On the Limits of Self-serving Behavior
- Inequality, Attention, and the Self-serving Bias
Master’s Thesis
MSc Business Economics - Track Neuroeconomics
Supervisor: Dianna Amasino
Rafaela Antunes Pinto
August 2021
Statement of Originality
This document is written by Student Rafaela Antunes Pinto who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
UvA Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Acknowledgements
“Nobody said it was easy”1. Especially during these difficult times. Therefore, I wish to show my gratitude to those who helped me along the way:
To the University of Amsterdam and its Professors, for the honor of studying in this admirable institution;
To Dianna Amasino, for all the advice and inspiration;
To my friends, for being good friends;
To El Curador, for the joy of meeting them;
To Lucas, Max, Rina, Tobi, and Vi, for sharing and caring. Strawberry fields forever2;
To Gonçalo, for the fellowship and laughter in every adventure;
À minha irmã, mãe, e pai, por serem casa.
1 Coldplay, Nobody Said it was easy, 2002.
2 The Beatles, Strawberry Fields Forever, 1967.
Abstract
Income inequality between rich and poor has widened in recent decades. Privileged people tend to explain their success on merit, neglecting the role that luck can play in their lives, keeping more to themselves and giving more to their wealthy peers. To examine the dichotomy of merit and luck as a justification for the self-serving bias and its consequences for redistributive preferences, an experiment involving a dictator game with previous production is conducted.
Dictators were randomly assigned a relatively advantaged or disadvantaged position by being paid more or less than their receivers and allowing luck to play a role in their contribution to the surplus. Explicit luck information on payment rates was provided to dictators, along with merit and outcome information, before they made division decisions. The extent to which people used their relative luck to justify self-serving divisions and inequality was investigated.
As the level of inequality in payment rates widened, relative attention to merit and luck information tended to increase. Importantly, it was found that an increase in relative attention to merit and luck information led to a reduction of self-serving divisions. These findings suggest that attention is a viable channel to mitigate the self-serving bias, reduce the polarization of political ideologies and potentially tackle the rise of populisms.
Keywords: fairness, redistribution, inequality, self-serving bias, merit, luck, attention, mouse- tracking.
Table of Contents
1. Introduction ... 6
2. Literature Review ... 8
3. Experimental Design ... 15
3.1. Surplus Production Phases ... 16
3.2. Allocation Decision ... 16
3.3. Dictator’s Status and Multiplier Treatments ... 17
3.3.1. Dictator’s Status ... 18
3.3.2. Multiplier ... 18
3.4. Fairness and Demographics Surveys ... 20
4. Hypotheses ... 20
5. Results ... 23
5.1. Status and Attention ... 24
5.2. Inequality and Attention ... 27
5.3. Attention and Allocations ... 31
5.4. Inequality and Allocations ... 36
5.5. Limitations of the Results ... 38
6. Discussion... 38
6.1. Information Avoidance and the Self-serving Bias ... 38
6.2. Gender Differences in Preferences for Redistribution ... 40
6.3. Attention to Self vs. Other’s Information and Allocation Behavior ... 41
6.4. Limitations ... 42
6.5. Future Steps ... 43
6.6. Main Findings and Conclusions... 44
7. Conclusion ... 47
8. Appendices ... 49
8.1. Additional Results ... 49
8.2. Experiment Instructions, Comprehension Questions and Surveys ... 53
9. References ... 63
“Many of the factors that separate winners from losers are arbitrary from a moral point of view.”
Michael Sandel, The Tyranny of Merit (2020, p. 51)
1. Introduction
Free market economies often lead to an unequal distribution of resources. On that account, economic and social inequalities are common in our daily lives. In fact, from the decade of 1970, the middle and lower classes saw their income slowly being reduced, while the income of the upper elites skyrocketed, resulting in a wider income gap (Stone et al., 2020). However, people generally perceive inequality differently than what the most commonly used economic axioms predict (Jancewicz, 2016). The rise of perceived inequality – not its actual level – is inseparably linked to the rise of conflict between the wealthy and the poor (Gimpelson &
Treisman, 2017). So, there is sometimes a dichotomy between what is just from a moral standpoint and what is democratically just, which leaves room for societal disagreements about how redistributive systems should be designed.
Perhaps due to our need to constantly compare ourselves to our peers, human beings seek for justifications for inequalities, and one’s socioeconomic status may affect those judgements. Affluent Americans tend to attribute their wealth to dispositional factors, such as intelligence and hard work, rather than situational factors, such as family and luck (Suhay et al., 2021). Accordingly, privileged individuals typically explain their success on work effort, not acknowledging the role of luck in their success, when compared to non-privileged subjects (Becker, 2013; Amasino et al., 2021). As a result, wealthy individuals tend to have a relatively weaker desire for redistribution, keeping more for themselves and giving more to their wealthy peers (Deffains et al., 2016). This shift toward a more selfish behavior based on one’s socioeconomic position is the self-serving bias. Therefore, it is important to investigate how this bias can be mitigated, to reduce societal disagreement and political polarization.
The dictator game with previous production is a useful experimental game to study the phenomenon of the self-serving bias seen outside the laboratory. In this game, two participants complete tasks to generate a reward, which one of them, the dictator, then chooses how to divide between the two. There is some consistent evidence on how people divide monetary resources when their contribution depends on various factors, such as merit and luck. People are more likely to allocate according to merit than luck (Cappelen et al., 2010; Becker, 2013;
Franco-Watkins et al., 2013; Lefgren et al., 2016). Indeed, there is neural evidence suggesting that people show inequity averse preferences, therefore disliking divisions that are not
proportional to one’s effort and tolerating inequalities derived from merit (Cappelen et al., 2014). To some extent, this could be due to people’s intrinsic selfishness to, or the desire to do better than others. On the other hand, there are also social norms against extreme inequality.
Accordingly, individuals usually show inequality aversion preferences in brain responses, as they disapprove of actions that increase inequality between the involved parties to some extent (Tricomi et al., 2010). There does not seem to be a clear neuroeconomic pattern for how people prefer to redistribute resources. However, people’s fairness judgments seem to ultimately depend on how luck benefits them. Indeed, people tend to change their behavior depending on the context, in order to protect their own interests (Rodriguez-Lara & Moreno-Garrido, 2012).
Therefore, one might wonder where the limits of self-serving behavior lie when luck plays a role in people’s success.
One possible channel through which the sources of the self-serving bias and differences in attitudes toward fairness can be studied is attention. Human attention is selective, meaning that we focus on certain information while ignoring other information. That enables the understanding of which source(s) of information we value most when asked to explain an inequality. Ideological beliefs, for instance, play a prominent role in shaping our attention to inequalities arounds us (Easterbrook, 2021). Individuals who adhere to egalitarian views are not only more inherently prone to, but also quicker, to notice inequality changes (Waldfogel et al., 2021). Moreover, when poverty is justified on situational causes, such as societal unfairness, as opposed to internal factors, people are more inclined to take action to reduce inequality (Piff et al., 2020). For these reasons, it seems to be of great importance to provide citizens with transparent and complete information on economic and social inequalities, so that the gap between perceived and actual inequality is lessened, as well as discrimination against the unfortunate.
It may be easy to reckon when inequality is too high, but it is certainly hard to decide on the level of inequality that might be motivating according to meritocratic standards.
Nonetheless, extreme inequality in general threatens social stability to an extent that perhaps no justifications can resolve. There is evidence for the self-serving bias when the advantaged participant’s payment rate is threefold that of the disadvantaged (Konow, 2000; Rodriguez- Lara & Moreno-Garrido, 2012; Amasino et al., 2021). Yet, the question that arises is whether people’s self-serving allocation behavior is equally influenced by different levels of inequality in payment rates, or, on the other hand, if people behave more altruistically when the inequality gap widens. Thus, this study aims to explore how the self-serving bias is influenced by different levels of inequality in payment rates, through attentional and allocative behavior patterns and
using a dictator game with previous production. Inspired by Amasino et al. (2021), and originally by Konow (2000), the design of this study incorporates explicit luck information available to the dictators before the division decision, along with merit and outcome information. This additional feature impacted dictators’ visual attention and allocation behavior, as it specifically illustrates the level of inequality in payment rates between the dictator and the receiver, which is due purely to random luck. Furthermore, it adds to previous literature in the sense that it assesses differences in how people engage with different information sources to explain their allocation decisions, not only by their status, but also across increasing inequality. Shortly, it questions the extent to which inequality can be justified and deserved.
At the attentional level, it was found that those who were randomly assigned an advantaged position by getting paid more than their pair spent relatively more time looking at merit information when inequality was extreme. In addition, disadvantaged subjects looked more at luck information when inequality increased. Yet, the most significant findings were that, when advantaged dictators looked relatively more at merit, possibly with the goal of increasingly following meritocratic norms, their allocations to their disadvantaged receivers increased. The same pattern is observed for an increase in relative time spent looking at luck information. Conversely, as disadvantaged dictators looked relatively more at merit or luck information, their allocations to their advantaged receivers decreased. These results corroborate recent others arguing that there are differences in attention between people with different socioeconomic status that motivate their differences in division preferences (Amasino et al., 2021), and extend them by suggesting that higher attention to certain sources of information might reduce self-serving divisions.
This thesis is organized as follows. Section 2 includes a survey of relevant literature.
Section 3 contains the experimental design. In Section 4, the hypotheses are outlined. The results are presented in Section 5. Section 6 consists of the discussion and Section 7 concludes the thesis.
2. Literature Review
This study falls under the literature on the role of merit and luck in redistributive fairness. Many studies use the dictator game to investigate how people make allocation decisions when there was no previous production. Standard economic theory would predict that dictators keep everything to themselves in that case, given their willingness to maximize their utility and the receiver’s absence of decisional power. There are studies showing that, usually, dictators share
earnings with their receivers (Camerer & Fehr, 2002, Forsythe et al., 1994, Cappelen et al., 2010). Nevertheless, there are also dictators who keep everything to themselves in those situations, possibly due to an endowment effect derived from getting their contribution to the surplus by pure luck. Notwithstanding the contribution of those studies for the understanding of fairness perceptions, it seems that a dictator game with previous production might exemplify more accurately in the laboratory how most redistributive decisions are to be made in reality, given the complexity of economic and social systems, and the fact that most people work for their money rather than receiving windfalls. In a dictator game with production, merit is introduced as a determinant to each player’s contribution to the surplus to be divided. In such circumstances, the dictators’ allocations will depend on their perception of which of those factors can effectively justify keeping more for themselves. In these cases, it is much rarer to observe dictators keeping the whole surplus to themselves than in dictator games without production. There is more sense of obligation to share when merit is introduced in the equation.
Fairness perceptions under inequality involving merit and luck are, therefore, the focus of this study and literature review.
Fairness views
Examining fairness views might provide a clearer answer on how people make monetary division decisions. The Egalitarian view claims that the surplus should be equally split between the involved parties, therefore it argues for a high level of redistribution and progressive taxes. The Meritocratic perspective states that the surplus should be divided proportionally to each party’s work effort, or, as it will be called from now onwards, merit, so inequalities based on luck should be compensated for. The Libertarian view asserts that the surplus should be split proportionally to each party’s monetary contribution, regardless of whether it is due to effort or external sources like luck, so it claims that there should be a lower level of redistribution and low income taxes. There is, however, converging evidence affirming a high level of heterogeneity in fairness views (Cappelen et al., 2007; Ubeda, 2014; Cappelen
& Tungodden, 2013). Moreover, individuals sometimes employ different fairness ideals depending on the context, with the aim of self-serving their interests or maximizing their payoffs (Rodriguez-Lara & Moreno-Garrido, 2012; Ubeda, 2014). For instance, Liebig et al.
(2015) show that if there is a cooperative environment, people will more likely be concerned about equality, whereas in an incentivized competition setting, they will opt for inequality more often. There are also spillover effects in allocation behavior when people make similar
decisions being involved, compared to when their decisions are impartial (Dengler-Roscher et al., 2018). Thus, society cannot speak with one voice regarding fairness views.
Merit versus Luck
Several studies show that varying the sources of inequality, such as merit and luck, has an influence on fairness perceptions and redistribution preferences. There is converging evidence suggesting that people are more likely to divide according to merit than luck when there is production. Effectively, when there is no previous effort, the tendency is to follow egalitarian fairness norms; when effort is exerted, people allocate monetary resources according to one’s previous work effort, or merit (Franco-Watkins et al., 2013). So, people generally hold each other accountable for aspects that lay within their control, only compensating others for factors beyond their personal control (Cappelen et al., 2010; Becker, 2013), which suggests that the most prevalent notion of fairness is meritocracy.
Recently, efforts have been made to investigate brain reactions to (un)fair situations, such as in the dictator game with previous production, as a way of understanding people’s redistribution willingness and its underlying motivations. Although its neural correlates are not yet known, it is believed that brain responses to inequality situations vary according to what was in the origin of that inequality, such as merit or luck (Vostroknutov et al., 2012). There is behavioral evidence, even if there are individual and contextual differences, answering how individuals choose to allocate monetary resources when those different factors produce an inequality. In line with the inequality aversion model, formulated by Fehr and Schmidt (1999), there is both behavioral (Korenok et al., 2012) and neural (Tricomi et al., 2010) evidence suggesting that dictators show inequality averse preferences, therefore disliking divisions that produce inequalities. Tricomi et al. (2010) demonstrated that brain areas highly involved in the valuation of social and non-social rewards, such as the ventral striatum and the ventromedial prefrontal cortex, showed more activation when monetary transfers that closed the inequality gap were made, both in advantageous inequality, when the subject is better-off than the other, and, strikingly, disadvantageous inequality, when the subject is worse-off than the other. Even though the subjects’ payoff depended solely on luck, as there was no previous surplus production, their findings suggest that there is a tendency to adopt inequality averse preferences when no effort was exerted.
Nevertheless, there is also wide approval of the equity theory or, as Homans (1961) called it in the early 1960s, “distributive justice” in literature. This theory suggests that inequalities based on merit should be compensated for. In their pioneer neuroimaging study,
Cappelen et al. (2014) prescanned participants as they earned money by working and, then, neuronally scanned them they made judgements about different possible distributions of their income. They found that people’s hemodynamic response in the ventral striatum, highly involved in the valuation of social outcomes, increases as they are shown distributions of income that increasingly deviate from distributions proportional to their merit, in accordance with the equity theory. Therefore, individuals consider dividing according to rules different from meritocratic ones a norm infraction, so they recognize a difference between fair inequalities, caused by merit differences, and unfair inequalities, possibly caused by luck differences, an idea that previous findings corroborate (Fliessbach et al., 2012). Some findings advocate that both advantaged and disadvantaged people deliberately vote against their self- interests when that signifies mitigating the gap between the others’ merit and their reward, showing their willingness to follow meritocratic norms (Lefgren et al., 2016). In a setting where subjects were involved in the production of the outcome, but also got an externally determined rate of return, the majority of participants took into account how much their pair had invested when distributing resources, which cannot be explained by inequality aversion (Cappelen et al., 2007). Besides, when the underlying causes of inequality relocate to external factors beyond individuals’ control, such as lucky events, then people are more prone to favor a higher level of redistribution (Wiwad et al., 2020). Given that most people are more willing to compensate each other for external factors rather than internal ones (Buser et al., 2020), it can inferred that merit is considered by most people the most plausible reason to explain a disparity.
The Self-serving Bias
Socioeconomic status can play a large role in determining one’s attitudes towards redistribution. Deffains et al. (2016) created a setting in which subjects played a real effort task and were then divided in two groups, successful and unsuccessful. Success was, however, in large part exogenously determined, as subjects were randomly assigned to either an easy or a hard task. Thus, there was some degree of ambiguity about whether success was due to internal or external factors. Their results suggest that successful people tend to attribute their success to internal factors, such as work effort, not acknowledging that external factors, such as luck, can also play a role in their success. Such observations are in line with the ones from a novel survey conducted to understand extremely wealthy Americans’ perceptions on the reasons behind their relative success (Suhay et al., 2021). Consequently, societal winners show a preference for less redistribution, keeping more to themselves and giving more to their affluent peers than to the societal underdogs (Deffains et al., 2016). The self-serving bias is the
phenomenon behind such deviation in allocation behavior caused by one’s social status, even when status is randomly assigned. In fact, results suggest that the neural response to unfairness is lower for disadvantaged subjects, perhaps because they are more used to getting an unfair treatment by society, which clarifies why they are more prone to accept unfair offers than advantaged people (Hu et al., 2014).
Nonetheless, it is important to point that, even if there is converging evidence for a general agreement that meritocratic norms should rule, differences in attention and ambiguity can lead to different outcomes. The line that separates work effort and luck is clearly outlined to participants in this thesis, contrarily to the ambiguity in Deffains et al. (2016). So, instead of creating space for a possible “moral wiggle room”, we can learn about people’s motivations underlying their willingness to redistribute and figure out whether people truly believe in different fairness norms, or if they seek out different information that they use to decide according to the same, perhaps meritocratic, norm. Crucially, it is also possible to investigate what happens when they are confronted with their privilege. That includes their possible preference for information avoidance to allow for keeping higher shares, in a setting in which information is provided in a complete way.
Some studies point out the idea that, ultimately, self-serving biases are choices that support social norms which are accepted as reasonable (Bicchieri & Mercier, 2013). The possibility that many people view meritocratic norms as just might rationalize why successful individuals believe they indeed deserve their fate, as it is socially justified by merit. Notably, Grimalda et al. (2016) show that, only by knowing that there is a chance of starting off in a more advantaged position than the other, people are instantly willing to approve a higher inequality gap. Thus, there might be space to investigate whether allocation biases influence one’s degree of inequality acceptance.
All the findings mentioned above seem to be in line with the egocentric bias found by Konow (2000), who demonstrated that privileged dictators, whose advantage was randomly determined, tend to keep more to themselves than their effort-based earnings in proportion to the total surplus, even if their own earnings are not affected by the decision.
Attention to Inequality
Whether one benefits from inequality plays an important role in determining individuals’ redistribution willingness. Perhaps not surprisingly, humans start to develop their conviction that social inequalities are due to merit differences in adolescence (Almas et al., 2010). Therefore, citizens tend to believe that the merit gap widens as the societal level of
inequality increases (Heiserman & Simpson, 2017). Nevertheless, inequality perception may depend on many aspects, such as the self-interested willingness to support one’s own political ideologies (Easterbrook, 2021).
One’s fairness ideal may also anticipate social values towards low-status and high- status groups. Some argue that egalitarians are usually more empathetic towards disadvantaged people, yet less empathetic than people with other fairness views towards advantaged, maybe because they look up for them (Lucas & Kteily, 2018). Hence, egalitarians are instinctively more prone to notice inequality differences in their environment, and to notice it faster (Waldfogel et al., 2021). So, attention to inequality is intrinsically molded by ideology.
However, when the underlying causes of inequality relocate to external factors beyond individuals’ control, such as luck, then people are more prone to follow egalitarian norms and accept more redistribution (Wiwad et al., 2020). That corroborates the idea that meritocracy is the guiding norm for most people. If the perception of wage inequality reaches high levels, people tend to discredit merit as being the reason behind wage differences (Kuhn, 2019). The majority of people does not perceive the level of inequality surrounding them as the most popular indicators predict (Jancewicz, 2016; Gimpelson & Treisman, 2018). In spite of one’s perception of inequality, it is also interesting to analyze the perceived fairness of inequalities.
A recent study that links Danish people’s survey data to some aspects of their life suggests that the fairness perception of inequalities is heavily associated with current social position, moves with shocks to that position, and, strikingly, changes when participants of an experiment are told about their positions (Stantcheva et al., 2020). In spite of the importance of investigating inequality perception and its perceived fairness, it remains an open discussion how governments can provide their citizens with transparent information about wage inequality, for instance, so that the gap between inequality perception and real inequality is lessened.
Human’s attention is selective and motivated. Only recently has the role of attention been explored by scholars, with the goal of examining information-seeking patterns to payoffs in economic decisions. Attention can be studied through eye tracking tools, quantifying eye attention to information, and mouselab WEB, which measures the dwell time that people spend hovering their mousepads over certain types of information so that they can reveal them. Other measures of attention are sometimes relevant to assess, such as the proportion of overall time looking at a certain type of information, or the number of transitions the individual did between different sources of information. These attentional measures are precious in telling researchers important underlying processes of economic decision making. Fiedler et al. (2013) demonstrate that the visual searching and processing of information by individuals differs in accordance
with their social value orientation, as people showing more pro-social orientations seek out information about others relatively more. Jiang et al. (2016) also reveal that eye movement patterns coincide with choices towards one’s social preferences. Furthermore, later results suggest that those interactions are also visible under the use of mouse-tracking techniques, such as mouselabWEB (Bieleke et al., 2020). Remarkably, manipulating attention by limiting its availability time leads to an expressive change in visual attention and in choices (Ghaffari &
Fiedler, 2018). Therefore, the study of attention can lead us precious insights on the underlying processes of economic decision-making.
On the other hand, Amasino et al. (2021) is, presumably, the first study contributing to the understanding of people’s attention to the determinants of economic production. Through mouselabWEB, they found significant differences in attention patterns between people who are randomly assigned with an advantage or disadvantaged in the form of payment rate per correct answer, as well as differences in their allocation behavior. Particularly, advantaged subjects pay relatively less attention to merit information than disadvantaged ones, which motivates them to keep more of the surplus to themselves. Besides, when participants are forced to look relatively more at merit information, there is a significant reduction in the effect of being advantaged on allocations. These findings substantiate the previously found self-serving bias, but they go a step further by exposing a correlation between attention and allocation behavior.
Inspired by Amasino et al. (2021), this study aims to expand insights on the role of attention to information relative to merit and luck. By explicitly including pure luck information on randomly determined payment rates, in addition to pure merit information on the participants’ work effort and combined outcome information on each one’s monetary contribution to the surplus available to the dictators, complete and transparent information is provided. It addresses the extent to which the self-serving divisions are observed under different levels of inequality in payment rates, and seeks for attentional patterns that justify self-serving decisions under inequality. The fact that dictators are presented with several possible scenarios not only allows for an analysis on attention and allocation behavior adopted in each situation, but also an analysis on how those differ between participants randomly assigned an advantaged or a disadvantaged economic status. Do those who are advantaged feel guilty if inequality increases due to their luck, or do they avoid learning about the role that luck plays in their success and simply take advantage of the inequality to keep more to themselves?
How, on the other hand, do those who are disadvantaged respond to that?
Ultimately, this study contributes to the understanding of how channels, such as attention and information presentation, might help in the mitigation of the self-serving bias. By having people agree more on redistributive policies, the polarization of political ideologies and conflict between social classes could be lessened.
3. Experimental Design
The experimental design of this study is outlined here.
Participants
The experiment was conducted online, using MouselabWEB. Data were collected between the 4th and the 12th of June3. A total of 115 individuals participated in the experiment.
Six participants were excluded from the study, because they answered two or more comprehension questions incorrectly, out of eight. Thirteen other participants were excluded, because they either reloaded pages or returned to previous pages during the experiment, contrary to what was asked in the instructions, resulting in more or fewer allocation decisions than the 9 rounds. Therefore, data from 96 participants were analyzed. Participants were informed that the study had previously been submitted to an Ethics Approval, that their participation was anonymous, and that their data would be matched with their email, if they wished to provide it to receive their payout. They could only begin the experiment if they had given their informed consent to participate in the experiment.
The experiment lasted 20 minutes, and it consisted of three main parts. The first part included three surplus production phases. The second part involved nine allocation decisions (dictator game). The third part included two surveys, one asking about fairness views and the other collecting demographics. The instructions, comprehension questions and surveys of the experiment are provided in Appendix B.
Before the first and second parts, the instructions for each respective part were displayed. All participants were assigned the role of dictator, except for two, who were assigned the role of receiver, mostly for incentivization purposes. All subjects participated in all three parts of the experiment, except for the participants assigned the role of receivers, who completed only the first part (the three surplus production phases) and the demographics
3 Answers to the study after the 12th of June were not included in the analyzes, but the experiment can still be consulted at this link: http://thesisrap.com/thesis_exp/instructions1.php
survey. Four participants - the receivers and two randomly selected dictators - were paired and received their payoff at the end of the experiment. The payoff consisted of the monetary amount that was allocated to the respective participant, either the dictator or the receiver, in one of the allocation decisions. Therefore, all dictators were informed that they would be making decisions for a variety of possible scenarios that could occur and that, if they were selected for the payoff, one of their decisions would be played out with another actual participant who most closely matched the task performance in that decision round. Thus, they were advised to treat each allocation decision as if it would be selected for payoff.
Figure 1: Overview of the Experimental Design.
3.1. Surplus production phases
Subjects played three production phases in succession, with an interval page in between each page containing information about their performance. The task was the same for all three production phases, and it consisted of moving sliders. Subjects were asked to move as many sliders as possible to the requested value during a 2 minute period.
3.2. Allocation decision
The production phases were followed by nine allocation decisions. Subjects were asked to make a division decision by choosing how to split the joint monetary surplus between themselves and a receiver. Subjects could explicitly observe how much they would keep for themselves and give to the receiver, by selecting the value they wanted to keep to themselves on a slider, that ranged from zero to the value of the joint surplus. The joint surplus to be divided by the dictator in each round was generated as explained in 3.3.
3.3. Dictator Status and Multiplier Treatments
Before each allocation decision, a screen was displayed with information about how the subject and the receiver contributed to the common account in that round. Three types of information were displayed, with respect to both the dictator (“You”) and the receiver (“Other”): the number of correct answers, the payment rate, and the monetary contribution, as shown in Figure 2. The number of correct answers is pure merit information, the payment rate is pure luck information, and the monetary contribution is a combination of both, consisting of outcome information. Therefore, looking at monetary contribution obscures the input of merit and luck, while looking at correct answers and payment rates highlights the meritocratic and random elements of the contribution, respectively, that participants may wish to distinguish.
Figure 2: Information Screen.
The information was hidden in boxes. Subjects could reveal the information in each box, by hovering their mouse over it. When they moved the mouse away from the box, the box closed again, so only one box could be open at a time. This information screen was available for 10 seconds. Within this time limit, they could decide which and how many boxes to open.
Boxes could be opened more than once. At the top of the page there was a countdown timer, and participants were automatically redirected to the next page when the time expired. There were 12 possible and randomly assigned orientations of the boxes, corresponding to the random arrangement of participant information (You/ Other) in its rows and contribution information (Correct answers/ Pay rate/ Monetary contribution) in its columns. To allow for the development of information-seeking patterns, each subject saw the same orientation throughout the whole experiment, including in the instructions and the practice round4.
4 Subjects in each box orientation (orientation – number of subjects): 1 – 6; 2 – 9; 3 – 5; 4 – 5; 5 – 11; 6 – 8; 7 – 7; 8 – 8; 9 – 10; 10 – 9; 11 – 10; 12 – 9.
Any dwell times shorter than 200 milliseconds were disregarded, as that is considered the minimum amount of time needed to process any kind of information in previous literature (Willemsen & Johnson, 2019; Pachur et al., 2018).
For each of the nine rounds, the number of correct answers (merit information) was randomly selected from one of the three production phases that the subject had previously played, to ensure that subjects could not predict which values would be presented and to avoid learning effects throughout rounds. In this way, subjects had to hover their mouse over that information box in each round if they were interested in that information. The receiver’s number of correct answers varied between three possibilities: approximately 20% less than the dictator’s number of correct answers (0.8*P), the same number of correct answers as the dictator (P), or approximately 20% more than the dictator’s number of correct answers (1.2*P).
This number is an approximation, because it was rounded to the closest integer.
3.3.1. Dictator’s status
Each dictator was randomly assigned one of two possible statuses, ‘advantaged’ or
‘disadvantaged’. In the ‘disadvantaged’ status, the dictator was always paid less per correct answer than the receiver. In the ‘advantaged’ status, the dictator was always paid more per correct answer than the receiver. The dictator’s status remained the same for each subject throughout the whole experiment (between-subject)5. The reason behind this is the importance of experimentally simulate the dictator’s status in this study and keep it constant. Subjects were informed of their status, i.e., whether they were getting paid more or less per correct answer than their receiver, after playing the production tasks to ensure that there were no differences in effort in the tasks due to their dictator status. The receiver and the information relative to the receiver were hypothetical, as they represented different possible scenarios for a strategy method approach. Only if the participant was selected to receive his/her payoff, was the payoff for the selected round based on real information and impacted a real receiver. To avoid deception, participants were clearly informed about this in the instructions.
3.3.2. Multiplier
As for the payment rate (luck information), subjects were assigned a multiplier in each round that determined how much they were paid per correct answer, relative to the recipient.
The multiplier could be 3, 5, or 7, for both ‘advantaged’ and ‘disadvantaged’ (within-subject).
5 There were 44 ‘advantaged’ subjects and 52 ‘disadvantaged’ subjects.
This means that, in each round, dictators were paid 3, 5, or 7 times more than the recipient (in case of ‘advantaged’), or 3, 5, or 7 times less than the recipient (in case of ‘disadvantaged’).
Thus, there was always an inequality between the dictator’s payment and that of the recipient.
The dictator’s payment rate and the receiver’s payment rate always added up to a total of 80 cents in each round, so that the average total payment was kept constant. The payment rates were 0.20€/0.60€ in multiplier 3, 0.13€/0.67€ in multiplier 5, and 0.10€/0.70€ in multiplier 7.
When the multiplier was increased, the total pie was kept constant and only the pure luck of each participant was changed. So, behavioral changes could only be due to changes in payment rate luck. In this study, multiplier 3 is considered the control condition, as previous findings suggest that the self-serving bias occurs at this level of inequality in payment rates (Amasino et al., 2021), and it is the goal of this study to examine the extent to which self-serving behavior is observed at higher levels of inequality.
Both the multiplier level and the number of correct answers could change between rounds, but the status of the dictator was always the same. Each of the three multipliers (3, 5, or 7) was combined with each of the three possible levels for the recipient’s number of correct answers, resulting in nine different allocation decisions for each participant. These combinations of multipliers and receiver production were randomly ordered for each participant.
The monetary contribution (outcome information) corresponded to the number of correct answers multiplied by the payment rate, for each round. It represented each players’
monetary contribution to the joint surplus to be divided by the dictator.
Table 1: Conditions per subject.
3.4. Fairness and demographics surveys
Participants were then asked to complete a fairness survey to assess their fairness views.
They were asked to rate how morally acceptable it is to divide the joint surplus according to the egalitarian (equal split), the meritocratic (division according to effort), and the libertarian (division according to monetary contribution) views, as well as to keep it all to oneself and to divide it according to the payment rates (pure luck). They were also asked to what extent information on payment rates (luck information) influenced their decisions and how fair they thought it was that luck played a role in determining each participant’s monetary contribution.
Finally, they were asked to fill in a brief demographics survey (age, gender, nationality, education level, political leaning, and income level).
4. Hypotheses
The primary aspiration of this study is to confirm previous findings on the self-serving bias which argue that allocation decisions are influenced by socioeconomic status and, more importantly, to find possible effects between one’s allocation decisions and the level of inequality. Attention serves as a mediator and predictor of those effects. In particular, the novel aspect of this study is to understand how allocation biases and attention shift when luck is made explicit, and inequality increases. The following schema illustrates the relationships that this study aims to examine.
Figure 3: Framework of the hypothesized effects.
Amasino et al. (2021) find that ‘advantaged’ subjects look relatively more at outcome information (monetary contribution) than at merit information (correct answers), when
compared to ‘disadvantaged’ subjects. In this study, it is hypothesized that ‘advantaged’
dictators look relatively more at outcome information than at merit information and luck information. This could be due to their discomfort looking at their pure luck or advantage in the payment rate, as that might make them feel obligated to keep a smaller share of the surplus for themselves. So, they are likely to try to avoid that information in order to keep more for themselves. As a result, they may feel entitled to their monetary contribution, which obscures the role of luck to some degree, therefore paying more attention to it and taking it into account when dividing. This tendency is expected to be observed for all levels of inequality (multipliers), although to varying degrees for each.
Motivated attention predicts that ‘disadvantaged’ dictators, on the other hand, look more at merit and luck information than at outcome information. Knowing in advance that luck information puts them in a disadvantaged position, they look for the most plausible source of information to justify a fair division, which is merit information. Moreover, it seems likely that those who look more at merit information also look more at luck information, especially as inequality rises, given that giving less to the ‘advantaged’ presumably means that allocations will be more merit-based and that subjects will attempt to compensate for the luck component of each participant’s contribution.
These predictions are consistent with previous literature and the notion that people employ fairness views in a self-interested manner (Rodriguez-Lara & Moreno-Garrido, 2012;
Ubeda, 2014; Dengler-Roscher et al., 2018).
Hypothesis 1 (Status and Attention).
1.1. For each multiplier, ‘advantaged’ dictators look relatively more at outcome information than at merit information and luck information, compared to ‘disadvantaged’ dictators.
1.2. For each multiplier, ‘disadvantaged’ dictators look relatively more at merit information and luck information than at outcome information, compared to ‘advantaged’ dictators.
When the level of inequality is extreme, people tend to discredit internal factors as an explanation for inequality and begin to believe that external factors play a role in wealth and poverty (Davidai, 2018). Thus, it is hypothesized that as the level of inequality (multiplier) increases, participants take luck information on payment rates more and more into account, which translates to looking at it more. In addition, the fact that the gap between payment rates widens as the multiplier increases could mean that people more likely to visually notice that disparity.
It is hypothesized that the increase in attention to luck information is observed to a lesser extent among ‘disadvantaged’ than among ‘advantaged’ participants, because
‘disadvantaged’ already look relatively more at luck information at the lowest multiplier. As mentioned above, it seems reasonable to hypothesize that paying more attention to luck information implies also paying more attention to merit information, since merit information can be viewed as the basis for allocation decisions as inequality widens.
Hypothesis 2 (Inequality and Attention).
2.1. As the multiplier increases, ‘advantaged’ subjects look relatively more at luck information and merit information.
2.2. As the multiplier increases, ‘disadvantaged’ subjects look relatively more at luck information and merit information, even if in a smaller proportion than in 2.1.
As the inequality gap widens, it is conjectured that inequality perception also rises.
Since inequality perception and people’s willingness to redistribute are positively related (Gimpelson & Treisman, 2018), it is hypothesized that allocations from ‘advantaged’ to
‘disadvantaged’ subjects will increase. As ‘advantaged’ subjects increasingly follow the meritocratic norms, their allocations to ‘disadvantaged’ will increase and the self-serving bias will likely decrease.
Naturally, it is hypothesized that allocations from ‘disadvantaged’ to ‘advantaged’ will not undergo a large change, since they already look relatively more at merit information in the lowest multiplier and divide the surplus accordingly more than ‘advantaged’ dictators.
Moreover, there might also be some self-serving bias effect in their behavior. Perhaps, they also tend to believe that ‘advantaged’ really deserved more than them, which led them to give more to ‘advantaged’, also showing some bias influence to some extent. If that effect occurs, it is likely to be attenuated as inequality rises and meritocratic norms are increasingly followed.
Hypothesis 3 (Attention and Allocations).
3.1. As attention to luck information and merit information increases relatively to outcome information, in general, allocations from ‘advantaged’ to ‘disadvantaged’ subjects increase.
3.2. As attention to luck information and merit information increases relatively to outcome information, in general, allocations from ‘disadvantaged’ to ‘advantaged’ subjects decrease, even if in a smaller proportion than in 3.1.
Hypothesis 3, which combines the first two, can be used to derive the answer to the research question of this study. The hypothesized tendency is that, as the inequality level increases, the self-serving bias will tend to be reduced. So, dictators’ self-serving allocations will be relatively more pronounced under the ‘multiplier 3’ condition than under the ‘multiplier 5’ condition, and even less under the ‘multiplier 7’ condition.
5. Results
In this section, the results are outlined. The summary statistics of the main variables of the study are presented, as well as the findings for each hypothesized effect. Finally, some limitations of the results are discussed.
First, participants’ effort in the production phase (moving sliders tasks) is analyzed, to ensure that subjects were engaged in the task. On average, participants set 30.73 sliders to the correct value. The average number of correct answers was 29.51 for ‘advantaged’ and 31.76 for ‘disadvantaged’ dictators. Although their difference is statistically significant at the 5%
level (p = 0.019, Wilcoxon rank sum test), dictators were only told about their advantaged or disadvantaged status after the production phase and it was determined randomly, so it is unlikely that status caused effort differences between the two groups. Given the limitations of this study such as the sample size, which will be discussed in more detail later in this section, it seems reasonable to assume that ‘advantaged’ and ‘disadvantaged’ subjects exerted a similar amount of effort in the task. Within each group, the number of correct answers increased throughout the three tasks (27.93, 29.61, 30.98 for ‘advantaged’ and 29.73, 31.83, 33.71 for
‘disadvantaged’), possibly due to learning effects. However, in the subsequent phase of the experiment, the number of correct answers was randomly picked from one of the three tasks, so any differences in task performance are randomized across rounds.
Table 2 summarizes the mean values of the main outcome variables in this study.
Table 2: Summary Statistics.
A more detailed analysis on the data is conducted below, but a few initial notes are indispensable. First, no dictator kept all to himself/herself in all rounds. Dictators kept the entire surplus to themselves in roughly 1.6% of the observed rounds. These findings are in line with previous literature that suggests that only a very a small portion of dictators behaves in a completely selfish manner in dictator games with production (Cappelen et al., 2010, Rodriguez- Lara & Moreno-Garrido, 2012). Moreover, only one dictator gave the entire surplus to the receiver in all rounds. The percentage of rounds in which dictators exhibited that type of altruistic behavior was also relatively insignificant, close to 1.7%. Another finding worth noticing is that, in general, ‘advantaged’ dictators kept significantly more to themselves (58.47%) than ‘disadvantaged’ dictators (40.56%) (p < 0.001, Wilcoxon rank sum test). That difference remains extremely highly significant for each treatment. Overall, the share of the surplus that dictators kept to themselves decreased as the multiplier increased (49.81%, 48.77%, 47.72%), a trend that might be driven by ‘disadvantaged’ keeping even less throughout multipliers.
Second, a few remarks on attention patterns should be highlighted. Both ‘advantaged’
and ‘disadvantaged’ dictators looked more at outcome than at merit or luck information in all multipliers, except for ‘disadvantaged’ dictators in multiplier 3 and ‘advantaged’ dictators in multiplier 7, where they looked more at merit information. Overall, ‘disadvantaged’ spent on average 2.49 seconds looking at merit information, whereas ‘advantaged’ spent 2.18 seconds on it, a difference that is highly significant (p < 0.001, Wilcoxon rank sum test). However, the difference between the two groups is neither statistically significant for the dwell time to luck information (1.97 seconds for ‘advantaged’, 2.07 seconds for ‘disadvantaged’) nor for dwell time to outcome information (2.51 for ‘advantaged’, 2.54 for ‘disadvantaged’). Moreover,
‘disadvantaged’ dictators spent, in general, 440 more milliseconds looking at information than
‘advantaged’ dictators, possibly suggesting that ‘advantaged’ dictators either actively tried to avoid information or were just less information-seeking. An analysis on information avoidance is conducted on the discussion section of this thesis.
5.1. Status and Attention
With the aim of testing Hypothesis 1, dictators’ status and their attention patterns are studied.
Particularly, the relationship between being randomly assigned an ‘advantaged’ position in payment rates and how dictators engage with the different types of information. As mentioned before, recent attentional findings (Amasino et al., 2021) suggest that ‘advantaged’ dictators
look relatively more at outcome than at merit information, as opposed to ‘disadvantaged’
dictators. In this thesis, explicit information on luck (payment rates) was introduced. For the reasons mentioned before, it is hypothesized that ‘advantaged’ subjects look relatively more at outcome information than at merit and luck information, and that ‘disadvantaged’ subjects look relatively more at merit and luck information than at outcome information.
Figure 4 illustrates how ‘advantaged’ and ‘disadvantaged’ dictators engage with outcome information, relative to merit and luck information. It is important to note that although the outliers are excluded from the distribution in the boxplots in Figure 4, they were included in all the analyzes in this study, because there was no statistically strong reason to exclude them. Thus, all the presented statistics and analyzes include all data, including outliers, except for the boxplots in the figures.
Figure 4: Overall Differences in Attention, by Dictator Status.
Attention O-M (s): difference in dwell time between outcome information (O) and merit information (M), in seconds. Attention O-L (s): difference in dwell time between outcome
information (O) and luck information (L), in seconds.
On average, ‘advantaged’ dictators’ difference in dwell time between outcome and merit information was 0.32 seconds, against 0.05 seconds for ‘disadvantaged’ dictators, a marginally significant difference (p = 0.087, Wilcoxon rank sum test). That is probably because, as mentioned above, ‘disadvantaged’ dictators spent significantly more time looking at merit information relative to ‘advantaged’ dictators, which might generate a relatively smaller difference in dwell times between outcome and merit information for ‘disadvantaged’.
Additionally, the difference in dwell times between outcome and luck information was also generally higher for ‘advantaged’ than for ‘disadvantaged’ participants, but that difference is not statistically significant (p = 0.74, Wilcoxon rank sum test).
To further analyze the relationship between dictators’ status and their attention patterns, Table 3 shows linear regressions which control for subject characteristics and include standard errors clustered at the individual level. The dependent variable is the difference in dwell time between outcome information (O) and merit information (M) (first panel), and the difference in dwell time between outcome information (O) and luck information (L) (second panel), once the goal is to observe possible differences in attention to outcome information relative to merit and luck information, by dictator’ status.
Table 3: The Effect of Status on Attention.
All models are linear regressions, with the standard errors clustered at the individual level in parentheses and controlling for individual characteristics. List of controls: age (5 categories), gender
(3 categories), nationality (12 categories), education (5 categories), income (6 categories), and political leaning (8 categories).
M3: Multiplier 3, M5: Multiplier 5, M7: Multiplier 7. Adv: Advantaged.
Column (1) shows that being ‘advantaged’ implied spending 600 milliseconds more looking at outcome information than at merit information, and 180 milliseconds more looking at outcome information than at luck information, compared to ‘disadvantaged’ dictators, although these findings are not statistically significant. Interestingly, both the difference in dwell time between outcome and merit information and the difference in dwell time between outcome and luck information decrease as the multiplier increases, even turning negative under
multiplier 7 for outcome and luck information. Column (2) of the first panel shows that the difference in dwell time between outcome and merit information is significant at the 5% level for multiplier 3, when inequality in payment rates is the lowest in multiplier 3. These results point out a general tendency of ‘advantaged’ dictators to look relatively more to merit and luck information as inequality in payment rates widens. This relationship is further investigated in the next subsection.
Result 1. ‘Advantaged’ dictators look relatively more at outcome information than at merit and luck information, compared to ‘disadvantaged’ dictators, except for outcome versus luck information in multiplier 7. This difference is significant at the 5% level for the difference in dwell times between outcome and merit information in multiplier 3, but does not hold across all multipliers.
5.2. Inequality and Attention
To test Hypothesis 2, the relationship between inequality in payment rates and attention is investigated. It is tested whether increasing inequality in payment rates solely determined by pure luck influences how dictators engage with different sources of information. It is hypothesized that, as the multiplier6 of payment rates increases, the difference in dwell time between outcome and merit information and between outcome and luck information decreases, as a possible consequence of a relative increase in attention to merit and luck information.
Figure 57 depicts the difference in dwell time between outcome and merit information, in seconds, for each multiplier and by dictator status. Overall, the difference in dwell time between outcome and merit information does not change significantly throughout multipliers, increasing from 0.23 seconds to 0.30 seconds from multiplier 3 to 5, and then decreasing to 0.00 seconds in multiplier 7. In the case of ‘advantaged’ subjects, that difference insignificantly decreased from multiplier 3 (0.50 seconds) to multiplier 5 (0.49 seconds) (p = 0.67, Wilcoxon matched-pairs test), then marginally significantly decreased in multiplier 7 (-0.10 seconds) (p
= 0.016, Wilcoxon matched-pairs test).
6 The multiplier corresponds to the number of times that the ‘advantaged’’ payment rate is greater than the
‘disadvantaged’’ payment rate in each treatment. For example, in multiplier 3, the ‘advantaged’ payment rate (0.60€/correct answer) equals threefold the receiver’s (‘disadvantaged’) payment rate (0.20€/correct answer).
7 It is worth mentioning once more that all the statistics and analyzes conducted in this thesis include all data, including the outliers, even though the outliers are excluded from the distributions in the boxplot figures.
Figure 5: Differences in Attention to Outcome versus Merit Information, by Status and Multiplier.
Attention O-M (s): difference in dwell time between outcome information (O) and merit information (M), in seconds.
On the other hand, there were no statistically significant changes in the difference in dwell time between outcome and merit information across multipliers for ‘disadvantaged’
subjects. It rose from -0.07 seconds in multiplier 3 to 0.14 seconds in multiplier 5 (p = 0.18, Wilcoxon matched-pairs tests), and then it decreased to 0.08 seconds in multiplier 7 (p = 0.97, Wilcoxon matched-pairs tests). When comparing the lowest (3) and the highest (7) multipliers, there are no significant changes in the difference in dwell time between outcome and merit information for ‘disadvantaged’ dictators (p = 0.47, Wilcoxon matched-pairs test), but there are highly significant changes in the case of ‘advantaged’ dictators (p = 0.0022, Wilcoxon matched-pairs test).
Figure 68 depicts the difference in dwell time between outcome and luck information, in seconds, for each multiplier and by dictator status. Overall, the difference in dwell time between outcome and luck information significantly decreases at the 5% level from multiplier 3 (0.78 seconds) to multiplier 5 (0.42 seconds) (p = 0.011, Wilcoxon rank sum test), decreasing again, although not significantly, to multiplier 7 (0.31 seconds). For ‘advantaged’ dictators, the difference in dwell time between outcome and luck information did not change significantly
8 It is worth mentioning once more that all the statistics and analyzes conducted in this thesis include all data, including the outliers, even though the outliers are excluded from the distributions in the boxplot figures.
across multipliers, although it decreased (0.90 seconds, 0.53 seconds, and 0.20 seconds in multipliers 3, 5, and 7, respectively).
Figure 6: Differences in Attention to Outcome versus Luck Information, by Status and Multiplier.
Attention O-L (s): difference in dwell time between outcome information (O) and luck information (L), in seconds.
For ‘disadvantaged’ dictators, the difference in dwell time between outcome and luck information decreased, on average, from 0.68 seconds to 0.33 seconds from multiplier 3 to multiplier 5, a shift that is statistically significant at the 5% level (p = 0.039, Wilcoxon matched-pairs test). From multiplier 5 to 7, it insignificantly increased to 0.40 seconds.
Analysing changes between multipliers 3 and 7, the most extreme ones, it is observed that there is a decrease in ‘advantaged’’ difference in dwell time between outcome and luck information, which is significant at the 5% level (p = 0.015, Wilcoxon matched-pairs test), probably because they spent more time looking at luck information. For ‘disadvantaged’ subjects, it is observed an insignificant but also negative trend across multipliers (0.68 to 0.40 seconds).
To further investigate the relationship between inequality in payment rates and attention to merit and luck information, Table 4 shows the results of linear regressions which control for subject characteristics and include standard errors clustered at the individual level in parentheses. The dependent variable is the difference in dwell time between outcome and merit information (Columns (1) and (2)) and the difference in dwell time between outcome and luck information (Columns (3) and (4)), to understand the impact that different levels of inequality in randomly generated payment rates might have on attention to merit and luck information relative to outcome information. Interaction terms between each treatment and the dictator’s advantage were included, to better understand the effect of each multiplier simultaneously with being ‘advantaged’ on dwell times.
Table 4: The Effect of Inequality in Payment Rates on Attention.
All models are linear regressions, with the standard errors clustered at the individual level in parentheses and controlling for individual characteristics. List of controls: age (5 categories), gender
(3 categories), nationality (12 categories), education (5 categories), income (6 categories), political leaning (8 categories), and total dwell time of individuals in Columns (2) and (4).
M5: Multiplier 5, M7: Multiplier 7. Adv: Advantaged. Attention All: total dwell time of individuals.
Attention O-M: difference in dwell time between outcome and merit information, in seconds.
Attention O-L: difference in dwell time between outcome and luck information, in seconds.
Column (1) shows that ‘advantaged’ dictators looked significantly more (970 milliseconds) than ‘disadvantaged’ dictators at outcome relative to merit information at the baseline multiplier 3. Moreover, multiplier 5 influences ‘advantaged’ dictators to spend 300
more milliseconds on merit information relative to outcome information. That difference increases to 830 milliseconds when at the multiplier 7 treatment, with the highest level of inequality in payment rates, and it is significant at the 5% level. Thus, the most extreme level of inequality, multiplier 7, brings them closer to ‘disadvantaged’, but it does not eliminate the difference between the two groups. Column (2) shows similar effects when controlling for total dwell time of individuals. Columns (3) and (4) present the analogue results for the difference in dwell time between outcome and luck information. There are no strong differences to be interpreted, suggesting that changes in inequality in payment rates did not affect attention to luck information as much as attention to merit information.
It is not surprising that the total dwell time of the subjects significantly influenced the difference in dwell time between outcome and merit or luck information, as it might be an endogenous regressor. Therefore, the results of the regressions that include this control variable were included in the analysis, but must be interpreted with caution in this subsection.
Result 2.1. ‘Advantaged’ dictators look relatively more at merit information than at outcome information, a trend across multipliers that is significant at the 5% level for multiplier 7. There are no statistically significant changes in the difference in dwell time between outcome and luck information throughout multipliers.
Result 2.2. For ‘disadvantaged’ dictators, there are no significant differences in dwell time to merit information relative to outcome information throughout multipliers. ‘Disadvantaged’
dictators tended to look relatively more at luck information than at outcome information across multipliers, a result that is significant at the 10% level for multiplier 5.
5.3. Attention and Allocations
To examine Hypothesis 3, the relationship between dictators’ attention to merit and luck information and allocation decisions is analyzed. Specifically, it is tested whether an increase in time spent looking at merit and luck information relatively to outcome information leads to higher shares of the surplus given to the receiver, in the case of ‘advantaged’, or lower shares of the surplus given to the receiver, in the case of ‘disadvantaged’, therefore reducing the self- serving bias. It is relevant to mention that the “other” participant, the receiver, always corresponds to an ‘advantaged’ subject for ‘disadvantaged’ subjects, and vice-versa.
Figure 7 helps to visually observe the effect of looking at merit and luck information as opposed to outcome information on the percentage of the surplus allocated to the other participant.
Figure 7: Attention to merit and luck information versus outcome information and the percentage of the surplus given to the receiver, by dictator status.
Attention O-M (s): Difference in dwell time between outcome (O) and merit (M) information, in seconds. Attention O-L (s): Difference in dwell time between outcome (O) and luck (L) information,
in seconds. % given to the other: Share of the surplus given by the dictator to the receiver.
Looking relatively more at outcome information may bias dictators’ allocations towards
‘advantaged’ keeping more and ‘disadvantaged’ keeping less. When the difference in dwell times between outcome and merit information decreases, the relative time spent looking at outcome information decreases as well. The relationships in figure 7 seem to corroborate that hypothesis, as ‘advantaged’ dictators looked relatively more at merit information, their allocations to their ‘disadvantaged’ receivers increased. Inversely, as ‘disadvantaged’ dictators look relatively more at merit information, their allocations to their ‘advantaged’ receivers decreased. Likewise, when the difference in dwell times between outcome and luck information decreases, the relative time spent looking at outcome information decreases. Once more, the relationship highlighted in figure 7 shows that, as ‘advantaged’ dictators looked
