Overhead aversion: Do some types of overhead matter more than others?

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

Contents lists available at ScienceDirect

Journal of Behavioral and Experimental Economics journal homepage: www.elsevier.com/locate/jbee

Overhead aversion: Do some types of overhead matter more than others? a,

Javier E. Portillo *, Joseph Stinn a b

T

b

Department of Political Science, Economics and Sociology, Birmingham-Southern College, Birmingham, AL, 35254, USA Department of Economics, Florida State University, Tallahassee, FL, 32306, USA

A R T I C L E I N F O

A B S T R A C T

Keywords: Charitable giving Overhead Philanthropy

Overhead aversion is an issue of great importance to nonprofit organizations. On one hand, overhead (nonprogram expenses) is vital to the operational abilities of a nonprofit; on the other hand, there is evidence that donors dislike paying for overhead costs. In this paper, we use an experiment to study overhead from the perspective of donors. First, we replicate the prior finding establishing the existence of overhead aversion. Then we obtain new results by extending this line of research, examining whether donors have an aversion to specific types of overhead (specifically, salaries and fundraising). Our results suggest that donor behavior may be dependent on their outside option. If an overhead-free donation is readily available, then the average donor in our experiment (70–80% of subjects) prefers that charity to receive the donation. However, if donations are not overhead-free, most (approximately two-thirds of subjects) prefer the donation go toward fundraising efforts instead of salary-related expenditures.

JEL classification: C9 L30

1. Introduction Various factors, such as rebate or matching programs (e.g., Eckel and Grossman, 2003) and third-party ratings (e.g., Yörük, 2016; Brown et al., 2017), have been shown to affect charitable giving decisions. A more recent literature argues that a charity's overhead expenditures can also play a prominent role in the donation decision. Specifically, some donors experience “overhead aversion” - a negative feeling donors have toward a charity's overhead costs.1 This feeling could come from several sources (e.g., believing that only money going “directly to the cause” has an impact (Duncan, 2004), or that overhead spending is wasteful or inefficient), but the effect is the same – charities with higher overhead are less appealing to donors than those with lower overhead. In this paper, we contribute to the growing literature on overhead aversion (Bowman, 2006; Meer, 2014; Caviola et al., 2014; Gneezy et al., 2014, hereafter GKG2). We make use of a series of experiments to determine whether specific types of overhead (specifically, salaries and fundraising) influence donations. This is, to our knowledge, the first experimental study to examine attitudes toward specific overhead costs.

Our results suggest that donor behavior may be dependent on their outside option. If an overhead-free donation is readily available, then the average donor in our experiment prefers to direct a donation to that charity. This result and its strength are consistent, regardless of the overhead level (20% or 50%) and overhead type (general overhead, fundraising, and salaries). However, if donations are not overhead-free, they would prefer the donation go toward fundraising efforts rather than salary-related expenditures. Related literature has argued that potential donors pay close attention to expenditure ratios as a means of inferring important information like how efficient a charity is (e.g., Frumkin and Kim, 2001; Caviola et al., 2014). Empirically, GKG (2014) demonstrated through a series of experiments that donation allocations decrease when overhead expenses (fundraising and administrative) increase, but this effect disappears when a third party covers overhead expenses. Caviola et al. (2014) have shown that there is a strong tendency to give money to charities with low overhead ratios, when such ratios are used as an evaluation tool.3 Similarly, Bowman (2006) finds negative associations between positive changes in overhead ratios and the number of contributing donors (and the amounts contributed). In contrast,

⁎

Corresponding author. E-mail addresses: [email protected] (J.E. Portillo), [email protected] (J. Stinn). In our study, we define “overhead” costs as all non-program expenses, including both fundraising and administrative costs. A non-exhaustive list of overhead costs would include inhouse and third-party fundraising, salaries, training, rent, and insurance. 2 GKG demonstrated experimentally that donations decrease when overhead expenses increase. We base our design off theirs while altering their instructions to make explicit to our subjects that the overhead costs are those of the charity, not of the experimenters. 3 Caviola et al. (2014) also suggest alternative evaluating criteria such as the number of saved lives per dollar. However, they also make the argument that criteria like overhead ratios can be easily computed and compared across charities, which largely explains their popularity. Karlan and Woods (2017) also suggest that other measures of “effectiveness” could provide better information and persuade donors differently. Nevertheless, these measures are not always easily developed. 1

https://doi.org/10.1016/j.socec.2017.11.003 Received 20 June 2017; Received in revised form 13 November 2017; Accepted 13 November 2017 Available online 21 November 2017 2214-8043/ © 2017 Elsevier Inc. All rights reserved.

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

Frumkin and Kim (2001) find no evidence that suggests that charities who spend less in the administrative category fared any better in the market for contributions than those who spend more. Meer (2014), using data from DonorsChoose.org (a platform that makes overhead expenditures salient to donors), finds that higher overhead costs decrease the likelihood that a program receives funding. Most of the literature, therefore, seems to suggest that the average donor wants her donation to be spent mostly (if not only) on program-related expenses. However, it is clear that nonprofit organizations must spend some resources maintaining their infrastructure to deliver their charitable services. Survey evidence reports that the most commonly-researched information about charities by prospective donors relates to how their donation is going to be used (Hope Consulting, 2010).4 The growth of charity watchdogs such as Charity Navigator further suggest donors do try to educate themselves before donating.5 While the expenditure ratios may not be the best metric to make a donation decision (RoseAckerman, 1982; Steinberg 1986, 1989), they are easy to compute and comprehend (Tinkelman and Donabedian, 2007; Caviola et al., 2014). Therefore, we believe our results can inform nonprofit managers’ decisions on how to appeal to donors and structure their organizational costs. GKG found that having seed donors pay for overhead costs is a successful fundraising tactic.6 In practice, these seed funds are likely limited and unable to cover all of a charity's overhead expenditures. Our results suggest that donors are less willing to have their donation used for salary expenditures. Therefore, this (stronger) aversion to salary expenditures suggests those limited funds may be better targeted to cover a specific type of overhead, like salaries, and assuring donors that their donation would not be used for those purposes. This approach might be more feasible than finding donors to cover all overhead expenditures, and would help nonprofits raise funds from the general public and deliver services to their intended beneficiaries. Alternatively, our results also provide a first step to discovering whether certain overhead categories are more important than others to prospective donors, and how different overhead categories affect the donation decision. Knowing this information can help nonprofit managers to know what categories of overhead expenditures may need to be monitored closely to better appeal to donors and avoid drops in donations.

Table 1 Experimental designs and treatments. “Across Charity” Design1 Treatment

Description

Zero (baseline) 50GEN 50FUND

0% of the donation to T×V is allocated to overhead 50% of the donation to T×V is allocated to General Overhead 50% of the donation to T×V is allocated to Fundraising Expenses 50% of the donation to T×V is allocated to Salary Expenses 20% of the donation to T×V is allocated to General Overhead 20% of the donation to T×V is allocated to Fundraising Expenses 20% of the donation to T×V is allocated to Salary Expenses

50SAL 20GEN 20FUND 20SAL

“Within Charity” Design2 Treatment

Description

20%

20% of the donation is allocated to a specific type of overhead by the donor 50% of the donation is allocated to a specific type of overhead by the donor

50%

1 2

CW is always the alternative in every treatment above. No overhead is associated with CW. We compare these decisions to a conjectural even split.

effects on subjects’ decisions, relative to both the (i) “fixed” charity, and (ii) “general overhead” (which includes both salaries and fundraising expenses).7 Our treated charity is Truth x Vision (T×V), and our charity with zero overhead is charity: water (CW). These organizations were chosen because of their similar missions; they both aim to help communities in developing nations gain access to safe drinking water.8 CW is wellknown in the nonprofit community for their 100% donation policy. A sister organization, The Well, is funded by private donors and pays for all of CW's overhead costs. This allows CW to advertise to the public that 100% of public donations go entirely to programs, and enables us to truthfully inform subjects that this is how the donation would be used if CW was chosen.9 T×V agreed to partner with us and allocate any donations received as a result of this experiment as specified in our treatments.10 This agreement allowed us to vary the level and type of overhead across treatments, without deceiving our subjects. Thus, if T×V was chosen to receive the donation, our agreement with T×V enabled us to earmark our donations and direct portions of that donation to specific expense categories.11 A note indicating how the donation was to be allocated

2. Methodology In this section, we provide an overview of our experimental designs. To study the impact of overhead on donation decisions we run two experiments summarized in Table 1. Similar to GKG, our first design (the “Across Charity” experiment) asks subjects to choose one of two charities as the recipient of a $100 donation. We vary the level of overhead associated with donations to the treated charity between 0% (baseline), 20%, and 50%, while fixing the other charity's overhead at 0% across treatments. This implies that for a $100 donation (if the treated charity is selected), either $100, $80, or $50 goes to the treated charity's program expenses, and the remainder to that charity's overhead. We then extend GKG by attempting to identify whether specific types of overhead (salaries and fundraising expenses) have different

7

Treatment names refer to the overhead of the treated charity. For more information, visit www.truthxvision.org and www.charitywater.org. 9 On their website, CW state “Private donors cover our operating costs so 100% of your donation will bring clean water to people in need.” They further add that “[CW] depend on private donors, foundations and sponsors to cover everything from staff salaries to basic office systems to office rent and supplies. These donors are some of our most dedicated: their investment fuels our long-term mission, our ability to scale as an organization and our mission to continue using 100% of public donations for water projects.” They also provide an auditor's validation of the 100% model. While the structure of CW and The Well was not mentioned to subjects, they could verify CW's overhead costs upon request. 10 In Appendix B, we have provided the letter given to us the T×V's founder where he confirms his intent to allocate donations as specified. 11 We did not (and cannot) change the way the charities allocate their spending, so we cannot change their underlying overhead ratios. Instead, subjects in the treatments with overhead were told that a portion of our donation was going to an overhead cost. Gneezy et al. (2014) found that, as the overhead of their variable charity increased, the proportion of subjects who chose their fixed zero-overhead charity increased. However, when their subjects were told a third party would cover the overhead costs, the proportion of subjects who chose their variable charity increased to the point it was not significantly different from the results of their zero overhead treatment. This result suggests that people do not care about the underlying overhead ratio of a charity. Rather, people care about the percentage of their donation that goes to overhead costs. 8

4 Hope Consulting (2010) reports that roughly a third of donors do research before giving. The report further states that “For better or for worse, Overhead Ratio is the #1 piece of information donors are looking for.” 5 For instance, in self-reported numbers (https://www.slideshare.net/CharityNav/ charity-navigator-financial-measures-and-beyond, accessed June 1, 2017), Charity Navigator states they had 18,259 unique website visitors when it first launched its website in 2002. By 2005, the number of unique website visitors had grown to 1.2 million and in 2006, the number of unique visitors more than doubled to over 2.7 million. Currently, Charity Navigator advertises they have over 10 million annual visitors. 6 GKG proposed convincing some donors to cover overhead costs, then advertise (to other donors) donations that are overhead-free. If different types of overhead have differing appeal to donors, this type of seed money could be used strategically.

41

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

experiment would be taking place for extra credit, and about an alternate extra credit opportunity (a short quiz) for those who could not participate. The experiment was done at the beginning of the class, with the approval of the instructors. As subjects entered, they received a consent form and a closed envelope containing their decision sheet. Subjects were informed they could opt out of the experiment and complete an alternative extra credit assignment by marking the bottom of their decision sheet. After the consent forms were completed and collected, subjects proceeded to receive verbal instructions explaining the decision task. A volunteer served as a monitor, who would accompany an experimenter to the nearest mailbox (a short walk within the campus) and mail the donation check after the experiment to reinforce that we would be paying the selected charity. The monitor's decision was not used. We also told our subjects that they could request a copy of the donation receipt, once available, and verification from T×V that the donation would be allocated as indicated, to reinforce the notion that the donation was used as specified. Subjects made their decision privately by filling out the decision sheet contained in the envelope (sample decision sheets are provided below). After subjects finished, they placed their decision sheets back in the envelopes, which were then collected. The monitor randomly chose one of the envelopes. The enclosed decision sheet was the binding decision; the charity chosen on that sheet received the $100 donation. A check was filled out to that charity and put it into a pre-addressed envelope, along with appropriate instructions for how the donation was to be allocated. At the end of class we accompanied the monitor to mail the check. In several classes we ran multiple treatments simultaneously. In these cases, we ran similar treatments (e.g., the 20% and 50% Within treatments) so we could still use specific language without deception.16 Experimental assistants made sure subjects did not communicate with each other to remove outside influence. In total, we ran sessions in five separate classes containing a total of 746 subjects. We omit observations from subjects who did not wish to participate, or who failed to fill the decision sheet completely (e.g., not picking a charity).17 After dropping these observations, our sample contains 729 decisions. While our design closely follows that of GKG, we also build on their original design in two main aspects. First, in GKG the “overhead” portion of the donation goes to the experimenters' costs. Our instructions make it clear to subjects that the entire donation is going to the charity itself. If the money is earmarked to be split into program and overhead costs, the portion going toward overhead goes toward the charity's overhead rather than the experimenters' costs. This mirrors real-life donation decisions more closely and focuses the issue on charitable overhead. Second, we also build on GKG by specifying different types of overhead. Discovering whether donors have different attitudes toward different types of overhead is important because it may enable nonprofits to make better use of seed money meant to cover overhead costs (GKG, 2014). For example, if donors showed little to no reaction when the overhead associated with the donation relates to fundraising expenditures (as opposed to salaries), then any seed money obtained by a nonprofit to cover overhead costs, but which cannot cover all non-program costs, can be strategically used to cover specific types of overhead expenditures such as salaries. This may potentially enable nonprofits to minimize the extent to which overhead aversion affects donations to their organization.

was included alongside the check for donations that were mailed to T×V. Our overhead rates were chosen as follows. 50% was chosen for two main reasons. First, we wanted a high level of overhead that was likely to display subjects’ overhead aversion. GKG found evidence of overhead aversion at this rate, which informed our decision to have a 50% treatment. Second, given GKG's finding, we wanted to replicate their result and confirm the existence of overhead aversion using our modified setup. Next, we wanted to test an overhead rate lower than 50%. We settled on 20% for several reasons. First, 20% is the rate of nonprogram expenditures that T×V advertised to donors in their 2016 Annual Report.12 Second, we used the Statistics of Income Files (SOI) compiled by the National Center for Charitable Statistics (NCCS) to observe what the mean (median) overhead rate of similar charities were from 1995–2011. We found that similar charities had a mean (median) rate of 18.11% (15.13%).13 Finally, 20% is close to the average overhead level found to be “acceptable” in a survey of American donors (Grey Matter Research & Consulting, 2008). The survey revealed that the average American believes it is reasonable for nonprofits to spend 22.4% on non-program expenditures. For these reasons, 20% seems like a reasonable choice. Our second design, the “Within Charity” experiment, uses the treated charity (T×V) as the sole recipient of a $100 donation and fixes the percentage of a donation going toward overhead at either 20% or 50%. Subjects decided which overhead cost (fundraising or salaries) that portion of the donation supports, with the remainder going toward programs. Assuming donors are indifferent, we would expect an equal split between fundraising and salaries. While there are other types of overhead, we wanted to choose types of overhead spending that would straddle the aversion subjects felt toward general overhead. We thought subjects would be particularly averse to salaries and less averse to fundraising. Salaries play a prominent role in media stories about charity scandals14 even though it is well-established that workers in the nonprofit sector tend to make less than similar workers in the for-profit sector (Cohen, 2010). Therefore, we expect that subjects would be particularly averse to funding salaries. We expected fundraising to create less of an aversion since fundraising endeavors could be perceived as investments that will lead to more donations (e.g., Rose-Ackerman, 1982, Steinberg, 1986). Money spent on fundraising brings in donations from other people, and these resources can be spent to help recipients. For example, Andreoni and Payne (2011) find that fundraising is a strong investment, with a marginal dollar spent on fundraising bringing an estimated marginal return of five dollars.15 While not all subjects may think this way, we expected that some would realize that fundraising was a productive way to allocate money, and therefore fewer would show fundraising aversion than general overhead aversion (Steinberg, 1986). To run our experiment, we used students from large introductory courses at a major public university as subjects and offered extra credit as compensation. Prior to the day of the experiment, students were informed through a class email (sent by the instructor) that the

12

A scan on this report can be found in Appendix C. The NCCS classifies charities using the National Taxonomy of Exempt Entities (NTEE) system. The SOI dataset contains financial information on some of the largest nonprofits (which are arguably some of the most successful organizations) from 1995–2011. CW is categorized as a “P - Human Services- Multipurpose or Other” charity in this system. Using the available observations for category P, we found that similar charities had a mean (median) rate of 18.11% (15.13%). This is based on 16,034 observations. 14 Some recent examples of charities subject to scandals about executive salaries are Wounded Warrior Project, Susan G. Komen, Goodwill Omaha, Save the Children, and the American Red Cross. 15 This result suggests that charities are not net revenue maximizers, contrary to earlier theoretical literature (e.g., Rose-Ackerman, 1982). 13

16

Verbal instructions and subject decision sheets can be found in Appendix A. 613 observations belong in the “Across” treatment. Out 613 decision sheets collected, only 17 were dropped due to requests to be excluded or because subjects failed to pick a charity. These left us with 596 decisions in the “Across” treatment. The remaining 133 observations belong to the “Within” design, where no observations were dropped. 17

42

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

Table 2 Proportion of subjects choosing T×V. Treatment

N

%T×V

p-value

Zero (baseline) 50% overhead 50GEN 50FUND 50SAL 20% overhead 20GEN 20FUND 20SAL

62

66.1%

67 76 69

26.9% 25.0% 26.1%

< .001 < .001 < .001

123 88 111

29.3% 21.6% 22.5%

< .001 < .001 < .001

p-values in the last column were obtained using Fisher's Exact test, comparing the proportion of subjects choosing T×V in each treatment to the baseline (Zero).

Fig. 1. Proportion of subjects choosing T×V. Each column represents the proportion of subjects choosing T×V (rather than CW) as the recipient of the $100 donation, for each experimental treatment in our ``Across'' design.

3. Results18 3.1. “Across” charities Results for the “Across” design are summarized in Table 2 and Fig. 1. Fig. 1 displays the proportion of subjects choosing T×V as the recipient charity in each treatment with corresponding standard errors. Table 1 shows the overhead associated with the treated charity in the first column, the number of observations in the second column, the percentage of subjects choosing T×V in the third column, and the pvalue from Fisher's exact test comparing our overhead treatments to the baseline in the fourth column. In the ZERO treatment, roughly 66% of subjects chose T×V as the recipient of the $100 donation.19 Despite the initial preference for T×V, as the overhead increased this preference was reversed. Moving from the baseline to the 50% (20%) general overhead treatment, the proportion of subjects choosing T×V drops to 26.8% (29.3%), a drop of almost 59% (56%) from the baseline. This decrease is statistically significant (p < .001 for both) according to Fisher's exact test. This drop is also steeper than what was observed in GKG's original study.20 We believe that the relatively homogeneous missions of our charities make the impact of overhead aversion even stronger. That is, given that subjects had two similar charities, most of them opted for the one with less overhead. Therefore, we replicate the first finding of GKG and provide additional evidence of overhead aversion. Contrary to our expectations, neither the type nor magnitude of overhead appear to matter. The specific overhead treatments saw donations rates to T×V of 26.1% in 50SAL, and 25.0% in 50FUND. Similarly, in 20SAL only 22.5% of subjects opted to donate to T×V, and 21.6% in 20FUND. Similar to the general overhead case, the difference between these rates and the baseline are highly statistically significant

for all treatments (p < .001).21 However, there is no statistical significance comparing 50% overhead treatments to 20% overhead treatments (e.g., 50GEN vs 20GEN), nor when a comparison is made between specific overhead treatments and the general treatments (e.g., 50GEN vs 50FUND, 20GEN vs 20SAL, etc.). The lack of statistical difference may be driven by the homogeneous nature of the charities. Indeed, the familiarity scores for the two charities were quite close (on a scale of 1 to 7, the average familiarity scores were 1.30 for CW and 1.16 for T×V). Given that subjects had a choice between a charity with no overhead and one with overhead and a similar mission, it might be the case that there will always be a stronger preference for an overheadfree donation regardless of the type or magnitude of overhead affiliated with the donation.

3.2. “Within” charity To get a cleaner test of whether there is an aversion to a specific type of overhead, we used the Within design. Here, we held the charity (T×V) and overhead portion (20% or 50%) constant, and the subject had to decide whether to allocate the overhead portion to salaries or fundraising. As observed in Fig. 2, the majority of subjects opted to allocate the overhead portion of the donation to fundraising. The magnitude of the overhead does not appear to play a role. When 20% of the donation was allocated for overhead, 67.2% (32.8%) of subjects wanted to earmark the donation for fundraising (salaries). Similarly, when 50% of the donation was taken, 63.6% (36.4%) of subjects earmarked it for fundraising (salaries). Using a test of proportions, we verify that the proportion of subjects choosing salaries is statistically lower than an equal split (p = .0025 for 20% treatment, p = .0134 for 50% treatment), which would be expected if subjects were indifferent between these types of overhead. Thus, it appears that there is some preference for fundraising expenditures over salaries. However, when there is another potential recipient with no overhead, as shown by the Across experiment, donors would rather donate to such organizations.

18 In Appendix D, we provide regression results that corroborate the findings discussed in this section. 19 A 50-50 split might be predicted ex ante since the charities have similar missions and no overhead associated with the donation. However, a test of proportions reveals that the proportion of subjects choosing T×V is statistically different from 50% (p = .011). In their decision sheets, subjects were also asked how familiar they were with each charity (see the decision sheets in our supplementary material), a factor that may influence their decision. However, the familiarity scores of the two charities do not reveal any statistical differences in this treatment, according to a t-test (p = .635), so we conjecture the result is due to a preference for the name or the slight difference in mission description (an aspect that is held constant throughout all our treatments). 20 In GKG, 73.33% of subjects donated to their variable charity initially. This dropped to 49.43% in their 50% overhead treatment, a 32.6% decrease.

21 We tested each of the treatments in Table 2 against a hypothetical 50-50 split in the ZERO treatment. As suggested by Figure 1, each treatment has a significantly lower proportion of subjects choosing T×V, with the largest p-value equaling .0002. This indicates overhead aversion is not driven by an anomalous result in our ZERO treatment.

43

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

overhead option, is a subject for further research.23 These results have important implications for the nonprofit sector. They suggest a tension between donors’ desires and the needs of charities – however, this could be alleviated in several ways. As GKG suggest, charities could work to find sympathetic donors to cover these overhead costs. Our results indicate that covering salaries is a potentially fruitful alternative. This tactic is controversial, as some believe that it perpetuates the “overhead myth” and serves as an admission that overhead is inherently bad. Another way to ease the tension is to educate donors of the necessity of overhead spending (Pallotta, 2008; Taylor et al., 2013). Alternatively, the option of directed giving has been shown to increase donations in some instances (Eckel et al., 2017), even when this option is largely unused. This may be a promising avenue for charities concerned with overhead-averse donors. Cost-effectiveness is generally considered a better measure of charity worthiness, but the early evidence (Karlan and Wood, 2017; Caviola et al., 2014) is mixed on whether donors care about this information. Further research should address this issue, as well as testing sustainable solutions for overhead aversion. Additional work identifying the causes of overhead aversion will help to inspire ideas for potential solutions.

Fig. 2. Proportion of subjects choosing to allocate overhead to fundraising and salaries.

4. Conclusion This paper experimentally examines attitudes toward charitable overhead costs. Overhead aversion is a significant issue in the nonprofit sector, as it creates a tension between the desires of many donors and the ability of charities to fulfill their missions. Our results replicate the prior finding that overhead aversion exists. The results further suggest that donor behavior may be dependent on their outside option. Subjects appear to prefer overhead-free donations when possible, but have a preference to contribute to fundraising over salaries when the donation is not overhead-free.22 The nature of these results, along with the consistent and non-negligible number of subjects who chose the higher-

Acknowledgements We thank Lora Holcombe, Brandon Brice, Jesse Wright, Michael Hammock, Fredrick Bedsworth, Arthur Nelson, and Kevin Willardsen for their cooperation with this experiment. We would also like to thank Associate Editor, Phil Grossman, as well as three anonymous referees for helpful comments and suggestions. The authors also thank Kris Asleson, founder of Truth x Vision, for his cooperation with this study. Any errors or mistakes are our own.

Appendix A. Experimental instructions and decision sheets Across Design Sample (50GEN) Verbal instructions Good afternoon and thank you for coming. Today you are taking part in an experiment on economic decision making. In return for your participation, you'll get extra credit. We don't want anyone to influence other people's decisions, so we ask that you please refrain from talking during the experiment. Also, please keep you cell phones silenced and put away while the experiment is running. If you have any questions at any point during the experiment, raise your hand and we will come help you in private. You should all have an envelope and two papers at your seat. One of the papers is your consent form – go ahead and start reading it while I continue. The consent form tells you your rights as a participant in our experiment. If you agree to participate, you have to sign your name (and date) on the back of the form, preferably in pen. We will collect these before we get started. If you don't want to participate, then you have the option to complete an alternate assignment. In this case, put your name and email on the bottom of the decision sheet and we will contact you later. (collect consent forms) Here's the basic outline of our experiment: we have a blank check for $100 that we will send to a charitable organization. Your decision is which of the two charities will receive the money. On your decision sheet, you can see the two charities: Truth × Vision and charity: water. After I'm done reading these instructions, you will choose one of the two charities on your sheet. Then you'll put your decision sheet into your envelope to keep your decision private. We'll collect the envelopes, and one will be randomly chosen. Whatever the decision is in that envelope is binding. To assure you that we are donating this money to one of the charities, we will have one of you act as a monitor who will follow us after the

22 We would like to thank an anonymous referee for suggesting that a possible reason for why subjects strongly differentiate between the two specific types of overhead (fundraising and salaries) in the “Within” design but not in the “Across” design may have to do with the way information is presented. Specifically, a strand of literature suggests that presenting subjects with joint decisions vs. separate decisions may affect outcomes (e.g., Bazerman et al., 1999). In the “Across” treatment, subjects made separate evaluations over the types (and levels) of overhead we consider (i.e., we did not explicitly inform subjects about other types of overheads outside their specific treatment), while subjects made a joint evaluation decision on which overhead type they prefer in the “Within” treatment (they knew that a portion of the donation would be allocated to either salaries or fundraising, while being aware of both categories). However, we argue that our overall results remain. Results from our “Across” charities experiment consistently show an overhead aversion to the types and levels we test. In our “Within” charities experiment, we test preferences between overhead types as cleanly as possible by explicitly asking subjects to choose the overhead type they would prefer to fund, and find a strong preference to fund fundraising efforts. Under the assumption that information matters in the “Across” design, we would anticipate that a similar trend to that found in the “Within” treatments would present itself- a stronger preference for fundraising efforts. Nevertheless, the way in which information is presented is an interesting avenue for future research. 23 A potential explanation for the (highly consistent) number of donors who choose the charity with overhead might be related to donor preferences. For instance, Chlaß et al. (2015) explore how donors react to intermediaries who can take some portion of the donation for themselves. They use their result to classify donors into three different types of donors: (1) those who donate more because of the intermediary, (2) those who donate less because of the intermediary, and (3) those who do not change their behavior in response to the intermediary. In as much as we consider intermediaries and their ability to take some portion of the donation to be related to overhead in charitable organizations, the third classification might help explain the behavior we observe. This is undoubtedly an area that warrants further research.

44

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

experiment and observe us mailing the check. (pick a monitor, someone who can come to the dropbox) The monitor will also choose the envelope. Once we have the check ready, we'll put it into a pre-prepared envelope. Then we will go to the mail dropbox (at the Union) and mail the check. If anybody else would like to come to the dropbox and verify we are mailing the check, you are welcome to do so. If you are interested in verifying that the charity received the check, feel free to contact us and we will forward the receipt once we receive it. Now turn to your decision sheet. On that sheet, you'll see a description of how the donation is going to be used. For one charity, 100% of the donation is going toward program costs. For the other charity, 50% of the donation is going toward the charity's programs, and 50% is going toward a specific organizational cost. If you want to verify this, we can provide documentation upon request, after the experiment is finished. Also note that you don't put your name on your decision sheet, so decisions are anonymous and private. Now, please read your decision sheet and complete all the questions. When you're done, put it into your envelope and we'll collect them when everyone is finished. (collect envelopes, pick one, write check, mail the check) Within Design Sample (20SAL vs. 20FUND and 50SAL vs. 50FUND) Verbal instructions Good afternoon and thank you for coming. Today you are taking part in an experiment on economic decision making. In return for your participation, you'll get extra credit. We don't want anyone to influence other people's decisions, so we ask that you please refrain from talking during the experiment. Also, please keep you cell phones silenced and put away while the experiment is running. If you have any questions at any point during the experiment, raise your hand and we will come help you in private. You should all have an envelope and two papers at your seat. One of the papers is your consent form – go ahead and start reading it while I continue. The consent form tells you your rights as a participant in our experiment. If you agree to participate, you have to sign your name (and date) on the back of the form, preferably in pen. We will collect these before we get started. If you don't want to participate, then you have the option to complete an alternate assignment. In this case, put your name and email on the bottom of the decision sheet and we will contact you later. (collect consent forms) Here's the basic outline of our experiment: we have a blank check for $100 that we will send to a charitable organization, called Truth x Vision. Your decision will specify how the $100 donation will be spent by the charity among its various costs. Then you'll put your decision sheet into your envelope to keep your decision private. We will collect the envelopes, and one will be randomly chosen. Whatever the decision is in the chosen envelope is binding. To assure you that we are donating this money to the charity, we will have one of you act as a monitor who will follow us after the experiment/ class and observe us mailing the check. (pick a monitor, someone who can come to the dropbox) The monitor will also choose the envelope. Once an envelope has been chosen, we'll make out the check, and put it into a pre-prepared envelope. Then we will go to the mail dropbox (at the Union) and mail the check. If anybody else would like to come to the dropbox and verify we are mailing the check, you are welcome to do so. If you are interested in verifying that the charity received the check, feel free to contact us and we will forward the receipt once we receive it. Now turn to your decision sheet. On that sheet, you'll see a description of how the donation is going to be used. The $100 will be allocated to Truth × Vision as described on your sheet. A certain percentage of the donation will be allocated towards Truth × Vision's program costs, while the remainder will be allocated to a specific organizational cost. Your decision is which specific organizational cost will be funded: salaries or fundraising. Truth × Vision has agreed to spend our donation as we specify. If you want to verify this, we can provide documentation upon request, after the experiment is finished. Also note that you don't put your name on your decision sheet, so decisions are anonymous and private. Now, please read your decision sheet and complete all the questions. When you're done, put it into your envelope and we'll collect them when everyone is finished. (collect envelopes, pick one, write check, mail the check) Across Design Sample (50GEN) Subject decision sheets In today's study we will ask you to allocate $100 to one of two non-profits. At the end of the study, we will randomly choose the decision of one participant and implement it (i.e., make his/her specific payment). Your choice is whether to allocate the $100 to “Truth x Vision” or to “charity: water.” Truth × Vision. Truth × Vision is a nonprofit organization that provides water drills to help bring water to communities in need of safe drinking water. There is 50% overhead (i.e., spending on administrative and fundraising costs) associated with this donation, so for every dollar donated, 50 cents will go to Truth × Vision's programs and 50 cents will go toward Truth × Vision's overhead costs. Charity: water. charity: water is a nonprofit organization that provides clean and safe drinking water to people in developing nations. There is no overhead (i.e., spending on administrative and fundraising costs) associated with this donation, so for every dollar donated, the entire $1 will go to charity: water's programs.

45

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

Please tell us which organization you would like to receive $100: (circle one)

Truth × Vision Additional questions: On average, how often do you donate money 0 1 2 How familiar are you with charity: water? 1 2 3 How familiar are you with Truth × Vision? 1 2 3 What is your gender? What is your age?

Charity: water

to nonprofits? 3 4

(0 = never to 6 = six or more times a year) 5

6 (1 = not at all to 7 = very familiar)

4

5

6

7

4

5

6

7

(1 = not at all to 7 = very familiar)

**Once you are finished, please fold this sheet in half to keep decisions private.**

___________________________________________________________________________________________________________________ ____ I choose not to participate. If you choose not to participate, please write your name below so we can contact you about the alternate assignment. Within Design Sample (20SAL vs. 20FUND) Subject decision sheets In today's study we will ask you to decide how $100 will be spent by a charity. At the end of the study, we will randomly choose the decision of one participant and implement it (i.e., make his/her specific payment). Your choice is to specify how ``Truth × Vision'' will allocate a $100 donation. Truth × Vision. Truth × Vision is a nonprofit organization that provides water drills to help bring water to communities in need of safe drinking water. There is 20% overhead associated with this donation, so for every dollar donated, 80 cents will go to Truth × Vision's programs and 20 cents will go toward Truth × Vision's overhead costs. Your choice consists on deciding what category of overhead will receive the 20%. You can choose that 20% go to (a) salary expenses or (b) fundraising costs. In other words, for every dollar received by Truth × Vision, 80 cents will go to Truth × Vision's programs and 20 cents will go to either Truth × Vision salaries or fundraising costs, depending on your choice. Only one category will be funded but not the other. Please tell us which expense you prefer to receive 20% of the donation: (circle one)

Salaries Additional questions: On average, how often do you donate money to nonprofits? 0 1 2 3 4 How familiar are you with Truth × Vision? 1 2 3 4 5 What is your gender? What is your age?

Fundraising

(0 = never to 6 = six or more times a year) 5

6

6

7

(1 = not at all to 7 = very familiar)

**Once you are finished, please fold and put this sheet in your envelope to keep decisions private.**

__________________________________________________________________________________________________________________ ____ I choose not to participate. If you choose not to participate, please write your name and email below so we can contact you about the alternate assignment.

46

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

Appendix B. Letter from T×V founder

47

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

Appendix C. T×V 2016 annual report

48

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

Appendix D. Additional questions summary statistics and regression results In their decision sheets, subjects were also asked some additional questions that could influence their choice.24 In our Across experiment, the first question asked how often the subject donated to charity in a given year using a scale from 0 to 6, with 0 indicating they make no donation in a year and 6 implying the subject donates 6 or more times a year. The next two questions pertained to how familiar a subject was with the two charities that were eligible to get the donation (CW and T×V). Subjects could choose a number between 1 and 7, where 1 meant they were not familiar with the charity at all and 7 indicating they were very familiar with the charity. We also asked for the age and gender (we coded being female as 1). Summary statistics of the responses to these questions can be seen in Table D.1 for the Across experiment. (We also show the number of responses for each question in parentheses since not all subjects answered the additional questions.) For the most part, it appears that most of our subjects were unfamiliar with both charities, but they did make 1–2 donations a year. The same questions were used for the Within experiment, with exception of their familiarity with CW since T×V was the only charity used in this experiment. Subject answers can be seen in Table D.2. As before, these Subjects mean responses suggest they donate 1–2 times a year to a nonprofit, but that they were not too familiar with T×V. In Table D.3, we present regressions results for the Across experiment, were we use Subject responses as controls to explore whether our result for overhead aversion holds. The dependent variable in these regressions is a binary variable equal to 1 if the Subject chooses Truth × Vision (T×V) as the intended recipient of the of the $100 donation, and 0 if they choose charity: water (CW) as the intended recipient. We pool all our treatments together and include Treatment variables as control variables. (The omitted category is the baseline where both charities had a 0% overhead rate associated with the donation.) We also include the measures collected through the decision sheet on how familiar subjects were with both CW and T×V, how often they donate to charity, their age and their gender. Column (1) presents results from a linear probability model and Column (2) present results from a probit model. Both models yield the same conclusions presented in the main text: the presence of overhead decreases the likelihood that a subject would chose T×V as the recipient of the $100 donation. All treatment variables are negative and statistically significant. Being familiar with CW makes a donation to T×V less likely, while being familiar with T×V makes a donation to T×V more likely. This result could partly explain why some of our subjects chose T×V, despite the presence of overhead costs. Our results also suggest that females are less likely to donate to T×V.25 Meanwhile, age and how often a subject claimed to donate to charity appear to have no effect on their decision.

Table D.1 Across experiment additional questions summary statistics.

Zero (baseline) 50% overhead 50GEN 50FUND 50SAL 20% overhead 20GEN 20FUND 20SAL

How Often

Familiar w CW

Familiar w T×V

Age

Gender

1.37 (60)

1.28 (61)

1.23 (61)

19.12 (57)

0.33 (57)

1.58 (67) 1.71 (76) 1.69 (68)

1.16 (67) 1.25 (76) 1.21 (68)

1.05 (67) 1.18 (76) 1.09 (68)

18.39 (67) 18.63 (72) 18.75 (67)

0.43 (67) 0.625 (72) 0.49 (67)

1.96 (123) 2.1 (88) 1.77 (111)

1.36 (123) 1.42 (88) 1.23 (111)

1.11 (123) 1.23 (88) 1.1 (111)

19.05 (120) 18.74 (84) 18.94 (110)

0.55 (119) 0.48 (84) 0.42 (110)

Number of responses shown in parentheses.

Table D.2 Within experiment additional questions summary statistics.

20% 50%

How often

Familiar w T×V

Age

Gender

1.99 (67) 1.91 (66)

1.15 (67) 1.08 (66)

18.95 (66) 19.27 (65)

0.39 (66) 0.46 (65)

Number of responses are shown in parentheses.

24

These questions were also asked by GKG in their experiment. We have also interacted the gender indicator with the different treatments. When this is done, the coefficients on the interactions, as well as on the gender indicator, are not statistically significant. Treatment variables remain statistically significant. 25

49

Journal of Behavioral and Experimental Economics 72 (2018) 40–50

J.E. Portillo, J. Stinn

Table D.3 Regression results.

50GEN 50FUND 50SAL 20GEN 20FUND 20SAL Familiar CW Familiar T×V Age Gender How often donate Constant

Observations R2/Pseudo R2

(1) Linear probability model

(2) Probit

−.326*** (0.0801) −.343*** (0.0789) −.348*** (0.0793) −.291*** (0.0715) −.399*** (0.0764) −.386*** (0.0717) −.0666** (0.0280) .136*** (0.0449) 0.0307 (0.0195) −.108*** (0.0373) −0.0129 (0.0118) 0.0323 (0.380)

−.223*** (0.0634) −.237*** (0.0622) −.238*** (0.0615) −.216*** (0.0615) −.270*** (0.0588) −.276*** (0.0574) −.0916** (0.0389) .159*** (0.0534) 0.0308 (0.0204) −.115*** (0.0395) −0.0138 (0.0127)

575 0.107

575 0.086

Standard errors in parentheses *** p < .01, ** p < .05, * p < .1 The dependent variable, Choice, is a binary variable equal to 1 if T×V is chosen as the recipient of the $100 donation, and 0 if CW is chosen as the recipient. Column (1) shows results for a linear probability model, while Column (2) present results for a probit model. Column (2) report marginal effects.

For our Within experiment, in the main text we compared the proportion of subjects choosing a given overhead category (SAL or FUND) to a hypothetical 50/50 split. To test whether the answers provided by subjects to the additional questions influenced the overhead choice we ran independent regressions for the 20% and 50% treatments. The dependent variable in these regressions was equal to 1 if the subject chose salaries as the category they would prefer to see funded. For the most part, none of the elicited variables are significant. The only marginally significant variable (10%) was gender, where females are less likely to fund salaries in the 20% Within treatment. However, there is no significant variable in our 50% treatments.26

Rev. 61, 266–275. Grey Matter Research & Consulting, 2008. Where'd my money go? American Perceptions of the financial efficiency of non-profit organizations. Phoenix, Arizona. Gneezy, U., Keenan, E.A., Gneezy, A., 2014. Avoiding overhead aversion in charity. Science 346 (6209), 632–635. Hope Consulting. (2010) Money for good: the U.S. market for impact investments and charitable gifts from individual donors and investors. Karlan, D., Wood, D.H., 2017. The effect of effectiveness: donor response to aid effectiveness in a direct mail fundraising experiment. J. Behav. Exp. Econ. 66, 1–8. Meer, J., 2014. Effects of the price of charitable giving: evidence from an online crowdfunding platform. J. Econ. Behav. Organ. 103, 113–124. Pallotta, D., 2008. Uncharitable: How Restraints on Nonprofits Undermine their Potential. Tufts University Press. Rose-Ackerman, S., 1982. Charitable giving and “excessive” fundraising. Q. J. Econ. 97 (2), 193–212. Steinberg, R., 1986. Should donors care about fundraising. In: Rose-Ackerman, S. (Ed.), The Economics of Nonprofit Institutions: Studies in Structure and Policy. Oxford University Press, New York, pp. 347–364. Steinberg, R., 1989. Economic perspective on regulation of charitable solicitation. Case Western Law Rev. 39 (3), 775–797. Taylor, A., Harold, J., Berger, K., 2013. Moving Toward An Overhead Solution. The Overhead Myth Web-Accessed: 19 Dec. 2016. Tinkelman, D., Donabedian, B., 2007. Street lamps, alleys, ratio analysis, and nonprofit organization. Nonprofit Manage. Leadership 18 (1), 5–18. Yörük, B.K., 2016. Charity ratings. J. Econ. Manage. Strat. 25 (1), 195–219.

References Andreoni, J., Payne, A.A., 2011. Is crowding out due entirely to fundraising? Evidence from a panel of charities. J. Publ. Econ. 95 (5-6), 334–343. Bazerman, M.H., Moore, D.A., Tenbrunsel, A.E., Wade-Benzoni, K.A., Blount, S., 1999. Explaining how preferences change across joint versus separate evaluation. J. Econ. Behav. Organ. 39 (1), 41–58. Bowman, W., 2006. Should donors care about overhead costs? Do they care? Nonprofit Voluntary Sector Q. 35 (2), 288–310. Brown, A.L., Meer, J., Williams, J.F., 2017. Social distance and quality ratings in charity choice. J. Behav. Exp. Econ. 66, 9–15. Caviola, L., Faulmüller, N., Everett, J.A., Savulescu, J., Kahane, G., 2014. The evaluability bias in charitable giving: saving administration costs or saving lives? Judgment Decis. Making 9 (4), 303. Chlaß, N., Gangadharan, L., Jones, K., 2015. Charitable Giving and Intermediation. Jena Economic Research Papers (No. 2015-021). Cohen, R., 2010. Nonprofit salaries: achieving parity with the private sector. Nonprofit Quarterly 17 (2) Nonprofit Information Networking Association, Boston, MA. Duncan, B., 2004. A theory of impact philanthropy. J. Publ. Econ. 88 (9), 2159–2180. Eckel, C.C., Grossman, P.J., 2003. Rebate versus matching: does how we subsidize charitable contributions matter? J. Publ. Econ. 87 (3), 681–701. Eckel, C.C., Herberich, D.H., Meer, J., 2017. A field experiment on directed giving at a public university. J. Behav. Exp. Econ. 66, 66–71. Frumkin, P., Kim, M.T., 2001. Strategic positioning and the financing of nonprofit organizations: is efficiency rewarded in the contributions marketplace? Publ. Admin.

26

We do not show these results here since they do not convey any meaningful information. Nevertheless, these results are available from the authors upon request.

50