Rubaltelli Asymmetric cost and benefit perceptions in willingness‐to‐donate decisions 2019 Accepted

Examination of the ratings indicated that the average perceived benefit for recipients was higher than the average perceived cost for the donor in the first five scenarios, and the scen

Asymmetric cost and benefit perceptions in willingness-to-donate decisions

Enrico Rubaltelli1,* , Dorina Hysenbelli1, Stephan Dickert2,3, Marcus Mayorga4, Paul Slovic4,5

1 University of Padova

Department of Developmental and Socialization Psychology

via Venezia, 8 – 35131 Padova, Italy

2 University of Klagenfurt

Department of Psychology

Universitätsstr 65-67 – 9020 Klagenfurt, Austria

3 Queen Mary University of London

School of Business and Management

Mile End Campus – London E1 4NS, United Kingdom

Acknowledgement

This material is based upon work supported by the National Science Foundation under Grant Nos 1227729 and 1427414 Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation

Asymmetric cost and benefit perceptions in willingness-to-donate decisions

Abstract

Charitable giving entails the act of foregoing personal resources in order to improve the conditions of other people In the present paper, we systematically examine two dimensions integral to donation decisions that have thus far received relatively little attention but can explaincharitable behavior rather well: the perceptions of cost for the donor and benefit for the

recipients In line with current theories in judgment and decision making, we hypothesize that people weigh these dimensions subjectively and perceive them asymmetrically, consistent with prospect theory Costs for the donor are typically perceived as losses, whereas benefits for recipients are perceived as gains In four studies, we presented several scenarios to participants inwhich both donation amounts (costs) and number of lives helped (benefits) were manipulated while keeping the ratio of costs and benefits constant Results from Studies 1 and 2 showed that willingness to help decreased as donation amounts and number of lives helped increased

Additionally, Studies 3 and 4 provide evidence for a solution to reduce the asymmetry and increase donation amounts as the number of lives at risk increase

(WORD COUNT: 180)

Keywords: Prosocial behavior; donation decisions; cost; benefit; prospect theory

Although people may help others for many different reasons (e.g., Bekkers & Wiepking, 2011), they usually forego personal resources (i.e., time or money) to improve someone else’s wellbeing It seems reasonable to assume that the benefactor will engage in a subjective appraisal

of the degree of benefit that can be achieved with a specific amount of resources The

psychological underpinnings of such an appraisal likely involve both emotional aspects (e.g., sympathy for the victims) as well as more deliberative ones (e.g., impact of the donation), and the comparison between benefits and costs is considered instrumental in establishing whether theact of helping is worth it or not This may especially be the case when personal resources are limited, and the donor needs a compelling reason to give them away In other words, the decisionwhether to help or not is likely influenced by the perceived tradeoff between the resources a donor must give up and her perception of the beneficial effect of the helping intervention The goal of the present paper is to investigate whether the subjective weights attached to the personal cost and benefit of the helping action can lead donors to evaluate lives at risk in a non-normative way (i.e., placing different value on lives depending on the number of individuals at risk;

Dickert, Västfjäll, Kleber, & Slovic, 2012; Slovic, 2007) In line with prospect theory

(Kahneman & Tversky, 1979), we present evidence that donors put more emphasis on the cost ofgiving up an increasing amount of money compared to the benefit of helping increasing numbers

of lives, leading to a decrease in their willingness to help

Perceived cost for the donor and perceived benefit for the recipients

We investigate whether a person’s decision to donate to a particular fund-raising campaign depends on their subjective judgments about the cost they incur, and the benefit provided to the recipients The tradeoff between these two constructs has thus far not received much attention in

research on donation decisions but should be taken into consideration to understand how a donor assesses the utility of a donation

In the current paper, we focus on the monetary cost people face when contributing to a charity organization1, and benefits are conceptualized as the positive difference a donation will make for the people in need People are less willing to donate when they perceive their donation

as a “drop-in-the-bucket,” that is, when a contribution is not seen as making much of a differencetoward solving the problem (Small, Loewenstein, & Slovic, 2007) Recent studies have

suggested that people’s tendency to help the highest proportion of individuals in need, rather thanthe most in absolute numbers, can be interpreted as a way to support the fundraising programs whose perceived effectiveness is bigger (Bartels, 2006; Erlandsson, Bjorklund, & Backstrom,

2014, 2015; Fetherstonhaugh, Slovic, Johnson, & Friedrich, 1997) In addition, helping can induce people to experience positive emotions because they are responsible for having a positive impact on others’ lives (Andreoni, 1990; Dickert, Sagara, & Slovic, 2011)

However, people may also experience negative affective reactions when spending money in general (Prelec & Loewenstein, 1998; Raghubir, 2006; Knutson, Rick, Wimmer, Prelec, & Loewenstein, 2007; Chatterjee & Rose, 2012), and also in decisions involving helping other people (Rubaltelli & Agnoli, 2012; Genevsky, Västfjäll, Slovic, & Knutson, 2013) Consistent with these findings, changing the frame used to present a donation appeal can influence people’s emotional reactions and, consequently, their willingness to help For instance, Breman (2011) demonstrated in a field study that people were more willing to donate when they were asked to pledge to donate but their credit card was charged at a later stage rather than immediately The explanation for this result was that people tend to discount a cost when it is postponed in time

As a result, they are more willing to help when the cost of the donation is not charged

immediately Sussman, Sharma and Alter (2015) found that framing a donation as exceptional (i.e., the only chance to help Charity X) increases willingness to help compared to framing it as

an ordinary expense (i.e., this year’s chance to help Charity X)

We hypothesize that costs and benefits are represented differently by donors, such that changes in costs may not be matched by changes in benefits when the number of lives at risk increases According to egalitarian moral perspectives and related forms of utilitarianism all lives should be valued equally (e.g., Baron & Szymanska, 2011; Sinnot-Armstrong, 2011) and additional lives at risk could be valued even higher if their loss threatens the survivability of an entire group (Dickert, Västfjäll, Kleber, & Slovic, 2015; Slovic, Fischhoff, & Lichtenstein, 1982).2 However, despite agreeing that every life should be valued equally, people often fail to anticipate the decrease in willingness to give when the cost for the donor increases Consistent with this notion, research on donation decisions and the underlying psychological processes demonstrated that people are more willing to give when just one (or only a few) lives are at stake

or when donation amounts are not too high (Kogut & Ritov, 2005a, 2005b; Small, Loewenstein,

& Slovic, 2007; Cameron & Payne, 2011; Rubaltelli & Agnoli, 2012) While this research has focused on psychological mechanisms underlying donation decisions, the subjective perception

of cost and benefit has not been measured with regard to increasing lives at risk

Based on the existing literature on charitable giving, the proportion of participants who decide to make a donation is expected to decrease as the amount of money required (i.e., cost) and the number of lives helped (i.e., benefit) increase at the same rate This prediction is

motivated by the properties of the value function of prospect theory, which is steeper in the loss domain than in the gain domain (i.e., showing greater sensitivity to losses than gains; Kahneman

& Tversky, 1979; Tversky & Kahneman, 1992) In the domain of charitable giving, the cost for

the donor falls in the loss domain, since the donor is giving away part of her resources and experiencing negative affect because of this (Rubaltelli & Agnoli, 2012; Genevsky et al., 2013) Conversely, the number of lives helped falls in the gain domain, since the donation allows improving the life of needy individuals As a result, contrary to most literature on prospect theory, in our case the loss is experienced directly by the donors, who donate their own money, whereas the gain is experienced by other people who are at the receiving end of the helping action Nonetheless, we still expect to find results that are consistent with prospect theory

prediction that perceived cost for the donor should follow a steeper increase than perceived benefit for the recipients as the number of lives in need increases.As a consequence, the

disutility of donating more money should be felt more than the utility of saving more lives Therefore, we hypothesize that:

H1: The slope of the donor’s perception of costs will be steeper than the slope of the

perception of benefits for recipients

H2: Participants will become less willing to help as the number of lives at risk and the cost

of helping increase at a constant ratio

Finally, we assume that people’s willingness to help is a function of perceived benefit for therecipients and cost for the donor, and we expect that people’s asymmetric valuation of cost and benefit leads to a point where benefit no longer outweighs the cost At this point, people’s

willingness to help should decline or abate altogether Within a range of values up to which someone is willing to donate money, one would expect willingness to help to remain constant when both cost and benefit increase at the same rate However, this should not be the case if

losses are perceived asymmetrically compared to gains. Based on the above reasoning, we expectthat:

H3: The asymmetric slopes of cost for the donor and benefit for the recipients will predict

participants’ willingness to help

Across four studies, we test these hypotheses and show that the differential perception of cost for the donor and benefit for recipients can predict the decrease in willingness to help as the number of lives at risk increases The first study provides initial evidence for this effect The second study extends the results measuring the cost/benefit tradeoff with a single bipolar

question rather than two separate questions The third study demonstrates that presenting

donation amounts as part of a bundle of different aid programs rather than a lump sum can be an effective way to increase willingness to help by changing people’s perceptions of the cost

relative to the benefit Specifically, each program in the bundle is associated with a smaller donation amount compared to the lump sum condition Similarly, each program in the bundle condition helps fewer lives If people have a differential perception of cost and benefit, then breaking down these two dimensions should reduce the perception of cost with less impact on the perception of benefit Therefore, the goal of Study 3 is to further understand how the

cost/benefit tradeoff impacts donation decisions, while at the same time showing an effective way to influence people’s weighting of the cost dimension Finally, the fourth study examines whether this effect is asymmetrically influenced by splitting the lump sum into several smaller donation amounts or the overall number of lives helped into several smaller groups

Study 1

In Study 1, we presented participants with seven scenarios in which they were asked to donate money to Kenyan children who live in conditions of extreme poverty The amount of money and the number of children varied for each scenario, but the amount donated to a single child (i.e., the ratio between cost and benefit) was kept constant across scenarios.

Method

Participants One hundred and fifty-one participants (43% female; mean age 32 years, range

from 18 to 66) took part in the study They were recruited on Amazon Mechanical Turk,

completed the questionnaire online, and received a fee of $0.40 All participants were located in the United States Amazon Mechanical Turk is commonly used to recruit adult participants for online studies and has been validated by Buhrmester, Kwang and Gosling (2011) and Paolacci, Chandler and Ipeirotis (2010) Since we have conducted donation studies on Amazon

Mechanical Turk in the past, we used the TurkGate system to filter out participants who had already taken part in previous studies (Goldin & Darlow, 2013)

Materials and procedure Participants completed an online questionnaire that was about

6-minute long to complete In the questionnaire, they were told to imagine they had been contacted

by a humanitarian aid organization that was raising funds to help extremely poor people living inthe Kakuma refugee camp in Kenya Participants were informed that their donation would be used to buy food, clothes and medicine for children living in the refugee camp

We presented seven different scenarios within-subjects These scenarios were presented in a different random order for each participant and for each scenario we repeated the same cover story outlined above Participants were instructed to consider each donation request separately and independently from other scenarios presented In each scenario, participants were asked

whether or not they wanted to make a donation and we varied the amount of the donation and thenumber of children helped (see Table 1 below)

After each scenario, participants were asked to answer two questions, which measured how much the donation was perceived as a cost for the donor and as a benefit for the recipients Both

were answered using 7-point scales ranging from 1 (not at all) to 7 (very) After completing all

scenarios, participants provided demographic information and were presented with a debriefing page explaining method, hypotheses, and expected results of the study (full materials are

available in the Supplementary Online Materials)

Table 1 Results

-Donation decisions Most participants (88%) were willing to donate when presented with a

scenario asking them to help two children with a $5 donation, whereas only a few (14%) were willing to donate when the scenario asked them to help ninety children with a $225 donation A substantial proportion of participants (60%) refused to make a donation when the cost reached

$50 for helping 20 children (see Table 2 for the proportion of donations in each scenario)

We ran a multilevel logistic model using R software (version 3.4.3; R Development Core

Team, 2012) and the lme4 package (Bates et al., 2015) to estimate participants’ random effects

Scenario was the within-subject predictor and donation decisions in each of the seven scenarios

served as the dependent variable Results revealed a significant effect of scenario, χ 2 (6) =

243.59, p < 00001 As expected, the percentage of participants who made a donation decreased

as the size of the donation (and the number of children helped) increased (see Figure 1)

Table 2

-Contrast effects showed that, for each scenario, participants’ willingness to help was lower than for the previous one (e.g., less people made a donation in the second scenario than in the

first, and so on; always p < 00001) This shows that the effect we found is already present when

cost is relatively low and not just when it is so high to make the donation amount almost

unrealistic or too high for many people to consider it

Figure 1

-Cost - benefit perceptions Examination of the ratings indicated that the average perceived

benefit for recipients was higher than the average perceived cost for the donor in the first five scenarios, and the scenario factor had a significant effect on both ratings such that they increased

as donation amounts and number of lives at risk increased Consistent with donation decisions, the percentage of people who rated benefit higher than cost decreased as cost and benefit of helping increased (see Table 2) Finally, perceived cost and benefit were significantly correlated

across the seven scenarios, but the correlation was higher for participants who did not donate (r = 55, p < 001) than for participants who did (r = 17, p < 001) The difference between the two correlations was statistically significant (z = 3.84, p = 0001).

In line with these findings, we ran a multilevel linear model with scenario, dimension (cost

vs benefit), and the interaction between scenario and dimension as predictors and participants’ ratings in each scenario as the dependent variable Results revealed a significant effect of

scenario, χ 2 (6) = 1173.06, p < 00001, and a significant effect of the dimensions, χ 2 (1) = 238.57,

p < 0001 Contrast effects showed that ratings were higher in all scenarios compared to the first one (t values = 6.44 or higher, ps < 00001) On average, ratings were higher for perceived benefit than perceived cost (t = 15.45, p = 04) In addition, we found a significant interaction effect, χ 2 (6) = 272.13, p < 00001 (see Figure 2).

Figure 2

-A slope analysis revealed that, although both cost and benefit increased as donation amount

(and number of children helped) increased, the function for perceived cost was steeper (b = 70,

p < 00001) than the function for perceived benefit (b = 23, p < 00001; see Figure 2) Contrast

effects showed that the interaction was always significant when comparing one scenario with the

next one (always p < 00001).

Prediction of donation decisions To assess whether the tradeoff between cost and benefit

predicted participants’ decisions in each of the seven scenarios, we computed the difference between cost and benefit ratings, thus obtaining a score ranging from -6 (benefit much higher than cost) to 6 (cost much higher than benefit) We then ran a final multilevel model with the scenario and the difference score as predictors and donation decisions as the dependent variable

We also controlled for the random effect of participants Based on the Bayesian Information Criterion (BIC) this model was the one showing the best fit (for further details see the

Supplementary Online Materials) Results revealed a significant effect of scenario, χ 2 (6) = 85.19, p < 00001, and a significant effect of the difference score, χ 2 (1) = 72.05, p < 0001

These results are reported in Figure 3 where, for ease of representation, we grouped the scenarios

in three different levels The pattern of the results was the same for all scenarios, namely, people were more likely to make a donation when the benefit clearly outweighed the cost This effect became progressively more extreme as donation amount (and number of children helped)

increased

Figure 3 Discussion

-Study 1 demonstrated that when donation amounts and number of lives helped increased with a constant ratio, perceived cost for the donor and benefit for the recipients did not increase identically Although ratings for both dimensions increased significantly across scenarios, as we expected, participants were more sensitive to increases in the cost for the donor (donation

amounts) than in the benefit for the recipients (number of lives helped) This resulted in an asymmetry between gains and losses such that benefit outweighed cost less as donation amount (and number of children helped) increased As a result, more people decided to make a donation when cost and benefit were low than when they were high Although the ratio between cost for the donor and benefit for the recipients remained objectively the same across scenarios, increases

in cost were perceived as greater than increases in the number of children helped

The present findings are consistent with previous work on scope insensitivity in charitable decision-making (Kogut & Ritov, 2005a, 2005b; Small, Loewenstein, & Slovic, 2007; Dickert, Västfjäll, Kleber, & Slovic, 2012; Västfjäll, Slovic, & Mayorga, 2015) because ratings of benefit

for the recipients showed a kind of all-or-nothing pattern whereby they were quite high in all

scenarios However, scope insensitivity cannot fully explain the results of Study 1 because, at least in our within-subjects randomized design, participants were sensitive to the number of children helped in each scenario Perceived benefit increased significantly as the number of children helped increased, despite the fact that that people became less and less willing to donate.Therefore, the present results can only be fully understood when considering the interplay

between perceived benefit and cost and the consequent impact that the tradeoff between these two dimensions had on people’s decision whether to make a donation or not

A possible explanation for the results of Study 1 could be that we asked for relatively high donation amounts (up to $225) and participants felt that making such a donation would be

outside their financial possibilities However, this interpretation cannot explain the fact that the same pattern of results is found even for the first few scenarios If we only consider the first four scenarios, cost increases from $5 to $50 (while the number of children helped increases from 2 to20) and the donation amounts do not seem unrealistic Even in these scenarios, we found that both cost and benefit increased significantly, but the cost function was significantly steeper therefore impacting the tradeoff between the two dimensions As a result, the percentage of participants who made a donation decreased from 76% to less than 50% The differential

increase in benefit and cost ratings across these scenarios led to a reduction in the difference between the two dimensions that made people less and less willing to donate even when donationamounts were probably not unaffordable (see Supplementary Online Materials for details on the analyses with only the first four scenarios)

Study 2

In Study 1, we measured cost and benefit using two separate questions in order to assess the shape of the gain (number of children helped) and loss (donation amounts) curves across the seven scenarios By using two separate questions, we were able to show that the loss curve (cost)was steeper than the gain curve (benefit) This result was consistent with our second hypothesis

as well as with prospect theory’s value function Therefore, the results of Study 1 supported the conclusion that participants were trading off cost and benefit

An alternative, and more direct way, to assess this tradeoff is to measure cost and benefit perceptions along a single continuum (e.g., a bipolar scale ranging from “donation is much more costly than beneficial” to “donation is much more beneficial than costly”) Study 2 was designed

to investigate whether our hypotheses could be supported when using such a bipolar scale, whichforces participants to trade off cost and benefit when evaluating each scenario If tradeoffs

between cost and benefit influenced donation decisions in Study 1, then we should find the same pattern of results in Study 2 when using a single bipolar scale This would support Hypothesis 2, namely that people increasingly weigh cost as more important than benefit when donation amounts (and the number of children helped) increase at a constant ratio In other words, for the first few scenarios, participants should perceive the donation more as a benefit for the recipients than as a cost to themselves, but this balance should switch as the donation amount increases Conversely, if asking for cost and benefit jointly (as done in Study 2) leads to a different

weighting of these dimensions and different donation decisions, then using a bipolar scale shouldmake less likely that the tradeoff of costs and benefits (and the asymmetry in the way these dimensions are perceived) that we found in Study 1 explains the donation pattern that we

observed in Study 1

Method

Participants Three hundred and sixty-seven participants (40% female; mean age 36 years,

range from 18 to 79) took part in the study They were recruited using the Decision Research subject pool, completed the questionnaire online, and received a payment of $1.00

Materials and procedure Participants completed an online questionnaire that included the

same scenarios presented in Study 1 They were told to imagine they had been contacted by a humanitarian organization that was raising funds to help extremely poor people living in the Kakuma refugee camp in Kenya and that their donation would be used to buy food, clothes and medicine for children living in the camp

Scenarios were presented within-subjects, the presentation order was randomized for each participant and the same cover story was used for each scenario We varied the amount of the donation and the number of children helped as in Study 1 and asked participants whether they

wanted to make a donation or not However, in Study 2, after each scenario and after deciding whether to make a donation or not, participants were asked to judge whether they perceived the donation more as a cost for themselves or as a benefit for the recipients on a 9-point bipolar scale(1 = much more costly than beneficial; 5 = equally costly and beneficial; 9 = much more

beneficial than costly) After these ratings, participants provided demographic information and were presented with a debriefing page explaining the method, hypotheses, and expected results

of the study (full materials are available in the Supplementary Online Materials)

Results

Donation decisions Most participants (73%) were willing to make a donation when

presented with a scenario asking them to help two children with a $5 donation, whereas only a minority (30%) stated they were willing to donate when the scenario asked them to help ninety children with a $225 donation (see Table 3 for the proportion of donations in each scenario)

We ran a multilevel logistic model with scenario as the within-subject predictor and donationdecisions in each of the seven scenarios as the dependent variable (we also controlled for

participants’ random effects) Results revealed a significant effect of scenario, χ 2 (6) = 288.67, p

< 00001 As expected, the percentage of participants who made a donation decreased as the size

of the donation (and the number of children helped) increased (see Figure 4)

-Cost/benefit tradeoff Overall, results showed that participants rated the donation to be more

of a benefit than a cost in the scenarios asking for low donation amounts (and helping a smaller number of children), whereas they rated the donation to be more of a cost than a benefit as the donation amount (and the number of children helped) increased In line with these ratings, the percentage of participants rating the benefit more important than the cost decreased as cost and benefit of helping increased (see Table 4)

Table 4

-We next ran a multilevel linear model with scenario as the predictor and participants’ ratings

on the bipolar scale as the dependent variable Results revealed a significant effect of scenario, χ 2

(6) = 553.12, p < 00001 Contrast effects showed that ratings were lower in all scenarios

compared to the first one (ts = 6.44 or higher, ps < 00001; see Figure 5) Contrast effects showed that comparisons were significant among each pair of scenarios (always p < 00001).

Figure 5

-Prediction of donation decisions We ran a final multilevel model with the scenario and the

tradeoff between cost and benefit as predictors and donation decisions as the dependent variable

We also controlled for the random effect of participants Based on the BIC this model was the one showing the best fit (for further details see the Supplementary Online Materials).Results

revealed a significant effect of scenario, χ 2 (6) = 116.63, p < 00001, and a significant effect of the tradeoff, χ 2 (1) = 201.66, p < 00001 These results are reported in Figure 6 where, for easy of

representation we grouped the scenarios in three different levels The pattern of the results was the same for all scenarios, namely, people were more likely to make a donation when the benefit clearly outweighed the cost This effect became progressively stronger as donation amount (and number of children helped) increased

Figure 6 Discussion

-In Study 2, we asked participants to rate cost and benefit on a single bipolar question, therefore making explicit the tradeoff between these two dimensions We expected that, based onprospect theory, people should perceive the donation more as a cost to themselves than a benefit for the recipients as donation amounts (and number of children helped) increased Results from Study 2 supported this hypothesis For donation decisions, we found a similar pattern of results

as in Study 1 In addition, participants’ ratings on the bipolar scale supported the hypothesis that donations are driven by a tradeoff between cost and benefit Results showed that participants perceived the donation more as a benefit for the children when the donation amount (and the number of children helped) was low, whereas they perceived the donation more as a cost for themselves when the donation amount was high As a consequence, the cost/benefit tradeoff measured on the bipolar scale predicted donation decisions, showing that the decreasing

willingness to give depended on whether participants perceived the donation more as a cost or as

a benefit As in Study 1, when we limited the analyses to the first four scenarios, in which

donation amounts were quite affordable, the results did not change (see Supplementary Online Materials for details)

high donation into several smaller ones, to be paid all at once, leads people to perceive costs as lower relative to benefits and thus be more willing to give For instance, one of the scenarios presented in previous studies asked participants to donate $150 to help 60 children In Study 3,

we tested a condition in which the $150 donation is divided into a bundle composed by several aid programs (e.g., $75 to help 30 children, $50 to help 20 children, and $25 to help 10 children).Testing the effects of bundling gives further insight into understanding how people appraise the cost of a donation When people are presented with a bundle, each aid program asks for a lower donation amount (e.g., $75, $50, and $25) than the overall lump sum required to fund the

corresponding single aid program (e.g., $150)

Note that in Table 2, conditions with these three donation amounts were all judged to have benefits greater than costs In Table 3, the first two donation amounts had positive benefit/loss ratings while the third showed close to equal ratings Thus, we expected that each of the single aid programs in the bundle would be characterized by an advantageous benefit/cost ratio

(perceived benefit is rated equal or higher than perceived cost) than the aid program

corresponding to their overall combination

Finally, the prediction of a higher willingness to help when donors are presented with a bundle rather than a lump sum is also consistent with averaging models of impression formation (Anderson, 1965; Anderson, 1981; Skowronsky & Carlston, 1989; Slovic & Lichtenstein, 1971) These models suggest that the overall reaction towards the aid programs included in the bundle should be more positive than the reaction to the single program solution because of the way people integrate perceived cost and benefit in each aid program In the bundle, each aid program

is characterized by a relatively favorable benefit/cost ratio since the reaction towards the number

of lives helped outweighs the negative reaction induced by the donation amount Conversely, the

cost/benefit ratio is unfavorable in the lump sum solution because benefit does not outweigh cost As a consequence, the average reaction toward the bundle should be construed starting fromthree rather favorable reactions and should be more attractive than the reaction to the overall, single donation amount In addition, if donors’ decisions are influenced by their initial

impression towards the fundraising request (several smaller aid programs more positive than a larger single program), the same effect should be found regardless of participants’ ability to compute the correct overall donation amount in the bundle condition

Method

Participants Six hundred and four participants (50% female; mean age 35 years, ranged

from 19 to 71) took part in this study They were recruited on Amazon Mechanical Turk,

completed the questionnaire online, and received a fee of $0.30 As in previous studies, we used the TurkGate system to filter out participants who had already completed questionnaires on donation decisions (Goldin & Darlow, 2013) All participants were located in the United States

Materials and procedure In Study 3, the questionnaire reported only one scenario and

participants were assigned to one of the four between-subjects conditions The four conditions were created by varying the amount of the donation ($150 vs $165) and the way the donation request was presented (lump-sum vs bundle) The first condition (N = 151) asked for a single donation of $150 to help 60 children and was actually identical to the sixth scenario used in the previous studies The second condition (N = 149) asked participants to make a single donation of

$165 to help 66 children The two bundle conditions were identical to the lump-sum ones in terms of the overall amount of the donation and number of children helped, but we divided both cost and benefit into a series of aid programs Participants were asked to fully fund the bundle of aid programs Therefore, if they agreed to donate, their contribution would have covered the

overall amount of money (i.e., even though several programs were listed in the bundle,

participants could not choose among them) The single aid programs within the bundle were chosen among the scenarios used in previous studies In the first bundle condition (N = 150), the overall donation amount was $150 to help 60 children and participants were presented with a bundle of three different aid programs: $75 to help 30 children, $50 to help 20 children, and $25

to help 10 children (corresponding to scenarios 3, 4, and 5 in previous studies) The aid programswere all helping children living in refugee camps in Kenya, identified as camps A, B, and C The second bundle condition (N = 154) was identical to the first one, but it asked for an overall donation amount of $165 to help 66 children We added an additional aid program to the three presented in the other bundle condition, which asked for a $15 donation to help 6 children in refugee camp D (corresponding to scenario 2 in previous studies) In each condition, participants had to decide whether they wanted to make a donation or not In addition, they were then asked

to rate the perceived cost of the donation as well as the perceived benefit for the recipients on thesame 7-point scales used in previous studies

In the bundle conditions, we also asked participants to answer two manipulation check questions: in the first question, they had to report how many aid programs they were asked to fund; in the second question, they were asked to report the overall donation amount they were asked to contribute (in both cases participants made a choice among four different alternatives) This result is, at least in part, consistent with the literature reported above suggesting that a monetary amount presented in partitioned format can be less clear to process because of

cognitive load and the need for people to engage in a series of arithmetic computations Finally,

we computed a score by summing up the number of correct answers for each participant This score could range from 0 (no correct answers) to 2 (all correct answers) Most participants did

not answer correctly to neither of the two manipulation check questions (83%, N = 251) (full materials are available in the Supplementary Online Materials).

Results

Donation decisions A majority of participants decided to make a donation when presented

with a bundle of aid programs (71% for the bundle totaling $150 and 68% for the bundle totaling

$165), whereas less than half of participants decided to make a donation when presented with a single donation amount (46% for the $150 single donation and 42% for the $165 single donation;see Table 5 for details) A logistic regression with condition (bundle vs single donation request), amount of donation request ($150 vs $165), and their interaction was performed with people’s donation decision as dependent variable Results revealed a significant effect of the bundling

condition, Exp(B) = 1.03, SE = 24, Wald = 4.23, p = 00002 (Nagelkerke = 09) No other effects were significant: Exp(B) = - 19, p = 41 for the donation amount and Exp(B) = 08, p = 82 for

the interaction effect (see, Figure 7)

Table 5 - Figure 7 -

-Cost - benefit perception Examination of the cost and benefit perceptions indicated that

perceived benefit for recipients was rated higher than perceived cost for the donor although this difference appeared to be larger in the bundle conditions than in the single donation conditions Similarly, a greater percentage of participants rated the benefit higher than the cost in the two bundle conditions than in the lump sum conditions (see Table 5) We investigated these

differences with a multilevel linear model with condition (single donation vs bundle), donation amount ($150 vs $165), and dimension (cost vs benefit) with the last factor within-subjects and participants’ ratings as the dependent variable Results revealed a significant difference between

cost and benefit ratings, χ 2 (1) = 19.82, p < 000001, and a significant effect of condition, χ 2 (1) = 24.12, p < 000001 The effect of donation amount was only marginally significant, χ 2 (1) = 3.84,

p = 05 In addition, we also found a significant interaction between dimension and condition,

χ 2 (1) = 10.14, p = 002, as well as a significant interaction between dimension and donation amount, χ 2 (1) = 4.95, p = 03 The interaction between condition and donation amount was not significant, χ 2 (1) = 1.22, p = 27 Finally, the three-way interaction was also not significant, χ 2 (1)

= 1.60, p = 21.

Additional analyses showed that perceived cost for the donor was significantly lower in the

bundle condition compared to the single donation condition (B = 1.03, p < 001) Perceived benefits for recipients were not significantly different in the two conditions (B = 01, p = 91;

Figure 8)

Figure 8

-Mediation analysis We then tested a mediation model in which cost and benefit were the

mediators for the effect of condition (lump sum vs bundle) on donation decisions Analyses wereconducted with R software (R Development Core Team, 2015) and the lavaan package (Rosseel,

2012) Results showed an indirect effect of perceived cost for the donor, Exp(B) = 113, SE = 02, Wald = 6.22, p < 00001, but not of perceived benefit for the recipients, Exp(B) = 001, SE = 01, Wald = 11, p = 91 However, when both mediators were included in the model, the effect of condition was still significant, Exp(B) = 14, SE = 04, Wald = 4.01, p < 00001, indicating a

partial mediation

Figure 9

-Additional analyses We ran a logistic regression to assess whether the way participants

answered to the manipulation check questions had an effect on the results Since these questions

were presented only in the bundle conditions, the model included the donation amount, the manipulation check score, and their interactions as predictors and donation decisions as the

dependent variable (Nagelkerke = 04) Neither the effect of the donation amount, Exp(B) = -

131, SE = 27, Wald = - 49, p = 63, nor the effect of the manipulation check score, Exp(B) = 15.43, SE = 715.50, Wald = 02, p = 98, were statistically significant The interaction was also not significant, Exp(B) = - 15.11, SE = 715.50, Wald = - 02, p < 98.

Finally, we checked the time participants took to complete the survey in order to assess whether those who failed the manipulation checks had spent less time If this was true it could indicate that they were paying less attention to the information they were presented with

Specifically, we ran an analysis of variance with the number of manipulation-check questions answered correctly as the independent variable and survey time as the dependent variable

Results showed that this effect was not significant, F(301,2) = 21, p = 81 Furthermore,

pairwise comparisons showed that none of the three comparisons were significant (p = 84 or

higher) In addition, we ran a logistic regression with condition, donation amount, survey time, all two-way interactions, and the three-way interactions as predictors and donation decisions as the dependent variable (Nagelkerke = 10) This analysis was performed to assess whether some

of the participants who took less time to complete the survey may have behaved differently because of a lack of effort Consistent with the results based on all participants, the results of this

last analysis revealed only a significant effect of condition, Exp(B) = 1.65, SE = 46, Wald = 3.58,

p < 00001 Participants in the bundle conditions were more likely to make a donation Relevant

to this analysis is the fact that completion time was not a significant predictor of donation

decisions, Exp(B) = 005, SE = 003, Wald = 1.73, p = 08 Likewise, none of the interactions were significant (p = 09 or higher).

Study 3 applied an intervention intended to reduce perception of cost for the donor relative

to the perception of benefit In particular, we hypothesized that, in the bundle condition, donors’ initial reactions towards separate aid programs are more favorable than the reaction towards the overall program presented in the lump sum condition Our findings supported this hypothesis showing that, in the bundle condition, participants perceived lower cost for the donor, whereas benefit for the recipients was not significantly different in the two conditions These results are consistent with an explanation based on the asymmetry between gains and losses and the value function (Kahneman & Tversky, 1979) Splitting the overall donation amount (and number of children helped) led to decreased perception of costs because this solution entailed lower losses associated with each program within the bundle At the same time, our manipulation did not influence perceived benefit because, in the gain frame, the function is not as steep as in the loss frame resembling relatively more scope insensitivity as benefits increase Splitting the overall number of children across several aid programs does not seem to change perception of benefit associated with the bundle condition compared to the single aid program condition These resultsprovide further evidence that when perceived benefits for recipients outweigh cost for the donor, people are more willing to help

Study 4

In Study 4, we again examined the different effect of the lump sum versus bundle

presentations, although with lower donation amounts In addition, we added two other conditions

in which we bundled either donation amounts (but not the number of children helped) or number

of children helped (but not donation amounts) The goal was to understand which dimension wasmore effective in producing the bundle effect described in the previous study In other words, we

were interested in assessing whether this effect was asymmetrically influenced by splitting the lump sum cost into several smaller donation amounts or the overall number of children into several smaller groups

Finally, we previously suggested that the prediction of a greater willingness to help when donors are presented with the bundle versus the lump sum should be consistent with averaging models of impression formation (Anderson, 1965; Anderson, 1981; Skowronsky & Carlston, 1989; Slovic & Lichtenstein, 1971) According to our previous studies, people’s willingness to help is more impacted by changes in donation amount than number of lives helped As a result,

in the bundle condition, each aid program is characterized by a favorable benefit/cost ratio since donation amount are all lower than in the lump sum condition The average reaction toward the bundle should therefore be based on three initially rather favorable reactions To this end, in Study 4, we asked participants directly to rate their affective reactions toward the donation request and expected to find that these should be more positive in the bundle condition than in the lump sum condition

Method

Participants Five hundred and fifty-four participants (64% female; mean age 34 years,

ranged from 19 to 86) took part in this study They were recruited on Amazon Mechanical Turk, completed the questionnaire online, and received a fee of $0.30 As in previous studies, we used the TurkGate system to filter out participants who had already completed questionnaires on donation decisions (Goldin & Darlow, 2013) All participants were located in the United States

Materials and procedure In Study 4, participants were assigned to one of the four

between-subjects conditions The four conditions were created by varying how donation amounts and number of lives helped were presented to participants The lump sum and bundle condition were

presented in the same way as in Study 3 However, in Study 4, we changed the overall donation amount and number of children at risk that participants could help with their donation In the lump sum condition (C1; N = 125), participants were asked to make a donation of $50 dollars to help 20 children In the bundle condition (C2; N = 126), participants were asked to fund a bundle

of three different aid programs: $30 to help 12 children, $15 to help 6 children, and $5 to help 2 children In two additional conditions, we manipulated only one of the two dimensions through abundle-like presentation of either donation amount or number of lives helped In the children bundle condition (C3; N = 147), donation amount was presented as an overall amount of $50, whereas lives helped where presented as three groups of 12, 6, and 2 children living in different refugee camps In the money bundle condition (C4; N = 156), participants were asked to help 20 children by giving $30, $15, and $5 to a bundle of three different aid programs In all conditions, participants were told their donations were going to help children in refugee camps in Kenya

In each condition, participants had to decide whether they wanted to make a donation or not

In addition, they were then asked to rate the perceived cost of the donation as well as the

perceived benefit for the recipients on the same 7-point scales used in previous studies Finally,

in all conditions, participants answered questions about how positive or negative a feeling they experienced towards each aid program There was only one question in the lump sum condition, but three different questions in the bundle conditions Participants answered the valence question

on a 5-point scale ranging from “very negative” to “very positive”

Furthermore, in all conditions, we also asked participants to answer three manipulation check questions: in the first question, they had to report how many aid programs they were asked

to fund (1, 2 or 3); in the second question, they were asked to report the overall donation amount they were asked to contribute ($50, $60, $75 or $90); in the third question, participants were