User Perspectives on Distance- and Time-Based Road Tolling Schemes: European Case Study

The anonymous questionnaire was designed with a view toward recruiting the desired respondent type (i.e., person from North Macedonia with both DB and TB scheme experience) and minimizing the participant dropout rate. Thus, the landing page of the questionnaire clearly stated that the questionnaire focused on the aforementioned respondent type, included basic definitions of DB and TB pricing concepts, and contained the initial control questions. A DB road pricing system was introduced to the respondents through the description of the ETC system, while a TB road pricing system was introduced through a description of vignettes. The first control question asked the respondent to select the ETC system used previously, with the following options: ETC with barriers, ETC without barriers, and have not used ETC. The second control question asked the respondent to answer whether she has used a vignette system and, in the case of an affirmative answer, to estimate the number of kilometers traveled and the time period when a vignette was used. After the landing page, if the respondent had answered affirmatively to both ETC and vignette questions, the respondent proceeded to the questionnaire proper, which had four parts. More information about sampling and filtering is in the next section, “Data Collection Process and Sample Characteristics.”

The first part of the questionnaire was focused on socioeconomic characteristics of the respondents as well as the patterns of their motorway usage. Questions were about gender (male/female/other), age group (18–29, 30–39, 40–49, 50–59, $60+$), education level (elementary school, high school, vocational school, university), personal average monthly income in euros ($<250$, 250–500, 500–750, 750–1,000, 1,000–1,250, $>\mathrm{1,250}$), frequency of motorway usage (everyday, several times per week, several times per month, several times per year), average monthly distance traveled on motorway in kilometers ($<30$, 30–60, 60–90, 90–120, 120–150, $>150$), the most common day of the week of motorway usage (workday, weekend) and the most common reason for motorway usage (commuting to work/school, leisure and shopping, work-related, other). The income categories were selected in line with standard income categories used in North Macedonia, while distance traveled categories were selected based on estimates from regional travel survey. This part of the questionnaire had an additional filter question that asked about the most frequently used vehicle types on motorways (motorcycle, passenger car, light-duty vehicle, heavy-duty vehicle, bus, semitrailer truck). This question, together with the question about the total amount of kilometers traveled, was used to filter out respondents who were not predominantly passenger car users (see the section “Data Collection Process and Sample Characteristics”).

The second part of the questionnaire focused on the WTP and MAP questions, similarly to previous studies (e.g., Odeck and Kjerkreit 2010; Chen and Wen 2014; Glavić et al. 2017; Gomez et al. 2017; Jou et al. 2012; Milenković et al. 2019). WTP and MAP questions were first asked in relation to DB pricing, and respondents were required to answer the question of how much they were willing to pay (WTP) and at what price they would stop using the motorway (MAP). These two question types had two versions, with one asking for a toll estimate taking into account only travel time savings in comparison to using a two-lane road, and with the other asking for a toll estimate taking into account all other benefits from using the motorway compared to a two-lane road, such as a higher level of service and higher level of traffic safety. For the estimate travel time saved, the value was described as 30–40 min per 100 km traveled. This value was determined based on the data from automatic traffic counters regarding the average travel speed of passenger cars on two-lane roads and motorways. For example, passenger cars using a national motorway network tend to have average speeds of approximately $110\mathrm{km}/\mathrm{h}$, while those using a secondary road network drove around $50\mathrm{km}/\mathrm{h}$. Thus, 30–40 min of travel time savings is a realistic difference for a trip of 100 km switching from a secondary road to the motorway network. These WTP and MAP questions were in line with the contingent valuation method for stated preferences, where respondents are asked to select one value among the option prices ranging from €1 to €10, in €1 increments. The values were selected based on a range of values around the current toll price of three euros, although skewed for higher values, in order to explore a wider set of values, while also reducing the cognitive load for the respondent if the respondent had to provide a value herself. This exploratory approach, as used in some of the previous road pricing case studies (Glavić et al. 2017; Jou et al. 2012; Milenković et al. 2019), was necessary given the dearth of research on road pricing in southeastern Europe, which makes it hard to conduct stated-choice experiments. Moreover, the stated-choice approach was not used because it would significantly increase the time and effort needed to complete the questionnaire. Finally, the objective of this study was not to derive the road price value for either direct policy implementation or further use in transport modeling, so the stated-choice approach was not considered necessary.

Similarly, after the WTP and MAP questions for the DB pricing scheme, the respondent was asked to answer WTP and MAP questions related to the TB pricing scheme. Respondents were asked to state how much would they would pay in euros for weekly ($<3$, 3–6, 6–9, $>12$), 10–day ($<5$, 5–8, 8–11, 11–14, $>14$), monthly (9–12, 12–15, 15–18, 18–21, 21–24, $>24$), or annual (30–50, 51–70, 71–90, 91–110, $>110$) vignettes. The same values were used on the MAP question about the vignette price at which the respondent would stop driving on motorways. The range of values for vignette prices were developed taking into account actual vignette prices from other southeastern European countries, namely, Bulgaria, Romania, Hungary, Czechia, Slovakia, and Slovenia (Table 6). An additional reason for asking about WTP and MAP values was to filter out respondents from the sample. That is to say, MAP values had to be equal to or greater than the WTP values, because this conformed with the expectations of rational respondent choice (see the section “Data Collection Process and Sample Characteristics”).

The third part of the questionnaire was related to users’ attitudes toward the characteristics of the toll system, as well as their explicit preferences for one of the two pricing concepts. The respondents were first asked to evaluate the importance of the pricing system characteristics, using a five-point Likert scale, with one being the least important characteristic and five the most important. Pricing technology characteristics were grouped into four categories to minimize the number of questions: technical, traffic, social, and environmental characteristics. Technical characteristics pertain to (1) the interoperability of the system for use in other countries; and (2) flexibility in using the pricing device for other purposes, such as parking or navigation. Traffic characteristics pertain to (1) toll collection without frequent or excessive stopping, (2) traveling at a desired speed and without losing travel time, and (3) safe driving conditions in connection with acceleration/deceleration and changing lanes. Social characteristics pertain to (1) easy procedure for buying and recharging the toll collection device, (2) the cost of the toll collection device, (3) fairness of pricing, and (4) the possibility of fraud or misuse by the toll collection party. Finally, environmental characteristics pertain to negative impacts, such as noise or air pollution or visual degradation of the environment due to road pricing devices. The last question of this questionnaire was a binary preference choice between DB and TB road pricing schemes and technology.

The data collection process was defined to shorten the time needed for completing the survey, aiming for a maximum of 10 min, owing to the expected low willingness to participate in these studies by the general public. To acquire a statistically significant number of respondents’ questionnaires, data collection was based upon two collection methods: an online survey and field surveys. In the first phase, an online questionnaire was sent to various companies, organizations of the employed, unemployed, or retired in order to expand the representativeness of the sample. In the second phase of data collection, a field survey was used to increase the sample size beyond the minimum threshold. This task involved the distribution of a paper-based questionnaire to respondents. A total of 327 questionnaires were collected where respondents had experience with DB and TB schemes and were predominantly passenger car drivers. Of the 327 total responses collected, 251 (76.8%) were collected online and 76 (23.2%) in the field. These 327 questionnaires were validated for respondents who had had experience with both DB and TB pricing schemes and further filtered for consistency of the respondents’ answers using a comparison of WTP and MAP values. That is, the consistency check involved evaluating whether MAP values were equal to or greater than WTP values. After this process, the final sample included a total number of 284 valid questionnaires. Out of 284 respondents included in the final sample, 216 (76.1%) were collected online and 68 (23.9%) in the field.

The final sample size of 284 respondents adequately represented the anticipated motorway passenger car driving population with respect to age, education, and income, as well as frequency of motorway usage. The total sample had 189 (66.5%) male and 95 female (33.5%) participants. The largest part of the sample consisted of motorway users aged 30–39 (23%), 40–49 (27%), and 50–59 (21%), but the sample also included the youngest 18–29 (19.4%), as well as the oldest population 60+ (9.5%). In addition, 18.7% of sample respondents earned a monthly income of less than €250, 40.5% earned between €250 and €500, 22.9% earned between €500 and €750, 9.5% earned between €750 and €1,000, and finally 8.5% earned monthly income greater than €1,000. In comparison, the motorization rate in North Macedonia from 2017 was 194 passenger cars per thousand inhabitants (Marinković 2019). The inhabitants aged 25–54 years make up the largest share of the population (44.3%), while the median age in this country is 37.8 years (Worldometers 2019). The average monthly net wage paid per employee in 2019 was €417 (Republic of Macedonia State Statistical Office 2019). We have also tested the differences between the responses of the field survey respondents and online questionnaire respondents (Table 1). The use of ANOVA determined that the preferences of the field survey respondents and online questionnaire respondents did not differ statistically significantly (for WTP values $p=0.98$, and for MAP values $p=0.895$). Consequently, this parameter was not included in SEM analysis, and two samples were merged into one.

Considering the aforementioned sample size, we must highlight that several previous surveys that examined users’ attitudes and preferences regarding road pricing schemes were based on samples of similar or even smaller sizes. For example, sample sizes included 211 (Farrell and Saleh 2005), 114 (Di Ciommo et al. 2013), 206 (Grisolía et al. 2015), 121 (Politis et al. 2020), and 155 respondents (Francke and Kaniok 2013). In the analyses of transportation system participants’ behaviors and attitudes, SEM analysis was used based on the sample size, which corresponds to the sample size in this research (further details about SEM can be found in the “DB and TB Analysis Framework” section). For example, Şimşekoğlu and Lajunen (2008) aimed to explain self-reported seat belt use among front seat passengers on rural and urban roads using SEM analysis and relying on a sample of 277 respondents. Xu et al. (2010) analyzed the determinants of travel information use using SEM analysis and a sample of 247 respondents. Kline et al. (1998a, b) proposed that a minimum sample size of 200 was needed to reduce bias to an acceptable level for any type of SEM analysis. Although there is little consensus on the recommended sample size for SEM (Sivo et al. 2006), Hoelter (1983), and Garver and Mentzer (1999) also proposed a critical sample size of 200. Thus, we have adopted this rule of thumb, where 200 is understood as the minimum threshold to provide a sufficiently statistically significant sample for adequate analysis. Also, the sample size of this research is appropriate for ANOVA (Ross and Willson 2017) and for binary logistic regression (van Smeden et al. 2019).

To formulate the analysis framework, the normality of distribution of the sample data was tested by visual evaluation of the histogram shape and by the Kolmogorov-Smirnov test. Since the distributions of all continuous variables (WTP and MAP values) did not statistically significantly deviate from the normal distribution (Table 2), and considering that one of the aims of this study was to explore the factors affecting user preferences related to TB and DB road pricing, the overall analysis framework consists of the following three components:

SEM was selected as the analysis method for understanding behavioral intentions and relationships among the variables impacting behavior (Bamberg and Schmidt 2001; Jakobsson et al. 2000; Chen et al. 2007; Kim et al. 2013). One of SEM’s strengths is its flexibility, which permits examination of several variables and their interrelationships simultaneously, the use of various types of data (e.g., categorical, dimensional, censored, count variables), and comparisons across alternative models (Hoe 2008; Wolf et al. 2013). Within this research, SEM coefficients were used to express the strength and the sign of causal paths between predictors. SEM testing used chi-square (${\chi }^2$), df, $p$-value, root-mean-square error of approximation (RMSEA), comparative fit index (CFI), Tucker–Lewis index (TLI), and normed-fit index (NFI), excluding nonsignificant variables and nonsignificant coefficients. Neither of the initial SEM models for DB and TB schemes, including all the variables for MAP DB and TB, had a good fit to the data, with ${\chi }^2=367.179$, $\mathrm{df}=110$, $p<0.000$, $\mathrm{RMSEA}=0.091$, $\mathrm{CFI}=0.913$, $\mathrm{TLI}=0.880$, $\mathrm{NFI}=0.882$, and ${\chi }^2=427.388$, $\mathrm{df}=110$, $p<0.000$, $\mathrm{RMSEA}=0.101$, $\mathrm{CFI}=0.892$, $\mathrm{TLI}=0.850$, and $\mathrm{NFI}=0.862$, respectively. When evaluating SEM results in an iterative fashion to obtain a final model, we followed best practice recommendations in the literature, highlighting, for example, that the ${\chi }^2$ test should be nonsignificant at $p\ge 0.05$, $\mathrm{RMSEA}\le 0.06$, $\mathrm{NFI}\ge 0.95$, $\mathrm{CFI}\ge 0.95$, and standardized root-mean-square residual $(\mathrm{SRMR})\le 0.08$ (Hu and Bentler 1998, 1999). The most common challenge identified in earlier research, and in our case as well, is that the ${\chi }^2$ test is significant, and one or more of the four indexes are unacceptable under the proposed cut-offs (Bagozzi 2010). Thus, Browne and Cudeck (1993) suggest that a value of about 0.08 or less for the RMSEA would indicate a well-fitting model. Thus, we followed a recommendation that the proposed cut-off values are too conservative under certain conditions, especially if research is geared toward exploratory analysis (Bagozzi 2010).

In an iterative process, two final SEM models were developed, one with the DB MAP as the dependent variable and the other with the TB MAP as the dependent variable. In both SEM models the predictor patterns consists of the most common day of motorway usage and purpose of motorway usage. In addition, the latent variables that refer to users’ attitudes toward the importance of toll-collection system characteristics are technical characteristics, traffic characteristics, and social characteristics, in line with the categories in the last part of the questionnaire. Environmental characteristics was another variable considered for SEM based on the questionnaire, but neither for DB nor TB pricing model were they statistically significant for inclusion in the final formulation. In addition to these latent variables, both models contain predictors referring to the monthly income of respondents and their frequency of motorway usage. The DB model contains a variable related to the ETC experience, while the TB model contains the variable vignette experience.

The study confirms an assumption of the difference between WTP and MAP values for both TB and DB road pricing, where MAP values are higher than WTP (Hanemann 1991; Small 2012; Glavic et al. 2017). Although previous research informs us that users can have a hard time understanding the costs and benefits of road pricing (Grieco and Jones 1994; Small 2012), WTP and MAP values are higher when users are asked to account not only for travel-time savings but also other advantages of using motorways. The obtained WTP and MAP values are in accordance with the values of DB and TB prices in southeastern Europe, while they are lower from prices in central Europe, where GDP per capita is several times higher. The key factors affecting the MAP values for DB and TB road pricing were users’ income and frequency of motorway usage, with those with higher income being more willing to pay and more frequent users being less willing to pay, similar to conclusions from Yusuf et al. (2014). Contrary to some previous studies that concluded that income does not influence WTP or acceptability (Bueno et al. 2017; Gomez et al. 2017; Jaensirisak et al. 2005), one must take into account the fact that users in North Macedonia might be poor in both time and money. Thus, income can have a significant effect on the very capacity to afford road pricing for certain social groups. In contrast, those groups with higher income might also have a negative relation to the intention to reduce car use (Jakobsson et al. 2000), thereby leading to a higher threshold for refraining from motorway use. In addition, more frequent motorway usage over a certain period results in greater expenses to the user, so it leads to lower MAP values. In the case of both SEM models, it was determined that the latent variable pattern had a statistically significant impact on MAP values. It was discovered that those who used motorways on working days and for work/school purposes had a higher MAP value than those who used motorways on weekends and for other purposes. These results are also in accordance with several previous studies (Jakobsson et al. 2000; Hensher and Goodwin 2004; Zmud and Arce 2008; Glavić et al. 2017). Unlike in Jou et al. (2012), no effect of trip distances on WTP was identified in this study, which could be associated with a rather small size of the national motorway network in North Macedonia.

When all respondents are taken in the account, DB road pricing was generally perceived as a more preferable option than TB road pricing. However, further analysis of user attitudes showed that user preferences for DB and TB road pricing depended on the frequency of motorway usage. Daily users preferred TB road pricing, while monthly and annual users preferred DB road pricing. In other words, comparative analysis of the results showed that users recognize the advantages of one or the other pricing scheme in a way that corresponded to their user group. Comparison of the normalized DB and TB MAP values also supports this user attitude. When converting TB MAP values to a comparable unit as DB MAP, the result was that DB road pricing was 2.14 times cheaper for monthly and 2.29 times cheaper for annual users, while TB road pricing was 1.49 cheaper for daily users.

The obtained results are logical considering the fact that daily motorway users travel significantly more kilometers than occasional travelers, as previously concluded by Raub et al. (2013). Therefore, if they are able to buy an annual vignette, they will save more over paying for actual annually traveled kilometers. In the case of monthly and annual users, the situation is different, because they usually pay more in the case of TB road pricing than they would pay based on actual distance traveled. These results can be contrasted with a hypothetical ratio between DB and TB road pricing of $3:2$, with the assumption that drivers take more risks when they are being charged for the use of road space based on time (Bonsall and Palmer 1997). Furthermore, we must take into account that there is some consensus in the literature that the more complex a pricing scheme is perceived to be, the less acceptable it is likely to be (Bonsall et al. 2007; Bueno et al. 2017; Vrtic et al. 2007; Zmud and Arce 2008). At the core of user perspective on the positive relationship between pricing scheme complexity and acceptability might be a high cognitive load required for its evaluation (Francke and Kaniok 2013). For example, this evaluation challenge might manifest as a flat rate bias, where one is trying to avoid and insure from future unanticipated costs, even if pay-per-use might be cheaper in total.

The analysis also showed that users’ MAP value depended on the experience of users with DB and TB concepts and the technology of road pricing. When it comes to the DB model, it was found that experience using ETC systems, which are typical representatives of the DB concept, had a positive impact on users’ attitudes toward the significance of the traffic, technical, and social characteristics of the system. In other words, the more familiar users are with the ETC system, the more significant they find the aforementioned characteristics, which is similar to the general findings of Chen et al. (2007) and Heras-Molina et al. (2019). When it comes to the TB model, it was found that the experience of using TB systems had a positive, statistically significant impact on the importance of traffic and social characteristics, but not of technical characteristics. This finding means that the more familiar users are with TB road pricing, the more significant they find traffic and social characteristics, while the impact on technical criteria was not statistically significant. The obtained result is logical keeping in mind the fact that technical characteristics referring to interoperability and flexibility are primarily related to the ETC system and not to vignettes. Finally, it was determined for the DB model that if technical and traffic characteristics were more significant for users, the users’ MAP values were also relatively higher. When it comes to the TB model, only the latent variable referring to traffic characteristics had a positive, statistically significant impact on MAP values. Familiarity with technology and the consequent evaluation of the importance of its characteristics is interestingly also related to higher income and more frequent usage, as well as to higher willingness to pay. This point is related to the previously raised point on learning about benefits and costs associated with different pricing concepts in the context of existing road pricing schemes. However, it is important to highlight that, contrary to some previous research where environmental awareness was related to acceptability (Jaensirisak et al. 2005), such a significant causation was not identified in this research. This effect might be due to specific norms in North Macedonia that might differ from those in Western Europe.

In the context of multiple, and even potentially conflicting, policy objectives in the diverse European context, bridging perspectives between pricing system planners and managers on the one hand and road users on the other hand can be essential for developing policy processes. As stated previously, EU-level policy favors the DB concept because it is related to a user-pays principle, where users pay for kilometers traveled (Raub et al. 2013). In contrast, a criticism of TB schemes is that a flat rate decouples purchase from consumption, reducing the perceived cost of activity, thereby failing to have significant effects on road use demand. Besides changes in the total demand, both DB and TB pricing schemes with pricing differentiation can somewhat affect decisions to use other modes or passenger vehicles without an internal combustion engine. However, such decisions are constrained in terms of alternative modes, such as railways and low purchasing power, both of which apply to North Macedonia.

The question of interoperability across Europe for both TB and DB pricing schemes remains open because it depends on both further technological and policy developments, with the latter related to funding and financing policy. DB-based ETC charging technologies require the installation and use of onboard units that are not always accepted by all users (Chen et al. 2007; Heras-Molina et al. 2019), although ETC devices could be used for other purposes (e.g., paying for parking, fuel- or cordon-based congestion pricing, navigation, vehicle theft prevention). Users can also have higher concerns regarding privacy and data security with ETC devices. In the case of sticker-based vignettes attached to the windshield, some users might perceive this as a visual degradation of the vehicle.

From an infrastructure management perspective, the use of DB pricing requires additional investment in roadside equipment and might lead to congestion at tolling points if vehicles are required to slow down (Glavić et al. 2017). In comparison, a TB-based vignette system can be introduced relatively quickly but might still entail operating costs for enforcement (e.g., automatic number plate recognition). Finally, in the case of motorways built on a public-private partnership model, implementing either a TB or a DB scheme should be evaluated for the effective collection and distribution of tolls to avoid the previously mentioned problem of “rat running” and associated negative feedback loops.

Although the EU policy leans toward a DB scheme, we must recognize that all stakeholders face a multitude of equity dilemmas related to a multitude of pricing schemes, technologies, and objectives. Thus, our conclusion is that there is a clear need for greater understanding of different user perspectives, which might help in reducing regressively distributed effects and achieving wider sociopolitical acceptance. In particular, the findings of this study highlight the need for understanding different, especially low-income and low-use-frequency, user perspectives on both road pricing schemes and pricing technology. Considering the diverse needs of users, an immediate policy question arises as to whether to develop a hybrid road pricing scheme, where daily users could pay using a TB scheme while less frequent users would pay using a DB scheme. The concept of a hybrid road pricing scheme has been discussed, with the travel distance as opposed to the frequency as differentiating parameter (Jou et al. 2012). This hybrid road pricing model could hold promise in terms of addressing the trade-off between road managers’ desire to maximize revenue and road users’ desire to maximize motorway usage at an acceptable price. In the context of time- and money-poor users, this scheme would potentially avoid “rat running” problems described earlier. With the advancement of road pricing technologies, technical limitations pose no challenge for introducing hybrid pricing. However, besides the question of technical feasibility, planning road pricing schemes raises a larger issue of user involvement in the development of policymaking processes and institutional capacity building. A special emphasis in such development should be on transparent informing and deliberation about the intentions and revenue use from tolling (Bueno et al. 2017; Odeck and Kjerkreit 2010; Ison 2017; Langmyhr 1997; Glavic et al. 2017). Even if previous research often concluded that users prefer less complex pricing schemes, participatory solutions achieved through deliberation could seek not just the simplest but also the most desirable and feasible scheme, which would be able to optimally deal with trade-offs between multiple policy and user objectives. Ultimately, the question of hybrid pricing is something that EU-level policymaking must consider, especially when it comes to transition countries, such as North Macedonia.

One of the potential limitations of this research is sample size. This limitation stems from the specific requirement that respondents need to satisfy, which restricts the number of eligible respondents. Future research needs to try to increase sample size in a quantitative and qualitative manner. More qualitative methods could be added to the analysis framework to explain some of the missing factors, such as environmental awareness, because these might have deeper cultural meanings related to the car as a symbol of success and independence in transition countries. In addition, although a significant part of the variability of the dependent variable was explained in the developed models (in the case of MAP DB it was 61.1%, while in the case of MAP TB it was 69.7%), a specific percentage of the dependent variable’s variability (MAP DB and MAP TB) remained unexplained. Therefore, future studies should examine additional factors that might have an impact on these values, such as the relative level of service on secondary road networks, revenue redistribution schemes, or vehicle occupancy from the perspective of flat-rate bias and mental accounting. Although we can anticipate an effect similar to that in previous research on this topic, such as that cost will decrease but in a nonlinear manner in the case of shared vehicles (e.g., Ho et al. 2016), this topic would also pertain to sharing and gender norms in North Macedonia. Similarly, the relation between motorway pricing and mobility as service schemes should also be further investigated (Liimatainen and Mladenović 2018). In relation to further reflection on the proposed cut-off values for SEM, future research could expand upon factors that would further explain the variance of MAP values.

From the standpoint of governance studies, and given the ongoing EU-level policy processes, questionnaire studies in the future should span more European countries. We can especially highlight the lack of research in countries currently having a TB pricing scheme. These kinds of pan-European questionnaires should also include questions about pricing fairness for occasional motorway users across national borders. Given the amount of diverse literature on the topic, an essential step in further development is a wide-scale systematic literature review and meta-analysis of factors and data collection methods. Further deepening of the relation between user perspectives and policymaking would benefit from the development of decision-support frameworks (e.g., Milenković et al. 2018). If road pricing research is to contribute to further institutional capacity building, there is a need for in-depth comparative analysis of road pricing policy styles across Europe. As we have seen in this case study, domino theory is not adequate to explain further governance challenges since most countries have adopted some road pricing scheme. Metaphorically speaking, the dominos have fallen, but the game is still on, and many decisions remain to be made. These further decisions about the transition of transportation systems must account for a multitude of governance contexts across Europe and the associated need to develop mechanisms of deliberative democracy in southeastern Europe.