When the Relatively Poor Prosper: The Underdog Effect on Charitable Donations

Abstract

In fundraising, it is common for the donor to see how much a charity has received so far. What is the impact of this information on (a) how much people choose to donate and (b) which charity they choose to donate to? Conditional cooperation suggests that people will donate to the charity that has received the most prior support, while the Underdog Effect suggests increased donations to the charity with the least support. Across two laboratory experiments, an online study (combined N = 494) and a qualitative survey (N = 60), a consistent preference to donate to the charity with the least prior support was observed. Thus, the Underdog Effect was supported. We suggest people will show a preference for the underdog if there are two or more charities to donate to, one of the charities is at a disadvantage, and people have little preexisting loyalty to either charity.

Keywords

social information charitable donations underdog effect conditional cooperation effect impact philanthropy

It is becoming increasingly common for charities to allow potential donors to be able to observe other contributors’ prior donations to the organization (Butt & Shah, 2012). It is believed that the visibility of previous donations provides social information that potential donors can use to help inform their own donation decisions. A number of experimental studies examining the effect of social information on previous donations have shown that people prefer to give to a charity/organization with a larger number of previous donations both in terms of the number of donations and the amount donated (Frey & Meier, 2004; Martin & Randal, 2008). However, many of these studies prevented direct comparison of such social information effects across charities by using between-subjects designs, whereby participants were presented with information about either one charity or another, but not both (Frey & Meier, 2004; Martin & Randal, 2008).

A more realistic scenario is that potential donors are able to compare information on previous donations across charities simultaneously. This is akin to online fundraising where potential donors can easily compare donation information for a number of charities. Therefore, we investigate whether under this donation scenario, donors would show a preference to donate to either the most supported or the least supported charity. If participants do prefer the least supported charity, we ask, is this because they are motivated by the desire to make their donation have a bigger impact on the charity (the impact donor) or are they motivated by a preference to support those at a disadvantage (the underdog donor)? Theoretically, this article explores the potential role of an Underdog Effect in charitable decisions. First, we review the theories where the preference is to give to a charity with greater prior support. We then review the theories supporting the preference to contribute to a charity with the least prior support. Finally, we explore how these preferences may be moderated by the observability of prior donations before outlining the current research.

Preferences for Giving to a Charity With Greater Prior Support

Conditional cooperation suggests that there should be a positive association between the amount others have already contributed to a cause and how much another individual will contribute (Fischbacher, Gachter, & Fehr, 2001). That is, if a large number of people are known to have donated to one charity compared with another, then conditional cooperators should be more likely to contribute to the charity with a greater number of prior donations. For example, Frey and Meier (2004) informed students that either 64% or 46% of past students contributed to two student funds. Those students were 2.3% more likely to donate to the charity to which they believed 64% of previous students had donated. Quality Signaling theory suggests that individuals give to charities that already have high levels of contributions because they are perceived to be (a) of higher quality, (b) more likely to use donations effectively, and (c) more likely to receive future donations (Vesterlund, 2003).

Preferences for Giving to a Charity With Least Prior Support

Duncan’s (2004) “Impact Philanthropy” model defines an alternative type of philanthropic donor, known as an impact donor. Impact donors derive satisfaction from knowing their contribution makes a real difference to the level of goods and/or services provided by a charity. If other donors have given substantially to a charity, then, potentially, any additional donations will have a smaller effect on increasing the level of goods and/or services that the charity provides. To test this model empirically, Borgloh, Dannenberg, and Aretz (2013) gave participants the option of donating to a charity that has a small (€40,000-€300,000) or large (€5-€11 million) annual revenue. They found that 73% of participants choose to give to the charity with the smaller revenue, which resulted in an additional €323 being donated to that charity. Borgloh et al. (2013) suggested donations to the charity with a smaller revenue would have greater impact as the donors’ contributions would have the largest effect on increasing the endowment of the charity.

An alternative explanation focuses on the “Underdog Effect.” The “Underdog Effect” is a robust phenomenon observed in voting behavior, brand loyalty, and sports spectators’ choice of team (Goldschmied & Vandello, 2009; Goldschmied & Vandello, 2012; Shirai, 2017) where one favors a previous unaffiliated entity, which is perceived to be at an undeserved disadvantage relative to others (Vandello, Goldschmied, & Richards, 2007). The Underdog Effect may then be plausibly extended to charitable decision making, predicting that donors should give to the charity with the least revenue because that charity is at a relative disadvantage (i.e., less funding). Indeed, the concept of the underdog is highly accessible (Kim et al., 2008) and a lay conception of the underdog is overly optimistic regarding the likelihood of the underdog succeeding (Goldschmied & Vandello, 2012). However, if the charity with lower revenue were perceived to be at a deserved disadvantage due to squandering donations (i.e., high staffing costs), then it would not be perceived as an underdog.

Effects of Observability on Donations

An ongoing, prominent debate in the charitable donation literature is whether the observability of a donation act increases the amount donated (Jones & Linardi, 2012; Mason, 2016). The evidence for observability is mixed, with some researchers finding evidence that it increases donations, both inside and outside the laboratory (Andreoni & Petrie, 2004; Soetevent, 2005; van Leeuwen & Wiepking, 2013), while other researchers find no correlation between the level of observability and the amount donated (Dufwenberg & Muren, 2006; Jones & Linardi, 2012). Thus, Studies 1 and 2 include an observability manipulation to contribute to the evidence on the effects of observability on donations.

Present Research

From previous empirical findings and theoretical models, two key questions emerge. First, when comparing across charities, do people donate to the charity with the least prior support (e.g., impact donors or underdog donors) or do they favor a charity with more support (e.g., conditional cooperators or quality signaling). Second, if the least supported charity is preferred, which of the competing explanations (“Impact Model” or the “Underdog effect”) best explains the behavior? Conversely, if the greater supported charity is favored, which explanation (conditional cooperation or quality signaling) is preferred? These questions are addressed in four studies.

Study 1

The first study explores whether participants donate more on average to a charity with more existing support than that with less existing support.

Method

Participants

A total of 156 students were recruited from the University of Nottingham through convenience sampling. Four cases were dropped due to invalid responses (i.e., allocating more money to charity than their endowment) and one further case was dropped due to technical problems. The final sample of 151 students consisted of 61 males (40.4%), 89 females (58.9%), and one who wished not to disclose (0.7%). Participants completed the study in groups of between four and 10 (M = 6.63, SD = 1.63), with each group being randomly assigned to conditions. Remuneration for the study was the amount participants earned minus the amount they chose to give to charity. The sample size, gender composition, average earnings, and average donations are reported in Table 1.

Table 1.

Comparison of Sample Size, Gender Composition, Average Endowment, and Earnings Across All Four Studies.

	Study 1	Study 2	Study 3	Study 4
Initial sample size	156	132	206	60
Gender composition (% female)	58.9	54	57.1	Not collected
Endowment per person	£3	£4	£10.00^a	No endowment
M earnings per person (SD)	£2.09 (£1.56)	£1.01 (£1.09)	£2.82 (£3.61)^a	No earnings
M amount donated per person (SD)	£1.90 (£1.56)	£1.98 (£1.09)	£7.22 (£3.58)	£10.49 (£10.91)
Proportion of donations (% of donors vs. total potential donors)	76.67%	87.9%	89.83%	100%

A conditional lottery was used in Study 3, so one in 10 participants had their decisions remunerated.

Design

This mixed design study was a 2 (observability: private vs. public) x 3 (degrees of support: 50:50 vs. 20:80 vs. 80:20) x 2 (the distribution of resources across 80:20 and 50:50) mixed design. Observability and degree of support are between-subjects factors and the distribution of resources is a within-subjects factor. Following Ariely, Bracha, and Meier (2009) and Kataria and Regner (2015), observability was manipulated by informing participants in the public condition that at the end of the experiment, they would be asked to read out their donation decisions to the rest of the participants in their group once everyone in their group had completed the study. Participants in the private condition were informed that their decisions would remain private and anonymous. Degree of support was manipulated via a screen titled, “Donations made by participants so far.” On the screen, there were two clear plastic jars filled with money. Each was labeled with the name of a charity (British Heart Foundation: BHF or Cancer Research UK: CRUK). The degrees of support in the jars varied across three conditions: (a) BHF is 80% full and CRUK is 20% full, (b) BHF is 20% full and CRUK is 80% full, and (c) BHF is 50% full and CRUK is 50% full (see Figure 1). The 50:50 condition was included to check whether one charity was preferred over another. Distribution of resources is the amount of money in the jars within each of the three conditions (20%:80%, 80%:20%, 50%:50%).

Figure 1.

Example of the 80%, 50%, and 20% full jars of coins.

Procedure

The study was conducted in three sequential stages: (a) a money-earning task, (b) a charity dictator game, and (c) participants filling out an online questionnaire. In the money-earning task, participants pressed the “z” then “x” key in that sequence for a period of 5 min (following Ariely, Bracha, & Meier, 2009). Participants had been told that to earn £4, they must make 200 “zx” responses, which all the participants did (zx responses: M = 962.62, SD = 247.49).

In the charity dictator game, participants were informed on the next screen whether their donation would be public or private. Participants then saw a screen showing “Donations made by participants so far.” Depending upon the condition they were in, the jars were 80% to 20%, 50% to 50%, or 20% to 80% full of coins. The donation screen displayed three slider scales, each of which went from 0 pence to 400 pence with the slider moving in single penny increments. The three slider scales were labeled: “The British Heart Foundation,” “Cancer Research UK,” or “Self.” Participants could donate as much or as little money as they liked to BHF, CRUK, or keep it for themselves, as long as total money allocated equaled 400 pence. This page also contained the pictures of the jars as a visual reminder. If participants were in the public condition, once everyone had completed the study, they each read out their donation decisions to the rest of the participants.

In the last stage, the questionnaire collected data on gender, course of study, and current level of study. It also included two subquestionnaires, the Moral Foundation Questionnaire (MFQ) and the Reluctant Altruism Scale, which are not of current interest, so they are not discussed further. Participants were reimbursed with any money they chose to keep, and all charitable donations were given to the relevant charities.

Results and Discussion

The donations in the charity dictator game were not normally distributed (skewness and kurtosis test: χ² = 992.86, p < .001); therefore, nonparametric analysis was performed (D’Agostino, Belanger, & D’Agostino, 1990).

Charity preference

The 50:50 conditions were included to identify any preference bias for donating to either the BHF or CRUK. There was no significant difference in either the amount donated to the BHF (M = £0.81, SD = £0.84) or CRUK (M = £0.86, SD = £0.75; Wilcoxon signed-rank test: z = −1.26, p = .21) or in terms of the frequency of donations to the BHF (number donated = 13) or CRUK (number donated = 16; chi-square goodness of fit test, χ² = 0.31, p = .58). There was also no difference in the average amount donated between the 20% and 80% condition and 80% and 20% charity conditions (Wilcoxon rank sum test: z = −1.42, p = .15).

Preference based on distribution of resources

As the 50:50 condition showed no charity preference effect, we compared the amount donated when a jar was 20% full compared with when a jar was 80% full, regardless of charity. These analyses were run on the data from 105 participants in the 80:20 and 20:80 trials with those in the 50:50 trials removed. The charity with the least distribution of resources (20%) received significantly more money (M = £1.64, SD = £1.08) than the charity with a greater distribution of resources (80%; M = £1.34, SD = £0.98; Wilcoxon signed-rank test: z = 2.89, p < .01; see Table 2). Over the course of the study, the charity with the least distribution of resources received an additional £30.12 (see Table 2).

Table 2.

Average and Total Donations to Least and Most Supported Charities in Studies 1 and 2.

		M (SD)	Sum
Study 1	Least supported charity (20%)	£1.64 (£1.08)	£120.31
Study 1	Well supported charity (80%)	£1.34 (£0.98)	£90.19
Study 2	Least supported charity (20%)	£1.25 (£0.91)	£151.45
Study 2	Well supported charity (80%)	£0.75 (£0.72)	£95.25

Effects of observability

There was no significant effect of observability between donations in the public (M = £1.78, SD = £1.54) and private conditions (M = £2.02, SD = £1.59; Wilcoxon rank sum test: z = 0.79, p = .43). Thus, the overall results showed a preference to donate to the least supported charity compared with the more supported charity, and there was no effect of observability on the participants’ donations.

Study 2

This study aims to replicate the findings of Study 1 regarding preference for the charity with least support, but with three design changes. First, to check that the lack of an observability effect was not due to a relatively small sample size, a large number of participants were recruited. Second, we reduce the amount earned in the earning task to £3 due to financial constraints. Third, we did not include the 50:50 condition because Study 1 showed no charity preference bias.

Method

Participants

A total of 132 students from the University of Nottingham were recruited, via convenience sampling. Eight cases were dropped due to invalid responses (i.e., allocating more money than they earned) leaving a final sample of 124 students. It contained 55 males (44.4%), 67 females (54%), and two who wished not to disclose (1.6%). Participants completed the study in groups of between three and 10 (M = 6.66, SD = 1.88), with each group being randomly allocated to a condition.

Design

The study used a similar design to Study 1 with a 2 (observability: public vs. private) x 2 (degree of support: 20:80 vs. 80:20) x 2 (distribution of resources across 80:20) mixed design. The study consisted of three sequential stages: (a) money earning task, (b) charity dictator game (same as Study 1), and (c) an online questionnaire. The questionnaire collected gender, educational attainment, past charitable behavior, level of risk, level of trust, and the reluctant altruism scale. The present research does not focus on these measures and they are not discussed further.

Procedure

First, participants completed the money earning task where they were told that to earn £3, they must make 200 “zx” responses (zx responses: M = 966.14, SD = 241.7). All participants made over 200 zx responses and earned £3. The rest of the procedure is identical to Study 1.

Results

Donations in the charity dictator game were not normally distributed (skewness and kurtosis test: χ² = 27.46, p < .001), so nonparametric methods were used.

Charity preference

There was no significant difference in either the amount donated to the BHF (M = £0.99, SD = £0.83) or CRUK (M = £1.00, SD = £0.84; Wilcoxon signed-rank test, z = 0.01, p = .99) or in terms of the frequency of donations to the BHF (number donated = 30) or CRUK (number donated = 33; chi-square goodness of fit test, χ² = 1.8, p = .18). There was also no difference in the average amount donated between the 20% and 80% and 80% and 20% charity conditions (Wilcoxon rank sum test, z = −1.12, p = .26).

Preference based on distribution of resources

Table 2 shows that the charity with the least distribution of resources received significantly higher average donations (M = £1.25, SD = £0.91) than the charity with the greater distribution of resources (M = £0.75, SD = £0.72; Wilcox signed-rank test, z = 4.76, p < .001). The charity with the least distribution of resources received £56.20 more than the charity with the greater distribution of resources.

Effects of observability

As in Study 1, there was no effect of observability on average amounts donated (z = 0.11, p = .91; private: M = £2.01, SD = £1.08 vs. public: M = £1.97, SD = £1.11).

Combined analysis of Study 1 and Study 2

Data from Studies 1 and 2 (N = 266) were combined to see whether a larger sample size could help identify the observability effect and to control for demographic variables (gender and past level of helping¹) that may be affecting charitable donations. First, a multilevel negative binomial regression was performed to see whether the distribution of resources and observability predicted the amount donated to charity. The multilevel model allowed us to account for the within-subject component of the design (participant can distribute their money across the 20% and 80% full jars). A negative binomial link was specified to account for the overdispersion present in the data (conditional overdispersion = 2.49). Table 3 shows significantly higher donations went to the least supported charity (incident rate ratio [IRR] = 0.68, SE = 0.1, p < .01) and no effect of observability was found (IRR = 1.00, SE = 0.15, p = .97). A multilevel logistic regression was conducted to explore whether the distribution of resources and observability predicted the frequency of donations to charity. Table 4 shows significantly more people donated to the least supported charity (OR = 0.47, SE = 0.13, p < .01) while no effect of observability was found (OR = 1.23, SE = 0.13, p = .62).

Table 3.

A Multilevel Negative Binomial Regression Model Exploring the Effect That the Underdog Effect Has on Donations Made to Charity During Studies 1 and 2 (N = 235).

	Model 1	Model 2	Model 3
	IRR (SE)	IRR (SE)	IRR (SE)
20% vs. 80% charity (0 = 20%, 1 = 80%)	0.68 (0.1)**	0.68 (0.1)**	0.68 (0.1)**
Observability (0 = private, 1 = public)		1.00 (0.15)	1.00 (0.15)
Gender (0 = male, 1 = female)			1.08 (0.16)
Past level of helping			0.97 (0.06)
Study 1 or Study 2 (0 = Study 1, 1 = Study 2)			1.28 (0.19)

Note. Model 1: χ²(1) = 7.02, p < .01; Model 2: χ²(2) = 7.02, p < .05; Model 3: χ²(4) = 7.66, p = .060. IRR = incident rate ratio.

p < .05. **p < .01.

Table 4.

A Multilevel Logistic Regression Model Exploring the Effect That the Underdog Effect Has on the Frequency of Donations to Charity During Studies 1 and 2 (N = 235).

	Model 1	Model 2	Model 3
	OR (SE)	OR (SE)	OR (SE)
20% vs. 80% charity (0 = 20%, 1 = 80%)	0.47 (0.13)**	0.47 (0.13)**	0.47 (0.13)**
Observability (0 = private, 1 = public)		1.14 (0.47)	1.23 (0.51)
Gender (0 = male, 1 = female)			1.92 (0.79)
Past level of helping			0.94 (0.15)
Study 1 or Study 2(0 = Study 1, 1 = Study 2)			1.71 (0.71)

Note. Model 1: χ²(1) = 7.17, p < .01; Model 2: χ²(2) = 7.26, p < .05; Model 3: χ²(5) = 10.87, p = .054. OR = odds ratio.

p < .05. **p < .01.

Discussion

Both Study 1 and Study 2 extended the previous literature by showing that in a more naturalistic setting for crowdsourcing in which prior charitable donations are simultaneously observable, participants preferred to donate to the charity with the least prior support. Previous work by Borgloh et al. (2013) showed that when participants could review total annual revenue secured by charities, they preferred to give to the poorer charity. Thus, we replicated the preference for a relatively underresourced charity in a different funding context.

The findings of Studies 1 and 2 also suggest that observability did not influence donating to a charity. This is at odds with the majority of the literature (Bereczkei, Birkas, & Kerekes, 2007; Haley & Fessler, 2005). However, the sample sizes were relatively l. As such, we combined the data from Studies 1 and 2 giving a total N of 266 which is larger than sample sizes reported in other studies that found observability effects (Ariely et al., 2009, N = 161; Kataria & Regner, 2014, N = 185); however, the nonsignificant effect remained.

Study 3

Studies 1 and 2 showed that participants gave significantly higher amounts to the charity with the least prior support. It is possible that participants were donating to the less well-supported charity to maximize the impact of their donations, or due to the Underdog Effect. In this study, we try to tease apart these two accounts. To do this, an analog of a threshold public good game is used where the charity will only receive donated money if a specified threshold is achieved (de Hoop, van Kempen, & Fort, 2012). The introduction of the threshold should only change the behavior of a donor motivated to maximize impact versus support the underdog. Before the presence of a threshold, impact donors should give to the charity that has the least support to maximize any increase in the charity’s revenue (Borgloh et al., 2013). For an impact donor, when the amount of support already given makes the specified threshold for a charity achievable, their donation will have a greater impact with respect to achieving the goal, than when the amount of support already given means the specified threshold is far from achievable. In contrast, donors motivated to help the underdog should be more likely to donate to the charity with the least prior support, even though their donation is likely to have little impact on the charity achieving its specified threshold. To test this, we set up a study whereby five schools all needed to reach the same specified threshold, but they varied in how close they were to achieving it. Any donor population will be heterogeneous and made up of both underdog and impact donors, among others. We can, therefore, estimate those who are impact donors and those who are underdog donors by identifying those who give the most to the school that is closest to achieving its specified threshold (impact donors) versus those who give to the school that is furthest away from the specified threshold (underdog donors). Thus, we can examine the relative proportion of impact versus underdog donors.

Method

Sample

A total of 206 participants from the University of Nottingham were recruited via convenience sampling to take part in an online study. The final sample was 184 participants, as 22 cases were dropped due to missing data on key items (e.g., a discrete choice task). The final sample consisted of 78 males (42.3%), 105 females (57.1%), and one (0.6%) who preferred not to disclose. Remuneration was made via a conditional lottery mechanism (see Fischbacher et al., 2001, for details); participants had a one in 10 chance of having their responses to the decision task remunerated. Participants earned up to £10 that they either could keep or donate some, none, or all of the money to a local school in the form of book tokens.

Design and procedure

Participants were shown a hypothetical crowd fundraising profile describing five schools’ fundraising efforts to raise £130 to buy 10 fruit trees to plant on its school grounds (based on a real-life campaign²). Participants were informed that five schools were running exactly the same campaign but had raised different amounts of money, and at that point, all schools had only 2 weeks left to raise the remaining amount (see Table 5). One school (School B) had the least previous support (£10) and was the furthest from the specified threshold of £130, meaning no single donor could help them achieve it. One school (School C) was £10 away from the specified threshold, meaning one donor could make up the difference. Thus, contributions made to schools B and C are key to establishing the relative proportions of underdog and impact donors.

Table 5.

The Amount Raised by the Five Different Schools Over the First Week of the Campaign.

The amount raised so far by the different schools are:
School A has raised £136.50
School B has raised £10.00
School C has raised £120.00
School D has raised £65.00
School E has raised £110.00

Participants were informed that if a school did not raise the required £130 within a 2-week period, they would not be able to buy the 10 fruit trees. From the presentation of stimuli, there is no reason to believe that one school is less deserving of a donation than another. For example, participants were informed that the participating schools were identical in size, Ofsted reports,³ and academic attainment. From this crowdfunding project, participants had to make decisions: (a) decide whether they would be willing to donate to one of the five schools (a discrete choice task), and if they decided to donate to one of the schools, then (b) decide how much of a £10 endowment they would like to give or keep for themselves.

Results

Ninety-one participants out of the 166 participants who donated to a school chose to donate (54.82%; see Table 6 and Figure 2) to the school with the least previous support and furthest away from the threshold. A chi-square goodness of fit test, comparing the frequency of donations across the five schools against an expected frequency of chance (20%), indicated significant differences between the schools, χ²(4) = 144.81, p < .01. The school that had the least support and was furthest from the specified threshold (School B) received a significantly higher number of donations than the school just below the threshold (School C, z = 8.18, p < .001) and all other schools (Bonferroni corrections were applied; School A: z = 12.96, p < .001; School D: z = 6.75, p < .001; School E: z = 11.39, p < .001; Sharpe, 2015). The total donated to the school that had raised the least (School B) was £688, which is more than double any other school’s total donations (see Table 6 and Figure 2). However, School B also had the second lowest mean amount donated (M = £7.64, SD = £3.07), although a Kruskal–Wallis test showed there was no significant difference in the average levels of giving across the five schools, H(4) = 6.06, p = .20.

Table 6.

Average and Total Donations to the Five Schools.

Schools target (£130.00)	Number who donated	M (SD)	Total donations
School A (£136.50)	4	£9.63 (£0.75)	£38.50
School B (£10.00)	91	£7.64 (£3.07)	£688.00
School C (£120.00)	26	£8.69 (£2.20)	£226.00
School D (£65.00)	35	£8.34 (£2.63)	£292.00
School E (£110.00)	10	£6.60 (£3.31)	£66.00

Figure 2.

The total frequency and total amount (£) participants donated to each school.

Discussion

The introduction of the threshold indicated that most donors had a preference to contribute to the school with the least previous support and furthest away from the specified threshold. These are underdog donors. A smaller proportion donated to the school nearest the threshold and these are impact donors. However, while this effect was found for the frequency of donations, it was not found for the average amount donated. We offer two explanations of why this was the case. First, this study uses “house money” rather than earned endowment leading to higher donations overall. Second, donors who want to maximize their impact must give £10 to push School C to the threshold, while donors wishing to support the least supported school cannot push it over the threshold with their contribution; hence, they have more choice about the amount to donate.

Study 4

Study 3 provides some evidence that the preference for the least supported charity is due to the Underdog Effect. However, the Underdog Effect implies that there is an assumption that other potential donors are less willing generally to donate to the underdog. Thus, this study presents exactly the same scenario used in Study 3 but as well as asking participants which school they would choose and how much they would donate, we ask participants which school they believe others would choose, and how much they expect others would be willing to donate. Open-ended free-response questions were also asked to assess why the participant chose to donate to the school they selected and why they expected other people to donate to the school they selected. If the Underdog Effect drives the preference for the least supported school, the open responses should include terms referring to its relative disadvantage (i.e., “has the least”; Vandello et al., 2007) whereas, if donations are due to impact donating, responses should include making a difference (e.g., “biggest impact”; Duncan, 2004). We choose a free-response format to avoid creating any demand characteristics and constraining participants within theoretical frameworks.

Method

Sample

Sixty Nottingham University students were recruited through convenience sampling.

Design and procedure

Participants were asked to read the same crowd fundraising scenario used in Study 3. They were then asked to answer the following six questions: (a) “what school do you think most people would give to?” (b) “how much do you think others would give?” (c) “why do you think most people would give to that school?” (d) “what school would you give to?” (e) “how much would you give?” and (f) “why would you give to this school?”

Results

Four participants did not respond to either “what school do you think most people would give to?” (n = 1) or “how much do you think others would give?” (n = 3); therefore, they were excluded from those analyses. Interrater reliability for the deductive content analysis on the qualitative responses was moderate (κ = .58, z = 8.26, p < .001) between the two raters. All disagreements were discussed between the two raters until an agreement was reached.

School preference

When asked about which school they believe others would choose and which school they would choose themselves, the largest group of participants (28 out of 59) showed a preference for the school with the least support and furthest from the threshold (see Table 7). A chi-square goodness of fit tests on the frequency of choices in terms of the school they believe others would choose, χ²(4) = 38.37, p < .01, and for which school they would choose themselves, χ²(4) = 49.67, p < .01, were significant. A comparison across schools showed that more participants donated to the least supported school (School B) compared with the school just below the threshold (School C; z = 4.42, p < .05), and more participants expected that others would also donate to the least supported school compared with the school just below threshold (z = 4.73, p < .05). Furthermore, participants donated more money to the least supported school than any other school (School A: z = 5.67, p < .05; School D: z = 6.35, p < .05; School E: z = 7.13, p < .01) and expected others to do the same (School A: z = 3.65, p < .05; School D: z = 5.69, p < .05; School E: z = 6.05, p < .05). These results replicate Study 3 and indicate that the Underdog Effect has a normative component.

Table 7.

Average and Total Donations to the Five Schools Across What Most People Would Do and What the Participant Would Do.

Schools target (£130.00)	What would most people do?			What would you do?
Schools target (£130.00)	No. who donated	M (SD)	Total donations	No. who donated	M (SD)	Total donations
School A (£136.50)	11	£24.18 (£58.40)	£266.00	7	£8.71 (£5.53)	£61.00
School B (£10.00)	28	£15.16 (£19.88)	£424.50	33	£11.23 (£9.99)	£370.50
School C (£120.00)	8	£9.19 (£4.24)	£73.50	12	£7.58 (£3.65)	£91.00
School D (£65.00)	5	£16.40 (£27.42)	£82.00	5	£15.40 (£27.76)	£77.00
School E (£110.00)	4	£12.50 (£8.66)	£50.00	3	£10.00 (£8.66)	£30.00

Qualitative analyses of relatively poor school preference

Responses to “why do you think most people would give?”

Of those who gave to the most supported school (School A: already over the threshold), the majority (eight out of 12; 67%) focused on the fact that it had raised the most money (see Table 8). Whereas those who donated to the least supported school (School B) were concerned most with the relative financial disadvantage the school had (24 out of 30; 80%). For example, one respondent wrote, “It is furthest away from the target.” Those who donated to the school just below the threshold (School C) did so to (a) make a difference (two out of eight; 25%) and (b) to help it reach its target (eight out eight; 100%). One participant wrote, “by donating the final £10 I feel like I am making a difference.”

Table 8.

Number and Percentage of Donors Whose Qualitative Responses Indicated Underdog, Impact Donor, or Equality Motivations Driving Their Choice of School.

	Theories
	Underdog	Impact donor		Equality	Quality signaling
Constructs	Relative disadvantage	Making a difference	Easy reach of target	Equality	Quality signaling
Examples	Least, fewer, etc.	Difference, change, etc.	Achieve, closest, etc.	Equal, same, etc.	Most raised
Why would most people give? (N = 59)
School A (n = 12)	0 (0%)	0 (0%)	0 (0%)	0 (0%)	8 (67%)
School B (n = 30)	24 (80%)	0 (0%)	0 (0%)	1 (3.33%)	0 (0%)
School C (n = 8)	0 (0%)	2 (25%)	8 (100%)	0 (0%)	1 (13%)
School D (n = 5)	0 (0%)	0 (0%)	3 (60%)	0 (0%)	0 (0%)
School E (n = 4)	0 (0%)	0 (0%)	1 (25%)	0 (0%)	0 (0%)
Why did you give to this school? (N = 60)
School A (n = 7)	0 (0%)	0 (0%)	0 (0%)	0 (0%)	2 (29%)
School B (n = 33)	22 (67%)	1(3%)	0 (0%)	4(12%)	0 (0%)
School C (n = 12)	0 (0%)	1 (8%)	10 (83%)	0 (0%)	1 (8.33%)
School D (n = 5)	0 (0%)	0 (0%)	3 (60%)	0 (0%)	0 (0%)
School E (n = 3)	0 (0%)	0 (0%)	1 (33%)	0 (0%)	0 (0%)

Note. The words used to record relative disadvantage were least/fewer/lower/less/furthest. The words used to record the construct impact were difference/change/impact/worthwhile. Ease of target was measured by achieve/closest/reach. Equality was captured by equal, same, and similar. Quality signaling was captured by the term raised the most.

Responses to “why you gave to this school?”

Choosing the most supported school (School A) was partly explained by quality signaling, with two out of seven (29%) focusing on the fact that it had raised the most. The dominant response for giving to the least supported school (School B) was due to its relative disadvantage (22 out of 33; 67%). For example, one respondent said, “They are furthest away from the total so I would want to help them more.” Other motives for giving to the least supported school were a desire to make things equal (four out of 33; 12%) and wanting to make a difference (one out of 33; 3%). The main reason for donating to the school just below the threshold (School C) was the ease with which it could reach its target (10 out of 12; 83%). Only one respondent mentioned the desire to make an impact (one out of 12; 8.33%).

Discussion

These results replicated the effects found in Study 3, that most people choose to donate to the least supported school compared with any other school. The open responses provide support for the idea that participants gave to the least supported school because they were the underdog at a relative disadvantage. A small number of participants (four out of 33; 12%) in response to why they, not most others, would give to the least supported school suggested they were motivated by equality so were donating to School B to reduce its shortfall in donations. This indicates that there are competing explanations for supporting the least supported school, with the dominant motive being the Underdog Effect. An unusual finding in this data concerns the donations made to School A, which has already reached its target. Participants’ qualitative responses revealed that people took the amount that “school A” had already raised as a sign of its quality. This motivation is consistent with Vesterlund’s (2003) idea that past contributions to charity can be perceived by potential donors as a sign of the charity’s quality and thus make it appear more attractive for them to donate to it.

One caveat to the findings of the qualitative analysis is that self-report responses are subject to bias (i.e., demand characteristics). However, the strength of Study 4 is that it allows us to move beyond inferring donor motivations from behavioral decisions, to directly recording stated motivations. Indeed, the results of Study 4’s choice task replicate those of Study 3; thus, we can have some confidence in the validity of the motivations reported.

General Discussion

In the absence of any specified threshold, or when a threshold is a long way from being achieved, the charity with least support benefits by receiving a higher level of average donations or a higher frequency of donations. This finding leads to the least supported charity receiving a nontrivial increase in total donations compared with the better supported charity. This result is important as many online fundraising campaigns set a threshold to be attained.

Two possible explanations for the preference of selecting the charity/organization with the least support have been suggested (i.e., “Impact Model” and the “Underdog Effect”). The results provide support for the heterogeneity of participants’ preferences, with the Underdog Effect being the most likely explanation for giving to a charity with least support.

However, other studies find that the charity with greater support is preferred. Why is this? Studies that find a preference for giving to the more supported charity have focused on charity campaigns that may have an immediate benefit to the donors. For example, the Frey and Meier (2004) study focused on students contributing to a student welfare scheme, the Shang and Croson (2009) study examined contributions from supporters/listeners to a local public radio station, and Martin and Randal (2008) explored support to an art gallery by visitors. In all these examples, the donor could immediately benefit and as such may be more inclined to help when they see others contributing as ultimately this donation will also help them (i.e., they receive the hardship payments, their public radio station stays open, and their gallery stays open).

In contrast, our study and the Borgloh et al. (2013) study do not have this immediate benefit to the donor. In the Borgloh et al. (2013) study, the charities were described in terms of general categories (e.g., medical research, etc.), and as such, the participants would be less likely to know whether they could personally benefit. In our studies, the charities were either specific medical charities (cancer or heart disease), from which young students are unlikely to have any immediate benefit. Similarly, in the school’s crowdsourcing study, the participants will have no immediate benefit from donating to the schools. So it may be the case that conditional cooperation operates when there is an immediate personal benefit favoring the better supported charity, and the “Underdog Effect” is observed when there is no immediate personal benefit (Kim et al., 2008). Therefore, we hypothesize that a charity based “Underdog Effect” will be maximal when (a) there is no threshold/or a threshold that is unattainable by a single donation; and (b) when donors do not directly benefit from their donations. This will be the case for a charity that donors see as deserving but disadvantaged with respect to resources.

Implications

This article is important because it shows that when people have social information of the levels of support a charity has received in the past and can compare this information across multiple charities, they prefer to donate to the least supported charity. This finding suggests that fundraisers should take care to present their campaigns in real world and online contexts where their charity will appear less supported compared with their competitors. Conversely, if a charity is well supported and cannot change the real world or online contexts in which it presents itself, it might benefit from highlighting aspects of its current situation that make it appear at an undeserved disadvantage relative to some of its competitors.

One way this strategy could be achieved is by developing an underdog biography (Nagar, 2017). The work of Paharia, Keinan, Avery, and Schor (2011) shows that businesses that managed to develop an underdog biography enjoyed increased consumer purchasing and greater brand loyalty. Theoretically, this article demonstrates that the Underdog Effect is an important and underexplored motivation that charities could target to increase support and enhance their revenues.

Conclusion

The current research shows that when people are given the choice between a charity with substantial prior support compared to one with relatively little prior support, they prefer to donate to the charity with the least prior support. The main motivating factor behind this finding is the “Underdog Effect” where people have a preference for the charity at an undeserved disadvantage. However, the “Underdog Effect,” as a motivating factor that drives charitable decisions may be limited to contexts where charities are easily comparable and equally deserving.

Further research is needed to explore how factors that are known to influence donation decisions like charities’ effective use of resources, their impact on the cause, their perceived worthiness, and their reputation moderate the “Underdog Effect” (Bekkers & Wiepking, 2011). In particular, Michniewicz and Vandello (2013) showed that unfairly disadvantaged job applicants were rated by participants as more attractive, compared with fairly disadvantaged job applications. Thus, future research could examine the additional impact of the fairness of the disadvantage in accentuating the “Underdog Effect.”

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Economic and Social Research Council (Award No. 1491185), by a PhD award to Alex Bradley.

Notes

ORCID iD

Alex Bradley

Author Biographies

Alex Bradley received a Bsc from the University of Warwick, an Mres from the University of Sussex before completed his PhD at the University of Nottingham. He is now a teaching associate at Univeristy of Nottingham.

Claire Lawrence is an associate fellow of the British Psychological Society (www.bps.co.uk), fellow of the International Society for Research in Aggression (www.israsociety.com), founding member of the British Society for the Psychology of Individual Differences (www.bspid.org.uk/) and member of the Society for Personality and Social Psychology ().

Eamonn Ferguson is a chartered health and occupational psychologist, a fellow of the Royal Society for Public Health, an associate fellow of the British Psychological Society, and co-founding president of the British Society for the Psychology of Individual Differences ().

References

Andreoni

Petrie

(2004). Public goods experiments without confidentiality: A glimpse into fund-raising. Journal of Public Economics, 88, 1605-1623. doi:10.1016/S0047-2727(03)00040-9

Ariely

Bracha

Meier

(2009). Doing good or doing well? Image motivation and monetary incentives in behaving prosocially. American Economic Review, 99, 544-555.

Ariely

Bracha

Meier

(2007). Doing good or doing well? Image motivation and monetary incentives in behaving prosocially. Retrieved from https://www.aeaweb.org/articles?id=10.1257/aer.99.1.544

Bekkers

Wiepking

(2011). A literature review of empirical studies of philanthropy: Eight mechanisms that drive charitable giving. Nonprofit and Voluntary Sector Quarterly, 40, 924-973. doi:10.1177/0899764010380927

Bereczkei

Birkas

Kerekes

(2007). Public charity offer as a proximate factor of evolved reputation-building strategy: An experimental analysis of a real-life situation. Evolution and Human Behavior, 28, 277-284. doi:10.1016/j.evolhumbehav.2007.04.002

Borgloh

Dannenberg

Aretz

(2013). Small is beautiful—Experimental evidence of donors’ preferences for charities. Economics Letters, 120, 242-244. doi:10.1016/j.econlet.2013.04.011

Butt

Shah

(2012). Why technology and innovation are key to increasing charity donations. Retrieved from http://www.theguardian.com/voluntary-sector-network/2012/nov/29/technology-innovation-increase-charity-donations

CrowdFunder. (2013). Fruitshare project. Retrieved from http://www.crowdfunder.co.uk/campaign/fruitshare

D’Agostino

R. B.

Belanger

D’Agostino

R. B.

Jr. (1990). A suggestion for using powerful and informative tests of normality. The American Statistician, 44, 316-321.

10.

de Hoop

van Kempen

Fort

. (2012). Do cash transfers crowd out community investment in public goods? Lessons from a field experiment on health education. Nonprofit and Voluntary Sector Quarterly, 41, 232-256. doi:10.1177/0899764011401414

11.

Dufwenberg

Muren

(2006). Generosity, anonymity, gender. Journal of Economic Behavior & Organization, 61, 42-49. doi:10.1016/j.jebo.2004.11.007

12.

Duncan

(2004). A theory of impact philanthropy. Journal of Public Economics, 88, 2159-2180. doi:10.1016/S0047-2727(03)00037-9

13.

Fischbacher

Gachter

Fehr

(2001). Are people conditionally cooperative? Evidence from a public goods experiment. Economics Letters, 71, 397-404.

14.

Frey

B. S.

Meier

(2004). Social comparisons and pro-social behavior: Testing in a field experiment “conditional cooperation.” The American Economic Review, 94, 1717-1722.

15.

Goldschmied

Vandello

(2009). The advantage of disadvantage: Underdogs in the political arena. Basic and Applied Social Psychology, 31, 24-31. doi:10.1080/01973530802659810

16.

Goldschmied

N. P.

Vandello

J. A.

(2012). The future is bright: The underdog label, availability, and optimism. Basic and Applied Social Psychology, 34, 34-43. doi:10.1080/01973533.2011.637726

17.

Haley

K. J.

Fessler

D. M. T.

(2005). Nobody’s watching? Subtle cues affect generosity in an anonymous economic game. Evolution and Human Behavior, 26, 245-256. doi:10.1016/j.evolhumbehav.2005.01.002

18.

Jones

Linardi

(2012). Wallflowers doing good: Field and lab evidence of heterogeneity in reputation concerns. Retrieved from http://www.linardi.gspia.pitt.edu/wp-content/uploads/2012/07/Wallflowers_JonesLinardi2012.pdf

19.

Kataria

Regner

(2015). Honestly, why are you donating money to charity? An experimental study about self-awareness in status-seeking behavior. Theory and Decision, 79, 493-515.

20.

Kim

Allison

S. T.

Eylon

Goethals

G. R.

Markus

M. J.

Hindle

S. M.

McGuire

H. A.

(2008). Rooting for (and then abandoning) the underdog. Journal of Applied Social Psychology, 38, 2550-2573. doi:10.1111/j.1559-1816.2008.00403.x

21.

Martin

Randal

(2008). How is donation behaviour affected by the donations of others? Journal of Economic Behavior & Organization, 67, 228-238. doi:10.1016/j.jebo.2007.08.001

22.

Mason

D. P.

(2016). Recognition and cross-cultural communications as motivators for charitable giving a field experiment. Nonprofit and Voluntary Sector Quarterly, 45, 192-2014. doi:10.1177/0899764015576408

23.

Michniewicz

K. S.

Vandello

J. A.

(2013). The attractive underdog: When disadvantage bolsters attractiveness. Journal of Social and Personal Relationships, 30(7), 942-952.

24.

Nagar

(2017). Support for the underdog brand biography: Effects on consumer attitude and behavior. Journal of Marketing Communications, 1-17. doi:10.1080/13527266.2017.1356349

25.

Paharia

Keinan

Avery

Schor

J. B.

(2011). The underdog effect: The marketing of disadvantage and determination through brand biography. Journal of Consumer Research, 37(5), 775-790.

26.

Shang

Croson

(2009). A field experiment in charitable contribution: The impact of social information on the voluntary provision of public goods. The Economic Journal, 119(540), 1422-1439.

27.

Sharpe

(2015). Your chi-square test is statistically significant: Now what? Practical Assessment. Research & Evaluation, 20, 1-10.

28.

Shirai

(2017). Underdog effects: The role of consumption domain and retail crowding. Journal of Consumer Marketing, 34, 384-392. doi:10.1108/JCM-07-2016-1872

29.

Soetevent

A. R.

(2005). Anonymity in giving in a natural context—A field experiment in 30 churches. Journal of Public Economics, 89, 2301-2323. doi:10.1016/j.jpubeco.2004.11.002

30.

Vandello

J. A.

Goldschmied

N. P.

Richards

D. A.

(2007). The appeal of the underdog. Personality and Social Psychology Bulletin, 33, 1603-1616. doi:10.1177/0146167207307488

31.

van Leeuwen

M. H. D.

Wiepking

. (2013). National campaigns for charitable causes: A literature review. Nonprofit and Voluntary Sector Quarterly, 42, 219-240. doi:10.1177/0899764012467084

32.

Vesterlund

(2003). The informational value of sequential fundraising. Journal of Public Economics, 87, 627-657. doi:10.1016/S0047-2727(01)00187-6