1. Introduction
An organization is an environment where institutions play an important role in structuring members’ behavior.Footnote 1 In this article, we combine notions from economics, sociology, and psychology to investigate institutional change within organizations. We consider both formal institutions (like formal rules) and informal institutions (like social norms)Footnote 2,Footnote 3 and employ Ostrom's (Reference Ostrom1990: 140) definition of institutional change as “A change in any rule affecting the set of participants, the set of strategies available to participants, the control they have over outcomes, the information they have, or the payoffs.” In particular, we study whether and how individuals change the formal rules of their organization. We focus specifically on the role of pre-existing formal rules, social norms, and social values in the process of this change.
In line with both the bounded rationality perspective and analytical sociology, our conceptual framework is based on the assumption that individuals act in ways that find optimal outcomes while being constrained or enabled by rules, norms and values. In other words, our starting point is that (institutional) structure and agency are not opposing forces but, instead, interdependent and mutually constitutive (e.g. Giddens, Reference Giddens1984), and that “the interplay of both is required to understand how institutions are formed and sustained” (Hodgson, Reference Hodgson2006: 19). The environment we consider is one where a team of individuals needs to cooperate in order to achieve a collective outcome that benefits them all.Footnote 4 The benefits of the collective outcome are available to everyone in the team, independent of the individual cooperation decision. Hence the team faces a social dilemma.
Our framework combines three levels of analysis. First, we investigate the effects of institutional structure on agency (the “macro-to-micro link”: Coleman, Reference Coleman1990) by focusing on how formal rules affect behavioral outcomes. The formal rules we consider are “cooperation rules,” which prescribe individual cooperation to achieve a team's collective goal. In this way, organizational rules may guide and structure the interaction within teams (Sewell, Reference Sewell1998). On the one hand, such rules might prescribe cooperative behavior toward a common goal. On the other hand, the rules might be lenient and allow individual team members to free ride on the cooperative behavior of others, leading to a breakdown of cooperation (Olson, Reference Olson1965). To account for both possibilities, our framework distinguishes between two organizational environments that vary with respect to the extent of mandatory cooperation (i.e. the extent to which an existing rule leaves the decision to act cooperatively at the discretion of the individual).
Second, we examine agency by focusing on interactions at the individual and team levels (the “micro-to-micro link”: Coleman, Reference Coleman1990). We assign an important role to individual social values in exploring these micro-level dynamics. These values have been argued to be important (e.g. Parsons, Reference Parsons1966) and “socially meaningful,” because they connect individuals with social structures (Hitlin and Pinkston, Reference Hitlin, Pinkston, DeLamater and Ward2013: 320).
Finally, we investigate the role of agency on institutional structure (the “micro-to-macro link”: Coleman, Reference Coleman1990) by looking at the effect of micro-level interactions on institutional changes from an existing formal cooperation rule to a new one.
The research centers on a laboratory experiment. The experimental approach enables us to disentangle formal rules, social norms, social values, and the underlying mechanisms of institutional change (Jackson and Cox, Reference Jackson and Cox2013). In particular, the experimental approach provides us with tools to eliminate confounding variables that hinder causal inferences in natural settings.
Our study relates to the existing experimental literature on sanctioning and rewarding systems in social dilemmas. Like ours, this literature addresses both first- and second-order social dilemmas (Ostrom, Reference Ostrom1990). It argues that when a group faces a social dilemma there is a potential role for an institution to promote cooperation. The success of the institution often depends on individual efforts, however, like the willingness to punish free riders. This is what may create a second-order dilemma, where individuals can free ride in the hope that others will make the effort required for the institution to work. The main focus of this literature is on enforcement mechanisms, such as formal and informal systems of sanctioning and rewarding, which can be endogenously chosen or exogenously imposed (e.g. Fehr and Gächter, Reference Fehr and Gächter2000; Reuben and Riedl, Reference Reuben and Riedl2013; Sefton et al., Reference Sefton, Shupp and Walker2007; Yamagishi, Reference Yamagishi1986).
Our contribution to this literature is threefold. First, we address a different second-order collective good dilemma. Instead of studying enforcement mechanisms, we focus on formal institutions themselves and the likelihood of their being changed in order to solve the first-order collective good dilemma.Footnote 5 Doing so recognizes an important aspect, which is that individuals interacting in social dilemmas do not operate in a vacuum; their actions are structured by pre-existing formal rules. Second, we try to unravel the social mechanisms underlying this institutional change by focusing on micro-level interactions, where individual social values take a prominent role. Our final contribution relates to the institutional literature (e.g. Brousseau et al., Reference Brousseau, Garrouste and Raynaud2011; Greif and Kingston, Reference Greif, Kingston, Schofield and Caballero2011; Kingston and Caballero, Reference Kingston and Caballero2009). This literature distinguishes between on the one hand institutions as rules that are kept distinct from their enforcement, and on the other hand institutions as equilibria of the interaction amongst individuals. By investigating why individuals follow or change rules, we bring together these “institutions-as-rules” and the “institutions-as-equilibria” perspectives as advocated by Greif and Kingston (Reference Greif, Kingston, Schofield and Caballero2011: 15).
2. Institutional change
Structural theories of path dependence have typically predicted institutions to be resistant to change, in particular due to the “entrapment of actors” (Pierson, Reference Pierson2000: 253). Agency theories, on the other hand, have argued that actors can endogenously change institutions. This is the starting point of our study. Note, however, that pursuing institutional change may be costly (Yamagishi, Reference Yamagishi1986). Think for instance of organizational sanctions such as a reprimand or a fine (Morrison, Reference Morrison2006), or informal sanctions such as ostracism or derogative reactions by other members of one's team (Nee and Ingram, Reference Nee, Ingram, Brinton and Nee1998). It is then not a priori obvious that change will occur. This is the situation we are interested in. For this reason, we make the initiation of institutional change costly. A first important question, is who, if anyone, is willing to bear these costs? To study this, we distinguish between two scenarios. In the first, all individuals agree that the rule change is desirable. In the second scenario, only some individuals feel this way.
First, assume that desirability of the rule change is shared by all. That means, for any individual, that if she were the only one who could initiate change she would do so. When more individuals can initiate change, however, the costs involved imply that each individual prefers that someone else takes the initiative. This then constitutes a volunteer's dilemma (Diekmann, Reference Diekmann1985). All agree that an act is desirable, but everyone would like somebody else to bear the costs (i.e. initiating institutional change). The question, then, is who will volunteer?Footnote 6 Previous studies have considered, inter alia, how volunteering depends on the (a)symmetry of costs and benefits of the act or on how cost sharing affects the number of volunteers (Chen et al., Reference Chen, Gross and Dieckmann2013; Weesie and Franzen, Reference Weesie and Franzen1998). We look beyond material motives and argue that an individual's inclination to volunteer for a rule change depends on her social value orientation and its relationship with the pre-existing rule she wants to change.
This takes us to the second scenario, where there is no agreement about the desirability of a rule change. Some prefer to keep everything as is, while others want to change the rule. In this case, the decision to attempt the rule change no longer constitutes a volunteer's dilemma. We will argue that here, too, social values play an important role in explaining who will initiate change.
In what follows, we therefore first introduce social values and then discuss for both scenarios how they predict who will initiate a change.
Social values
Individual social values (a.k.a. internalized moral norms: Greif and Kingston, Reference Greif, Kingston, Schofield and Caballero2011) are typically defined as people's generalized beliefs regarding the desirability of conducts or end-states (e.g. Hitlin and Pinkston, Reference Hitlin, Pinkston, DeLamater and Ward2013). They are formed through socialization within families, friendships, social class, or working environments (Parsons, Reference Parsons1966). It is important to note that values and preferences are distinct concepts. While individuals’ value orientations capture their general principles, which can be applied to a range of situations, their preferences may differ depending on particular outcomes, specific actions, or context (e.g. Tao and Au, Reference Tao and Au2014). Values may affect one's preferences or attitudes but are more general than that. Relevant examples of values are self-transcendence (‘prosocial’) values, which reflect concern for others’ welfare and motivate cooperative behavior (Van Lange et al., Reference Van Lange, De Bruin, Otten and Joireman1997). When applied to specific environments, these values may lead to social or other-regarding preferences, as these are typically understood in the experimental economics literature (see Brousseau et al., Reference Brousseau, Garrouste and Raynaud2011 for an overview). Self-enhancement (“proself”) values, on the other hand, relate to pursuing self-interest, and motivate the accumulation of personal wealth and power. Values have been shown to be vital in guiding evaluation of alternatives and shaping behavioral choices (see for an overview: Tao and Au, Reference Tao and Au2014).
Studies in social psychology argue that when an individual faces social values and formal rules that are incongruent, she may experience a conflict between personal and social identity, making the shared social identity less salient (Turner et al., Reference Turner, Oakes, Haslam and McGarty1994). This, in turn, may diminish support for the formal rules that embody the social identity. We conjecture that this situation increases the likelihood that such individuals will attempt to change the formal rules they are subjected to.
First, we consider the volunteer's dilemma that occurs when everyone agrees that an existing rule should change (but, rationally speaking, no one wants to volunteer to initiate this change). Applying the general reasoning from social psychology to this scenario implies that individuals experiencing a conflict between their values and an existing rule are more likely to volunteer to bear the costs of initiating a change in this rule. If, for example, an existing rule leaves the decision to act cooperatively at the discretion of the individual, then those with a proself value have no conflict and might refrain from such actions as long as the rule is in place. Prosocials, however, will experience a conflict between this rule and their intrinsic value to act cooperatively. We can then expect that it will be the prosocials and not the proselfs who will attempt a rule change. More generally, we propose that those with a higher conflict between the existing rule and their value are more likely to volunteer in a volunteer's dilemma. Though the prediction that prosocials are more likely to initiate change in an uncooperative rule seems intuitive, to the best of our knowledge no previous study has addressed the question of how volunteering in a volunteer's dilemma correlates with one's value orientations.Footnote 7
For the second scenario, where only some individuals would like to see the rule changed, we again consider the case where an existing rule leaves the decision to act cooperatively at the discretion of the individual. Proselfs will typically have no desire to change the rule, because it enables them to pursue their self-interest with no restrictions.Footnote 8 The prosocials, on the other hand, are more concerned about the enhancement of joint outcomes, which are less likely to be realized under a less mandatory cooperation rule. Such a rule is therefore in conflict with their values and this conflict will make the prosocials more likely than the proselfs to initiate a change. Hence, also for this scenario, it is those with a conflict between the existing rule and their value orientation that are most likely to initiate a rule change.
This reasoning leads to our first theoretical prediction (TP1):
Theoretical Prediction 1. Individuals whose social values are in conflict with the formal rules are the ones most likely to attempt to change these rules.
TP1 predicts who will initiate institutional change. At first sight, one might think that those whose values are in harmony with the existing rules have no reason to change the latter. Recall from our discussion above, however, that values can be different from preferences. Those who experience value–rule harmony may still have a preference for a different environment because this might make them better off in a specific context. In our experiment, values indeed capture generalized principles. For example, a proself value describes a general tendency to choose the best for oneself, irrespective of what others do. This is in harmony with a rule that allows for low contributions. Still, there are cases where a proself might initiate a change to a more mandatory cooperation rule. This could occur, for example, if she expects that very few people will contribute under an uncooperative rule, hence she anticipates to earn more under a more cooperative rule, even if this requires her to give up the option of contributing little. TP1 implies, however, that such rule-change attempts by those with value–rule harmony are less likely than attempts by individuals with a conflict.
We now proceed to discuss how they do so. For this, the social norms governing behavior in an organization play a central role.
Social norms
An organization's social norms and formal rules form the structure that specifies not only how individuals in the organization ought to interact to receive rewards like status or salary, but also how they “compete” for the reproduction of those institutions (Rand and Nowak, Reference Rand and Nowak2013: 413). Thus, to understand how individuals with conflicting values attempt to change existing formal (cooperation) rules, we also need to take into account how the organization's social norms interact with its formal rules.
We thus focus specifically on the individuals experiencing a conflict between their individual values and the organization's formal rules. These are the individuals that we have predicted to attempt to change the formal rule (TP1). Following Nee and Ingram (Reference Nee, Ingram, Brinton and Nee1998: 35), we consider two possibilities with respect to the prevailing social norm. Either this norm aligns with the individual values or it is in agreement with the formal rule. In the former case, Nee and Ingram argue that individuals will typically find a way to circumvent the formal rules via “a decoupling of the informal norms and the formal rules of the organization.” Thus, both are maintained, but formal rules exist only de jure and social norms de facto. In practice, no formal institutional change is needed because individuals informally agree that one ought to behave differently than formally prescribed.
The second possibility is that existing social norms are in agreement with formal rules but in conflict with individual values. In this case, individual interactions within organizations may give rise to “opposition norms,” which are resistant to the formal rules (Nee and Ingram, Reference Nee, Ingram, Brinton and Nee1998). Opposition norms might undermine formal rules, and this might lead to institutional change. Thus, an individual facing a value–rule conflict, where the prevailing norm sides with the rule, has two options. She can initiate a change in the rule itself or she can try to make the rule irrelevant by changing the social norm into an opposition norm. The former will face more resistance than the latter, precisely because one is not only addressing an adverse rule, but also an adverse existing norm. We therefore predict that individuals in this situation will first try to change the norm into an opposition norm.
The attempted change to an opposition norm might fail, however. In this case, individuals with a conflict may attempt to change the formal rule directly if they “have substantial autonomy” (Ostrom, Reference Ostrom1990: 21). This is most likely in organizations characterized by a democratic structure where all individuals have the ability and the right to participate in rule making (Jacob Reference Jacob2015). In such a democratic environment, an individual can directly challenge an existing rule via voting (Walker et al., Reference Walker, Gardner, Herr and Ostrom2000). When binding for all involved, voting on rule change may provide a successful way to establish socially optimal outcomes (Hauser et al., Reference Hauser, Rand, Peysakhovich and Nowak2014: 220). TP2 summarizes this discussion:
Theoretical Prediction 2
a) Individuals who have a value–rule conflict, in an environment where the social norm aligns with the formal rule, first address this conflict by attempting to change the corresponding social norm.
b) If the social norm does not change, they will attempt to change the rules directly by voting.
Communication
Finally, we consider communication as a characteristic of the environment that may affect the chances of success after individuals initiate a rule change. Communication plays an important role in achieving common goals. Experimental research shows that it improves the ability of groups to overcome issues of cooperation (for an overview, see: Ostrom, Reference Ostrom2005). Moreover, one of the prerequisites for institutional change is the presence of a minimum coalition in favor of change (DiMaggio, Reference DiMaggio and Zucker1988; Ostrom, Reference Ostrom1990). The formation of a coalition is facilitated by the ability to communicate not only with the like-minded but also with opponents (to persuade them). We therefore propose that any attempt to change a rule is more likely to succeed if individuals can contest it through debate, which can be facilitated by communication. This leads to the following prediction:
Theoretical Prediction 3. A rule change is more likely when individuals are able to communicate than without communication.
3. Experimental procedures and design
The experiment was conducted in June and October 2014 at the CREED Laboratory of the University of Amsterdam and consisted of a pre-study (explained below) and a “main experiment.” Two hundred and twenty individuals participated in the main experiment.Footnote 9 We have data from a total of 215 participants (124 men and 91 women, M age = 22.62, SD = 3.16). They participated in one of 17 sessions of the main experiment, in which they were randomly assigned to one of our four treatment cells. Sessions lasted approximately 90 minutes and participants earned on average €29.30 including a €7 show-up fee.
Procedures and tasks
Upon arrival, participants were randomly seated at separated computer cubicles, and were asked to complete multiple tasks, before participating in two public goods games (PGG1 and PGG2). They were informed that the experimental session consisted of multiple tasks and rounds, but they did not know how many rounds or which tasks would follow. Instructions were given at the beginning of each task.Footnote 10 All participants received the same general instructions, aside from treatment-specific information.
The first task measures participants’ risk attitudes by asking “How willing are you to take risks, in general?” with answers on a 11-point Likert scale ranging from 0 (not at all) to 10 (very much).Footnote 11 We measure risk attitudes because an individual's willingness to take risks may positively influence the likelihood of organizational rule breaking (Morrison, Reference Morrison2006). The second task assesses participants’ social value orientations using the triple-dominance measure (TDM; Van Lange et al., Reference Van Lange, De Bruin, Otten and Joireman1997). The TDM consists of nine items, each containing three distinct outcome distributions assigning points to oneself and to an (anonymous) paired other.Footnote 12 For example, a participant is asked to make a choice between (own earnings; other's earnings) options A = (480;80), B = (540;280), and C = (480;480). Depending on their choices, the TDM classifies participants into one of three categories – competitive (in this case, option A), self-interested (B), or prosocial (C). To isolate the value–rule conflict, on the one hand, from “internal” conflicts between an individual's values, on the other, we focus on participants that show “consistent” value orientations, i.e. those who make the same kind of choice nine out of nine times. We therefore categorize a participant as consistently competitive, proself or prosocial if all of the choices can be attributed to the motive concerned. We measure the participants’ social value orientations twice, before and immediately after PGG1, which we refer to as SVO1 and SVO2.
The third task measures the social norm, that is a general convention (Miller, Reference Miller2008), of what one “ought to do” in a PGG. We do so by employing an adapted version of the method introduced by Krupka and Weber (Reference Krupka and Weber2013).Footnote 13 To avoid the measurement of a social norm interacting with participants’ decisions in the main experiment, we conducted a pre-study in which participants were presented with the instructions of the PGG to be used in the main experiment.Footnote 14 We asked the participants in the pre-study to predict which contribution to the team project would be considered most “socially appropriate” by indicating a number between 0 and 10 points. Any participant who chose the modal response earned 500 points. By rewarding an estimation of the modal response, we did not elicit the participants’ own preferences but rather asked them to match the responses of others (i.e. we asked them to indicate what they thought was a “shared understanding” of what one ought to contribute to the PGG). In this way, the participants play a coordination game in which they have to anticipate the extent to which others rate an action as most socially appropriate (Krupka and Weber, Reference Krupka and Weber2013). Our norm measure captures two important features of a social norm. The first concerns what is considered socially appropriate and the second addresses the social consensus (Bicchieri, Reference Bicchieri2006; Krupka and Weber, Reference Krupka and Weber2013).
The pre-study social norm was that one ought to contribute five points (out of 10). For the norm measurement in the main experiment, we present participants with the same pre-study instructions and ask them to indicate their own beliefs about which contribution to the team project the pre-study participants had regarded as most socially appropriate. Again, we provide a 500-point bonus for correctly estimating the pre-study's modal response. This task serves to measure the beliefs held by participants in the main experiment about the prevalent norm. Here again, the modal response was five points. Subsequently, we disclose to all participants in the main experiment that a social norm of contributing five points to the team project had been considered most socially appropriate by those who took part in the pre-study. This revelation of the modal response serves to create a shared understanding of the prevalent norm for the environment of the main experiment, which captures the third important feature of a social norm (Fehr and Gächter, Reference Fehr and Gächter2000).
Public goods game
The core of our experiment consists of two public goods games –PGG1 and PGG2– of 10 rounds each. After finishing the tasks described above, the participants are randomly assigned to a team of five for PGG1 (anonymous matching), and remain in this team throughout all rounds as well as in PGG2. Participants are unaware during PGG1 that a second public good game will follow. In each round of every game, each individual is endowed with a fixed 10-point resource, which they can (partially or completely) contribute to a “team project.” The term “team project” was purposefully chosen to mimic a simple organizational environment. In each round the total team contribution is announced to the team members, but individual contributions remain unknown to others. Participants are informed that each point contributed to the team project yields a marginal per capital return (MPCR) of 0.4 points to every team member, irrespective of their own contributions. This MPCR of 0.4 yields individual earnings per round of:
10 points – # points contributed to the team project + 0.4*total team contribution.
Treatments
We use a full-factorial 2 x 2 between-subject design. In the first treatment – Contribution rule – we vary the formal rule prescribing a minimal mandatory contribution level in PGG1. The second treatment – Communication – varies in whether or not communication with other group members is allowed.
In PGG1 the Contribution rule is either Rule 2, where a minimum contribution of 2 points is imposed in each round; or Rule 8, where the minimum contribution is 8 points. These formal rules are strictly enforced: participants are unable to contribute fewer points than the treatment calls for.Footnote 15 Note that the Rule 8 environment leaves the participants with very little freedom to choose (they can contribute eight, nine, or ten points per round). As clarified below, this provides us with a benchmark to which we can compare the Rule 2 environment.
After participating in PGG1, we first measure SVO2. Subsequently, all participants are given the opportunity to attempt a change of the existing contribution rule (Rule 2 to Rule 8, or vice versa, depending on the treatment) by initiating a call to vote. This is implemented by asking the participants to indicate “yes” or “no” to the question of whether they would like to call for a vote to change the existing contribution rule. In line with our theoretical discussion, a single individual “volunteering” to initiate the vote suffices. The initiation of a vote costs 10 points, irrespective of others’ decisions. This amount reflects the observation that challenging an existing rule in the world outside of the laboratory is costly. If no team member calls for a vote, no voting procedure takes place and the minimum contribution level of PGG1 stays in place. If at least one team member initiates a call to vote –which is our measure of rule-change attempt – then a voting procedure starts, where the minimum contribution level of PGG2 is decided by majority rule (three out of five votes in favor of the change are required). Our measure of an actual rule change is thus defined as the switch from the rule implemented in the PGG1 to the opposing contribution rule.
In the treatments with communication, team members are allowed to communicate for 90 seconds through a chat box before the actual vote takes place. Participants are informed about this chat opportunity before they are given the possibility to call for a vote. In the treatments without communication, participants do not receive any additional information and vote without the possibility of communicating beforehand.
After a call to vote and the voting itself, PGG2 starts, with the same contribution rule as before if there was no call to vote or if the change option was rejected by a majority, and with a new contribution rule if there was a call for vote and the majority supported the change.
Each session concludes with a short questionnaire on the participants’ socio-demographic characteristics. The experimenter then calls each participant separately to the back of the room to receive their payment anonymously. Figure 1 summarizes the timeline of the experiment.
Timeline of the experiment
Payoffs
Participants are paid in cash privately at the end of each session. Earnings are in “points” and exchanged for euros at a rate of €1 per 100 points. Earned points consist of: a possible bonus for the norm measurement; one randomly determined choice for each of the SVO measures; points earned in PGG1 and PGG2; and (negative) points if they initiated a call to vote.
Testable hypotheses
The theoretical predictions presented in the previous section straightforwardly yield hypotheses (H) that can be tested with the data from our experiment.Footnote 16 As argued above, we categorize those participants with consistently proself or prosocial choices as experiencing a conflict.
H1 (based on TP1)
In Rule 2, consistent prosocials are more likely to call for a vote than all others. In Rule 8, consistent proselfs are more likely to call for a vote than all others.
Note that we are not arguing that the proselfs (prosocials) never have a conflict in Rule 2 (Rule 8). For example, our discussion following Theoretical Prediction 1 (applied to the experimental environment in fn. 16) makes clear when proselfs might want to change Rule 2. Like TP1, H1 is a comparative static prediction comparing the likelihood of rule change attempt between the types.
H2 (based on TP2)
a) In Rule 2, consistent prosocials will first try to change the existing norm to higher contributions than five.Footnote 17
b) If their norm-change attempt does not succeed, consistent prosocials will call for a vote.
H3 (based on TP3)
The existing Rule is changed more often when communication is allowed.
4. Results
DescriptivesFootnote 18
Social value measure
Our two measures of social value orientations allow us to assess whether social values are affected by interactions and behavioral outcomes in PGG1. Of the 215 participants, 174 (81%) were categorized equally at time 1 (SVO1) and time 2 (SVO2). For this reason, we base our analyses on the first measure, SVO1.
Our categorization reveals that no participant made competitive choices, leaving only prosocials and proselfs. Twenty participants (9.3%) remain uncategorized because they made fewer than six consistent choices. As is customary when applying this SVO measure (Stouten et al., Reference Stouten, de Cremer and Van Dijk2005), these participants are not considered in the analyses. This leaves us with a total of 195 participants (112 men and 83 women, M age = 22.65, SD = 3.24). Of these, 94 (48.2%) are consistently proself and 52 (26.7%) are consistently prosocial.
We created a variable conflict between social values and the existing contribution rule that indicates when a participant's value orientation is opposite to what the rule requires. Table 1 summarizes how this variable is created, and shows that 28 (23.5%) out of 119 participants in the Rule 2 treatment have a value–rule conflict. These participants are consistently prosocial while the formal rule is “proself.’ In comparison, 34 (44.7%) out of 76 participants in the Rule 8 treatment have a consistent proself value orientation while facing a “prosocial” formal rule. We find no significant differences between the fractions of consistent proselfs (and prosocials) in Rule 2 versus Rule 8 (Fisher exact test, p = 0.177).
Value–rule conflict
Note: The final column assigns individuals into one of three categories: (1) prosocial individuals with a value–rule conflict because they have a consistent value (column 3) that conflicts with the rule in column 1 (Rule 2); (2) proself individuals with a value–rule conflict because they have a consistent value (column 3) that conflicts with the rule in column 1 (Rule 8); and (3) participants without a value–rule conflict. Participants who cannot be classified as either prosocial or proself (N = 20) are not included.
Social norm measure
Because we make the social norm of contributing five points common knowledge, the aggregate group contributions revealed at the end of a round in PGG1 provides participants with a first indication of the extent of norm compliance in the team. Those who do not conform to the social norm might be signaling a desire for an alternative norm (Knight and Ensminger, Reference Knight, Ensminger, Brinton and Nee1998). We therefore consider individual contributions in the first round as a signal to the team members of behavior that deviates from the shared norm. We thus measure norm-change attempts as first-round individual contributions in excess of the norm of five points.Footnote 19
The norm of contributing five points was correctly estimated by 30% of the pre-study participants.Footnote 20 An individual's belief about the prevalent norm is not affected by her own values; although prosocials report a slightly higher norm (M = 6.58, SD = 2.01) than proselfs (M = 6.43, SD = 2.31), the difference is far from statistically significant (two-tailed Mann–Whitney, z = –0.912, p = 0.362).
Analysis of institutional change
Value–rule conflict
Our analysis begins with a test of H1, which predicts that consistent prosocials are most likely to attempt to change Rule 2 while consistent proselfs are most likely to try to change Rule 8.
We start with the Rule 2 treatment to investigate whether having a conflict affects the likelihood of calling for a vote to change Rule 2 to Rule 8. Table 2 provides the results of a logistic regression of the call to vote on being consistently prosocial (that is, having a value–rule conflict), and on risk attitude, gender and age. The results reveal that a value–rule conflict is a strong and significant predictor of a rule-change attempt from Rule 2 to Rule 8 (second and third columns). None of the controls has a significant effect on the call to vote.
Rule-change attempts
Notes: In Rule 2 (Rule 8) the consistently prosocial (proself) have a conflict.
*/** p < 0.05/p < 0.01 (two-tailed tests).
A similar analysis for the Rule 8 treatment shows no significant effect of a value–rule conflict on the attempts to change to Rule 2 (forth and fifth columns of Table 2), that is, consistent proselfs are not more likely to attempt change than all others. The regression also shows that attempting this rule change is positively correlated with one's willingness to take risks.
The results in Table 2 show that the role of a value–rule conflict in individual attempts to change formal rules depends on the institutional environment. When the latter is based on a formal rule requiring a low level of contribution, we find that a value–rule conflict is the main driver of formal institutional change. H1 is thus partially supported.
Social norm and rule-change attempt
H2a predicts that consistent prosocials in Rule 2 will first try to change the existing norm. A norm-change attempt refers to first-round contributions above the social norm of five points. We have 28 cases with a value–rule conflict and 91 without (cf. Table 1). Of those with a conflict (the consistent prosocials), 17 (60.7%) attempted to change the norm, while 34 (37.4%) of those without a conflict attempted to do so.Footnote 21 This difference in proportions is significant at the 5% level (Fisher's exact test, p = 0.048). Therefore, we conclude that individuals with a value–rule conflict do indeed attempt to change the social norm first, thus providing support for H2a.
At first sight, one might think that this result simply replicates the known effect that prosocials contribute more to a public good than proselfs (e.g. Offerman et al., Reference Offerman, Sonnemans and Schram1996, Van Dijk et al., Reference Van Dijk, Sonnemans and van Winden2002). This is not obviously the case, however, because there are only two possible motivations for prosocials to do so, neither of which can explain what we observe in our data. On the one hand, prosocials might unconditionally contribute more than proselfs, e.g. for reasons of pure altruism. If this were the case, they would not only contribute in the first round of PGG, but in all rounds. This is not observed in our data. On the other hand, prosocials might start to contribute and continue to do so, conditional on others contributing. In our design, such “conditional cooperators” (Keser and van Winden, Reference Keser and van Winden2000) have no reason to contribute fully in the first round, however, because they have been informed that the social norm is to contribute five points. We conclude that our observation that prosocials contribute more than five points can be interpreted as signaling a desire to change the social norm.
Next, we investigate whether a failed norm-change attempt affects the likelihood of a subsequent attempt to change the existing formal rule (H2b). We consider a norm change to be successful if groups reach a mean contribution of eight points or more in all ten rounds of PGG1.Footnote 22 In Rule 2, attempts to change the norm were observed in 25 of the 26 groups. Only three groups succeeded in this attempt (12%). Of the 14 individuals involved in these three groups, only two called for a vote after PGG1 as opposed to 47 of the 105 participants in groups where the norm did not change.
H2b predicts that individuals with a conflict try to change the contribution rule directly via voting, if they previously tried to change the norm and failed. To test this conjecture, we only consider teams where the norm did not change. This reduces the number of observations to 105. Table 3 provides the results of a logit regression of the call to vote (Rule 2 to Rule 8) on the same set of regressors as in Table 2, and includes two dummy variables and their interaction. The first reflects participants with a conflict (that is, the consistently prosocial) and the second captures individuals who have attempted a (failed) norm change. The interaction term between these two dummies allows us to isolate the group that H2b is concerned with. Including this interaction term means that the first dummy reflects individuals with a conflict who did not attempt a norm change, while the second concerns individuals without a conflict who attempted a (failed) norm change. For these two dummies, we do not expect an effect on inititating rule change. Indeed, the coefficients are not significantly different than zero. The hypothesis is concerned with individuals with a conflict who have attempted a norm change and failed. To test the effect on initiating rule change, we need to test whether the sum of the coefficents for individuals with conflict, for those who have attempted to change the norm and failed, and for whom the interaction term is different from zero. We find strong evidence that this is the case; the null of no effect is rejected (χ2 = 8.56; p = 0.003). This evidence that attempts to change the rule are preceded by attempts to change the norm provides strong support for H2b.
Direct rule change attempts (Rule 2 to Rule 8)
Notes: N = 105; *p < 0.10 (two-tailed test).
The role of communication
Finally, we expected a change in the contribution rule to be facilitated by communication (H3). Whether communication has this effect may depend on the institutional environment. For example, participants in Rule 8 might feel hesitant to communicate because their wish to change it to Rule 2 would signal a proself value (which they may not want to share in a “prosocial” environment). We therefore considered the two rule treatments separately. To fully utilize the individual-level data that we have, we consider individual votes on whether to change the rule (treating the cases where no one called to vote as a vote against the change). This approach is based on the idea that the likelihood of a vote in favor of change and the likelihood of an actual rule change are highly correlated.
In Rule 2, 79.7% of the (59) participants in the treatment without communication are in favor of a rule change, and 86.7% of the (60) participants in the treatment with communication are in favor. The difference is not significant at conventional levels (Fisher's exact test, p = 0.337). The percentages are, respectively, 7.9% and 2.6% in Rule 8, but again not significant (Fisher's exact test, p = 0.615). These results do not support H3.
5. Concluding discussion
We address agency-driven institutional change by experimentally investigating the micro-processes of individuals’ social values, social norms and actions within the constraints and opportunities implied by exiting formal rules. In particular, we consider two formal institutional environments that vary with respect to the extent of mandatory cooperation (i.e. the extent to which an existing rule leaves the decision to act cooperatively at the discretion of the individual). Doing so provides empirical evidence to the macro–micro–macro link in analytical sociology (Coleman, Reference Coleman1990). Our results show that formal institutional change is most likely to be initiated by individuals who experience a conflict between prosocial values and existing rules allowing for low cooperation.
Moreover, if such individuals with a value–rule conflict face a social norm they disapprove of, they start their attempts at institutional change by trying to change this social norm. They do so by signaling a desire to deviate from it. Note that attempting a norm change by deviating from it is not a priori obvious. This is because doing so may be costly due to possible social pressure. In our experiment, deviating from the social norm in a more prosocial direction is costly because of the nature of the public goods game: while the income of other team members increases, the deviator's income decreases.
The results show that although individuals make many attempts to change the social norm, they often fail. When this occurs, individuals try to change the formal rule directly. In our experiment this is undertaken by calling for a vote to decide about a rule change. Note that a rule-change attempt after a failed norm-change attempt comes at an additional loss of income because of the tangible costs associated with it (e.g. in our experiment, these are the costs related to calling for a vote). Nevertheless, many individuals in our study not only deviate from the existing social norm in an attempt to change it, but also follow through with a rule-change attempt after the norm change has proven unsuccessful. The results also show that for the few cases where a norm-change attempt was successful, it was rarely followed by a subsequent rule-change attempt. This confirms theoretical arguments proposed in the literature that new social norms can become self-reinforcing and can render a change in formal rules unnecessary (Nee and Ingram, Reference Nee, Ingram, Brinton and Nee1998; Ostrom, Reference Ostrom1990). Taken together, these findings are in line with Hodgson's (Reference Hodgson2006) conclusion that while some institutions can be self-enforced, others need a more formalized sustainment.
We do not find evidence that communication facilitates an actual rule change, which is rather surprising because communication between team members has been shown to solve cooperation problems (Janssen et al., Reference Janssen, Tyson and Lee2014; Ostrom, Reference Ostrom2005). In fact, we observe that a majority of those voting are in favor of a formal rule change toward more mandatory cooperation, irrespective of whether they were able to communicate beforehand. One possibility is that participants are familiar with the environment characterizing the social dilemma. A call to vote for a change in the less cooperation rule may be informative in itself without the need for more discussion. If this is the case, communication will have no effect. More research is needed to further investigate the role of communication in this setting.
There is an asymmetry in our results across the two formal institutional environments that we examine. Rule-change attempts are both more frequent and more successful when changing a rule from low to high cooperation than vice versa. Very few individuals with a conflict between their proself values and a (prosocial) high cooperation rule attempt to change the latter. This result may not be surprising if we consider the nature of the public good game in our experiment. A proself value implies acting in a way that maximizes its own (economic) well-being. In our study, the formal rule requiring high cooperation ensures not only an equal and prosocial distribution of wealth but also high earnings for all. A change to a less stringent rule might reduce the earnings even of a proself individual who would typically contribute less after the change. This could motivate her not to initiate a change and to maintain the high cooperation rule. On the other hand, proself individuals will want to reduce the minimum contribution level if they expect to benefit from doing so. This will be the case if proselfs expect that sufficient other team members will maintain high contribution levels even after the required minimum has been reduced. In that case, free riding on the prosocials’ contributions would be a profitable alternative to being forced to contribute highly. The high cooperation rule, however, does not allow the proselfs to collect information about others’ willingness to contribute under less restrictive contribution rules. In short, while prosocials in the low contribution requirement can contribute above the minimum to see how others respond, proselfs cannot contribute less when high contribution is mandatory. This asymmetry caused by the rules might explain the asymmetry in our results.
We have focused here on institutional change within teams, the simplest form of an organization. The laboratory environment allows us to draw causal conclusions on underlying mechanisms of institutional change, a task that would be more difficult under more complex organizational structures and virtually impossible with observational field data. This high internal validity leaves open the matter of the generalizability of our results (i.e. the external validity). Many organizations outside the laboratory are more involved than teams and do not allow for detailed analysis of the complex interactions between values, norms, and rules. Our study intends to provide a benchmark for further investigation under more complicated settings.
One line of future research may systematically introduce new features into our laboratory design to directly check the robustness of our results. For instance, while we believe the public goods game to be a tool that is well-suited for studying the agency of institutional change in social dilemma situations, institutional change might have different patterns in other interactive environments. Similarly, our experiment enables individuals to directly vote on formal institutional change. Further research should also aim to investigate institutional change in alternative decision-making environments. Another line of future research could involve moving from the laboratory to the field. This can be done by running field experiments along lines similar to our design, or by collecting observational data in the field (for example via surveys amongst members of organizations). The latter will not allow for conclusions about causality, but their high external validity would provide a powerful addition to the causal inferences we have been able to derive from our experiments.
Finally, by investigating the interplay between agency and institutional structure, our study was able to show empirically that the institutions-as-rules and the institutions-as-equilibria perspectives “become entwined” (Hodgson, Reference Hodgson2006: 21). The main takeaway of our study may well be the importance of individuals, their (non)material motives, and group-level processes for understanding the evolution of rules in organizations.
Acknowledgements
Financial support from the NWO (ASPASIA-VIDI #015.007.048) is gratefully acknowledged. We would like to thank seminar participants at the 2014 ESA International Meetings in Hawaii and the participants at the 2015 IMEBESS conference in Toulouse who contributed to earlier versions of this work with helpful comments.
Open access
