PRELIMINARY DRAFT: Please do not cite without permission How Low Can You Go? An Investigation into Matching Gifts in Fundraising Sara Helms Department of Economics, Finance, and QA Brock School of Business Samford University shelms1@samford.edu Timothy Diette Department of Economics Williams School of Commerce Washington & Lee University DietteT@wlu.edu Betsy Bugg Holloway Department of Management, Marketing, and Entrepreneurship Brock School of Business Samford University bbhollow@samford.edu Abstract There is a rich and varied set of research papers that considers the impact of matching offers on the behavior of donors to nonprofit organizations. Prior work includes both laboratory and field experiments, though the general consensus is turning toward field experiments as the better instrument. We evaluate the impact of matching gift offers included in a nonpartisan nonprofit organization s holiday mail fundraising drive. We add to the existing literature in two ways. First, our use of a nonpartisan nonprofit is uncommon. Second, prior literature establishes that more generous matches (beyond $1:$1) generally do not increase donations in a cost-effective way. However, few studies consider less generous offers. We evaluate the impact of a $1:$10 matching offer. Our results suggest that nonprofit organizations donors are not susceptible to matching offers, but that the 1:1 match led to a higher response rate than the 1:10 match for existing donors. We find no other statistically discernible impact of the matching offers. Our results suggest that matching offers may not work for all types of nonprofit organizations, and that the response to the offers differs across regular donors to the organization compared to other donor types. JEL Code: C93 Field Experiments; D64 Altruism; Philanthropy; L31 Nonprofit Institutions, NGOs; M31 Marketing Keywords: charity; donations; nonprofit; experiment; matching offers; incentives Acknowledgements: Diette is appreciative of the financial support provided by the Lenfest Summer Research Grant through Washington and Lee University. All errors are the responsibility of the authors. 1
Introduction The third sector in the United States economy captures nonprofit organizations. Charitable giving by individuals in the US amounts to nearly $300 billion annually and volunteering is done by more than 25 percent of the population (Giving USA 2011; BLS 2012). Nonprofit organizations exist, in no small part, as a result of donations made by individuals. How to raise money is at the forefront of many nonprofit organizations minds, and there is an entire industry focused on how to efficiently raise funds. Economists entered into this area of study using standard econometric techniques, laboratory experiments, and field experiments to understand which factors drive giving behavior, and if there are predictable ways to increase giving at low cost. Despite the abundance of existing studies, there are many questions yet to be answered. In 2010 we conducted a randomized control experiment that altered the incentives for giving for a mailing sent to potential donors. We provide a full account of the prior experiment in Helms, et al (2013). The following year, 2011, we conducted a follow-up study with the same nonprofit organization. Using their mailing lists for existing contributors to the nonprofit organization, we implemented a match-grant scheme. We used match rates of 100% (up to $100) and 10%. We aim to better understand the impact of high and low matching grants, in the context of an organization that is nonpartisan. Both aspects of our study address gaps in the existing literature. Background and Literature Review Both laboratory and field experiments are part of the existing research body; we focus here on field experiments. List (2008) promotes the use of field experiments over laboratory experiments when considering donor behavior, arguing that field experiments get much closer to 2
the true decision process. For a detailed history of the study of the economics of charity, see List (2011). Nonprofit organizations use a variety of incentive schemes in their quest to inspire donations matching offers, gifts, and social pressure, to name a few. Given the setup of our field experiment, we focus on matching offers in the existing literature. The basic matching offer involves a third party matching gifts made by a donor, typically at something less than 100 percent. For example, for each $1 a donor gives, the third party gives an additional $0.50 to the organization, often with some cap. Charities must consider the theory behind offering matching or seed grant incentives to donors. Andreoni (2006) develops a theory which addresses the motivation for advertising seed gifts. He argues that large gifts announced in conjunction with fundraising drives, which he terms leadership giving, provide a signal to donors about the credibility of the charity. List and Lucking-Reiley (2002) test Andreoni s theory using a university fundraising campaign, and find that increasing seed money increases contributions. As explained in Karlan and List (2007), matching grants are effective because they lower the price of giving, they serve as credibility signals, and they signal that now is the right time to give. Similarly, Bekkers and Wiepking (2011) describe matching offers as signals for the efficacy of the gifts, and as having a legitimizing effect indicating the trustworthiness of the organization (p. 943). One of the most important studies on matching grants is by Karlan and List (2007). While their organization, unlike ours, is politically motivated, they examine the effect of three different match schemes. They consider the effect of a $1 match for $1 donated (written as $1:$1), and the more generous matches of $2:$1 and $3:$1. While they find that introducing a match increases 3
both the likelihood that a donation is made and the revenue per solicitation, there is no additional benefit to more generous matches (the $2:$1 and $3:$1 conditions). We extend their study by moving in the other direction. If larger matches do not improve fundraising, perhaps even smaller matches can yield similar results. In contrast, Eckel and Grossman (2008) find that matching gifts in a local public radio station fundraising drive increase total contributions mainly through inspiring additional gifts, and not larger gifts from existing givers. Using data from the Netherlands, Bekkers (2005) similarly finds that the matching offers increase giving through additional donations, but not larger gifts. However, existing Sierra Club donors exposed to a matching gift did not exhibit increased giving (Rondeau and List, 2008). Other work focuses on the impact of social cues on giving. Knowing what others gave changes a donor s gift (Croson and Shang, 2008). Martin and Randal (2008) find that, when donations are current and visible in a glass box, larger initial donations beget larger donations, and smaller initial donations beget smaller donations. 1 Even the attractiveness of the solicitor matters in door-to-door fundraising (Landry et al, 2006)! Mail-based fundraising campaigns are a common research tool. They are rather low-cost to implement, and they provide many opportunities for varied treatments. Hager, et all, (2003) consider the responsiveness of the organizations themselves to a mail-based survey, and explore what factors might encourage nonprofit executives to return a survey. They find that a $5 gift (given regardless of response), the $5 gift plus the promise of a $50 to the organization, and no such incentive do not lead to differential response rates. 1 We enclose initial in quotation marks because the initial donations were staged by the researchers, and not the result of true prior giving. 4
Experiment Design We worked with the local office of a large scale, nationally-recognized, health-related nonprofit organization whose work is viewed as nonpartisan. Using the organization s database of existing contributors, we designed a randomized experiment to consider the influence of matching offers on giving behavior. In the fall of 2011, we included our offers (and a control group offered no match) in the annual holiday mailing. The matching gifts were funded by support from a local foundation which chose to remain anonymous. Given prior studies focus on generous matches of $1:$1 and higher, we aimed to consider the impact of a smaller match, and the impact of a match on a nonpartisan organization s fundraising efforts. In assigning treatment and control conditions to mailing recipients, we randomized the three conditions across three separate mailing lists. Table 1 shows the division of our sample between the three mailing lists and the three conditions. Our control group was provided no additional incentive to donate money to the organization, other than the usual holiday fundraising letter (N = 2,129). The second group is our low-match group, which received the offer that each $1 donated would warrant an additional $0.10 gift [funded through the foundation we worked with, which preferred to remain anonymous] (N = 2,119). The third group is our highmatch group, which was told that their donations would be doubled, a standard $1:$1 match (N = 2,122). The first mailing list, Non-Donor, included individuals who had not donated money to the organization, but were affiliated in some other way (for example, as a volunteer) (N = 2,970). The second list, Existing, included existing givers who already gave in 2011, prior to the holiday mailing which included our experiment (N = 2,315). All households on this list had given prior to 2011 as well. The third list, Newly Acquired, made their first donation to the organization in 2011 (N = 1,085). For each of the three mailing lists, we 5
randomly assigned one of three conditions to each mailing address. As shown in Table 1, the size of the incentive groups is similar across all lists. In the mailings that went to each household, control or treatment, the needs of the organization are stated identically. In the control group, the letter was similar to the holiday mailing used by the organization each year. What set apart the matching offers was the additional paragraph which for the 1:1 match stated: Thankfully (and just in time!), a local foundation has stepped in with a special Matching Grant offer. That means your gift today will DOUBLE in impact by being matched dollar-for-dollar. Your gift of $30 becomes $60 a gift of $50 becomes $100 to help more <redacted> and a gift of $100 becomes $200 to help provide <redacted>! I ll be honest these kinds of opportunities don t come along often. And I m so grateful when they do. I m also grateful for friends like you. So please give to our Year-End Fund Drive your gift will go TWICE as far to help Can I count on you for a tax-deductible gift by December 31? In the control group s letters, it simply read: And now, more <redacted> are coming to us for <redacted>. That s why we ve launched our Year-End Fund Drive and why I am turning to you again today. Can I count on you for a tax-deductible gift by December 31? Given the random assignment of households across conditions, we can consider the causal impact of a matching gift offer on the behavior of long-term donors, newly-acquired donors, and households which had not previous donated money. Results Unconditional Gifts by Mailing List and Incentive Group We first consider the impact of the matching gift offer on the unconditional average gift amount, by condition and mailing list. By unconditional, we mean that we include in the average 6
those households that did not give. In a sense, this unconditional average gift measures the return for the nonprofit of the mailing, averaged over the number of letters mailed. Table 2 presents the results, and includes tests for equality of average gift size across conditions but within each mailing list. In column (1) of Panel A, we see an overall average gift of $0.66 for non-donor households without a matching offer, with larger average gifts for the 1:10 match ($1.51) and 1:1 match ($0.87). In Panel B, we report the results of tests for equality across condition, and tests for similar distributions (measured using the Mann-Whitney test). While the difference across conditions appears to be large, we cannot reject the null hypothesis that there is no difference across condition for non-donor households. In column (2), we repeat the analysis for existing donors. We see a different pattern in this list, as the largest average gift is for the 1:1 match ($2.13), followed by the no incentive group ($1.40) and 1:10 match group ($1.02). While most of the apparent differences are not statistically significant, we confirm that the 1:1 match is larger than the gift in the 1:10 match condition at the 0.05 level of significance. In column (3) we report results for newly acquired donors. We find a third pattern with this group; the largest unconditional average gift is $2.05 in the no incentive condition, followed by the 1:1 match at $1.78, and the 1:10 match at $1.51. None of the differences are statistically distinguishable from zero. We report the pooled results for all lists in column (4), in part out of concern that the smaller samples for the individual lists lead to the difficulty in establishing statistically significant results. We see the anticipated pattern of largest gift for the 1:1 match ($1.48), then the 1:10 match ($1.33) and the no incentive group ($1.16). Given the disparate results for the 7
three mailing lists, it is not surprising that we find no statistically discernible difference across the conditions. Response Rates by Incentive Group We separate the unconditional average gift amount into two analyses to better understand the impact of the matching incentive on donor behavior. First, we consider the response rates of the incentive groups for each mailing list and for the mailing list as a whole. We report these results in Table 3; we include the calculations in Panel A, and test for differences of values and distributions in Panel B. It is not surprising to see that the non-donor list has the lowest response rates, and that for each mailing list the 1:1 match condition has the highest response rate. Somewhat surprising, we see higher response rates for the no incentive group for each mailing list including prior givers. In column (1), we show that the response rate of 1.11% for the 1:1 match group is higher than the 0.81% response rate for the no incentive group at the 90% level of confidence. For existing donors, the only statistically distinguishable difference is that the 1:1 response rate of 5.57% is greater than the 4.15% response rate for the 1:10 match. None of the differences found for newly acquired donors is statistically significant. With the pooled sample, we find (again) that the 1:1 response rate of 3.58% is greater than the 1:10 rate of 2.78%. Average Gift Amount among When we calculated the unconditional average gifts, we mixed the effects of the response rate and the average gifts for households giving a donation. To round out our simple analysis, we show in Table 4 the average gift amount, conditional on giving a donation, for each incentive group and mailing list combination. Among non-donor households, the no incentive group 8
donors gave on average $81.25, the 1:10 match donors $135.91, and the 1:1 match donors $61.07. None of these differences are statistically significant. 2 Among existing donors, we nearly reject the equality of the conditional average gift for the 1:1 match ($38.19) and the 1:10 match ($24.56). The no incentive group average of $29.24 falls between the two match groups. Among the newly acquired donors, the average gifts are nearly identical for the 1:10 and 1:1 match groups ($33.75 and $33.84, respectively), and both are lower than the no incentive group ($41.39). However, none of these differences are close to statistically distinguishable from zero at the 90% level of confidence. Once again, the pooled sample has no clear pattern of conditional average gift between incentive groups, and once again, this appears to be due to the varying patterns across the mailing lists. Multivariate Regression Analysis In order to test for common differences across mailing lists, we follow our simple analysis of means with a series of regressions that simultaneously control for match condition and mailing list. In Table 5, we report results for a linear probability model that uses likelihood of responding to the mailing as the dependent variable. Consistent with our earlier results, we find no results within each mailing list. When we pool our sample, we find no difference across conditions but we find that the existing donor mailing list is 3.7 percentage points more likely to give than the no incentive group, and that existing donors are more likely to respond to the 2 There was one gift over $500 that resulted from this holiday mailing, and it fell in the 1:10 match, non-donor list. This large outlier gift may be skewing our results. 9
mailing than the newly acquired donor group. Table 6 repeats the analysis using a probit model in order to account for the binary dependent variable. Our results mimic those of the linear probability model. In Table 7 we report the results from an ordinary least squares (OLS) regression with the unconditional donation amount as the dependent variable. Within each mailing list we find no statistically distinguishable result between the no incentive groups and either of the matching groups. The coefficient of 0.725 indicates that existing donor households who got the 1:1 match offer gave $0.725 more than the no incentive group, while the 1:10 match group gave $0.382 less. We find that the coefficient of 0.725 on the 1:1 match group in the existing donors list is statistically different from the -0.382 coefficient on the 1:10 match group. In other words, among existing donors the unconditional gift amount is larger for the 1:1 match than the 1:10 match. Our last multivariate regression analysis includes donors only, and uses the (conditional) donation amount as the dependent variable. While we find no statistically distinguishable effects across the matching conditions, we once again find that the existing donor mailing list gives less than the non-donor household list. 3 Moreover, we find that the existing donor list also gives less than the newly acquired donor mailing list. Discussion and Conclusions Our results show that households in the non-donor mailing list generally did not alter their behavior as a result of the matching gift offer. While this group is the only of the three that did not have a prior gift recorded for 2011, the list is comprised only of individuals who had never donated money to the organization, but who had instead volunteered or otherwise been 3 Though we once again are concerned that one large gift among the non-donor households skews our results. 10
involved with it. We find that the generous $1:$1 match induced a larger unconditional average gift when compared to no incentive, though, suggesting that an aggressive campaign could move non-(monetary)-donors to (monetary) donors. That said, we find no impact on the propensity to give or the conditional average gift size, suggesting that matching offers will not move these households to give when they otherwise are not inclined to give. Similarly, newly acquired donors do not have a long history of giving to the organization; only in 2011 did they begin donating. They, too, were unresponsive to the matching offer. There are many theories for why this might be. The households may not realize that the offer is unusual for the organization, or they may be less attached to the organization and its cause. The discussion above contrasts with the observed response by existing donors who have a giving history with the organization that extends before 2011. These households are most responsive to the matching gift, and exhibited a large reaction an at least 10% increase in the propensity to give when offered a $1:$1 match. That said, the households were not responsive to the $1:$10 match. One goal of this study was to examine the effect of smaller matches than typically used in existing studies. Prior studies generally lower match rates only to $1:$2. We find that lowering the match rate to $1:$10 did not lead to any measurable changes in the behavior of donor households. Thus, it turns out that it is possible to go too low with the match offer. We describe our evidence as weak, however, since our results are not as stark as other existing studies on generous matching offers. Taken together, our results suggest that matching gifts are not a panacea for nonprofit organizations. When large enough, they increase the giving of existing donors. We find a large, 11
statistically significant increase in giving for existing donors when they are offered a $1:$1 match. However, we find no effects for newly acquired donors or non-donor households. Works Cited Andreoni, James. 2006. Leadership giving in charitable fund-raising. Journal of Public Economic Theory 8(1): 1-22. Baker III, R.J., J.M. Walker, and A.W. Williams. 2009. Matching contributions and the voluntary provision of a pure public good: Experimental evidence. Journal of Economic Behavior and Organization 70(1-2): 122-134. Bekkers, R. 2005. When and Why Matches are More Effective Subsidies Than Rebates. http://igitur-archive.library.uu.nl/fss/2006-0727-201343/bekkers_05_when-and-why-matchesare-more-effective.pdf (downloaded 19 June 2013). Bekkers, R. and P. Wiepking. 2011. A literature review of empirical studies of philanthropy: Eight mechanisms that drive charitable giving. Nonprofit and Voluntary Sector Quarterly. 40(5): 924-973. Bureau of Labor Statistics. 2012. Volunteering in the United States 2011. United States Department of Labor. USDL-12-0329. Chen, Y., X. Li and J.K MacKie-Mason. 2006. Online fund-raising mechanisms: A field experiment. The Berkeley Electronic Press Contributions to Economic Analysis and Policy. 5(2) Article 4: 1-37. Last accessed July 13, 2011. Croson, Rachel, and Shang, Jen. The Impact of Downward Social Information on Contribution Decisions. Experimental Economics 11(3): 221-233. Eckel, C. and P.J. Grossman. 2008. Subsidizing charitable contributions; a natural field experiment comparing matching and rebate subsidies. Experimental Economics 11(3): 234-252. Giving USA Foundation. 2011. The Annual Report on Philanthropy for the Year 2010: Executive Summary. Giving USA and The Center on Philanthropy at Indiana University. Last accessed 11/10/2012. Hager, Mark A., Sarah Wilson, Thomas H. Pollak and Patrick Michael Rooney. Response Rates for Mail Surveys of Nonprofit Organizations: A Review and Empirical Test. Nonprofit and Voluntary Sector Quarterly 2003 32: 252, DOI: 10.1177/0899764003032002005. 12
Helms, Sara, Timothy Diette, and Betsy Bugg Holloway. 2013. Acquiring the New Donor, Motivating the Prior Giver: An Empirical Examination of Incentive-Based Strategies for Nonprofit. Under review. Karlan, D. and J. List. (2007). Does price matter in charitable giving? Evidence from a largescale natural field experiment. The American Economic Review. 97(5): 1774-1793. Landry, C., Lange, A., List, J., Price, M., and Rupp, N. 2006. Toward an understanding of the economics of charity: Evidence from a field experiment. The Quarterly Journal of Economics 121(2):747-782. List, John. 2008. Introduction to field experiments in economics with applications to the economics of charity. Experimental Economics 11(3): 203-212. List, John. 2011. The Market for Charitable Giving. Journal of Economic Perspectives 25(2): 157-80. List, John, Lucking-Reiley, David. 2002. The effects of seed money and refunds on charitable giving: Experimental evidence from a university capital campaign. Journal of Political Economy 110(1): 215-233. Martin, R. and J. Randal 2008. How is donation behaviour affected by the donations of others? Journal of Economic Behavior and Organization 67(1): 228-238. Meier, Stephan. 2007. Do subsidies increase charitable giving in the long run? Matching donations in a field experiment. Journal of European Economic Association 5(6): 1203-1222. Rasul, Imran, and Huck, Steffen. 2010. Transactions costs in charitable giving: Evidence from two field experiments. The B.E. Journal of Economic Analysis & Policy 10(1): Article 31. Rondeau, Daniel, and List, John. 2008. Matching and challenge gifts to charity: evidence from laboratory and natural field experiments. Experimental Economics 11(3): 253-267. 13
Table 1: Observations by Incentive Group and Donor Relationship Non-Donor Existing Newly Acquired No Incentive 991 772 366 2,129 Match 1:10 991 771 357 2,119 Match 1:1 988 772 362 2,122 2,970 2,315 1,085 6,370 14
Table 2: Comparison of Unconditional Average Gift Amount by Incentive Group and Donor Relationship Panel A: Means and Standard Deviations Non- Donor (1) Existing (2) Newly Acquired (3) (4) No Incentive $0.66 $1.40 $2.04 $1.16 (8.812) (7.943) (11.215) (8.991) Match 1:10 $1.51 $1.02 $1.51 $1.33 (32.260) (5.761) (8.035) (22.570) Match 1:1 $0.87 $2.13 $1.78 $1.48 (9.994) (16.334) (9.533) (12.621) $1.01 $1.52 $1.78 $1.32 (20.155) (11.008) (9.688) (15.795) Observations 2,970 2,315 1,085 6,370 Panel B: Tests of Differences Non- Existing Donor Newly Acquired 1:10 Match vs. No Incentive Mean Difference $0.85 -$0.38 -$0.52 $0.17 Pr(T<t) 0.789 0.140 0.236 0.625 Pr( T > t ) 0.422 0.280 0.472 0.750 Pr(T>t) 0.211 0.860 0.764 0.375 Mann-Whitney Test 0.492 0.536 0.769 0.728 1:1 Match vs. No Incentive Mean Difference $0.21 $0.73 -$0.26 $0.32 Pr(T<t) 0.690 0.866 0.369 0.827 Pr( T > t ) 0.621 0.267 0.737 0.347 Pr(T>t) 0.311 0.134 0.632 0.173 Mann-Whitney Test 0.199 0.475 0.858 0.254 1:1 Match vs. 1:10 Match Mean Difference -$0.64 $1.11 $0.26 $0.15 Pr(T<t) 0.275 0.962 0.656 0.604 Pr( T > t ) 0.549 0.076* 0.689 0.792 Pr(T>t) 0.725 0.038** 0.344 0.396 Mann-Whitney Test 0.542 0.184 0.648 0.138 15
Table 3: Comparison of Response Rates by Incentive Group and Donor Relationship Panel A: Means and Standard Deviations Non- Donor (1) Existing (2) Newly Acquired (3) (4) No Incentive 0.81% 4.79% 4.92% 2.96% (0.090) (0.214) (0.217) (0.170) Match 1:10 1.11% 4.15% 4.48% 2.78% (0.105) (0.200) (0.207) (0.165) Match 1:1 1.42% 5.57% 5.25% 3.58% (0.118) (0.230) (0.223) (0.186) 1.11% 4.84% 4.88% 3.11% (0.105) (0.215) (0.216) (0.174) Observations 2,970 2,315 1,085 6,370 Panel B: Tests of Differences Non- Donor Existing Newly Acquired 1:10 Match vs. No Incentive Mean Difference 0.30% -0.64% -0.43% -0.17% Pr(T<t) 0.755 0.271 0.391 0.367 Pr( T > t ) 0.490 0.542 0.782 0.733 Pr(T>t) 0.245 0.729 0.609 0.633 Mann-Whitney Test 0.489 0.542 0.782 0.733 1:1 Match vs. No Incentive Mean Difference 0.61% 0.78% 0.33% 0.62% Pr(T<t) 0.902 0.754 0.580 0.873 Pr( T > t ) 0.196 0.491 0.839 0.254 Pr(T>t) 0.098* 0.246 0.420 0.127 Mann-Whitney Test 0.196 0.491 0.839 0.254 1:1 Match vs. 1:10 Match Mean Difference 0.31% 1.42% 0.77% 0.79% Pr(T<t) 0.729 0.903 0.683 0.930 Pr( T > t ) 0.541 0.195 0.633 0.139 Pr(T>t) 0.271 0.098* 0.317 0.070* Mann-Whitney Test 0.541 0.195 0.633 0.139 16
Table 4: Comparison of Average Gift Amount by Incentive Group and Donor Relationship: Conditional on Giving a Donation Panel A: Means and Standard Deviations Non- Donor (1) Existing (2) Newly Acquired (3) (4) No Incentive $81.25 29.24 41.39 39.32 (59.146) (22.682) (31.239) (35.355) Match 1:10 135.91 24.56 33.75 47.81 (287.983) (15.085) (19.279) (127.834) Match 1:1 61.07 38.19 33.84 41.32 (60.198) (59.061) (26.022) (53.260) 90.91 31.34 36.38 42.62 (171.065) (39.788) (25.958) (79.352) Observations 33 112 53 198 Panel B: Tests of Differences Non- Donor Existing Newly Acquired 1:10 Match vs. No Incentive Mean Difference $54.66 -$4.68 -$7.63 $8.50 Pr(T<t) 0.697 0.162 0.202 0.694 Pr( T > t ) 0.606 0.325 0.405 0.613 Pr(T>t) 0.303 0.838 0.798 0.306 Mann-Whitney Test 0.397 0.740 0.528 0.666 1:1 Match vs. No Incentive Mean Difference -$20.18 $8.94 -$7.54 $2.00 Pr(T<t) 0.228 0.806 0.215 0.600 Pr( T > t ) 0.456 0.389 0.429 0.799 Pr(T>t) 0.772 0.194 0.786 0.400 Mann-Whitney Test 0.232 0.390 0.415 0.981 1:1 Match vs. 1:10 Match Mean Difference -$74.84 $13.62 $0.09 -$6.49 Pr(T<t) 0.176 0.896 0.505 0.345 Pr( T > t ) 0.351 0.207 0.991 0.689 Pr(T>t) 0.824 0.104 0.495 0.655 Mann-Whitney Test 0.777 0.252 0.429 0.698 17
Table 5: Multivariate Estimation Results Linear Probability Model Likelihood of Response with Donation Non- Donor (1) Existing (2) Newly Acquired (3) (4) Match 1:10 0.003-0.006-0.004-0.002 (0.520) (0.557) (0.786) (0.753) Match 1:1 0.006 0.008 0.003 0.006 (0.196) (0.477) (0.836) (0.239) Existing Donor 0.037*** (0.000) Newly Acquired Donor 0.000 (0.942) χ 2 Tests Match 1:10= Match 1:1 0.515 0.194 0.634 0.136 Existing=Newly Acquired 0.000*** Observations 2,970 2,315 1,085 6,370 R-squared 0.001 0.001 0.000 0.012 Notes: significant results in bold; *** p<0.01, ** p<0.05, * p<0.1 18
Table 6: Multivariate Estimation Results Probit Model Likelihood of Response with Donation Non- Donor (1) Existing (2) Newly Acquired (3) (4) Match 1:10 0.003-0.007-0.004-0.001 [0.492] [0.542] [0.782] [0.797] Match 1:1 0.006 0.007 0.003 0.006 [0.206] [0.491] [0.839] [0.226] Existing Donor 0.042 *** [0.000] Newly Acquired Donor 0.000 [0.950] χ 2 Tests Match 1:10= Match 1:1 0.543 0.197 0.633 0.143 Existing=Newly Acquired 0.000*** Observations 2,970 2,315 1,085 6,370 Notes: significant results in bold; *** p<0.01, ** p<0.05, * p<0.1 19
Table 7: Multivariate Estimation Results OLS Model Amount of Donation: Unconditional on Donating Non- Donor (1) Existing (2) Newly Acquired (3) (4) Match 1:10 0.853-0.382-0.523 0.170 (0.347) (0.495) (0.469) (0.726) Match 1:1 0.209 0.725-0.259 0.317 (0.817) (0.195) (0.718) (0.513) Existing Donor 0.506 (0.248) Newly Acquired Donor 0.261 (0.653) χ 2 Tests Match 1:10= Match 1:1 0.478 0.048** 0.716 0.762 Existing=Newly Acquired 0.777 Observations 2,970 2,315 1,085 6,370 R-squared 0.000 0.002 0.000 0.000 Notes: significant results in bold; *** p<0.01, ** p<0.05, * p<0.1 20
Table 8: Multivariate Estimation Results OLS Model Amount of Donation: Conditional on Donating Non- Donor (1) Existing (2) Newly Acquired (3) (4) Match 1:10 54.659-4.681-7.639 5.037 (0.502) (0.627) (0.400) (0.719) Match 1:1-20.179 8.943-7.547-1.224 (0.795) (0.318) (0.386) (0.926) Existing Donor -59.379*** (0.000) Newly Acquired Donor 4.925 (0.702) χ 2 Tests Match 1:10= Match 1:1 0.292 0.145 0.992 0.640 Existing=Newly Acquired 0.005*** Observations 33 112 53 198 R-Squared 0.038 0.021 0.020 0.076 Notes: significant results in bold; *** p<0.01, ** p<0.05, * p<0.1 21