1. Introduction
There has been a significant rise in research focusing on charitable giving in the past decades. Notably, field experiments have gained prominence as a methodological approach, following the pioneering work by Karlan and List (Reference Karlan and List2007) on donation matching. Field experiments have enabled researchers to combine fundamental inquiries into the nature of giving with practical, direct applications in collaboration with charitable organizations. These collaborations have been instrumental in refining fundraising strategies, leading to more effective and targeted approaches employed by many charitable organizations. Our study contributes to this evolving landscape by addressing a relatively underexplored aspect of fundraising: the impact of donor appreciation strategies, specifically examining the effectiveness of different forms of acknowledgment in donor communications.
Through two field experiments conducted in collaboration with one of Germany’s leading opera houses, we investigate how saying ‘thank you’ impacts efficient and sustainable fundraising. The first experiment investigates whether thanking in advance in solicitation letters affects donation decisions, and the second investigates the role of handwritten versus printed thank you postcards in communications with past donors.
While there is a fairly large body of literature on expressions of gratitude in psychology,Footnote 1 the number of studies that examine the role of gratitude for giving behavior is rather small. Andreoni and Serra-Garcia (Reference Andreoni and Serra-Garcia2021) show that thank you notes that follow pledges to donate reduce the rate of reneging. In contrast, Samek and Longfield (Reference Samek and Longfield2023), who examine whether thank you calls after a donation affect the propensity to give again, only find a null result. Bruttel et al. (Reference Bruttel, Nithammer and Stolley2022) document an adverse effect of expressions of gratitude in the context of a dictator game in the laboratory, where recipients can send a message to the dictator before the dictator decides: those who thank in advance receive less than those who do not. The negative impact of thanking in advance could be due to the perception that the message sender has an unconditional expectation that their request will be fulfilled, which may cause recipients to feel coerced. Dwyer (Reference Dwyer2015) shows that gratitude expressions may not only increase compliance but can also backfire.
There is also some inconclusive literature on thank you gifts. Falk (Reference Falk2007) documents that the frequency of donations increased by 17% with a small gift and by 75% with a large gift. In Eckel et al. (Reference Eckel, Herberich and Meer2016) donors are twice as likely to give when they receive an unconditional, high-quality gift. In contrast, Chao (Reference Chao2017) found a reduction of giving probability through a thank you gift. The inconclusiveness in the literature can partly be attributed to differences in gift-giving strategies. In Chao (Reference Chao2017), individuals received a thank you gift only if they made a donation, whereas in other studies, gifts were given unconditionally.
There is some evidence on the effects of thank you messages on recurring donations. Moussaoui et al. (Reference Moussaoui, Herren-Vichet, Baldelli, Waldvogel and Desrichard2019) demonstrate that sending thank you text messages to blood donors increases donor return rates in a way that is statistically significant, but the overall effect is rather small. Fosgaard et al. (Reference Fosgaard, Hansen, Jacobsen, Sørensen, Romose and Ullum2020) show that informing blood donors per text message about the use of their donation increased subsequent donations. As all their text messages also included a thank you statement, it is not possible to disentangle the effects of feedback information and the gratitude expression. In an earlier study, Gemelli et al. (Reference Gemelli, Carver, Garn, Wright and Davison2018) also showed that text messages about blood donation use increased recurring donations. However, their texts did not include any thank you notes.
Like Andreoni and Serra-Garcia (Reference Andreoni and Serra-Garcia2021) and Samek and Longfield (Reference Samek and Longfield2023), our first experiment studies donor behavior in the context of an actual fundraising campaign carried out by a charitable organization. In addition, it shares with Bruttel et al. (Reference Bruttel, Nithammer and Stolley2022), the feature that it focuses on an anticipatory expression of gratitude, a ‘thank you in advance’ phrase in a solicitation letter. We have two competing hypotheses regarding its impact on donations. Both are related to the observation that the phrase increases the pressure that the organization exerts on the potential donor. If this pressure operates like social or peer pressure, which has been shown to increase giving (see, for example, DellaVigna et al., Reference DellaVigna, List and Malmendier2012 or Meer, Reference Meer2011), the phrase should be effective. If, in contrast, it is perceived as a threat to freedom of choice, it might stimulate an adverse reaction (‘reactance’ in the terminology of Brehm and Brehm, Reference Brehm and Brehm1981) and thus turn out to be counterproductive. This reaction could seriously impede the success of fundraising efforts, as shown by Adena and Huck (Reference Adena and Huck2022) – trying to push donors into higher giving categories might create backlash – and by Adena and Huck (Reference Adena and Huck2020) and Adena and Huck (Reference Adena and Huck2023) – who examine adverse customer reactions to facilitated-giving campaigns.
In our second study, we investigate the effectiveness of handwriting in expressions of gratitude.Footnote 2 Clark and Kaminski (Reference Clark and Kaminski1988), for example, find a large positive effect of handwritten letters on survey participation. Garner (Reference Garner2005) finds that handwritten post-it notes significantly increase the likelihood of participation in surveys. He suggests that handwritten requests are perceived as a request for a personal favor. In contrast, Moss and Worthen (Reference Moss and Worthen1991) find that typed salutations were more effective than handwritten salutations for survey participation. Despite this small set of marketing studies, we are not aware of any study that examines handwriting in the context of charitable giving. We study the role of handwriting for notes that say thank you to previous donors and ask for a new donation. In contrast to the ambiguous predictions for the first experiment, we have a clear one-sided hypothesis for this second experiment. Handwritten thank you notes should be more effective than printed notes simply because they signal more effort and appreciation that should, through the channel of reciprocity, enhance giving (see, Dufwenberg and Kirchsteiger, Reference Dufwenberg and Kirchsteiger2004; Falk and Fischbacher, Reference Falk and Fischbacher2006 for theories of reciprocity; Fehr and Gächter, Reference Fehr and Gächter2000 for a review of the early experimental literature on reciprocity).
2. Methods
2.1. Data collection and treatments
We conducted two natural field experiments in conjunction with a large opera house in Germany.Footnote 3 The opera house regularly engages in charitable activities in areas such as arts education through two affiliated arts education foundations based in the region.
Based on anonymized customer and donor data provided by the opera house, we selected individuals to receive variants of letters based on our treatments. The opera house ensured that no individuals or couples in the selected sample of recipients received more than one fundraising letter. The opera house subcontracted a company to print and send fundraising letters in May 2023. The letter was similar to the previous letters that the opera house had used to raise funds but included a slight variation. The additional thank you card was not distributed in the past. Table 1 summarizes our experimental treatments. The exact formulation of the letters and cards is presented in Appendix A4.
Designs of the two studies
Note: Visuals of all materials and letters can be found in the Appendix A4.
The top panel of Table 1 presents our ‘Thank you in advance’ study, which involved two treatments. While the baseline letter did not include the phrases ‘Thank you in advance’ or ‘Thank you for your support,’ the treatment group received a letter that concluded with these expressions. To maintain the context of support, the letter specifically included ‘Thank you for your support’ to prevent alternative interpretations, such as merely thanking the recipient for reading the letter. The opera house sent 8,800 letters to examine the impact of thanking in advance. The bottom panel of Table 1 presents our personalization study, which also involved two treatments. In addition to the fundraising letter (Thank-you-in-advance version as used in the previous campaigns), a separate group of donors received a postcard that had either a printed or a handwritten message thanking them for their past donations. In total, 1,132 past donors received these cards to test the effectiveness of handwritten notes. All 569 cards were individually handwritten and not copy-printed. We explicitly decided to use a pen that required some pressure and left deeper marks on the paper. Additionally, we specifically asked for the writing not to be perfect, meaning that all letters should not have the same shape, and the lines should not always be exactly horizontal. We are quite sure that, at least upon closer inspection, it was evident to the recipients that the handwriting was not a serial copy. An example of the handwriting can be found in Appendix A4, Table A8.
We obtained ethical approval from WZB Berlin (2023/02/195) for both of the studies. We wrote the pre-analysis plan before receiving the data and registered it at Open Science Framework (https://osf.io/vsjw6).Footnote 4
2.2. Sample and randomization
The gross sample was based on an anonymized customer database that we received from the opera house. It included individual customers who had bought tickets in the last 18 months prior to our studies or donated to one of the past fundraising campaigns (2015–2018, 2022). We combined these data with the data from the customer database that we received in the previous joint study campaigns.Footnote 5 These data include information related to ticket purchases and individual characteristics alongside past donation behavior.
Given that there is substantial heterogeneity in the sample with respect to the amount spent on tickets, frequency of visiting the opera house, and the amounts donated, imbalances in the randomization procedure can hinder inference. Hence, we made sure to achieve a good ex-ante balance between the treatment groups. In addition, given that charitable giving is a low-frequency behavior, we selected the individuals who were most likely to respond to the campaign in order to ensure maximum power. Based on our previous experience with the opera house, we anticipated that the highest response rate would come from past donors (warm list), followed by loyal Opera customers, and then by Opera customers holding academic titlesFootnote 6 (Adena & Huck, Reference Adena and Huck2022). Consequently, we predominantly selected these customers for the pool of letter recipients. In contrast, the response rate from customers with high spending on opera tickets was typically much lower than that of the other groups. Even though the amounts donated by them were generally higher, the resulting return per letter was lower, so we placed less emphasis on ticket spending when selecting letter recipients. Appendix A1 presents the power calculations. Detailed explanation of the sampling and balancing procedures can be found in Appendix A2. The overall budget of the campaign allowed for sending 10,000 letters. Because 1,200 past donors were pre-allocated to Study 2, Study 1 comprised 8,800 letters.
2.3. Primary hypotheses
2.3.1. Study 1: Thank you in advance
Given that the charities commonly use the phrase ‘Thank you in advance,’ and there is evidence of a positive impact of thanking on helping and donations (see the literature review above), we hypothesize that this phrase increases charitable giving on both the extensive and intensive margins.
H1a: Including a ‘Thank you in advance’ phrase in a donation letter increases the number of donations.
H1b: Including a ‘Thank you in advance’ phrase in a donation letter increases the donation amounts.
Although our hypotheses are directional, we also acknowledge the potential for opposing effects.Footnote 7 The thank you phrase may increase pressure on potential donors. If this pressure resembles social or peer pressure, it could boost donations, as noted by DellaVigna et al. (Reference DellaVigna, List and Malmendier2012) and Meer (Reference Meer2011). Conversely, if seen as a threat to freedom of choice, it may trigger ‘reactance’ (Brehm & Brehm, Reference Brehm and Brehm1981), leading to counterproductive outcomes. Such reactions could hinder fundraising efforts, as demonstrated by Adena and Huck (Reference Adena and Huck2022) and by Adena and Huck (Reference Adena and Huck2023).
2.3.2. Study 2: Handwritten versus printed cards
As the literature on personalization of fundraising letters is thin, the directions of our hypotheses are based on the literature suggesting that personalization of invitation letters to surveys increases participation (see the literature review above). We expect that more involved personalization and, hence, effort displayed by a fundraiser increases charitable giving.
H2: A handwritten thank you note generates higher donations than a printed note.
3. Results
In our statistical analyses, we strictly follow our pre-analysis plan, which specifies all variables and the entire analysis. As prespecified in the pre-analysis plan, we excluded those individuals whose letters were not delivered due to relocation or death (213 in total).
Figure A1 in the Appendix presents the correlation matrix for the variables used in the analysis separately for Study 1 and Study 2.
3.1. Primary hypotheses tests
Given the randomization procedure, for the tests of our hypotheses, we use a chi-square test (for the extensive margin) and a Mann-Whitney U-Test (MWU; for the extensive and combined margins). Table 2 presents the non-parametric analyses. The top panel presents the results of Study 1 and the bottom panel presents the results of Study 2. Given that charitable giving is a low-frequency, high-variance behavior, we might lack precision in both tests, especially when analyzing the combined margin that is characterized by a large proportion of zero donations, such that an MWU test lacks power in such a context. Therefore, we also run simple OLS regressions with robust (Huber-White) standard errors. We consider significance thresholds of p < 0.05* and p < 0.01** throughout the study.
Non-parametric analyses
Note: Standard errors in paratheses.
Following the literature on charitable giving, we study extensive, intensive, and the combined margin of responses. That is, we look at the probability to donate, the value of the donation (excluding zeros), and the return per letter (donation value including zeros). The latter two variables have highly skewed distributions. In line with the literature and our pre-analysis plan, we take the log of the donation value (excluding zeros) when studying the intensive margin. We take the log of the donation value plus one (including zeros) when studying the total effect.
We run OLS regressions for all three dependent variables. For each dependent variable, we run two regressions: one without control variables and one including controls for blocks of randomization alongside all characteristics used in the process of the randomization. This all adds up to six models for each of the two studies in the regression table.
Our models without control variables involve only a binary variable that defines treatment variations. Following regressions without control variables, we include independent variables in our models as they were used in the randomization. Table A2 in the Appendix explains all the independent variables used in the regression analyses in both studies. Table 3 presents regression results.
Regression results
* Notes: Robust standard errors in parentheses; 95% confidence intervals in brackets; * p < 0.05, ** p < 0.01; a – 150 donors belong to 148 randomization blocks such that we cannot add block of randomization FE.
In Study 1, the response rate was 1.6% in the baseline treatment and 1.8% in the treatment with the ‘thank you in advance’ phrase. Neither the chi-square test nor a parametric regression with or without control variables show any significant differences. In Table 3, we also include the 95% confidence intervals (in square brackets), which show that, if there is any difference, it is very close to zero. This shows that the zero estimates are not due to a power problem but are precisely estimated.
The average positive donation was €111 in the baseline and €80 in the treatment with a thank you in advance. Neither the MWU test in Table 2 nor parametric regressions in Table 3 shows statistical differences between treatments.
The revenue per letter was €1.82 in the baseline and €1.45 in the thank you in advance treatment. However, we do not find any significant differences, either using non-parametric or parametric tests. The zero effect in Table 3 is quite precisely estimated.
In Study 2, among past donors, the response rate was much higher, with 18.7% in the printed version and slightly higher with 19.3% in the handwritten version of the card. The differences are again not significant, and the zero effect is quite precisely estimated, as can be seen in Table 3.
The average positive donations are similar in Study 2. The average donation was €114 with the printed card, and €87 with the handwritten card. Again, the differences are not significant. The revenue per letter was €21 in the printed card treatment and €17 in the handwritten card treatment, which is again not significant.
We present some additional robustness checks in the Appendix – heterogeneity with respect to previous experience with solicitation letters (Table A6, only study 1) and to gender (Table A7). We did not find any evidence of heterogeneous treatment effects.
4. Conclusion and discussion
In a preregistered, large-scale field experiment, we show that neither thanking in advance in solicitation letters nor sending handwritten thank you postcards matters for the success of a fundraising campaign. Importantly, our zero results are quite precisely estimated, removing any concerns related to power.
While we draw our conclusions from the specific context of our study, we do not think that our results are necessarily only limited to the specific context of opera giving in Germany. Indeed, our study is in line with the null effects of thanking donors on subsequent giving by Samek and Longfield (Reference Samek and Longfield2023) in the context of radio donations and the US. Of course, applications to non-Western cultures might be limited and more research in this context is needed. Several studies indeed show that there are cultural differences in thanking and interpretation of politeness norms (e.g., Ohashi, Reference Ohashi2003; Park & Lee, Reference Park and Lee2012).
Despite limited direct evidence on gratitude in fundraising, models of social preferences imply gratitude, especially in reciprocal actions. Dufwenberg and Kirchsteiger (Reference Dufwenberg and Kirchsteiger2004), alongside Falk and Fischbacher (Reference Falk and Fischbacher2006), show that individuals are more inclined to reciprocate when they perceive substantial effort, deliberate intent, or personal investment from another party. In this regard, the null result in Study 2 – where handwritten cards did not correlate with increased donations – is counterintuitive. Handwritten notes, which intrinsically demand greater effort and convey a higher degree of personal attention compared to printed cards, could have been construed by donors as indicative of commitment. Therefore, reciprocity-based theoretical frameworks would predict that such a demonstrative act of effort would elicit a correspondingly enhanced donation response. The psychology literature on emotions similarly highlights the importance of gratitude in social networks (Algoe et al., Reference Algoe, Dwyer, Younge and Oveis2020).
The null result in our second study may be explained by the fact that although the cards were handwritten, they were not perceived as personal because they were sent by an institution rather than by a genuine peer. Evidence regarding differences between human–human and human–computer interactions can support this view (see, for example, Nass and Moon, Reference Nass and Moon2000; Sanfey et al., Reference Sanfey, Rilling, Aronson, Nystrom and Cohen2003; Engel, Reference Engel2011). Similarly, Algoe et al. (Reference Algoe, Kurtz and Hilaire2016) show that personalization should extend beyond merely expressing gratitude and include personal praise – for example, saying ‘you are so generous.’ In light of the null results of our study, we caution fundraisers regarding the allocation of their limited resources to personalization of solicitation letters. Rather than fine-tuning conventional solicitation letters, investing in adaptation to technological advancements, such as artificial intelligence (AI), might yield higher returns in the near future. For example, Namkoong et al. (Reference Namkoong, Park, Park and Lee2023) show that AI chatbots’ gratitude expressions increased people’s willingness to donate, especially when the AI provided a human identity cue. Future studies can further investigate the role of AI in fundraising campaigns in relation to human–AI interactions and other means of personalization and expressions of gratitude.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/eec.2025.10033.
Acknowledgements
We thank Jonathan Meer for valuable comments, Katharina Dorn, Lenna Simmat, and Marius Werz for research support. We thank the opera house for making this project possible. Maja Adena and Levent Neyse acknowledge the financial support of the German Research Foundation (DFG) through the project CRC TRR 190 Rationality and Competition (project number 280092119). Maja Adena acknowledges the financial support of the Leibniz Association through projects P146/2022 and K595/2024 and of the German Research Foundation (DFG) through project 518302089. Levent Neyse acknowledges the financial support of the Leibniz Association through projects SAW Lab Square, SAW10868, and K523/2023. The replication material for the study is available at https://doi.org/10.7802/2907.
AI Use Statement
We used AI-based tools (i.e. ChatGPT by OpenAI) solely for language editing and grammar correction after completing the writing of this manuscript. No AI system was used for generating content, conducting analysis, or drawing conclusions.
Open access
