Extending the theory of planned behavior to predict and nudge toward the subscription to a public transport ticket

Motor car use in urban areas is connected to numerous ramifications for the environment and human health, including greenhouse gas emissions, fine-particle and noise pollution. Cars stuck in traffic, e.g. during rush hours, emit even more of these damaging substances, and simultaneously decrease the viability of public transport [47]. One approach to overcoming these issues is achieving a modal shift, i.e., convincing private car users to switch to more sustainable means of transport, such as public transport or bicycles [32]. Strategies aimed at achieving this modal shift typically follow the push-and-pull approach: They either try to increase the attractiveness of more sustainable travel or decrease the attractiveness of car use [42, 45]. As sustainability concerns are on the rise, advertising for environmentally friendly travel is increasingly becoming part of transport marketing strategies as well. But, within the frame of a liberal society, changing behavior, especially habituated behavior like car use, is an ambitious and difficult task [7].

To be able to change behavior, it is important to understand what drives said behavior. One of the most widely used and accepted model to explain behavior is the theory of planned behavior (TPB) [1], which uses attitudes, subjective norm and perceived behavioral control as predictors. A previously unexplored attempt to achieve a behavioral change, i.e. the decision to switch from private car to public transport, is addressing the predictors of this theory using nudges [44]. The concept of nudging [44], developed in the field of behavioral economics, has proven effective in incentivizing voluntary behavior change across numerous fields. Nudges are interventions that support or direct the choices that people make (e.g. healthy, sustainable, etc.) without restricting freedom of choice. Nudging is an umbrella term for numerous interventions, and as such, its definition suffers from lower levels of granularity, and a coherent “nudging theory” still needs to be developed [35]. A step toward resolving this could be the integration of nudges into the theory of planned behavior.

First, this study investigates if the theory of planned behavior extended by environmental concern is able to predict the decision to subscribe to a public transport ticket. Second, this study aims to test if two nudges (a default and a social nudge) aimed at enticing employees to purchase a public transport ticket could increase subscription numbers. Finally, this study analyses if the nudges can be integrated into the theory of planned behavior.

As mentioned above, the theory of planned behavior by Ajzen [1] is one of the most widely used models to predict behavior is the theory of planned behavior (TPB) [1]. In the field of environmental psychology, studies found that the predictors of the TPB affect the intention of car use as well as the decision to use public transportation or other transport means instead of the car [8, 9, 20, 23]. Donald et al. [18] argue that the prediction of mode choice can be enhanced by environmental concern, and found that it has an indirect effect on car use. In an extended version of the theory of planned behavior [36], environmental concern [19] functions as a fourth predictor of behavioral intention as well as of the other three predictors (Fig. 1).

Attitude (AT) is defined as the extent to which a person has a favorable or unfavorable evaluation of the behavior. Subjective norm (SN) is defined as the perceived social pressure to perform or refrain from the behavior. Two kinds of social norms can be differentiated [2]: injunctive norms refer to what others (dis)approve of, while descriptive norms refer to the actual behavior of others. Perceived behavioral control (PBC) is defined as the perceived ease or difficulty of performing the behavior.

Environmental concern (EC) scales the awareness of environmental issues and the personal effort to contribute to their solution [19]. It is targeted specifically at current environmental issues regarding the transportation sector, based on the assumption that “attitudes toward specific environmental topics are ultimately reflections of a single, broad environmental attitude —what is sometimes referred to as environmental concern” [16, p. 2], [19]. Donald et al. [18] argue that the prediction of mode choice can be enhanced by environmental concern, as it provides these additional beliefs beyond the three classic TPB constructs, and they indeed found that EC has an indirect effect on car use. We thus adapted the model from Paul et al. [36], who too, found environmental concern to be of predictive value (regarding green consumption), and included environmental concern in the model used in this study. However, there is an ongoing discussion on the proper measurement of environmental attitudes in the literature. Harland et al. [23], for example, use items that are phrased very similarly to our conceptualization of environmental concern (e.g. “I am worried about the condition of the environment”), for their measurement of environmental involvement. This topic will be resumed in the discussion of this paper.

Nudges [44] are interventions that direct the choices that people make (e.g. healthy, sustainable, etc.) without restricting freedom of choice. Examples include design interventions (like spatial arrangements that place healthy food in cafeterias or supermarkets at eye-level) or default settings with opt-out options (like being an organ donor by default vs. having to sign up as an organ donor) [29, 37].

The relevance of nudges with regard to sustainable transportation derives from travelers not being a rational homo economicus: Cost–benefit calculations alone do not determine which transportation mode travelers choose because their behavior is subject to limited cognitive resources and bounded rationality. Nudging travelers can “help them to make better decisions for themselves, to improve the performance of the overall transport system, and to reduce some of the external costs (economic, environmental, societal) associated with choices made by individual travelers” [6], p. 2. Among the most promising nudges are default settings and social nudges. While often addressed on a theoretical basis, according to Byerly et al. [13] both have been used rarely as interventions to change mobility behavior.

The effectiveness of some nudges is explained with generic theories (e.g. dual process theory [30]; reflective-impulsive model [41]), but a coherent “nudging theory” still needs to be developed [35]. This paper tries to connect the two nudges mentioned above to the theory of planned behavior, thus hoping to contribute to the development of a more coherent nudging theory. This derives from the assumption that setting a default is expected to simplify the perceived ability to perform the behavior, and that adding information of the social norm acts as a manipulation of the predictor subjective norm in the extended TPB.

The nudging concentrates on the purchasing process; consequently, the model targets the decision for a public transport ticket, and not the intention to use it. To be consistent, this also includes the PBC predictor, but we recognize that PBC is, perhaps even more than other predictors, also strongly related to the use of public transport. This paper focuses on the specific case of the Job ticket offered in a German city (names blinded for peer review) (ca. 550,000 inhabitants) by the local transport association in collaboration with the local university (ca. 8,000 employees). The Job ticket is a price-reduced monthly subscription to the local public transport and offers several advantages over a comparable monthly subscription ticket, e.g. by being 20% cheaper and usable as a family ticket. In 2018, only 16% of university employees used the Job ticket to ride public transport [48]. There is room for improvement, defined as higher subscription rates.

The target population contained any adult either working at the university or likely to soon start working there (i.e. Master students). An invitation to the experiment’s web link appeared in student newsletters and was emailed via the administrative board. Participants were unaware of the aim of the study and were led to believe that it focused on the feasibility of digital working contracts. 373 people participated in the study (58.1% female; age M = 34.47, SD = 10.10), out of which 161 participants indicated that they already owned the Job ticket or a similar subscription ticket. To be able to identify the effectiveness of the nudges on those that did not yet possess such a ticket, we separately analyzed the subsample of N = 212 (51.9% female; age M = 38.01, SD = 9.64). For more descriptive data, see Tables 2 and 3 in the analysis and results section.

Participants were randomly and unknowingly assigned to one of four between-subject conditions. Following the data security information, they were asked to imagine that they had just received their working contract and to click through it as if they are just about to start working at the main campus (which offers relatively good public transport connections). After a few unrelated standard contract paragraphs to increase realism, participants encountered the Job ticket paragraph. The ticket information was identical to the actual Job ticket arrangements. Depending on the condition, this page either contained both the default and social nudge (condition DSN), only the default (condition D), only the social nudge (condition SN), or no nudge at all (control).

In conditions in which the default nudge was active, the information about the Job ticket came with the notice that the participant, as a future employee, was automatically subscribed to it. If he/she wished to unsubscribe, he/she had to un-check the preselected “Yes, I want to subscribe to the Job ticket” and instead select the “No, I do not want to subscribe”. In conditions in which the default was not active, this information stated that, if the participant wished, he/she could subscribe to the Job ticket by checking the respective button. In conditions in which the social norm nudge was active, a bright yellow banner at the top-right corner of the webpage informed the participant that 76% of future colleagues had already subscribed to the Job ticket. In conditions in which no social norm nudge was active, this information was missing. The investigated behavior was the subscription decision regarding the Job ticket at the bottom of the page. The layout of the webpage with the example condition DSN can be found in “Appendix 1”.

After clicking through this Job ticket paragraph, the mock-up contract ended and participants encountered a questionnaire. The questionnaire items provided information on participants’ demographic data, their mobility-related behavior, and on the extended theory of planned behavior.

Demographic, geographic, and mobility related descriptive data (adapted from Mobilität in Deutschland; Nobis and Kuhnimhof [34]) is presented for the complete sample (N = 373) and for the subsample of participants (N = 212) in Table 1, who reported that they already possess a Job ticket or a similar public transport subscription ticket. After scanning this data, it was concluded that there are no substantial differences between the four experimental conditions (DSN, D, SN, Control) in the complete sample, nor in the subsample. This impression was supported by test statistics (see Appendix 2).

The questionnaire’s design followed Ajzen’s [2] recommendations and included adapted items from Paul et al. [36], who established the extended TPB. The questionnaire measured attitude and injunctive as well as descriptive norm towards the purchase of public transport tickets, perceived behavioral control of purchasing a public transport ticket, and environmental concern with a focus on the environmental issues caused by motor car use, with 5 to 7 items on a 7-point Likert scale each (see Appendix 3). Items regarding purchase intention were omitted, as our experiment was designed to record purchase behavior within the scenario directly.

Scale reliability was ensured through computation of Cronbach's α using SPSS 26. Two items (PBC5 and EC6) did not meet the threshold value of 0.50 [16] in the corrected item-to-total correlation and were thus removed from further analysis. After exclusion of these two variables, Cronbach’s α of all constructs was greater than the optimal 0.80 [43] (Table 2).

To check if purchase behavior could be predicted with the extended version of the theory of planned behavior, a two-step analysis comprised of a confirmatory factor analysis (CFA) followed by structural equation modeling (SEM) using maximum likelihood estimation was performed. CFA was used to determine if the questionnaire adequately measured the four latent constructs, and therefore to assess the reliability and validity of the measurement model. SEM was used to determine the causal relationships between these constructs and the outcome variable, and therefore to assess the validity of the structural model.

To evaluate the goodness-of-fit of the measurement model as well as the structural model, a range of recommended indicators was used. A good model fit was considered when χ²/df was between 2 and 3; when goodness of fit indicators (GFI (goodness-of-fit index), CFI (comparative fit index), and TLI (Tucker-Lewis-Index)) > 0.90; and when RMSEA < 0.07 [21, 28].

To check if the nudges affected purchase behavior, a binomial logistic regression with dummy variables was performed for the complete sample (N = 373). The raw choice data per condition is displayed in Fig. 3. The least tickets were purchased in the conditions in which the social nudge was active (54.0% and 54.3%). The most tickets were purchased in the default condition (61.9%). In the control condition, 58.0% of people purchased the ticket.

To gain insight into the effect of the nudge on those that had not previously purchased a Job ticket or a comparable public transport subscription ticket (and were thus the main target for this study), the participants who indicated they already owned such a ticket were excluded. With the resulting subsample of N = 212, the analysis was performed again. In these results, the trend visible for the complete sample intensified (Fig. 3).

To see if there are any statistically significant differences, a logistic regression with dummies for each variable (the control condition being omitted) was performed for both the complete sample and the subsample. Results are reported in Table 7. No condition had a significant effect on purchasing behavior. The model explained 0.06% (Nagelkerke R²) of the variance in purchasing behavior and correctly classified 57.4% of cases in the complete sample. In the subsample, too, no condition had a significant effect on purchasing behavior. Here, the model explained 2.5% (Nagelkerke R²) of the variance in purchasing behavior and correctly classified 61.3% of cases.

The extended theory of planned behavior [36] predicted the purchasing decision well. It did, however, not prove to be of higher utility than the original theory of planned behavior by Ajzen [1]. Commonly, subjective norm is described as the weakest predictor of behavior [2, 38]. In this experiment, however, subjective norm was the strongest predictor of the subscription decision. If a subject believed others appreciated the ticket (injunctive) and would purchase it as well (descriptive), he/she was likely to purchase the ticket, too. As expected, attitude toward purchasing a ticket predicted the subscription decision as well. Perceived behavioral control, however, had no statistically significant influence on purchasing decision. This is likely due to the questionnaire items revealing slight skewness and kurtosis, meaning that most participants found it similarly easy to purchase the ticket in this study. After all, it took only a simple mouse click to decide for or against the ticket. Undoubtedly, perceived behavioral control still plays a significant role when it comes to using public transport, as shown in several studies (e.g. Donald [18, 24, 25]), however, this was not investigated in the present study.

Lastly, environmental concern had an indirect effect on behavior which was mediated by each of the other predictors of the TPB, but had no direct effect on the subscription decision. On the one hand, this finding is in line with Heimlich and Ardoin [26], who summarize that pro-environmental attitudes rarely lead to actual behavioral changes. In the traffic sector (and numerous others), behavior change is often induced by marketing/advertising for the environmental benefit or sustainability of a transport mode. However, according to our results, sustainability concerns did not drive the decision to buy the ticket on its own, so it might be worthwhile to interlock environmentally-focused marketing with subjective norms. On the other hand, the lack of a direct effect on behavior could, in part, be explained by the use of environmental concern as opposed to other constructs measuring environmental attitudes. For example, personal (environmental) norm, as found in Schwartz’s [40] norm-activation theory, targets self-expectations based on internalized values and might have yielded different results.

We investigated whether the effectiveness of nudges stretches to the transport sector. While trends were observed, statistically significant results were not obtained, which provides an important understanding of the limits of the effectiveness of nudges in the transport sector.

The results presented here fall in line with the summarized findings of Byerly et al. [13], i.e., that in changing the environmentally relevant behavior of transportation choice, defaults and norms (as well as education) have no effect whatsoever—even though it was established with the structural equation modelling, that subjective norms do have a strong effect on the purchase decision. It seems, thus, that it is very hard for interventions to influence this predictor. Interventions targeting sustainable transportation using commitments, salience, and finances did previously show mixed to promising effects. Still, it is claimed that “the nudging of travelers could be one of the most promising approaches to deal with the need for a radical and urgent behavioral change” ([6], p. 15). If this is the case, social nudges and defaults, as designed in the presented experiment, do not seem to be the appropriate choice.

Even though subjective norm was the strongest predictor in the model, the social nudge was not at all effective, producing even less subscriptions than the control condition. There has previously been evidence that people display reactance to social nudges [4]. Considering that the social nudge used here was quite obvious (a bright yellow banner), and mentioned a high number of purchased tickets compared to the real numbers (73% vs. 16%), reactance might be an explanation. Since social nudges were not effective in other transport studies as of yet, one can assume reactance cannot be the sole factor. Social nudges do work, on the contrary, in the sector of reducing waste [22] and water use [11]. These studies focused on refraining from “bad” behavior (using less water, reducing paper waste), while this experiment focused on incentivizing “good” behavior (buying a ticket, using public transport). The social nudge used here further targeted highly habituated behavior and involved monetary costs, which probably increased the cost of changing behavior.

The involvement of habituated behavior and monetary costs might also explain the results of the default nudge. Even though there was a trend following our expectations, the result was not robust. First, commuting is highly habituated. Commuters have traveled this route countless times and changing the mode or route is thus connected to possibly uncomfortable alterations and mental workload. The targeted behavior in successful nudging studies using the default (e.g. choosing an energy contract, Momsen and Stoerk [33] is typically not routine behavior. Why should subjects invest in changing something that already works, and potentially lose money while doing so?

These significant monetary costs could have led to bigger resistance toward purchasing the ticket. Losses weigh higher than gains [46], which might have activated conscious thinking (system 1 of the dual process theory), and nudges are believed to attack system 2 (i.e. unconscious processing). On top of that, opportunity and comfort costs may arise. If various leisure activities are planned for after work, a car is a symbol of freedom, it provides flexibility and spontaneity which people might think public transport cannot offer. Our default might have been designed too softly to consider all factors.

As previous studies found, nudges can be effective. In this experiment, where there was monetary cost involved, switching the default was simple, the social nudge was quite obvious, and the targeted behavior was very habituated, nudging seemed to hit its limit, and these limits still require systematic analysis. In future studies, it seems worthwhile to investigate exactly how harsh nudges should be designed to ensure their impact in the transport sector.

Since the nudges themselves were not effective, they could not be tied to the predictors of the model. We therefore urge that connecting nudges to model predictors should be retested with effective nudges to facilitate the search for a comprehensive framework regarding nudging theory. However, in light of these results, it is debatable if there is a good enough reason to keep using the general term of nudging or if it is more promising for future research to revert back to looking at interventions and their mechanisms individually.

When interpreting the above-mentioned results of the structural equation model, it is important to remember that the questionnaire was targeted at simply the purchase of the ticket, in line with Paul et al. [36], who targeted the purchase of green products. However, the purchasing decision of a public transport ticket would naturally be influenced by attitude, social norm, and self-efficacy toward using public transport, as well. To get a more holistic picture of the issue, it would be valuable to consider this in future studies.

Another suggestion for future studies is taking the experiment out “on the road”. Participants were asked to imagine the scenario and to act accordingly. While this procedure offers great feasibility and is fairly common, it can obviously not be guaranteed that it would produce the same results as in a field study.

While promoting public transport seems like a good idea to increase sustainability, it entails an important drawback: instead of attracting car drivers, promotions tend to entice walkers and bikers, who are already moving sustainably. In the city of Hasselt (Belgium), 16% of public transport users stated that they had left a car at home, while 21% would have walked or biked instead [15]. To avoid this, ways to target such strategies especially and directly at car drivers need to be found.

While the theory of planned behavior could predict the purchasing decision toward the Job ticket well, environmental concern did not directly affect it. This is an important finding regarding the advertisement of sustainable transport while also implying considerations regarding the measurement of environmental attitudes. The lack of effectivity of the nudges provides insights into the limitations of nudging theory for the transport sector. Here, nudging is different to—and apparently more difficult than in—the fields of behavioral economics, in which the concept initially boomed. Rather than a single or once in a lifetime decision, transport mode choice is habituated behavior, enforced daily, and monetary costs are involved. For policy makers, it is important to consider that social nudges could potentially trigger reactance in people, and that softly designed nudges might only have very slight effects in the transport sector, which needs to be weighed against the (typically low) cost of implementing them.