Abstract
Whilst environmental issues and sustainability rise on the political agenda, the shadow side of car usage is becoming more visible. Governments struggle with solving traffic congestion problems and reducing CO2 emissions as car usage continues to increase. An interesting case offers in the Netherlands, where bicycling is a convenient and practical alternative to the car. However, the same issues as elsewhere in the world continue to exist in the Netherlands. This irrational behaviour in the Netherlands is studied in this thesis, with use of insights of behavioural economics. The effect of social norms on the transportation mode choice for commuting is studied. In particular, the influence of prestige and opinions of friends. The results confirm with earlier studies, showing that prestige and opinions of friends have a significant effect on the transportation choice, however quite small.
Table of content
Abstract ... 2 1. Introduction ... 4 2. Theoretical framework ... 6 2.1 Biases ... 6 2.2 Information ... 8 2.3 Social norms ... 82.3.1 Social norms: descriptive versus injunctive ... 9
2.4 Other factors influencing transportation choice ... 10
2.5 Hypotheses... 10
3. Data and methodology ... 11
3.1 The sample ... 11
3.2 Operationalization of social norms ... 11
3.2.1 Opinions of friends and/or peers ... 12
3.2.2 Prestige ... 12
3.2.3 Control variables ... 12
3.3 Methodology ... 12
4. Analysis ... 14
4.1 Overview of the sample: descriptive statistics ... 14
4.2 Assumption tests and likelihood ratio test ... 16
4.3 The effect of social norms on home-to-work commute: logistic regressions... 18
4.3.1 The influence of friends on the transportation mode from home-to-work ... 18
4.3.2 Influence of prestige on home-to-work commute ... 19
4.3.3 Combined model: prestige and opinions of friends ... 20
4.4 Robustness checks ... 23
5. Conclusion and discussion ... 28
1. Introduction
Ever since cars have been introduced to the public, its usage has increased immensely (CBS, 2017). But by the increasing usage of cars, the shadow side of cars has been revealed (Moriguchi, Kondo & Shimizu, 1993). Not only the environment is negatively impacted by rising C02 emissions, but the life cycle of a car includes the processing of raw materials, the assembly, use and maintenance and eventually the final disposal as well (Moriguchi, Kondo & Shimizu, 1993). Furthermore, impacts of construction and maintenance of road infrastructure should be considered as well (Moriguchi, Kondo & Shimizu, 1993). Beside these impacts, the massive amount of car usage leads to another issue concerning worldwide traffic congestion (INRIX, n.d.). Considering all the above (negative) impacts that cars have on the environment and individual lives, what drives people to still use the car?
The negative impacts of cars did not escape government its sight. Western European countries took the United States as a role model for policies adapting cities to car travel during the 1920s until the 1960s. But in the 1960s European countries and cities started to focus their policies towards promoting cycling, walking and public transportation (Buehler, 2014). More recently, as issues regarding the environment and traffic congestion start to burst at the seams, these issues have risen on the political agenda (ANWB, 2018). Despite policies focussing on promoting other transportation modes than the car, in the European Union solely, one out of two persons own a car (van Exel & Rietveld, 2009). Moreover, 80 to 90 per cent of all passenger kilometres are travelled by car (van Exel & Rietveld, 2009). Next to this, the Kennisinstituut voor Mobiliteitsbeleid (2018) shows that car usage rates continue to increase. It presents the complicated task for governments to reduce car usage and stimulating more sustainable transportation modes.
A fascinating case offers in the Netherlands, where bicycling is a safe, convenient and a practical way to get around cities (Pucher & Buehler, 2008), but where traffic congestion is problematic as well (Kennisinstituut voor Mobiliteitsbeleid, 2018). The Netherlands, moreover, has one of the highest levels of car ownership in the world (Pucher & Buehler, 2008). An initiative by the Dutch government called ‘Beter Benutten’ that has run from 2011 until 2018 has led to 80,000 less vehicles during peak hours (Rijkswaterstaat, 2018). Nonetheless, the Netherlands continues to cope with daily traffic congestion, although bicycling in many cases is a viable alternative.
Beside the difficult task set aside for government, the behaviour of individuals is interestingly as well. Standard micro-economic theory explains transportation mode choice as the “optimal decision to maximise individual utility subject to time and budget constraints” (Innocenti, Lattarulo & Pazienza, 2013: 158). It requires travellers to not be influenced by biases and to process all information correctly (Loewenstein, 1999). However, how can travellers meet to these requirements at all times? Standard microeconomic rationality theory seems to fail to explain decision-making in this aspect. How can commuting to work by car be rational behaviour when it does not maximize utility subject to time and budget constraints? Behavioural economics, on the other hand, incorporates aspects of psychology and
economics together, to explain irrational behaviour (Loewenstein, 1999). The body of literature on transportation behaviour including aspects of behavioural economics is growing. Different themes, such as biases, information provision and social norms are addressed by focussing on different modes of transportation. Exel (2011) for example showed that an alternative transportation mode only becomes a good alternative when the individual is sufficiently enough informed about the details, such as routes or stations. Other studies focus on the societal aspect of owning and using a car. Research has shown that cars have a special status in society. Children are at a very young age already influenced by the different possible transportation modes (Haustein, Klöckner & Blöbaum, 2009).
The continuing issues regarding car usage and the interesting case in the Netherlands provides the core of this thesis. Alongside practical factors influencing transportation choice, the irrationality of transportation choice allows for further research on this topic. Therefore, the following research question is answered: To what extent do social influences influence transport behaviour for commuters in the Netherlands (in 2016)?
Transportation and insights of behavioural economics is quite specific. Many studies within the existing body of literature focus on one type of transportation mode, or study how to reduce usage of a specific transportation mode rather than studying the requirements or drivers to transfer to another mode of transport. A study of the Join Research Centre of the European Commission (Lourence et al., 2016) furthermore shows that elements regarding mobility interventions are very limited. Therefore, studying and understanding the effects of a specific part of behavioural economics – in this case social norms – can be of use in further policy development and tackling global issues regarding sustainability. This research will thus contribute to filling the scientific gap within transportation research combined with behavioural insights. Its societal relevance of the study is more complex, as the issues regarding transportation are of interest for society and governments as well. As climate change, sustainability and environmental damages are topics rising on the political agenda all over the world, people are becoming more aware of the problems as well. Making more sustainable choices can influence our living environment and may have positive effects on (Dutch) society. Furthermore, relieving the traffic congestion problem in the Netherlands can be an important contribution to a more sustainable living. A less car-dependent lifestyle and reducing car usage is believed to be achievable by means of nudging.
Following this introduction, the second chapter explores the concepts and theories of behavioural economics, within the context of transportation mode choice. The third chapter focusses on how the sample has been constructed and explains how the research is conducted. The fourth chapter provides an extensive overview of the empirical findings of the logistic regression. The conclusion provides a reflection of the findings, elaborates on the implications of the findings by providing policy recommendations, shows the limitations of the study and provides suggestions for further research.
2. Theoretical framework
Behavioural economics uses insights from psychology to explain economic phenomena (Loewenstein, 1999). Within this specific field of economics, the concept of rationality plays an important role. Where many economic theories explain behaviour in terms of rationality, behavioural economics explains how people do not always follow the path of rationality. Moreover, behaviour in many cases is boundedly rational, or moreover irrational. The objective of behavioural economics, thus, is to figure out what really influences our daily life decisions.
Ever since transport modelling has been introduced in the 1950s, travel behaviour has received a lot of attention of researchers and practitioners. Arguments can be made that these transport models emerged from neoclassical economic concepts, where people are represented rational choice makers, who interact to achieve equilibrium (Avineri, 2012). But, as described above, research in behavioural economics has shown that people do not always make rational, consistent and efficient choices. Yet, where transport modelling has been a great area of interest, synergy between transportation research and behavioural economics has been relatively little (Gaker, Zheng & Walker, 2010). Metcalfe & Dolan (2012) state that this might be caused by the fact that in transportation research, behavioural economics is often not directly considered.
Nonetheless we find a growing body of literature on behavioural economics incorporated in transportation behaviour. In predicting transportation planning, it is utterly important to understand human behaviour. How can transportation behaviour be predicted and are we able to influence transportation behaviour over time to a maybe more sustainable choice? The body of literature in this area ranges from topics on the motives for mobility in the first place (work or social activities) to studies examining the drivers of car-use behaviour or car ownership.
The lessons in behavioural economics are enormous, but also popularity has risen immensely. Thinking, Fast and Slow (Kahneman, 2011), Nudge (Thaler & Sunstein, 2008), Predictably Irrational (Ariely, 2008) are a few of the many popular books that have been published relatively recently. Their conclusions have led to many more research, within different areas in the scientific arena. For the purpose of this study, which is understanding car-use behaviour, several themes within the behavioural economics are of particular interest. First, the topic of biases that cause people and individuals to systematically divert from rational behaviour. Second, the power of information and third, the influence on our behaviour by social norms and our peers.
The theoretical framework elaborates on the three main themes of behavioural economics and aim to summarize main ideas within the specific areas. Where possible, a body of literature is discussed that has connected specific themes within the transportation area.
2.1 Biases
Tversky and Kahneman’s research to three heuristics in the early 70s evolved in a research programme which studied how people make real-world judgments and under which conditions those judgments are
unreliable. Where the basic economic idea assumes that people make rational decisions, their study undermines this idea and states that decisions are not always made rationally. In their article “Judgment under Uncertainty: Heuristics and Biases”, the authors present “three heuristics that are employed to assess probabilities and to predict values” (Tversky & Kahneman, 1974: 1124). The three heuristics that Tversky & Kahneman presented are: (1) Representativeness, (2) Availability and (3) Adjustment and Anchoring.
The first heuristic Tversky & Kahneman (1974) present is representativeness. It is a question of probability and seeks to answer “how we value and weigh the attributes of alternatives” (Tversky & Kahneman, 1974: 1124). Tversky & Kahnemann (1974) showed that people are insensitive to prior probability of outcomes and sample size. Furthermore, their study showed that people have a misconception of Chance and Regression. It means for ‘chance’ that people see a normal event and think of it as being rare. For regression, it is the other way around; people see a rare event and think of it as being normal.
Literature that combines the representativeness heuristics and transportation is little. However, an individual’s beliefs can influence how people gain information and later behaviour in transportation mode choice. Because this bias is present in mobility choices, we must distinguish between the transportation alternatives that an individual is available of and the alternatives that are considered by the individual (Exel, 2011). For example, car drivers can believe that public transportation is not a viable option, where it could actually be a good alternative. Goodwin (1995) showed for example that there is a gap between subjective and objective car-dependence.
The second heuristic, availability, is the bias that we believe events are more probable the more easily we can recall an instance of them happening. Tversky & Kahneman (1974) describe that we assess certain events by the frequency of it, or the probability of the event happening. However, next to probability and frequency, there are a lot more factors that influence availability. Ergo, we are able to predict certain biases. An example that is used by Tversky & Kahneman (1974) is that people draw conclusions on the availability of their memories. This happens for instance when train and airplane crashes are sensationalized in the news. As these memories are more available, rather than everyday crashed on the highway, people are more likely to draw the conclusion that train and airplane travelling is safer than driving the car on the highway, where actually the opposite is true (Tversky & Kahneman, 1974).
Gaker & Walker (2011) state that “Many of our decisions are consequences of what we don’t know instead of what we do know, independent of the availability of that information” (Gaker & Walker, 2011: 109). It shows, for example, that if you do not know how the bus system works, that driving habits are reinforced. In addition, how we value certain characteristics of alternatives is influenced by the lack of perception of the true costs of a decision. In terms of transportation, this is shown in research by Parry, Walls & Harrington (2007). Their study researched the externalities of car-use, namely local
pollution, global pollution, oil dependence, traffic congestion and traffic accidents (Parry, Walls & Harrington, 2007). Their study shows that most people are not aware of the externalities, both in terms of social costs (externalities) or private costs (personal time and monetary expenses). According to the authors, we should deviate from the traditional economic policies implemented to reduce carbon emissions, thus car-use, to try to push car-users to a more sustainable choice.
The third heuristic presented by Tversky & Kahneman (1974) is adjustment and anchoring. “In many situations, people make estimates by starting from an initial value that is adjusted to yield the final answer” (Tversky & Kahneman, 1974: 1128). With Tversky & Kahneman’s findings as a foundation, many other behavioural economists continue uncovering biases that influence decision-making. One famous example is that of Ariely (2009), where the power of free is tested. Does reducing the price of two commodities by the exact same amount reverse consumer preference of one over the other? Basic economic principles state that this is not possible. The opposite seems to be true (Ariely, 2008). With reference to transportation decisions, Gaker, Zheng & Walker (2010) continued research in anchoring and framing in the field of transportation decisions. The authors showed in an experiment with college undergraduates that subject valued the power of free alternative beyond simply having no cost.
2.2 Information
The second theme within the behavioural economic research is information. Literature suggests that receiving complete information about alternatives is important in decision-making. In addition, it is important that people receive feedback about the consequences of their actions. In Thaler & Sunstein’s ‘Nudge’ (2008), for example, they found that providing information at time of consumption is important, rather than informing people ex ante. The particular case they used was the impact on health impacts of nutritional choices.
Little information is available that ex ante knowledge is influential in transportation choices. Gaker, Zheng & Walker (2010) studied to what extent information and feedback and social influences are of influence on behaviour modification. They found that trip- or individual-specific information on the environmental impact has potential to significantly alter behaviour towards a more sustainable choice. Their study confirms that social norms influences most powerfully transport behaviour.
2.3 Social norms
Human beings are social by nature. A long history of studies exists where the influence of our peers is examined. Jon Elster (1989) defined social norms by the contrary feature of rationality: social norms are not outcome-oriented. Specifically, Elster describes that norms become social when “they are shared by other people and partly sustained by their approval and disapproval” (Elster, 1989: 99). Many other social scientists thereafter sought to explain why social influences are so common. The result is that scientists have many different opinions on how social norms influence individuals or groups. Sherif (1936, cited in Gaker & Walker, 2011), for example, states that groups are supported in accomplishing its goals by social norms. Hence, norms survive on the basis of their success. In contrast, Berger &
Luckmann (1967) say that social norms in group accomplishments work the other way around. Norms rather define the group, instead that the group defines the norm. Another explanation of social norms is that social norms are of powerful use due to acceptance by the group (Solomon, Greenberg, Psyszczynski, 1991).
2.3.1 Social norms: descriptive versus injunctive
The social norms approach origins from a marketing perspective. As norms influence individual behaviour, it is extremely helpful for marketing purposes to influence consumer choices. Social norms can be divided into two main categories: descriptive and injunctive. Within the social norms marketing approach, there are two consistent findings. The first states that the greater part of individuals overestimate the prevalence of undesirable behaviours. Many studies have proved this by studying alcohol use among peers (see for example Prentice & Miller, 1993). The second finding states that individuals use their perceptions of peer norms as a standard against which to compare their own behaviour (see for example Perkins & Berkowitz, 1986). Social norms marketing campaigns aim to correct the individuals’ misperceptions regarding the behaviour its prevalence. Cialdini et al. (1991) and Reno, Cialdini, & Kallgren (1993) defined the perception of prevalence as the descriptive norm governing a behaviour. Alongside descriptive norms, there are also injunctive norms. Where descriptive norms refer to perceptions of common behaviour in a specific situation, injunctive norms refer to the perceptions of what is approved or disapproved within a society (Reno, Cialdini, & Kallgren, 1993). Action is thus motivated by social sanctions. According to Cialdini et al. (1991) both norms can be dominant in certain situations. Therefore, individual actions are directed by the salient norm in a particular situation.
The findings in behavioural economics literature show the importance of social norms in influencing decision-making, thus behaviour. Where studies mostly focus on social behaviour, such as health-care decision or alcohol use (Barr, 2012; Kahneman, 2003), social norms have shown to be influential within the transportation field as well. Many studies within this field focus on different aspects of the car, such as car ownership or car use. It has been shown that cars have a high symbolic and affective function, especially for frequent drivers (Steg, 2005). Steg (2005) researched the commuters in the Netherlands and showed that car use is not only for instrumental use. Furthermore, Sigurdardottir et al. (2013) for example studies the intentions of adolescents to commute by car or bicycle. Their findings conclude that intentions to use the car is related to positive car experience, general interest in cars and car ownership norms. On the other hand, cycling intentions are related to positive cycling experiences, willingness to accept car restrictions, negative attitudes towards cars and bicycle-oriented future vision. Their research furthermore shows differences in individual characteristics in intentions to use car/bicycle. Parallel to this study is the study of Haustein, Klöckner & Blöbaum (2009) who study the role of travel socialization. Children are at a very young age already shown how to differentiate between different modes of transportation and different levels of prestige. Several studies have shown that for example
owning a driver’s license is one of the most important life events regarding transportation modes (Klöckner, 2004). Another study of Schönhammer (1999, cited in Haustein, Klöckner & Blöbaum, 2009) shows that owning a car is most often more associated with the social aspect, rather than the technical and practical aspects. Kormos, Gifford & Brown (2015) studied the impact of descriptive social norm on self-reported reduction of private vehicle use. Their findings show that highlighting more prevalent descriptive social norms increased sustainable transportation behaviour for commuting, but not for non-commuting purposes.
2.4 Other factors influencing transportation choice
Next to influences from behavioural economics, such as social norms, there are many other factors that may be of influence in transportation decision-making. Akar, Fischer & Namgung (2013) studied the underlying differences in bicycling choice. Their case-study – Ohio State University – showed that women have different perceptions on safety and feasibility of alternative transportation modes. Encouraging women to bicycle, according to Akar, Fischer & Namgung (2013), need different policy and infrastructural changes. Next to gender, Steinbach et al. (2011) showed that ethnicity can have an impact on healthy transport choices. Different cultural populations have different affinities with bicycling. Steinbach et al. (2011) showed that the symbolic and aesthetic goals of cycling are more attractive to largely white men and women. People from other class and ethnic backgrounds found cycling to be less attractive (Steinbach et al., 2011). Beside personal characteristics, work-related factors have also shown to influence transportation choice. Heinen, Maat & van Wee (2013) have shown that positive cycling attitudes, colleagues’ expectations of cycling and presence of bicycle storage increase the likelihood of regularly commuting to work by bike. Working hours have shown to be of influence in Heinen, Maat & van Wee (2013) their study, when individuals needed a vehicle during working hours.
2.5 Hypotheses
The literature above describes the many different forms in which behavioural economics is expressed in the scientific field. For the purpose of this study, social norms are used to study travel behaviour of commuters in the Netherlands. Steg (2005) has shown that cars are not only used for its instrumental function, but have a highly symbolic and affective function. Together with the injunctive norm that individual behaviour is influenced towards the socially acceptable norm, the following hypothesis is tested: individuals whose friends have positive opinions towards car ownership and car use prefer to use the car for home-to-work commute. Furthermore, as Steg (2005) also shows, people have the opinion that cars have a means of showing prestige and someone’s status in society. Therefore, the following hypothesis is tested alongside: Individuals who are of the opinion that cars express a certain social status are more likely to commute by car.
3. Data and methodology
The following chapter describes in detail the research design and the method of data collection. As the dataset has had a major influence on the design of the research, the sample collection process is described first. Next, the dependent, independent and control variables are operationalized, showing how the hypotheses are measured. Lastly, the research strategy is set out, in which also the common threats of inference are elaborated on.
3.1 The sample
To construct the sample, the Netherlands Mobility Panel (MPN) is used. This is the most recently available data on travel behaviour in the Netherlands. The MPN is a household panel, which main objectives are to establish short-run and long-run dynamics in travel behaviour of individuals and households, and to determine how changes in personal and household characteristics and in other travel-related factors (e.g. economic crisis, reduced taxes on sustainable transport, changes in land-use or increases availability and use of ICT) correlate with changes in travel behaviour (see Hoogendoorn-Lanser et al. (2015) for more details).
Starting July 2013, respondents of 12 years or older from 2,500 complete households record their travel data using a three-day travel diary. For each respondent, the diary provides information about all trips (stages) the respondent made (transport modes, trip purposes, travel companionship, delays, parking costs). At the same time, respondents fill out different questionnaires offering a large amount of background information on the respondents and their households. This will be repeated at least yearly with the same respondents.
The MPN dataset distinguishes several sub-datasets of which each focusses on a different aspect. For the purpose of this study, the W4_Pdata and W4_Pdata_bijzonder are used. These two datasets are merged accordingly, using the unique person ID. As the study focusses on commuters, some filters were applied to create an appropriate sample. First, relevant characteristics were defined for commuters. To do this, the definition of commuters was used “a person who travels some distance to work on a regular basis” (CBS, 2016). Therefore, the respondents in the dataset must be part of the labour force, have a permanent working address outside of their home and be on a payroll. The first step in doing this is to drop respondents who are not part of the labour force by age. As the dataset only covers age by age categories, the age categories that are most closely to the cut-off of being in the labour force is used. The sample therefore only exists of people who are in between 17-70 years old. The sample is further specified to include only people who own a car and/or bike. It is without further notice that those people who do not own such a vehicle are unable to commute by means of car and/or bike. The above is summarized in Table 1 Sample collection process.
3.2 Operationalization of social norms
Literature on behavioural economics offers many interesting insights into travel behaviour. With regard to behavioural economics, this study focusses on social norms. Elster (1989) describes that the essential
elements of a social norm is that it is shared by other people. In light of that, social norms with respect to travel behaviour is conceptualized twofold: prestige and the opinions of friends and/or peers.
3.2.1 Opinions of friends and/or peers
The definition of social norms has shown that it is because of other people accepting a certain type of behaviour, that individuals are often urged to behave similarly. In studying social norms and transport behaviour, several aspects can be addressed. As for this specific study, there are three statements of use from the MPN dataset for this study. These statements regard car use, car ownership and the congestion problem in the Netherlands. The statements are as follows:
- Congestion problem: “My friends believe that the traffic congestion problem in the
Netherlands is greatly exaggerated”
- Car ownership: “My friends advised me to purchase a car”
- Car use: “My friends believe that you must only use a car when it is really necessary”
3.2.2 Prestige
Literature has shown that injunctive norms influence decision-making to the acceptable norm of society. As societies often deal with different societal hierarchies and the car is often associated with other means than its instrumental use only, one way of operationalizing a social norm for this study is by means of prestige. Within the MPN dataset there are several statement to which respondents answered whether they agree or not with the statement. The statements are as follow:
- Prestige statement: “Travelling must give me prestige”
- Personal style: “A car says a lot about someone's personal taste / sense of style”
- Cycling prestige: “Cycling gives me prestige”
- Car prestige: “Travelling by car gives me prestige”
- Status in society: “A car says a lot about a person's status in society”
3.2.3 Control variables
To control for other factors that could influence the most used transportation mode, some control variables are added to the models. Literature has shown that gender influences the likelihood to cycle, due to safety reasons and feasibility (Akar, Fischer & Namgung, 2013). Gender might influence the influence of social norms and is therefore added to the model. As the model aims to explain the use of cars and bicycle, it is important for the respondents to own such vehicles. Two control variables are added to control for the ownership of these vehicles: car ownership and bike ownership.
3.3 Methodology
This research aims to analyse the effect of social norms on the choice of transportation mode for commuting. The following research question will guide this research: To what extent do social influences influence transport behaviour for commuters in the Netherlands (in 2016)? The study is conducted in retrospective, allowing for correlations to be revealed.
In composing the research question, there are several elements that have played an important role. The study was aiming to research transportation mode decision-making towards a more sustainable mode. Sustainable transportation modes are not only cycling, but for example public transportation and electric vehicles such as cars may also be categorized as more sustainable choices. However, as the research question was highly dependent on the available data, this research focusses on the Netherlands. The Netherlands has a very unique infrastructure, with bicycles incorporated as one of the main modes of transportation. It was therefore found to be very interesting to study solely the car and bicycle, rather than incorporating other sustainable transportation modes. Secondly, the choice to only study commuters is related to the traffic congestion problems in the Netherlands (Krijger, 2010). Peak hours in the Netherlands are becoming more and more problematic. This research may therefore contribute to relieve traffic congestion.
The research is solely based on quantitative data, received from Mobiliteitspanel Nederland (MPN). Because this study within this field is quite specific, the research question was very dependent of the available data. The chosen dependent variable is binary. An ordinary least squares (OLS) regression would therefore not fit the data. To test correlation a binary logistic regression is conducted. Different from an OLS regression, the binary logistic regression predicts the probability of the dependent variable occurring based on the values of the independent variables. Testing the suitability of the dataset for a logistic regression is based on several assumptions. These results are presented in Table 3 and Table 4.
The effect of social norms are tested in three different models. One model tests only the effect of prestige, the other only the opinions of friends and lastly one model tests both operationalizations of social norms in one model.
Table 1 Sample collection process
The Mobility Panel Netherlands (MPN) N
Merge W4_Pdata and W4_Pdata_bijzonder using person ID 6,786
Delete observations whose most used transportation mode is not the car or bike 4,172 Delete observations who are not aged between 17 and 70 years 3,908 Delete observations without working address outside home 3,820
Exclude observations who are not on a payroll 3,302
Delete observations who are self-employed and/or unemployed 3,021 Delete observations who do not own any transportation modes (neither car/bike) 3,005 Delete observations who are not the driver of the car (car-sharing) 2,947
Merge with W4_HHdata 2,947
4. Analysis
4.1 Overview of the sample: descriptive statistics
Table 2 shows descriptive statistics of the sample. The descriptive statistics in Table 2 provide more insight into the dependent, independent and control variables. Frequencies of the categories of the variables are presented, alongside the percentages of the sample. Expectedly, the majority of the sample (around 70 percent) has the car as their most used transportation mode for their home-to-work commute. For the independent variables, there are several categories distinguished, ranging in importance and agreement with several statements. The descriptive statistics show that for the statements on the congestion problem in the Netherlands, the extent to which cycling or riding the car provides prestige, the extent to which a car shows your status in society, the statement on car use and the extent to which a car represents a personal style the majority of the sample shows a neutral opinion. Reversely, the majority of the sample has a stronger opinion on the extent to which travelling in general provides prestige and on the statement on car ownership. In both cases, the majority of the sample strongly disagrees with the statement.
Table 2 Descriptive statistics of dependent, independent and control variables
Variable Categories N Percentage of sample
Most used transportation mode for home-to-work commute
Car 2,104 71.39
Bicycle/e-bike 843 28.61
Total 2,947 100
Prestige travelling statement Very unimportant 1,338 45.40
Unimportant 755 25.62
Not important, not unimportant
544 18.46
Important 260 8.82
Very important 50 1.70
Total 2,947 100.00
Personal style Strongly disagree 295 10.01
Somewhat disagree 562 19.07
Neutral 1,099 37.29
Somewhat agree 864 29.32
Strongly agree 127 4.31
Total 2,947 100.00
Cycling prestige Strongly disagree 506 17.17
Do not agree, do not disagree 1,052 35.70 Agree 325 11.03 Strongly agree 112 3.80 No opinion 221 7.50 Total 2,947 100.00
Car prestige Strongly disagree 466 15.81
Disagree 670 22.73
Do not agree, do not disagree 993 33.70 Agree 411 13.95 Strongly agree 224 7.60 No opinion 183 6.21 Total 2,947 100.00
Status in society Strongly disagree 573 19.44
Disagree 789 26.77
Do not agree, do not disagree
1,009 34.24
Agree 502 17.03
Strongly agree 74 2.51
Total 2,947 100.00
Car ownership statement Strongly disagree 961 32.61
Somewhat disagree 560 19.00 Neutral 571 19.38 Somewhat agree 216 7.33 Strongly agree 91 3.09 No opinion 548 18.60 Total 2,947 100
Car use statement Strongly disagree 449 15.24
Somewhat disagree 852 28.91
Neutral 1,338 45.40
Somewhat agree 247 8.38
Strongly agree 61 2.07
Total 2,947 100
Congestion problem statement Strongly disagree 394 13.37
Neutral 1,569 53.24
Somewhat agree 141 4.78
Strongly agree 53 1.80
Total 2,947 100
Car ownership Yes 2,652 89.99
No 295 10.01
Total 2,947 100
Bicycle ownership Yes 2,163 73.40
No 784 26.60
Total 2,947 100
Gender Male 1,444 49.00
Female 1,503 51.00
Total 2,947 100
4.2 Assumption tests and likelihood ratio test
In order to run a logistic regression there are several assumptions that must be met. First, the dependent variable must be ordinal or binary, dependent on the type of regression that is preferred. For this study, the dependent variable is binary (either car or bike) and thus a binary logistic regression is conducted. Second, observations must be independent of each other. Third, there needs to be little to no multicollinearity among the independent variables. This is tested by means of a correlation matrix, presented in Table 3. Tabachnick, Fidell & Ullman (2007) suggest that correlations under 0.9 will meet the assumptions. The correlations in Table 3 show no findings higher than 0.9, therefore, multicollinearity will not affect the results of the logistic regression. Another means of testing multicollinearity is by assessing the pseudo R2 and the standard errors, which are both presented in the results in Table 4. When both values are very high with insignificant results, multicollinearity may be present. In neither of the cases does this case present. The fourth assumption is that the sample size must be large. The sample for this study consists of 2,947 observations; thus being large enough. The last assumption states that the independent variables are linearly related to the log odds.
Alongside the assumptions, it is necessary that the models are fitted for the logistic regression. To test this, a likelihood ratio test was conducted. The likelihood ratio test shows the best model between two nested models. The results of this test are presented in Table 4. Two test are conducted, a likelihood ratio test where model 1 is nested in model 3 and a second where model 2 is nested in model 3. The p-values for both tests show p-values smaller than the confidence level of two percent. Therefore, it means that adding predictor variables together in model 3, which counts both for model 1 and 2, results in a significant improvement of the model fit.
Table 3 Correlation matrix Variables Mo st u se d tran sp o rt m o d e S tate m en t ca r o wn ersh ip S tate m en t ca r u se State m en t co n g esti o n p ro b lem Trav ell in g g iv es p re sti g e S tate m en t p erso n al st y le Trav ell in g b y b ik e g iv es p re stig e Trav ell in g b y ca r g iv es p re stig e S tate m en t so ciety sta tu s Ge n d er Ca r o wn ersh ip Bik e o wn ersh ip Most used transport mode 1.000 Statement car ownership 0.025 1.000
Statement car use 0.006 0.191*** 1.000
Statement congestion problem 0.091*** 0.152*** 0.279*** 1.000 Travelling provides prestige -0.087*** 0.106*** 0.216*** 0.125*** 1.000 Statement personal style -0.072*** 0.095*** 0.106*** 0.099*** 0.208*** 1.000 Travelling by bike gives prestige 0.019 0.183*** 0.158*** 0.165*** 0.261*** 0.075*** 1.000 Travelling by car gives prestige -0.068 0.184*** 0.156*** 0.069*** 0.431*** 0.227*** 0.472*** 1.000 Statement society status -0.014 0.115*** 0.174*** 0.140*** 0.323*** 0.556*** 0.130*** 0.320*** 1.000 Gender 0.114 -0.023 -0.003 0.027 -0.085*** -0.118*** -0.020 -0.088*** -0.106*** 1.000 Car ownership -0.374*** -0.103*** -0.021 -0.050** 0.033 0.022 0.006 -0.005 -0.030 -0.045** 1.0000 Bike ownership 0.125*** -0.002 -0.025 -0.013 -0.083*** -0.009 -0.043* -0.033 -0.005 -0.005 -0.105*** 1.0000 Note. *p<0.05; **p<0.01; ***p<0.0001.
4.3 The effect of social norms on home-to-work commute: logistic regressions
The following section describes the results of the logistic regression on the most used transportation mode for home-to-work commute. Model 1 shows the results for the regression where the independent variables on prestige are tested. Model 2 presents the results for the regression on the second social norm; the opinions of friends/peers. Subsequently, model 3 presents the results of both social norms incorporated in one model.
As a logistic regression is conducted, the interpretation of the coefficients differs from a regular regression. Furthermore, the variables being categorical allows for further explanation. The coefficients in Table 4 present the likeliness of an event at a certain level of the predictor than at the reference level. Table 4 presents the reference level as the first row of a predictor left empty. Logically, negative coefficients indicate that the event is less likely to occur than at the reference level.
4.3.1 The influence of friends on the transportation mode from home-to-work
Model 1 presents the coefficients and odds ratios of the predictor variables on the opinions of friends. Three statements are presented in the model differing from subject. The coefficients for the statement on car ownership show negative coefficients, except for the category neutral. Interestingly, the neutral category is the only one showing a significant coefficient. The coefficients show the probabilities of having the bike as most used transportation mode for home-to-work commute compared to strongly disagreeing with the statement. Respondents who strongly disagree with the statement were little influenced by their friends to purchase a car. Thus, respondents who strongly agree with the statement are influenced by their friends to purchase a car. As the results are showing negative coefficients, the probability of having the bike as most used transportation mode for home-to-work commute is lower than the most used transportation mode being the car, disregard the opinion of your friends. However, those respondents who have a neutral opinion on their friends influencing to purchase a car have a higher probability than all other categories to mostly use the car. As described above, this is the only significant influence within this model. The odds ratios for the car ownership statement in model 1 are presented in column 2. The ratios show all positive results, meaning that among a one unit increase in the predictor variable the odds of using the bike increases. However, only the category neutral shows a significant ratio.
The coefficients for the statement on the congestion problem in the Netherlands show negative coefficients as well, however not significant. Respondents who strongly agree with the statement have friends that are of opinion that the congestion problem in the Netherlands is strongly overrated. Contrarily, respondents who strongly disagree do not in general have friends who share this opinion. The results for the variable show negative coefficients amongst all categories. Therefore, the reference level is more likely to occur: friends are in general not of the opinion that the congestion problem is overrated and mostly use the bike for home-to-work commute. The odds ratios are presented in column 2. The ratios show positive results, meaning that for example the change from strongly disagree to
slightly disagree changes the odds of the bike as most used transportation mode by 0.939. Nonetheless, neither of the coefficients or odds ratios show significant results.
Other than the previous two statements, the coefficients for the statement on car use show positive coefficients, where the slightly agree category is even significant on the 99% confidence level. The categories neutral and strongly agree both show significant results as well. People who strongly agree with the statement are of opinion that the car must only be used when it is really necessary. On the other hand, people who strongly disagree with the statement are of opinion that the car may be used more often than only in necessary cases. All coefficients show positive results, with the categories neutral, slightly agree and strongly agree being significant. The coefficients show an increasing trend from strongly disagree to strongly agree, where strongly disagree shows the highest significant coefficient. Respondents who strongly agree with the statement, compared to strongly disagreeing, have a higher probability to use the bike for their home-to-work commute. The odds ratios for model 1 are presented in column 2 in Table 4. The odds ratios for neutral, slightly agree and strongly agree show positive significant results. Meaning that a significant change in the odds occurs when changing from the reference level.
The control variables in model 1 show all three a significant result at a 99% confidence level. Gender shows a positive significant result, which means that it is more likely to use the bike as a female for home-to-work commute than males. Owning a car shows a negative significant result. Compared to the reference level, therefore, owning a car decreases the likelihood of using the bike for home-to-work commute.
4.3.2 Influence of prestige on home-to-work commute
Model 2 in Table 4 presents the coefficients of the effect of different statements that elaborate on social prestige whilst travelling or using a specific transportation mode. Model 2 test five different statements. The first statement distinguishes whether people are of opinion that travelling must give them prestige. Compared to the reference level, very unimportant, the event is less likely to occur when a change is made from the reference level to any other category, as all coefficients show negative results. Within these categories, only the category important shows a significant result. The odds ratios in column 4 show positive ratios, which means that a positive change in the odds occurs when we move away from the reference level.
The second statement in model 2 shows the extent to which people are opiniated that a car says a lot about someone’s personal taste/sense of style. Respondents who strongly disagree are of opinion that a car does not say a lot about someone’s personal taste/sense of style. The coefficients for this predictor variable all show negative results, where slightly agree is significant. The odds of using the bike significantly decrease when we move away from the reference level strongly disagree, towards slightly agree.
The third variable shows whether people are opinionated that cycling gives them prestige. The results shows significant positive coefficients for neutral, slightly agree and strongly agree. Therefore,
the odds of using the bike for home-to-work commute increase significantly when moving away from the reference level where people are of opinion that cycling does not give them prestige.
The coefficients for the fourth statement, whether driving the car gives prestige, show negative coefficients. The categories neutral and strongly agree show negative significant coefficients. Therefore, people are less likely to use the bike for home-to-work commute when they are of opinion that the car gives them prestige, compared to the reference level.
The fifth statement elaborates on whether a car shows a person’s status in society. Column 3 show insignificant positive coefficients for this statement. The odds of using the bike increase as people agree more that a car says a lot about a person’s status in society, compared to the reference level.
The control variables for model 2 show similar coefficients as in model 1. Again, all coefficients and odds ratios for the control variables are significant at a 99% confidence level. Compared to model 1, the coefficients for all three control variables slightly decreased.
- Prestige statement: “Travelling must give me prestige”
- Personal style: “A car says a lot about someone's personal taste / sense of style”
- Cycling prestige: “Cycling gives me prestige”
- Car prestige: “Travelling by car gives me prestige”
- Status in society: “A car says a lot about a person's status in society”
4.3.3 Combined model: prestige and opinions of friends
Model 3, presented in column 5 and 6, show the results of a logistic regression where both social norms are tested. Starting with the statement on car ownership, model 3 shows negative coefficients for all categories, unlike model 1. Furthermore, the significant result for the neutral category disappears when other predictor variables are added to the model. None of the categories of the statement on car ownership show significant coefficients. Other than this, the statement on car use show similar coefficients like model 1. Adding predictor variables, however, leads to less significant results for the categories neutral and slightly agree, but still remain significant. The significant coefficient for strongly agree in model 3, disappears in model 3. The coefficients for the car use statement are all positive. Therefore, a change from the reference level increases the odds of using the bike. People who have friends that are opiniated that you must only use the car when necessary are more likely to use the bike for home-to-work commute. The variable on the congestion problem in the Netherlands shows only insignificant results. Furthermore, the coefficients are all negative: people who agree that the congestion problem in the Netherlands is exaggerated are less likely to use the bike, compared to the reference level where people are opiniated that the congestion problem is not exaggerated.
The variable on prestige for travelling in general shows similar results in model 3 as in model 2. The results shows a negative significant coefficient for the category important. Therefore, people who find prestige for travelling in general important, compared to people who find this very unimportant, are more likely to use the bike for home-to-work commute. Second, the variable on the statement whether a car says a lot about someone’s personal taste/sense of style. Compared to model 2, the coefficients in
model 3 have slightly increased. The category slightly agree remains significant when adding more predictor variables. The odds for people who slightly agree that a car says a lot about someone’s personal taste/sense of style are significantly lower to use the bike. The effect of whether cycling gives prestige is measured in the variable bicycle prestige. The coefficients for this variable are positive, except for the category no opinion. Significance remains, alike model 2, for the categories neutral, slightly agree and strongly agree. People are more likely to use the bike when they are of opinion that cycling gives them prestige, compared to the reference level where people think that cycling does not give them prestige. Interestingly, people who do not have an opinion are less likely to cycle to work, compared to the reference level. The variable on whether driving the car gives prestige shows negative coefficients, where the category strongly agree is significant at a 99% confidence level. Therefore, people who strongly agree that driving a car gives them prestige are less likely to use the bike to work, than people who strongly disagree that driving a car gives them prestige. The statement on the status in society shows positive coefficients, however all are insignificant.
As the case in model 1 and 2, the control variables in model 3 again show significant coefficients and odds ratios at a 99% confidence level. Gender shows a positive significant coefficient, meaning that females are more likely to use the bike to work than males. Furthermore, owning a car shows a significant less likely probability of using the bike to commute, compared to people who do not own a car. Lastly, owning a bike increases the probability of using the bike to commute by 1.703.
Table 4 Results of logistic regression
Model 1 Model 2 Model 3
Variable B Odds ratio B Odds ratio B Odds ratio
Constant -0.036 (0.264) 0.968 (0.256) 0.466 (0.274) 1.594 (0.437) 0.300 (0.295) 1.351 (0.399) Travelling should give me prestige Very unimportant Unimportant -0.144 (0.123) 0.866 (0.106) -0.132 (0.125) 0.876 (0.109) Not important, not
unimportant -0.279 (0.149) 0.757 (0.113) -0.262 (0.152) 0.770 (0.117) Important -0.821*** (0.217) 0.440*** (0.096) -0.814*** (0.220) 0.443*** (0.097) Very important -0.174 (0.388) 0.841 (0.327) -0.200 (0.391) 0.819 (0.320) Car and personal
style Strongly disagree Slightly disagree -0.166 (0.187) 0.847 (0.158) -0.195 (0.190) 0.823 (0.157)
Neutral -0.203 (0.181) 0.816 (0.147) -0.244 (0.184) 0.784 (0.144) Slightly agree -0.527** (0.191) 0.590* (0.113) -0.565** (0.193) 0.568** (0.110) Strongly agree -0.379 (0.320) 0.684 (0.219) -0.386 (0.323) 0.680 (0.220) Bike prestige Strongly disagree Slightly disagree 0.117 (0.171) 1.124 (0.192) 0.105 (0.174) 1.110 (0.193) Neutral 0.375* (0.166) 1.455 (0.242) 0.346* (0.169) 1.414* (0.239) Slightly agree 0.826*** (0.203) 2.283* (0.464)*** 0.766*** (0.207) 2.151*** (0.444) Strongly agree 1.746*** (0.269) 5.732*** (1.541) 1.662*** (0.271) 5.268*** (1.430) No opinion -0.217 (0.267) 0.805 (0.215) -0.264 (0.270) 0.768 (0.208) Car prestige Strongly disagree Slightly disagree -0.197 (0.169) 0.822 (0.139) -0.168 (0.170) 0.845 (0.144) Neutral -0.371* (0.175) 0.690 (0.120) -0.318 (0.176) 0.728 (0.128) Slightly agree -0.339 (0.213) 0.713 (0.152) -0.286 (0.215) 0.751 (0.162) Strongly agree -1.144*** (0.274) 0.319*** (0.087) -1.066*** (0.276) 0.344*** (0.095) No opinion -0.129 (0.270) 0.879 (0.237) -0.100 (0.272) 0.905 (0.246) Status in society Strongly disagree Slightly disagree 0.035 (0.150) 1.035 (0.156) 0.040 (0.154) 1.040 (0.160) Neutral 0.081 (0.156) 1.084 (0.169) 0.053 (0.159) 1.055 (0.167) Slightly agree 0.265 (0.182) 1.303 (0.237) 0.251 (0.184) 1.285 (0.237) Strongly agree 0.650 (0.374) 1.915 (0.716) 0.584 (0.381) 1.794 (0.683) Statement car ownership Strongly disagree Slightly disagree -0.211 (0.133) 0.810 (0.108) -0.151 (0.139) 0.860 (0.120)
Neutral 0.345* (0.135) 0.708* (0.096) -0.270 (0.143) 0.763 (0.109) Slightly agree -0.286 (0.189) 0.751 (0.142) -0.125 (0.197) 0.882 (0.174) Strongly agree -0.282 (0.281) 0.754 (0.212) -0.164 (0.292) 0.849 (0.247) No opinion -0.133 (0.131) 0.875 (0.115) -0.091 (0.136) 0.913 (0.125) Statement car use
Strongly disagree Slightly disagree 0.230 (0.158) 1.259 (0.199) 0.240 (0.164) 1.271 (0.208) Neutral 0.518** (0.150) 1.678** (0.252) 0.502** (0.155) 1.652** (0.257) Slightly agree 0.716*** (0.199) 2.047*** (0.408) 0.672** (0.207) 1.958** (0.405) Strongly agree 0.832* (0.327) 2.298** (0.752) 0.606 (0.340) 1.833 (0.623) Statement congestion problem Strongly disagree Slightly disagree -0.063 (0.160) 0.939 (0.151) -0.031 (0.166) 0.969 (0.161) Neutral -0.110 (0.150) 0.896 (0.134) -0.052 (0.155) 0.949 (0.148) Slightly agree -0.164 (0.251) 0.848 (0.213) -0.109 (0.263) 0.896 (0.235) Strongly agree -0.048 (0.363) 0.953 (0.346) -0.031 (0.380) 0.969 (0.368) Gender 0.500*** (0.090) 1.648*** (0.149) 0.469*** (0.093) 1.599*** (0.148) 0.462*** (0.093) 1.587*** (0.148) Car ownership -2.522*** (0.155) 0.080*** (0.012) -2.491*** (0.158) 0.083*** (0.013) -2.512*** (0.161) 0.081*** (0.013) Bike ownership 0.578*** (0.109) 1.782*** (0.194) 0.532*** (0.111) 1.703*** (0.189) 0.536*** (0.112) 1.709*** (0.191) Likelihood ratio test Chi square: 20.23
(0.0896)
Chi square: 97.94 (0.0000)
N 2,947 2,947 2,947
Pseudo R2 0.1300 0.1520 0.1577
Note. *p<0.05; **p<0.01; ***p<0.0001. Standard errors in parentheses.
4.4 Robustness checks
As literature on transportation choice describes, decision-making is paired with many other factors than only social norms. Driving the car is dependent of whether you are in possession of a driver’s license, and whether you are cycling to work may be dependent of the distance to be travelled. Furthermore,
cycling is one of the Netherlands’ main mode of transportation, whereas in other countries this is not the case. Having a different heritage may influence your choice of using the bike, as you may not even know how to cycle. In addition, working hours might influence transportation mode choice. People might wish to travel by car, but choose the bicycle to avoid traffic during peak hours. To control whether other factors might influence the transportation mode decision, Table 5 presents the coefficients of a logistic regression in which the factors heritage, whether the person started travelling at different times and whether the person started travelling at different times when their working hours changed are incorporated.
Column 2 of table 5 presents the coefficients of the model with variables on the opinions of friends with moderators. The model shows a negative insignificant constant. Looking at the car statement variable, the only coefficient showing a significant effect is for the neutral category. This means that people who are neutrally opinionated towards the statement whether their friends are of the opinion that you should only use the car if necessary are significantly less probable to commute by bike. Compared to the model in Table 5, the coefficients and results show similar results. The control variables heritage and new working hours do not show significant results. The variables new travelling hours, however, shows a positive significant effect. Therefore, people who started travelling at a different time are significantly more likely to travel by bike. The odds ratios for model 1 in Table 5 show similar significant effects as for the coefficients. Significant and relatively large positive odds ratios are seen for the categories neutral, slightly agree and strongly agree for the statement on car use. For strong agreement, the log of odds changes by 2.295 by a change in the dependent variable – the choice of car or bike. The control variable on new travelling times is the only control variable showing a significant effect. The log of odds changes significantly by 1.143 when the travelling hours have changed. Furthermore, gender, bike ownership and car ownership also show significant odds at a 99% significance level.
Column 3 and 3 of table 5 presents the coefficients and odds ratios of the model with variables that tell something about how people are opinionated towards prestige. For the variable about whether travelling should give prestige, the only category showing significant results is ‘important’. Therefore, people who find it important that travelling should give prestige are 0.833 less likely to travel by bike. The second variable on the extent to which a car tells something about personal style shows significant effects for the category ‘slightly agree’. The coefficient for this category is negative, meaning that people who slightly agree are less likely to travel by bike. Interestingly, also people who slightly disagree are less likely to travel by bike. However, the other coefficients do not show significant results. The results on whether travelling by bike gives prestige, show significant results for the categories slightly agree and strongly agree. The results are positive and significant at a 99% confidence level. People who strongly agree that cycling gives prestige are 1.718 more likely to travel by bike. Strikingly, people who have a neutral opinion towards this statement are less likely to travel by bike, whereas people who slightly disagree show a positive coefficient. The coefficients for whether driving the car gives prestige
are all negative. Nonetheless the opinion of driving the car, you are less likely to cycle to work. However, only the category ‘strongly agree’ shows a significant negative coefficient. People who strongly agree that driving the car gives prestige are 1.090 significantly more likely to travel by car to work. The coefficients on whether the car shows your status in society show positive, but insignificant results. For all categories, people are more likely to travel by bike. For model 2 in Table 5 the control variables gender, bike ownership and car ownership continue to present highly significant coefficients. Being a woman increases the likelihood of travelling by bike by 0.508. Whereas owning a car decreases the likelihood of travelling by car with 2.503. Ownership of a bike, on the other hand, increases the likelihood of travelling by bike. The only control variable added as robustness check showing a significant results is the new travelling time. People who started travelling at other times of the day have a higher likelihood to travel by bike. The control variables of new working hours and heritage show very small negative coefficients, but insignificant.
Column 6 and 7 of Table 5 combines both operationalizations of social norms – prestige and opinions of friends – in one model with control variables as a robustness check. The constant remains insignificant in the third model. All variables showing significant results in model 1 and model 2, continue to present significant coefficients and odds ratios in the model incorporating all variables. However, whereas the category ‘neutral’ of the car ownership statement shows a significant result in model 1, it fails to present significance in model 3. This implies that when controlling for other factors in the model, the significant relationship does no longer withstand.
Table 5 Robustness checks
Model 1 Model 2 Model 3
Variable B Odds ratio B Odds ratio B Odds ratio
Constant -0.185 (0.281) 0.831 (0 .234) 0.316 (0.293) 1.371 (0.402) 0.142 (0.313) 1.152 (0.361) Travelling should give me prestige Very unimportant Unimportant -0.143 (0.124) 0.866 (0.107) -0.133 (0.125) 0.875 (0.110) Not important, not
unimportant -0.282 (0.151) 0.754 (0.114) -0.265 (0.153) 0.767 (0.118) Important -0.833*** (0.221) 0.435*** (0.096) -0.829*** (0.223) 0.436*** (0.097) Very important -0.157 (0.389) 0.854 (0.333) -0.186 (0.392) 0.831 (0.326) Car and personal
style Strongly disagree Slightly disagree -0.156 (0.188) 0.856 (0.161) -0.189 (0.192) 0.828 (0.159)
Neutral -0.202 (0.182) 0.817 (0.149) -0.247 (0.182) 0.781 (0.145) Slightly agree -0.527** (0.192) 0.591** (0.114) -0.569** (0.195) 0.566** (0.110) Strongly agree -0.376 (0.322) 0.686 (0.221) -0.392 (0.325) 0.675 (0.220) Bike prestige Strongly disagree Slightly disagree 0.108 (0.172) 1.114 (0.192) 0.096 (0.175) 1.100 (0.192) Neutral -0.328 (0.176) 1.385 (0.232) 0.297 (0.170) 1.346 (0.229) Slightly agree 0.795*** (0.205) 2.215*** (0.453) 0.735*** (0.208) 2.085*** (0.434) Strongly agree 1.718*** (0.269) 5.572*** (1.500) 1.636*** (0.272) 5.136*** (1.296) No opinion -0.263 (0.268) 0.769 (0.206) -0.312 (0.272) 0.732 (0.199) Car prestige Strongly disagree Slightly disagree -0.194 (0.170) 0.824 (0.140) -0.168 (0.171) 0.845 (0.145) Neutral -0.328 (0.176) 0.720 (0.127) -0.276 (0.177) 0.759 (0.134) Slightly agree -0.291 (0.215) 0.748 (0.160) -0.240 (0.217) 0.787 (0.171) Strongly agree -1.090*** (0.275) 0.336*** (0.093) -1.016*** (0.278) 0.362*** (0.101) No opinion -0.081 (0.271) 0.922 (0.250) -0.051 (0.273) 0.950 (0.259) Status in society Strongly disagree Slightly disagree 0.041 (0.151) 1.041 (0.158) 0.044 (0.154) 1.045 (0.161) Neutral 0.060 (0.157) 1.062 (0.166) 0.033 (0.160) 1.033 (0.165) Slightly agree 0.275 (0.183) 1.316 (0.241) 0.257 (0.185) 1.293 (0.239) Strongly agree 0.638 (0.376) 1.892 (0.712) 0.572 (0.384) 1.771 (0.680) Statement car ownership Strongly disagree Slightly disagree -0.204 (0.134) 0.815 (0.109) -0.141 (0.140) 0.868 (0.122)
Neutral -0.334* (0.137) 0.716* (0.098) -0.259 (0.145) 0.772 (0.112) Slightly agree -0.249 (0.190) 0.780 (0.148) -0.086 (0.198) 0.918 (0.182) Strongly agree -0.265 (0.285) 0.767 (0.219) -0.159 (0.296) 0.853 (0.253) No opinion -0.129 (0.132) 0.879 (0.116) -0.085 (0.137) 0.919 (0.126) Statement car use
Strongly disagree Slightly disagree 0.241 (0.159) 1.272 (0.202) 0.254 (0.165) 1.289 (0.212) Neutral 0.500*** (0.151) 1.650*** (0.250) 0.494* (0.156) 1.639 (0.256) Slightly agree 0.715*** (0.201) 2.044*** (0.411) 0.676** (0.209) 1.966 (0.410) Strongly agree 0.831* (0.328) 2.295* (0.753) 0.618 (0.341) 1.855 (0.632) Statement congestion problem Strongly disagree Slightly disagree -0.066 (0.161) 0.936 (0.151) -0.035 (0.167) 0.965 (0.161) Neutral -0.111 (0.150) 0.895 (0.135) -0.050 (0.156) 0.951 (0.149) Slightly agree -0.166 (0.255) 0.847 (0.216) -0.118 (0.267) 0.889 (0.237) Strongly agree -0.045 (0.363) 0.956 (0.347) -0.028 (0.380) 0.972 (0.369) Gender 0.536*** (0.092) 1.709*** (0.157) 0.508*** (0.094) 1.661*** (0.156) 0.498*** (0.095) 1.646*** (0.156) Car ownership -2.534*** (0.157) 0.079*** (0.012) -2.503*** (0.159) 0.082*** (0.013) -2.519*** (0.162) 0.081*** (0.013) Bike ownership 0.571*** (0.109) 1.771*** (0.194) 0.533*** (0.112) 1.703*** (0.191) 0.533*** (0.112) 1.704*** (0.191) New travelling time 0.134*
(0.053) 1.143* (0.060) 0.133* (0.054) 1.143* (0.062) 0.133* (0.054) 1.142* (0.062) New working hours -0.001
(0.001) 0.999 (0.001) -0.001 (0.001) 0.999 (0.001) -0.001 (0.001) 0.999 (0.001) Heritage Native Dutch Western migrant -0.056 (0.203) 0.946 (0.192) -0.003 (0.205) 0.997 (0.205) -0.005 (0.207) 0.995 (0.206) Non-Western migrant -0.489 (0.448) 0.613 (0.274) -0.371 (0.465) 0.690 (0.321) -0.342 (0.474) 0.711 (0.337)
Likelihood ratio test Chi square: 95.10 (0.000) Chi square: 19.07 (0.1210) N 2,918 2,918 2,918 Pseudo R2 0.1325 0.1543 0.1597
Note. *p<0.05; **p<0.01; ***p<0.0001. Standard errors in parentheses.
5. Conclusion and discussion
This thesis aims to explain why people travel to work by car or by bike with insights from behavioural economics, specifically social norms. By running a logistic regression, the relationship between the most frequently used transportation mode and the effect of social norms was assessed. Theory suggested that social norms influence day-to-day decision-making of people, including their transportation decision. In particular, the opinions of friends and the extent to which prestige is found important by individuals, shows in previous studies to have impacted transportation mode choice. This thesis aims to elaborate on these findings by studying the effect of social norms on commuters in the Netherlands by car or bicycle. This study has found that social norms can significantly influence the most frequently used transportation mode for home-to-work commute. The study finds significant results for hypothesis 1 and 2. Two operationalizations for social norms have been studied: opinions of friends and prestige. Starting off with the opinions of friends, the results of this study show that people strongly agreeing to the statement whether their friends advised them to buy a car are less likely to travel by bike. Secondly, the results show that for people who have friends that share the opinion that cars must only be used when necessary are more likely to travel by bike. These findings are at odds with hypothesis 1, which predicted that positive opinions of friends towards car usage should increase the likelihood of using the car for home-to-work commute. A third statement was tested, which shows insight in the opinion towards the congestion problem in the Netherlands. People who strongly agree with the statement share the opinion that the traffic congestion problem in the Netherlands is greatly exaggerated. The findings show that nonetheless the opinion of your friends, people are less likely to travel by bike. This finding, too, seems at odds. It would be expected that if individuals are influenced by the opinion of their friends, they would show a higher likelihood to travel by car if they would agree with the opinion and a higher likelihood to travel by bike if they would disagree with the opinion. However, the results do show significant results for people who (slightly and strongly) agree with the statement. Altogether, the results show significant proof for hypothesis 1.
The second operationalization of social norms considered whether prestige plays a role in travelling and the effect on the most frequently used transportation mode. It was expected that individuals who are of the opinion that cars express a certain social status are more likely to commute by car. Several variables were used to test the hypothesis. First, travelling in general should give me prestige. The findings show that people are significantly less likely to travel by bike if they find it important that travelling by bike should give prestige. People who find it less important that travelling
