Introduction
When providing products, public sectors often face challenges stemming from citizens’ excessive appropriation due to the contradiction between short-term-oriented individual interests and long-term-oriented collective interests (Godwin and Shepard, Reference Godwin and Shepard1977). To address these challenges, public sectors may employ various governance means, including market, leviathan and community (Welch, Reference Welch1983; Oates, Reference Oates1985; Ostrom, Reference Ostrom1990). As behavioral public policy (BPP) and behavioral public administration (BPA) gain prominence in these decades, scholars realize that the behavioral approach offers an alternative or supplement to traditional governance means (Thaler and Sunstein, Reference Thaler and Sunstein2008; Oliver, Reference Oliver2017; Zhang and Li, Reference Zhang and Li2018).
In addition to comparing behavioral approaches with traditional governance means, BPP scholars also discuss behavioral instruments as a set of techniques to ensure effective social change, distinguishing from traditional policy instruments such as carrots, sticks and sermons (Bemelmans-Videc et al., Reference Bemelmans-Videc, Rist and Vedung2010). While the diverse behavioral instruments can be broadly termed as nudging – a label for noncoercive methods of influencing people – this wide-ranging concept may obscure many meaningful distinctions (Gigerenzer, Reference Gigerenzer2015), such as shaping behavioral environments through architectural nudging, sludge or facilitating (Michaelsen and Esch, Reference Michaelsen and Esch2022; Sunstein, Reference Sunstein2022a, Reference Sunstein2022b); improving knowledge or competence through de-bias, boost or technocognition (Grüne-Yanoff and Hertwig, Reference Grüne-Yanoff and Hertwig2016; Kozyreva et al., Reference Kozyreva, Lewandowsky and Hertwig2020; Banerjee et al., Reference Banerjee, Grüne-Yanoff, John and Moseley2024); mixed type such as self-nudging or nudge + (Reijula and Hertwig, Reference Reijula and Hertwig2022; Banerjee and John, Reference Banerjee and John2024); and budging that involves different implementers (Oliver, Reference Oliver2015). Among these instruments, we follow the ‘nudging versus boosting’ category of Grüne-Yanoff and Hertwig (Reference Grüne-Yanoff and Hertwig2016), adopting a narrow definition of nudging (i.e., architectural nudging), to avoid the potential disadvantages of a broader definition. Architectural nudging mainly emphasizes properties of choice architecture rather than relying on educative information (Hansen, Reference Hansen2016; Münscher et al., Reference Münscher, Vetter and Scheuerle2016; Hertwig and Mazar, Reference Hertwig and Mazar2022; Sunstein, Reference Sunstein2022b; Münscher, Reference Münscher2024). Its political philosophy assumes that the choice architect determines the best choice for people and steers people’s behaviors to the desired direction without incentives, coercion, or education (Gigerenzer, Reference Gigerenzer2015).
This research focuses on comparing and combining nudging and boosting since these two representative instruments well embody the different aspects of bounded rationality. The nudging strategy focuses on the irrationality and limitation aspects of bounded rationality, whereas the boosting strategy focuses on the improvability aspect of bounded rationality (Hertwig and Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017; Liu, Reference Liu2022). Apart from theoretical discussions, empirical comparisons between these two behavioral instruments are still relatively rare, especially in non-Western contexts (Lazaric and Toumi, Reference Lazaric and Toumi2022; van Roekel et al., Reference van Roekel, Reinhard and Grimmelikhuijsen2022; Igei et al., Reference Igei, Kurokawa, Iseki, Kitsuki, Kurita, Managi, Nakamuro and Sakano2024). Due to cultural differences, we cannot assume the universality of past evidence. Therefore, further studies are necessary to enhance our confidence in this category.
In this research, we conduct two experiments to empirically test the differences and complementarities between nudging and boosting from China. The results show that both nudging and boosting strategies guide individuals to use public products reasonably; while nudging has a greater immediate effect (i.e., the effect observed right after the intervention), boosting has a greater delayed effect (i.e., the remaining effect after a certain period following the intervention). Furthermore, the combination of the two interventions generally outperforms either strategy when used alone.
Our research contributes to empirically examining the respective effects of nudging and boosting on the reasonable use of public products in non-Western contexts, as well as the differences between them. More importantly, our verification of the better effects of combined intervention suggests that nudging and other ‘improved’ behavioral instruments (such as boosting) can complement each other’s strengths, providing insights for future exploration on how to combine various behavioral instruments more organically (He et al., Reference He, Li and Liang2018). Furthermore, we extend our discussion to illustrate how the combination of nudging and boosting comprehensively embodies the limitation and improvability of bounded rationality, as well as its potential to be integrated into the framework of behavioral governance (Zhang et al., Reference Zhang, Wang and Qin2024).
Theory and hypotheses
The limitation and improvability aspects of bounded rationality
The category of nudging and boosting traces its origins to bounded rationality. ‘Rationality is concerned with the selection of preferred behavior alternatives in terms of some system of values whereby the consequences of behavior can be evaluated’ (Simon, Reference Simon1948, 75). When discussing bounded rationality, ‘bounded’ has two meanings. One is that ‘rational’ should be modified with appropriate adverbs to clarify the criteria used to judge whether a behavior is rational. For example, objectively rational directs to the correct behavior that in fact maximizes given values in a given situation, while subjectively rational refers to maximizing attainment relative to the actual knowledge of the subject (Simon, Reference Simon1948, 75–77). Objective rationality is judged by the factual effectiveness of behaviors, while subjective rationality is judged by the decision-maker. The other meaning of bounded is that individuals’ rationality is limited. Individuals may lack knowledge due to cognitive constraints, mishandle correct information due to emotional interferences, or receive incorrect information because of an inadequate environment. These limitations, stemming from cognition, noncognitive factors (such as emotion or motivation), and the environment, restrict individuals’ rationality. Consequently, they are often unable to fully recognize alternatives, consequences and values, which frequently leads to inconsistencies between subjective rationality and objective rationality (Liu, Reference Liu2022).
The complexity of bounded rationality reflects its two aspects: limitation and improvability. On the one hand, the bounds of individual rationality are limited, meaning individuals may only be able to consider a fraction of the available options within a given search space (Mills, Reference Mills2025). Within these bounds, individuals sometimes fail to achieve subjective rationality due to the influence of noncognitive factors such as emotion or motivation (Liu, Reference Liu2022). And when their behavior is evaluated against objectively rational criteria, it is more difficult to be considered rational. On the other hand, this limited aspect does not mean a pessimistic view that the masses lack rationality (Gigerenzer et al., Reference Gigerenzer and Todd1999, 27). On the contrary, the rationality of individuals is improvable. Individuals can achieve subjective rationality by improving how they process information or overcoming the interference of noncognitive factors within the bounds of rationality (Hertwig and Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017). Additionally, organizations and institutions can help individuals achieve a high degree of objective rationality by expanding their bounds of rationality or providing a choice environment of ‘givens’ for them (Mills, Reference Mills2025; Simon, Reference Simon1948, 67, 79).
As two behavioral instruments, nudging and boosting strategies can both help individuals achieve objective rationality, meaning directing them to good decisions, while they adopt different approaches.Footnote 1 The typical nudging strategies mainly concentrate on the limitation aspect of individual rationality, viewing systematic biases as ingrained irrational errors and advocating steering people’s behavior directly through external choice architectures in a ‘stimulus-response’ pattern. During this process, individuals’ subjective rationality remains unchanged. The typical boosting strategies, however, mainly concentrate on the improvability aspect of individual rationality, viewing simple heuristics and motivational interventions as ecologically rational decision-making methods and advocating empowering people indirectly through internal competence improvement in a ‘stimulus-cognitive-response’ pattern. During this process, boosting enhances individuals’ subjective rationality levels through cognitive improvement (Simon, Reference Simon1948, 75–77; Grüne-Yanoff and Hertwig, Reference Grüne-Yanoff and Hertwig2016; Hertwig and Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017; Zhang et al., Reference Zhang, Wang and Zhou2018; Hansen, Reference Hansen2016; Liu, Reference Liu2022). These differences between the two instruments lead to variations in the speed and sustainability of behavioral changes and suggest the potential for combining these instruments.
Nudging: default options and the reasonable use of public products
Thaler and Sunstein (Reference Thaler and Sunstein2008) defined nudging as ‘any aspect of the choice architecture that alters people’s behavior in a predictable way without forbidding any options or significantly changing their economic incentives’. Nudging is based on insights into the dual-system cognitive structure, which comprises an intuition system and a reasoning system (Kahneman, Reference Kahneman2003, Reference Kahneman2011; Evans, Reference Evans2008). On the one hand, nudging interventions utilize stable patterns of cognitive biases (systematic errors) within the intuition system, which is fast, automatic and effortless (Thaler and Sunstein, Reference Thaler and Sunstein2008; Kahneman, Reference Kahneman2011). On the other hand, such interventions emphasize directly shaping people’s behaviors without engaging in the reasoning system, which is slow, serial and effortful (Thaler and Sunstein, Reference Thaler and Sunstein2008; Kahneman, Reference Kahneman2011). Nudging becomes an effective and inexpensive behavioral instrument owing to its alignment with the features of the dual-system cognitive structure; people usually make decisions on the basis of the intuition system while maintaining the reasoning system in a comfortable low-effort mode to save psychological resources, and those systematic errors within the intuition system are difficult to correct via the reasoning system (Kahneman, Reference Kahneman2011; Zhang et al., Reference Zhang, Wang and Zhou2018).
Among the numerous nudging strategies, meta-analysis has shown that the default option is one of the most commonly used and effective instruments (Zhao et al., Reference Zhao, Liu, Li and Zheng2022). The default option is a preset strategy that is taken unless the decision maker actively makes a change (Johnson and Goldstein, Reference Johnson and Goldstein2003; Michaelsen and Sunstein, Reference Michaelsen, Sunstein, Michaelsen and Sunstein2023). As many studies have found, individuals prefer inaction or to stick with existing choices (Kahneman et al., Reference Kahneman, Knetsch and Thaler1991; Huang et al., Reference Huang, Song, Shao, Li and Liang2018) because the default option may hint at policy-makers’ recommendation, minimize cognitive effort, or help avoid the regret associated with loss aversion during decision-making tradeoffs (Johnson and Goldstein, Reference Johnson and Goldstein2003). The effectiveness of the default option has been widely demonstrated in environmentally related behaviors, such as opting for green electricity (Pichert and Katsikopoulos, Reference Pichert and Katsikopoulos2008), using double-sided printing to save paper (Egebark and Ekström, Reference Egebark and Ekström2016), or reducing disposable cutlery (He et al., Reference He, Pan, Park, Sawada and Tan2023). Similarly, we hypothesize the following:
Hypothesis 1: When receiving a nudging intervention (i.e., default option), participants more reasonably use public products.
Boosting: future orientation and reasonable use of public products
Compared with nudging, which changes individuals’ external environments, boosting improves individuals’ decision-making strategies, skills and knowledge through simple heuristics or motivational interventions and thus enhances their subjective rationality and indirectly enhances the objective rationality of their behaviors (Hertwig and Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017; Zhang et al., Reference Zhang, Wang and Zhou2018; Simon, Reference Simon1948, 76). A simple heuristic refers to simplified cognitive strategies such as one-reason decision-making, whereas a motivational intervention focuses on autonomous motivation adjustment, cognitive control and self-control through expressive writing or other external interventions (Gigerenzer et al., Reference Gigerenzer and Todd1999, 73–188; Ramirez and Beilock, Reference Ramirez and Beilock2011; Hertwig and Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017). On the basis of the former, the latter helps individuals become sufficiently motivated to implement newly acquired skills, thereby improving their competence and enabling them to make satisficing decisions under conditions of limited computational capability and information and environmental uncertainty (Grüne-Yanoff and Hertwig, Reference Grüne-Yanoff and Hertwig2016).
As a psychological tendency toward future time dimensions in decision-making at a given time (Bergadaà, Reference Bergadaà1990; Holman and Silver, Reference Holman and Silver1998; Zimbardo and Boyd, Reference Zimbardo and Boyd1999), future orientation is one of the most suitable boosting interventions for the reasonable use of public products. In many situations involving the use of public products, individuals fail to realize the aggregated future consequences (Kaenzig and Wüstenhagen, Reference Kaenzig and Wüstenhagen2010; Stankuniene et al., Reference Stankuniene, Streimikiene and Kyriakopoulos2020). Future orientation, however, can improve individuals’ competence to vary their sense of connection with their future self or their competence to mentally bridge long time horizons (Shipp et al., Reference Shipp, Edwards and Lambert2009; Hershfield et al., Reference Hershfield, Goldstein, Sharpe, Fox, Yeykelis, Carstensen and Bailenson2011; Zaval et al., Reference Zaval, Markowitz and Weber2015). Considering that ‘present–future’ separation significantly influences individuals’ opportunistic and pro-environmental behaviors (Ostrom, Reference Ostrom1990, 33–38; Joireman, Reference Joireman, Strathman and Joireman2005; Milfont et al., Reference Milfont, Wilson and Diniz2012), numerous studies have investigated the effectiveness of future orientation and found that this intervention promotes pro-environmental behaviors, prosocial behaviors and behaviors for long-term collective benefits (Strathman et al., Reference Strathman, Gleicher, Boninger and Edwards1994; Joireman et al., Reference Joireman, Lasane, Bennett, Richards and Solaimani2001; Zaval et al., Reference Zaval, Markowitz and Weber2015). Considering China's high score on the long-term orientation index (Hofstede et al., Reference Hofstede, Hofstede and Minkov2010, 235–276; Ma et al., Reference Ma, Ding, Shen, Kuang, Yang, Xu and Li2022), we expect that future orientation intervention can be easily accepted in promoting reasonable use of public products (Liu et al., Reference Liu, Chen, Tao, Li and Zheng2024) and therefore formulate the following hypothesis:
Hypothesis 2: When receiving a boosting intervention (i.e., future orientation), participants more reasonably use public products.
Comparing and combining the nudging and boosting
The different cognitive bases and core interventional aspects of nudging and boosting lead to their varying effects (Hertwig and Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017; Zhang et al., Reference Zhang, Wang and Zhou2018). Specifically, nudging theory posits that cognitive bias is a systematic and inherent phenomenon stemming from the automatic and rapid operation of System 1, which is difficult to recognize and reject (Kahneman, Reference Kahneman2011). Accordingly, nudging theory harnesses bias to design a choice architecture that aligns with individual cognitive and motivational processes rather than attempting to avoid or rectify the bias. Throughout the intervention process, individuals may change their behaviors without being aware of the influence of cognitive bias (Grüne-Yanoff and Hertwig, Reference Grüne-Yanoff and Hertwig2016; Rasmussen et al., Reference Rasmussen, Lindekilde and Petersen2024). Therefore, nudging intervention leads to individuals’ reflexive behaviors, and the intervention effect appears immediately; however, once the choice architecture changes, the effect may also disappear rapidly (van Roekel et al., Reference van Roekel, Reinhard and Grimmelikhuijsen2022).
Boosting theory, however, adopts a different intervention approach. Although boosting theory acknowledges the existence of heuristics, it does not view them as cognitive errors (biases) deviating from rationality but rather as ecologically rational strategies that match the characteristics of the decision-making environment (Hertwig and Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017; Zhang et al., Reference Zhang, Wang and Zhou2018). Accordingly, boosting theory aims to improve individuals’ various competences, maintain their subjective initiative, and enable them to make wise decisions independently. Since boosting interventions focus on the development of individual cognitive and motivational competence rather than direct behavioral change, individuals may need more time to acquire new skills, and the effectiveness of these interventions may not be immediately apparent. However, once competence is established, it tends to stabilize over time and persist even after the interventions are removed (Hertwig, Reference Hertwig2017; Caballero and Ploner, Reference Caballero and Ploner2022; Paunov and Grüne-Yanoff, Reference Paunov and Grüne-Yanoff2023). Therefore, we hypothesize the following:
Hypothesis 3: The immediate effect of the nudging intervention is greater than that of the boosting intervention. Specifically, participants receiving the nudging intervention more reasonably use public products immediately after the intervention than those receiving the boosting intervention.
Hypothesis 4: The delayed effect of the boosting intervention is greater than that of the nudging intervention. Specifically, after the behavioral interventions are removed, participants who have received the boosting intervention more reasonably use public products than those who have received the nudging intervention.
Additionally, there is limited evidence on the combined effects of nudging and boosting interventions, with only a few exceptions (Lazaric and Toumi, Reference Lazaric and Toumi2022; Igei et al., Reference Igei, Kurokawa, Iseki, Kitsuki, Kurita, Managi, Nakamuro and Sakano2024). In addition to offering practical significance by providing more policy instrument combinations, evaluating the combined effects also holds theoretical significance. Such evaluation not only helps eliminate potential crowding-out effects (Howley and Ocean, Reference Howley and Ocean2022) but also allows the strengths of nudging and boosting interventions to complement and offset their respective disadvantages (He et al., Reference He, Li and Liang2018). Thus, we expect that combining nudging and boosting may achieve a better effect, and we hypothesize the following:
Hypothesis 5: Compared with participants who experienced a single nudging intervention, participants who experienced both nudging and boosting interventions more reasonably use public products.
Hypothesis 6: Compared with participants who experienced a single boosting intervention, participants who experienced both nudging and boosting interventions more reasonably use public products.
We conducted two experiments to verify these hypotheses. These two experiments adopted the approach of conceptual replication (Crandall and Sherman, Reference Crandall and Sherman2016), examining the robustness of hypothesis-testing conclusions by modifying experimental scenarios and operationalizations. We chose reasonable use of electricity in talent apartments and toilet paper in public restrooms as two specific scenarios. Both public products are supplied by the public sector for free, and citizens’ excessive appropriation may lead to great waste.Footnote 2 Traditional governance means, however, may be less advantageous compared to the behavioral approach due to political factors or high administrative costs.Footnote 3 Moreover, these scenarios are very common in China, which makes the experimental task easier to understand.
Study 1: a laboratory experiment on using free electricity
Experimental design
Study 1 used the free electricity in a talent apartment as an operationalization of public products within a virtual context. We conducted a laboratory experiment using a 4-level single-factor between-participants design, in which participants were randomly assigned to one of the four experimental conditions (i.e., nudging condition vs. boosting condition vs. combination condition vs. control condition). To test the immediate and delayed effects of nudging and boosting, participants were asked to report their electricity use at two time points – right after the interventions and after a certain time interval. A randomized laboratory experiment could achieve higher internal validity for testing theoretical hypotheses. Specifically, it also facilitates the control of the inter-task interval to examine delayed effects in this research.
Participants
Since this experiment has two steps and the interval time needs to be controlled, we chose an offline student sample rather than an online representative sample. In addition, students at universities are potential future users of talent apartments, which alleviates the external validity problem caused by student samples (Peterson, Reference Peterson2001).
We used G*Power3.1.9.7 to conduct a prior estimate of the sample size before the experiment (Faul et al., Reference Faul, Erdfelder, Lang and Buchner2007). The total required sample size for a four-level one-way analysis of variance between participants was at least 180 (effect size f = 0.25, significance level α = 0.05, statistical power 1−β = 0.8, number of groups = 4). After the pilot study and modification, we recruited 228 student participants. After excluding those participants who did not follow the guidelines or who did not participate in the posttest, data from 224 participants were included in the final analysis. Table 1 shows the descriptive statistics of the main variables. Most participants were female (60.3%) and 18–25 years old (96%), had an undergraduate degree (77.7%), and had medium or high income levels.
Table 1. Descriptive statistics of the main variables
Note:
a Considering the difference in the experimental tasks between Step 1 and Step 2, the two electricity consumption (unit: kWh) cannot be compared directly.
Procedures and experimental tasks
Figure 1 shows a summary of the procedures in this experiment, with Step 1 aiming to investigate the immediate effect of the behavioral interventions and Step 2 aiming to investigate the delayed effect 90 minutesFootnote 4 after participants were subjected to the intervention. All the participants read the experimental instructions and independently performed the experimental tasks through mobile phones or laptops in a laboratory or a classroom. The participants were asked to imagine that they successfully applied for a 60 m2 talent apartment in city G, in which the government fully covers electricity bills.
Figure 1. Summary of the laboratory experiment in study 1.
The participants in each randomly assigned group were required to report the usage of various electrical appliances in two given scenarios. In the first scenario, participants had to decide whether to turn on various lights while working or studying alone at home from 8 PM to 10 PM and, if so, for how many hours. In the second scenario, they were required to report their air conditioner usage while at home from 8 PM to 8 AM the next day in the summer. The participants subsequently completed questionnaires, including items related to manipulation checks, control variables and demographic variables. After a 90-minute interval, participants proceeded to Step 2. In Step 2, the interventions were removed and participants were directed to imagine a slightly different yet similar scenario, where they had to decide whether to turn on the lights and, if so, for how long they would keep them on while having dinner at home from 6 PM to 7 PM, without any interventions. We used a scenario distinct from that in Step 1 to minimize practice effects.
Manipulation and measures
Manipulation. In the nudging group, participants were presented with a default option for each experimental task item. For example, in the first scenario, supposing one works or studies alone at home from 8 PM to 10 PM, the default option was to turn on the ceiling lamp and desk lamp in the bedroom while turning off all other lights. In the boosting group, participants were asked to write down five key words related to ‘the future impact of energy conservation and emission reduction’ before the experimental task to prime their future orientation. The combination group involved participants writing keywords for priming the future orientation before the task and receiving default options during the task. The control group involved no intervention (see Supplementary Appendix 1 for a complete description of the laboratory experiment).
Dependent variables. All the lights and the air conditioner were labeled with their rated power levels. We calculated the total electricity consumption for each participant as the dependent variable (electricity consumption = rated power × open hours, unit: kWh). We believe such an operation decreases the reduction in internal validity and in ecological validity caused by participants calculating electricity consumption themselves, which rarely occurs in daily life and thus may trigger deliberation.
Other variables. The manipulation checkFootnote 5 for future orientation consisted of three items adapted from the concept of future consequences (Strathman et al., Reference Strathman, Gleicher, Boninger and Edwards1994). If participants who received the future orientation intervention scored significantly higher, then the manipulation was effective. The control variables included resource concerns (Stern et al., Reference Stern, Dietz, Abel, Guagnano and Kalof1999) and comfort preferences (Mi, Reference Mi2011), both of which were measured on 5-point Likert scales and demonstrated good reliability and validity (see Supplementary Appendix 2 for the scale items). The demographic variables included gender, age, education and family income.
Results
There are no significant differences in control variables between the groups; p all > 0.1 (see Supplementary Appendix 3), indicating that the sample allocation was randomized. The manipulation was proven effective considering that participants in the boosting (M = 4.07, SD = 0.39) and combination groups (M = 4.09, SD = 0.45) scored significantly higher on future consequences than those in the control group (M = 2.6, SD = 0.47); p all < 0.001 (see Supplementary Appendix 4).
Table 2 presents the results of the analysis of covariance (ANCOVA) for the task in Step 1, with control variables as covariates. We also conducted Bonferroni multiple comparisons of means. Figure 2 clearly illustrates these results by group. The participants in both the nudging group (M = 3.90, SD = 1.35) and the boosting group (M = 4.79, SD = 1.48) reported significantly lower electricity usage levels than those in the control group (M = 6.10, SD = 1.57); p all < 0.001. Thus, Hypotheses 1 and 2 are confirmed. The difference in the degree of electricity usage between the participants in the nudging group and those in the boosting group reached significance (p < 0.01). Thus, Hypothesis 3 is confirmed. Furthermore, participants in the combination group reported significantly lower electricity usage (M = 3.03, SD = 0.90) than those in the nudging group (p < 0.01) and those in the boosting group (p < 0.001). Thus, Hypotheses 5 and 6 are confirmed.
Figure 2. Electricity consumption with the intervention.
Table 2. Results of ANCOVA in Step 1
Note: The dependent variable is electricity consumption; the sample size N = 224; for the ANCOVA, the noninteger values in the factor variables are rounded to the nearest integer values; SS: sum of squares (type III), DF: degrees of freedom, MS: mean square.
After an interval of 90 minutes, all participants completed a decision task without any intervention. As shown in Table 3 and Figure 3, participants who experienced the boosting intervention in Step 1 reported significantly lower electricity usage (M = 0.091, SD = 0.056) than those who experienced the nudging intervention (M = 0.131, SD = 0.043), p < 0.001. Thus, Hypothesis 4 is confirmed. In addition, participants assigned to the boosting group in Step 1 reported significantly lower electricity usage than those in the control group (M = 0.141, SD = 0.044), p < 0.001. There was no significant difference between the participants in the nudging group and those in the control group. While participants in the combination group reported significantly (p < 0.001) lower electricity usage (M = 0.082, SD = 0.053) than those in the control group and nudging group, their reported electricity usage level was not significantly different from that in the boosting group. These results generally align with our theoretical expectations.
Figure 3. Electricity consumption after the interventions were removed.
Table 3. Results of ANCOVA in Step 2
Note: The dependent variable is electricity consumption; the sample size N = 224; for the ANCOVA, the noninteger values in the factor variables are rounded to the nearest integer values; SS: sum of squares (type III), DF: degrees of freedom, MS: mean square.
Study 2: a field experiment on using free toilet paper
We conducted another field experiment to examine actual usage behavior, as intentions in the laboratory may not fully align with real-world actions (Webb and Sheeran, Reference Webb and Sheeran2006). To strengthen confidence in our theoretical assumptions through conceptual replication (Crandall and Sherman, Reference Crandall and Sherman2016), we adapted the operationalization of both nudging/boosting interventions and of reasonable use of public products. The experiment was set in a real-world context: using public toilet paper in a school building.
Experimental design
This field experiment employed a 3-level single-factor (intervention: nudging vs. boosting vs. combination) repeated-measures design. The experiment was conducted in a school building at Sun Yat-sen University. Owing to concerns for privacy and the limitations of measurement techniques, we could not directly observe the public toilet paper usage of each person. Therefore, we could use only restrooms as the observation unit. We randomly assigned restrooms to the three intervention groups and took the per capita paper usage of one restroom on an observation day as the analysis unit.Footnote 6
For the restrooms in each intervention group, we conducted approximately four observations, including an observation before the interventions (pretest), an observation with the interventions (posttest 1), an observation one day after the interventions were removed (posttest 2), and an observation one week after the interventions were removed (posttest 3). The purpose of setting up four observations is to investigate the immediate and delayed effects of various interventions.
Participants
To ensure the comparability of pre- and post-measurements, it was necessary to maintain a consistent population across four observations.Footnote 7 The current experimental setup largely met this requirement for the following reasons. First, the personnel composition in the selected school building was simple, consisting mainly of faculty members, doctoral students, full-time graduate students and master of public administration (MPA) students from the School of Government, with minimal interference from unrelated individuals. Second, these individuals followed a set schedule for their study or work in the building, allowing us to ensure consistency by conducting observations on specific dates, thus observing the same group of people each time. Third, during the observation period, these individuals studied or worked in fixed locations. Therefore, they typically used restrooms near their classrooms or offices instead of those in other areas, which prevented the same group of people from being exposed to different interventions.
Finally, each intervention group had at least six data points at the restroom level for each observation (except for the combination group in posttest 2; see Supplementary Appendix 5b for more information). Although the number of data points at the restroom level was not particularly large, the number of individual visits during the observation period reached 20,546.
Procedures
Figure 4 shows a summary of the procedures of Study 2. From March 11 to November 17, 2023, we selected specific dates for observations on the basis of the schedules of faculty and students. Initially, we conducted a pretest in each restroom to establish its baseline of daily per capita paper usage. One week later, on the same day, with the same participants working or studying in the same area, we implemented the interventions. The per capita paper usage of that restroom during that day was measured as posttest 1, with the aim of assessing the immediate effects of the interventions. The per capita paper usage of that restroom was subsequently observed one day (posttest 2) and one week (posttest 3) after the interventions were removed to measure their delayed effects.
Figure 4. Field experimental design of Study 2.
Manipulation and measures
There is a tissue box and a toilet roll box next to the sink in every restroom, which is stocked with free toilet paper provided by the university. The toilet paper is standardized to have the same size and weight. During the experiment, we took over the supply of toilet paper from the administrator to avoid any interference.
Manipulation. In the restrooms with the nudging intervention, we placed a white sign with scales beneath the toilet roll, sized it to match the roll’s width, and marked it to indicate four pieces of toilet paper. The default options for toilet roll usage were labeled ‘Two for Pee and Four for Poop’ in women’s rooms, with a note at the bottom stating ‘Enough when Enough’. In men’s rooms, the same note appeared at the three-piece length on the toilet roll, serving as the default option. And all tissue boxes were labeled with a green sign with the caption ‘Only One for Once’. These interventions determine the best choice or desired direction and comply with the mechanism of default option (Johnson and Goldstein, Reference Johnson and Goldstein2003; Gigerenzer, Reference Gigerenzer2015). In the restrooms with the boosting intervention, a poster was placed above the middle of the two boxes to activate individuals’ future orientation. The poster depicted a stump resembling a toilet roll on parched land, with the caption ‘One Roll of Paper, One Year of Trees. Save Toilet Paper, Better for the Future’. In the combination group, we combined the above interventions (for more details, please see Supplementary Appendix 6).
Dependent variable. Since individual-level paper usage could not be observed, we used the per capita paper usage of a restroom in one day (total paper usage of that restroom on that day/the number of individuals entering that restroom on that day) as the dependent variable. We calculated the total paper usage of one restroom on one specific day by measuring the total weight of the standardized tissues and toilet rolls consumed on that day. Additionally, the number of individuals entering the restroom each day was counted via hidden infrared digital counters fixed at the entrance (see Supplementary Appendix 5). The difference in paper usage between pretest and posttest 1 reflects the immediate effects of the interventions. Differences between pretest and posttest 2, as well as between pretest and posttest 3, reflect the delayed effects of the interventions one day and one week later, respectively.
Results
Table 4 shows the descriptive statistics of the main variables of Study 2. Since we used per capita paper usage of a restroom in one day as the dependent variable, the sample size was too small to conduct a parametric test. Thus, nonparametric tests, which are suitable for small sample analysis (Siegel, Reference Siegel1956, 32), were conducted to test the hypotheses. Hypotheses 1 and 2 are confirmed by the experimental data. Figure 5 shows the Wilcoxon matched-pairs signed-rank test results for each intervention group. The nudging (∆ = 0.37, p < 0.01), boosting (∆ = 0.34, p < 0.01) and combination (∆ = 0.71, p < 0.01) interventions all significantly reduced per capita toilet paper usage.
Table 4. Descriptive statistics of the main variables of Study 2
Note: Per capita paper usage = Usage of toilet paper (g) / Number of visits of one restroom on an observation day, the mean of this variable is independently calculated rather than by dividing the means of the previous two variables; for detailed information on the observation of each group, please refer to Appendix 5.
Figure 5. Wilcoxon matched-pairs signed-rank test for each intervention group.
Hypotheses 3 and 4 are not confirmed. Wilcoxon rank-sum tests revealed that the immediate and delayed effects of the nudging intervention were not significantly different from those of the boosting intervention. Considering the importance of reporting null effects and that the trends meet our expectations, we still report these results (Vogel and Homberg, Reference Vogel and Homberg2021). As Figure 6 shows, the immediate effect of the nudging intervention is numerically greater than that of the boosting intervention (∆ = 0.02), whereas its delayed effects are smaller than those of the boosting intervention (∆’delayed effect 1 = −0.22; ∆’delayed effect 2 = −0.12).
Figure 6. Wilcoxon rank-sum tests for the immediate effects and delayed effects of the three interventions.
Hypotheses 5 and 6 are partly confirmed. Wilcoxon rank-sum tests revealed that the immediate effect of the combination intervention was significantly greater than that of the nudging (∆ = 0.35, p < 0.01) and boosting interventions (∆ = 0.37, p < 0.01). While the trends are similar for the delayed effect, the differences are not significant at the 0.05 level (see Figure 6).
Discussion
Effective and cost-efficient behavioral instruments exhibit unique advantages in promoting reasonable use of public products when traditional governance means are limited. While the nudging strategy attaches importance to harnessing individuals’ deeply rooted irrational biases and directly influences their reasonable use of public products, the boosting strategy emphasizes the utilization of individuals’ cognitive malleability and improvement of their competences, considering the finiteness and dynamicity of bounded rationality (Grüne-Yanoff and Hertwig, Reference Grüne-Yanoff and Hertwig2016; Zhang et al., Reference Zhang, Wang and Zhou2018; Liu, Reference Liu2022).
Through a laboratory experiment and a field experiment, we basically verified our hypotheses regarding the effects of two primary behavioral instruments. We found that all nudging, boosting and combination interventions significantly promoted reasonable use of public products. In comparison, in the laboratory scenarios in which electricity is used in talent apartments, the nudging intervention has a significantly greater immediate effect, whereas the boosting intervention has a significantly greater delayed effect. Although these differences do not reach significance in the field experiment, the trends are still consistent with our theoretical expectations. These results provide more empirical evidence for the effectiveness of behavioral instruments and the differences between the nudging and boosting strategies in a Chinese context (van Roekel et al., Reference van Roekel, Reinhard and Grimmelikhuijsen2022). Our research didn’t specifically focus on cultural differences in behavioral interventions, but inspired by China’s cultural dimension of long-term orientation, we designed the future orientation intervention, which may work better in such contexts. Further empirical exploration of cultural differences is promising (Schimmelpfennig and Muthukrishna, Reference Schimmelpfennig and Muthukrishna2025). For example, the impact of communication frames on civic behavior may be related to individualism – collectivism dimension (Wang et al., Reference Wang, Zhang and Zhang2025), while power distance is linked to lure of choice, mental accounting and overconfidence (Hoffmann and Anwar, Reference Hoffmann and Anwar2024).
Furthermore, the combination intervention had a significantly greater immediate effect than any single intervention in both experiments, and it tended to have a greater delayed effect, although this difference was significant only in the laboratory experiments. These results expand the literature by investigating the combined effects of two primary behavioral instruments through a laboratory experiment with a high level of internal validity and a field experiment with a high level of ecological validity, which demonstrates the promising prospects of combining these instruments’ complementary advantages and exploring boundary conditions and causal mechanisms (He et al., Reference He, Li and Liang2018; Lazaric and Toumi, Reference Lazaric and Toumi2022; Igei et al., Reference Igei, Kurokawa, Iseki, Kitsuki, Kurita, Managi, Nakamuro and Sakano2024; Li and Wang, Reference Li and Wang2024). As more behavioral instruments beyond nudging are explored, it is necessary to assess the combined effects of different behavioral instruments to achieve better outcomes and avoid potential crowding-out effects (Howley and Ocean, Reference Howley and Ocean2022). The discovery of a greater combined effect renders the classification of nudging and boosting more valuable in practical applications, as the two are not mutually exclusive. It also suggests a more holistic inheritance of bounded rationality through this classification, embodying the duality of limitation and improvability (Liu, Reference Liu2022).
Finally, the scenarios and interventions in this research hint a possibility of bridging BPP/BPA with governance theory through behavioral governance, applying behavioral science insights to the public governance field (Straßheim and Korinek, Reference Straßheim, Korinek, Wilsdon and Doubleday2015, 155–162; Gofen et al., Reference Gofen, Moseley, Thomann and Weaver2021; Zhang et al., Reference Zhang, Wang and Qin2024). First, from a perspective of bounded rationality, behavioral governance has the same function of improving the objective rationality of individual behaviors as traditional governance mechanisms, addressing where individual self-serving short-term gains harm collective long-term benefits (Godwin and Shepard, Reference Godwin and Shepard1977; Simon, Reference Simon1948, 67, 79). This article only focuses on an appropriation problem, and there remains ample opportunity to explore other forms of appropriation challenges and supply problems such as promoting co-production through behavioral approach (Ostrom, Reference Ostrom1990, 42–50; Wang and Zhang, Reference Wang and Zhang2024; Liu and Lin et al., Reference Liu, Lin, Yuan, He and Zhang2024). Second, behavioral instruments have unique advantages in certain governance contexts. As mentioned earlier, the restriction on market means due to political attributes of Chinese policy implementation in the public electricity scenario, and the high costs of administrative means in the public toilet paper scenario, both highlight the advantages of behavioral instruments in terms of compatibility and affordability. Third, behavioral instruments such as nudging and boosting exemplify the uniqueness of behavioral mechanisms that distinguish them from traditional governance means. In addition to architectural nudging and boosting, there are many meaningful behavioral instruments such as nudge +, self-nudging, etc. (Reijula and Hertwig, Reference Reijula and Hertwig2022; Banerjee and John, Reference Banerjee and John2024). Empirical research and practical application will build a behavioral governance bridge to integrate behavioral science into core public activities (Hallsworth, Reference Hallsworth2023).
Limitations
First, we recommend caution when comparing the methods of the two studies. The comparability would be maximized if we conducted them using the same scenarios, experimental interventions and measurements, with the only differences being their experimental contexts and the participants involved. However, our initial idea was to enhance people’s confidence in theoretical assumptions by conceptual replication rather than enhance people’s confidence in a certain operationalization by empirical extension (Crandall and Sherman, Reference Crandall and Sherman2016). Therefore, we chose different scenarios with various operationalizations since the effects of behavioral interventions are likely limited to the specific choice contexts (Mertens et al., Reference Mertens, Herberz, Hahnel and Brosch2022). We also further tested real behavior considering its potential gap with intention (Webb and Sheeran, Reference Webb and Sheeran2006). These strategies are aimed at robustly verifying the theoretical differences in the effects of nudging, boosting, and combination interventions on promoting reasonable use of public products (including public electricity and toilet paper). Concurrently, we acknowledge that our strategy may indeed sacrifice some comparability. Therefore, based on the confidence in theoretical assumptions established by this preliminary research, future application studies can be further extended empirically to enhance people’s confidence in certain operationalizations.
The second limitation concerns our technical operations. The immersion of laboratory experiment can still be improved. In Study 1, we conduct the experiment in laboratory while using self-reported measurement of dependent variable. We chose laboratory experiment because it tests theoretical hypotheses with higher internal validity by controlling environmental interference and the inter-task interval. But due to considerations of cost and the experimenter effect, we ultimately used self-report in hypothetical scenarios to measure the dependent variable, rather than behavioral measurement in an established ‘talent apartment scene’ in the laboratory. Although self-report is acceptable as we crafted the questions skillfully and conducted a conceptual replication using behavioral measurement in Study 2, the sense of immersion in laboratory experiments could still be enhanced (Anderson and Edwards, Reference Anderson and Edwards2015). Evolving virtual reality (VR) technology offers new opportunities to address this issue. VR can improve task immersion with lower cost, providing behavioral data that are closer to real-life situations than laboratory experiments (Diemer et al., Reference Diemer, Alpers, Peperkorn, Shiban and Muehlberger2015; Brookes et al., Reference Brookes, Warburton, Alghadier, Mon-Williams and Mushtaq2020).
Finally, participant sampling is limited primarily to individuals from Sun Yat-sen University. We maintain that using a student sample does not significantly impact the conclusions, as students are the main (potential) users of talent apartments and university public restrooms. However, when generalizing and applying our conclusions, it is crucial to take social and cultural backgrounds into account, and ethical factors such as autonomy or transparency should also be considered (Sunstein, Reference Sunstein2015; Vugts et al., Reference Vugts, van den Hoven, de Vet and Verweij2020; van Roekel et al., Reference van Roekel, Giurge, Schott and Tummers2023).
Conclusions
The ‘behavioral turn’ has been one of the most popular trends in public administration and policy over the past few decades (Gofen et al., Reference Gofen, Moseley, Thomann and Weaver2021; Zhang and Guo, Reference Zhang and Guo2021). Among various behavioral instruments, we compare the effects of nudging and boosting instruments in promoting the reasonable use of public products, grounded in a comprehensive review of the finiteness and improvability of bounded rationality (Liu, Reference Liu2022). Our experiments demonstrate the effectiveness of both behavioral instruments in promoting the reasonable use of public products, and the immediate and delayed effects of nudging and boosting are indeed different. We further validated the greater effectiveness of the combined intervention, which suggests the theoretical and practical potential of combining behavioral instruments.
Considering that public products are often allocated for specific policy goals, such as attracting talent or providing public welfare, individuals’ public product use tends to be negligible and difficult to observe, and the users of these products are often flowing crowds. Consequently, traditional market, administrative and community mechanisms often fall short in promoting reasonable use of public products. The behavioral approach, however, provides a viable way to address these challenges.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/bpp.2025.10008.
Acknowledgments
We thank Associate Professor Jie Zhou from the Institute of Psychology, Chinese Academy of Sciences and Assistant Professor Xin Chen from the School of Government, Sun Yat-sen University, for their suggestions on this paper.
Funding statement
This study was supported by the National Natural Science Foundation of China [72474237], the National Social Science Fund of China [23CGL056] and the Fundamental Research Funds for the Central Universities, Sun Yat-sen University [24wkqb05].
Competing interests
The authors declare none.
Data availability statement
The data underlying this article are available in the article and its online supplementary material.
Open access
