Legitimacy from Decision- Making Influence and Outcome Favourability: Results from General Population Survey Experiments

Transcription

1 667956PSX / Political StudiesArnesen research-article2016 Article Legitimacy from Decision- Making Influence and Outcome Favourability: Results from General Population Survey Experiments Political Studies 2017, Vol. 65(1S) The Author(s) 2016 Reprints and permissions: sagepub.co.uk/journalspermissions.nav DOI: journals.sagepub.com/home/psx Sveinung Arnesen Abstract Democracies are typically considered more legitimate than other types of regimes because they allow the citizens to participate in the policy decision-making process. Others argue that the policy output matters most, and citizen influence plays a lesser role. This study presents two survey experiments on the micro foundations of these two sources of political legitimacy, thus contributing to an emerging literature that experimentally investigates the effects of democratic procedures in small-scale settings. Respondents who saw the decision going in their favour found the decision much more acceptable than the respondents who preferred another outcome. Conversely, decision-making influence generally did not serve as a legitimising factor among the respondents. This result supports the argument that citizens prefer a stealth democracy where they are minimally involved in democratic decision-making processes. Keywords input legitimacy, output legitimacy, survey experiment, stealth democracy Accepted: 11 July 2016 This study presents two survey experiments that aim to uncover the effect of perceived decision-making influence on receptivity to specific decisions. They also measure to what degree the favourability of the outcome matters to the evaluation of the decisions. As such, the study contributes to an emerging political science literature on experiments with decision-making processes and their impact on legitimacy beliefs (Esaiasson et al., 2012, 2016; Persson et al., 2013). These micro-level empirical studies take a novel and complementary approach in examining a central question in studies of democracy that tend to Department of Comparative Politics, University of Bergen, Bergen, Norway Corresponding author: Sveinung Arnesen, Department of Comparative Politics, University of Bergen, P.O. Box 7800, 5020 Bergen, Norway. sveinung.arnesen@uib.no

2 Arnesen 147 focus on the system level, investigating how political legitimacy is influenced by the policy makers willingness to listen to citizens preferences and their ability to deliver just and favourable policies, which is sometimes called input versus output legitimacy (Easton, 1965; Rothstein, 2012; Scharpf, 1999). The studies presented here construct specific decision-making situations that intend to compare input and output legitimacy from the citizens perspectives. I investigate what is more important to people when deciding how to spend a common good upon a single issue: that they feel they are influencing the outcome or that the outcome is consistent with their preferences. Two survey experiments addressing this issue are conducted in a general population web panel with >4200 respondents. The data-generating infrastructure for both experiments is the Norwegian Citizen Panel (NCP), an academically run general population web panel. Using real money, the respondents in the first experiment are faced with a situation where 5000 Norwegian Kroner will be spent either on the respondents themselves with one winner being drawn from the pool or donated to the charitable organisation Doctors Without Borders. The full factorial design has two treatment dimensions with two levels each. The first dimension splits the respondents in two, separating those who are told they can influence the decision from those who are told they do not have any influence. Regardless of their ability to influence the outcome, all respondents are asked to state which outcome they prefer. The second dimension again divides the respondents into two groups, where one group s preferred outcome occurs, and the other group faces an outcome opposite their stated preference. The second experiment supplements the first by presenting similar scenarios, but the second scenarios are hypothetical and not a real decision. Several vignette variations are tested on the respondents to measure whether changes in the framing of the situation or the treatment affect the respondents decision acceptance. The respondents clearly react more negatively to the decision when the outcome differs from their stated preference. Furthermore, whether the respondents have an influence on the decision does not have a significant impact on how they perceived the decision. The results are consistent when the respondents have a greater degree of influence on the decision or the question wording is changed. Although these results are suggestive rather than conclusive, they support observational studies that find weak or no evidence between decision-making influence and decision acceptance (Grimes, 2005; Hibbing and Theiss-Morse, 2002; Skitka et al., 2003). Thus, I ask whether emphasising direct citizen decision-making influence in political matters is misguided. Participants tend to be heavily skewed towards resourceful individuals (Rosenstone and Hansen, 2002; Verba et al., 1995), and their preferences may have weighed too heavily in the democratic debate at the expense of other dimensions that kindle perceptions of fair procedures. Input legitimacy is not only about direct decision-making influence, but it may also emerge from other types of procedures that assure the public that political decisions have been made in a fair and unbiased manner, such as procedural transparency and representation of relevant social and interest groups in the decision-making bodies (Tyler and Lind, 1992). This study first presents a theoretical framework for the experiment, yielding two hypotheses on how influence on a decision and its outcome affect its legitimacy. The two experiments are then presented with descriptions of their design, data-generating infrastructure and results. Next, the analysis section describes the statistical analyses of the experimental effects. The final section discusses the implications of the findings.

3 148 Political Studies 65 (1S) Decision-Making Influence and Outcome Favourability as Legitimising Factors Actions are facilitated when the affected individuals comply. Compliance can be achieved through the use of various forms of power, such as money, social status or the use of force. Another way of facilitating action occurs when the affected individuals expect and consent to the action taking place. The term legitimacy conveys this concept of wilful compliance towards an action. Max Weber defined legitimacy as a conviction on the part of persons subject to authority that it is right and proper and that they have some obligation to obey, regardless of the basis on which this belief rests (see Uphoff 1989). Norman Uphoff (1989: 312) formulates it that: [t]o accord legitimacy to a regime, to a role, to an incumbent, to a policy, or simply to the outcome of a decision process is to grant a type of political credit which can be drawn on by persons in authority. It is in the self-interest of the regime, institution or any other authority to maintain a relationship to its citizens that maximises its support among them. Without legitimacy, other sources of powers must be used to enforce citizen compliance. These powers are often undesirable for a democratic regime to apply and always costlier than when the people wilfully comply. One of the classical merits of democracy is that it yields more legitimate decisions than those made by individual or small groups of rulers (Dahl, 1989). When citizens have a say in decisions, either directly or through elected representatives, the regime is perceived as more legitimate because its policies have not been imposed on the public but are decisions of the people. This concept is a core element advanced by protagonists of participatory democracy and deliberative democracy (Elster, 1998; Habermas and Rehg, 1996; Pateman, 1975). Effective participation of and by citizens is a defining characteristic of democracy. Reconsidering the classic works of Jean-Jacques Rousseau and John Stuart Mill, Carole Pateman (1970) distinguishes two models of participatory democratic procedures. Rousseau propagates a direct participatory system in which citizen participation, first and foremost, is an aggregation of citizens independent preferences. In this model, deliberation among citizens is unwelcome because it may confound individuals independent preferences and empower those preferences that are perceived to be in the majority. Conversely, Mill portrays communication between the citizens as the main component for acquiring decision acceptance. Increased civic participation in decision-making has been argued to enhance democratic legitimacy (see Parkinson, 2006; Smith, 2009; Stoker, 2006). However, electoral studies indicate that participation alone is insufficient if the outcome is unfavourable. A consistent winner loser gap exists in satisfaction with democracy, where those who voted for the losing parties in national elections are generally less satisfied with the way democracy works compared to the electoral winners (Anderson et al., 2005; Dahlberg and Linde, 2016). Anderson et al. define the losers as the crucial players in the democratic game because they will find consenting to being governed by those with whom they disagree most difficult. As Richard Nadeau and André Blais (1993: 553) argue, losers support for the system requires the recognition of the legitimacy of a procedure that has produced an outcome deemed to be undesirable. In the end, the viability of electoral democracy depends on its ability to secure

4 Arnesen 149 the support of a substantial proportion of individuals who are displeased with the outcome of an election. Losers are produced with every political decision made, not only in the electoral context, and many have argued that the really significant source of legitimacy originates on the output side. When the authority is able to deliver what the people want, it will enjoy support (Dahlberg et al., 2015). In other words, policy outcome matters. The policy outcomes should be favourable to the affected citizens (Estlund, 2009) or at least perceived by the citizens as fair and unbiased (Hibbing and Theiss-Morse, 2001, 2002; Rothstein, 2009). Thus, a distinction in democratic theory exists between those who emphasise input legitimacy and those who focus on the relevance of output legitimacy. Input factors include whether citizens are able to express their preferences and can expect the authorities to listen and whether the decision-making procedure is fair that is, transparent, unbiased and unfolds as expected. Output legitimacy, on the other hand, comes from the authorities ability to deliver policies consistent with citizens preferences and/or that citizens perceive to be fair (Easton, 1965). The input output framework originally came from systems theory (Von Bertalanffy, 1956) and is now used within several subfields of political science. For example, the terms were introduced in European Union (EU) studies by Fritz Scharpf (1999, 1970), who defined output-oriented legitimacy as EU institutions ability to govern effectively for the people and input-oriented legitimacy as involving political participation by the people. 1 In parallel to the system-level literature in democratic theory, previous research on procedural and distributive fairness or justice originating from social psychology has studied input and output effects from a micro-level perspective (Lind and Tyler, 1988; Thibaut and Walker, 1975). These studies include measurements of influence and outcome favourability effects on the evaluation of decision legitimacy. Decision-making influence is regarded as a legitimising factor, as a subcomponent within the broader concept of procedural fairness or justice (Tyler, 2006; Tyler and Fagan, 2008; Tyler and Lind, 1992). In a meta-analysis of 89 articles on the procedural and distributive fairness or justice literature, Skitka et al. (2003) found that influence had a small but significant effect on how people evaluated an outcome. Outcome favourability in the procedural and distributive justice or fairness literature also affects how people evaluate an outcome (Muller, 1979; Skitka et al., 2003; Tyler, 2000; Van der Toorn et al., 2011). The outcome evaluation takes the form of a range of dependent measures, including decision acceptance, evaluation of the decision-making authority, and job performance. Several political science studies have approached the effects of democratic procedures from an empirical and experimental micro-level perspective (Karpowitz and Mendelberg, 2011; Thompson, 2008). Michael Morrell (1999: 294) experimentally examined citizens evaluations of participatory democratic procedures, concluding that the procedures by which participation takes place can significantly influence the effects of participation. Neblo et al. (2010) experimented with Americans willingness to deliberate and found that many are positively inclined to take part in such activities. More recently, a series of experimental studies investigated the effects of procedures on political legitimacy (Esaiasson et al., 2012, 2016; Persson et al., 2013). Persson et al. (2013) studied the effects of direct voting and deliberation on legitimacy beliefs in a field experimental study of small-group decision-making. They found that both voting and deliberation increased the legitimacy of a decision compared with a decision-making procedure where neither of these mechanisms were present. In another study by the same authors, a field

5 150 Political Studies 65 (1S) experiment mimicked decision-making arrangements in large-scale democracies (i.e. participatory democracy, participatory constitution-making and procedural fairness) and tested how these arrangements generated legitimacy beliefs (Esaiasson et al., 2012). Their main finding was that personal involvement in the decision-making process increased the legitimacy of the decision. Finally, Esaiasson et al. (2016) present a tripartite typology about how decision-makers can demonstrate responsiveness to the citizens in the decision-making process and generate higher levels of acceptance to listen, explain and adapt. Each of these actions signals to policy losers that their preferences were considered despite the unfavourable outcome. These novel experiments represent an important contribution to the study of the mechanisms behind political legitimacy, as they allow for the sources of legitimacy to be studied in isolation. In experiments, the researcher has direct control of the variables in question, that is, the variables of interest are manipulated by the researcher while keeping other factors constant (Shadish et al., 2002). This study seeks to build on these recent contributions to experimental knowledge about decision-making procedures and legitimacy. In particular, this study shares many features with Esaiasson et al. s (2016) study, where the goal is to combine input and output factors that are thought to increase decision acceptance. Reformulated into hypotheses, the theoretical discussion above leads us to hypothesise that (1) an outcome that is favourable to an individual increases the legitimacy of the decision and (2) having an influence on a decision increases the legitimacy of the decision. An added value of our study is that the participants are representative of the population, and the study is comparatively large scale. The data-generating infrastructure is the NCP, a web-based survey of Norwegians opinions towards important societal matters (Ivarsflaten et al., 2014a, 2014b, 2015). 2 The panel participants represent a cross section of the Norwegian population randomly selected from the national registry, and they are invited several times per year to provide their opinions on important questions regarding Norwegian society and politics. As the experiments are conducted on a general population panel, the respondents answers reflect the decision acceptance of young and old, men and women, rich and poor. The fact that the study reflects society is of particular importance on an issue such as political legitimacy, where the target population is the entire citizenry. Although previous studies have focused on the effects of variations in decision-making procedures on legitimacy beliefs, this study is concerned with the generic dimensions of influence on a decision and whether the outcome is favourable. As will become clear when the experimental design is presented, the type of influence is closer to the direct participatory model than the deliberative model because participants have no opportunities for communication with each other. The type of influence the participants can have on the outcome is also considered, and participants have the perception that their preferences are taken into account before the decision is made. In the second experiment, the influence dimension takes a more instrumental character and envisions scenarios where the participants are both heard and influence the outcome. This study presents results from two experiments; the second experiment builds on the first. The first experiment is a real-money experiment completed in two survey waves, allowing time for an actual decision to occur between waves. The respondents in the first wave were asked their preferences on how to spend a sum of money they were given, and they learn the outcome of the decision in the second wave. The second experiment has a similar general organisation. We test the robustness of the findings in the first experiment by varying question formulations and treatment dose.

6 Arnesen 151 The Experimental Designs In the first experiment, the respondents are informed about a concrete decision that is about to be made. Specifically, they are presented with a situation where money will be distributed either to the respondents in the form of a randomly selected winner of a travel gift card or donated to the charitable organisation Doctors Without Borders. At time t1, a randomised half of the respondents are informed that they can influence the decision by stating their preference for the outcome: A researcher linked to the Citizen Panel has NOK 5,000 at his disposal. These funds will either be drawn as a gift card for one of the Citizen Panel s participants, or they will be donated to the charitable organisation Doctors Without Borders. You as a participant have a say in this decision, and it is therefore interesting to know your preference for how this money should be used. What do you prefer? A prize draw to select one Citizen Panel participant who will receive a NOK 5,000 gift card A donation of NOK 5,000 to Doctors Without Borders No opinion Members of this group the treatment group are told they can influence the outcome by stating their preference about how to spend the money. The phrase have a say in this context implies that the decision-maker will listen to the preferences of the respondents, without promising that the majority position of the respondents will be adopted. As Esaiasson et al. (2016) demonstrate, this type of influence may be as effective in winning decision acceptance as having an instrumental voting influence on the outcome. The control group receives an almost identical text, except they are told that their preferences will not be considered in the decision-making process. The respondents answers are recorded. At time t2 the second survey round they are reminded of the question at time t1. The respondents are again randomised and receive two different but equally true messages about the outcome of the decision. 3 One group is informed that the money went to Doctors Without Borders, and the other group is told that a lucky respondent will receive it. The experiment thus constitutes a 2 2 full factorial design. One dimension is the ability to influence the decision, with the levels yes and no. The other dimension is the favourability of the decision outcome. Note here that the two levels on this dimension are not how the money is spent but rather whether the way the money was spent accorded with the respondents stated preferences at time t1 ( yes or no ). Having been informed about the decision outcome, the respondents are asked to rate how acceptable they found the decision in the following manner: 4 It has now been decided that NOK 5000 will be given to [Doctors Without Borders]/[a randomly selected respondent]. To what degree do you find this decision acceptable? 1. To a very high degree 2. To a high degree 3. To some degree 4. To a low degree 5. Not at all The legitimacy concept is here operationalised as the degree to which the decision can be accepted in recognition that the term legitimate is not commonly used among the

7 152 Political Studies 65 (1S) general Norwegian population. In Norwegian, akseptere (accept) can mean to avfinne seg med (comply). In this respect, the term should serve as a satisfactory operationalisation of the definition of legitimacy as previously defined. It is also consistent with an operationalisation in other studies on legitimacy (e.g. see Grimes, 2005; Van der Toorn et al., 2011). As mentioned, listening to the people s will can be a more effective responsiveness strategy for gaining acceptance for an outcome than promising to follow majority opinion. However, testing whether a more instrumental form of decision-making influence would have a positive impact on decision acceptance would also be fruitful. Given that the first experiment is a real-money experiment, introducing several vignette variations using real-money payments for each situation would be too expensive. Thus, the first experiment is accompanied by a second, hypothetical experiment in which it is possible to introduce a larger set of vignettes. Hainmueller et al. (2014) demonstrate that survey experiments perform remarkably well in recovering the effects of a comparable natural experiment, so the second experiment serves as a sound test for replication and robustness of the findings in the real-money experiment. 5 The benefits of following the real-money experiment with a hypothetical one is that we can test whether some specific question wording or framing is driving the results. For instance, one could object that the respondents did not remember whether they had an influence on the decision, although they were presented with the question wording from t1 when they received the outcome at t2. In the second experiment, the outcome is presented in the same wave and directly after they have been presented with the question. This follow-up experiment is a full factorial vignette experiment where the respondents are presented with the same scenario as in the first experiment. Complementary to the situation in the real-money experiment, some hypothetical scenarios add another phrase to emphasise that the respondents have full ownership of the money. The follow-up experiment tests whether such changes in the design will affect decision-making influence. Therefore, the wording used for half of the respondents was changed to The panelists receive NOK 5000 at their disposal as compensation for participating in the Norwegian Citizen Panel, whereas the other group receives the same wording as in the first experiment: One of the researchers linked to the Citizen Panel has NOK 5,000 at his disposal. Furthermore, we introduce a more instrumental operationalisation of the influence treatment in the second experiment. Two levels on this dimension are identical to the first experiment to ensure comparability between the two experiments. One group of respondents is told that they have no influence on the decision and another that they do have an influence. For the remaining four groups, the degree of influence is explicitly stated by indicating that they are among a limited number of people who will decide how to spend the money. One group of respondents is told that they are 1 of 1000 decision-makers. Another group of respondents is told that they are 1 of 100, the penultimate group learns that they are 1 of 10, and the last group is told that they are 1 of 3 people to decide on the matter. By explicitly stating the number of decision-makers, we are able to intensify the instrumental decision-making influence. The following section presents the combined results from both experiments. Outcome Matters but Influence Does Not Table 1 displays the average scores for how acceptable the respondents found the decision. 6 The higher the score given, the more the respondents found the decision acceptable.

8 Arnesen 153 Table 1. Real-money experiment. Overall Influence No influence ATE influence Favourable outcome 4.63 (781, 0.024) 4.64 (380, 0.034) 4.62 (400, 0.034) Unfavourable outcome 2.62 (780, 0.042) 2.57 (394, 0.059) 2.68 (386, 0.059) Overall 3.59 (774, 0.049) 3.66 (786, 0.047) 0.07 ( 0.99) ATE outcome 2.01 (41.5) Acceptability scores by decision-making influence and outcome favourability. Cells show mean scores on a unipolar scale where respondents rate to which degree they find the decision acceptable, from 1 = Not at all to 5 = To a very high degree. Number of respondents and standard errors, respectively, are shown in parentheses. The average treatment effects (ATE) are displayed for the main treatments, with t-values in parentheses. The four rows represent the 2 2 design, with two treatment dimensions and two levels in each dimension. This table demonstrates a significant difference between the two groups on the outcome favourability dimension. As noted, this dimension separated respondents who had their wishes fulfilled regarding the decision outcome from those who did not. When the decision went against their preferences, the respondents reacted strongly. On a scale from 1 to 5 where 1 is Not at all and 5 is To a very high degree the group who received a favourable outcome scored an average of 4.63, but the group that did not receive their preferred outcome scored an average of On a 5-point scale, a difference of two points is quite remarkable. The result that outcome favourability increases legitimacy is consistent with the earlier theoretical discussion. As such, it is not a surprising result, although the magnitude of the effect may appear larger than anticipated. The lack of effect from the decision-making influence was more surprising. The results indicate that respondents who had an influence on the outcome did not regard the outcome as more acceptable than those who did not have an influence. The average rating that the respondents who had an influence on the decision gave was 3.59 (Table 1), which was not significantly different from the other group s average of No significant interaction effects were measured between the influence and outcome favourability treatments. Among the respondents who did not receive their preferred outcome, the mean score was slightly lower for those who had an influence on the decision, but this observed difference is likely coincidental. Therefore, we subsequently focus on the main treatment effects. To check whether the result could be attributed to some features of the experimental design, we proceed directly to analysing the second experiment. The replication of the first experiment with identical question wording indicates that the effects of both the outcome favourability and influence dimensions are very similar in the two experiments. As in the first experiment, the respondents react negatively when the outcome does not go in their favour. The same is true for the difference between having an influence and not having an influence on the decision. Hence, the effects we find in the vignette variations of the hypothetical experiment validate the original experiment (Table 2). In fact, and somewhat curiously, the respondents are equally unhappy with an imagined decision outcome they did not prefer as with a real decision. Again, outcome favourability is the dimension that clearly changes the respondents perceptions of legitimacy. When the outcome is consistent with their preferences, they rate the decision at 4.68 on a scale from 1 to 5, where 5 is the best possible score. When they oppose the outcome, the mean score

9 154 Political Studies 65 (1S) Table 2. Comparison of treatment scores in real-money and hypothetical experiments. Real-money experiment Hypothetical experiment Difference of mean (95% confidence interval) t-value Unfavourable outcome ( 0.26, 0.04) 1.0 Favourable outcome ( 0.13, 0.03) 1.0 Influence ( 0.33, 0.17) 0.7 No influence ( 0.31, 0.27) 0.2 Comparison of the first experiment and the second vignette with exactly the same question wording as in the first experiment. Cells show mean scores on a unipolar scale where respondents rate to what degree they find the decision acceptable, from 1 = Not at all to 5 = To a very high degree. t-test statistics calculated with the Welch two-sample t-test. Figure 1. Decision acceptance depending on outcome and perception of influence on decision. drops to 2.73, almost two points. Figure 1 indicates the magnitude of the effect. This figure presents the mean scores of the groups and their 95% confidence intervals calculated from a t distribution. In contrast to the significant effects of outcome favourability, decision-making influence seems to have no effect on the respondents evaluation of the decision. Rephrasing the vignette so that the respondents feel more ownership of the money does not change this result. As previously described, the original experiment wording specified that a researcher has some money at his or her disposal and asks how the respondents think it should be spent. By changing the wording to emphasise the respondents ownership of the money, respondents could imagine that the money was theirs and form stronger preferences about how it was spent. Framing the money as compensation for participating does not have an effect on the outcome, as the average treatment effects are estimated to be ± Furthermore, as the graph displays, the effects of influence are similar in all the treatment groups. The only group that deviates are the respondents told that they are one of three groups to decide. However, this group s confidence interval also overlaps all the other groups, including the respondents who were told they had no influence on the decision. The average treatment effects are small and fail to achieve statistical significance.

10 Arnesen 155 Discussion The lack of effect that decision-making influence has on the acceptance of a decision defies the expectations stated in the theoretical discussion and the experiments on the effects of voting and decision-making procedures on legitimacy beliefs referenced in that section. Those studies were designed somewhat differently, which may or may not explain the differences. Several factors varied between the experiments. They were conducted in different contexts and countries, with different subject pools and with different operationalisations of the dependent variable. Our experiment was a survey experiment on a representative sample of the Norwegian population, and the earlier decision-making experiments were field experiments conducted on Swedish high school pupils. In Peter Esaiasson s (2010) study, the degree of involvement was higher among the participants. The high school students were involved in direct voting on the outcome and face-to-face deliberation activities with each other. The intent in recruiting high school pupils as participants in the experiment was to study individuals who were forced to take collective responsibility for the consequences of their decisions, thus mimicking large-scale democratic decision-making. This contrasts with our survey respondents involvement, in which no interaction with other respondents occurred before the decision was made. The participants in the high school field experiments were also more homogeneous in background, while our survey respondents were more diverse in terms of age distribution and area of residence. The field experiments are more time-consuming and demand more effort from the subjects in the decision-making process than is required from respondents in a survey experiment. When Persson et al. (2013) find evidence of increased legitimacy from non-deliberative decision settings in which the subjects have a direct influence on the outcome, the difference in the experimental result might be due to the higher degree of involvement that naturally occurs through the differences in field and survey experimental designs. Esaiasson et al. s (2016) study is the most similar study to our research, as it is a survey experiment that investigates decision acceptance both in relation to outcome favourability and variations of citizen input into the decision-making procedure. Both the studies find that outcome clearly matters most. However, the respondents in that study increase their acceptance for the decision when the hypothetical decision-making scenario includes a component where the decision-makers listen to citizens preferences. Several reasons for this difference in findings are possible. One likely explanation is the different levels of treatment effect priming. Their vignettes were very explicit about the actions the decision-makers were taking, signalling loud and clear that the politicians had been listening and explaining during the decision-making process or that they had not (Esaiasson et al., 2016: 10). Our two survey experiments are subtler in comparison. Thus, differences in experimental designs may be why we found little evidence of increased decision acceptance among those who were told they had an influence, although both research designs are justifiable. Although individual-level experimental studies are better at detecting mechanisms and causal relationships than observational studies tend to be, they are less accurate in reflecting where the actual political authority operates and its legitimacy is assessed. Therefore, limits exist on how extensively the implications of these results of single collective decisions can be extrapolated to judge mechanisms of decision acceptance of political authority. External validity is the issue of concern for most social science experiments, and it is relevant also in our case to condition the implications of the results on the context in which they were produced. With such qualifications in mind, we nevertheless argue that

11 156 Political Studies 65 (1S) the experiments indeed do have value for our knowledge about the mechanisms that influence political legitimacy at the general level they are intended to capture. The experiments are about making collective decisions about how to spend a shared resource, and the subjects involved represent a cross-section of Norwegian citizens. The major difference in the acceptance rate between those whose preferences accorded with the outcome compared to those whose did not indicates that respondents did care about the issue that was put forth. When measured in isolation, decision-making influence is strongly dominated by outcome favourability in our setting, that is, output legitimacy trumps input legitimacy. The results from the two experiments differ somewhat from previous field experiments but are consistent with other studies that find a weak or no relationship between legitimacy and decision-making influence (Skitka et al., 2003). They support those who argue that decision-making influence is rather unimportant for political legitimacy or at least less important than is commonly perceived. In an in-depth study of the modernisation of a Swedish railway line, Marcia Grimes (2005) asks whether influence or procedure matters most to citizens in collective decision-making, and her answer is clearly procedure. It matters whether a decision has been made consistent with unbiased and transparent procedures, but it is less important whether the citizens have been able to influence the process themselves. Similar findings lead John Hibbing and Elizabeth Theiss-Morse (2002) to argue that American citizens want what they call stealth democracy, that is, a decision system that is built on democratic ideals, but in which they need not take part except in unusual circumstances. The average person, they say, does not want to be involved in the decision-making processes. They want the political system to work, the processes to be fair and the policy outcomes to be favourable to them or at least perceived as fair, but they do not want to participate in decision-making themselves. Thus, in the authors words, why should getting people to do something they do not want to do make them feel the system is more legitimate? (Hibbing and Theiss-Morse, 2002: 5). Thus, a timely question to ask is whether citizen involvement in decision-making procedures may have been overemphasised with regard to their legitimising potential. Has citizen participation been high on the democratic agenda because people who take an interest in politics find it important to be part of the decision-making, while other citizens really cannot be bothered? We know that political and civic participation varies considerably between groups of different socio-economic background, skills, motivations and networks, and it is heavily skewed towards those with high scores on political interest and generally high social resources (Lijphart, 1997; Rosenstone and Hansen, 2002; Strømsnes, 2001; Togeby, 2003; Verba et al., 1995). Indeed, Katarina Eriksson (2007) states that the research on political participation and socio-economic resources is so unambiguous that it may be classified as an empirical truth: Those who have resources are more actively participating in society, influencing politics and other social arenas. Key background variables that explain participation tend to be education, age, gender and to some extent income. Participation is also associated with skills and motivations such as political efficacy, political sophistication and political interest. Can it be that these factors connected to political and civic engagement also divide the population with regard to how they appreciate having an influence on a decision-making process? Making the fair assumption that political scientists and politicians are interested in politics more than average citizens, does this group of people grant other people attitudes they do not possess? The conceptual leap is short: Having an influence on the decision increases the legitimacy of the decision among those who tend to be decision-makers in society and vice

12 Arnesen 157 versa. Those who do not participate in society also do not value having influence as a legitimising factor. A recent study by Åsa Bengtsson and Henrik Christensen (2014) of the Finnish Election Study is relevant to this idea. The authors find that the citizens have heterogeneous democratic ideals. Separating democratic conceptions into three types (representation, expertise and participation), the authors find that those who favour participatory democracy have legitimacy as their central democratic value. The central value of those who favour representative democracy is accountability, whereas those who prefer to let experts decide are mostly concerned with output quality. Their actions also correspond to these attitudes so that the participatory democrats are engaged extensively in many forms of political participation, whereas the representative democrats merely vote in elections and expert democrats display a lower level of engagement in all forms of participation. These findings are supported by Carolina Johnson (2015), who finds that people who report greater participation in any of a range of local public decision-making activities allocate stronger importance to the values of the democratic process. One of the central critiques against participatory democracy is that those who engage are those who possess greater than average resources in the form of time, money, skills and networks. If these are also the same people who propagate participation to increase the legitimacy of collective decisions, they may be suffering from a wishful thinking bias (Babad, 1997). This view is a somewhat controversial inference to make, but it is not the first time it has been suggested. Hibbing and Theiss-Morse (2002) argue that a gap exists between what the elites think the people want and what most people actually want. Perhaps it is all wishful thinking, as Hibbing and Theiss-Morse suggest, where the values of those interested in politics are projected onto those who are less politically engaged. There might be good reasons to let citizens participate in political decision-making in between elections. One potential reason is that when citizens are involved, the policy decisions tend to improve because the decision-makers are then sensitive to citizen preferences. For instance, the core argument of deliberative democracy is that people need to participate in political discussions so that the best arguments will surface and prevail over inferior arguments. From this perspective, citizen participation is a necessary prerequisite for the deliberative process. However, the important distinction is that citizen participation is only a tool to enhance policy outcomes; it is not a goal in itself. This analytical distinction is easily blurred. If little or no political legitimacy can be gained from including people in the decision-making process, the implication is that citizen participation should not be introduced into political processes for its own sake. Participation may have been credited as a source of legitimacy when actually it is the consequence of participation better policy outcomes which is the real source of legitimacy. Could citizen participation have benefitted from being associated with good policy outcomes and then mistakenly have become a goal in itself? It is a question worth asking. When decision acceptance increases, the reward emerges in the form of a more compliant citizenry that in turn makes policy implementation more efficient. Although making policies that are favourable to as many as possible is important, political decisions are bound to produce some winners and some losers, and it is the losers we must care about. We have observed the considerable gap in decision acceptability in the two experiments, which is consistent with much of the research on outcome favourability. Accepting a decision consistent with one s own preferences is easy, but accepting an unfavourable outcome is not easy. Given that unfavourable political outcomes are inevitable, how can the losers misfortune be alleviated? The results from this experimental study on a representative sample of the Norwegian population indicate that decision-making influence makes

13 158 Political Studies 65 (1S) no difference with regard to decision acceptance. However, input into the democratic process is not only about direct citizen influence. As Tom Tyler (2001) explains, legitimacy is rooted in citizens assessments of the fairness of the decision-making procedures used by authorities and institutions. People value fair treatment by others because fair treatment communicates respect and dignity, encourages a feeling that authorities are neutral and generates trust in the motives of the authorities with whom one is dealing (Tyler, 2001: 427). Decision-making procedures that ensure fair treatment, such as transparency, and representation of interests and social groups in the decision-making bodies should, therefore, also matter in the legitimacy of a decision. As most policy decisions are made by elected representatives or appointed experts who never consider allowing public consideration, much could be gained from studying input legitimacy beyond perceived direct decision-making influence, experimentally investigating how other types of fair procedural features relate to the willingness to accept a decision. Acknowledgements Earlier versions of this article have been presented at the 2014 NOPSA Conference in Gothenburg, the 2014 ECPR General Conference in Glasgow, and to the Citizen, Opinion and Representation research group at the Department of Comparative Politics, University of Bergen. I would like to thank the participants there for valuable feedback and the anonymous reviewers for their insightful comments. I would also like to thank the panel respondents at the NCP who, through their participation in the experiments, made this research possible. All errors and omissions remain the responsibility of the author. Funding The author(s) received no financial support for the research, authorship and/or publication of this article. Supplementary Information Additional supplementary information may be found with the online version of this article. Table A1: Statistics on the experimental dimensions and their levels Figure A1: Group balance gender 1st experiment Figure A2: Group balance gender 2nd experiment Figure A3: Group balance political interest 1st experiment Figure A4: Group balance political interest 2nd experiment Notes 1. Throughput legitimacy was introduced later, referring to the efficacy of the institutional rules and processes, their accountability and transparency (Schmidt, 2013). This concept makes sense to the extent that fair procedures are important both in the decision-making process (input) and during the implementation of the outcome (output). 2. Data and codebooks are freely available to all academics upon request to the Norwegian Social Science Data Archive. 3. The respondents must be randomly designated to the two potential outcomes to maintain a proper experimental design. If only one of the potential outcomes occurs, we cannot know whether those who prefer that option find the decision more acceptable because the outcome is favourable or because of some other difference between those who prefer a charitable organisation and those who prefer the gift card. Thus, the wording about how the respondents influence the decision must be somewhat vague so that they are not being promised a type of influence they do not have. With this format, the respondents are not being deceived because if one of the options is preferable to none of them, it will not occur. However, in practice, their influence on the outcome is limited to such an extent that it might be more correct to speak of perceived influence rather than actual influence. 4. Note that the scale is reversed in the analysis section to allow for a more intuitive interpretation of the results.

14 Arnesen Care was taken to ensure that no survey panelists who participated in the first experiment took part in the second experiment. 6. Most respondents preferred to give the money to Doctors Without Borders, but a small minority preferred that the money be given to the respondents as a gift card. To comply with ethical standards and minimise deception, both Doctors Without Borders and a randomly drawn respondent received NOK 5000 from the experiment. 7. See Table A1 in the online Appendix for details. References Anderson CJ, Blais A, Bowler S, et al. (2005) Losers Consent: Elections and Democratic Legitimacy: Elections and Democratic Legitimacy. Oxford: Oxford University Press. Babad E (1997) Wishful Thinking among Voters: Motivational and Cognitive Influences. International Journal of Public Opinion Research 9 (2): Bengtsson Å and Christensen H (2014) Ideals and Actions: Do Citizens Patterns of Political Participation Correspond to Their Conceptions of Democracy? Government and Opposition 51: Dahl R (1989) Democracy and Its Critics. New Haven, CT: Yale University Press. Dahlberg S, Linde J and Holmberg S (2015) Democratic Discontent in Old and New Democracies: Assessing the Importance of Democratic Input and Governmental Output. Political Studies 63: Dahlberg S and Linde J (2016) Losing Happily? The Mitigating Effect of Democracy and Quality of Government on the Winner-Loser Gap in Political Support. International Journal of Public Administration 39 (9): Easton D (1965) A Systems Analysis of Political Life. New York: Wiley. Elster J (1998) Deliberative Democracy, vol. 1. Cambridge: Cambridge University Press. Eriksson K (2007) Spelar adressen någon roll? En studie av områdeseffekter på medborgares politiska deltagande. Umeå: Umeå University. Esaiasson P (2010) Will Citizens Take No for an Answer? What Government Officials can do to Enhance Decision Acceptance. European Political Science Review 2 (3): Esaiasson P, Gilljam M and Persson M (2012) Which Decision-Making Arrangements Generate the Strongest Legitimacy Beliefs? Evidence from a Randomised Field Experiment. European Journal of Political Research 51 (6): Esaiasson P, Gilljam M and Persson M (2016) Responsiveness beyond Policy Satisfaction Does It Matter to Citizens? Comparative Political Studies. Epub ahead of print 27 January. DOI: / Estlund DM (2009) Democratic Authority: A Philosophical Framework. Princeton, NJ: Princeton University Press. Grimes M (2005) Democracy s Infrastructure. Göteborg: Department of Political Science, Göteborg University. Habermas J and Rehg W (1996) Contributions to a Discourse Theory of Law and Democracy. Cambridge: Polity Press. Hainmueller J, Yamamoto T and Hangartner D (2014) Do Survey Experiments Capture Real-World Behavior? External Validation of Conjoint and Vignette Analyses with a Natural Experiment. Working Paper. Available at: Hibbing JR and Theiss-Morse E (2001) Process Preferences and American Politics: What People Want Government to Be. American Political Science Review 95 (1): Hibbing JR and Theiss-Morse E (2002) Stealth Democracy: Americans Beliefs about How Government Should Work. Cambridge: Cambridge University Press. Ivarsflaten E, Andersson M, Arnesen S, et al. (2014a) The Norwegian Citizen Panel: Wave 1 Data. Bergen: Norwegian Citizen Panel. Ivarsflaten E, Andersson M, Arnesen S, et al. (2014b) The Norwegian Citizen Panel: Wave 2 Data. Bergen: Norwegian Citizen Panel. Ivarsflaten E, Andersson M, Arnesen S, et al. (2015) The Norwegian Citizen Panel: Wave 3 Data. Bergen: Norwegian Citizen Panel. Johnson C (2015) Local Civic Participation and Democratic Legitimacy: Evidence from England and Wales. Political Studies 63: Karpowitz CF and Mendelberg T (2011) An Experimental Approach to Citizen Deliberation. In: Druckman JN, Green DP, Kuklinski JH, et al. (eds) Cambridge Handbook of Experimental Political Science. Cambridge: Cambridge University Press, pp