specific mode choice for the last mile after parking a car in a shared parking lot. The attributes used were travel time, cost, shared micromobility availability at departure, and weather conditions. The case studied by Cao, Zhang, Chua, Yu, & Zhao (2021b) comprises the usage of shared e-scooters to replace short-distance public transportation trips in Singapore, where public transportation users face several stops and transfers to make short displacements. In this case, students, white-collar workers, and tourists faced hypothetical choices with attributes such as public transportation stops, public transportation transfers, public transportation access within walking distance, public transportation fare, public transportation travel time, e-scooter travel time, and e-scooter fare. In Paris, Gioldasis, Christoforou, & Seidowsky (2021) studied the impacts of specific human factors that influence shared e-scooter riding. For this stated choice experiment, users of shared e-scooter services faced hypothetical choice scenarios comprising attributes such as mode choice criteria (i.e., comfort, safety, cost), e-scooter use frequency, use frequency for strolling, and the primary motivation for using the service (i.e., leisure activities, services, commute, visit family) Then, Krauss, Krail, & Axhausen (2022) deployed a stated choice experiment in Germany to identify the willingness to use shared micromobility modes (e.g., shared bicycles, shared e-scooters) for short or medium-distance leisure trips. The attributes were the length of the trip, travel time, access time, egress time, availability of the shared micro vehicle, and cost. In addition, a synthesis of the main attributes used in all foremost studies is presented in Table 20. Besides the attributes used in stated choice experiments that value micromobility and shared micromobility as a way to complement trips as a last-mile option and the willingness to shift from current modes of transportation, some latent variables and sociodemographic information are also considered. Latent variables are regarded as further opportunities to enhance the behavioral realism of shared micromobility choices. They are represented by attitudes and perceptions on mode-choice decisions (W. Li & Kamargianni, 2019), while sociodemographic data from respondents are an essential source of explanation for the choices, as well as a way to check whether the sample collected matches known characteristics of the population of interest (Hensher et al., 2005b). Chapter IV – Case Study: City of Braga 114 The main latent variables used in the twenty references used in this section regard the respondents’ perspective on micromobility safety, management, the newness of this mode of transport, the possible positive environmental outcomes from the usage of micromobility, flexibility, convenience for short trips, operation of the micro vehicle or the app to unlock and pay for the service, trip purpose, experience using shared micromobility (e.g., involvement in accidents, smartphone usage while riding), physical activity promoted by micromobility, and traffic safety. Besides the latent variables, all studies collected sociodemographic information from respondents. This information regards their gender, age, employment status, monthly income, educational level, if they have a mobile data plan, car ownership, bicycle ownership, possession of a driver’s license, possession of a public transportation card or monthly pass, frequency of micromobility usage, and general state of health. Impacts of Shared E-scooters on Urban Transportation Systems 115 Table 20: Main attributes used in SC experiments for micromobility Authors Year Type of Micro vehicle Context Attributes Travel time Cost Availability Weather condition Parking Parking cost Dedicated infrastructure Walking time Campbell et al. 2014 Shared bike, shared e-bike Factors influencing the switch from the respondent’s preferred mode to a new hypothetical mode X X Campbell et al. 2016 Shared bike, shared e-bike Factors influencing the choice to switch from a current mode of transportation to bike sharing X X X Lin, Wang & Feng 2017 Shared bikes Relationship between shared bike pricing and usage as a transfer mode from metro X X X Li & Kamargianni 2018 Shared bikes Choice of a mode of transportation in short-distance trips in heavily air-polluted cities X X X X X OrozcoFontalvo et al. 2018 Shared bikes Identification of variables that influence inhabitant’s modal choice and their perception of the trips they make by bicycle X X X Adnan et al. 2019 Shared bikes Factors influencing the usage of shared bicycles for the last mile of journeys through the rail X X X X X X Li & Kamargianni 2019 Shared bikes Mode choice behavior on shared mobility services X X X X X Manca, Sivakumar & Polak 2019 Shared bikes Choice of shared bikes if there is a public transportation strike X X X X Masoumi 2019 Shared bikes Explain the causality of urban travel mode choice X X Politis et al. 2020 Shared bikes Mode choice between current mode and shared bicycle X X X Cao et al. 2021 Shared escooters To what extent shared e-scooters can replace shortdistance trips made by public transport X X X Chapter IV – Case Study: City of Braga 116 Authors Year Type of Micro vehicle Context Attributes Travel time Cost Availability Weather condition Parking Parking cost Dedicated infrastructure Walking time Baek, Lee & Kim 2021 Shared escooters How people value shared e-scooters as a last-mile mode of transportation X X X Gioldasis, Christoforou & Seidowsky 2021 Shared escooters Impacts of specific human factors on e-scooter riding X Awad-Núnez et al. 2021 Shared bikes, shared escooters Respondent’s willingness to use and pay for public transportation and shared e-scooters given a set of Covid-19 safety measures X van Kuijk et al. 2022 Shared bikes, shared escooters Preference for shared modes as first and last mile connection in local public transport X X Krauss, Krail & Axhausen 2022 Shared bike, shared escooter Willingness to use shared modes for short or medium-distance leisure trips X X X X McQueen & Clifton 2022 Shared escooters Willingness to take multimodal (PT and e-scooter) trips to the university X X X Hong, Jang & Lee 2023 Shared bikes, shared escooters Mode choice for the last mile after parking a car in a shared parking lot X X X X Nikiforiadis et al. 2023 Shared escooters Willingness to use shared e-scooters for intermodal trips X X Yan et al. 2023 Shared escooters Enhancement of public transportation by e-scoot and ride option X X Impacts of Shared E-scooters on Urban Transportation Systems 117 IV.3.3. Stated choice design for commute trips The data collected in the literature review on stated choice experiments regarding the usage of micromobility in cities, as well as the results from the revealed preference survey in Braga, allows better analysis and refinement of the structure and the attributes that can be used in the SC experiment in Braga for shared e-scooters. IV.3.3.1. Context for the stated choice experiment for commute trips In order to increase the respondents’ familiarity with the context of the stated choice experiment, a wellknown route in Braga was chosen. This route is where respondents can experience the changes in their commute according to the attributes and attribute levels of the experiment. The route connects Braga’s train station and the University of Minho. It is 4.5 km long and composed of Caires Street, Conde Dom Henrique Avenue, Imaculada Conceição Avenue, João XXI Avenue, and João Paulo II Avenue (Figure 40). Figure 40: The selected route for the stated choice experiment for commute trips (own author) This route was also chosen because it has the potential to be a direct connection from the train station to the University of Minho by micromobility. Currently, shared e-scooters and shared e-bikes have restrictions on using this route due to the high speeds allowed for cars. The geofence strategy allows Chapter IV – Case Study: City of Braga 118 shared micromobility vehicles to reach the maximum speed of 6 km/h on the streets that compose the route selected. The possibility of creating a green route in this very same location would allow residents to be displaced safely from two important trip generator hubs in Braga. This would also contribute to generating equity in this area's public space since the redistribution allows different types of users to use it. The context of the experiment also regards the adoption of a BUS lane (Bus Rapid Transit – BRT) to increase the realism of the choice experiment since the city of Braga has plans to implement a rapid bus lane that connects the two hubs studied here, namely the train station and the University of Minho. The current state of the route prioritizes private vehicles for displacements. Imaculada Conceição Avenue, João XXI Avenue, and João Paulo II Avenue have five to six car lanes, with no dedicated infrastructure for buses or micromobility. In addition, there is a lack of accessibility on the route since pedestrians can only cross the roads using overpasses and tunnels (Figure 41). Figure 41: Current state of João Paulo II Avenue (own author) Thus, the creation of a green route for micromobility and the implementation of a BUS lane in the route selected to connect the train station to the University of Minho would allow for a better distribution of Impacts of Shared E-scooters on Urban Transportation Systems 119 public space and the support for the use of more sustainable modes of transport, therefore reducing the modal share of private cars in the city of Braga. IV.3.3.2. Alternatives, attributes, and attribute levels To compare the likelihood of using different modes of transportation in Braga, namely private cars, buses, shared e-scooters, and shared e-bikes, some attributes are selected to constitute the stated choice experiment. Travel time and the cost of the trip are crucial elements that need to be considered in this type of experiment. The following attribute selected is the availability of dedicated infrastructure (Adnan et al., 2019; Campbell et al., 2016b; Lin et al., 2017; Masoumi, 2019; Orozco-Fontalvo et al., 2018), which is corroborated by the results from the revealed preference survey with non-users of the services since the implementation of cycle lanes and zoning is likely to increase the number of potential users of shared e-scooters in the future in Braga. In addition to the increased safety for shared micromobility users promoted by the implementation of dedicated infrastructure and zoning, employing intelligent sensors in the micro vehicle would be an asset to detect and inform riders of dangerous riding contexts. The technology for Dangerous Riding Detection (DRD) has already been implemented in Canada, Australia, and the United Kingdom, where e-scooters have this newly developed technology that allows high accuracy location, rapid geofence detection, and an array of multi-function sensors that detect unsafe behavior. The sensors monitor real-time dangerous habits for riders, such as sidewalk riding, aggressive swerving, skidding, tandem riding, and curb jumping (Onag, 2021a) (Figure 42). The implementation of an obstacle detection system in shared e-scooters and e-bikes in Braga, as it has been made in the countries mentioned above, would improve the awareness of riders for the most common dangers that provoke injuries among the users of this mode of transportation (Nikolaj et al., 2019; Ragot-Court et al., 2021) Chapter IV – Case Study: City of Braga 120 Figure 42: Dangerous riding sensors in e-scooters (Onag, 2021b) In addition, the provision of onboard Wi-Fi in buses is seen as a technological advancement that increases the quality of the service provided to the population (Rupprecht Consult (editor), 2019). This technology has already been used in some cities, such as Oxford, to support urban mobility principles (Department for Transport, 2020). Wi-Fi as an attribute in stated choice experiments has already been used to access users' preferences for bus services, according to studies performed by Chalak, Al-Naghi, Irani, & AbouZeid (2016). Therefore, the stated choice experiment deployed in Braga is composed of four different attributes that are relevant according to the literature and the data collected in Braga through the revealed preference survey, and they are:  Travel time;  Travel cost;  Presence of dedicated infrastructure;  Presence of technology (i.e., the presence of Wi-Fi in buses and the presence of DRD technology in shared e-scooters and e-bikes). The four attributes mentioned above were used in four alternatives to provide the population with a realistic stated choice experiment. Each attribute has two or three different levels of attributes, as they encompass the current situation of the modes of transportation and other hypothetical situations. The Impacts of Shared E-scooters on Urban Transportation Systems 121 trips were considered to be 4.5 km long, which comprises the distance between the Braga’s train station and the University of Minho campus in Gualtar.  Car For this Stated Choice experiment, the travel time and cost of current commute trips are considered the same as those of private cars. The trip in a private vehicle between the train station and the University of Minho campus in the selected route takes 8 minutes using Google Maps (© Google) and costs 3.50 EUR. The cost of the private car trip is measured according to the reference for a Volkswagen Golf 1.0, estimated to be EUR 0.83 per km traveled, which was also used in the study by Gössling, Kees, & Litman (2022). The cost per kilometer traveled comprises private costs (e.g., vehicle depreciation, operating costs, fixed costs, repairs, and maintenance) and social costs (e.g., health costs, infrastructure, subsidies, and environmental costs). The price of the reference level is estimated to be the cost of a 4.5 km long trip, which comprises the route selected for the stated choice experiment.  Bus The travel time for a bus to depart from the train station and arrive at the University of Minho through the selected route is 15 minutes on Google Maps (© Google), a base level for the experiment. Then, level 3, which represents the optimal condition for the bus to travel between the train station and the University of Minho, counts for the travel time in a dedicated infrastructure on an arterial road (Hook & Wright, 2017). Level 2 represents a mid-point between the most extended trip versus the quickest trip scenario. The base scenario for the bus trip cost represents the pre-paid single-trip ticket, which costs EUR 0.75 in Braga. Level 2 represents the daily cost of bus usage for adults who pay for the monthly bus pass, which means a decrease in the trip cost for the stated choice scenario (TUB, 2023). The last level for this attribute presents the bus fare with no cost for the population to incorporate the Sustainable and Smart Mobility Strategy launched by the European Commission (European Commission, 2020b) as well as the new trend of free public transportation already taking place in Luxembourg, Malta, Hasselt (Belgium), Tallinn (Estonia), Dunkirk (France), and Cascais and other cities in Portugal. Technology in buses refers to providing Wi-Fi service to the population. In this case, there are only two attribute levels: the existence or nonexistence of Wi-Fi in the bus. Chapter IV – Case Study: City of Braga 128 Category Subcategory N Sample % Sample % Braga* Household monthly income Up to EUR 760 13 3.1% - EUR 761 – EUR 1000 54 12.7% - EUR 1001 – EUR 1500 106 24.9% - EUR 1501 – EUR 2000 68 16.0% - EUR 2001 – EUR 2500 49 11.5% - Household monthly income Above EUR 2501 80 18.8% - *Population of Braga considering the minimum age of 18 years old (minimum age to have a driver’s license and to rent an e-scooter or e-bike The respondents also stated that 41.6% of their households have two cars. Also, 59.1% of the respondents never use the buses in Braga, and 83.1% do not have a monthly public transportation pass. Regarding the previous usage of shared micromobility, 76% of respondents said they never used an e-scooter in Braga. In comparison, almost all respondents (97.2%) said they had not used an e-bike in the city. The overall choice for modes of transportation in the scenarios presented showed that the preferred sustainable mode to replace the car in Braga would be the bus. Shared micromobility is still not preferred in relation to the vehicle. Figure 45 shows the modes of transportation in the stated choice experiment for commute trips in Braga. Figure 45: Choice of modes of transportation in the SC experiment for commute trips 57,5 17,6 14,5 10,4 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Modes of transportation Bus Car E-bike E-scooter Impacts of Shared E-scooters on Urban Transportation Systems 129 It is possible that the bus is the preferred mode of transportation chosen by the respondents to switch from the car to achieve sustainable mobility in Braga. The bus even surpassed the statical preference to continue using the vehicle, which would not be expected due to decision inertia, which is the tendency to repeat the same previous choices (i.e., using a car) regardless of the outcome (Jung et al., 2019). Table 23 also shows the preference for buses over cars and the effects of attributes in choosing a mode of transportation in Braga. Table 23: Base logit model for the commute SC experiment Parameter Parameter estimate Rob. std err Rob. t-test Rob. p-value ASC_BUS 1.009 0.080 12.55 0.0 ASC_CAR -0.517 0.065 -7.89 2.88e-15 B_COST -0.085 0.023 -3.60 3.08e-04 B_INFRA 0.004 0.000 10.22 0.0 B_TECH 0.112 0.042 2.66 7.74e-03 B_TIME -0.076 0.005 -14.65 0.0 Summary statistics Number of observations = 5100 𝓛(𝚶) = -7070.101 𝓛(𝜷) = -5647.564 𝝆−𝟐 = 0.201 For estimation purposes and considering that using the model with the alternative-specific constants to cars and buses gave the highest rho-square values, the alternative-specific constants for e-scooters and e-bikes were normalized to zero. The estimated values for the alternative-specific constants ASC_CAR and ASC_BUS show that all else being equal, there is a preference for the bus concerning the other modes of transportation available (i.e., car, e-bike, e-scooter). As expected, both the travel time and the cost coefficients have negative signs. The higher the travel time or the cost of an alternative, the lower the related utility. The positive estimate of the infrastructure and technology coefficient (B_INFRA and B_TECH) indicates that the utility is higher when more infrastructure and technology are offered. Next, sociodemographic characteristics are added to the investigation results to capture the heterogeneity of preferences in the population. Table 24 shows the results, considering a dummy variable (SENIOR) that identifies people who are 40 years old or older. This age frame for SENIOR was selected following the pilot results, which showed that people over 40 would preferably choose the bus in all scenarios. The base age category in the model specification was considered to be from 18 to 39. Chapter IV – Case Study: City of Braga 130 Table 24: Logit model with the sociodemographic characteristic age Parameter Parameter estimate Rob. std err Rob. t-test Rob. p-value ASC_BUS 0.811 0.084 9.62 0.0 ASC_CAR -0.523 0.065 -7.96 1.7e-15 B_COST -0.085 0.023 -3.63 0.0 B_INFRA 0.004 0.0 10.3 0.0 B_SENIOR -0.479 0.059 -8.05 8.88e-16 B_TECH 0.114 0.042 2.7 0.006 B_TIME -0.076 0.005 -14.7 0.0 Summary statistics Number of observations = 5100 𝓛(𝚶) = -7070.101 𝓛(𝜷) = -5614.35 𝝆−𝟐 = 0.206 The estimation results for this model show a negative sign of the age coefficient (dummy variable SENIOR), reflecting that people over 40 are less likely to use shared micromobility in Braga compared to younger people. This aligns with the expectations that older people would not be willing to shift to micromobility due to the lack of experience in riding these two vehicles and the perceived high lack of safety in a two-wheeled vehicle. The influence of the respondents' gender was also tested in the SC experiment model for commute trips in Braga. Table 25 presents the results considering a dummy variable, FEMALE, describing people who identify themselves as female. The base gender category in the model specification was considered to be the male one. Table 25: Logit model with the sociodemographic characteristic gender Parameter Parameter estimate Rob. std err Rob. t-test Rob. p-value ASC_BUS 0.712 0.086 8.24 2.22e-16 ASC_CAR -0.45 0.068 -6.54 6.26e-11 B_COST -0.088 0.023 -3.76 0.0 B_FEMALE -0.409 0.057 -7.09 1.38e-12 B_INFRA 0.002 0.0 5.34 9.19e-18 B_TECH 0.114 0.042 2.71 0.0 B_TIME -0.073 0.005 -14.4 0.0 Summary statistics Number of observations = 5040 𝓛(𝚶) = -6986.924 Impacts of Shared E-scooters on Urban Transportation Systems 131 Summary statistics 𝓛(𝜷) = -5622.743 𝝆−𝟐 = 0.195 The negative sign of the gender coefficient (referring to the FEMALE dummy variable) reflects a rejection from females towards shared e-scooters and e-bikes in Braga. This finding is aligned with the expectation considering the RP results for the city, dictated probably by safety reasons concerning these shared modes of transport, and since women usually are responsible for carrying children to school and grocery shopping, which are two activities that are enabled by modes of transportation that can carry only one person at a time, and do not have space to carry shopping bags (UN Women, 2019). The possession of a bus monthly pass was also tested in the model to assess its effect in the usage or not usage of shared micromobility in Braga. Table 26 shows the results of the model considering a dummy variable (BPASS) determining respondents that own a bus monthly pass in Braga (i.e., TUB pass). Table 26: Logit model with the characteristic of possession of a bus monthly pass Parameter Parameter estimate Rob. std err Rob. t-test Rob. p-value ASC_BUS 1.03 0.087 11.8 0.0 ASC_CAR -0.476 0.068 -6.98 3.02e-12 B_BPASS -0.152 0.055 -2.75 0.006 B_COST -0.085 0.024 -3.53 0.0 B_INFRA 0.004 0.0 10.2 0.0 B_TECH 0.112 0.043 2.61 0.0 B_TIME -0.077 0.005 -14.6 0.0 Summary statistics Number of observations = 5100 𝓛(𝚶) = -7273.872 𝓛(𝜷) = -5697.455 𝝆−𝟐 = 0.217 The estimation results for this model show a negative sign of the pass coefficient (referring to the BPASS dummy variable), which indicates that people who own a bus monthly pass in Braga are less likely to choose shared e-scooters and e-bikes for commuting compared to people who do not own a bus monthly pass. This seems to be a plausible conclusion since bus pass owners prefer to take the bus instead of incurring extra costs when using other modes of transport. Chapter IV – Case Study: City of Braga 132 The respondents' professions were also tested in the model to assess how this variable affects the choice of shared micro vehicles in Braga. For this, a dummy variable EMPLOYER and a dummy variable STUDENT were considered. In this model, the base category considered was to be an employee. Table 27 highlights the results from the interaction of the dummy variables in the model. Table 27: Logit model with the sociodemographic characteristic profession Parameter Parameter estimate Rob. std err Rob. t-test Rob. p-value ASC_BUS 0.99 0.081 12.1 0 ASC_CAR -0.519 0.065 -7.91 2.44e-15 B_COST -0.085 0.023 -3.61 0.0 B_EMPLOYER 0.314 0.137 2.28 0.022 B_INFRA 0.004 0.0 10.2 0.0 B_STUDENT -0.19 0.076 -2.5 0.012 B_TECH 0.113 0.042 2.68 0.007 B_TIME -0.076 0.005 -14.7 0.0 Summary statistics Number of observations = 5100 𝓛(𝚶) = -7070.101 𝓛(𝜷) = -5641.243 𝝆−𝟐 = 0.202 The positive sign of the profession coefficient for the dummy variable EMPLOYER shows this group is more likely to use shared e-bikes and shared e-scooters in Braga than employees. This result contributes to the literature confirming the usage of shared micromobility, mainly shared e-scooters, by businesspeople who usually work in the city's central areas. On the other hand, the negative sign of the profession-related coefficient estimate referring to the dummy variable STUDENT reveals that this stratification is less likely to use shared e-bikes and shared e-scooters for commute trips (compared to employees), meaning they prefer other modes of transportation to go to classes. This can be related to the latest measures taken by the Municipality, and the public enterprise that provides buses in Braga, which started providing free monthly bus passes for students of all ages in Braga. Next, the education status of the respondents was tested in the model to capture the heterogeneity of taste across different categories. For this, five dummy variables were used: a dummy variable (DOC) to identify people who have a doctoral degree, a dummy variable (HIGH) to identify people who finished high school, a dummy variable (MAST) to identify people who completed their master’s degree, a dummy variable (MID) identifying people who have finished middle school, and a dummy variable (TECHNI) Impacts of Shared E-scooters on Urban Transportation Systems 133 identifying people who have accomplished a technical course. In this model, the base educational status category in the specification was considered to be the bachelor’s degree. Table 28 presents the results from the model with these dummy variables. Table 28: Logit model with the sociodemographic characteristic educational status Parameter Parameter estimate Rob. std err Rob. t-test Rob. p-value ASC_BUS 0.913 0.087 10.5 0.0 ASC_CAR -0.523 0.065 -7.96 1.78e-15 B_COST -0.085 0.023 -3.61 0.0 B_DOC -0.638 0.203 -3.14 0.001 B_HIGH -0.33 0.073 -4.49 6.99e-06 B_INFRA 0.004 0.0 10.3 0.0 B_MAST -0.147 0.074 -1.97 0.049 B_MID -0.042 0.305 -0.137 0.891 B_TECH 0.113 0.042 2.67 0.007 B_TECHNI 0.62 0.127 4.9 9.74e-07 B_TIME -0.076 0.005 -14.7 0.0 Summary statistics Number of observations = 5100 𝓛(𝚶) = -7070.101 𝓛(𝜷) = -5613.222 𝝆−𝟐 = 0.206 For commute trips in Braga, it is possible to infer that people with a technical course are more likely to use shared micromobility modes compared to people with a bachelor’s degree since the parameters estimate for this dummy variable presents a positive sign. Nevertheless, people with a doctorate, master’s degree, high school, and middle school are less likely to use a shared e-bike or shared e-scooter for their commute trips in comparison to people with a bachelor’s degree; the negative signs in the parameter estimate for these dummy variables attest to this. When the family size is investigated to capture its heterogeneity effects in the population, four dummy variables are used: dummy variable (ONE) for a unipersonal family, dummy variable (THREE) for families with three people, dummy variable (FOUR) for families of four people, and dummy variable (MORE) for families with more than four people. In this model, the base category is a family with two people. Table 29 presents the results from the model with family-size dummy variables. Chapter IV – Case Study: City of Braga 134 Table 29: Logit model with the sociodemographic characteristic family size Parameter Parameter estimate Rob. std err Rob. t-test Rob. p-value ASC_BUS 0.871 0.092 9.4 0.0 ASC_CAR -0.519 0.065 -7.91 2.66e-15 B_COST -0.085 0.023 -3.6 0.0 B_FOUR -0.085 0.079 -1.08 0.28 B_INFRA 0.004 0.0 10.2 0.0 B_MORE -0.252 0.12 -2.1 0.035 B_ONE -0.088 0.098 -0.903 0.367 B_TECH 0.112 0.042 2.65 0.008 B_THREE -0.329 0.078 -4.2 2.66e-05 B_TIME -0.076 0.005 -14.6 0.0 Summary statistics Number of observations = 5100 𝓛(𝚶) = -7070.101 𝓛(𝜷) = -5637.358 𝝆−𝟐 = 0.203 It is possible to infer that, due to the negative signs for dummy variables ONE, THREE, FOUR, and MORE, people in these family sizes are less likely to use shared micromobility to commute in Braga compared to families with two people. It would be expected that people in bigger families would not prefer to use shared micromobility because commute trips, in this case, could involve multiple stops (e.g., a stop at school to leave children). Therefore, a shared micro vehicle would not be a suitable option. However, in Braga, people from small families (i.e., one person) would also be less likely to ride shared e-scooters and e-bikes than people in families with two people. The last sociodemographic characteristic tested to capture the heterogeneity of taste in the population is the family monthly income of the respondents. In this case, a dummy variable (INCOME) was used to identify family monthly incomes higher than EUR 1,251, which represents a higher income than the average for the city of Braga, which is EUR 1,146 (Instituto Nacional de Estatística, 2022a). Table 30 presents the results from the model considering the dummy variable INCOME. Table 30: Logit model with the sociodemographic characteristic family income Parameter Parameter estimate Rob. std err Rob. t-test Rob. p-value ASC_BUS 0.755 0.090 8.31 0.0 ASC_CAR -0.553 0.07 -7.9 2.89e-15 Impacts of Shared E-scooters on Urban Transportation Systems 135 Parameter Parameter estimate Rob. std err Rob. t-test Rob. p-value B_COST -0.088 0.024 -3.56 0.0 B_INCOME -0.356 0.062 -5.74 9.56e-09 B_INFRA 0.004 0.0 9.94 0.0 B_TECH 0.081 0.044 1.82 0.069 B_TIME -0.078 0.005 -14.1 0.0 Summary statistics Number of observations = 4440 𝓛(𝚶) = -6155.147 𝓛(𝜷) = -4952.082 𝝆−𝟐 = 0.195 The negative sign of the income coefficient (referring to the INCOME dummy variable) reflects that people with higher income tend is less likely to choose shared micromobility to make their commutes compared to people with income lower than EUR 1,250. It seems a reasonable conclusion, dictated probably by the fact that higher income people would prefer to continue using their private cars to commute due to comfort reasons. When the income increases, there is a preference to continue using the vehicle, or if the mode of transportation needs to be replaced, the bus is chosen. After the test of the sociodemographic characteristics of the sample in the model to understand how they affect the preference for micromobility and sustainable modes of transportation in Braga for commuting trips, the incremental effects of the models were compiled in Table 31. Table 31: Incremental effects for the choice of micromobility in the commute SC experiment Variable/Description of the interaction effect Expected sign* (+) or (-) Incremental effect is statistically significant Model represents a statistical improvement Age Age of the respondent (senior)  (-) No Yes Gender Gender of respondent (female)  (-) No Yes Pass Possession of a bus pass  (-) No Yes Profession Student  (+) No Yes Employer  (+) Yes Degree Middle school  (-) No Yes High school  (+) No Technical course  (+) Yes Chapter IV – Case Study: City of Braga 136 Variable/Description of the interaction effect Expected sign* (+) or (-) Incremental effect is statistically significant Model represents a statistical improvement Degree Masters  (-) No Yes Doctorate  (-) No Family size One person  (+) No Yes Three persons  (-) No Four persons  (-) No More than five persons  (-) No Income Above 1251 EUR  (-) No Yes *(Ghasri et al., 2024; Nigro et al., 2024; Nikiforiadis et al., 2024) In Braga, the effects of sociodemographic characteristics align with the available literature for commute trips. Even though students, people with high school degrees, and single-person families were expected to prefer shared micromobility, the other factors followed the expected results. The age of the respondents (80% being 25 years old or older) also plays a key role in shared micromobility choice since it is expected that with the increase in the age of the respondents, the preference for these modes of transportation will decrease. Thus, even if the person has a high school degree, which does not mean that the person is young and live by themselves (single-person family), it does not reflect in a higher preference for commuting in shared e-scooters or e-bikes. However, the preference for the bus as a sustainable option for commuters is also a great advantage for the city to invest in this mode of transportation to promote modal shifts. In more than 50% of cases, people would choose the bus instead of any other mode of transportation if better traffic conditions and comfort were available for the riders. If only shared micromobility is regarded, it is seen that e-bikes overpass e-scooters in preference by respondents. Even if Braga shared e-bikes are unavailable, people would prefer this mode to e-scooters. These research findings indicate that e-bikes are preferred in relation to e-scooters for commute trips if they are available and safe conditions for riding are provided. Impacts of Shared E-scooters on Urban Transportation Systems 137 IV.4. Stated choice experiment for leisure trips IV.4.1. Objectives of the SC experiment for leisure trips The stated choice experiment for leisure trips aims to assess how different attributes can affect the choice of sustainable modes of transportation in Braga for this type of trip. The literature has evaluated the determinants of shared e-scooter usage and their implications for commuting. However, since one of the main reasons for using shared e-scooters is that they are fun and pleasant, people tend to use them for leisure trips. Therefore, the possible modal shift from cars to sustainable modes of transport, such as shared escooters, e-bikes, and walking, must be studied. This stated choice experiment evaluates how different attributes in the built environment can affect the modal shift from cars. Unlike the stated choice experiment for commute trips, the experiment for leisure trips considers that the modal shift occurs in short-distance trips, for walking, shared e-scooter, or shared e-bikes compared with the car. In this case, the bus alternative was substituted for walking, since short-distance trips are considered that can be made by this mode of transportation. This stated choice experiment is closely related to the following research questions:  RQ2: What are the main drivers/determinants of shared e-scooter usage for commute and leisure trips?  RQ3: How do transportation planning and traffic management influence the safe usage of shared e-scooters?  RQ4: What are the impacts of shared e-scooter usage on the sustainability of the transportation system?  RQ5: Are shared e-scooters more competitive than shared e-bikes for short and medium-distance trips? IV.4.2. Walking as a mode of transportation Of all journeys made daily in European Union countries, 20-40% travel by bicycle or on foot (European Commission, 2023). The average length of walking trips varies from just under 1 km to 2.8 km. However, the extent of coverage of short trips may differ from country to country (European Commission, 2023). Walking is a way of traveling used mainly for two purposes: short trips to specific destinations such as 240 241 242 243 244 245 246 247 248 249 256 257 258 259 260 261 262 263 264 265