Legitimate and ambiguous entitlements in the ultimatum game

(1)

THESIS

LEGITIMATE AND AMBIGUOUS ENTITLEMENTS

IN THE ULTIMATUM GAME

Submitted by

Maria Carolina Pena Madeira Gouveia de Campos, Student number: 10604731 28/06/2015

Master of Science in Economics

Specialisation: Behavioural Economics and Game Theory

Faculty of Economics and Business, University of Amsterdam, The Netherlands Supervisor: Dr. Joël van der Weele

(2)

Statement of originality

This document is written by Student Maria Carolina Pena Madeira Gouveia de Campos who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

“What could not easily be anticipated is how socially, motivationally, and emotionally rich already the very simple ultimatum game turned out to be.” (Güth & Kocher, 2014)

(4)

An Ultimatum Game where the proposer earns the right to split the endowment and this surplus is endogenously produced by the responder is studied. This competitive setup represents a context where

both parties are differently entitled to the bargaining object, so their entitlements become hard to compare. This ambiguous treatment failed to produce any significant differences in players’ behaviour.

Nevertheless, when it was clearer both were entitled to the surplus, proposers more frequently leaned towards equality and responders became more demanding. A cooperative even split assigns equal weight to both entitlements and seems to be the solution that gains acceptance on both sides of the

(5)

1. Introduction

The ultimatum game (UG) has been extensively studied in economics to uncover the myriad of motivations behind players’ behaviour. This game involves two players: a proposer and a responder. The proposer is endowed with a certain amount of money or tokens to split with the responder. The responder can either accept or reject the offer made by the proposer: acceptance means a split in accordance with the proposer’s offer, rejection yields a payoff of zero for both players. Rationality dictates the recipient will accept any positive offer (as any amount, however small, is greater than zero) and thus the proposer will offer the smallest amount possible, keeping virtually all of the initial pie1_.

Game-theoretic predictions seem nevertheless to fail in economic experiments. Proposers offer more than the minimum required by backward induction and responders reject low offers:

“On average, players in the game tend to offer around 40-50% of the pie in the standard version of the game. Such offers are almost always accepted. Responders’ acceptance rates decrease with smaller offers, and they approach zero quite quickly for offers below 20%.”

(Güth & Kocher, 2014, p.398)

Rejection is triggered by a sense of unfairness that comes from a low share of the total endowment and a mixture of fairness concerns and fear of rejection explain proposers’ high offers. Responders' rejection of small offers appears to have an emotional nature too: the anger that may follow a perception of unfairness can cause spite and a need to hurt the proposer (Pillutla & Murnighan, 1996). On proposers’ side, offers cannot be explained by fairness concerns alone, but also by strategic considerations on the amount their counterpart will reject. In fact, were proposers’ sense of fairness to be the only explanation for their offers in the UG, one should not observe

1_{The words “pie”, “endowment”, “amount”, “resources”, “surplus”, “stakes” are used in the text as synonyms and refer to}

(7)

2 significant differences in behaviour between the UG and the dictator game (DG), a game where strategic concerns do not play a role (since the recipient has to passively accept any offer):

“If players give away money only because of a desire to be fair, the distributions of proposals in dictator and ultimatum games with equal pies would be identical. This clearly did not happen in our games with pay, where the fairness hypothesis is strongly rejected.”

(Forsythe, Horowitz, Savin, & Sefton, 1994, p.362).

Traditionally, experiments with the UG are silent about the source of the proposer's endowment and how the roles came about (roles are usually randomly assigned). These omissions can result in different interpretations and prompt participants to look for familiar analogues of the game, especially given the artificial and neutral context in which the game is usually presented. In a cross-cultural study where the game was played in different small societies, behaviour was closely linked to patterns of interaction in daily life (Henrich et al., 2005). The degree of pro-social behaviour observed in the game greatly depended on social and organisational norms specific of every society, namely the level of market integration and the value attached to cooperation. Besides, allocating UG roles by chance also has important implications on perceptions of entitlement: “Participants who received their position in such a way [by chance] may not feel entitled to exploit its strategic possibilities” (Güth, 1995, p.331).

Many authors have filled in this gap in UGs or DGs through mechanisms such as property rights (Leliveld, van Dijk, & van Beest, 2008), joint production of an endowment (Bediou, Sacharin, Hill, Sander, & Scherer, 2012; Feng et al., 2013), earned role in the game through a performance in a quiz (Hoffman, McCabe, Shachat, & Smith, 1994; Hoffman, McCabe, & Smith, 1996), and earned endowment (Cherry, Frykblom, & Shogren, 2002; Oxoby & Spraggon, 2008; Ruffle, 1998; Carr & Mellizo, 2013).

(8)

3 cases where this surplus is not earned or deserved. These studies emphasise entitlement on one side of the game only. However, bargaining and distributive issues usually involve parties with legitimate and competing arguments for the demands they make. In other words, real strategic and distributive interactions tend to bring together parties with simultaneous and legitimate claims to the bargaining object. This happens in negotiations between employees and managers, political parties forming a coalition2_{, a brand and a retailer drawing up a contract. The recent exchange of arguments between} Greece and its creditors also clearly illustrates a negotiation involving parties with coexisting demands that are both legitimate and conflicting.

The introduction of coexisting and legitimate entitlements has not been investigated in either the DG or the UG. In fact, behaviour in these games has only been studied in settings where certain criteria or principles make one side clearly more entitled to the endowment (for instance, the proposer earns the right to his or her position or the proposer (and/or responder) gets the chance to produce an endowment at a stage prior to the game). What happens when both sides of the UG have legitimate and conflicting entitlements to the bargaining object? This thesis investigates proposer and responder behaviour in a conflicting and ambiguous UG where both parties are legitimately entitled to the pie. In particular, in this scenario:

1) the proposer earns the right to decide over the resources and 2) the responder produces the endowment.

The first question to be answered is the following: how do proposers behave when both parties have entitlements to the endowment? The answer to this question is far from straightforward as proposers may be faced with conflicting motivations. On the one hand, earning the right to be the proposer induces a sense of entitlement over the pie that can only be acknowledged through a low offer. On the other hand, the fact that the responder is the one producing the pie pulls the proposer in the opposite direction, i.e. towards more generous offers that honour responder’s entitlement. The

(9)

4 second research question regards responder behaviour: are responders more or less willing to accept low offers from an endowment they have themselves produced and made by proposers who have legitimately earned the right to split these resources?

By creating a realistic ambiguous scenario, this thesis builds on previous work that has varied entitlements of either player in a DG or UG. In particular, players’ behaviour has been studied in contexts where the responder produced the resources or the proposer earned his or her role. In the conflicting case studied, these legitimate entitlements are simultaneous and therefore, become hard to compare. At this point, it makes sense to justify the choice of the UG (over the DG) to study behaviour in this case. In real life, power of decision-makers is not unlimited: it can be challenged and resisted. Also in the UG, responders have the power to reject an offer, so proposers are required to foresee the entitlements and claims of second players. Finally, the introduction of entitlements and property rights has received more attention in DGs than in the UG (see, for instance, Cherry et al. (2002), Oxoby & Spraggon (2008) and others’ focus on DGs).

In short, proposer and responder behaviour will be analysed in an UG where the proposer earns the right to split resources that have been produced by the responder. The results will shed light on bargaining outcomes when both negotiating parties have their own distinct and subjective entitlement to the object. As in the definitions used by Hoffman & Spitzer (1985), “entitlements” are understood as claims to resources and “rights” as “morally justified entitlements” (Hoffman & Spitzer, 1985, p.260). The entitlements or claims of the responder arise from this player’s direct production of the endowment and the proposer’s “morally justified” right to split the resources comes from his or her successful performance in a skill task. Insofar as the proposer can only express his or her right through a monetary claim, the word “entitlement” as a claim is also used for this player. Said

(10)

5 differently, both players have a legitimate entitlement (i.e. claim) to the endowment: the proposer earns a “legal” entitlement; the responder has a “moral” one3_.

2. Literature review

Ultimatum Game overview

The UG resembles the last stage of a bargaining process where a “take-it-or-leave-it” offer is made. A proposer makes an offer (O) up to X currency units to the responder, who can either accept or reject the offer: acceptance means each party is paid according to the proposed split (O for the responder and X-O to the proposer); rejection means a payoff of zero for both players. Assuming a pie or endowment X that can be expressed up to the smallest unit (cent), backward induction determines that almost all of the initial pie will end up in the hands of the proposer, if both players play the game rationally4_.

The game was first tried experimentally in the late 1970s and the reported results were very different from the rational predictions above. In fact, Güth, Schmittberger, & Schwarze (1982) concluded that the proposer’s advantage in the game could not be fully exploited, as rejection happens if this player keeps a large share. Since then, research efforts have not only tried to explain consistent deviations from rationality (and even from justice ideals), but also to identify the conditions that ensure convergence to the game-theoretic equilibrium.

Responders appear to be concerned about other aspects than mere absolute payoff (Güth & Kocher, 2014). They care about relative payoff and punish proposers who make greedy offers. However, according to Pillutla & Murnighan (1996), fairness concerns alone are not enough to explain

3_{In Gächter & Riedl (2005), a “moral property right” referred to claims earned prior to negotiation and a “legal property}

right” pointed out players’ equal ability to negotiate. Players were “strategically equal”, since bargaining was “free-form” and both players would earn nothing if they did not reach an agreement (Gächter & Riedl, 2005, p.252). The authors concluded moral property rights preceded legal ones. In other words, despite both parties’ equal “legal” rights, “moral” rights acquired before bargaining shaped final outcomes. However, in Gächter & Riedl’s (2005) paper, equal legal rights were assigned to both players in the negotiation (rather than earned).

4_{For the game-theoretic equilibrium to occur both players are rational and this is common knowledge. This considers a}

(11)

6 responder’s behaviour in the UG. Rejection happens only when unfairness triggers anger and anger urges the need for revenge. The authors brought the emotional element to the analysis of responder behaviour as the “critical link” between fairness and behaviour. Besides, Pillutla & Murnighan (1996) claim that “intentionality” intensified responders’ concern for relative payoffs. Evidence nevertheless suggests that responders’ concern fairness is self-interested (Pillutla & Murnighan, 2003).

As for proposers, a combination of fairness reasons and strategic concerns seem to guide their behaviour. Differences in behaviour observed between the DG and UG suggest nevertheless that proposers are not guided by a “desire to be fair” (Forsythe et al., 1994). Other papers argue that proposers’ behaviour is not really driven by fairness. What guides proposers is rather a desire to appear fair, as abiding by socially acceptable rules will grant them acceptance from the other party. Fear of rejection seems then to better explain offers than fairness. Indeed, studies that varied information available to responders have shown that proposers are strategic and make use of their informational advantage when making their offer (for example, Pillutla & Murnighan, 1995; Croson, 1996; Kagel, Kim, & Moser, 1996). These studies have toned down the importance of fairness concerns in explaining proposers’ behaviour: ignoring responders’ expectations might result in conflict, which is why offers seem to reflect a self-interested strategy of proposers to the prospect of rejection5_.

The current review does not intend to be exhaustive. In fact, the works of Thaler (1988), Camerer & Thaler (1995), Güth (1995), Camerer (2003), De Oliveira & Eckel (2011) and Güth & Kocher (2014) provide already comprehensive overviews and summaries of important findings with UG variations that attempt to isolate different reasons and motivations behind both players’ behaviour in the game. Camerer (2003) presents five categories of variables tested in the UG: methodological, demographic, cultural, descriptive and structural. According to the author, “structural variables are the most useful to study because they connect simple games to richer economic structures (e.g., adding

(12)

7 competition) and also provide the most direct clues to the psychology underlying social preference” (Camerer, 2003, p. 75/76). Camerer (2003) identifies “entitlement”, a crucial concern of this thesis, as a structural variable. The following section aims at summarising some of the main findings of this line of research.

Entitlements in the Ultimatum Game and the Dictator Game

An equal split between both players is the “fair” solution in an UG with random assignment of roles. However, introducing entitlements challenges this notion of fairness. As Smith (1991) writes: “equal-split bargaining results [referring to Hoffman & Spitzer’s (1985) results later explained] may be due, generically, to an important treatment thought to be benign, namely, the standard use of random devices to allocate subjects to initial conditions.” (p. 883). Yet, “rights in the economy do not arise by random assignment, but are acquired through some economic or political process” (Smith, 1991, p. 883). Fairness explanations become then more complex once entitlements are introduced, as these create differences between players that may indeed justify departures from equal splits. The following papers suggest that fairness assessments are context-dependent and adjustable to attributes of the decision environment6_.

Güth & Tietz (1986) were one of the first to introduce some kind of entitlement to the surplus in the UG. In their paper, both positions of the game were auctioned before the game and the winner got the game’s earnings minus the price paid. According to Hoffman et al. (1994), this procedure greatly reduced offers7_.

Another pioneer study in entitlement literature is that of Hoffman & Spitzer (1985). In their experiment, participants took part in a game similar to the UG. A “controller” could decide to keep $12 for himself and 0 for the recipient or choose to cooperate with the recipient and jointly agree on the

6_{Frey & Bohnet (1995) identify important “institutional variations” affecting fairness in games. Property rights are one of}

them.

7_{The original paper of Güth & Tietz (1986) could not be accessed. However, it was mentioned several times in other}

(13)

8 division of $14. The Nash equilibrium calls for cooperation and a split of $13 for the controller and $1 for the recipient. The authors wanted nevertheless to understand the equality of splits ($7 for each player) observed in earlier experimental results. They compared behaviour of participants across different ways of assigning property entitlement and roles. In particular, the experiment had a 2×2 design with two variables: moral authority (i.e. framed as earning the role of controller) vs. no moral authority (i.e. simply being “designated” as controller) and earning the role through performance in a task/game (vs. coin flip). Over 90% of subjects opted for cooperation (maximisation of aggregate profits), but there were differences in splits. The amount kept by the controller was higher when this player had “moral authority” and when the position was earned rather than randomly assigned8_. However, moral authority was more powerful in changing behaviour than the game variable. Only when the difference between players was not “morally relevant” did participants tend to an egalitarian distribution. Hoffman & Spitzer (1985) argue that the effect of earning the role of controller (through a task) on splits shows subjects’ respect for a Lockean principle of just desert. The effect of moral authority is more ambiguous, as it can also simply reflect a respect for authority: “subjects can be prodded to regard entitlements as rights, without reference to a value-oriented theory of fairness (authoritarian explanation)” (Hoffman & Spitzer, 1985, p. 283).

Allowing the first mover to earn his or her position has also been extended to the UG. In Hoffman et al. (1994), proposers earned their right through performance in a quiz. Subjects took part in a contest that determined their role and their paired partner (the best player was paired with the worst one, the second highest was paired with the second lowest and so on). The authors suggest that common knowledge on the acquisition of (earned) property rights9_{harmonises expectations and} therefore lowers proposers’ concern with rejection. As a matter of fact, offers were lower when the

8_{The “greed index” (difference from equal split) was highest when moral authority was combined with the game treatment}

(Hoffman & Spitzer, 1985).

9_{The authors provide a simple and concise definition of a property right: “a guarantee allowing action within guidelines}

defined by the right” (Hoffman et al., 1994, p. 350). It legitimises the actions of its holder and therefore constrains “punishment strategies which might otherwise be used to insure cooperative behavior” (Hoffman et al., 1994, p. 350).

(14)

9 proposer earned his/her right in comparison to random assignment of roles10_{. This happened without} an increase in rejection rates (these remained low), which reveals that first movers correctly anticipated responders’ willingness to accept lower offers (Hoffman et al., 1994).

Introducing property rights in the UG has also been done in Leliveld et al. (2008). In their paper, unlike that of Hoffman et al. (1994), assignment of property rights was done through placement of endowment chips in different locations of the (bargaining) table separating proposer and recipient. Three types of UG were compared: “giving” (property is initially with the proposer), “splitting” (traditional UG with chips in the middle) and “taking” (initial property belongs to the receiver). The highest offers were observed in the “taking” condition and the lowest in the “giving” one. According to the authors, proposers' high offers resulted from a genuine concern with receiver’s entitlement – called “do-no-harm principle” – and low offers were explained by a sense of entitlement (rather than self-interest), even after controlling for strategic considerations.

Entitlements are crucial in bargaining and greatly influence UG outcomes, in particular. Entitlement perceptions appear nevertheless to guide behaviour in a self-interested manner. The studies of Bediou et al. (2012) and Feng et al. (2013)11_{provide evidence on this self-serving bias. In} these studies, both players contributed to a joint endowment that was later distributed by the proposer12_{. Bediou et al. (2012) concluded that proposers “flexibly” applied justice principles,} depending on their relative contribution, instead of consistently using the same principle. When producing less than their counterpart, proposers used equality (or in other words, kept more than what they had generated). Interestingly, when equity would lead to very low offers, proposers adjusted and tried to anticipate responders “expectations”: responders may not have performed well, but they invested effort (Bediou et al., 2012). In this case, despite the fact that proposers’ performance was

10_{This effect was stronger when the game was framed as an exchange (buyer/seller interaction) instead of a division}

problem.

11_{Both studies used deception in their design.}

(15)

10 better, offers were in between equality and equity (that is to say, proposers kept a bit less than what they had produced). It is uncertain whether behaviour here was fair (as a true concern with responders’ entitlement) or strategic. Offers that respected equity or equality tended to be accepted by responders, regardless of their relative performance. Only offers that clearly disregarded any of the two principles were rejected. Feng et al. (2013) reached similar conclusions. The authors concluded that offers reflected each party’s entitlement, but these concerns were selfish: proposers allocated more to themselves when performing better than their counterpart than they allocated to the counterpart when this player had had a better performance. In fact, proposers valued entitlements more heavily when performing better than their partners. As responders, tolerance for “disadvantageous unequal offers” was higher when performance was worse than that of their counterpart, but “the acceptance rate of advantageous unequal offers for the better-performance condition was higher than that of disadvantageous unequal offers for the worse-performance condition” (Feng et al., 2013, p. 14).

In the last two studies, one party was clearly more deserving than the other, but both players had (unequal) claims to the surplus. Both had produced a joint pie. Carr & Mellizo (2013) also introduce entitlements through production, but on only on one side of the negotiation13_{. Following a} design already used with the DG, Carr & Mellizo (2013) conducted an experiment with a treatment where the responder was the only one producing the pie at stake. In their full UG, offers increased in this case for every pie level (in comparison to the exogenous treatment), but remained below the total amount at stake. Indeed, “unproductive proposers” felt they deserved a share of the endowment only because of their role in the game. Besides, in the endogenous endowment treatment, percent offers increased with endowment size. Rejection rates barely changed, but were slightly lower in the

13_{This was also done in the above-mentioned paper of Hoffman et al. (1994), but not through production. The proposer}

earned the right to split the endowment, but did not produce it. Besides, different from Carr & Mellizo (2013) and Ruffle (1998) – further detailed in the text – Hoffman et al. (1994) increased the perceived merit of the proposer, not the responder.

(16)

11 responder-generated surplus treatment. In the mini UG14_{, when the responder produced the} endowment, rejection rates of low offers were lower (than those of the exogenous treatment) and increased with endowment size15_{: “the larger the endowment the more responders care about} distribution when responders produce the endowment, but not enough to raise rejection rates above the exogenous endowment treatment” (Carr & Mellizo, 2013, p. 8).

Ruffle (1998) also conducted a UG experiment with an endogenous endowment produced by the responder. The results were compared with those of a DG. In the DG, offers were higher when the recipient successfully produced the highest possible amount (“skillful” receivers) in comparison to the exogenous treatment (i.e. “lucky” recipients who got the highest amount after a coin flip). However, recipients who produced only a low pie (“losing” receivers) were offered less than those who got that amount by chance. This “punishment” observed in the low stakes DG was nevertheless a “mild” one – because some “allocators” recognised their partner’s effort as a source of deservingness – and it was even eliminated in the UG version (which shows proposers’ adjustment to the strategic nature of the game). Besides, offers to “skillful” responders (UG) did not differ from those to “skillful” recipients (DG), suggesting sincere fairness concerns, rather than strategic ones (Ruffle, 1998)16_{. The author} further contends that these results are likely to be amplified in real-life contexts, where people and “gifts” are identifiable: in reality, “social pressures, moral imperatives, and the warm glow of giving are likely to magnify the reward to skillful or deserving Recipients” (Ruffle, 1998, p. 259). The author also

14_{A mini UG restricts the offers the proposer can make. Here, proposers could offer 20% or 80% of the pie.} 15_{In the exogenous treatment of the mini UG, rejection rates decreased with endowment size.}

16_{The explanation that Ruffle (1998) provides for fair offers towards “winning” recipients and responders resembles, in}

fact, one of elimination of cognitive dissonance: “tension or distress (…) by causing inequity” is what prevents dictators or proposers from keeping the entire pie (Ruffle, 1998, p. 256). The author suggests that a fair offer is the price to pay to terminate distress and preserve a self-image of a “fair or just” person. Developed by Leon Festinger, the theory of cognitive dissonance from social psychology focuses on the ways individuals reduce or eliminate conflicting and inconsistent motivations (for instance, fairness and self-interest). Kowol (2008) presents a review and summary of this theory.

(17)

12 identifies some real situations resembling the bargaining/distribution scenario created such as principal-agent problems and tipping.

Before proceeding to a brief review on entitlements in the DG, it is important to mention another study with the UG. This study shows the interest of political science in the UG. Categorising it as falling under the “entitlement” line of research is questionable, but it is nevertheless important (particularly given the research question of this thesis on responder behaviour). Smith, Larimer, Littvay, & Hibbing (2007) varied the reasons behind the assignment of proposers’ position in an UG in order to assess acceptance behaviour. In their experiment, all subjects were actually responders and the offer they reacted to was a low one17_{. The authors intended to test whether low offers were more} or less acceptable depending on how the proposer got his or her role: “craved power” (responders were told that the proposer wanted his or her role more than the participant) or “deserved power” (responders were informed that the proposer had earned his or her position). “Trusting” people rejected the (low) offer more often when power was “craved” rather than “earned”18_{. Their results} suggest people care about process (and how power is acquired) and not simply about outcomes.

Entitlements have been studied in the UG, but have received more attention in the DG. In fact, as far as distribution is concerned, the DG – by controlling for strategic motives – provides a better account of fairness preferences, which is why it is important to review the most important papers that introduce entitlement in this game. Cherry et al. (2002) and Oxoby & Spraggon (2008) both varied the source of the endowment in a DG. Cherry et al. (2002) allowed the dictator to produce the wealth of the game. This really affected dictators' behaviour: in stark contrast with the baseline, there was a very high frequency of zero offers (for both high and low stakes). This “self-interested” effect was

17_{The authors used deception.}

18_{In fact, the authors make a distinction between “trusting” and “distrusting” people in their experiment. This distinction}

was made based on questions about trust. Smith et al. (2007) explain that people who tend to trust others are also more sensitive to the characteristics of decision-makers and the way they obtained power (Smith et al., 2007). According to Smith et al. (2007), evolutionary reasons are behind people's concern for procedural justice, their “aversion” to those who desire power and even “differences in sensitivity to leader traits” (Smith et al., 2007, p. 296).

(18)

13 even reinforced under anonymity conditions. Oxoby & Spraggon (2008), building on the work of Cherry et al. (2002), allowed either the dictator or the receiver of the game to generate the wealth. When receivers earned the wealth, dictators respected receivers’ entitlement and made higher offers than in baseline or the dictator-produced wealth treatment (where all dictators made zero offers). This reduction in self-interested offers was greater when the receiver managed to generate a higher level of wealth (when receivers generated the highest amount possible, even offers of 100% were observed)19_{. In fact, there was no significant difference between baseline and “receiver earnings} treatment” when the lowest level of wealth was generated. In other words, receivers who failed to “exert a verifiable level of effort” were not treated any differently than those in baseline (Oxoby & Spraggon, 2008, p. 708). Since the lowest wealth level was given when receivers answered between 0 and 8 questions correctly, dictators were uncertain of whether earnings were due to low effort or low ability (Oxoby & Spraggon, 2008)20_{. Property rights (that come from legitimising the game's wealth)} explain the experiment's results.

Other studies with the DG attempt to investigate how different justice principles are applied. These studies go further than those introducing endowment legitimacy and create more intricate production situations before the distribution phase in order to study fairness. In particular, they try to distinguish the effect of production factors within and beyond one's control in distributive issues (for instance, Konow, 2000; Cappelen, Sørensen, & Tungodden, 2010).

19_{As Carr & Mellizo (2013) point out, it is interesting to observe that even though offers were high when receivers}

produced the wealth (in comparison with traditional DGs), they were lower than the amount the dictators kept for themselves when they were the ones producing the endowment. This suggests that (unproductive) dictators feel they deserve a share of the pie just because of their role as dictators: “we [Carr & Mellizo] interpret their [Oxoby and

Spraggon’s] results as subjects feeling entitled to a claim on the surplus because of their stated role in the game” (Carr &

Mellizo, 2013, p. 4).

20_{The idea that (negative) reciprocity should be based on low effort (not low skill) may have refrained dictators from}

punishing these receivers (Oxoby & Spraggon, 2008, p. 709). As the authors point out, these results differ from Ruffle's (2008). Here, the existence of dictators negatively reciprocating low skill produced a “mild” punishment effect.

(19)

14 In sum, even if not exhaustive, this literature review already reveals that entitlements matter in bargaining and distributive issues. Again, the current thesis builds on previous research about entitlements in the UG by creating a conflict scenario. Both parties have entitlements to what is at stake, but – unlike those of many of the studies mentioned above – these entitlements come from different sources and therefore become difficult to compare. Similar to Hoffman et al. (1994), proposers earn their role in the game and, in line with Carr & Melizzo (2013) and other DGs studies, responders generate the game’s endowment. Also in real life there are bargaining and distribution situations where parties at the negotiating table are both entitled to the bargaining object, and yet their entitlements or property rights are of ambiguous comparison. The entitlements of both players mainly come from either skill or effort. The next section details the experimental design.

3. Experimental design and Methodology

In order to answer the research questions, I ran an experiment with the following three treatments:

Proposer earns his/her role Roles randomly determined Endowment is generated by the responder C - conflict RE - responder entitlement

Endowment is randomly determined PE - proposer entitlement -

In treatment PE, the proposer earns his or her position in a phase that precedes the stage where the UG is played. In this group, there is a “windfall” endowment of 5 or 10 euros, as in standard UG experiments. In treatment RE, roles are randomly determined (proposer does not earn his or her position) and the responder is given the chance to produce an endowment prior to the game stage. Finally, treatment C is the group where the proposer earns the right to be the proposer in the game and then the responder generates an endowment21_{. The standard case (with random roles and an} exogenous endowment) was not run, due to limited access to a larger subject pool. Besides, not only is this the case that has received more attention in UG experiments, the research questions on the

21_{The instructions of every treatment can be found in English in the Appendix. The experiment was conducted with}

(20)

15 intersection of entitlements called for comparisons with cases where only one of the parties is entitled to the pie (i.e. treatment RE and PE). A timeline of the experiment with the steps in every treatment can be found later.

Every treatment was conducted in two sessions, with 10 participants each (5 proposers and 5 responders). At the end of every session, every participant was paid 1 euro of participation fee plus experimental earnings from the game.

Skill task – Matrix task

The skill task chosen for the current experiment to award the role of proposer was a Matrix task, similar to the one used in Ariely, Gneezy, Loewenstein, and Mazar (2009). Participants were given a set of 24 matrices: each one was a 3×3 matrix and had 9 numbers with 2 decimal cases. Participants had to find the only two numbers in each matrix that added up to 10, as in the next example:

Example

6.75 6.33 4.75 5.23 3.25 6.51 0.56 2.53 8.18

This task was used in treatments PE and C, the two treatments where proposers earned their position. In these sessions, all 10 participants in a room started the experiment by taking part in the Matrix task. Subjects were told that they would earn the role of proposer in the UG and got to split 5 or 10 euros if they performed better than the person they had been paired with for the game. They had 3 minutes for this.

This task requires ability and skill rather than effort, which gives a better justification for ranking players in every pair. Besides, one could expect enough variation to rank them. Indeed, participants were told that in the unlikely event that their performance was the same as their partner's, one of them would be randomly chosen to take the proposer’s position. Getting the exact same number of correct matrices was possible, but it was foreseen as an unlikely scenario, as the task has been chosen to

(21)

16 also create variability across subjects22_{. The word “unlikely” was intentionally used in the instructions} and aimed at diminishing any uncertainty that could arise when earning the role of proposer. Unless this word was used, proposers could doubt whether they had truly earned their role or whether they had been randomly chosen.

Effort task – Decoding task

The effort task used for responders to generate an endowment was a Decoding task23_. Responders were given a set of 10 words. Each word was an 8-letter word presented in numbers. A decryption key was provided to convert the numbers into letters and figure out the entire word24_. Responders were given 3 minutes for this task. The words chosen – nouns or (conjugated) verbs – were completely unrelated.

This task required more effort than ability and it was not as “fun” or challenging as the Matrix task. The conversion between correct words and euros went as follows:

Between 0 and 4 correct words 5 euros 5 or more correct words 10 euros

In order to avoid proposers' assessment and judgement of responders' performance, proposers did not see the task25_{and did not know these conversion rates. Only responders knew these rates.} Proposers were only told that responders' correct responses would be converted in either 5 or 10

22_{This was later confirmed in the experiment. Only in 3 pairs (out of 20 in treatments PE and C) was performance the}

same. One member of the pair was then randomly chosen to take the role of proposer.

23_{Both the Matrix task and the Decoding task can be found in the Appendix.} 24_{The decryption key was taken from Charness, Masclet, & Villeval (2010).}

25_{Other papers with a responder-generated endowment differ about this point. Oxoby & Spraggon (2008) provided a copy}

of the production task to dictators and the exact conversion between correct answers and pie size produced. The dictator did not know the exact score of the recipient, but was informed of the monetary amount this player had produced. In Ruffle (1998), first movers did not see the task taken by recipients: “Doing so would have introduced an unobservable variable upon which Allocators may have conditioned their offers, namely, their perceptions about the level of difficulty of the quiz” (p.251).

(22)

17 euros. The idea that higher effort would lead to a higher endowment was nonetheless common knowledge.

In treatment RE, participants knew their role in the game already at the start of the experiment. The experiment then began with responders taking the Decoding task. In treatment C, only after the roles of the game were known (i.e. after the Matrix task) did responders take the Decoding task to determine the endowment at stake.

In treatments RE and C, there was the possibility that responders would decide not to exert any effort at all in the Decoding task. After all, they knew already they would not be the ones splitting the pie. This effect could be even stronger in treatment C, with responders' anticipating proposers' higher sense of entitlement. The possibility of absence of effort in the production phase could be minimised by revealing roles and stakes only after all players had taken part in the Decoding task (treatment RE) or in both tasks (treatment PE). However, this would have been more time-consuming and evidence suggests responders will actually exert effort to produce the pie26_{. Oxoby & Spraggon} (2008) authors identified possible reasons behind responders' choice to make an effort in the endowment generation phase: responders may expect dictators to honour their property entitlements, they may have a preference for efficiency and they may even enjoy the task (Oxoby & Spraggon, 2008, p. 705). The fear that responders would not exert any effort was unfounded (as the experiment later confirmed), which is why a design where only the responders performed the task to generate an endowment was created for both treatments RE and C27_.

Another concern when designing the experiment was that responders in treatments RE and C – particularly C – would suffer from extra tiredness when the time to make a decision came. This

26_{Even in DG experiments with a receiver-generated wealth phase, receivers exerted effort in spite of their role in the}

game (see, for instance, Oxoby & Spraggon, 2008).

27_{Besides, when responders produce a pie to be split by proposers, this production phase becomes similar to a trust}

game. Again, experimental results of the trust game suggest that the first mover sends a positive sum to the second one (rather than the rational zero transfer) as explained in Holt (2006) about the experiments of Berg, Dickhaut and McCabe in 1995.

(23)

18 could affect their behaviour in these two treatments. One possible way to overcome this issue would have been to allow all players in all treatments to take part in both tasks. However, introducing a task in treatments where such task is unnecessary may have undesirable consequences28_{: not only does it} make it hard to justify its existence to participants, but also participants may suspect of deception if asked to do something irrelevant. In treatments RE and C, responders may have been more tired when deciding. However, they had plenty of time to rest while the experimenter counted their number of correct answers in the Decoding task and prepared the acceptance sheets for every responder. Besides, the Decoding task only took 3 minutes and it was relatively easy to answer the minimum number of words required to get 10 euros.

Making an offer and Accepting or Rejecting an offer – Minimum Acceptable Offer and the Strategy Method

Proposers were given a proposals’ sheet and asked to choose one out of 11 possibilities to divide 5 euros and to choose one out of 11 possibilities to divide 10 euros. They had to do this before endowment size was known. They also knew only one of the offers would be actually implemented once the endowment level was determined: again, in treatments RE and C this depended on responder performance in the Decoding task; in treatment PE, stakes were randomly determined (die throw). In all treatments, proposers moved first: in treatments RE and C, proposers were asked to state their offers while the responder was taking the Decoding task; in treatment PE, proposers were asked to state their offers while the experimenter determined the stakes in each pair.

As far as responders are concerned, they were given an acceptance sheet. In this sheet, responders had to state their minimum acceptable offer (MAO) for a given endowment (i.e. the minimum amount below which they would reject an offer). The endowment level presented was randomly chosen for treatment PE. For treatments RE and C, the level presented was the one the

28_{For instance, introducing the Decoding task also in treatment PE where such a task is actually redundant could cause}

(24)

19 responder had managed to produce. Once responders were given the acceptance sheet, proposers’ offers had already been collected (and were not observable to responders).

This method to elicit responder behaviour – known as the MAO method – has been preferred over the specific-offer method29_{. In fact, if one just confronted responders with specific offers, data on} rejection behaviour would be scarce. This method has also been preferred over asking for an accept/reject decision for every possible offer. Two reasons explain this choice. First of all, admitting proposers to state whether they accept or reject every possible offer may give rise to three types of responders: those who reject low offers and accept after a threshold, those who reject very high or low offers and those who randomly change between accept/reject30_{. The risk of facing three different} profiles should be avoided in the current design, especially given the reduced number of observations. The second reason for avoiding a format where responders make a decision for every possible offer regards the potential loss of the intrinsic emotional nature of responders' decision. This may also be the case with the MAO method. However, a MAO represents the minimum one is willing to accept at the negotiation table and below which bargaining is over. For this reason, this eliciting method seems to be more in line with the traditional one-shot accept/reject decision, particularly in the current design, with MAOs being elicited after offers have been collected31_.

Indeed, in spite of the fact that the MAO method was used, the sequential nature of decisions was kept. Asking for offers and MAOs simultaneously can turn the UG into a coordination game. Indeed, as Weber, Camerer & Knez (2004) explain, only when the UG is sequentially played does the

29_{Economists have nevertheless been uneasy about the MAO method, even though there is no clear difference between}

both methods in eliciting responder behaviour, as pointed out by Camerer (2003).

30_{For instance, Bahry & Wilson (2006) identified these three “responder profiles”. In their experiment, most responders}

were either “monotonically rational” (accepting after a threshold) or “hyper fair” (rejecting high or low offers).

31_{See Appendix – Additional explanations on design choices (MAO method) – for an additional reason why the MAO}

(25)

20 game have a single subgame perfect equilibrium. This equilibrium is the only rational benchmark in the UG and its singularity may be compromised if sequence is neglected32_.

Stating a MAO as the second move (even if offers are not observable) was chosen for the current design. It respects the sequence of decisions of the original UG and its rational benchmark equilibrium. In sum, responder’s best strategy after the proposer makes a non-observable offer is to identify a low MAO (0 or 0.5/1 euros). It is questionable whether this method makes responders more demanding than when faced with specific offers – especially considering sequence has been maintained (see Weber et al., 2004 in the Appendix). However, the same method was applied in all treatments, so any differences in behaviour can only be attributed to the treatment in question.

The method used here to elicit proposer and responder behaviour resembles the pure strategy method. The strategy method, as explained in Brandts & Charness (2011), is a method “in which subjects make contingent decisions for all nodes at which they may have to play” (p. 376)33_{. In the} UG, the pure strategy method would require both players to state their decision(s) before the game is actually played (and even before knowing their roles). This is different from the “direct-response method” usually used in UG experiments where subjects observe previous moves and make a decision when it is their time to do so in the game.

The strategy method raises possible concerns of realism. Thinking about hypothetical behaviour can be very different from actual behaviour. In Oxoby & McLeish (2004), proposers’ behaviour did not change between the strategy method and the sequential one34_{. But rejections of} small offers seemed to be lower under the strategy method, which might be explained by emotions35_, as pointed out by Brandts & Charness (2011).

32_{See Appendix – Additional explanations on design choices (Timing of decisions).}

33_{As Brandts & Charness (2011) explain, this method – introduced by Selten in 1967 – requires subjects to first become}

acquainted with the game (through the usual “direct-response method”) and only after that are they asked to state their complete strategies (Brandts & Charness, 2011, p. 376/377).

34_{This is the standard game where proposers made an offer and responders accepted or rejected it.}

(26)

21 The current method to elicit behaviour differed from the pure strategy method in important ways: players knew their role in the game when deciding (and made decisions only under this role), which means they were not asked for a “complete strategy” in advance of knowing their position; decisions were sequential (not simultaneous) and responders did not have to fill in an accept/reject decision for each possible offer.

However, the method used here to elicit behaviour has similar advantages to those of the strategy method. First of all, more information was collected from both proposers (for every proposer an additional offer that was not implemented was collected) and responders (a MAO rather than acceptance or rejection of a specific offer). Second, the current design prompted more serious thinking about choices. Thirdly, proposers had something to do in those treatments where responders produced the endowment. Finally, and most importantly, the sequential nature of the game was kept.

Incentives and Stakes

Even if the use of monetary incentives in economic (and UG) experiments is a matter of debate36_{, the current experiment was incentivised. Participants were rewarded according to the} outcome of the game, once one of the proposer’s offers was matched with the MAO of the corresponding responder. As Croson (2005) clearly explains, “Economic theory makes no predictions of what individuals will say they would do, only what they will actually do when faced with a given decision and the resulting payoffs” (p. 133). The UG represents a bargaining situation over a given monetary endowment, so subjects should also be rewarded in monetary units. They may not even take the experiment seriously, if they are bargaining over something that does not really exist. Real monetary consequences become particularly relevant here, given the somewhat artificial setup already created by an economic experiment.

The question of whether stake levels actually matter in the UG is also not settled. The “high-stakes argument” suggests that by increasing the cost of rejection, “high-stakes strengthen responders’

(27)

22 willingness to accept and therefore, proposers’ offers will move towards the rational equilibrium. However, as Hoffman et al. (1996) explain, rejection may become more salient to the proposer as stakes increase or this player may even display “constant relative risk aversion” (which means the percentage offered will not change).

In fact, the effect of stakes in UG behaviour seems to be “surprisingly weak” (Camerer, 2003). Studies that vary stakes show only a small effect of this variable. For instance, in Hoffman et al. (1996), there were no significant differences in offers between 10 and 100-dollar stakes and rejection rates declined as stakes increased. In the current experiment, stake levels (5 or 10 euros) were not expected to have an effect in proposer or responder behaviour in treatment PE. Previous results indicate that players’ behaviour does not differ between such levels. Nevertheless, stakes could impact both players in treatments RE and C where the endowment is endogenously generated by the responder. This idea is further developed in the hypotheses section.

Finally, experiments should not only be incentivised according to what is being tested, but also remuneration should cover opportunity costs. The opportunity cost is what participants would earn elsewhere had they not taken part in the experiment. In Portugal, where the experiment was conducted, an hourly wage of approximately 5 euros for someone with a university degree at an entry-level position seems to be a widespread practice. In this experiment, participants could potentially earn up to 10 euros and every session lasted around 30 minutes. The average final payoff in this experiment (1 euro participation fee plus additional earnings) was almost 4.60 euros, which exceeded the opportunity cost.

Terminology

Usually, economic experiments provide little context to their participants37_{. Croson (2005) gives} some reasons for that. One of the reasons is that whatever is being tested should be applied to several situations. Another one is the variance that context may introduce, depending on individual

(28)

23 interpretations of such context. However, as also pointed out by the author, “experiments with context have more external validity, cueing subjects to behavior that we might more often observe in the real world” (Croson, 2005, p. 137). If context is not provided, the experiment might create a very artificial setting. As Babcock & Loewenstein (1997) explain, an experiment without context is “illusory”: for instance, in UG experiments, “subtle variations” in this context (such as the way roles are determined) greatly affect behaviour (p. 123).

The experimental instructions made use of meaningful terminology. This intended to provide a context and accentuate differences between treatments.

Post-experiment questionnaires

A different questionnaire for each player type and treatment was designed. Questionnaires can been found in the Appendix. Questions on demographic variables (such as age and gender) were first, followed by questions on the difficulty of tasks (when this made sense for the treatment in question), difficulty of decisions (making an offer, indicating a MAO) and a general question on entitlement (based on Leliveld et al., 2008). In the final section of the questionnaire, participants had to rate their level of agreement with certain sentences about entitlement, fear of rejection/fear of a low offer and other motivations behind decisions. Treatment C questionnaires were practically a merger of PE and RE questionnaires.

Experiment's Timeline

To summarise, the experiment's timeline went as follows in every treatment:

Steps / Treatments PE RE C

All 10 Participants in the room are randomly paired (before the experiment starts). x x x Participants take part in a Matrix task. The best scorer in every pair earns the role of proposer. x x Participants are given a piece of paper with their role (proposer or responder) in the game. x x x Responders take part in a Decoding task that generates the game's endowment (5 or 10 euros).

While responders take part in a Matrix task, proposers note down their offers for two possible endowment sizes (5 or 10 euros) in a PROPOSALS SHEET.

(29)

24

The experimenter determines the endowment size (5 or 10 euros) for every pair by throwing a die. While the experimenter does this, proposers note down their offers for two possible endowment sizes (5 or 10 euros) in a PROPOSALS SHEET.

x

The Decoding task is collected and the number of answers is converted in 5 or 10 euros. x x Once the endowment is determined, responders are given an ACCEPTANCE SHEET to note

down their MAO for the pie size at stake. x x x For every pair, one of proposer's offers is matched with the responder's MAO. While the

experimenter determines individual earnings, participants fill in a questionnaire. x x x All participants in the room are paid according to the final result of the game. x x x

4. Hypotheses

Main hypotheses on proposer behaviour

Hypothesis 1: Proposers in treatment PE are the most selfish ones and proposers in treatment RE are the least selfish. In other words, offers are lowest in treatment PE and highest in RE. The offers observed in treatment C are between those observed in PE and RE: offer RE > offer C > offer PE.

Treatment PE is the treatment with the strongest incentives to behave selfishly. When the proposer earns his or her role, offers are lower as this player feels entitled to most of the endowment (Hoffman et al., 1994). An analogous effect can also be seen when the proposer contributes to the endowment of the game (Bediou et al., 2012; Feng et al., 2013) or when this player is given property rights over it (Leliveld et al., 2008). The results with the DG also corroborate these entitlement ideas (Cherry et al., 2002; Oxoby & Spraggon, 2008). While low offers are expected in treatment PE, zero offers are not anticipated, given the strategic considerations of the UG.

Legitimately earning the role in the game increases opportunities for selfish behaviour, but an endowment produced by the responder calls upon proposers to behave in a less selfish manner and honour their counterpart’s entitlements. Evidence from DG and UG experiment suggests that offers are higher when the responder/receiver produces the endowment in comparison to an exogenous endowment case (Carr & Mellizo, 2013; Ruffle, 1998; Oxoby & Spraggon, 2008). This is also in line with the idea that claims (arising prior to negotiation) generate “moral property rights” that take precedence over “legal” ones in the bargaining process (Gächter & Riedl, 2005). The respect for the other party’s entitlement is likely to be even more pronounced in the UG, where fear of rejection plays

(30)

25 such a vital role. In the current experimental design, treatment RE is the one that maximises responder entitlement to the endowment. Offers equal to the total amount involved are not expected, but may occur (these have been observed in Oxoby & Spraggon's (2008) DG experiment).

In sum, when players have legitimate entitlements, they get a larger share of the endowment in comparison to the “unearned” case. Fairness perceptions seem to depend on the “behaviors leading up (and the characteristics of) a decision environment” (Oxoby & Spraggon, 2008, p. 704).

It is hard to hypothesise on the closeness of treatment C’s offers to those observed in treatments PE and RE. Proposers in treatment C face conflicting motivations: on the one hand, they successfully and legitimately earn their role; on the other hand, the responder is the one generating the endowment at stake. Indeed, their sense of entitlement from winning the Matrix task is probably attenuated by the fact the endowment comes from the effort of their counterpart. Comparing the offers of group C to those observed in treatments PE and RE will allow one to understand which “force” wins the internal conflict proposers face. Given the self-serving manner in which justice principles tend to be applied, one can expect offers in treatment C to be closer to the “selfish” ones observed in treatment PE.

Hypothesis 2: In treatments RE and C, percent offers are higher when the responder manages to produce a 10-euro pie rather than a 5-euro pie. In other words, stakes have an effect in treatments where the amount at stake is endogenously generated by the responder. In treatment PE, stakes do not have an effect.

The proposer is likely to show reciprocity towards a responder that has managed to successfully produce a high pie rather than a low one, especially since he/she knows that two pie sizes are possible depending on performance. This stakes effect – increase in mean percent offers with stakes in responder-produced endowment treatments – has been observed in Carr & Mellizo (2013) and in the DG of Oxoby & Spraggon (2008).

In treatment PE, the only treatment with an exogenous endowment, stakes are not likely to have an effect on percent offers. Evidence on effect of stakes in the UG is mixed, but these play a

(31)

26 minor role with the amounts considered in the current experiment (see Experimental Design and Methodology – Incentives and Stakes).

The stakes effect expected in treatments where the responder generates the endowment may also mean that the difference between offers of RE or C (endogenous endowment) and those of PE (exogenous endowment) will be larger when 10 euros, rather than 5, are at stake. In Oxoby & Spraggon (2008), receivers of a DG who produced the lowest amount possible got the same treatment as those in the baseline scenario (windfall amount) and in Ruffle (1998), “losing” recipients were even “mildly” punished. This punishment effect was nevertheless eliminated in the UG version, which shows “proposers’ ability to adapt to the strategic environment of the ultimatum game” (Ruffle, 1998, p. 254). This suggests that stake levels may not have such a considerable impact on the difference in percent offers between endogenous endowment treatments (RE and C) and the exogenous case (PE) in the current UG experiment.

Main hypotheses on responder behaviour

Hypothesis 3: Responders in treatment PE are the least demanding ones and proposers in treatment RE the most demanding. In other words, MAOs will be lowest in PE and highest in RE. MAOs observed in C will be in between those of PE and RE: MAO RE > MAO C > MAO PE.

Evidence suggests that dictators’ and proposers’ offers are made in accordance to each party’s entitlement to the amount. One can therefore also expect this to be the case for responders’ behaviour. Responders are then likely to be more demanding when they generate the endowment and the less demanding when the proposer earns his/her role.

In the current experiment, responders state their MAO without knowing the offers of their corresponding proposer. This may result in more demanding responders in comparison to the case where these players are faced with a specific offer. The specific offer method has been more widely used in the UG experiments than the MAO method. For instance, in Hoffman et al. (1994), rejection rates did not change from the case where the proposer was randomly chosen to the case where the proposer earned his or her right. In fact, rejections were very low in both cases. The authors argue

(32)

27 that a treatment with proposer entitlement created “common expectations” in every pair: proposers offered less and correctly anticipated responders would accept this.

The results of Feng et al. (2013) suggest that responder behaviour is likely to also depend on relative performance. Responders showed greater tolerance towards “disadvantageous unequal offers” when they performed worse than their partner in the resource production phase that preceded the game. This result points to the observation of the lowest MAO in treatment PE, where the responder falls behind in the Matrix task. Besides, there is evidence that responders may tolerate a low offer coming from a proposer that earns his/her role (Smith et al., 2007). Treatment RE, on the other hand, is the treatment where the responder has the strongest claims to the pie at stake. This means the MAO observed will be the highest in this treatment38_.

The MAO in treatment C is expected to fall in between that of treatment PE and RE. In treatment C, responders generate the endowment, but their sense of entitlement is likely to be diminished by the fact that they lost against the proposer in the Matrix task. It is however quite hard to hypothesise on responder behaviour, given the scarcity of studies that report on responder behaviour in an UG where this player has (completely or partially) produced the endowment.

Finally, the relationship between MAOs and stakes is likely to differ between treatments where endowment is exogenous and those where the responder produced it (see Carr & Mellizo, 2013, in Literature review). However, in the current experiment, the number of MAOs collected for the low-stake level (5 euros) in responder-generated endowment treatments is likely to be very low. This means that the experiment will not capture any stakes effect for responders’ behaviour.

38_{Evidence from Carr & Mellizo (2013) tells something different. In their full UG, rejection rates were very similar between}

exogenous and endogenous endowment treatments, but slightly lower in the latter case. In their mini UG, rejection rates were considerably lower when the responder produced the pie. The authors write the following: “Responders are less willing to destroy the surplus if they produce it themselves” (Carr & Mellizo, 2013, p.3). Again, responders in their experiments were confronted with a specific offer rather than asked for a MAO. In the current experiment, responders are asked for a MAO without knowing the actual offers.

(33)

28

5. Analysis of results

The experiment was conducted between April 28th_{2015 and May 7}th_{2015 with 60 students of} four different tutorial groups of the Faculty of Economics of the University of Coimbra (Portugal)39_{. All} participants but three were students of Economics. The average age of participants was 21.9 years old40_.

There were six experimental sessions. Each lasted approximately 30 minutes. Every session was conducted in a classroom41_{and instructions were read out loud so as to ensure common} knowledge. All pairs in every session were in the same room42_{. Every subject took part in the} experiment only once.

Proposer behaviour

The minimum and maximum offer per treatment and endowment size as well as mean offers (and standard deviations) can be found in table 1:

39_{This Faculty – in my hometown – is where I completed my Bachelor studies in Economics in 2013. I approached former}

professors who welcomed the idea of sparing the final or beginning part of a class to let their students participate. Students of this faculty are generally unfamiliar with experimental approaches in economics, a reason why this experiment was also thought of a good opportunity for students to contact with a different way of doing research in economics. Before the sessions took place, I entered the classroom, introduced myself and asked for the participation of students in an experiment. In this introduction, participants were told that they would be given a set of instructions that would be read out loud and they would make decisions. Besides, they were told that they would earn a 1-euro participation fee and additional earnings up to 10 euros.

40_{The students stated their age and gender in the post-experiment questionnaire distributed at the end of the experiment}

while earnings were calculated. There were 27 male participants, 23 female participants and 10 did not answer the gender question.

41_{Students were placed in two parallel rows. They could see everybody in the room. Since they were close to each other,}

verbal communication was possible, but rarely happened.

42_{Having subjects in the same room reduces doubt about the true existence of a counterpart (see, for instance, Frohlich,}

Oppenheimer, & Kurki, 2004, p. 94 and 110). This doubt was probably even eliminated here, because participants of every session could see each other and notice different player types in the room (for instance, in treatment RE or C sessions, proposers could see there were responders taking the production task).

Legitimate and ambiguous entitlements in the ultimatum game

THESIS