In recent years, many institutions and governmental agencies have tried to improve the diffusion of alternative means of transportation, and an important target seems to be the diffusion of bicycle use for active commuting (ECF, 2018; Ministry of Environment, 2020). Understanding the factors that influence and determine the use of bicycles in cities could help to enhance air quality, reduce the number of automobiles in circulation, and improve the levels of physical activity in the population. Active travel is defined as methods of traveling to and from work that involve physical activity (walking and cycling; Jones and Ogilvie, 2012). According to this definition, active commuting can involve traveling only by walking or cycling, or traveling by walking or cycling in combination with motorized travel (e.g., using a combination of walking and car or train and cycling; Jones and Ogilvie, 2012). Psychology has provided a great number of useful models and tools to understand the factors that determine this behavior, especially considering the high complexity of the factors involved [for a review of the most widely used models, see Lanzini and Khan (2017)]. One model, in particular, has shown excellent fit in identifying key variables for active commuting behaviors: the Theory of Planned Behavior (TPB, Ajzen, 1991). After the dispersal of COVID-19, the number of studies on bicycle use for active commuting in the Italian context has increased (e.g., Torrisi et al., 2021); however, none of these studies have focused on the TPB framework. In a recent review, Caballero et al. (2019) observed that the model explained 25–73% of the variance in intention to use a bicycle. The model assumes that behavioral intention is determined by three fundamental factors: attitudes, subjective norms, and perceived behavioral control (PBC). Moreover, studies have indicated that the role of habits in determining the intention to use a bicycle is unclear (de Bruijn et al., 2009; Heinen et al., 2011; Rondinella et al., 2012; Muñoz Lopez et al., 2013; Caballero et al., 2014), as well as the role of financial incentives in decision-making processes regarding commute mode choice (Bamberg and Schmidt, 2000; Riggs, 2017). The present study, which was carried out in the Italian context, aims to investigate the relative importance of TPB variables, habits, and financial incentives in determining the intention to use a bicycle for active commuting.

Table 1 shows the means, standard deviations, and correlations between the investigated variables. The results showed that for descriptive norms, participants did not have many models that used a bike for daily commuting, while for prescriptive norms, a moderate level was observed. The attitude toward the behavior is slightly positive, whereas the perceived behavior control is low. With regard to habits, participants rarely used bikes before lockdown. Concerning financial incentives, the analysis revealed a low propensity for their use and low intentions to begin to use a bike for daily commuting. Regarding correlations, intention was positively correlated with all predictors considered; moreover, all predictors were positively correlated with each other except for habits and prescriptive norms, financial incentives and both descriptive norms and behavioral control.

To determine the relative importance of the variables considered in this study (descriptive norms, prescriptive norms, attitude, behavioral control, habits, and financial incentives) in predicting intention to use a bike, a dominance analysis was performed (Azen and Budescu, 2003).

Table 2 shows the 2⁶ = 64 subset models evaluated and their corresponding R², additional contributions of excluded variables (columns labeled X_i), and the average of additional contributions to all subset models of a given size (rows labeled k).

To establish complete dominance, the additional contribution of one variable must be compared with the additional contribution of another variable across all subset models. For example, to establish complete dominance between attitude and habits, you have to compare the columns of Table 2, respectively labeled X₃ and X₅. If the additional contribution of one predictor is always greater than that of the other predictor, then it can be said that one predictor completely dominates the other. Results indicate that habits (X₅) completely dominate descriptive (X₁) and prescriptive norms (X₂), attitude (X₃), and behavioral control (X₄); financial incentives (X₆) completely dominate descriptive norms (X₁), prescriptive norms (X₂), and behavioral control (X₄); attitude (X₃) completely dominates descriptive norms (X₁) and behavioral control (X₄); prescriptive norms (X₂) completely dominate descriptive norms (X₁). In all other cases, complete dominance cannot be established; indeed, the relative importance of variables changes based on the subset model used for making the comparison.

To establish conditional dominance, the average additional contribution of one predictor within each model size must be compared with the average additional contribution of another variable. For example, to establish conditional dominance between attitude and habits, you must compare the entry under columns labeled X₃ and X₅ in each k row of Table 2. If the average contribution of one predictor is greater than that of the other predictor for each model size, then it can be said that one predictor conditionally dominates the other. Results indicate that prescriptive norms (X₂) dominate behavioral control (X₄) while habits (X₅) dominate financial incentives (X₆); for the remaining case, conditional dominance still cannot be established. Indeed, the average additional contribution, within each model size, in some cases is greater for one predictor while in other cases is greater for another.

To establish general dominance, one predictor’s overall averaged additional contribution must be compared with the overall averaged additional contribution of another predictor. For example, to establish general dominance between attitude and habits, you must compare the entry under columns labeled X₃ and X₅ in the last row of Table 2 (Overall average). If the overall averaged contribution of one predictor is greater than that of the other predictor, then it can be said that one predictor generally dominates the other. Results indicate that attitude (X₃) dominates prescriptive norms (X₂) and financial incentives (X₆) while behavioral control (X₄) dominates descriptive norms (X₁).

Taken together, these results indicate that habits are the stronger predictor of intention to use a bicycle for daily commuting, followed by attitude and incentives, which have almost the same weight, while prescriptive norms, PBC, and descriptive norms are the weaker predictors.

Following Azen and Budescu (2003), to assess the generality and robustness of the obtained results, the degree to which the dominance pattern obtained in the sample is reproduced in 1,000 bootstrap samples was measured (Table 3). If the mean of dominance values (mean of D_ij) is closer to one, X_i dominates X_j; if the mean of dominance is closer to zero, X_j dominates X_i; if the mean of dominance is closer to 0.5 is not possible to establish the dominance. Another method of evaluating the generality and robustness of the results is the reproducibility, which could be intended as the confidence level; for example, if the reproducibility is 0.96, the result was replicated in 96% of bootstrap samples, then we can be 96% confident in concluding that a predictor dominates another in the population.

With regard to complete dominance, the results confirmed those obtained by previous analyses. The means of the dominance values were close to one, and the reproducibility was good. Only in the case of dominance of habits (X₅) on attitude (X₃), the dominance values (0.7830) and reproducibility (59.7%) were lower. Even in the case of conditional dominance, the results confirmed those previously obtained, although the means of dominance values and reproducibility were lower. Finally, general dominance confirmed previous results and reduced the number of indeterminacies except for the dominance of attitude (X₃) on financial incentives (X₆), which show low values of dominance (0.5100) and reproducibility (51%), indicating that in half of all samples, dominance between these two variables is undetermined.

This study aimed to establish the relative weight of habits, financial incentives, attitudes, norms, and PBC in influencing the intention of using a bicycle for daily commuting using the dominance analysis approach. The literature on this topic usually uses the traditional regression model approach (e.g., Caballero et al., 2019) instead of dominance analysis. However, dominance analysis is better than other approaches in determining the relative importance of predictors. Usually, researchers intend to explain an outcome variable from various predictors. This process can be divided into two phases: model selection and predictor comparison (Budescu, 1993; Azen et al., 2001). In the first phase, researchers identify the model and consider various predictors that better describe the outcome variables. The second phase evaluates predictor importance by comparing the contributions made by each independent variable included in the model in predicting the dependent variable. Compared to multiple regression, the dominance analysis approach can complete the second stage by clearly defining the importance of predictors. While other approaches, such as regression, infer the importance of predictors by interpreting a statistical coefficient, dominance analysis has a clear definition of importance that facilitates the identification of predictor importance: an independent variable is considered more important than another if it contributes more than the other to the prediction of the dependent variable (Azen and Budescu, 2003). Moreover, by assigning a weight to each variable, this procedure could be closer to how people decide whether or not to use the bike.

The results of our study indicated that habits are the stronger variable that influences the intention to use a bicycle for daily commuting, followed by attitude and incentives, which seem to play almost the same role. Finally, the less important variables related to intention are both types of norms and control. Concerning the role of habit, the results are similar to those of other studies carried out in different contexts (Rondinella et al., 2012; Muñoz Lopez et al., 2013). While, considering financial incentives, this variable has shown good power in predicting intentions, which is contrary to previous studies (Savelkoul and Peutz, 2017) and similar to other ones (Bamberg and Schmidt, 1998; Hunecke et al., 2001; Thøgersen, 2009; Martin et al., 2012; de Krujif et al., 2018; Moser et al., 2018; Ciccone et al., 2021). Financial incentives are nearly equivalent to the attitude toward bicycle use, thus confirming its important role as a predictor of bicycle use (Dill and Voros, 2007; Heinen et al., 2011; Milkovic and Stambuk, 2015; Frater et al., 2017; Yu et al., 2018). It is interesting to observe the relationship between financial incentives and subjective norms: in our study, in opposition to other researches (Hunecke et al., 2001; Riggs, 2017), financial incentives dominate subjective norms. It is an important result if we consider the Italian context: this means that the kind of incentive used by the government is quite efficient to determine a change in the intention, despite the absence of strong support from significant others. Moreover, it should be useful to reflect on the relation between habits and incentives: other studies have observed their interplay in particular moments as relocation, showing the strength of incentives in changing behavior (Bamberg, 2006). However, in our research, habits dominate incentives, stressing the importance of practice creation before financial incentives.

In relation to PBC, unlike other studies (Caballero et al., 2014; de Souza et al., 2014; Acheampong and Siiba, 2018; Si et al., 2020), the present research shows a very low impact of this variable in determining intention, together with subjective norms. These results may be because Italy is less “bike-friendly” than other countries. For example, in the 2019 Bicycle Cities Index (Coya, 2019) results, which analyzed the conditions for cycling in 90 cities across the world, Italian cities are ranked toward the bottom of the ranking (Milan is in 65th place and Rome is in 70th place). However, these results may also depend on the sample’s composition, which was not representative of the Italian population and was mainly composed of female participants. Obviously, the sample was affected by the self-selection, probably because the data were collected by an online questionnaire distributed over social networks, and the participation was voluntary. It would be desirable to repeat the study with a well-balanced sample and identify subgroups of populations to explore possible differences between them (Winters et al., 2017). Furthermore, regarding PBC, we do not have information about the conformation of the territory in which the participants live. Indeed, using a bike could be simpler living on plains than living on mountains or hills (see, for example, Heinen et al., 2010; Tyndall, 2020). Further research should also consider this aspect of behavioral control. Finally, the last limitation is the self-report measure used, which involves the possibility of response bias.

However, despite these limitations, this study suggests that bicycle use is affected by various factors that need integration, as suggested in other contexts (e.g., Savan et al., 2017) and provides some important indications for promoting bicycle use by acting on variables that showed the greatest predictive power. Both habits and attitudes underly the importance of establishing a bike use culture in cities. Thus, both habits and attitudes should be supported through communication campaigns focusing on positive aspects of using cycles for daily commuting (e.g., time savings and/or health benefits). Regarding financial incentives, they seem to be an interesting method of improving the use of bicycles as a means of transportation. It would be advisable to offer financial incentives regularly and not as an exception; for example, the incentives given by the Italian government led to a 17% increase in bicycle sales in 2020 compared to the previous year (ANCMA, 2020). The role of incentives leads us to an important consideration: financial incentives together with attitude could represent an important method of promoting the shift toward bicycles as a means of transportation. The Italian government has reserved many resources to increase bicycle purchases to create and maintain healthy mobility habits.

In this study, the dominance analysis was used to establish the relative importance of the predictors; however, similar results could be obtained using the relative weight analysis (Johnson, 2000). To calculate the relative importance of the predictors, dominance analysis performs a series of regressions; on the other hand, the relative weight analysis transforms the independent variables into a new set of independent variables that are orthogonal to each other but still correlated with the original ones. Despite the differences in statistical rationale, both analyses arrive at nearly identical results (Johnson, 2000; LeBreton et al., 2004).

Future research could also provide insights on behavior maintenance in individuals who used financial incentives to buy a bicycle. Furthermore, given the limited effects of the norms and perceived control, probably due to the Italian context and the sample, the study should be repeated considering other samples and more “bike-friendly” contexts.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Internal Ethic Review Board (IERB), Department of Educational Sciences (Psychology Section), University of Catania. The patients/participants providedtheir written informed consent to participate in this study.

VB: conceptualization, methodology, data collection, and writing—original draft. ZH: supervision, methodology, data analysis, and writing—original draft. FS and CD: writing—reviewing and editing. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.