Research articlesThe expressive function of public policy: renewable energy mandates signal social norms
Abstract
Addressing collective action problems requires individuals to engage in coordinated and cooperative behaviours. Existing research suggests that individuals' propensity to work together depends in part on their belief that others support the cause in question. People form their expectations about prevalent beliefs and behaviours from many sources. To date, most of the literature has focussed on how social norm perceptions are inferred from peers or summary statistics. We explore an understudied source of norm information: the passage of policies by democratically elected institutions. Institutional signals, such as the setting of defaults, national laws or policies, can act as coordination devices, signalling or prescribing social norms to large audiences. However, their expressive function is likely to depend on whether the institution is seen as accountable to the public. In two highly powered, pre-registered experiments (N = 11 636), we examine the role of policy signals as a source of social norm information. In Study 1, Americans randomly assigned to learn that their state passed a 100% renewable energy mandate believe that a greater percentage of their state's residents support such a mandate. In Study 2, we replicate this effect for national policy and show that the influence is moderated by information about whether the government represents the will of the people.
This article is part of the theme issue ‘Social norm change: drivers and consequences’.
1. Introduction
Addressing collective action problems like global climate change requires that individuals engage in coordinated and cooperative behaviour. Past and ongoing research finds that individuals' propensity to work together with others in these challenges is dependent in part on their understanding that others support collective action (i.e. perceived norms around an issue; for a review, see [1,2]). At the same time, emerging research spanning different contexts and geographies finds that people often systematically misperceive prevailing social norms in many domains, including climate change [3–5]. This phenomenon, known as pluralistic ignorance, can inhibit collective action and result in self-silencing (e.g. [6]), which can in turn create a feedback loop that reinforces or entrenches social norms, even when they depart from popular preferences. Understanding how perceptions of social norms are updated and how systematic misperceptions can be overcome is thus critical.
Perceptions of social norms are shaped by different sources of normative information, including individual behaviour, summary information about a group and signals from formal institutions [7]. Here, we examine the ‘expressive’ role of institutional signals in shaping perceived support for a cause (i.e. perceived norms). We focus on the role of political signals, specifically the passage of policies by democratically elected institutions, which could serve as a widely broadcasted indicator of what is popular. We hypothesize that policy signals may be seen to convey information about social norms if people infer that elected political representatives are accountable to their constituents and responsive to public opinion. In such cases, the passage of a policy might be interpreted as a signal of public preferences, leading individuals to update their beliefs about prevailing social norms. If the policy indeed reflects widely held attitudes and these attitudes were previously underestimated, it could decrease misperceptions (i.e. decrease pluralistic ignorance) as individuals recalibrate their perceptions of popular support for the issue.
Here, we test this theory and a potential mechanism in two large, highly powered and pre-registered studies by examining whether exposure to information about a 100% renewable energy mandate (Study 1) or Federal investment in renewable energy (Study 2) increases Americans' estimates about the extent of popular support for these issues. In Study 2, we test a potential moderator of the influence of policy signals on perceived social norms: beliefs about the accountability of the government. Importantly, when people do not believe that a governing body that has passed a policy is accountable to the public, the policy signal may no longer be seen as a meaningful indicator of social norms. In such cases, even the passage of popular policies might ultimately fail to shift norm perceptions or reduce pluralistic ignorance.
By expanding on the scientific understanding of the influence of institutional signals on norm perceptions, we highlight their potential as a catalyst for rapid social change. To the extent that new policies act as widespread signals of prevailing or emerging social norms, we might expect increasing political action on climate change to have indirect effects, in addition to direct effects, on mitigation efforts through its influence on perceived social norms, which might in turn instigate or shore-up collective action.
(a) Norms, social change and collective action
Our behaviour is, in part, influenced by normative beliefs and our own attitudes [8]. There is evidence that it may be easier to intervene on perceptions of social norms than to try and directly change people's attitudes, as the latter are often shaped by years of experiences [1] and do not guarantee changes in accompanying behaviours [9]. Instead, practitioners who seek to promote social change have increasingly turned to interventions that aim to shift social norm perceptions, as these appear to be more malleable over a shorter time frame and to influence behaviours [7].
The definition of social norms varies substantially both across and within academic disciplines. Here, we use perceived social norms to refer to perceptions of prevalent behaviours or beliefs in a particular group that influence one's own actions (for similar definitions, see [10–13]). Social norm information can pertain to the behaviours that others commonly do—referred to as descriptive norms—and what they commonly approve or disapprove of—referred to as injunctive norms [11]. Some define norms such that either perceptions of prevalence or social approval regarding behaviour or attitudes constitute norms of different types, as noted above (e.g. [2,14]), while other definitions contend that a behaviour or belief must both be perceived as common and have anticipated reputational consequences to be a norm (e.g. [15]).
Observing social norm information can produce conformity through a variety of psychological means: norms can shape individuals' perceptions of phenomena in the world [16], can change one's judgements [17] or personal beliefs [18], and they can change one's intentions and behaviours through both deliberative processes [19] and unconscious processes like mimicry [20]. Conformity to norms is driven by powerful motives, including a desire to accurately perceive the world around us and make wise choices (referred to as informational influence), as well as to help us strengthen or maintain social ties with others (referred to as reputational influence [19,21]).
The ability of norms to influence people's behaviours has been documented across different domains. For instance, norm-based interventions that highlighted a relevant social norm, directly or implicitly, were effective in promoting timely tax payments [22,23], increasing organ donation registration [24], increasing voting in an election [25], reducing harassment in a school-setting [26], reducing intergroup conflict and increasing intergroup contact [27,28], increasing female labour force participation [29,30] and promoting a range of sustainability-related behaviours (for a review see [2]). An added strength of norm-based interventions is that they can be communicated through short messages (e.g. flyers, emails or other relevant materials) that can be easily disseminated (e.g. [31,28,32]).
However, interventions that focus on highlighting prevailing norms are not always successful, and existing social norms can be persistent, self-reinforcing and resistant to change [33]. For instance, interventions highlighting normative information can backfire if the person exposed to the information is already adhering to the norm [34–36]. Additionally, there is growing evidence that many commonly held beliefs and behaviours are not only subjective estimates that can depart from actual rates of behaviour or opinion, but they can be systematically misperceived. This may be particularly true for dynamic and rapidly changing issues, such as climate change, where perceived norms may lag behind actual changes in opinions and behaviours, creating pluralistic ignorance [37]. This lag may be explained or reinforced by pervasive and often incorrect indicators of majority opinions (e.g. lack of substantive policy on issues with broad public support such as gun control and, until recently, climate change; Shamir & Shamir [38]). Although not a focus of the present investigation, research has also examined intergroup dynamics involved in norm perceptions and misperceptions, finding that partisans often both misperceive outpartisan norms [39] and hold exaggerated perceptions of outpartisan attitudes [40]. Notably, conformity by partisans is driven more by adopting ingroup attitudes than by opposing outgroup attitudes and norms [41]. Norm-based interventions may be particularly effective when there is a common misperception about a norm [42]. In such cases, perceived social norms may change quickly as new public information becomes available.
Signals from institutions are a potentially important and understudied source of information about social norms. Under the right circumstances, interventions may be able to leverage institutional signals to broadly and efficiently communicate information about social norms.
(b) Institutional signals and norm perception
According to Tankard & Paluck [1], there are three main sources of information that people rely on to learn about new norms and to update their perceptions of existing norms: observation of individual behaviour, learning summary information about a group, and institutional signals. Forming accurate perceptions of widespread public opinions or behaviours (e.g. at a national level) is hard (or arguably impossible) to achieve based solely on first-hand observations of others given the scale, lack of access to information about what a random or unbiased sample of others do or think, and the lack of visibility surrounding many behaviours and opinions, including policy preferences or political behaviours. In these cases, signals by formal institutions may act as a mechanism for coordinating the public by offering a salient and highly visible source of norm information—especially when those institutions are viewed as accountable to the public (e.g. democratic governments) or otherwise dependent on the public (e.g. media organizations).
One potentially important and understudied source of information about social norms are signals from institutions that are perceived to have some level of accountability to the public because (1) they are likely to reflect and/or shape public opinion, (2) are widely broadcast and (3) there is shared or common knowledge that many people are simultaneously exposed to the signal. Institutions can both create norms and be effective communicators of norms—they thus have an ‘expressive’ function [43]. At the same time, they are also themselves shaped by societal norms [44–46]. For example, a law passed by a governing body may directly shift behaviours through taxation or subsidies, indirectly shift behaviours by signalling that there is underlying support for the proposed policy, and it may have been voted into law due to popular support.
The effectiveness with which an institution signals a particular norm through its actions or statements is likely to hinge on whether the institution is legitimate and whether it is seen to be, in some way, accountable to the public. Scholars operationalize legitimacy as the perception that a particular institution reflects, to its best capacity, the interests of the population it represents (e.g. [47]). Thus, one potentially important and understudied source of information about social norms is signals by democratically elected bodies as they may be perceived to be accountable to the public and thus reflect public preferences. In particular, the passage of policies by elected bodies that are seen to be legitimate might be understood to signal information about social norms among constituents.
Scholars have proposed a cyclical process that determines a democratic government's ability to function well. If the government is trustworthy then its citizens acknowledge this condition, cooperating with the policies and measures it implements. In turn, a government's trustworthiness is directly dependent on the well-being and flourishing of its citizens [48]. A trusted government, namely one that upholds an implicit social contract to serve the needs of its citizens, is in turn met with increased deference to and acceptance of its policies [49,50]. In this cycle, legitimacy and trust are often identified as key mechanisms [49,51,52]. Trust in the government can exert a powerful influence on citizens' willingness to cooperate with its measures. For instance, research suggests that liberals and conservatives who trust the government exhibit similar preferences regarding Federal spending and taxation despite differing political beliefs, showing trust to be a valuable resource for policy makers, particularly in ideologically contested domains [53].
Prior empirical research has found that formal institutions are able to shift norm perceptions through their statements and actions. For example, when the supreme court ruled in favour of gay marriage in 2015, there was an immediate increase in perceptions that Americans generally supported gay marriage, even in the absence of a change in personal attitudes [7]. When survey participants are told that partisan elites support a policy (or that the opposing party opposes it), they infer partisan norms, increasing perceptions that voters in the party generally support that issue or policy [54]. Relatedly, surprising referenda or election results can lead to rapid and widespread shifts in behaviours and opinions. For example, the election of Donald Trump in the USA increased individuals’ willingness to publicly express xenophobic opinions while also reducing the sanctioning of such views, especially in areas where those views were perceived to be more popular [29,30]. One mechanism by which these changes may happen is that the election of Donal Trump may have signalled the popularity of such prejudiced views among the American public. However, experimental evidence on the influence of signals from elected political institutions on norm perception is fairly sparse [1].
Institutions (much like individuals) are more likely to influence norm perceptions under certain conditions. A study run in the USA in 2018—a period of substantial government backsliding on environmental policy—found that commitments to climate action by coalitions of businesses and scientists were more effective in signalling public norms than a coalition of local governments [55]. Research has also found that individuals who identify with the source communicating the normative information have a stronger inclination to align their views and behaviours with that of the group (e.g. [56]).
In keeping with the literature on trust and institutional incentives, credibility of the institution and how accountable it is perceived to be to the public are likely to influence whether it is seen as a signal of public preferences [48,50]. Let us consider a national government that passes a policy into a law. If citizens believe this government represents the will of its constituents—that is, it is trusted and perceived to be legitimate—then we might expect that perceived support for related issues will increase as people learn that a policy has passed. Conversely, if a government is seen as corrupt and illegitimate, its decisions are unlikely to be taken as reflections of public will and may even signal the opposite.
(c) The current studies
We build on prior research by (i) providing causal evidence regarding the effect of the passage of new policies related to dynamic issues (i.e. climate change and clean energy transition) on perceived norms, and (ii) testing the hypothesis that the efficacy of institutional signals in communicating social norms depends critically on whether the institution is perceived to be accountable to the public. We focussed on the issue of renewable energy in the U.S.—a salient and rapidly evolving issue in American politics (e.g. the Inflation Reduction Act includes provisions relevant to renewable energy; [57]). Renewable energy is also a domain where Americans share widespread and systematic underestimation about the level of support for certain policies (see [5]), despite high actual levels of support [58]. In two large, pre-registered survey experiments with American respondents, we sought to answer the following questions: Does the passage of policies by state (Study 1) and national (Study 2) governments increase perceptions of support for related issues? Does such an effect vary if the policy's passage is attributed to the legislature or executive branch (Study 1)? Does the effect of policy passage on norm perceptions vary with information about whether the government is accountable to the public (Study 2)? Code, data and surveys for the studies are available on the Open Science Framework (OSF): https://osf.io/fvb6s/?view_only=7db3bc28b6834296b71f3f42780dada4.
2. Study 1
We first examined how policy signals, attributed to different branches of US state governments, influence perceived norms. Specifically, we examine whether information that a clean energy policy was enacted influences respondents' perceptions of public support for related issues. We identified US states that had passed mandates requiring that the state generate 100% of its electricity using only renewable energy sources in the near future. Notably, state level governments are fairly well-trusted (relative to the federal government; [59,60]) in the USA.
In exploratory analyses, we compared whether attributing the policy to different branches of government had differential effects on norm perceptions. We compared the executive and legislative branches, as well as the ‘State’, as the passage of policies is often attributed to any and all of these in popular media. Past research provides some basis for expecting that signals from different branches of government may have differential impacts on norm perceptions. For example, Thomas & Fink [61] found that a larger number of social models demonstrating a behaviour increases perceptions of norm prevalence more than a smaller number does. As such, actions by the legislative body—which is made up of a large number of representatives—may be interpreted as a stronger norm signal than the actions of a single individual—the governor. Moreover, evaluating the influence of different source frames offers practical insights for understanding potential indirect influences of political discourse and communication about public policy.
(a) Methods
The methods, sample size, hypotheses and analysis plan, including sample exclusions, were pre-registered: https://aspredicted.org/74Z_HKW.
(i) Participants
We collected data using Prolific, an online research platform. Based on an a priori power analysis, we determined that we would need a sample of 699 participants per cell to detect a small effect size (d = 0.15) with 80% power for the comparison of two treatment arms (i.e. comparing each experimental manipulation to the control) using a two-tailed test. With four experimental cells (legislature, governor, state, control) and adding roughly 7% for potential exclusions, we recruited a sample of 3000 participants from 13 states that passed legislation requiring that 100% of their electricity come from renewable sources. The states are: Arizona, California, Colorado, Hawaii, Illinois, Maine, Minnesota, New Mexico, New York, Oregon, Rhode Island, Virginia and Washington. We excluded participants who failed a basic attention check, resulting in 2982 participants. In this sample, 1488 participants identified as men, 1441 as women, and 53 as non-binary or self-identified as another category. The average age was 38.05 years (s.d. = 13.55).
(ii) Procedure
Participants were randomly assigned to one of four conditions. In three of these conditions, they were provided with a message based on true information about a recent renewable energy policy that their state had passed. In the Control condition (n = 742), they were not given this information and simply proceeded with the rest of the survey. Participants in the Legislature condition (n = 748) read that in their state ‘elected representatives from the legislature passed a policy mandating that [their state] generate 100% of its electricity using renewable energy sources (e.g. wind, solar, hydropower) in the coming years.’ While those in the Executive condition (n = 742) read that their state's ‘elected governor pushed through’ the same policy. Those assigned to the State condition (n = 750) simply read that their state passed this policy, without specifying the branch of government (see electronic supplementary material for the full survey text).
There were five key considerations in designing the policy statement. First, we aimed to use a level of detail to describe the policy that could be matched across states, thus we could not indicate by which specific year this mandate would come into effect. Second, we kept the policy signal portion of the message the same across conditions to ensure that any differences between conditions could be attributed to the branch of government implicated in each statement. Third, we emphasized the agency of different actors or branches of government in enacting the policy—so while saying a governor ‘signed’ a piece of legislation would be accurate, it is arguably agnostic as to whether one should attribute the passage to the governor. By contrast, ‘pushed’ clarifies the governor's agency. Fourth, the language used was consistent with how these branches affect policy (i.e. legislatures ‘pass’ laws, but governors do not—though they can exert influence on legislative priorities and policies). Fifth, we tried to maintain external validity by using common phrasing that one might see in news articles (hence the colloquial tone).
(iii) Dependent and Independent Variables
Main outcome: descriptive norms. Participants responded to the question: ‘What percentage of people living in [their state] do you think support a 100% renewable energy mandate for your state?’ Participants were provided with a free response field where they could enter a number between 0 and 100.
Exploratory outcomes. Participants were asked to express their personal support for a 100% renewable energy mandate for their state on a 7-point Likert scale (1 = Strongly oppose, 7 = Strongly support). They were also asked to estimate the percentage of people in their state who support a related political action and a related environmental policy: signing a petition to reach zero greenhouse gas emissions or support a carbon tax—a question borrowed directly from the Yale Program for Climate Change Communication (YPCCC) [62]. For both estimates, they were given a free response and asked to indicate a number between 0 and 100%. For the carbon tax item, subtracting participants' estimates from the YPCCC polling for each participants’ state (i.e. the real value) yielded a misperception score. Note that we could not do this for renewable energy mandates, which would be more directly related to our intervention, as this is not a question regularly asked by the YPCCC.
In addition to these items, participants also indicated whether they thought their state's legislature, governor, and their state's policies reflect the preferences of the state's population. Each measure was assessed on a 5-point Likert-type scale (1 = Not at all, 5 = Very much). Trust towards the state governor and legislature were measured using the same 5-point Likert scale.
(iv) Pre-registered hypotheses and analyses
We hypothesized that participants in a policy signal condition would perceive greater support for 100% RE among residents of their state as compared to those in the control condition (H1). We hypothesized that this effect would be present for each of the three policy signals, such that those in the Legislature condition (H2a), the Executive condition (H2b) and State condition (H2c) would each perceive greater public support for 100% RE when compared with the control in pairwise contrasts. Importantly, we had no a priori hypothesis about differences between the executive branch, the legislature and the state conditions.
Following our pre-registered analytical plan, we conducted mixed model linear regressions predicting norm estimates (and exploratory outcomes) using participants' state as a random intercept. The effect of any policy signal conditions (H1) or each specific policy signal condition (H2) to the control were assessed using dummy-coded variables for condition.
In exploratory analyses, we examined whether the condition effects were moderated by demographic indicators and whether they extended to other policy outcomes (signing a zero emissions petition and carbon tax misperception scores). Although not pre-registered as an exploratory hypothesis, we also explored whether the information treatments impacted trust towards the state legislature and governor, and the perception that they represent the preferences of the state's constituents.
(b) Results
(i) Effects on descriptive norms
Receiving a policy signal message increased perceptions of support for renewable energy when compared with the control, b = 2.43, 95% C.I. [0.88, 3.98], p = 0.002, d = 0.11 (Hypothesis 1). This effect was also present when comparing each information treatment to the control, including pairwise contrasts for H2a) Legislature versus control: b = 2.88, 95% C.I. [0.97, 4.77], p = 0.003, d = 0.11, H2b) Governor versus control: b = 2.01, 95% C.I. [0.11, 3.92], p = 0.038, d = 0.08, and H2c) State versus control: b = 2.40, 95% C.I. [0.50, 4.29], p = 0.013, d = 0.09 (Hypothesis 2). These effects remained significant when including covariates for education, age, income, political ideology and gender (see electronic supplementary material, tables S1 and S2 in the SOM). The differences observed represent a 2–3% shift in norm perceptions.
(ii) Exploratory analyses: additional effects of policy signals
No significant condition effect was observed for any of the following exploratory outcomes: personal support for renewable energy, perceived norms surrounding signing of a petition, perceived support for a carbon tax and misperceptions of carbon tax popularity (see electronic supplementary material, table S4 in the SOM). However, each of these items did positively correlate with our main outcome (see electronic supplementary material, table S3 in the SOM) and the effects are all in the hypothesized direction with marginally significant effects (p = 0.099) on personal support.
We also examined whether there were any differential effects on norm perceptions by attribution condition. Participants in all policy signal conditions (combined) reported greater endorsement of this view compared to the control. Pairwise condition effects suggest that this was largely driven by the Governor and State conditions (table 1). Additionally, in exploratory analyses, we found that only the State condition impacted trust, increasing trust in the state legislature and governor (table 1).
| estimate | 95% CI | p-value | d | |
|---|---|---|---|---|
| legislature represents population | ||||
| all policy signals versus control | 0.14 | 0.06, 0.21 | 0.003 | 0.13 |
| legislature versus control | 0.09 | −0.01, 0.18 | 0.062 | 0.07 |
| governor versus control | 0.10 | 0.02, 0.20 | 0.020 | 0.09 |
| state versus control | 0.21 | 0.12, 0.30 | <0.001 | 0.16 |
| governor represents population | ||||
| all policy signals versus control | 0.14 | 0.05, 0.21 | <0.001 | 0.11 |
| legislature versus control | 0.06 | −0.04, 0.16 | 0.263 | 0.04 |
| governor versus control | 0.14 | 0.04, 0.24 | 0.006 | 0.10 |
| state versus control | 0.19 | 0.09, 0.29 | <0.001 | 0.14 |
| state policies reflect population | ||||
| all policy signals versus control | 0.13 | 0.05, 0.20 | 0.001 | 0.12 |
| legislature versus control | 0.08 | −0.01, 0.18 | 0.080 | 0.06 |
| governor versus control | 0.11 | 0.01, 0.20 | 0.023 | 0.08 |
| state versus control | 0.19 | 0.10, 0.28 | <0.001 | 0.15 |
| trust in state governor | ||||
| all policy signals versus control | 0.13 | 0.00, 0.20 | 0.001 | 0.07 |
| legislature versus control | 0.03 | −0.09, 0.15 | 0.625 | 0.02 |
| governor versus control | 0.09 | –0.03, 0.21 | 0.154 | 0.05 |
| state versus control | 0.18 | 0.06, 0.30 | 0.004 | 0.11 |
| trust in state legislature | ||||
| all policy signals versus control | 0.08 | −0.01, 0.17 | 0.077 | 0.06 |
| legislature versus control | 0.05 | −0.06, 0.16 | 0.356 | 0.03 |
| governor versus control | 0.04 | −0.07, 0.14 | 0.502 | 0.02 |
| state versus control | 0.15 | 0.04, 0.26 | 0.006 | 0.10 |
(iii) Exploratory analyses: moderation by political orientation
In a pre-registered, exploratory analysis, we examined whether the effects of condition on perceived descriptive norms were moderated by participants' political ideology (measured on a 1–7 Likert scale, ranging from 1 = liberal to 7 = conservative). When comparing the presence of a policy signal to control, there was an interaction between information treatment and political orientation, b = −1,07, s.e. = 0.41, p = 0.023, R2 = 0.002, such that the effect was stronger for liberals than conservatives (figure 1). The effect of the message, b = 5.62, 95% C.I. [2.40, 8.83], p < 0.001, d = 0.13, and the effect of political ideology, b = −1.21, 95% C.I. [−2.00, −0.41], p = 0.003, R2 = 0.003, were also significant.
Figure 1. Expressive function of institutional signals by political ideology. Plot of policy signal treatment interacted with self-reported political ideology. The solid line indicates the policy signal treatment, while the dashed line indicates control. Grey shading indicates 95% confidence intervals. Political ideology is measured on a 7-pt Likert-scale from Liberal (1) to Conservative (7). (Online version in colour.)
(iv) State-level effects
In two exploratory analyses suggested by one of our reviewers, we assessed whether the level of support for renewable energy mandates in the state and the political leaning of the state attenuated or strengthened the effects of the policy signal. To test this, we estimated two multilevel models: an interaction between the treatment condition (policy signal present versus control) and state-level political leaning or the treatment condition and pre-existing norms around renewable energy policy as predictors.
To measure partisan leaning of the state, we used the percentage of the 2020 vote for Joe Biden in that state. This statistic was obtained from the CNN website.1 Measures of support for renewable energy policies were taken from the YPCCC.2 Using the average of three items that asked about support for renewable energy policies: 1. funding research into renewable energy sources; 2. requiring utilities to produce 20% electricity from renewable sources; 3. tax rebates for people who purchase energy-efficient vehicles or solar panels. Scores reflect the percentage of the adult population in a given state who at least somewhat support the three items.
In the model that includes the Biden vote share, we found significant positive effects of the policy signal and state political orientation, and a significant negative interaction between vote share and the policy signal treatment. These results suggest that the policy signal had a larger effect on perceived social norms in states with a smaller share of votes for Joe Biden (i.e. states with more votes for Republican candidate Donald Trump).
3. Study 2
Results from Study 1 suggest that when participants are exposed to a policy signal from their local government, they adjust their perceptions of social norms to be more aligned with the policy. While this experiment was tightly controlled and well-powered, the effect was also relatively weak, resulting in a 2.5 per cent average increase in perceived support for renewable energy mandates. This is unsurprising given the light-touch information intervention. It might also be the case that participants, including those in the control condition, may have had some awareness about the policy, which might weaken the effect of information. Exploratory analyses suggest that there may also be heterogeneous treatment effects. We found that the effect is larger among liberals than conservatives. This finding is in keeping with existing theory on the effectiveness of norm interventions: when individuals already share the opinion expressed by an institutional signal, they are more likely to shift their perceptions of social norms following exposure to that signal [1]. In this case, liberals, who tend to be more supportive of renewable energy (e.g. [63]), were also more likely to update their perceptions of social norms surrounding support for renewable energy in their state. Further, these effects were more substantial for those in more conservative states. Taken together, this might suggest the effect is largest for liberals in more conservative contexts who might, in the absence of such signals, be more pessimistic about public support for renewable energy.
In our second study, we sought to expand upon these findings by testing a potential psychological mechanism underlying the expressive function of policy signals. Specifically, we examined whether the belief that a government body is accountable to the public influences whether social norms are inferred from the policy signal. If policy passage signals social norms because people believe the government represents public preferences, then providing respondents with information from academic research highlighting the lack of accountability of the American government should weaken the effect. We examine this possibility at the level of national politics (i) to replicate results from Study 1 at a national level and generalize them to a related policy, (ii) because Americans’ view the federal government with greater skepticism (e.g. [59]), making it more feasible to manipulate perceptions of how accountable the government is to its constituents, and (iii) because of a recent empirical paper demonstrating that public opinion does not influence legislative decision-making in the USA [64], which formed the basis of our government accountability treatment.
(a) Methods
The methods, sample size, hypotheses and analytical plan for this study are pre-registered, https://aspredicted.org/S6K_G6C.
(i) Participants
We again collected data on Prolific. Based on a mixed model power simulation using ‘powerSim’ in the R package ‘simr’ with 2000 simulations and data from Study 1, we determined that to power a 2 × 2 interaction, we required a sample of 2100 participants in each of the four cells (i.e. a total N of 8400 participants). We recruited an additional 100 participants per condition to account for potential exclusions.
We excluded participants who failed a basic attention check item, as per our pre-registration; however, these participants were excluded directly on Prolific (mid-survey) and thus did not count towards our final N. Applying our additional pre-registered exclusion criteria resulted in 8654 participants. In this sample, 3947 participants identified as men, 4521 as women and 183 as non-binary or self-identified as another category. The average age was 38.07 years (s.d. = 13.60).
(ii) Procedure
Participants were randomly assigned to one of four conditions. We employed a 2 (policy signal versus no signal) × 2 (unaccountable government versus no information) between-subjects design. This design allowed us to: (a) conceptually replicate and generalize the results of Study 1 to federal policy signals, and (b) examine whether information about the accountability of the institution (i.e. the government) attenuates the expressive function of the policy signal. The policy signal, which was meant to describe the renewable energy component of the Inflation Reduction Act passed in 2022, was phrased as follows: ‘Recently, a national policy was passed by elected officials that includes the largest economic investment ever made to help the US transition to using renewable energy (e.g. solar and wind power) as its primary energy source.’
One-quarter of participants were also randomly assigned to an information condition describing a recent study that concluded that the U.S. government does not represent the will of its constituents [64]. This information was adapted from a website covering the study3 and included a figure using the data from Gilens & Page [64]. This figure is a line graph, with the x-axis showing popular support for a policy and the y-axis showing the likelihood that the policy is voted into law. A purple line shows how accountability would look in an ‘ideal’ democratic society, in which a government passes laws that reflect public preferences. A flat orange line shows the dramatically different actual observed pattern in the USA (see SOM for details). Participants were thus randomly assigned to one of four conditions: (a) policy signal + no information (n = 2211); (b) policy signal + unaccountable government (n = 2109); (c) no signal + no information (n = 2216); (d) no signal + unaccountable government (n = 2118).4
(iii) Dependent and Independent Variables
For our primary dependent variable, participants responded to the question ‘What percentage of people living in the USA do you think support using federal funding to help transition the U.S. to using renewable energy (e.g. solar and wind power) as its primary energy source?’ Participants were provided with a free response field, and could enter a number between 0 and 100.
For exploratory analyses, participants were asked to express their personal support for federal funding to help transition the USA to using renewable energy (e.g. solar and wind power) as its primary energy source, on a 7-point Likert scale (1 = Strongly oppose, 7 = Strongly support). Perceived norms and personal support were highly correlated, r = 0.47, p < 0.001.
In addition to this item, participants also indicated whether the USA federal government's policies reflect the preferences of people living in the U.S. on a 5-point scale, ranging from 1 = Not at all – 5 = Very much.
(iv) Pre-registered hypotheses and analyses
Our pre-registered hypothesis was that the effect of the institutional signal would increase the perceived support for investments into renewable energy among the American public (H1a), but that this increase would be attenuated by information about the lack of accountability of the government to its constituents (H1b-c,H2).
We ran linear regressions to predict perceived social norms (or exploratory outcomes) with dummy-coded variables comparing the effect of the policy signal + no information condition relative to all other conditions (H1a-H1c), or with the 2 × 2 interaction term included (H2). We ran analyses with and without demographic controls, which included age, gender and party identification (captured on a 7-point scale, with higher scores indicating identification as a Republican).
(b) Results
(i) Effect of policy signal on norm perceptions
Participants in the policy signal + no information condition perceived significantly higher support for federal funding of a transition to renewable energy compared to all other conditions, which supports H1a-H1c (table 2). Comparing the policy signal + no information condition to the no signal + no information condition, we see a perceived norm increase of around four percentage points. These results remain significant after accounting for gender, age and political ideology (see electronic supplementary material, table S8 in the SOM).
| hypothesis | Group 1 M (s.d.) | Group 2 M (s.d.) | comparison | effect size | per cent increase |
|---|---|---|---|---|---|
| H1a: policy signal + no information > no signal + no information | 52.75 (17.46) | 48.74 (17.66) | b = 4.01, p < 0.001, 95% C.I. = [2.98, 5.05] | d = 0.23 | 4.01% |
| H1b: policy signal + no information > no signal + unaccountable government | 52.75 (17.46) | 49.28 (17.75) | b = 3.48, p < 0.001, 95% C.I. = [2.43, 4.52] | d = 0.20 | 3.47% |
| H1c: policy signal + no information > policy signal + unaccountable government | 52.75 (17.46) | 51.02 (17.41) | b = 1.74, p < 0.001, 95% C.I. = [0.69, 2.79] | d = 0.10 | 1.74% |
To understand how information about lack of government accountability influences the expressive function of the policy, we ran a regression including the interaction of policy signal treatment with accountability information treatment. In this model, a significant signal by accountability interaction (b = −2.27, p = 0.003, 95% C.I. [−3.75, −0.79]) was observed, such that the effect of the policy signal was significantly smaller when accompanied by information about the lack of accountability of the U.S. government to its constituency in terms of passing policies that supports the public (figure 2).
Figure 2. Interaction between policy signal and information about lack of government accountability. Black indicates the no information condition, and grey indicates the lack of accountability treatment. Error bars are 95% confidence intervals.
(ii) Exploratory analyses
Consistent with Study 1, we find no significant effects of the treatments on personal levels of support (all ps > 0.05; see electronic supplementary material, table S9 in the SOM). Personal support was high for all conditions, with scores ranging from 5.54 to 5.61.
Participants in conditions without any information about accountability rated government information policies as more reflective of public preferences relative to the condition. Additionally, while not a manipulation check, participants in the Policy Signal + No Information condition scored significantly higher than those in the No Signal + No Information condition, suggesting that the policy signal itself may drive perceptions of accountability (especially when considering that support for the policy was high). See electronic supplementary material, table S10 in the SOM for these results.
To replicate the significant but exploratory interaction observed in Study 1 between political orientation and the policy signal, we ran a linear regression interacting political affiliation with a dummy for the policy signal & no information condition (1) relative to the no signal & no information (0). We did not find a significant interaction (b = 0.06, p = 840, 95% C.I. [−0.50, 0.61), failing to conceptually replicate the exploratory results from Study 1.
We replicated the same analyses looking at local political and normative context on the expressive function of the policy signal. To do so, we ran analyses only for participants in the no information conditions. We find significant effects of the policy signal after accounting for state-level norms (i.e. support of renewable energy policies) and political leaning of the state. Both state-level norms and Biden vote share predicted increased perception of descriptive norms in the policy signal condition. The interactions were significant in both models, showing that the effects of the policy signal were larger in states with a smaller Biden vote share (i.e. states with a higher vote share for Republican candidate Donald Trump) and with lower pre-existing support for renewable energy policies.
4. Discussion
Taking the recent passage of renewable energy policy in the USA as a case study, we show that policy signals can inform public perceptions of related social norms. In two well-powered and pre-registered experiments, we find that true information about the passage of a recent policy increases respondents' perceptions of public support for related issues. In Study 1, we find that this holds for state-level policy signals, regardless of whether the policy is attributed to the legislature, governor, or the ‘state’. Study 2 generalizes the findings of Study 1 to a national-level policy signal. We find that information about recently passed federal energy policy similarly increases perceptions of related social norms, in this case shifting the perceived support from a minority position (48.7%) to a majority position (52.7%). By randomizing exposure to an additional information treatment based on recent research, Study 2 shows that the expressive function of policy signals hinges on beliefs that one's government represents the will of the public.
Renewable energy policy, and climate change more broadly, are issues for which there is documented sizable and systematic underestimation of popular support, concern and associated actions [6,4]. Our results suggest that broadly communicating about the passage of policies that are in fact aligned with public preferences and behaviours may be one way of ameliorating misperceptions, with the potential for widespread impacts given the public nature and wide reach of policy signals.
Conversely, norm misperceptions may be created or reinforced when one's government does not act on issues that the public supports, as has largely been the case for climate change in the USA until the passage of the Inflation Reduction Act in 2022. Inaction by a governing body that is believed to represent the will of the public may suggest limited public appetite for action on that issue. Similarly, the passage of unpopular policies could lead the public to wrongfully infer a greater prevalence of support for that policy than exists. For example, there is widespread underestimation of support for reproductive rights, which may be reinforced by the 2022 Dobbs Ruling by the U.S. Supreme Court, which overturned the constitutional right to an abortion. Notably, informing people that one's government does not represent the will of the public can weaken the expressive function of policy, and thus curtail norm misperceptions in such situations, though there might also be other negative consequences that cannot be easily reversed.
The broader social and political context in which individuals are situated influences the effectiveness of policy signals. In exploratory analyses, we show that the policy signals were more effective at increasing perceived support in states with a greater Republican vote share in the 2020 Federal election and states with lower apriori support for renewable energy policies based on existing polling data. Confirmatory investigations of these effects are needed to ensure they are robust.
In summary, this work provides evidence for an indirect expressive or signalling function of policy that may reinforce its direct effects since we know from a large literature that people are often influenced by and act in accordance with the social norms they perceive in their surroundings [2,7]. Institutional signals are often broadly communicated, and there is shared knowledge that this is the case, suggesting that they may be a powerful and widespread signal of public preferences. At the same time, we show that their signalling function hinges critically on the perceived legitimacy or accountability of the institution. This research suggests two complementary intervention approaches to ameliorate norm misperceptions and help solve collective action problems: when popular policies are enacted but their support is underestimated, broadly inform people about the passage of these policies—and when popular policies are not enacted (or unpopular ones are) highlight the government's lack of accountability to the public in legislative decision-making in these particular instances.
(a) Limitations and future directions
Despite the use of large, highly powered samples and a pre-registered design and analysis, there are a few key limitations of this work that are worth noting. First, it is unclear how generalizable the findings observed here are to other populations or policy domains. That said, we do find that the expressive function of the policy signal generalizes across branches and levels of government. Furthermore, our findings on the psychological process provide a basis to suggest that these results might replicate in contexts where there is widespread belief that the government is accountable to the preferences of its constituents. In contexts where large publics perceive their governments to act with disregard for public preferences, we might assume a much weaker interpretation of newly passed policies as signals of social norms. Indeed, perceptions of local and national governments, including perceived legitimacy, accountability, corruption and trust, differ substantially across countries as well as within them. Similarly, the type of political system—whether it is a democracy, which kind of democracy, etc.—also shapes how accountable the government is to the public and differs across countries (e.g. [65]).
The USA may be somewhat unique when it comes to climate change and energy-transition-related policies as there are significant partisan differences in support for renewable energy (e.g. [63]). As such, we might expect that for partisan issues, the passage of a policy may only be seen to signal information about related social norms among co-partisans of the party in power at the time the policy was passed. Despite this possibility, we observe bipartisan effects of the policy signal. However, we also do see partisan differences, with effects stronger among liberals in Study 1 (so those whose views are presumably already aligned with the policy measure) and in contexts with higher Republican vote share, potentially because the signal is more surprising and informative in those contexts. One mechanism that could explain the bipartisan effects is that the public may infer not only that the policy signals existing social norms but also they might anticipate that it will quickly shift social norms. Research that varies the partisanship of the issue might find different effects.
A second limitation worth acknowledging is the potential for a demand effect in Study 2. Participants in the lack of government accountability conditions were told about historical research that finds that public preferences have little bearing on the policies enacted by the U.S. government. It is possible that participants' responses conformed to experimental demands. However, the instructions included language that we believe mitigates these concerns as they were written to ensure that participants did not feel pressure to respond in a given way (i.e. a reminder that their responses are anonymous to the researchers, and that the goal of the study was to understand their thoughts). Further, the main dependent variable, normative perception, focussed on participants’ thoughts about others rather than themselves (i.e. what percentage of Americans do you think…). Finally, the materials used in the government accountability interventions were based on and described as evidence from a 2014 paper by other scholars (i.e. not the present research team [64]). Notwithstanding these efforts, future work may benefit from measures to reduce demand effects, such as financially incentivizing participants for accuracy.
There are several additional paths for future research to expand on the current work. There are various factors that can affect the connection between the public and politicians, but also the misconnections (Sherman & van Boven, 2023). Future work might consider not only how policy signals shape perceived social norms, but also how government inaction on issues or the failure to pass a policy might decrease perceptions of related norms. Relatedly, work along these lines could consider how perceptions of social norms and of the legitimacy of government are impacted by the passage of unpopular policies, especially in cases where that discrepancy is communicated to the public via protests or media, as was the case with the 2022 Supreme Court ruling overturning the constitutionally protected right to an abortion.
Future work might also consider the influence of institutional signals on perceptions of how social norms are trending. Existing work on institutional signals focuses on static norms about policy support. However, it is possible that signals by formal institutions shape people's perceptions of the direction of change in social norms, including recent changes signalled by the passage of the policy and anticipated changes due to the introduction of the policy. Another promising avenue for future work is to consider the potentially differential influence of different types of policy signals on perceptions of social norms. Our work finds consistent effect sizes across two types of policies—one that includes restrictions (i.e. a renewable energy mandate) and one that focuses instead on incentives to promote renewable energy—and different branches and levels of government (i.e. governor, legislature, state, federal government). Although promising, these results could be expanded to assess the impact of policy signals spanning different issues and domains and that vary in whether they take a ‘carrots’ or ‘sticks’ approach, how aggressive they are in terms of goal setting (e.g. variation in deadlines for mandates) and the distribution of partisan or bipartisan support surrounding the issue.
This future work might also consider the effects of partisan policy signals, highlighting whether policies were passed or rejected by specific political groups, a factor that the current investigation did not address. This line of work could offer clarity on whether effects are stronger for co-partisan signals, and whether the expressive function of such partisan signals is circumscribed to members of that party. Additionally, research focussed on these topics might also aim to disentangle whether the passage of policy is understood to signal an increase in support among constituents, among members of a specific party, or rather the relative number of Democrats or Republicans. Understanding this may be important in the context of polarized issues because if the passage of a new policy is understood to signal the relative size of an aligned political group, this may instead increase resistance among the non-aligned group.
Our policy signals did not shift personal support or attitudes directly, which is consistent with prior research. Existing research on the influence of the 2015 Supreme Court ruling on same-sex marriage found that while the ruling increased perceived support for same-sex marriage among the public, it did not shift their personal attitudes [7]. Additionally, work on institutional functions [49,50] finds that governments that are perceived as legitimate and inspire trust are more likely to instill adherence to passed policies, suggesting that the actions of governments that represent the preferences of their constituencies are likely to have stronger indirect influence on perceived social norms but also direct effects on the behaviours and beliefs of the public. Thus it is likely that in the longer term personal attitudes and behaviours will change as well.
While our aim in these studies was to show a causal effect of policy signals, and we thus opted for a controlled and highly powered survey experiment, future research might aim for greater ecological validity. Designs that involve advertisements that remind participants of the institutional signal in settings outside of a survey or laboratory might be particularly effective. Further, these studies could also ‘close the loop’ by assessing not only the impact of institutional signals on perceived social norms but also on individual behaviours. Thus, we recommend that future researchers use these findings as a stepping stone to assess behavioural outcomes and the impact of policy signals on perceived social norms in more naturalistic settings [66].
5. Conclusion
We find that signals by formal, democratic institutions—specifically information about the passage of a policy—influences perceptions of social norms around related issues. This is true of state- and national-level policies, for policies attributed to the legislative branch, executive branch, or to the government, and for policies that introduce restrictions and incentives. We also find that the expressive function of policy signals—that is, their signalling of social norms—depends critically on the belief that the institution in question is accountable to public preferences in its legislative decision-making. Similarly to gun control and reproductive rights, renewable energy policy is a domain where widespread, large and systematic underestimations of public support have been documented, perhaps in part due to inaction or even outright rejection of progressive policies on such issues by governments. Our work suggests that as more renewable energy legislation passes, perceptions of associated social norms may increase. For those seeking to correct norm misperceptions, this work also suggests two courses of action: inform people that a policy has passed when the public underestimates support for issues related to that policy; or, when popular policies are not enacted or unpopular ones are, provide people with information about the government's lack of response to public sentiment in those instances.
Ethics
This study was approved by the Institutional Review Board at Northeastern University. All participants provided informed consent before participating in the study.
Data accessibility
The data, survey and R code are all available from the OSF repository: https://osf.io/fvb6s/ [67].
Supplementary material is available online [68].
Declaration of AI use
We have not used AI-assisted technologies in creating this article.
Authors' contributions
S.S.: conceptualization, data curation, formal analysis, methodology, writing—original draft; G.S.: conceptualization, methodology, supervision, writing—original draft; S.M.C.: conceptualization, data curation, funding acquisition, methodology, supervision, writing—original draft.
All authors gave final approval for publication and agreed to be held accountable for the work performed therein.
Conflict of interest declaration
We declare we have no competing interests.
Funding
We received no funding for this study.
Footnotes
Endnotes
4 Despite the random assignment in our design (see below), significantly more participants ( = 66.80, p < 0.001) were excluded in the conditions that included the unrepresentative government information (N = 177) compared to the conditions that did not include this information (N = 56). In absolute terms, this difference is about 1.5% of the total sample, which should greatly mitigate concerns about sampling differences. As an additional precaution, we reanalyse our data testing H1 and H2 including covariates for demographics to mitigate the possibility that this introduced sample differences (including controlling for political orientation, gender and age). These results are included in the SOM.
