Doing nothing is all you've ever wanted to do : does simply benefiting affect not only our desire to change the status quo... but how we view it too?

(1)

Doing Nothing is All You’ve Ever

Wanted to Do

Does simply benefiting affect not only our desire to

change the status quo... but how we view it too?

By Ciaran Downey

MSc Economics: Behavioural Economics & Game Theory

UvA Student No: 1177845 ECTS: 15

Abstract

Status quo bias refers to the observed tendency of humans to make choices within our environment which remain as close as possible to those previously selected by ourselves or others. It has been found in certain experimental set-ups to account for a large proportion of the choices made by subjects. In this paper we sought to observe how the status quo bias manifests itself when sub-jects make choices about whether to change (meaningfully or not) pre-selected distributions selected by others depending on whether they themselves ben-efit from the distribution. We also attempted to observe how it interacts with another bias, namely the self-serving bias, to justify subjects’ decisions about whether or not to act. We found that while status quo bias was more pronounced when you benefit from not changing it, it did not cause subjects conception of what is fair to alter via a self-serving bias.

(2)

Statement of Originality

This document is written by Ciaran Thomas Downey 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.

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

(3)

1 Introduction

How many of us like to assume that we behave primarily according to a set of princi-ples and that our own moral judgements have been arrived at through greater under-standing, empathy and/or rationality? Could something as simple as a pre-selected option radically alter our distribution preferences while leaving us completely un-aware that it has affected us at all? Could it even go further and affect our very conception of what is fair?

Seeking to answer some of these questions and inspired by the findings within the existing literature surrounding status quo bias – about not only how it could affect our distribution preferences without us being aware (Dhingra, Gorn, Kener, & Dana, 2012), how we as humans become overly reliant upon it in the face of complexity (Boxall, Adamowicz, & Moon, 2009), and how it can be used to improve policy outcomes (Li, Hawley, & Schnier, 2013; Thaler & Benartzi, 2004), but also how it can affect our own moral attitudes – we hoped to determine the extent and power of the status quo bias. To do this we conducted an experiment according to a 2x2 factorial design with our two alternating factors being whether someone benefits or not from the status quo (which was decided by somebody else) and whether they can meaningfully change it or not. We sought to observe how subjects behaved according to the presence or not of these factors as well as to whether they caused the stated beliefs of what constituted a ‘fair’ distribution to differ between treatments.

However should simply benefiting affect not only our desire to change the status quo but also how we view it, we believe that this would be particularly pertinent and relevant within the field of institutional economics. If unfair procedures and norms could be tacitly supported and remain unopposed despite not promoting greater economic efficiency, innovation or indeed social mobility, we believe this could be to the detriment of the wider society as a whole.

In order to explore this topic we created our very own variant of the dictator game popularly used within the field of experimental economics (Camerer & Fehr, 2004). Our variant expanded the standard game to three players one fulfilling the role of a dictator, who chooses how to split a group endowment between the players; another player being put in the role of their ‘payoff’ partner, which meant that they received exactly the same amount of money as the dictator received; and the final role being a standard passive player who receives any remaining money not distributed to the dictator and their partner. In a first stage dictators selected a distribution and then in a second stage, to see the behaviour of those who benefited,

(5)

we observed and analysed how subjects acted when assigned to role of the dictator’s partner and being offered either the option to meaningfully change the distribution or ,alternatively in a different treatment, hypothetically indicate what they would have changed the distribution to. In order to observe the behaviour of those who didn’t benefit we assigned half of the subjects to a new outsider witness role who were observing the dictator’s choice without themselves benefiting from it - they were then offered the same meaningful/hypothetical choice as those assigned to the role of the dictator’s partner.

We believed that should subjects opt to maintain the status quo when they themselves benefited, that they may seek to support the decision as being ‘fair’ via the unconscious effect of a self serving bias, whereby a subject seeks to maintain their own self-image by justifying and explaining the ‘valid’ motivations behind making such a decision. Indeed within the existing literature there is already a great deal of evidence to suggest the effect this has on our subjective formations as to what is fair. As an example, (Andreoni & Sanchez, 2014) found that subjects stated beliefs as to the trustworthiness of their partner could conveniently differ from their revealed beliefs if it was to their own self-interest, and (Rodriguez-Lara & Moreno-Garrido, 2012) found that the moral frameworks that subjects chose to abide also differed dependent on again whether it was in their own self interest.

What we found was that while benefiting did indeed alter what inequalities sub-jects were willing to accept, whether they could alter the distribution meaningfully or not did not seem to have any effect whatsoever - which was surprising for one of our treatments. Moreover we found that the moral judgements of subjects did not seem to differ significantly by treatments, suggesting that the power the self serving bias is not strong enough in all cases to force subjects to justify being inert in the fact of the status quo.

(6)

2 Literature Review

In our experiment we make use of a variant on the standard dictator game where a second player can choose whether or not to alter the choice of the dictator afterwards either meaningfully, in a payout relevant way, or hypothetically. For this reason we first examine the literature surrounding this experimental set-up before exploring the heuristics and biases which we believe could prove to have an effect on our results.

2.1 Dictator Games & Observed Behaviours

Dictator games are used in a variety of settings and can be used to attempt to observe both altruistic behaviour and distribution preferences (Camerer & Fehr, 2004). Normally they take the form of two ‘players’ with one player (the dictator) choosing how to distribute an amount of money between themselves and another person who has no choice but to accept the offer (the passive player). For this reason they do not constitute an experimental game in the usual sense as they do not contain an interaction between the players as only one individual has an active role in deciding the distribution. The reason it is thought of as being better able to test altruistic and egalitarian preferences is due to the fact that it does not suffer from players expectations with regards to reciprocity or stategic behaviour, which can be a problem with other games such as the ultimatum game1 _{(Hoffman, McCabe,}

& Smith, 1996).

According to the classical economic conception of human beings as rational util-ity maximisers (more commonly referred to as ‘homo economicus’), in this set-up dictators should seek to maximise their own payoff at the expense of passive partic-ipants by distributing all of the available payoff to themselves. Yet in experimental settings, when dictators do not know who they are playing with, we do not find this. Forsythe, Horowitz, Savin, and Sefton (1994) found that only 36% of the sub-jects they observed chose to completely take everything, while 22% decided to give the passive player an equal share of the money or better. In another paper, with a smaller sample size (Burnham, 2003), a majority of participants opted to take every-thing (54%) but 19% still gave 1/4 or more of the total distribution available to their

1_{The ultimatum game is similar to the dictator game with the only exception being that the}

player who does not choose the distribution can choose to reject the offer should they so wish; however, if they do both players will receive nothing. In such a setting the player who chooses how to distribute the payoff may seek to allocate an amount that they believe will be agreeable to the other player so as to ensure they do not reject the offer – see Thaler (1988) for more detail.

(7)

partners. When looking at gender disparities, there is also evidence to suggest that women are far more generous than their male counterparts, with one study finding women, on average, offer twice as much to their partners as their male counterparts do (Eckel & Grossman, 1998).

However do these results indicate that subjects are inequity averse? An ex-pansive paper by Dana, Weber, and Kuang (2007) questioned the degree to which fairness considerations affected results in dictator games. In their study they found three separate factors which greatly decreased equitable distributions (from 78% to 34-38%) by dictators compared to that seen in a normal condition. The three factors which caused this decline were 1) the introduction of uncertainty, where the partner was made aware what they would always know what they would receive in a distri-bution but had to choose whether to reveal what the passive player would receive2_;

2) sharing accountability for a decision with another dictator; and, finally 3) there being plausible deniability about why a selfish decision was chosen (in the experi-ment, this was a time limit forcing a choice). The second factor is especially relevant to our experiment as we make use of a three-person dictator game, and suggests that people are far more likely to accept an inequality if they share responsibility for the outcome with another person.

In our experiment subjects are also unaware who they are playing with and anonymity itself can have an important impact on behaviour in dictator games. Bra˜nas-Garza (2006) found that by providing dictators with more information about the personal financial position of the passive player (specifically that they were poor) increased the amount they offered this player. Moreover, Burnham (2003) found that dictators were significantly more likely to distribute money equally if they were provided with a photograph of the person they were playing with; this is in line with other literature that emphasis the importance of identification on equitable distributions in dictator games (Bohnet, 1999).

There is also evidence to suggest that one of the reasons why people choose equi-table distributions is due to their awareness that their action is being perceived and therefore desire to be seen to abide by a social-norm. In order to explore the degree to which participants may aim to protect how others see them, researchers employed the use of a double-blind in two-player games, such as the dictator game, where even the experimenter is unaware of participants’ choices. When these experimental con-ditions are introducted the altruistic behaviour of dictators falls (Hoffman, McCabe, Shachat, & Smith, 1994). Note, however that the participant are still fully aware of

2_{This paper found that a number of participants chose not to act and kept the partners pay}

(8)

what choice they are made.

2.2 Inequity Aversion: Theories and Models

As our experiment makes use of a experimental design which can contain the pres-ence of an inequality we briefly cover some of the existing theories as why in-dividuals seek to reduce inequality within distribution games even if they them-selves can benefit from an inequality. Challenging the model of self-interest pur-ported by classical economics, Fehr and Schmidt (1999) developed their own indi-vidual utility model, which sought to explain why indiindi-viduals do not simply max-imise their own outcomes in experimental settings and often opt for equitable out-comes. In a game with n participants, the model takes the form of the following: Ui(x) = xi − αi_n−11

P

j6=imax{xj − xi, 0} − βi_n−11

P

j6=imax{xi − xj, 0}. Here xi

represents the amount that a player i receives, the second term refers to i’s utility loss from disadvantageous inequality, and the third term refers to their utility loss from advantageous inequality. Generally, for most people we would expect αi ≥ βi

as individuals are far more likely to opt to an inequality that is to their advantage than to their disadvantage, at least in experimental settings (Loewenstein, Thomp-son, & Bazerman, 1989). Such a model could be used to explain the behaviour that is witnessed in dictator games, especially when the dictator has been given no information about who they are playing with. If the participant is inequity averse, according to this model, they should seek to divide the money equally in our ex-periment among the three payoff eligible participants. If they have no aversion to inequality they should seek to maximise their own payoffs, which in our experiment would also maximise the payoff of their partner all at the expense of the passive player who would receive nothing.

Taking a different approach, Bolton and Ockenfels (2000) developed a model to explain not only equitable choices, but also reciprocal and competitive behaviour. The motivation function for their model is vi = vi(yi, σi), where yi represents the

payout to player i and σi being their share of the total payout to all players. vi is

maximised if σi = _n1, with n again referring to the number of participants that are

taking part in the game. Therefore an individual is less likely to be concerned on how the money is distributed so long as their own share is as close as possible to the average amount given. In our experiment, the only way for a partner to get as close as possible to the average amount given to each player is by the money being equally distributed amongst the three players. One limitation of the theory, as the authors admit, is that it is a theory of local behavior meaning it can only be used to explain the stationary patterns seen in simple experimental games, especially those

(9)

played over brief time periods, however our experiment does fulfil these criteria. There are however some further criticisms of both the models above which will limit their predictive power with regards to our experiment. Engelmann and Strobel (2004) tested these models using a series of experimental designs which could disas-sociate various components from each other and looking at the role that inequality aversion truly has on the behaviour observed, along with other considerations such as efficiency and maximin3 considerations (which in our experiment would involve splitting the distribution equally as this is the only way to maximise the lowest pay-off given to all the paricipants). While they state that the Fehr and Schmidt model fits the results seen more closely than the ERC model, they claim that efficiency and maximin considerations more readily predict the results in their experiments than either of these two models. More glaringly, they state a key problem with both of the above models is that they do not account for how perceived ‘intentions’ affect our decision making. Outcome based models such as F & S and ERC are inconsistent with findings such as those found in trust games where subjects can be asked their beliefs about the intentions of the player who decides to place their trust in another player (McCabe, Rigdon, & Smith, 2003).

2.3 Status Quo Bias

In our experiment, subjects were asked whether they wanted to – or would have wanted to – change a decision, for that reason it is important to visit the litera-ture surrounding status quo bias (also referred to as default bias), which has been extensively covered within the fields of psychology, economics and sociology. Put simply, status quo bias is the observed tendency of humans to act (or rather not act) in a way which requires the least amount of effort, such as sticking to a previous choice or a choice already made by another – this could be a default option on an online form, for example (Kahneman, Knetsch, & Thaler, 1991). It was shown by Samuelson and Zeckhauser (1988) that status quo bias is present in a number of real-world scenarios (such as keeping retirement funds in accounts which generate a lesser rate of return than could be achieved elsewhere) and that it can prevent us from making changes which ultimately benefit us and others. Such is the perceived power of status quo bias that policy makers have even sought to use it to improve so-cially desirable outcomes. Li et al. (2013), for example, showed the efficacy of using an opt-out as opposed to opt-in scheme as a way to increase organ donation rates significantly; similar initiatives have been used successfully to increase retirement

(10)

saving (Thaler & Benartzi, 2004).

In an experimental setting, and using a dictator game, Dhingra et al. (2012) found that by simply having a distribution pre-selected, but not locked, that this caused subjects final distribution to be pulled towards that value. Therefore if an equal split was pre-selected it would increase the frequency of equitable choices being selected, whereas if the dictator takes all option had been selected, this would increase the frequency of selfish distributions chosen by dictators. Interestingly, the authors reported that the majority of the subjects were entirely oblivious as to how the pre-selected option had affected their choice, and denied it had played any role at all in their final decision.

Status quo bias is also relevant to our experiment because it has been found to alter our own moral judgements. In our paper we are not only looking to see how behaviour changes but also how subjects’ moral perceptions might change as a result of being assigned to a particular treatment group. For example, Kay et al. (2009) showed that, generally, individuals have a tendency to justify the status quo to support what they view as their own social, economic and political systems -even more so when they view those systems as being under threat. Within the field of political psychology, this is more commonly referred to as system justification theory. Jost, Banaji, and Nosek (2004) performed a review of the literature in this field and stated that their was indeed evidence to suggest that people are not only motivated to view their own groups favourably but also their own systems as well; this even appeared to be found when individuals were at a disadvantage within their own political or economic system.

One further point worth mentioning is that the increased complexity of choice architecture has been found to increase subjects’ use of the status quo bias (Boxall et al., 2009); however we do not envision this being the case in our experiment as we utilise a fairly simple experimental design.

2.4 Self-serving Bias

One of the key purposes of our experiment is to see if the moral attitude of subjects changes depending on the two altering factors used in our 2x2 design: whether you benefit financially from a distribution and whether you can meaningfully change it. For this reason it is especially important to focus on the effects of self-serving biases. Self-serving has been described as “any cognitive or perceptual process that is distorted by the need to maintain and enhance self-esteem” (Sherrill, 2007). In our experiment a self-serving bias could manifest itself in the form of an individ-ual seeking to justify the distributions they have selected themselves, or agreed to

(11)

through inertia, as being ‘fair’.

An example of self-serving bias in practice can be seen in a paper by Andreoni and Sanchez (2014). They created an experiment which tested participants actions in a trust game according to their ‘stated’ beliefs and then again in a variant which was designed so as to reveal their actual beliefs. They found that participants would state that their partner would act selfishly more often when they themselves chose to act selfishly, but then in the revealed condition chose to trust players as opposed to have their payment chosen probabilistically (which should have been more attractive to them according to their ‘stated’ belief of how trustworthy their partner was). The authors suggest that this shows that when choosing the selfish option their stated beliefs are used to protect their own image, both to themselves and other people (in this case the experimenters)4_.

Specifically looking at the effects of self-serving bias within the context of dictator games, Rodriguez-Lara and Moreno-Garrido (2012) questioned what the altruistic behaviour seen in dictator games should be attributed to. They used an altered two-person dictator game where both players had, separately, completed 20 questions in an earning phase and were rewarded for each correct answer. In one treatment group dictators were rewarded more per correct answer, and in another, less. The dictator then had to decide how to distribute their collective earnings (with full knowledge of how much each participant had contributed). They found that dictators behaviour could not be explained by any single justice principle (egalitarian, accountability or libertarian) and that the way that the distribution was split by the dictator would change to be in line with one of these principles if it was in the dictators interest.

Outside of these standard experimental games Babcock and Loewenstein (1997) found that individuals choose to selectively evaluate information to determine how much a defendant should have to pay out to a plaintiff when their payout is depen-dent on how much they state one of the two parties should receive. They observed this bias in both artificial environments created by the experimenters such as the one above, and it more common everyday environments such as asking people to judge how many fouls were committed by their team and the team they were playing against and comparing this with the same questions put to opposing fans. Unsur-prisingly the results between both sets of fans differed greatly.

In a working paper by Cassar and Klein (2016) they have shown how experience can impact redistribution preferences and also beliefs. In a repeated experiment

4_{Interestingly beliefs themselves can have a great impact on redistribution preferences if subjects}

believe that people in society are responsible for their own weak economic position they are far less likely to support strong redistribution (Fong, 2001)

(12)

where payoff was determined by a tournament/lottery, they found that those indi-viduals who had experienced being a loser (i.e. had done worse of than the other participants in the tournament/lottery) stated that they had higher redistribution preferences in subsequent games and attributed more of the initial distribution ef-fects to chance. Our experiment does not have a repeated element, but this is just one more example of how the moral psychology of subjects can seemingly change depending on their circumstances within the experiment.

(13)

3 Methodology

In order to explore the research question posed the desire was to create an experiment which could focus and isolate the effect of two factors and how they affect stated distribution preferences: whether a person benefits financially from a given scenario , and, whether they can meaningfully change the distribution. As a result the experiment made use of a 2x2 factorial design (see Figure 3.1). It was therefore opted to design an experiment where these conditions could be observed with minimal difference between their essential elements.

Figure 3.1: 2x2 Factorial Design Showing Treatment Names

3.1 Experimental Design

In order to be able to observe the four separate conditions, an experiment was de-vised which centred around an altered sequential three-person dictator game com-prised of three roles within the game: a dictator (referred to as Role X in the experiment), the dictator’s partner (Role Y), and a totally-passive player (Role Z). Similar to other dictator games, in the first stage a dictator has the responsibility to choose how to distribute an amount of money (a group endowment) between all of the players. However, the following conditions were imposed on the dictator:

1. The partner must receive the same amount as the dictator

2. The lowest amount they can choose for themselves is where all players receive the same amount of money

3. The amount of that each player receives should be by Euro (e) integers Therefore, the dictator (Role X) chooses an amount for themselves X according to X ∈ Z : (Endowment3 ≥ X ≥

Endowment

2 ) with the partner (Role Y) also receiving

(14)

the experiment, the total endowment given to the group wase72, which meant that the most the dictator could choose for themselves and their partner wase36 (where the passive-player receives nothing) and the minimum e24 (where all three players receive the same amount).

In the first stage of the experiment we would collect choices from real partici-pants acting as dictators so as to credibly be able to inform subjects in our treatments that they were witnessing the real choice of another human participant; the dictators choices are not what we are analysing in this experiment. We conducted this stage according to the rules above and once we had been provided with enough unique distributions selected by dictators (in the end we had four unique distributions that were well varied5_{) we proceeded to the second stage of the experiment.}

In the second stage subjects would be each randomly assigned across the four treatment groups indicated by Figure 3.1. In accordance with the recommendations of Zizzo (2010) we did not inform subjects of the other possible treatment conditions they could have been assigned to before or after they themselves were assigned to a treatment group in order to obfuscate the purpose of the experiment and the underlying research question and reduce experimenter demand effects.

If subjects were assigned to one of the treatment groups in which they benefit, they would be informed at the beginning of the experiment that they had been assigned to Role Y (the dictator’s partner); so in this treatment group there were three participants (X, Y and Z). All subjects were told that one group would be selected for payment according to the final amount set for that group after two stages – this would, of course, be just what the dictator (Role X) had selected had the subject been assigned to one of the non-consequential treatment groups, but could be different in the consequential treatment groups if the subject in stage two for that group opted to change the distribution (which would be the new amount paid, if selected). Those assigned to the treatment groups where they didn’t benefit were given a new role (Role W). Subjects in this role would be witnesses to the distribution meaning they did not stand to benefit from the distribution (or indeed the experiment) in any way; therefore in this treatment group there would be four roles (X, Y6, Z and W). They would still, however, be asked the consequential and non-consequential questions depending on whether they were assigned to NB-C

5_{From the four unique dictator choices the difference between what X/Y received and what Z}

received wase36 (e36, e36, e0), e24 (e32, e32, e8), e12 (e28, e28, e16), and e0 (e24, e24, e24).

6_{In the non-benefit treatments role Y would not be fulfilled by someone who had the same}

options available to them as in the benefit treatments, but would be a passive player (like Z) in that they had no action to fulfil.

(15)

Figure 3.2: Example of Bar Chart Used to Display Distribution

or NB-NC. These questions would have the same potential effect on the final set distribution as for those who benefited (see Appendix II.II for more detail).

All subjects would be shown one of the dictators’ choices and, in consequential treatment groups (B-C & NB-C), be asked if they would like to change the distri-bution; in the non-consequential treatment groups (B-NC & NB-NC) they would simply be asked a purely hypothetical question as to whether they would have changed the distribution, if they had been given the chance. The distribution choice of person X would be displayed via a bar chart (see Figure 3.2) to ensure that the inequality was more readily understood by subjects.

For those subjects who indicated that they wanted to or would have wanted to change the distribution, they were taken to another screen in which the could indicate exactly how they wanted to/would have wanted to change the distribution. On this screen they would use an interactive bar chart (as in Figure 3.2) to submit their choice. Before submitting they would be shown their choice and asked to confirm that this was indeed what they had selected.

After providing their answers, the next section showed subjects all four of the possible unique distributions (so the one they had originally been shown and the three others) and asked them a Yes/No question as to whether they regarded each separate distribution as ‘fair’. This was an important part of the experiment as it could be used to establish if there was a difference in the stated moral attitudes towards the distributions based on what treatment group participants were assigned to.

(16)

Finally, in order to explore the explanatory power of demographic characteristics, subjects were asked at the end of the survey to provide information pertaining to their age, gender and where they generally viewed themselves on left-right political spectrum. They were also asked to what extent they agreed/disagreed with two statements the first being “A society should aim to equalise the incomes of its citizens as much as possible” and the second being “Generally, people are only concerned with doing what’s in their own self-interest”. Again, this was done to explore the correlation between the subjects’ stated beliefs and the final distribution that they settled on.

3.2 Methodological Choices

3.2.1 Use of Context-free Environment

Many papers in the field seek to test the factors we are focusing on by providing a particular context behind the experimental questioning or environment (Kay et al., 2009), in this experiment we have opted not to provide such framing. This was a deliberate choice and the primary motivation was seeking to see if subjects attitudes (especially their moral considerations) and decisions were significantly different be-tween treatment groups in such context-free conditions as we believed this could expose the strength of various biases.

3.2.2 Recruitment of Subjects & Condition Group Sizes

As has been stated previously, the first stage data was collected solely for the pur-pose of being able to assure subjects that they were observing the choice of a real human participant when they looked at the distribution given. For that reason data was collected from this group until it was felt that there were enough differing dis-tributions of thee72 chosen in order to be better able to gauge subjects’ sensitivity to this in the second stage. After eight participants participated in stage one we had four unique distributions, which were fortuitously each e12 apart in terms of

Table 3.1: Experimental Design

Subject’s Role Stake Available Choice Payable Distribution

B-C Partner (Y) Inside Distribution Payoff Relevant Subject’s Choice B-NC Partner (Y) Inside Distribution Hypothetical Dictator’s (X) Choice

NB-C Witness (W) None Payoff Relevant Subject’s Choice

(17)

the inequality between what X/Y received and what Z received. Participants at this stage were recruited on the University of Amsterdam campus and an experimenter was present (but not supervising, to allow privacy) at all times during their partici-pation. They were also informed, to ensure that there was no deception, that there was a possibility that the final distribution could be different to that which they had selected.

In the second stage, the desire was to collect between 15 to 25 subjects for each Condition in order to have an adequate sample size with which to analyse. In the end 79 observations were collected in total: B-C (18 subjects.), B-NC (21 subjects), NB-C (22 subjects), and NB-NC (18 subjects). It is important to note that we believed subjects would be sensitive to subjects having their behaviour observed by an experimenter (Bateson, Nettle, & Roberts, 2006) so, after observing trial behaviour, it was decided to distribute the experiment online with the caveat that the participants should not know the experimenter personally. To achieve this the experiment was distributed by other students at UvA via their own social media platforms. As the experiment was online, at the end of the instructions page it was made clear to subjects that if they did not understand the basic instructions of the experiment – such as what their role was and the choice architecture imposed on the dictator – that they should leave the experiment uncompleted (and thus unrecorded)7.

3.2.3 Payment

Subjects were informed that one of the groups would be selected and that the fi-nal distribution for the group would be paid out to its participants. For the non-consequential conditions (B-NC & NB-NC) that meant that the initial distribution picked by Person X (the dictator) would be paid out; however, for the consequential conditions (B-C & NB-C) this could be different if the participant chose to change the distribution, this would be the amount paid - this was made clear to them during the experiment (see Appendix II). Also,e72 was selected as the total group endow-ment as it was viewed as being a significant enough amount of money to override other social preferences8_.

7_{On completion we realised that the use of control questions may have been more effective in}

confirming comprehension - see Limitations in this paper for more detail.

8_{In a preliminary trial version of this experiment the total group endowment to be distributed}

by Person X was e18. What quickly became apparent was that the difference between what a dictator receives choosing a self-interested decision (e9, e9, e0 ) and an equitable decision ( e6, e6, e6 ) was only e3, which was judged to be too low to make the choices truly dominant. The total group endowment was therefore increased toe72, which meant that the difference between

(18)

3.2.4 Use of Participant Numbers

One further methodological choice which was viewed to be extremely important, was ensuring that subjects could have confidence that real participants would be fulfilling the other roles they could see in the experiment. Frohlich, Oppenheimer, and Moore (2001) have spoken about how anonymity can raise doubts about the authenticity of pairings in the subjects mind and thus affect their behaviour. Aware of this, we wanted to prevent subjects from being tempted to behave differently due to the belief that their decisions would not impact other human-players in any way. In order to reduce this effect we believed the use of assigning participant numbers to each subject at the beginning of the experiment and then pairing them up with other players who were represented by their participant numbers within the dictator game. Our hope was that this would increase the credibility of the experiment, as opposed to simply referring to all participants by the their role title (e.g Role X, Y...).

3.3 Main Hypotheses

Hypothesis One: Simply benefiting will lower rates of deciding to change the status quo

In consequential treatments self-interest dictates that those that benefit (B-C) will have a lower incentive to reduce inequalities when faced with them than those who don’t (NB-C) as subjects in B-C are hurt financially by adjusting the distribution in a way that reduces the inequality whereas this is not the case for those in NB-C. However, following on from the literature by Dana, Weber and Kuang (2007), we believe the fact that an inequality has been selected by another player for these subjects will make it more acceptable than would otherwise be the case. We believe that the frequency and size of selected (or accepted) inequalities in these groups (B-C & B-NC) will be the result of opting for the status quo more frequently than in the non-benefit groups (NB-C & NB-NC), or that by changing the distribution they will reduce the inquality by a lesser degree (Dhingra et al., 2012).

Hypothesis Two: Consequences matter more when you benefit

Following on from Hypothesis One, subjects in B-C have far more of an incentive to either leave an inequitable outcome or change it so that an inequality still exists than those in B-NC, despite the fact that both groups benefit. Either by a desire to

(19)

be seen to abide by the observed social norm to share, or to protect their own self-image (Babcock & Loewenstein, 1997). Subjects in this treatment group will likely indicate that they would have changed the distribution towards a more equitable one far more often than those that were actually realised in B-C. This would be in accordance with the research by Hoffman et al. (1994). Following on from this, as self-interest and protection of self-image are not factors at all between treatments NB-C and NB-NC, we believe there will be no significant difference between these treatments.

Hypothesis Three: Fairness considerations could change depending on treatment group

Finally, we believe there is a possibility that those subjects who have been assigned to treatment groups where they benefit and where an inequality exists (which in this game is further to their benefit) will be more likely to characterise the inequitable distributions as fair given that it is a system which they themselves benefit from (Kay et al., 2009; Jost et al., 2004). However, it is possible that the context-free environment of the experiment will prevent this moral alteration to occur as subjects may find their is limited subjective information available to justify the choice (Rodriguez-Lara & Moreno-Garrido, 2012).

(20)

4 Results

Throughout this section we will make use of the term difference which we have de-fined as the inequality between what X/Y received individually and what Z received. So in a distribution such as the followinge25 (X), e25 (Y), e22 (Z) the difference would be e3. In order to test our hypotheses two variables were created Difference Shown (which is the distribution inequality present within the distribution they are shown made by Player X at the beginning of the experiment) and Final Difference (which is the inequality that exists by the end of the experiment). If subjects did not opt to change the distribution – either hypothetically or consequentially – this would be set according to the original decision of the dictator for their group, otherwise this would set according to the change they made to the distribution9_.

4.1 Descriptive Statistics

4.1.1 Population Breakdown & Treatment Randomisation Check

Focusing solely on participants in Stage Two (i.e. those assigned to the roles of Y and W) 79 participants completed the survey with all those who didn’t being excluded from further analysis. Of those 79, 43 were women (54.43%) and 36 men (45.57%); in terms of age, 70 participants were between the age range of 19-29 (88.61%), 6 were between 30-39 (7.59%) and 3 between 50-59 (3.8%).

Regarding politically orientation, only 11 (13.92%) participants stated that their politically views were on the right side of the political spectrum, with all of these 11 stating that they were on the ‘Centre-right’ as opposed to ‘Right’; 11 (13.92%) placed themselves in the ‘Centre’, 37 (46.84%) placed themselves on the ‘Centre-left’, and 20 (25.32%) placed themselves on the ‘Left’.

Between treatment groups the subjects did not differ significantly by age [F(3,75) =

0.18, p = 0.9119], gender (χ2

(3) = 5.2056, p = 0.157), or political orientation

[F(3,75) = 0.12, p = 0.9510], suggesting that the comparisons between them are

statistically valid.

4.1.2 Summary Statistics

In the first stage we collected data until we had enough unique distributions to be able to show subjects in the second stage of the experiment. After collecting data

9_{Please note that final difference does not constitute the distribution that would be paid, as}

those assigned to non-consequential treatments would not be able to change the payable distribution (which would be Player X’s initial choice).

(21)

from eight participants we had four unique distributions which were well distributed. Five dictators (62.5%) chose to split the money equally (so a difference ofe0), with one dictator each (12.5%) of the remaining three selecting the following:

• e28 (X), e28 (Y), e16 (Z) - Difference of e12 • e32 (X), e32 (Y), e8 (Z) - Difference of e24 • e36 (X), e36 (Y), e0 (Z) - Difference of e0

In the second stage we attempted to ensure that each of the above unique distri-butions were displayed to subjects by treatment group evenly. The mean values of what final difference subjects agreed to by both the difference they were shown and their treatment group is displayed in Figure 4.1. As can been seen, all participants who were shown a difference of e0 (complete equality) did not opt to change the distribution. However as the inequalities that were displayed to subjects increased the final difference was generally higher in groups that benefited (B-C and B-NC) than those who did not. Aggregating the data further – only focusing on treatment groups – the Final Difference mean and standard deviation statistics show that

(22)

Figure 4.2: Differences Between X/Y and Z Subjects Believed Fair

benefiting appears to increase the inequality that people are willing to accept: B-C [M = 11, SD = 13.69], B-NC [M = 13, SD = 14.92], NB-C [M = 4.23, SD = 10.42], and NB-NC [M = 4, SD = 11.64].

By treatment group the number of subjects who chose to change the distribution was 6/12 of B-C (%33), 10/21 of B-NC (48%), 14/22 of NB-C (64%), and 12/18 of NB-NC (67%). This seems to indicate that benefiting reduces the desire of subjects to change the distribution and may help to explain why inequalities are higher in those treatment groups where subjects benefit.

Regarding what participants regarded as fair, as you can see from Figure 4.2, the majority of all treatment groups regarded the equal treatment (where the difference equalled e0) as fair, in fact only one participant (who was in NB-C) regarded it as unfair. At levels that were inequitable, interestingly B-C & NB-C were the two groups that viewed inequities as fairer, but the trend was negative for all groups except Treatment I as the inequity was introduced and got bigger.

4.2 Analysis & Hypotheses

4.2.1 Hypothesis One

Looking at the final difference settled on by groups and performing two-sample t-tests for differences in means between B-C and NB-C the difference is positive [M D = 6.77] with those benefiting settling for higher inequalities on average, but this difference is not significant at the 5% confidence level [t(31) = 1.73, p > 0.05].

Comparing between B-NC and NB-NC the difference is again positive [M D = 9] but is, in this case, significant [t(2) = 2.11, p < 0.05].

(23)

Table 4.1: OLS on Final Difference

(1) (2) (3)

Final Difference Final Difference Final Difference

Benefits 8.788∗ 8.983∗ 9.096∗

(2.24) (2.28) (2.28)

Consequential Choice 0.723 0.720 0.992

(0.19) (0.18) (0.24)

Benefits × Consequential Choice -2.219 -2.183 -2.343

(-0.41) (-0.40) (-0.42) Difference Shown 0.356∗∗∗ 0.364∗∗∗ 0.374∗∗∗ (3.60) (3.66) (3.71) Female 0.197 0.563 0.770 (0.07) (0.20) (0.26) Age 0.194 0.188 (0.97) (0.93)

Political Orientation (Right to Left) -1.288 -0.944

(-0.91) (-0.63)

S1: Equalise Agree -1.035

(-0.94)

S2: People Selfish Agree -0.367

(-0.36) Constant -2.804 -7.119 -6.715 (-0.69) (-1.05) (-0.90) Observations 79 79 79 t statistics in parentheses ∗ _{p < 0.05,}∗∗ _{p < 0.01,}∗∗∗ _{p < 0.001}

our dependent variable and regressed demographic factors and experimental factors (such as the difference they were exposed to) on it. Table 4.1 shows the results of this regression and we can see in this regression that simply benefiting did indeed have a significant effect across all three of our models [t ≥ 2.24, p < 0.05] on what subjects did or would have settled on either via action or inaction.

(24)

Exploring further we performed a logit regression looking at what factors affected the probability of changing the status quo (in our experiment the dictator’s choice). Table 4.2 uses as two dependent variables consequential choice and hypothetical choice to show that benefiting appeared to significantly reduce the likelihood of changing the dictator’s decision if the decision was consequential [z = −2.13, p < 0.05], and also appeared to reduce the probability of indicating that you would change the distribution when the decision was hypothetical - although this was not significant [z = −1.33, p > 0.05].

Finally as an upper and lower limit was put on the amount that could be assigned to the X/Y players, a Tobit regression model was also run taking into account these limitations which again showed that benefiting significantly increased the final distributions inequality [t ≥ 2.22, p < 0.05](see Appendix I).

Therefore we state that there is evidence to support Hypothesis One: that simply benefiting negatively affects our desire to change the status quo, or indeed indicate that we would, if possible.

4.2.2 Hypothesis Two

To test our second hypothesis we need to compare between B-C and B-NC, if our hypothesis is true B-C will have a significantly greater final difference than B-NC. Performing a two-sample t-test we can see that their is no significant difference between the two groups [t(37) = −0.44, p > 0.05] and that the difference in means

is actually negative [M D = −2] suggesting that those subjects in B-NC actually accepted higher inequalities. As we predicted, between NB-C and NB-NC there was also no significant effect when making a mean comparison [t(35) = 0.06, p >

0.05, M D = 0.23].

This is also reflected in our regression analysis as there appears to be no signifi-cant effect of either having had a consequential opportunity to change the distribu-tion [t ≤ 0.24, p > 0.05], or indeed the specific interacdistribu-tion of having both benefited whilst also being offered such an opportunity [ | t | ≤ 0.42, p > 0.05]. Again, our results actually indicated that those who were assigned to B-C were more likely to settle on less inequitable distributions than those in B-NC (although this effect was not significant).

These results are obviously not in line with our hypothesis and existing literature such as those found by (Babcock & Loewenstein, 1997). We therefore have no evidence to support Hypothesis Two and it is thus rejected, consequences did not seem to significantly affect the final distribution which subjects settled on when they benefited.

(25)

Table 4.2: Logistic Regression: What Affects Status Quo Bias?

(1) (2)

Opted to change Would change

Benefits -1.796∗ -1.117 (-2.13) (-1.33) Difference Shown 0.0937∗∗ 0.0984∗∗ (2.86) (2.90) Age -0.0231 0.0103 (-0.51) (0.18) Female 0.513 0.221 (0.66) (0.27) Constant -0.502 -1.297 (-0.35) (-0.65) Observations 40 39 χ2 _14.91 _13.15 z statistics in parentheses ∗ _{p < 0.05,}∗∗ _{p < 0.01,}∗∗∗ _{p < 0.001} 4.2.3 Hypothesis Three

Finally we created an additional dependent variable from the subjects’ answers as to what distributions they regarded as fair reflecting the maximum distribution difference they stated was fair10_{. So if the maximum inequality a subject regarded}

as acceptable was e12 this would be the value used.

Using this measurement we can see keeping the benefit state constant11 _{we can}

see no statistical difference [t(33)= 0.99, p > 0.05] between B-C [M = 4, SD = 5.82]

and B-NC as to what they regard as fair [M = 2.29, SD = 4.83]. There is however a significant difference [t(23) = 2.56, p < 0.05] between NB-C [M = 8.73, SD =

14.43]and NB-NC [M = 0.67, SD = 2.83] with those in NB-C regarding larger

10_{As there were only two observations in our sample where the fairness considerations of these}

subjects were not linearly consistent (i.e. they regarded total equality fair and the maximisation of X/Y payoffs at the expense of Z but nothing in between) we decided to drop these observations from this stage of the analysis.

(26)

inequalities far more acceptable than those in NB-NC [M D = 8.06]. Keeping the consequential state constant we can find no significant difference between B-NC and NB-NC [t(33) = 1.30, p > 0.05, M D = 1.62] or between B-C and NB-C [t(29) =

−1.40, p > 0.05, M D = −4.73].

We can find no reason within the existing literature why the only significant difference exists between NB-C and NB-NC and note that all other comparisons within the 2x2 design do not produce significant results. That being said, tentatively, we can neither reject or accept Hypothesis Three that moral attitudes were altered by the conditions imposed on their treatment group12.

4.2.4 Other Findings

There did not appear to be any significant difference over what values men and women chose to settle on (see Table 4.1), like those seen in other papers (Eckel & Grossman, 1998). Political orientation, Age and the extent to which they agreed with the two statements that they were shown also appeared to have no explanatory power.

12_{Due to an oversight in our experimental design our analysis in this section is further hindered}

(27)

5 Discussion

5.1 Implications

In our experiment simply benefiting seemed to increase subjects willingness to agree to an unfair outcome, but primarily through inertia. Perhaps most surprising is that subjects also appeared to express less of a willingness to reduce an inequality when they were asked whether they would hypothetically change the distribution without it having any impact on the actual amount payable. Evidence from the field seems to suggest that more people act in their own self-interest against social norms than otherwise would if they feel like they are not being observed (Beaman, Klentz, Diener, & Svanum, 1979; Bateson et al., 2006; Hoffman et al., 1994), but all of these experiments require subjects to act against the social norm in order to maximise their own payout. Although subjects were also not being observed in our experiment this latter condition was not imposed on subjects in the particular conditional treatment we are referring to (B-NC).

If we were expand these results to situations outside of the experimental environ-ment it would suggest that people may be more likely to tacitly justify institutional economic frameworks (such as pay scales within a company, terms of economic trade or certain norms surrounding property) solely because they benefit from them de-spite the fact that they do not require their support. This has obvious implications to the field of institutional economics for example as it may explain how flawed insti-tutional practices or norms, which could harm both productivity and social mobility (as examples), may maintain support or, at the very least, face little opposition from relevant populations.

However we also ought to point out another finding from our paper, which was that there was no evidence to suggest that the moral judgements of subjects differed based on the two alternating factors of whether they benefited or whether they could meaningfully change the status-quo. In our experiment we provided a simplified and relatively context-free environment as we believed that should there be any biases regarding such moral judgements revealed within our results, that this could potentially expose the strength of such biases. This provides some indication that perhaps even if individuals were willing to act in their own self-interest within the context of our experiment that they were still shared generally the same social norms as those who did not benefit.

(28)

5.2 Methodological Limitations

One of the key flaws of the experimental design employed in this experiment was not to randomise when the questions were asked surrounding what distribution levels participants regarded as ‘fair’ . In our experiment, this was only asked after the participants had been exposed to the distribution choices of X players to see whether the treatment group that had been assigned to had affected their moral perspective. However, as the questions were only asked at the end of the experiment it would prove difficult to prove the direction of any causal relationship. Do our fairness considerations inform what we agree to settle on in this experimental setting, or vice versa? One point worth noting, however, is that in our experiment we did not find any significant difference between the treatment groups, so it is difficult to see how asking the questions before would have had impacted the answers regarding fairness. Even so, regrettably, we should have sought to confirm those intuitions.

Secondly, the experimental design of the dictator game has been called into ques-tion by some who doubt the external validity of conclusions derived from results which have utilised it. Frohlich et al. (2001) have brought attention to the fact that the attitude of subjects towards the dictator game may be, in fact, to view it simply as a game, and thus act in a way that would be inconsistent with how they would approach distribution decisions in their everyday life. We agree with the authors that this is indeed a possibility and have purposefully sought to restrain the conclu-sions we have arrived at using our results for that reason. This phenomenon could also explain why subjects in treatment group B-NC did not behave as expected. That being said, we wish to highlight that however the experiment may have been perceived by subjects its essential elements were described to the subjects in the same manner and therefore we believe that a good deal of interpretation between the treatments is still valid.

As is often the case with university experiments, the external validity of the experiment is effected somewhat by the limited demographic of the subject pool. While, in this experiment, there was a well mixed population with regards to gen-der, the population was skewed to the left when it came to self identified political orientation of participants with no subjects identifying themselves as being on the ‘right’ of the political spectrum. On age, the vast majority of subjects were in the age range 18-30.

The decision to distribute the experiment online also had the disadvantage of not allowing us to control the experimental environment were in when subjects took the experiment and to check for subjects comprehensive understanding of the

(29)

experiment. Potentially the use of control questions could have been utilised before the experiment to check understanding. This was not opted for in this experiment with the view of making the experiment as quick for participants to undertake so as to increase participation, but in order to have greater confidence in the results this certainly should be employed in the future.

5.3 Recommendations for Future Research

Self-serving bias has been shown to be present in situations where there are factors which can be more readily interpreted in a subjective manner, such as in those found by Babcock and Loewenstein (1997). In our experiment the endowment had been given to the group without any effort being exerted by anyone within it and could therefore have been viewed as a ‘gift’ to the group – which may have made inequalities harder to justify (at least by the moral standards of what is fair). Frey and Bohnet (1995), for example, have stated that the introduction of institutional factors such as property rights can alter the fairness norm for participants. Therefore any future research could seek to introduce income effects such as having the group endowment earned primarily or totally by the dictator but seek to make use of the 2x2 design employed in this experiment. Rodriguez-Lara and Moreno-Garrido (2012) have partly already done this, but they only observed the behaviour of dictators who could meaningfully set the distribution as opposed to change that set by somebody else (meaningfully or otherwise).

In the real world we rarely set the conditions and norms which we abide by every day but largely follow those set by others before us. It is for that reason that we believe that experimental economists should follow the lead of those within the field of political psychology and continue to seek to better understand our adherence to existing economic frameworks.

(30)

References

Andreoni, J., & Sanchez, A. (2014). Do beliefs justify actions or do actions jus-tify beliefs? an experiment on stated beliefs, revealed beliefs, and social-image manipulation (Tech. Rep.). National Bureau of Economic Research.

Babcock, L., & Loewenstein, G. (1997). Explaining bargaining impasse: The role of self-serving biases. Journal of Economic perspectives, 11 (1), 109–126. Bateson, M., Nettle, D., & Roberts, G. (2006). Cues of being watched enhance

cooperation in a real-world setting. Biology letters, 2 (3), 412–414.

Beaman, A. L., Klentz, B., Diener, E., & Svanum, S. (1979). Self-awareness and transgression in children: Two field studies. Journal of Personality and Social Psychology, 37 (10), 1835.

Bohnet, I. (1999). The sound of silence in prisoners dilemma and dictator games. In Economics as a science of human behaviour (pp. 177–194). Springer. Bolton, G. E., & Ockenfels, A. (2000). Erc: A theory of equity, reciprocity, and

competition. American economic review , 90 (1), 166–193.

Boxall, P., Adamowicz, W. L., & Moon, A. (2009). Complexity in choice ex-periments: choice of the status quo alternative and implications for welfare measurement. Australian Journal of Agricultural and Resource Economics, 53 (4), 503–519.

Bra˜nas-Garza, P. (2006). Poverty in dictator games: Awakening solidarity. Journal of Economic Behavior & Organization, 60 (3), 306–320.

Burnham, T. C. (2003). Engineering altruism: A theoretical and experimental investigation of anonymity and gift giving. Journal of Economic Behavior & Organization, 50 (1), 133–144.

Camerer, C. F., & Fehr, E. (2004). Measuring social norms and preferences us-ing experimental games: A guide for social scientists. Foundations of human sociality: Economic experiments and ethnographic evidence from fifteen small-scale societies, 97 , 55–95.

Cassar, L., & Klein, A. (2016). A matter of perspective: How experience shapes preferences for redistribution.

(CESifo Working Paper No.6302)

Dana, J., Weber, R. A., & Kuang, J. X. (2007). Exploiting moral wiggle room: ex-periments demonstrating an illusory preference for fairness. Economic Theory, 33 (1), 67–80.

Dhingra, N., Gorn, Z., Kener, A., & Dana, J. (2012). The default pull: An exper-imental demonstration of subtle default effects on preferences. Judgment and

(31)

Decision Making, 7 (1), 69.

Eckel, C. C., & Grossman, P. J. (1998). Are women less selfish than men?: Evidence from dictator experiments. The economic journal , 108 (448), 726–735.

Engelmann, D., & Strobel, M. (2004). Inequality aversion, efficiency, and maximin preferences in simple distribution experiments. American economic review , 94 (4), 857–869.

Fehr, E., & Schmidt, K. M. (1999). A theory of fairness, competition, and cooper-ation. The quarterly journal of economics, 114 (3), 817–868.

Fong, C. (2001). Social preferences, self-interest, and the demand for redistribution. Journal of Public economics, 82 (2), 225–246.

Forsythe, R., Horowitz, J. L., Savin, N. E., & Sefton, M. (1994). Fairness in simple bargaining experiments. Games and Economic behavior , 6 (3), 347–369. Frey, B. S., & Bohnet, I. (1995). Institutions affect fairness: Experimental

investiga-tions. Journal of Institutional and Theoretical Economics (JITE)/Zeitschrift f¨ur die gesamte Staatswissenschaft , 286–303.

Frohlich, N., Oppenheimer, J., & Moore, J. B. (2001). Some doubts about measuring self-interest using dictator experiments: the costs of anonymity. Journal of Economic Behavior & Organization, 46 (3), 271–290.

Hoffman, E., McCabe, K., Shachat, K., & Smith, V. (1994). Preferences, property rights, and anonymity in bargaining games. Games and Economic behavior , 7 (3), 346–380.

Hoffman, E., McCabe, K., & Smith, V. L. (1996). Social distance and other-regarding behavior in dictator games. The American economic review , 86 (3), 653–660.

Jost, J. T., Banaji, M. R., & Nosek, B. A. (2004). A decade of system justification theory: Accumulated evidence of conscious and unconscious bolstering of the status quo. Political psychology, 25 (6), 881–919.

Kahneman, D., Knetsch, J. L., & Thaler, R. H. (1991). Anomalies: The endowment effect, loss aversion, and status quo bias. Journal of Economic perspectives, 5 (1), 193–206.

Kay, A. C., Gaucher, D., Peach, J. M., Laurin, K., Friesen, J., Zanna, M. P., & Spencer, S. J. (2009). Inequality, discrimination, and the power of the status quo: Direct evidence for a motivation to see the way things are as the way they should be. Journal of personality and social psychology, 97 (3), 421. Li, D., Hawley, Z., & Schnier, K. (2013). Increasing organ donation via changes

in the default choice or allocation rule. Journal of health economics, 32 (6), 1117–1129.

(32)

Loewenstein, G. F., Thompson, L., & Bazerman, M. H. (1989). Social utility and decision making in interpersonal contexts. Journal of Personality and Social psychology, 57 (3), 426.

McCabe, K. A., Rigdon, M. L., & Smith, V. L. (2003). Positive reciprocity and intentions in trust games. Journal of Economic Behavior & Organization, 52 (2), 267–275.

Rodriguez-Lara, I., & Moreno-Garrido, L. (2012). Self-interest and fairness: self-serving choices of justice principles. Experimental Economics, 15 (1), 158–175. Samuelson, W., & Zeckhauser, R. (1988). Status quo bias in decision making.

Journal of risk and uncertainty, 1 (1), 7–59.

Sherrill, M. (2007). Self-serving bias. Am Psychol , 41 , 954–969.

Thaler, R. H. (1988). Anomalies: The ultimatum game. Journal of Economic Perspectives, 2 (4), 195–206.

Thaler, R. H., & Benartzi, S. (2004). Save more tomorrow: Using behavioral economics to increase employee saving. Journal of political Economy, 112 (S1), S164–S187.

Zizzo, D. J. (2010). Experimenter demand effects in economic experiments. Exper-imental Economics, 13 (1), 75–98.

(33)

APPENDIX I - Tobit Regression

Table 5.1: Tobit Regression on Final Difference

(1) (2) (3)

Final Difference Final Difference Final Difference

Benefits 47.81∗ 46.78∗ 46.48∗

(2.23) (2.22) (2.23)

Consequential Choice 11.29 9.645 9.432

(0.54) (0.47) (0.45)

Benefits × Consequential Choice -13.53 -11.99 -11.41

(-0.52) (-0.47) (-0.44) Difference Shown 1.653∗∗ 1.642∗∗ 1.656∗∗ (2.87) (2.88) (2.93) Female 1.047 1.849 2.734 (0.08) (0.15) (0.22) Age 0.598 0.604 (0.60) (0.61)

Political Orientation (Right to Left) -1.678 -0.0136

(-0.27) (-0.00)

S1: Equalise Agree -3.848

(-0.80)

S2: People Selfish Agree -0.797

(-0.18) Constant -79.88∗∗ -92.39∗ -91.21∗ (-2.74) (-2.32) (-2.21) var(e.finaldiff) 1668.8∗ 1638.8∗ 1586.9∗ (2.27) (2.27) (2.27) Observations 79 79 79 t statistics in parentheses,∗ p < 0.05,∗∗ p < 0.01,∗∗∗ p < 0.001

(34)

APPENDIX II - Experiment Scripts

Please note, fixed participant numbers have been used in the scripts below - this is just to illustrate how the experiment would appear to the participants. In the actual experiment the participant numbers would be generated and assigned to participants.

II.I - Stage One: Person X Script

————— =⇒ start of experiment ⇐= —————

Thank you for taking the time to take part in this study.

The study consists of two stages, but your input is only required for the first stage. In total this should only take 3-5 minutes of your time.

During the study it is necessary that you do not interact with anyone around you except the experimenter, who will be available should you have any questions.

On the next page the details of the study will be explained to you as well as the conditions of who will be paid.

IMPORTANT

• You can have complete confidence that the decisions in this quiz will be kept private and the findings will be entirely anonymised

• You can only participate in this study once, so if you know or suspect that you have already participated in this study please do not continue

When you are ready to begin, click “Next page”

(35)

Instructions

Please read these instructions carefully

This is a simple group experiment in which you will be paired with two other human participants in which one of you will be asked to distribute e72 between the three of you. The three roles are as follows:

• Person X: This individual will choose how to distributee72

• Person Y: This individual will receive the same amount as Person X in any given distribution

• Person Z: This individual will receive the remaining money (i.e. e72 minus the amounts that Person X and Person Y receive)

Restrictions on Person X’s choice:

The maximum that Person X can choose for themselves is e36 as Person Y must receive the same as them and the total distribution across all subjects must equal e72. With the minimum that Person X can choose for themselves being e24.

Finally, Person X’s allocation has to be an integer (i.e. e27 or e30 not e27.50 or e30.50)

(36)

Payment:

After all data from both stages has been collected, one group will be chosen at ran-dom to be paid out and contacted by email to arrange payment.

Please be aware that in SOME cases the distributions may be changed in Stage 2 -but the rules of how the money can be distri-buted will remain the same.

On the next page you will be given your role and a candidate number.

————— =⇒ new section ⇐= —————

You have been assigned to the role of: Person X

Your participant number is: X124

On the next page, you will be asked to choose the distribution. You can do this by clicking on the top of the bar and dragging up or down (as shown below)

(37)

If you have any questions or do not understand the instructions, ask the experi-menter now.

Once you are ready click next.

————— =⇒ new section ⇐= —————

As you are Person X, you have been chosen to select the

dis-tribution.

Use the graph below to decide how you would like to distribute thee72

Remember: to change the distribution, click on the top of the bar to change the value dynamically

(38)

X124 (You):

e

Y567:

e

Z922:

e

————— =⇒ new section ⇐= —————

You have selected the following distribution:

• X124(You):

e(output)

• Y567:

e(output)

• Z922:

e(output)

Is that correct? Yes No

N.B. - If participant selects ’No’ they are taken back to the section where they choose distribution.

————— =⇒ new section ⇐= —————

Thank you. On the next page you will be asked for some basic demographic infor-mation.

————— =⇒ new section ⇐= —————

Email Address:

It’s important this is correct so you can be contacted if your group’s distribution is selected to be paid

(39)

Age: * Gender: Female Male Other

Prefer not to say

In political matters, people talk of the left and the right. How would you place your views on this scale, generally speaking?

Left

Centre-left

Centre

Centre-right

Right

To what extent do you agree with the following statements:

“A society should aim to equalise the incomes of its citizens as much as possible”

Strongly agree

(40)

Somewhat agree

Neither agree nor disagree

Somewhat disagree

Disagree

Strongly disagree

“Generally, people are only concerned with doing what’s in their own self-interest”

Strongly agree

Agree

Somewhat agree

Neither agree nor disagree

Somewhat disagree

Disagree

Strongly disagree

(41)

II.II - Stage Two: Person Y & W Script

As there were multiple treatment groups with different roles and possible actions (i.e. consequential/non-consequential) the instructions needed to differ slightly. This dif-ference in script between treatment groups is indicated by separating columns below.

————— =⇒ start of experiment ⇐= —————

Thank you for taking the time to take part in this study. In total this should only take 3-5 minutes of your time.

On the next page the details of the study will be explained to you as well as the conditions of who will be paid.

IMPORTANT

• You can have complete confidence that the decisions in this quiz will be kept private and the findings will be entirely anonymised

• You can only participate in this study once, so if you know or suspect that you have already participated in this study please do not continue

Doing nothing is all you've ever wanted to do : does simply benefiting affect not only our desire to change the status quo... but how we view it too?