University of Amsterdam
Faculty of Economics and Business Roetersstraat 11
1018 WB Amsterdam The Netherlands
Telephone: +31 20 525 5387 http://www.uva.nl/en/contact/
Title: The price elasticity of charitable giving: the role of wilful ignorance
Author: Stefana Oros (10621555)
Assignment: Master’s thesis
Supervisor: dr. J.J. (Joël) van der Weele
MSc Program: Economics
Specialization: Behavioural Economics and Game Theory
Number of ECTS: 15 ECTS
The price elasticity of charitable giving:
the role of wilful ignorance
Master’s Thesis Stefana Oros August 2017
This paper investigates the price elasticity of charitable giving with and without the opportunity for self-deception. Participants can choose to avoid information about charity characteristics which may morally urge them to behave more generously. The effects of the cost of giving and wilful ignorance are addressed individually by previous research, however this study has the purpose to analyse the prosocial consequences of the interaction of these factors. The econometric analysis provides strong evidence that the characteristics of a charity highly impact donor’s response and that the desire to conform to social norms determines subjects to donate more, as well as to search for justifications for more selfish behaviour. However, the increase in the cost of giving does not appear to influence giving, when it coincides with an increase in income, and the opportunity for self-deception does not seemingly impact the price elasticity of giving.
Statement of Originality
This document is written by Stefana Oros who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Contents
1. Introduction ... 4
2. Related Literature ... 6
2.1. Giving – one of the most common means of prosocial behaviour ... 7
2.2. The cost of giving – helping others versus helping oneself ... 9
2.3. Self-deception – finding excuses for giving less ... 10
3. Methodology ... 13
3.1. Data set and experimental design ... 13
3.2. Hypotheses and econometric procedures ... 19
4. Results and Analysis ... 26
4.1. Charity evaluation ... 26
4.2. Overview of charitable donations data by experimental groups ... 28
4.3. Mean difference tests results - Results of the hypotheses tests ... 33
4.3.1. The influence of perceived differences in charity characteristics ... 33
on charitable giving ... 33
4.3.2. The impact of the cost of giving on average donations ... 35
4.3.3. Self-deception and information selection: Differences in generosity between treatment and control groups ... 36
4.3.4. Ignorance rate and self-deception results ... 38
4.3.5. Price elasticity of giving with and without the opportunity for self-deception: Difference-in-difference estimator - Regression results ... 40
5. Discussion ... 43
6. Conclusions ... 45
Appendix ... 48
References ... 50
1. Introduction
Charitable giving is one of the most common forms of prosocial behaviour. Philanthropic activities do not provide any personal payoff, but request sacrificing a share of own welfare in order to promote and aid a good cause. However, several factors influence donors’ behaviour and motivation, which constitutes a relevant aspect because the largest share of charity funds are obtained from private contributions.
It is argued that philanthropists are motivated to give by the outcome their donations accomplish (“making a difference”, Duncan, 2004), the positive feeling they experience from giving (the warm-glow effect, Crumpler & Grossman, 2008), and the desire to act according to social norms (Andreoni & Bernheim, 2009). Therefore, charitable giving might not only be prompted by altruism and sincere care for the less fortunate ones. Specifically, studies suggest that the trade-off between personal gain and improving social welfare is an ongoing process for the decision maker, and that prosocial decisions might often be the outcome of the pressure of completing a moral duty (Gebauer, Riketta, Broemer, & Maio, 2008).
Moreover, the main factors interposing charitable giving are: the cost of giving, that is the level of income forgone to benefit the charity, personal wealth, charity characteristics, such as notoriety, the personal preference or affinity for particular charitable activities (Breeze, 2013), and the urgency and emotional impact of the cause promoted by the charity. Therefore, conforming to social norms by making contributions can be the result of the desire to maintain a positive self-image, such as reputation concerns, and aversion to negative feelings like guilt or remorse (Haley & Fessler, 2005).
Consequently, when given the opportunity, individuals may exploit ambiguous or imprecise contextual situations in order to justify more selfish decisions. Self-deception provides opportunity for self-interested behaviour, while upholding the false belief that such actions are based on solid moral principles (Tenbrunsel & Messick, 2004). Among other contexts, self-deception can occur when individuals avoid obtaining unwanted information that would compel them to act against their wish, and use this lack of knowledge to defend unfair actions (Sweeny, Menyk, Miller, & Shepperd, 2010).
Moreover, as mentioned before, charitable giving comes at a personal cost. This cost of giving, measured in the personal welfare sacrificed to help the charity, can negatively impact the incentives to give and the ultimate amounts donated. For instance, the price of giving can represent the share of the donation used by the charity to cover administrative and operational
costs, money which does not help the cause directly. Therefore, the higher the price of giving, the lower the ability but also the desire to contribute.
All things considered, although there exists a vast literature addressing both the topics of price elasticity of giving and self-deception, there is limited evidence on the effects of the interaction of these two factors on charitable donations. Specifically, as an increase in the price of giving would expectedly decrease donations to charity, the opportunity to avoid potentially inconvenient information might determine less generous donors to give even less than they would if they had to confront this information.
Attempting to fill this evidence gap might prove to be important, because it can provide insights on whether the price of ethical actions negatively impacts the incentives to give. That is, if the cost of giving increase determines individuals to opt more frequently to remain ignorant to information, as well as to subsequently donate less, then one can conclude that given the negative cost effect, the occurrence of self-interested behaviour rises. For instance, an increase in a well-regarded charity’s overhead costs may generally make donors less willing to contribute or determine them to donate less. These effects could occur because contributors want their donation to directly benefit the charity’s promoted cause to an as large extent as possible. Thus, the increase in the charity’s administrative expenses could be negatively perceived, because a particular donation would now directly help the cause by a lower extent than prior to the organization’s policy change.
This study aims at analysing whether this could be indeed the case, by trying to measure and compare the price elasticity of prosocial behaviour when individuals can or cannot avoid information. Therefore, the research question this paper aims to answer is:
To what extent does the cost of giving affect charitable donations when there is room for self- deception?
In the experiment conducted within this study, participants act in the role of a dictator, who can share part of his endowment with one of two charities, while the cost of donating is either high or low. In the treatment group, self-deception is introduced by giving the subjects the choice to reveal which charity benefits from their donation. Because each charity’s promoted cause significantly differs in their urgency of receiving donations, participants should reasonably have a lower incentive to donate to the less deserving charity. Therefore, more self-interested subjects might self-deceive by choosing to stay ignorant to the name of the charity which benefits from the donation, and subsequently act more selfishly.
This research contributes to the existing literature relating the determinants of charitable giving. Namely, it confirms that the charity characteristics are an important measure in donor’s evaluation regarding how much to help: positively perceived charities receive larger donations than those which are less well-regarded. Moreover, although inconclusively, the findings of this study support the arguments brought to the fore by related literature regarding the effects of self-deception on giving. Specifically, subjects often opted for ignorance regarding information concerning the identity of the recipient charity, and subsequently donated less than participants who could not avoid this information. However, apparently, the change in the cost of giving did not influence charitable donations, probably due to the fact that, given the experimental setting, the price rise coincided with an income increase. Therefore, the opportunity for self-deception does not seem to impact significantly the price elasticity of giving.
The remainder of this thesis is organized as follows: firstly, section 2 will comprise the existing literature addressing the topics which were presented in the introduction. The collected data, experimental design, and the hypotheses developed in this research together with the economic procedures employed to test them will be discussed in section 3. Section 4 will contain the results of the study obtained from the statistical analysis of the data, followed by the discussion of these outcomes in section 5. Finally, conclusions together with suggestions for future research will be formulated in section 6.
2. Related Literature
This section summarizes some of the most relevant research addressing the topics of current interest and provides an extensive outline of the elements of this study. Firstly, in sub-section 2.1., the main findings of the available literature regarding charitable giving are described and discussed, in order to reveal the principal behavioural aspects that establish this common prosocial action. Secondly, sub-section 2.2. comprises the key evaluations concerning the price of giving and its major consequences on the giving behaviour. Lastly, sub-section 2.3. presents how given such cost-related constraints, deception emerges as a form of self-interest defence, and hence modulates charitable actions.
2.1. Giving – one of the most common means of prosocial behaviour
Standard economic theory assumes that people only act in their own interest, with no regard to the welfare of others (Homo economicus), however, practice shows that individuals often act prosocially and obtain utility from providing benefits to other people (Henrich et al., 2001). Within the wide range of actions that qualify as being prosocial, the most common are giving, helping, comforting, sharing, and cooperating (Batson & Powell, 2003). Concerning giving, Adam Smith (1759) stated that people show an interest in the fortune of others, and often derive pleasure from making those in need better off. This prosocial behaviour is triggered by altruism, the wish of benefiting others rather than oneself, and compassion, the emotional impact the misery of others has on our own self (Leiberg, Klimecki, & Singe, 2011). Compassion may lie at the basis of one of the most common means of giving and helping others: charitable donations. Philanthropists engage in offering monetary donations to organizations which fight for a good cause, most usually by helping other people in need. Most charities are public, thus they can be regarded as a public good to which people can voluntarily contribute with monetary donations or with labour (Wunderink, 2002). Also, the transparency and facility of donating to charity make it for donors a favoured target for generosity. This fact is supported by Eckel and Grossman (1996) who conducted an anonymous ultimatum game experiment and found that legitimate charities received higher donations than unknown recipients.
The existing literature provides useful insights into the criteria subject to which people select which charitable organizations to support and the monetary amounts they want to contribute to the cause those charities promote. Notably, Breeze (2013) argues that decisions made by charitable donors are not entirely rational, but that they emphasize personal preferences based on the analogy with own experience or emotions. Therefore, philanthropists do not pick charities subject to the urgency of the promoted cause, but according to personal taste and developed heuristics.
Moreover, the effectiveness of a charity’s activities is regarded as a relevant, but not as the most important criterion in the choice of a charitable organization. In line with this finding, Niehaus (2014) argued that donors prefer to think that their contributions are effectively administered and that they tend to protect this belief by avoiding seeking for actual evidence or feedback in the aftermath of their donation.
Actually, the primary concern of donors is the type of charity (van Iwaarden, van der Wiele, Williams, & Moxham, 2009; Metzger & Günther, 2015). For instance, Kajackaite
(2015) observed that people are averse to giving to negatively perceived charities, and that they are even willing to forego some of their own benefit in order to ensure that undeserving charities receive no contribution. Furthermore, previous studies generally agree that the more deserving of aid a charity is mutually considered to be, the higher the donations it receives (Eckel & Grossman, 1996; Fong, 2005; Brown, Meer, & Williams, 2017). Fong (2005) found this outcome surprising because participants with a high pre-established level of unconditional altruism should have demonstrated their inequity aversion and high preference for fairness by making equal donations, independent on how deserving the recipient was.
Therefore, this available evidence implies that altruism is not the sole factor that determines donor behaviour. For instance, the paper of Lee and Chang (2007) reveals that although volunteering is indeed determined by empathy and generosity, monetary donations are caused by socioeconomic and demographic conditions, such as income, age, gender, or family status. However, their paper does not include charity characteristics, which as discussed above, are so important in shaping donor behaviour.
Foremost, numerous studies have investigated the less apparent reasons why people give to charity, and found that individuals donate either because they feel compelled to fulfil a moral duty, because they obtain personal satisfaction from the act of giving itself (warm-glow effect), or because they want to make a pertinent change and win social acclaim (Andreoni, 1990; Duncan, 2004). First, DellaVigna, List, and Malmendier (2012) proved that social pressure is a major element affecting donation choices, as people do not like to refuse making a humanitarian gesture and consequently damage their self-image. Specifically, they found that within a door-to-door fund raising activity, informing households beforehand about the time and purpose of the campaign reduces the occurrence of donations as well as the amounts contributed. Thus, fear of building a bad reputation, guilt, social exclusion, punishment, or other sorts of negative reciprocity might pressure people into acting against their instinctive will (Haley and Fessler, 2005).
However, while not having purely altruistic incentives, agents may still increase their utility when donating, by experiencing the warm-glow effect of giving (Andreoni, 1990). Crumpler and Grossman (2008) found that the gratification derived from the act of helping those in need does condition giving. Nonetheless, the warm-glow giving phenomenon has beneficial effects both on the welfare of the recipient charity, as well as on the utility of the donor. However, it discloses the behavioural aspect that rather than the concern for the welfare of the receiver alone, the somewhat selfish incentive of maintaining a favourable self-image significantly triggers giving.
All things considered, giving is not a predictable behaviour, as donors’ decisions are not always governed by rational considerations and ingenuous intentions. Giving is not at all times utility enhancing, because donors often face a trade-off between prosocial and personal interests. This aspect has several implications for policy makers and charitable organizations in establishing collaborative relationships as well as attracting potential donors, and motivating people to contribute.
2.2. The cost of giving – helping others versus helping oneself
Giving often entails incurring a cost by sacrificing a share of personal welfare in order to financially benefit another party. The costs of acting prosocially can be represented by effort, time, or money invested in helping others than oneself. More generally, Meer (2014) defines the cost of giving as the amount a donor has to forgo in order to increase the revenue of a charity by one dollar. Existing literature argues that the price of giving influences this behaviour, as people are not pure altruists and they are concerned about their own welfare. In this respect, numerous studies explore the decision-making process when subjects have the opportunity to both act prosocially and obtain private monetary gains or increase their welfare. Within their noteworthy research, Andreoni and Miller (2002) find that individuals react to monetary incentives and that their prosocial actions are shaped accordingly. Specifically, they observe that aggregate giving drops when the cost of acting generously increases. Also other studies reach a consensus that there exists a negative relationship between the cost of generosity and prosocial behaviour (Wagner & Wheeler, 1969). For instance, Iyer and Kanekar (1991) find that the willingness to help decreases as the cost of helping increases. Similarly, given a higher price of giving, the probability of contributing to the funding of a project decreases (Meer, 2014).
By contrast, Gneezy, Imas, Brown, Nelson, & Norton (2012) are proponents of costly prosocial behaviour, by asserting that when one’s initial generous actions come at a cost, this will likely signal one’s prosocial identity and subsequently stimulate such behaviour; however, if recent prosocial actions are costless, prosocial behaviour is still positive, but it is less likely to be consistent. However, their study does not investigate the effects of a varying cost of giving on prosocial behaviour, and this thesis aims at providing insight into this matter.
One could argue that other than the price of giving as such, also the aggregate level of income of an individual is relevant in shaping giving behaviour. However, in this respect, the
available evidence is inconclusive. Although one might expect that a higher level of income will increase generosity, since for richer people giving is more affordable, Mayo and Tinsley (2009) find that higher income households underestimate their own fortune but also their potential recipients’ poverty. These perceptual biases lead to a decrease in the warm-glow effect experienced by wealthy donors, and subsequently to lower contributions to those in need than otherwise expected.
Moreover, Exley and Terry (2016) conduct an experiment on how effort levels respond to wage changes. Interestingly, when subjects’ labour supply is equivalent solely to private earnings, a higher wage motivates people to exert more effort. However, when individuals are assigned to earn money for a charity, they attribute a lower value to their earnings and subsequently exert less effort when the wage increases. Nonetheless, if people opt for voluntarily benefitting a charity, they do not exhibit a negative effort response to an increase in wage.
This body of facts is compatible with the conjecture made by Andreoni and Miller (2002) that several adjoining factors influence the price elasticity of giving. For instance, in a study about children’s prosocial behaviour, Eisenberg-Berg and Neal (1981) found that the younger the children, the more self-interestedly they act, and that a higher cost of acting prosocially has a higher negative impact on young compared to older children. This result suggests that age and giving are positively related, and also that receiving a higher level of education increases prosocial actions.
Overall, there is a general agreement within the literature that the cost of giving affects donating to charities. However, there is less consistency across studies regarding the magnitude of the price effects on giving, as well as on the interplay between cost and other regarding interests when behaving prosocially.
2.3. Self-deception – finding excuses for giving less
The drawbacks of giving, such as the cost of acting generously and less favourable assessments of the recipient type, may overcome the benevolence of potential donors. Indeed, there is evidence that although people may be motivated to act prosocially, their behaviour can be unethical. Bersoff (1999) explained that individuals might ignore a social obligation and act more self-interestedly, because a break occurs between the initially-perceived moral judgement and their subsequent behaviour regarding a given context.
In line with this statement, Tenbrunsel and Messick (2004) attribute unethical decisions to self-deception, a process that implies a biased update of beliefs that allows one to act selfishly while still enabling moral justifications for such behaviour. They argue that self-deception enables people to behave unethically and take responsibility for their selfish decisions without incurring negative feelings such as guilt or remorse.
Generally, self-deception has evolved out of deception, in order to facilitate social interactions and to avoid the detection and retribution of deception (McKay, Mijovic-Prelec,& Drazen Prelec, 2011). Self-deception can be socially advantageous, as it reduces the costs of deceiving others, and it increases people’s confidence (Von Hippel & Trivers, 2011). However, self-deception also has personal and social disadvantages; for instance, a study on the health risk behaviour finds that people demonstrate a self-deceivingly optimism by underestimating and even denying the dangers their health may be exposed to (Schwardmann, 2017).
Regarding giving, and specifically donating to charities, the available literature has been substantially interested in the determinants of self-deception and in people’s mental capability of hosting falsehoods. In this regard, it has been generally agreed that ambiguity and uncertainty host self-deception (Sloman, Fernbach, & Hagmayer, 2010). People might exploit unclear situations and behave more selfishly, and thereafter defend their unethical decisions by speculating on contextual vagueness.
Ambiguity can arise out of having subjects facing risky choices or inexplicit information. First of all, Haisley and Weber (2010) found evidence of preference for ambiguous risks, when the decisions made under such contexts enable justifications for unethical actions. In a similar effort, Exley (2015) finds that people use risk as an excuse to give less to charity. Specifically, participants in her study evaluate the risk faced by a charity and that faced by themselves; with no monetary trade-offs between the two parties, individuals similarly estimate charity- and self-risk. However, when their actions imply a compromise between money for the charity and for themselves, and they do have a reason to search for justifications to act more selfishly, their behaviour highlights a higher aversion to the risk faced by the charity and a lower aversion to their own risk.
Second of all, individuals can take advantage of eventual (lack of) information in order to avoid giving or to contribute less. For instance, Exley (2015) observes that people may speculate on the provided information regarding the effectiveness of charities by exaggerating their dislike to the less efficient ones and by subsequently using this lack of taste as an excuse for making significantly lower donations. Moreover, there is comprehensive research conducted on people’s prudent choice to remain ignorant to information that might compel
them to act against their will. Although holding knowledge is usually advantageous, people often choose to remain ignorant to information out of fear (Howell & Shepperd, 2013). This fear is generally caused by the emotional and behavioural implications of the respective information, such as urging an undesired change in beliefs and inconvenient subsequent actions, as well as the undergoing of negative emotions (Sweeny et al., 2010).
In the context of giving, ignorance is opted for out of the fear of openly making a compromise between the personal and the charity’s monetary payoffs (Grossman & van der Weele, 2017). For instance, Dana, Weber, and Kuang (2007) provide evidence for the illusionary preference for fair outcomes in the Dictator games, by proving that when inconvenient information can be avoided, people behave self-interestedly. They find that, by removing transparency, subjects feel less compelled to act according to social norms and they make more selfish choices. Furthermore, even with unequivocal information, people yet evoke irrelevant reasons in order to justify more selfish decisions (Exley & Kessler, 2016). Such findings hint to the fact that some people are not willing to give at all, but that their actions reflect the social pressure of acting according to some well-established norms, and the consequent incentive to preserve a good self-image.
The interest of this research is to contribute to the related literature by observing the effects of the interplay between self-deception and the cost of giving on philanthropic actions within a Dictator game. Specifically, there is limited available evidence on the differences between the price elasticity of giving with and without the opportunity of self-deception. A first contribution of this study is that it attempts to cumulate the factors discussed separately in the literature which determine charitable giving. These elements have been presented in the paragraphs above, and in summary they are: the characteristics of the charities, such as type, cause, or notoriety, the cost of giving, as well as the income of the donor, the social pressure and self-image concerns relative to one’s actual will, and the opportunity to self-deceive by avoiding potentially inconvenient information in order to mitigate such external pressure and justify eventual unethical decisions. By merging all these different factors, this study may help to better predict donors’ charitable involvement, as well as to improve the approach of philanthropic organizations and policy makers.
Moreover, this study differs from the existing analogous research in several aspects. Firstly, it includes two different types of recipient charities, one being significantly more deserving of receiving aid than the other. Secondly, similarly to the study of Dana et al. (2007), the opportunity for self-deception is introduced by making available the possibility to deliberately choose to avoid potentially inconvenient information. However, within this study,
the information concerns the type of recipient, namely the more or the less deserving charity. Therefore, the focus is on people’s actions subject to different types of receivers and not to varying experimental settings. Also, the option to remain ignorant by avoiding information can be directly chosen, if applicable, by the participants themselves, in order to ensure that ignorance occurs wilfully and not due to misunderstandings or lack of attention.
Nevertheless, in this setting, distinct from other studies, the donor does not need to exert any effort to earn money, but does obtain his or her income as a gift. Accordingly, this research is orchestrated in a one-shot Dictator game, where subjects face the dilemma of whether, and if yes to what extent, to share a gift with a charity. Moreover, this study focuses primarily on the differences in philanthropic behaviour given various costs of giving. Therefore, this experimental framework helps to mitigate the effect of additional variables or biases which are beside the point of this study, such as the ability of subjects to perform in a given task, their view on the relationship between effort and earning money, or their post-effort evaluation of charities.
Provided the outcome of these findings, this research aims at supplying additional insights into people’s reasoning and price elasticity of generosity directed to charitable organizations. The experimental design and the hypotheses to be tested throughout this research, as well as the econometric methods employed are presented in the section below.
3. Methodology
This section contains the research methods used in this paper. First, sub-section 3.1. describes the design of the experiment conducted during this research, as well as the data set collected. Next, in sub-section 3.2. the procedures implemented to operationalize and analyse the data are explained, followed by the formulation of the working hypotheses which this research aims to test.
3.1. Data set and experimental design
The data utilized in this study were collected within an online experiment, designed as a survey in Qualtrics, a software capacitating users to collect various types of data1. Moreover, the 418
participants in the experiment, out of which 175 were female and 243 male, were recruited from Amazon’s Mechanical Turk (MTurk), a crowdsourcing online marketplace where human intelligence is made available to businesses and developers2.
All things considered, whether the online data collection method (implemented in this study) leads to similar and at least as valid results as laboratory or field experiments becomes an important issue for this research. However, the related literature is conclusive. Namely, Casler, Bickel, and Hackett (2013) find that the results of a behavioural study are almost identical when the data is gathered online from MTurk or in person. Moreover, the consensus is that online experiments are at least as valid as those run in the laboratory and that running an experiment online is significantly less costly and less time-consuming compared to an in-person experiment (Horton, Rand, & Zeckhauser, 2011). Specifically, MTurk is a feasible platform for conducting research which provides highly diverse and large cohorts, cheap and fast recruitment of subjects, and reliable data (Buhrmester, Kwang, & Gosling, 2011). Therefore, one can fairly assume that the method in which the data were collected does not pose any additional threats to the reliability and validity of this study, compared to alternative traditional methods of conducting research.
Further on, participants recruited from MTurk accessed the online survey and took part of the experiment. Besides the participation fee, subjects were given an endowment of 50 cents, out of which they could donate either nothing, or any multiple of 5 cents up to the maximum of 50 to one out of two charitable organizations, while the difference between the initial endowment and their donation was assigned to them. Therefore, the subjects played a version of a one-shot Dictator game, while being in the role of the dictator and the charities representing the receiver.
Participants were informed about both charities, namely UNICEF and The Oxford University Yacht Club, which highly differed in the mutually accepted degree of deserving monetary donations. Specifically, UNICEF defends the rights of children and uses the cash donations for the improvement of policies and services that protect and help children, whereas The Oxford University Yacht Club provides yachting and dinghy sailing opportunities to
Oxford students and it depends on donations to finance these activities both for leisure and at a competitive level3, 4.
The reason why these two different charities were chosen as respondents in the experiment is that the interest of this study is to analyse donors’ decisions given contrasting types of potential beneficiaries. Specifically, it was made obvious to the subjects that the recipients of their donation differ in the urgency of the cause they are promoting, namely that The Oxford University Yacht Club supports students engaging in leisure activities, whereas UNICEF fights against an actual serious problem. Thus, one could easily realise that UNICEF was generally regarded as highly deserving of receiving aid, whereas the other charity was, in comparison, significantly more negatively perceived. Therefore, acting more selfishly when The Oxford University Club was the charity benefiting from one’s donation could be considered as somewhat reasonable, whereas ungenerous behaviour directed to UNICEF would be socially regarded as inappropriate.
Furthermore, subjects were divided into two control groups and a treatment group. Within the first control group, participants were informed that UNICEF would benefit from their contribution, while the participants of the second control group were explicitly briefed that the Oxford University Yacht Club would receive their donation. However, in the treatment group, subjects were not told beforehand which charity would receive their contribution, but that each one of the two was equally likely to obtain it. Participants in the treatment group were asked later in the survey to choose between revealing and not revealing the name of the charity which would receive their monetary gift. This option made room for self-deception, as the more self-interested people could choose to avoid the possibility of obtaining the inconvenient information that the more deserving charity, UNICEF, would benefit from their donation. As revealing such information might have morally pressed them to act more generously, choosing not to reveal the name of the charity provided them the opportunity to act more selfishly by donating a lower amount.
Moreover, as this study aims at analysing the price elasticity of prosocial behaviour when there is opportunity for self-deception compared to when there is not, the cost of acting prosocially varied across groups. Specifically, each of the three groups was divided into two prices of giving conditions, a low (baseline) and a high one (treatment). These costs were
3 More details about UNICEF can be found on: https://www.UNICEF.org 4 Additional information about The Oxford University Yacht Club is available on:
applied across participants by using a multiplier to convert the amount assigned by subjects to themselves into their final bonus payoff. Namely, in the low price of giving group, one’s bonus payoff was formed by the difference between the initial 50 cents and one’s donation, difference multiplied by 0.5, while in the high cost of giving group this difference was multiplied by 0.9. These multipliers were intentionally selected to be below 1, in order to eliminate the incentive to act totally selfishly during the experiment, but at the impetus of altruism, by thereafter privately donating the whole maximized amount to the charity. Table 1 summarizes these main elements of this experiment.
Table 1: Elements of the experiment
Group Price of giving: multiplier Choice Information Choice Donation Beneficiary charity
Control: Low Price
(81 participants) 0.5 -
0, 5, 10, …, 45, 50 UNICEF 0, 5, 10, …, 45, 50 The Oxford University
Yacht Club
Control: High Price
(85 participants) 0.9 -
0, 5, 10, …, 45, 50 UNICEF 0, 5, 10, …, 45, 50 The Oxford University
Yacht Club
Treatment: Low Price
(129 participants) 0.5
Not Reveal Charity
0, 5, 10, …, 45, 50 Either UNICEF or The Oxford University Yacht
Club (50% chance each)
Reveal Charity
0, 5, 10, …, 45, 50 UNICEF 0, 5, 10, …, 45, 50 The Oxford University
Yacht Club Treatment: High Price (123 participants) 0.9 Not Reveal Charity
0, 5, 10, …, 45, 50 Either UNICEF or The Oxford University Yacht
Club (50% chance each)
Reveal Charity
0, 5, 10, …, 45, 50 UNICEF 0, 5, 10, …, 45, 50 The Oxford University
Yacht Club
Moreover, Table 2 depicts all possible payoffs that participants could obtain, depending on their decision regarding how much to donate and on the cost of giving condition that they were assigned to. Notably, one can observe that the subjects who incurred higher costs of acting prosocially obtained, however, higher final payoffs compared to the participants facing a low
price of giving. This aspect of the experimental setting is different from the general view that lower costs of giving are associated to higher levels of income (Mayo and Tinsley, 2009). As an increase in income and a decrease in the price of giving might reasonably lead to higher donations, a positive relationship between these two factors has less foreseeable effects on generosity.
Table 2: Participants’ all possible final bonus payoffs, depending on cost of giving condition and donation decision
Initial
endowment Donation
Amount assigned to
oneself
Low price of giving High price of giving Multiplier Final bonus
payoff Multiplier Final bonus payoff 50 0 50 0.5 25 0.9 45 5 45 22.5 40.5 10 40 20 36 15 35 17.5 31.5 20 30 15 27 25 25 12.5 22.5 30 20 10 18 35 15 7.5 13.5 40 10 5 9 45 5 2.5 4.5 50 0 0 0
Note: All outcomes presented in this tables are amounts in eurocents, except the multipliers 0.5 and 0.9, which are proportions.
Further on, Figure 1 displays the timeline and organization of the experiment. It also highlights how subjects were randomly assigned to different experimental conditions. Initially, 418 individuals started the experiment, and were afterwards equally divided between the two costs of giving conditions. Furthermore, the participants were assigned into the information conditions, namely, either into the control group where the information about the beneficiary’s identity was fully disclosed, or into the treatment group where this information was originally hidden. Moreover, more participants were distributed to the treatment than the control group; this feature was intentionally designed in order to increase the likelihood of the occurrence of ignorance, and to balance the number of choices made under each condition. Figure 1 shows that this randomization worked accordingly, and that the distribution of the observations is reasonable enough to make this data set a reliable material for this study.
Figure 1: Timeline and organization of the experiment
All things considered, the design of this experiment can be represented in a 4x2 matrix, with the rows constituting the information conditions, and the columns the cost-of-giving states. Table 3 depicts this matrix, as well as the abbreviation of each experimental group which will be used as terminology in the remainder of this paper. As already described above, the participants incurred either a low or a high cost of giving, and each charity had equal likelihood of collecting donations, regardless of whether subjects knew precisely which charity was their donation’s recipient or not. Within the abbreviation, the first letter, C or T, represents the information group, control and treatment respectively; the second letter stands for the name of the charity, U for UNICEF and Y for The Oxford University Yacht Club; the third letter corresponds to the price of giving, low (L) or high (H). For instance, the cell corresponding to the first column and the third row, TUL, represents the group which was not given the information about the recipient charity’s identity beforehand. However, all the participants within this group were assigned to donate to UNICEF indifferently on whether they chose to reveal this information or to stay ignorant. Lastly, this group incurred a low cost of giving to charity.
Table 3: Experimental groups matrix and group abbreviation (terminology)
Price of Giving
Low Price of Giving High Price of Giving Group, UNICEF Control, UNICEF CUL CUH
Treatment, UNICEF TUL TUH
Group, The Oxford University Yacht Club
Control, The Oxford
University Yacht Club CYL CYH
Treatment, The Oxford
University Yacht Club TYL TYH
Finally, at the end of the experiment, participants were asked to fill in a questionnaire asking about their gender, age, level of education, the frequency they donate to charity, which charity out of the two they consider more deserving of receiving aid, and the extent to which they find each charity worthy of obtaining donations. These data were collected in order to serve as control variables in the following analysis5.
3.2. Hypotheses and econometric procedures
Taking into account the insights as well as the gaps in the related literature presented and analysed in the previous sections, several hypotheses can be formed. Furthermore, throughout this study, provided the structure of the experiment and the data collected, these hypotheses will be econometrically tested and interpreted.
Firstly, it is of high importance to observe whether the distinction between the two charities influences donations. Specifically, UNICEF should be regarded as more deserving of receiving donations than The Oxford University Yacht Club, because it promotes a more urgent and touching cause. This fact might determine subjects to donate more to UNICEF, either
5 For a preview of the experiment, access:
because they want to or because they think that that would be an appropriate action, whereas the incentives to give to The Oxford University Yacht Club might be sensibly lower and the donations likewise. However, whether the majority of subjects had a similar perception regarding this aspect should be demonstrated by their evaluations regarding each charity made within the questionnaire, as well as by higher donations made to UNICEF than to The Oxford University Yacht Club. Therefore, the first hypothesis of this study can be formulated as:
Hypothesis 1: Within each price of giving group, donations made to UNICEF will be higher than those made to The Oxford University Yacht Club. Formally,
𝐻": 𝜇&'( > 𝜇&*( 1
𝜇&', > 𝜇&*, 2
𝜇.'( > 𝜇.*( 3
𝜇.', > 𝜇.*, 4
where and 𝜇 represents average donations, and the subscripts are the abbreviations of the experimental groups described above in Table 3.
Each of these hypotheses will be tested against the null hypothesis that the type of charity has no effect on the amounts donated, namely that the relationship between the average donations within each equation is of equality. Notably, it is essential to measure the effects of charity characteristics only within groups belonging to the same price of giving conditions, in order to mitigate the cost of giving effects. This hypothesis represents the premise of this study, as its outcome is crucial for the remainder of the analysis.
Moreover, as mentioned above, one can expect that the costs of giving will cause differences in donations. Specifically, a higher price of giving might generally make subjects less willing to contribute to charity (Wagner & Wheeler, 1969; Iyer & Kanekar, 1991). Thus, an increase in the cost of giving might lead to lower average contributions, because in spite of the positive incentive to aid UNICEF, subject may not want to incur large drops in their income. However, as generosity aimed at The Oxford University Yacht Club is expected to be at low levels, the change in the price of giving should lead to minimal, if not null, effects on average donations. Therefore, the second hypothesis of this research is:
Hypothesis 2: Within each information group, the amounts donated to UNICEF will be on average higher when the price of giving is low, compared to when the cost of giving is high. That is, an increase in the price of giving will negatively affect contributions made to UNICEF. Formally,
𝐻1: 𝜇&'( > 𝜇&', 5 𝜇.'( > 𝜇.', 6 .
Therefore, this research attempts to observe the effects on giving of the nominal changes in income, rather than the impact on prosocial actions caused by variations of the price of giving relative to income. More precisely, referring to Table 2, one can notice that for each price of giving conditions, increasing donations by 5 cents requires sacrificing a tenth of the total maximum income; however, this research focuses on the consequences of the aggregate difference between the two costs of giving, namely 0.4 (=0.9-0.5).
The above hypotheses indicate that several factors might determine people to wish to donate less. Particularly, besides a personal preference, subjects may also feel socially-compelled to contribute (more) if UNICEF is assigned to be their recipient charity. Thus, some subjects may exhibit a level of prosocial behaviour which does not reflect their actual intentions and degree of generosity. Under these circumstances, opting for ignorance, and therefore avoiding the possibility to obtain the information that UNICEF is indeed the beneficiary of the donation, might “shelter” a more selfish behaviour, by alleviating negative feelings or external pressures, such as guilt, remorse, image-concerns, or inappropriateness. Therefore, a fair hypothesis, which has already been extensively tested by numerous studies (Dana et al., 2007;
Grossman & van der Weele, 2017) is:
Hypothesis 3:The aggregate level of giving will be lower within the group where the information that UNICEF is the recipient charity will be hidden, compared to the group which will be provided this information beforehand. Formally,
𝐻4: 𝜇.'( < 𝜇&'( 7 𝜇.', < 𝜇&', 8 .
Differently stated, this hypothesis implies that the group of subjects who are not explicitly told which one of the two charities will collect their contributions, although in fact
the charity which is assigned to be their recipient is UNICEF, will contribute on average less than the group which is given this information at the beginning of the experiment. The former group will donate on average less than the latter, regardless of whether participants choose to reveal or not this information. Again, these average donations are only compared across groups within the same price of giving condition, in order to exclude the cost of giving effects.
However, one can reasonably expect that, within the hidden information groups, the prosocial behaviour exhibited by the participants who choose to reveal the identity of the recipient charity will be different from that of those who opt to remain ignorant to this information. This can be inferred from the fact that subjects consciously select whether or not they want to reveal the name of the charity which will collect their donations, that is, their prosocial intentions are already shaped by the time they make this choice.
Firstly, when UNICEF is the recipient charity, some subjects might feel compelled to make higher donations. However, in the hidden information conditions, participants who opt for ignorance will avoid disclosing this potentially inconvenient information and use this lack of knowledge as an excuse for lower contributions. Therefore, one can expect that the average donations made by these subjects will be lower than the mean donations made by participants who are given this information beforehand.
Secondly, subjects who choose to disclose the information that UNICEF collects their contributions will probably not act in the same way as those who obtain this information passively (Grossman & van der Weele, 2017). This expectation is formed because the control group (complete information) comprises various types of donators, whereas within the treatment condition (hidden information) only the highly altruistic subjects choose to reveal this information. Thus, participants who choose to reveal the information that UNICEF is the recipient charity should donate on average more than those who obtain this information passively.
All things considered, regarding Hypothesis 3, equations 7 and 8 encompass the idea that compared to the control condition, the decrease in giving caused by ignorance is expected to overweigh the increase in average donations due to the wilful disclosure of the information that UNICEF is the recipient charity. That is, relative to complete information conditions, the aggregate giving will drop in the hidden information conditions.
However, when The Oxford University Yacht Club is assigned to be the recipient charity, the average contributions made by the participants within the hidden information groups should not significantly differ from those made by the subjects belonging to the control (complete information) groups. The reasoning of this expectation is that as incentives to donate
to this charity ought to be anyway low, creating opportunity for self-deception should not have a notable impact on average donations.
Specifically, opting for ignorance should be a deliberate action, taken in order to avoid having to wittingly donate to UNICEF. Therefore, a more selfish behaviour could be defended by the argument that The Oxford University Yacht Club could have been the recipient charity. Also, subjects who reveal that The Oxford University Yacht Club will collect their donations are not expected to behave significantly different from those who obtain this information passively.
Further on, it is important to note that ignorance is not equivalent to indifference. That is, subjects who are indifferent between which one of the two charities they help should choose to reveal information. Information avoidance occurs when subjects are aware of the perceived differences between the two charities and they are highly concerned about their self-image. Therefore, some expectations can be formed regarding the ignorance rate.
Firstly, while some subjects may behave as homo economicus by giving nothing to UNICEF, some other marginal types might still exhibit some degree of generosity in order to avoid being included in the former group (Grossman & van der Weele, 2017). However, allowing for information avoidance, may prompt these marginal types switch into acting more selfishly, which would decrease average donations compared to the control group. This aspect can be formulated in the hypothesis:
Hypothesis 4: When UNICEF is the recipient charity, the rate of completely selfish choices (that is, donating nothing), when subjects opt for ignorance within the hidden information group, will be higher than the rate of completely selfish choices within the complete information control group. Formally:
𝐻8: 𝑠.'( < 𝑠&'( 9
𝑠.', < 𝑠&', 10
where 𝑠 represents the rate of totally selfish choices.
Secondly, as stated in Hypothesis 2, when the price of giving increases, subjects should become less willing to contribute; therefore, more of them might choose not to reveal the potentially inconvenient information. Therefore, one would expect that the ignorance rate will increase when the cost of giving increases:
Hypothesis 5: When the cost of giving is high, more subjects will opt for ignorance in the hidden information condition, compared to when the price of giving is low. Formally,
𝐻<: 𝑖.'( < 𝑖.', 11 where 𝑖 stands for the rate of ignorance.
Nevertheless, the main focus of this study is to observe whether the average effects of the price of giving on the amounts donated are higher if the subjects have the opportunity to self-deceive by remaining ignorant to potentially inconvenient information. The reasoning is that although the increase in the cost of giving should negatively impact donations regardless of the information condition, the fact that participants can use ignorance to behave more self-interestedly should lead to a larger drop in aggregate giving relative to the case when subjects know beforehand that UNICEF is collecting their contributions. Therefore,
Hypothesis 6: When UNICEF is assigned to obtain one’s donation, the price elasticity of giving is higher when one can self-deceive by avoiding information relative to when one cannot. That is, when the price of acting generously increases, the negative effect on contributions will be larger for self-deceivers compared to subjects who do not have this opportunity. Formally,
𝐻>: 𝜇.',− 𝜇.'( < 𝜇&', − 𝜇&'( , 𝜇.',− 𝜇.'( > 𝜇&', − 𝜇&'( (12).
All things considered, in the following section, these hypotheses will be initially individually tested using one-sided Mann-Whitney U tests in the case of equations (1), (2), (3), (4), (5), (6), (7), and (8) and one-sided exact z-tests for equations (9), (10), and (11). Furthermore, the majority of these hypotheses together with equation (12) and additional control variables will be incorporated in a regression model, in order to provide a more accurate estimate of the effects on charitable giving. The regression model this paper proposes is as such:
𝑫𝒊 = 𝜷𝟎𝒊+ 𝜷𝟏𝒊∗ 𝑷𝒓𝒊𝒄𝒆 𝒐𝒇 𝑮𝒊𝒗𝒊𝒏𝒈𝒊+ 𝜷𝟐𝒊∗ 𝑮𝒓𝒐𝒖𝒑, 𝑼𝒏𝒊𝒄𝒆𝒇𝒊+ 𝜷𝟑𝒊∗
𝑷𝒓𝒊𝒄𝒆 𝒐𝒇 𝑮𝒊𝒗𝒊𝒏𝒈𝒊∗ 𝑮𝒓𝒐𝒖𝒑, 𝑼𝒏𝒊𝒄𝒆𝒇𝒊 +𝜷𝟒𝒊∗ 𝑮𝒆𝒏𝒅𝒆𝒓𝒊+ 𝜷𝟓𝒊∗ 𝑨𝒈𝒆𝒊+ 𝜷𝟔𝒊∗ 𝑬𝒅𝒖𝒄𝒂𝒕𝒊𝒐𝒏𝒊+ 𝜷𝟕𝒊∗ 𝑵𝒂𝒕𝒊𝒗𝒆 𝑬𝒏𝒈𝒍𝒊𝒔𝒉𝒊+ 𝜷𝟖𝒊∗ 𝑭𝒓𝒆𝒒𝒖𝒆𝒏𝒄𝒚 𝑫𝒐𝒏𝒂𝒕𝒆𝒊+ 𝜷𝟗𝒊∗ 𝑫𝒖𝒓𝒂𝒕𝒊𝒐𝒏𝒊+ 𝜺𝒊 (13)
where 𝑫𝒊 represents the amount donated to UNICEF by individual 𝑖, regardless of whether he or she knew that UNICEF was the recipient charity; 𝑷𝒓𝒊𝒄𝒆 𝒐𝒇 𝑮𝒊𝒗𝒊𝒏𝒈𝒊 is a dummy variable representing the price of giving condition (see Table 3), namely, it is equal to 0 or 1 if the choice was made given a low or high price of giving, respectively; 𝑮𝒓𝒐𝒖𝒑, 𝑼𝒏𝒊𝒄𝒆𝒇𝒊 is also an
indicator variable equal to 0 if participant 𝑖 is in the control group donating to UNICEF, and 1 if he or she is in the treatment group contributing to UNICEF (indifferent on if he or she chose to reveal this information or not). Moreover, 𝑷𝒓𝒊𝒄𝒆 𝒐𝒇 𝑮𝒊𝒗𝒊𝒏𝒈𝒊∗ 𝑮𝒓𝒐𝒖𝒑, 𝑼𝒏𝒊𝒄𝒆𝒇𝒊 is an interaction term and also a dummy variable, which is 1 if individual 𝑖 is in the treatment group donating to UNICEF and makes his or her choice given a high price of giving, and 0 in all other three cases.
Furthermore, the remainder are control variables, namely, 𝑮𝒆𝒏𝒅𝒆𝒓𝒊 is a binary variable equal to 0 if subject 𝑖 is female and 1 if he is male, 𝑨𝒈𝒆𝒊 is his or her age, 𝑬𝒅𝒖𝒄𝒂𝒕𝒊𝒐𝒏𝒊 is the level of education of individual 𝑖, namely, less than highschool, highschool, some college, Bachelor’s degree, Master’s Degree, or Doctorate (represented also as a dummy variable), 𝑵𝒂𝒕𝒊𝒗𝒆 𝑬𝒏𝒈𝒍𝒊𝒔𝒉𝒊 is a binary variable equal to 1 if individual 𝑖 is a native English speaker and 0 if not, 𝑭𝒓𝒆𝒒𝒖𝒆𝒏𝒄𝒚𝑫𝒐𝒏𝒂𝒕𝒆𝒊 stands for the rate with which participant 𝑖 stated that he or she donates to charities in real-life, specifically, never, sometimes, or often (also an indicator variable), and 𝑫𝒖𝒓𝒂𝒕𝒊𝒐𝒏𝒊 represents the number of seconds it took individual 𝑖 to complete the experiment.
The interpretation of the beta coefficients is as follows: 𝜷𝟎𝒊 and 𝜷𝟎𝒊+ 𝜷𝟏𝒊 represent the average donations made within the control group (donating to UNICEF) given a low and a high price of giving, respectively; these coefficients are related to equations (5) and (6) of Hypothesis 2. 𝜷𝟎𝒊+ 𝜷𝟐𝒊 and 𝜷𝟎𝒊+ 𝜷𝟏𝒊+ 𝜷𝟐𝒊+ 𝜷𝟑𝒊 constitute the mean contributions made to UNICEF by the treatment group under a low and high price of giving, respectively. Therefore, 𝜷𝟐𝒊 encompasses the single difference in giving between the treated and the control group donating to UNICEF given a low price of giving and 𝜷𝟐𝒊+ 𝜷𝟑𝒊 represents the same difference but under the high price of giving condition (equivalent to equations (7) and (8) of Hypothesis 3, respectively). Of the highest importance to this study, 𝜷𝟑𝒊 represents the
difference between these two differences and it termed as the difference-in-difference estimator. It represents the impact of the higher price of giving for the treatment versus the control groups donating to UNICEF relative to the effect of the lower price6; it highlights the relationship expressed equation (12) of Hypothesis 6. Nevertheless, 𝜷𝟒𝒊, 𝜷𝟓𝒊, 𝜷𝟔𝒊, 𝜷𝟕𝒊, 𝜷𝟖𝒊, and 𝜷𝟗𝒊 are regression coefficients encompassing the structural relationship between the control variables and charitable donations, and 𝜀o is the error term. Additionally, the error term
𝜀o is assumed to be normally distributed with a mean 0 and a constant variance 𝜎1, which is
the residual sample variance.
The outcomes obtained from the experiments, as well as the results of the hypotheses’ tests are presented and analysed in the section below.
4. Results and Analysis
This section is organized as follows: firstly, sub-section 4.1. encompasses the charity evaluations made by subjects within the questionnaire at the end of the survey, assessments which represent the most important prerequisite for this study. Secondly, sub-section 4.2. provides an overview of the data obtained per treatment group (information about the recipient charity and price of giving conditions) regarding charitable donations, outlined in a comprehensive table and depicted in explanatory graphs. Furthermore, the outcomes of the tests of the first 5 hypotheses (respectively, the first 11 equations) against their corresponding null hypotheses will be presented and discussed in sub-section 4.3. Lastly, sub-section 4.4. will comprise the outputs of the regression developed above (see equation (13).
4.1. Charity evaluation
As already mentioned in the previous section, this study was built on the presumption that the characteristics of a charity influence donor’s behaviour. In this setting, two charities, which differed to a high extent in their degree of urgency of receiving help, were presented to the subjects. UNICEF fights for the rights and protection of unfortunate children, whereas The
6 For more information about Difference-in-Difference estimators and their analysis in the statistics program
Oxford University Yacht Club promotes leisure sailing activities for students. In order to observe whether this difference between charities was indeed generally considered to be significant, subjects were asked at the end of the experiment to assess the degree to which they found each of the organizations of being deserving to receive aid.
As Figure 2 shows, subjects indeed had a different perception on the two charities: 382 of them considered that UNICEF is more deserving of receiving donations than The Oxford University Yacht Club, while the remainder of 36 found otherwise. Therefore, the majority of participants agreed that UNICEF highly deserved to collect their donations, whereas The Oxford University Yacht Club was highly negatively perceived. Moreover, the gradual increase in positive opinions about UNICEF and more negative ones about the second charity suggests that the rest of the participants tended to be in consent with the majority.
Figure 2: Charity’s Degree of Deserving Receiving Donations Assessment
Note: The vertical line measures the number of participants choosing a particular option. The vertical bars represent the selected assessment regarding each charity. The blue one stands for UNICEF, whereas the orange one for The Oxford University Yacht Club.
This finding is of high importance to the remainder of the analysis conducted in this paper. The clear trend in people’s preferences implies that subjects were significantly less willing to donate to The Oxford University Yacht Club than to UNICEF. Moreover, this consensus provided a collective requirement to donate (an adequate amount) to UNICEF, whereas acting less generously towards the other charity was not socially regarded as inappropriate. Therefore, subjects who were more interested in increasing their own rather than
the charity’s welfare, but who also had high concerns for their self-image, might have donated more solely due to this social pressure. However, within the treatment (hidden information) groups, the opportunity to avoid obtaining the information that UNICEF will receive their donation allowed these subjects to behave according to their intrinsic will. The following subsections will provide evidence that this is indeed the case.
4.2. Overview of charitable donations data by experimental groups
This subsection comprises some of the most relevant descriptive statistics of this study. Table 5 and Figure 3 depict the average donations, as well as the standard deviation of the contributions made across groups. It indeed seems that people donated more to UNICEF than to The Oxford University Yacht Club, because within each price of giving condition, average contributions made to the former charity were larger than those directed to the latter. At this stage, the first hypothesis formulated in this study appears to be realistic, as equations (1), (2), (3), and (4) appear to be confirmed in the Mean column of Table 5, as well as by the differences between the blue and yellow and the green and red lines, respectively, in Figure 3.
Moreover, within the treatment groups, there is a significant proportion choosing for ambiguity regarding the identity of the recipient charity (69% and 75% under the low and high cost of giving, respectively). This suggests that, as explained in the previous subsection, the generally perceived characteristics of the charity, as well as the desire to behave according to social norms, play an important role in donor’s actions. Thus, the opportunity to remain ignorant to inconvenient information can represent a safe haven for more selfish subjects.
However, the findings regarding donations given a low versus a high price of acting prosocially are relatively at odds: in almost all the groups the donations made in the former case are lower than those made in the latter. The second hypothesis, namely equations (5) and (6), does not seem to be supported by the data: donations made in the CUH group are slightly higher than those made in the CUL one; the same is valid for the TUH and TUL groups. This aspect suggests that the increase in income probably overcame the associated increase in the cost of giving; however, these differences across groups are minimal.
Moreover, the aggregate average contributions made by the participants assigned to the hidden information groups, namely TUL and TUH, were moderately higher than those made by the subjects belonging to the control groups helping UNICEF (CUL and CUH, respectively). However, this finding does not imply that no self-deception occurred; rather, it
suggests that participants who wilfully revealed that UNICEF was the recipient charity provided a higher positive impact on aggregate donations than the negative effect caused by subjects who opted for ignorance. This conjecture is supported by the outcomes presented in the disaggregate hidden information treatment displayed in Figure 4, where, relative to CUL or CUH, the decrease in the contributions made by participants who chose to avoid information was lower than the increase in the amounts donated by the subjects who revealed UNICEF.
Nevertheless, as expected, donations directed to The Oxford University Yacht Club do not appear to differ significantly across groups. It however seems that the aggregate giving slightly drops in the treatment conditions compared to the control groups, when subjects are already aware that The Oxford University Yacht Club is the recipient charity, suggesting that wilfully revealing this charity leads to strong negative consequences on contributions (see Figure 4). Nonetheless, the standard deviations are similar across treatment groups, inferring that the randomization of subjects across conditions was efficient to provide unbiased results. An extended discussion regarding subjects’ characteristics and their interaction with the randomization is provided in Table 4 of the appendix.
Table 5: Descriptive Statistics: Donations
Group Charity Notation Mean Standard deviation N
Control group Low Price
UNICEF CUL 18.293 17.557 41
The Oxford University
Yacht Club CYL 11 15.2 40
Control group High Price
UNICEF CUH 18.415 16.219 41
The Oxford University
Yacht Club CYH 11.364 14.361 44
Treatment group Low Price Not Reveal - 11.125 16.112 40 Reveal UNICEF
The Oxford University Yacht Club
22.045 16.612
44
8.444 15.219 45
Hidden
information UNICEF TUL 18.984 16.954 64
The Oxford University
Yacht Club TYL 8.923 15.499 65
Proportion revealing the charity 89/129 (69%)
Treatment group High Price Not Reveal - 16.935 17.449 31 Reveal UNICEF 20.217 14.868 46
The Oxford University
Yacht Club 9.239 12.692 46
Hidden
information UNICEF TUH 19.194 15.738 62
The Oxford University
Yacht Club TYH 11.311 14.257 61
Figure 3: The effects of an increase in the price of giving on donations across experimental groups
Notes: The vertical line measures average contribution made to charity; the horizontal line stands for the price of giving, low (L) when the multiplier is 0.5 and high (H) when the multiplier is 0.9.
The coloured lines represent the change in mean donations given the cost of giving change for every information condition: CU represents the control (complete information) group donating to UNICEF, CY the control group contributing to The Oxford University Yacht Club, TU the treatment (hidden information) group assigned to donate to UNICEF, and TY the treatment group assigned to donate to The Oxford University Yacht Club.
Figure 4: The effects of an increase in the price of giving on donations across the hidden information groups by information choice
Notes: The vertical line measures average contribution made to charity; the horizontal line stands for the price of giving, low when the multiplier is 0.5 and high when the multiplier is 0.9.
The coloured lines represent the change in mean donations given the cost of giving change within the hidden information treatment group disaggregated by information choice: the Not Reveal line represents average donations made under wilful ignorance, Reveal U the contributions made after revealing that UNICEF is the recipient charity, and Reveal Y stands for the mean donations made to The Oxford Yacht Club after disclosing this information.
