Social value orientation and sanctioning attack decisions in an economic contest game.

2

Abstract

The aim of this study was to investigate how principals sanction their agent’s attempts (attacker) at exploiting another party (defender) through attack, and whether sanctions are

influenced by the outcomes of the agent’s contest decisions (win/loss and victory/no victory) and the principal’s social value orientation. The results showed that principals have higher sanction values (towards reward) when the outcomes of the attacker are win or victory (compared to loss and no victory). In addition, the results showed that principals have higher sanction values towards win outcomes, no matter if these outcomes are victory or not. There were no significant effects found of the social value orientation, which does not have an influence on the sanctioning behaviour. Altogether, this study deepens the understanding of sanctioning behaviour in humans in conflict situations where people perform actions for someone else who will in response receive a reward or punishment.

3

Index

Sanctioning: punishment and reward ... 11

Present study ... 12

Exploratory study: the social value orientation ... 15

Method ... 18

1. Design ... 18

2. Participants ... 19

3. Procedure ... 20

3.1 Contest version of the attacker-defender game ... 22

3.2 Sanctioning task of the principal ... 23

3.3 Social value orientation: Slider Measure ... 24

3.4 Risk preferences: Gamble Task ... 25

4. Statistical analysis ... 25

4.1 Assumptions ... 26

4

Results ... 28

Participant information ... 28

Part 1: Assumptions ... 28

Part 1: Within-subject effects ... 29

Part 2: Assumptions ... 31

Part 2: Two-way repeated measures ANOVA ... 31

Part 3: Assumptions ... 33

Part 3: Mixed ANOVA ... 34

Discussion ... 36

Conclusion ... 36

Limitations & future research ... 37

5

Introduction

From an evolutionary perspective humans are social animals who are usually quite cooperative (Tomasello, 2014). Cooperation in humans is generally based on social norms, legal laws and rules of society (Fehr & Fischbacher, 2004a). There are a lot of studies that focus on the human incentive of acting only in self-interest (e.g. Jensen, 1994; Miller, 2001; Schwartz, 1986), nevertheless, within psychology studies there is a growing attention for more prosocial

motivations that drive individuals (De Cremer & Van Lange, 2001). Prosocials do not only focus on what is best for the self, but also care about the tendency to enlarge equal and good outcomes for others (De Cremer & Van Lange, 2001; Kelley & Thibaut, 1978). The social value orientation looks at this by examining different kinds of orientations of individuals (prosocials & proselfs) in the weights they assign to the outcomes of others and for the self (Van Lange, 1999). Altogether, this could play an important role (for example) in economic decision-making processes, where the outcomes are not only important for the individual making the decision, but are also important for the bigger picture: for the group, employer or company.

However, conflicts are likely to arise if people have different social orientations and different goals to achieve (Oberschall, 1978). In many conflict situations, one party wants to change the status quo (also known as the attacker), whereas the other party wants to defend the status quo (also known as the defender) to protect oneself against loss (De Dreu & Gross, 2019). The status quo can be seen as a reference point from which attackers try to gain something and from which defenders try not to lose (Samuelson & Zeckhauser, 1988). What would happen if a third party is added to the conflict, who can sanction (punish and reward) the attacker based on

6 their outcomes? Does the social value orientation have a role in this sanctioning behaviour of the third party?

A brief overview

The social value orientation is often used in economic decision making games and social dilemma games, like the prisoners’ dilemma, where conflict often occurs (e.g. Balliet, Parks, & Joireman, 2009; Fehr & Fischbacher, 2004a; Kwaadsteniet, Rijkhoff & Dijk, 2013; Murphy, Ackermann, & Handgraaf, 2011). De Dreu, Giacomantonio, Giffin, & Vecchiato (2019) examined the social value orientation in their first experiments using the contest version of the attacker-defender game: a game where two players have to choose how much to invest in attack (as attacker) or defence (as defender). The results showed that individuals with strong prosocial preferences are less willing to attack (not less willing to defend). It also showed that attack and how much to invest in attack was moderated by prosocial preferences. The present study will make use of this contest version of the attacker-defender game (AD-G) (De Dreu et al., 2019), but will instead add a third party to the game (also known as the principal). This third party is able to sanction (reward or punish) the players (attacker or defender) in response to the way they acted in the game (Kopelman, Weber, & Messick, 2002). The attacker (and defender) in the contest version of the AD-G will use resources out of a given endowment, to choose how much to invest in attack. That decision will lead, for example, to victory or no victory, which will only affect the third party (principal) and not the players itself.

Real-life economic decision making processes do not only consist of two individuals using resources to act for themselves, but in most cases consists of making decisions that affect others or the whole company (Allison & Messick, 1990; Salancik & Pfeffer, 1974). By adding

7 the third party (principal), it could be investigated how they would act upon the outcomes of the attacker (agent) in the contest version of the AD-G. This third party creates the setting of a principal-agent model within the contest version of the AD-G (Coleman, 1990; Williamson, 1975). The players (agents) act in the interest of the principal, as the outcomes will only affect the principal, who in response has the ability to sanction.

The main goal of this study is to investigate how principals sanction their agent’s attempts (attacker) at exploiting another party (defender) through attack, and whether sanctions are

influenced by the outcomes of the agent’s decisions (win/loss and victory/no victory) in a contest version of the AD-G. Does the sanctioning behaviour depend on the outcomes (win/loss), and does it matter whether the outcomes are victory or not? The outcomes of the attacker are defined in units gained, ranged from 1-19 units. A “win” outcome is defined as units gained whenever the outcomes of the attacker are > 10 units. A “loss” outcome is defined as units gained whenever the outcomes of the attacker are < 11 units. The starting capital of 10 units is taken as a reference point to define these win and loss outcomes. The attacker is victorious when he or she invested more units than the defender in the contest version of the AD-G. Whenever the attacker invests the same or less units than the defender, this is defined as no victory. To clarify, this means that a gain outcome (>10) is necessarily also a victory, but for some loss outcomes (<11) it can be both a victory or no victory. In the current research the main focus will be on the outcomes made by the attacker and the sanctioning behaviour towards the attacker (not the defender) to not

overcomplicate the research design and analysis.

8 The present study will also explore if the social value orientation affects the principal’s decision-making in sanctioning the attacker. For example, to see if prosocial principals would sanction victory outcomes of attackers differently than proselfs do. This has an exploratory direction, as there is limited research available about the influence of the social value orientation in a similar design setting.

Research investigating the differences in sanctioning behaviour of third parties (principals) towards attackers (agents) in this contest version of the AD-G is missing in the literature. Prosocial and proself orientations of principals could influence sanctioning behaviour towards players and will be investigated exploratory. The results of this study could be of value for all different kinds of economic settings (e.g. employers, lawyers, stakeholders) to know how third parties respond to outcomes made by others, to see if they act differently in regard to win/loss outcomes, to victory or no victory and to see if different social value orientations enlist different punishment and reward behaviour in principals. Altogether, this study deepens the understanding of human sanctioning behaviour in conflict situations.

Attacker-defender game

When two parties are negotiating and acting in their own interest, conflict is almost inevitable (Benhabib & Rustichini, 1996; Pruitt & Carnevale, 1993; Oberschall, 1978). A few examples of negotiating parties could be: two companies trying to merge, two opposing attorneys acting in their own clients interest or two countries being at war with each other. In most cases, one party is looking to maximize their reward, while the other party tries to protect themselves against exploitation (De Dreu & Gross, 2019). This divides the two parties in an attacker and a defender, who both have different interests in the attacker-defender game (AD-G).

9 The AD-G is different from other social decision making games (e.g. the prisoners’

dilemma) as it is an asymmetric game (De Dreu & Gross, 2019). The outcomes are not symmetrical opposites for each player involved (see Figure 1.0). The position of attacker or defender is of importance as it determines the reason for choosing a certain action. The attackers can choose to invest in attack (defect: D) or not (cooperate: C) and the defenders can choose to invest in defence (D) or not (C). If both players want the best outcome for themselves, the attacker will choose to defect, whereas cooperating is the most attractive for the defender (see Figure 1.0). It is most beneficial for the attacker to mismatch its defender’s action (outcome DC or CD), whereas it is most beneficial for the defender to match its attacker’s action (outcome DD or CC) (De Dreu & Gross, 2019; Goeree, Holt, & Palfrey, 2003).

Figure 1.0. The attacker-defender game. Two players, an attacker and a defender, can choose to

cooperate (C) or to defect (D) with both its own (assymetrical) beneficial outcomes.

The AD-G can be transformed into a contest game, also known as the predator-prey contest (De Dreu et al., 2019). The attacker and defender have to choose how much resources to invest out of a given endowment. The attacker chooses how much to invest in attack (x), while

10 the defender chooses how much to invest in defence (y). They both use resources out of an equal endowment (e) (e.g. money). The attacker wins when the investment in attack is bigger than the investment of the defender (x > y). The attacker gains what is left of the endowment minus the investment of the defender (e – y). Added to the remaining endowment (e – x), the attacker has a total gain of 2e – x – y, while the defender is left with 0. If the investment of the defender is equal or higher than the investment of the attacker (y ≥ x), the defender wins, which leads to a gain of

e – y for the defender and a total gain of e – x for the attacker.

A third party (also known as a principal) could be added to this contest version of the attacker-defender game. The players (attacker and defender) will use resources (out of given endowment) in the contest AD-G to get outcomes that will only affect the principal, and not the players itself. So if the attacker wins, he or she wins the outcomes for the principal. The design of the contest version of the attacker-defender game with the given endowment is further explained in the method section. This implementation of a third party in the contest AD-G can be explained by the principal-agent theory.

Principal-agent theory

The principal-agent theory consists of the principals’ agent (e.g. doctor, lawyer,

employee) who is trying to achieve the objectives of the principal (e.g. patient, client, employer) (Coleman, 1990; Williamson, 1975). The principal disposes a few resources (e.g. money) to the agent, in hope that the agent is willing to further the interests of the principal (Coleman, 1990). The agent is often chosen because the principal lacks the skills, abilities or time to effectively perform the given tasks (Petersen, 1993). The agent carries out the corresponding actions to fulfil the objective, which leads to certain outcomes, which will affect the principal. A principal can,

11 based on the outcomes, use sanctions to reward or punish the agent’s behaviour (McGillivray & Smith, 2000). It occurs in most cases that the principal would punish exploitative behaviour and reward behaviour that is positive for the principal (Fehr & Fischbacher, 2004b; Fehr, Gächter, & Kirchsteiger, 1997). Translated to the present research with the contest AD-G, this means that the players are the agents (attacker or defender) who are choosing to invest out of the given

endowment, which outcomes affects the principal. The principal can sanction (punish or reward) the agent’s outcomes.

Sanctioning: punishment and reward

Sanctioning systems can increase cooperation in social dilemma games (Fehr & Gächter, 2000; Fehr, Fischbacher, & Gächter, 2002; Mulder, Van Dijk, De Cremer, & Wilke, 2006; Wit & Wilke, 1990). The choice of defection is less attractive for participants through the use of

punishment, causing the cooperation choice to be more attractive. Research showed (Fehr & Gächter, 2000; Fehr et al., 2002) that the action of cooperation is more attractive in a contest game that implements punishments for non-cooperative behaviour and rewards for cooperative behaviour, compared to a contest game without sanctions. This result was also shown in research of McGillivray & Smith (2000) who used punishments in the agent-principal game. It showed that individuals also keep thriving for the best outcome in the future, knowing that punishments may return. In addition, research of Mulder et al. (2006) showed that if there are no punishments present in a social decision game, where individuals have to make decisions for third parties, that those players are less willing to cooperate. There is no concrete evidence that punishments are more effectively than rewards, as studies found results indicating both punishments as well as rewards to be more effective (Komorita & Barth, 1985; Rapoport & Au, 2001).

12 Research of de Kwaadsteniet, Rijkhoff and van Dijk (2013) showed that third parties use the equality rule as a measure to reward or punish behaviour of others acting in the third party’s interest. This means that a third party will sanction based on what he or she sees as an equal, fair response to the invested resources of players leading to the (preferable) outcome (Fehr &

Schmidt, 1999). Sanctions do encourage individuals to not only focus on the outcome in their self-interest, but also to look at the bigger picture of a collective benefit (Balliet, Mulder, & Van Lange, 2011; Gächter, Renner, & Sefton, 2008). That is why sanctions are implemented by the principal in the contest version of the AD-G and will be investigated in the present study, to examine how principals would respond (sanction) towards different outcomes (win/loss and victory/no victory) of attackers.

Present study

Based on the previous topics discussed, it would be an interesting direction to look at the sanctioning behaviour of the principal (third party) towards the attacker (player/agent) in the contest version of the AD-G. Does the sanctioning behaviour (punishment and reward) depend on the outcomes (win/loss) of the attacker, and are there differences whether the outcomes of the attacker are victory or not?

The following hypotheses will be investigated:

Hypothesis 1.0: The principal would (on average) A) punish losses and reward win outcomes of the attacker and B) would have higher sanction values towards the attacker in win outcomes compared to loss outcomes.

13 Hypothesis 2.0: The principal would (on average) A) punish no victory outcomes and reward victory outcomes of the attacker and B) would have higher sanction values towards the attacker in victory outcomes compared to no victory outcomes.

Hypothesis 3.0: The principal would have higher sanction values towards win outcomes of the attacker than loss outcomes, no matter if the outcomes are victory or not.

The sanctioning behaviour of the principal is defined in points given to the player (attacker) after the investments and outcomes of each trial. These points can be positive to function as a reward (+1 to +3), be negative to function as punishment (-1 to -3) or be neutral (0) where no reward or punishment is given. The outcomes of the attacker are expressed in units gained, ranged from 1-19 units. A “win” outcome is defined as units gained whenever the

outcomes of the attacker are > 10 units. A “loss” outcome is defined as units gained whenever the outcomes of the attacker are < 11 units. The starting capital of 10 units is taken as a reference point to define these win and loss outcomes. The attacker is victorious when he or she invested more units than the defender in the contest version of the AD-G. Whenever the attacker invests the same or less units than the defender, this is defined as no victory. This means that a gain outcome (>10) is necessarily a victory, but for some loss outcomes (<11) it can be both a victory or no victory. For example, if the attacker invests 9 units and the defender 8 units, the attacker is victorious, but will only gain 3 units (2e – x – y), which is defined as a loss.

Research of different economic games (e.g. public goods game & third-party prisoners’ dilemma) with a third party involved showed that this party will reward successful outcomes and punish failure (Balliet et al., 2011; Fehr, & Fischbacher, 2004b; Kwaadsteniet, Rijkhoff, & Dijk,

14 2013; Nikiforakis, & Mitchell, 2014). In regard to the contest version of the AD-G and the

principal, this could mean two things: i) if the attacker invests more than the defender, he or she succeeds (victory) and will be rewarded for this behaviour. If the attacker invests less than the defender, he loses and will get punished; ii) if the attacker gains more units than the reference point of 10 units, he or she will be rewarded for this. If the gained outcomes are less than 11 units, he or she will be punished for this. This sanctioning behaviour would most likely occur when the agent is acting in the interest of the principal (Koford & Penno, 1992).

In most cases it would be that the principal rewards behaviour that is positive for the principal and punishes behaviour that is not favourable for the principal (Fehr & Fischbacher, 2004b; Fehr, Gächter, & Kirchsteiger, 1997). However, the tendency to punish can be inhibited by the do-no-harm principle, other-concern or feelings of empathy (e.g. Baron, 1995; De Dreu et al., 2019). Fehr et al. (2002) showed that reciprocity is a strong perception that guides people in their social behaviour. So when a person is cooperating, the reply to this behaviour is (most of the time) also cooperative behaviour. In the present study this could be translated to that if the

attacker wins, showing positive behaviour and outcomes towards the principal, the player will receive something positive in return: rewarding behaviour.

Research of Fehr & Fischbacher (2004a) looked at the strength of third party sanctions to players in a prisoners’ dilemma game. They looked at the preferences of the third party, if they preferred cooperation or defect more of the players. Results showed that if the third party

preferred cooperation more, the players that chose the defect-option got stronger punishments. If the third-party preferred the defect-option more, players that chose to defect were less punished. Other research of Fehr & Fischbacher (2004b) showed in the dictator game that third parties

15 punished individuals more when they invested less than half of their given endowment. Higher, positive outcomes let to less punishment behaviour of the third party.

These hypotheses imply that principals will sanction (use punishment or reward), to see if there is a difference between win or loss outcomes and if it differs when there are victory (or no victory) outcomes. However, the question remains what could influence this sanctioning

behaviour of principals? Are there differences between a principal who has a proself or a

prosocial orientation? That is why the present study will also look exploratory if the social value orientation affects the principal’s decision-making in sanctioning the attacker and if differences between prosocial and prosocial principals can be found.

Exploratory study: the social value orientation

The social value orientation looks at how much an individual cares about the outcomes for themselves and others (Van Lange, 1999). Most individuals care about positive outcomes for the self and try to only maximize their own gains (proself). They differ from prosocials, who care about the losses and gains of others. These two orientations influence people’s opinions and actions (Miller, 2001). In this exploratory study it goes further into the hypotheses and research question mentioned before, where the main focus will be on the outcomes made by the attacker and the sanctioning behaviour of the principal towards the attacker. The question asked is if there are differences in sanctioning behaviour between prosocial and proself principals towards the win/loss outcomes and victory (and no victory) outcomes of the attacker?

16 To elaborate on the previous hypotheses, the following hypotheses will also be

investigated:

Hypothesis 4.0: The principal would (on average) A) punish losses and reward win outcomes of the attacker and B) would have higher sanction values towards the attacker in win outcomes compared to loss outcomes. This is especially the case for prosocial principals compared to proself principals.

Hypothesis 5.0: The principal would (on average) A) punish no victory outcomes and reward victory outcomes of the attacker and B) would have higher sanction values towards the attacker in victory outcomes compared to no victory outcomes. This is especially the case for prosocial principals compared to proself principals.

Hypothesis 6.0: The principal would have higher sanction values towards win outcomes of the attacker than loss outcomes, no matter if the outcomes are victory or not. This is especially the case for prosocial principals compared to proself principals.

Prosocials might punish attackers when they are aggressive towards the defender, which means attackers exploit (invested in maximum attack) the defender (Fehr & Fischbacher, 2004b; Fehr, Gächter, & Kirchsteiger, 1997). In addition, prosocials will reward success behaviour that does not exploit the defender and does not fully “harm” its endowment. Prosocials do not only care what is best for the self, but also care about equal and good outcomes for others (De Cremer & Van Lange, 2001; Kelley & Thibaut, 1978). Proselfs on the other hand might punish weak, failure behaviour of the attacker more (compared to prosocials), as they want to maximize their own gain. They will only reward aggressive, exploiting behaviour of the attacker, as this gives

17 the best outcome for themselves. It is about striving for the best outcome for the self, by giving rewards or punishments to the other party involved, that will benefit the self (Miller, 2001; Schwartz, 1986).

These hypotheses are in line with different studies about prosocials and proselfs (e.g. De Cremer & Van Lange, 2001; Kelley & Thibaut, 1978; Miller, 2001). The hypotheses still have an exploratory direction, as there is limited research available about the influence of the social value orientation in a similar design setting like the current study. Research of De Dreu et al. (2019) looked at the role of prosocial preferences towards players in the attacker-defender game. Prosocial preferences were measured by the amount of empathy and other-concern. The results showed that attackers with a prosocial orientation were less willing to attack and also invested less in aggressive attacks. It also showed that the quantity of attack was moderated by prosocial preferences. The prosocial orientation showed no direct effects on defending of defenders. The current research will explore whether similar findings can be found in principals in regard to their sanctioning behaviour.

18

Method

1. Design

The present study is based on a social decision-making lab experiment among participants playing as attacker or defender (agents) and as a principal (third party) in the contest version of the AD-G. Each participant is either an attacker or a defender throughout all the tasks. As participant you make investments for yourself as an attacker (or defender), then you make investments on behalf of someone else as an attacker (or defender) and then you evaluate an attacker who made investments with your resources as a principal. This social decision-making lab experiment with the corresponding data collection was executed by another MSc thesis project group of De Dreu in 2019. The data of this previous lab experiment are used, as a new lab experiment cannot be conducted due to circumstances of the Corona virus in the Netherlands.

This study makes use of a within-subject design, where the sanctioning behaviour of principals in two conditions (win/loss outcomes and victory/no victory) will be analysed. The main goal of this research is to investigate the sanctioning behaviour (punishment and reward) of principals towards the outcomes (win/loss and victory/defeat) of the attacker in the contest version of the AD-G. In addition, the aim is also to look exploratory at the social value

orientation, if this influences the principal’s sanctioning behaviour, which will be analysed with a between-subject design. The lab experiment did not involve any deception, as participants really played the contest version of the AD-G with each other. The procedure of how the lab experiment was conducted and which questionnaires were used will be explained in more detail.

19

2. Participants

Participants took part in the social decision-making lab experiment of 2019. This experiment consisted of three phases: i) the first phase consisted of participants playing the contest version of the attacker-defender game, playing as the attacker or defender, making investments that would affect their own outcomes; ii) the second phase was the same as the first phase, the only difference there was that the players were investing resources of the principal and the outcomes were affecting the principal; iii) the third phase consisted of the participants playing as the principal (third party) who could reward or punish the given outcomes made by a player acting for the principal in the contest version of the AD-G. Half of the participants (N = 114) performed all three phases as an attacker, as the other half performed them as a defender. The participants participated in dyads, where they were randomly assigned to i) other participants in the opposite role (attacker with a defender); ii) to a principal; iii) an agent who made decisions that would affect the principal.

At the end of all phases the Slider Measure (Murphy et al., 2011) for the social value orientation was given as well as the Gamble Task to measure risk preferences (Holt & Laury, 2002). The experiment ended with a short questionnaire about the participant’s demographic information. Participants took part in all three phases. The present study will only look at the third phase of the experiment, together with the data of the Slider Measure for the social value orientation.

The only inclusion criteria for the experiment were proficient in English and people between the age of 16 and 35 years old. All participants were informed with the informed consent beforehand and were voluntarily participating in the lab experiment. At the end of the experiment

20 participants were able to write down their bank account administration so they could be

compensated a few days later for their participation. Participants were recruited via Sona and the personal social network of the lab conductors.

3. Procedure

First of all, professor C.K.W. De Dreu submitted the overall research proposal to the

Leiden University Ethical Committee. After this approval the lab experiment in 2019 was

executed.

The experiment took place at the Level Leiden University Lab. The data of the lab experiment was collected by computers in separate cubicles where participants entered their replies. The participants were not able to communicate with each other during the experiment. About 5 dyads of participants were able to participate together at the same time in the lab. After informing participants about the setup of the experiment and signing the informed consent, they began reading the instructions for the first phase of the experiment: the contest version of the attacker-defender game, playing as attacker or defender. Beforehand, it was stressed to the participants that their decisions were made anonymous and were kept in confidence. A short summary task followed to see if they understood the instructions. The participants participated in dyads, where one was assigned as the role of defender (role A) and the other as the role of attacker (role B). Participants were matched with the same opponent in each trial. The first phase of the experiment consisted of 30 trials where participants had to choose how much to invest in attack (or defend) from the given endowment of capital. The given capital consisted of 10 units. The given endowment was of the player itself, as well as the outcomes that followed, for example a gain in 3 units are for the attacker (or defender) itself. A detailed description of how this

21 investment process in the contest version of the AD-G works is given in section 3.1. After each decision the results of a participant’s own investment is shown, their counterpart’s investment, and the total outcome (in units) of the contest game for both parties.

After that, the second phase of the experiment was started. The same dyads as before were playing as attacker or defender again in the contest version of the AD-G, but now with a given endowment of 10 units that were from a third party, the principal. Again 30 trials were given to both players who had to decide how much to invest in attack (or defence). The outcomes did not affect the players, but only the principal. Each player (both attacker and defender) had to choose again how much to invest in attack or in defence. Again, after each decision the results of the participant’s given investment were shown, their counterpart’s investment, and the total outcome (in units) of the contest game for both parties that would affect the principal.

Subsequently, the third phase started. In this phase, the same participants as before played now as the role of the principal. If a participant was assigned as the role of an attacker in the previous phases of the experiment, that participant would get a list of 66 possible investments and outcomes (0, 1, 2, 3 … 19) an attacker could make in the contest version of the AD-G. If a

participant was assigned as the role of the defender, that participant would get the 21 possible investments and outcomes (0, 1, 2, 3 … 10) a defender could make in the contest version of the AD-G. The participant (as principal) had to make a sanction decision based on these investments and outcomes. A more in depth description of how these sanction decisions were made is given in section 3.2.

22 At the end of the experiment, the Slider Measure (Murphy et al., 2011) and the Gamble Task (Holt & Laury, 2002) were given as well as a short questionnaire about the participant’s demographic information. The Slider Measure measured the participant’s social preference and the Gamble Task measured the participant’s risk preference. Both measures will be explained in section 3.3 and 3.4. After the participants filled in their demographic information, the participants were done with the experiment and the debriefing of the whole experiment took place. The participants received € 6,50 (or 2 credits in SONA) as a fee for their participation. An additional earning (ratio units to euro’s 100:15) was based on the decisions made in all three phases of the experiment, by randomly selecting outcomes of the trials in phase 1 and phase 2. In addition, the sanctioning points (expressed in Monetary Units) of the principal could be added or subtracted from this additional earning. The amount of money was transferred to their bank account a couple of days after the experiment. The total duration of the execution of the experiment was about 60 minutes on average.

3.1 Contest version of the attacker-defender game

The contest version of the attacker-defender game goes further than just choosing to cooperate or defect like in the normal AD-G. Two players, the defender (role A) and attacker (role B), have to choose how much resources to invest out of a given endowment (De Dreu et al., 2019). Participants do not get the name as “defender” or “attacker”, but are just assigned to the name of role A or B with the suited instructions for each role. The attacker chooses how much to invest in attack, whereas the defender chooses how much to invest in defence. The given

endowment capital is explained in units. Both players will receive 10 units for each trial. In every trial a question will appear asking the participant to choose how much to invest between 0 till 10 units out of the given endowment. If the participant is the defender (role A) and invests more

23 units (or the same amount) than the other participant as the attacker (role B), the defender wins and they both keep what is left of their own endowment. For example: the defender invests 4 units, while the attacker invests 3 units. The defender wins, and is left with 6 units, whereas the attacker is left with 7 units. If the defender invests less units than the attacker, the attacker wins and only the attacker receives the units of the defender that are left of their endowment plus the attacker’s own leftover endowment. For example: the defender invests 4 units, while the attacker invests 5 units. The attacker wins, and receives 6 units that are left of the defender and 5 units that are left of the attacker itself (total units = 11). As the current study focusses on the third phase of the experiment, the outcomes made by the attacker (or defender) affect the principal only. So the principal will receive the units that are gained by the attacker.

3.2 Sanctioning task of the principal

In the third phase of the experiment participants are asked to sanction (reward or punish) 66 (or 21 for defenders) possible decisions the attackers could have made during the contest version of the AD-G. These participants were now in the role of the principal (third party). A list of all possible investments and outcomes of which the attacker and defender could make were shown to the participant. The outcomes of this phase of the contest AD-G affected the principal only and not the players. The principal made a sanction decision to the agent (attacker or defender) who made decisions on behalf of them. The participant had to sanction the agent (attacker) and could choose to reward (points given from +1 to +3), to punish (point given from -1 to -3) or to do nothing (0) based on the given decisions. These sanction points were expressed as Monetary Units (MU). If a principal would assign MU’s to a certain investment or outcome made by the agent, the agent would receive it threefold its value (e.g. if the principal gives +2 MU’s to an agent, this leads to a gain in +6). In the end one sanctioning decision would be

24 chosen at random and entered for actual payment for both participants (principal and agent). The term punishment was not used in the experiment and was replaced by assigning “deduction points”.

3.3 Social value orientation: Slider Measure

Several different measurements to measure the social value orientation have been developed, like the Ring Measure (Liebrand, 1984) or the Triple-Dominance Measure (Messick & McClintock, 1968). Murphy et al. (2011) introduced a third measurement called the Slider Measure. The Slider Measure in the current study consisted of 15 question items that each showed nine different ways to distribute a certain amount of units between the participant and another player. The participant had to choose which distribution he or she preferred. Each participant was randomly matched with another participant from the experiment. It was stressed beforehand that all decisions made would be anonymous and kept in confidence. The outcomes of these decisions are computed into 4 orientations of the Slider Measure: (i) altruistic; (ii) prosocial; (iii) individualistic; (iv) competitive. (i) The altruistic orientation is expressed by people who will attempt to minimize the difference between the outcomes of oneself and the other. (ii) The prosocial orientation focusses on maximizing the collective outcome. (iii) The individualistic orientation is mainly focussed on maximizing the own outcome. (iv) The competitive orientation attempts to make the difference as big as possible between the own outcome and the outcome of the other. The current exploratory study is interested in the

differences between proself and prosocial orientations of principals. The proself orientation will be identical to the individualistic orientation of the Slider Measure (Murphy et al., 2011). The conversion rate for the Slider Measure questionnaires was 100:6.

25

3.4 Risk preferences: Gamble Task

The Gamble Task in this study consisted of 32 questions (Holt & Laury, 2002). One question consisted of choosing between two options, option A or option B. Option A consisted of a 50/50 gamble decision with different sizes of outcomes, whereas option B was a certain, fixed outcome. Only one decision was randomly selected at the end of the experiment. The conversion rate for the Gamble Task was 100:50. The results of the Gamble Task will not be discussed further in the present study, as the focus of the study is on the sanctioning behaviour and the social value orientation of the principal.

4. Statistical analysis

The statistical analysis are done with IBM SPSS. The data of the lab experiment of 2019 was collected by computers at the Level Leiden Lab. This data was directly transferred to an SPSS-file without the possibility of human error. Hypothesis 1.0 & 2.0 & 3.0 will be investigated by two one-way ANOVA analyses and a 2 x 2 (within-subjects) repeated measures ANOVA design. The within-subjects factors are Gain: which focusses on the distinction between a “win” outcome and a “loss” outcome in units gained by the principal, and Victory: which focusses on the distinction between victory or no victory outcomes of the attacker. For the variable Gain, a “win” outcome is defined as units gained whenever the outcomes of the attacker are > 10 units. A “loss” outcome is defined as units gained whenever the outcomes of the attacker are < 11 units. The starting capital of 10 units is taken as a reference point to define these win and loss

outcomes. For the variable Victory, the attacker is victorious when he or she invested more units than the defender in the contest version of the AD-G. Whenever the attacker invests the same or less units than the defender, this is defined as no victory. The dependent variable is the amount a principal invests in sanctioning (punishment or reward from -3 to +3 and 0) towards the attacker,

26 which will be looked at in both Gain (win versus loss) and Victory (yes versus no) factors. Both factors have two levels, which will be computed into new variables in the SPSS file. For both

Gain (with levels GainLoss-Sanction and GainWin-Sanction) and Victory (with levels

YesVictory-Sanction and NoVictory-Sanction) the average sanction score for each participant will

be calculated. After all analyses are done, two separate one-sample t-tests will be performed to see if the sanctioning behaviour of the principals, in the different conditions, differs from zero.

4.1 Assumptions

Before statistical conclusions can be made, the assumptions should be checked. First by exploring the data, looking for missing data, outliers and influential points. Outliers will be checked by looking at the standard deviations from the mean and by looking at the leverage value. If the standardized residuals are more than 3 standard deviations from the mean, the outlier will be removed from the data (Pallant, 2013). Next to that the leverage value should be lower than < 3(k+1)/N to not have any outliers on the independent variables. If the Cook’s distance is larger than 1, there are influential data points. The assumption of normality should be checked by looking at the number of participants in each group. If the N is reasonably large (N > 15) in each condition, the F is robust against violation. To check the sphericity assumption, the degree of sphericity is used by investigating epsilon (e). If the value of the Greenhouse-Geisser is ≤ .75, the GG-corrected F test should be used. If the value is higher than .75 the Huynh-Feldt should be used for F correction. If the analyses consists of variables with only two levels, the sphericity assumption is (in most cases) met (Pallant, 2013).

27 4.2 Exploratory analysis

Hypothesis 4.0 & 5.0 & 6.0 will be investigated by a 2 x 2 x 2 (within-subjects and between subjects) mixed ANOVA design. The within-subjects factors are Gain: which focusses on the distinction between “win” (gain) outcomes and “loss” (gain) outcomes, and Victory: which focusses on the distinction between victory or no victory outcomes of the attacker. The between-subjects factor is the SVO (social value orientation) with the between between-subjects SVO-prosocial and

SVO-proself. The dependent variable is the investment in sanctioning, punishment or reward (-3

to +3), of the principal. Before statistical conclusions can be made, the same assumptions as section 4.1 will be checked.

28

Results

Participant information

In total, 114 participants took part in the lab experiment, consisting of 88 females and 26 males between the age of 17 and 32. The average age for attackers was M = 20.91, SD = 2.57, and for the defenders it was M = 21.35, SD = 2.53. The 114 participants participated in dyads, so

N = 57 for attackers (includes the corresponding principals as well) and N = 57 for defenders.

Looking at the identification of the social value orientation, 71 participants were categorized as social and 43 as self. In regard to the attackers the distribution was 35 participants pro-social and 22 pro-self.

Part 1: Assumptions

Before all analyses were done, the data was explored and the assumptions were checked of the two separate one-way repeated measures ANOVA. Firstly, the data was explored of factor

Gain (with levels GainLoss-Sanction and GainWin-Sanction) and factor Victory (with levels YesVictory-Sanction and NoVictory-Sanction). The distribution of the data of these factors is

shown in Figure 2.0. There were no missing cases found. One (extreme) outlier on the dependent variable was found in both factors who had a standardized residual value outside the range (< -3 or > 3), so this person was removed from the data (Pallant, 2013), which resulted in N = 56. Next to that, the leverage value was .02, which is lower than .16 (3(2+1)/56), so there were no outliers on the independent variables. The largest Cook’s distance for all levels was .17, which is smaller than 1, so there were no influential data points.

29 Next to that, the F is robust against violation of the normality assumption since the N is reasonably large (N > 15) in each condition (N = 56) (Pallant, 2013). The value of the

Greenhouse-Geisser for both factors is 1.0, as both factors have only two levels, which means the sphericity assumption is met. Thereafter, the one-way repeated measures ANOVA for both factors were separately executed first.

Figure 2.0. Boxplot of the distribution of the data of the two factors with both their two levels.

On the horizontal axis the factors Gain (win/loss) and Victory (yes/no). On the vertical axis the distribution of the sanctioning value of the principal.

Part 1: Within-subject effects

A one-way repeated measures ANOVA was conducted for the factor Gain. The amount of sanctioning differed statistically significantly between Gain outcomes (F(1,55) = 122.491, p < .001). This means there is a statistically significant difference between the means of the two

30 levels of the within-subjects factor Gain (GainLoss-Sanction and GainWin-Sanction). The post hoc test using the Bonferroni correction showed that GainWin-Sanction was significantly different from GainLoss-Sanction (p < .001), where the sanctioning in GainWin-Sanction is higher (M = 1.173, SD = .797) than in GainLoss-Sanction (M = -.033, SD = .714). This means that the principal’s sanctioning behaviour is higher (more in the direction of reward) towards attackers who have “win” outcomes, by gaining more than 10 units for the principal, than the sanctioning behaviour towards attackers who have “loss” outcomes, by gaining less than 11 units for the principal (more in the direction of punishment). In addition a one-sample t-test was performed to see if the means of both levels differed from zero. Principals in the

GainWin-Sanction condition reported higher sanctioning values (M = 1.173, SD = .797), in the direction of

reward, which differed significantly from zero (t (55) = 11.008, p < .001). Principals in the

GainLoss-Sanction condition reported lower sanctioning behaviour values (M = -.033, SD =

.714), in the direction of punishment, but this does not differ significantly from zero (t (55) = - .347, p = .730).

Another one-way repeated measures ANOVA was conducted for the factor Victory. The amount of sanctioning differed statistically significantly between Victory outcomes (F(1,55) = 81.112, p < .001). This means there is a statistically significant difference between the means of the two levels of the within-subjects factor Victory (NoVictory-Sanction and

YesVictory-Sanction). The post hoc test using the Bonferroni correction showed that YesVictory-Sanction

was significantly different from NoVictory-Sanction (p < .001), where the sanctioning in

YesVictory-Sanction is higher (M = .527, SD = .640) than in NoVictory-Sanction (M = -.094, SD

= .763). This means that the principal’s sanctioning behaviour is higher (a little more in the direction of reward) towards attackers who are victorious, by investing more units than their

31 opponent (defender), than the sanctioning behaviour towards attackers who are not victorious, by investing the same or less units than their opponent (a little more in the direction of punishment). In addition a one-sample t-test was performed to see if the means of both levels differed from zero. Principals in the YesVictory-Sanction condition reported higher sanctioning values (M = .527, SD = .640), in the direction of reward, which differed significantly from zero (t (55) = 6.166, p < .001). Principals in the NoVictory-Sanction condition reported lower sanctioning behaviour values (M = -.094, SD = .763), in the direction of punishment, but this does not differ significantly from zero (t (55) = - .924, p = .360).

Part 2: Assumptions

Thereafter the 2 x 2 repeated measures ANOVA was performed with the two previous within-subject factors Gain and Victory. The assumptions were checked again. There were no missing cases found and no outliers on the dependent variable, as both factors had standardized residual values within the range (< -3 or > 3). Next to that, the leverage value was the same as explored before (.02), which is lower than .16 (3(2+1)/56), so there were no outliers on the independent variables. The largest Cook’s distance for all levels was .19, which is smaller than 1, so there were no influential data points. There are no sphericity results, as there are only two levels of each within-subjects factor, so we assume the sphericity assumption is met.

Part 2: Two-way repeated measures ANOVA

The results of the two-way repeated measures ANOVA revealed that there was a

significant main effect of Gain on the sanctioning behaviour of the principal (F(1,55) = 121.143,

p < .001). This means that a principal who sanctions an attacker in regard to “win” outcomes does

32 principals show higher sanctioning values towards “win” outcomes of the attacker (M = .757, SD = .089) than towards “loss” outcomes (M = .121, SD = .090). In addition, a significant main effect was found of Victory on the sanctioning behaviour of the principal (F(1,55) = 53.762, p < .001). This means that a principal who sanctions an attacker when they are victorious does significantly differ from sanctioning no victorious outcomes made by the attacker. Principals show higher sanctioning values towards victory outcomes (M = .508, SD = .085) than towards no victory outcomes (M = .370, SD = .085). Next to that, the interaction effect between Gain *

Victory was significant (F(1,55) = 133.773, p < .001). This shows that the sanctioning behaviour

of principals was the highest in “win” outcomes (more towards reward), no matter if the attacker was victorious or not, compared to “loss” outcomes (more in between reward and punishment). This can be seen in the descriptive statics (Table 1.0) as well as in the estimated marginal means plot (Figure 3.0). What also stands out, looking at the descriptive statistics (Table 1.0), is that the sanctioning behaviour is higher in the GainWin-NoVictory condition (M = .786) compared to the

GainWin-YesVictory condition (M = .729).

Table 1.0

Descriptive statics of factors Gain (Win/Loss) with Victory (Yes/No).

M SD N

GainWin-YesVictory .729 .660 56

GainWin-NoVictory .786 .681 56

GainLoss-YesVictory .288 .647 56

33 Figure 3.0. Estimated Marginal Means plot of Victory and Gain. On the horizontal axis the

factor Victory, with an attacker being victorious or not. On the vertical axis the estimated marginal means of the sanctioning value of the principal. The red line represents the win outcomes condition of the factor Gain and the blue line represents the loss outcome condition.

Part 3: Assumptions

Of the N = 56 used in the analyses, 22 principals were pro-self and 34 pro-social.

A 2 x 2 x 2 mixed ANOVA was performed with the within-subject factors Gain and Victory and the between-subject factor SVO-Attacker (pro-self / pro-social social value orientation of the principal towards the attacker). Beforehand the assumptions were checked again. The F is robust against violation of the normality assumption since the N is reasonably large (N > 15) in each condition (N = 56). In addition, the F is robust with regard to the homogeneity of variance assumption, since Levene’s test of equality of error variances is non-significant for all factors (p > .458). There were no missing cases found and no outliers on the dependent variable, as all

34 factors had standardized residual values within the range (< -3 or > 3). Next to that, the leverage value was .05, which is lower than .21 (3(3+1)/56), so there were no outliers on the independent variables. The largest Cook’s distance for all levels was .30, which is smaller than 1, so there were no influential data points. There are no sphericity results, as there are only two levels of each within-subjects factor and also for the between-subject factor, so we assume the sphericity assumption is met.

Part 3: Mixed-ANOVA

The 2 x 2 x 2 mixed ANOVA is used to determine whether a change in sanctioning behaviour of the principal is the result of the interaction between the within-subjects factors (Gain and Victory) and the between-subject factor (SVO-Attacker). The results showed similar significant effects for the factors Gain and Victory as before and also for the interaction effect

Gain * Victory (see Table 2.0). It did not show any significant effects of SVO-Attacker with the

within-subject factors (see Table 2.0). Also, the test of between-subjects effect shows a

nonsignificant result (F(1,54) = 2.072, p = .156). This means it cannot be said that the change in sanctioning behaviour of the principal is due to the interaction between Gain, Victory and the social value orientation of the principal towards the attacker.

35 Table 2.0

Test of within-subject effects of Gain and Victory and their interaction effect with each other and the Social Value Orientation of the principal towards the attacker.

F df df error p-value

Gain 111.922 1 54 < .001

Gain * SVO attacker .181 1 54 .672

Victory 53.438 1 54 < .001

Victory * SVO attacker .641 1 54 .427

Gain * Victory 104.750 1 54 < .001

36

Discussion

Conclusion

The aim of this study was to investigate how principals sanction their agent’s attempts (attacker) at exploiting another party (defender) through attack, and whether sanctions are influenced by the outcomes of the agent’s contest decisions (win/loss and victory/no victory). In this study the main focus was on the outcomes made by the attacker and sanctioning behaviour towards the attacker (not the defender), to not overcomplicate the research design and analysis. The results showed that principals do on average punish losses and reward win outcomes made by the attacker, where they have significantly higher sanction values (in the direction of reward) towards the attacker in win outcomes compared to loss outcomes (hypothesis 1.0). Next to that, the results also showed that principals do on average punish no victory and reward victory outcomes made by the attacker, where they have significantly higher sanction values (in the direction of reward) towards the attacker in victory outcomes compared to no victory outcomes (hypothesis 2.0). The significant interaction effect showed that principals have higher sanction values towards win outcomes, where the principal gains more than 10 units (with 10 units of the given endowment as reference point), no matter if the attacker is victorious or not (hypothesis 3.0).

The present study also explored if the social value orientation affected the principal’s sanctioning behaviour towards the attacker. To see if there were differences between prosocial and proself principals. The results showed no significant results of the social value orientation towards the sanction behaviour of principals in the win/loss conditions, as well as in the

37 victory/no victory conditions. This means that all exploratory hypotheses (hypotheses 4.0 & 5.0 & 6.0) should be rejected.

To conclude, the current study shows how principals sanction towards attackers in a contest version of the AD-G. The principals’ sanctioning behaviour towards the attacker does differ between win and loss outcomes, as well as between victory and no victory outcomes. In general, principals have higher sanction values (more in the direction of reward) towards win outcomes compared to loss outcomes, no matter if the outcome is a victory or no victory outcome. Finally, the social value orientation does not seem to have any influence on the sanctioning behaviour of the principal (in all outcome conditions).

Limitations & future research

The significant results in regard to hypotheses 1.0 & 2.0 & 3.0 will be discussed. After that the non-significant results in regard to hypotheses 4.0 & 5.0 & 6.0 will be reviewed. In addition, some critical notes on the design of the current study will be discussed and some advice for future research will be suggested.

To begin with, the principal does indeed sanction (in direction of reward) win outcomes (success) and victory outcomes, whereas loss outcomes (failure) and no victory outcomes are sanctioned more in the direction of punishment, which previous studies suggested in other economic games (Balliet et al., 2011; Fehr & Fischbacher, 2004b; Gächter & Kirchsteiger, 1997; Kwaadsteniet et al., 2013; Nikiforakis & Mitchell, 2014). Next to that, in general, win outcomes, where the principal receives more units (> 10) than what the agent (attacker) originally started with, have higher sanctioning values of the principal (in the direction of reward) compared to loss

38 outcomes, no matter if the agent was victorious in the contest version of the AD-G or not. This could be due to the fact that in the win and loss outcomes the distinction is more clear in terms of units gained and more certain for the principal what he or she will get. In the case of victory, the principal could still get less units than the reference point of 10 units. Participants may have only focussed on gaining units, no matter if they are victorious of not (Miller, 2001; Schwartz, 1986). Future research might want to investigate if similar results would occur in sanctioning behaviour towards defenders. Moreover, it would be an interesting study to see if there are differences between sanctioning behaviour towards attackers who act in the interest of the principal compared to attackers who act in their own interest, or in both interests (both attacker and principal).

In addition, the exploratory study investigated if the social value orientation of principals could be of influence in their sanctioning behaviour and if there would be differences between prosocial and proself principals. The results showed no significant results, which means that the change in sanctioning behaviour of the principal towards the attacker is not due to the interaction between win/loss outcomes or victory/no victory outcomes. This may be due to the exploratory approach of the social value orientation’s influence in this particular design setting, as the results have been found in other economic decision games and social dilemma games (Kuhlman & Wimberley, 1976; Kwaadsteniet et al., 2006; Parks, 1994) where the social value orientation influenced participant’s decision making. Nevertheless, in research of De Dreu et al. (2019) the social value orientation did influence the player’s investment behaviour in the contest version of attacker-defender game. Although that research is focussed on the players (attacker/defender), similar findings could have been found in principals as well. Furthermore, there is a possibility that the Slider Measure is not sufficient in capturing the right social orientations of the principals.

39 De Dreu et al. (2019) used the levels of other-concern and empathy to measure someone’s social preference. Also the Ring Measure (Liebrand, 1984) or the Triple-Dominance Measure (Messick & McClintock, 1968) are often used in studies to measure the social value orientation, which may be able to capture someone’s social preferences even better in future research.

Next to that, the design of the current study could have had an influence on participant’s decision behaviour. Even though each participant finished the short summary task to see if they understood the instructions of the lab experiment, it still was quite a complex setting with a lot of information and instructions to process. In total, 114 participants took part in the lab experiment in the year 2019. It should be advised to replicate the study with a larger group of participants to increase reliability and generalizability of the population (Cohen, 1988). The design of the study was (somewhat) underpowered (especially in regard to the social value orientation measurement), which may have led to the lack of discovering real findings. Low statistical power reduces the main purpose of the study by reducing the change of detecting a real effect (Button et al., 2013). If future research replicates the study using the Ring Measure (Liebrand, 1984) or the Triple-Dominance Measure (Messick & McClintock, 1968) for example, this may lead to different results. By increasing the sample size, the more one can generalize the results, and the more the power increases of the study (Cohen, 1988). In addition, the participants consisted of 88 females and 26 males between the age of 17 and 32. This also questions the generalizability and should be done in future research with equal groups (Franz & Loftus, 2012). More research is needed in relation to social-psychological mechanisms and the effect on sanctioning behaviour of principals to get a better understanding which variables influence this behaviour (e.g. age, sex).

40 The above questions surrounding the design of the study and the overall recommendations should be implemented in future research. The current study showed how principals would sanction their attacker (agent) and that different outcomes (win/loss and victory/no victory) would influence this sanctioning behaviour (in general more in the direction of reward). If people sanction in line with their social value orientation is not confirmed and should be investigated further. However, the results of this study are valuable for economic decision-making settings (e.g. employers with an employee) to see how a third party would respond to outcomes made by others that would influence the third party itself. Altogether, this study deepens the understanding of human sanctioning behaviour in conflict situations. This will help future research to predict and influence people’s sanctioning behaviour in a similar setting like an economic contest game.

41

References

Allison, S. T., & Messick, D. M. (1990). Social decision heuristics in the use of shared resources. Journal of Behavioral Decision Making, 3(3), 195-204.

Balliet, D., Mulder, L. B., & Van Lange, P. A. (2011). Reward, punishment, and cooperation: a meta-analysis. Psychological bulletin, 137(4), 594-614.

Balliet, D., Parks, C., & Joireman, J. (2009). Social value orientation and cooperation in social dilemmas: A meta-analysis. Group Processes & Intergroup Relations, 12(4), 533-547.

Baron, J. (1995). Blind justice: Fairness to groups and the do‐no‐harm principle. Journal of

behavioral decision making, 8(2), 71-83.

Benhabib, J., & Rustichini, A. (1996). Social conflict and growth. Journal of Economic

growth, 1(1), 125-142.

Button, K. S., Ioannidis, J. P., Mokrysz, C., Nosek, B. A., Flint, J., Robinson, E. S., & Munafò, M. R. (2013). Power failure: why small sample size undermines the reliability of neuroscience. Nature Reviews Neuroscience, 14(5), 365-376.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates.

42 Coleman, J. S. (1990). Foundations of social theory. Cambridge, Mass: Belknap Press of Harvard

University Press.

De Cremer, D., & Van Lange, P. A. M. (2001). Why prosocials exhibit greater cooperation than proselfs: The roles of social responsibility and reciprocity. European Journal of

personality, 15(1), 5-18.

De Dreu, C. K., Giacomantonio, M., Giffin, M. R., & Vecchiato, G. (2019). Psychological constraints on aggressive predation in economic contests. Journal of Experimental

Psychology: General, 148(10), 1767-1781.

De Dreu, C. K., & Gross, J. (2019). Revisiting the form and function of conflict: Neurobiological, psychological, and cultural mechanisms for attack and defense within and between groups. Behavioral and brain sciences, 42, 1-66.

Fehr, E., & Fischbacher, U. (2004a). Social norms and human cooperation. Trends in cognitive

sciences, 8(4), 185-190.

Fehr, E., & Fischbacher, U. (2004b). Third-party punishment and social norms. Evolution and

human behavior, 25(2), 63-87.

Fehr, E., Fischbacher, U., & Gächter, S. (2002). Strong reciprocity, human cooperation, and the enforcement of social norms. Human nature, 13(1), 1-25.

43 Fehr, E., & Gachter, S. (2000). Cooperation and punishment in public goods

experiments. American Economic Review, 90(4), 980-994.

Fehr, E., Gächter, S., & Kirchsteiger, G. (1997). Reciprocity as a contract enforcement device: Experimental evidence. ECONOMETRICA, 65, 833-860.

Fehr, E., & Schmidt, K. (1999). A theory of fairness, competition and cooperation. Quarterly

Journal of Economics, 114, 817–868.

Franz, V. H., & Loftus, G. R. (2012). Standard errors and confidence intervals in within-subjects designs: Generalizing Loftus and Masson (1994) and avoiding the biases of alternative accounts. Psychonomic bulletin & review, 19(3), 395-404.

Goeree, J. K., Holt C. A., & Palfrey, T. R. (2003) Risk averse behavior in generalized matching pennies games. Games and Economic Behavior, 45, 97–113.

Holt, C. A., & Laury, S. K. (2002). Risk aversion and incentive effects. The American Economic

Review, 92(5), 1644-1655.

Jensen, M. C. (1994). Self‐interest, altruism, incentives, and agency theory. Journal of applied

44 Kelley, H. H., & Thibaut, J. W. (1978). Interpersonal relations: A theory of interdependence.

New York: John Wiley & Sons.

Koford, K., & Penno, M. (1992). Accounting, principal-agent theory, and self-interested behavior. The Ruffin Series in Business Ethics, 1, 127-142.

Kopelman, S., Weber, J. M., & Messick, D. M. (2002). Factors influencing cooperation in commons dilemmas: A review of experimental psychological research. The drama of

the commons, 2, 113-156.

Kuhlman, D. M., & Wimberley, D. C. (1976). Expectations of choice behavior held by

cooperators, competitors, and individualists across four classes of experimental game.

Journal of Personality and Social Psychology, 34, 69-81.

de Kwaadsteniet, E. W., van Dijk, E., Wit, A., & de Cremer, D. (2006). Social dilemmas as strong versus weak situations: Social value orientations and tacit coordination under resource size uncertainty. Journal of Experimental Social Psychology, 42(4), 509-516.

de Kwaadsteniet, E. W., Rijkhoff, S. A., & van Dijk, E. (2013). Equality as a benchmark for third-party punishment and reward: The moderating role of uncertainty in social dilemmas. Organizational Behavior and Human Decision Processes, 120(2), 251-259.

45 Liebrand, W. B. G. (1984). “The effect of social motives, communication and group size on

behaviour in an n-person multi stage mixed motive game”. European Journal of Social

Psychology, 14(3), 239–264.

McGillivray, F., & Smith, A. (2000). Trust and cooperation through agent-specific punishments. International Organization, 54(4), 809-824.

Messick, D. M., & McClintock, C. G. (1968). "Motivational Bases of Choice in Experimental Games". Journal of Experimental Social Psychology, 4, 1-25.

Miller, D. T. (2001). The norm of self-interest. The next phase of business ethics: Integrating

psychology and ethics, 3, 193-210.

Mulder, L. B., Van Dijk, E., De Cremer, D., & Wilke, H. A. (2006). Undermining trust and cooperation: The paradox of sanctioning systems in social dilemmas. Journal of

Experimental social psychology, 42(2), 147-162.

Murphy, R. O., Ackermann, K. A., & Handgraaf, M. (2011). Measuring social value orientation. Judgment and Decision making, 6(8), 771-781.

Oberschall, A. (1978). Theories of social conflict. Annual review of sociology, 4(1), 291-315.

Pallant, J. (2013). SPSS survival manual: A step by step guide to data analysis using IBM SPSS (4th ed.). Crows Nest, NSW: Allen & Unwin.

46 Parks, C. D. (1994). The predictive ability of social values in resource dilemmas and public

goods games. Personality and Social Psychology Bulletin, 20(4), 431-438.

Petersen, T. (1993). The Economics of Organization: The Principal-Agent Relationship. Acta

Sociologica, 36(3), 277–293.

Pruitt, D. G., & Carnevale, P. J. (1993). Negotiation in social conflict. Thomson Brooks/Cole Publishing Co.

Rapoport, A., & Au, W. T. (2001). Bonus and penalty in common pool resource dilemmas under uncertainty. Organizational behavior and human decision processes, 85(1), 135-165.

Salancik, G. R., & Pfeffer, J. (1974). The bases and use of power in organizational decision making: The case of a university. Administrative Science Quarterly, 2, 453-473.

Samuelson, W. & Zeckhauser, R. (1988). Status quo bias in decision making. Journal of Risk and

Uncertainty, 1, 7-59.

Schwartz, B. (1986). The battle for human nature. New York: Norton.

Tomasello, M. (2014). The ultra-social animal. European Journal of Social Psychology, 44(3), 187-194.

47 Van Lange, P. A. M. (1999). The pursuit of joint outcomes and equality in outcomes: An

integrative model of social value orientations. Journal of Personality and Social

Psychology, 77, 337–349.