To What Extent Is Trust a Prerequisite for Charitable Giving? A Systematic Review and Meta-Analysis

Literature searches were run in June 2018 across four databases: the two largest multidisciplinary academic databases (Scopus and Web of Science) plus content-relevant databases in business (ABI/Inform) and psychology (PsycINFO). We used search terms nested around three required concepts: charitable giving (donation* OR donate OR donor* OR philanthrop* OR charitable), nonprofits (“not for profit*” OR “non profit*” OR nonprofit* OR NGO* OR “non governmental” OR “third sector” OR “charit*”), and trust (trust OR confidence OR integrity OR efficac* OR competen*). The search was limited to articles published in peer-reviewed journals between 1988 and 2018. The initial batch of citations were later supplemented through snowballing from reference lists, citation alerts of new publications meeting the search criteria, and open calls for unpublished data distributed through Twitter and the Association for Research on Nonprofit Organizations and Voluntary Action, Society for Personality and Social Psychology, and European Association of Social Psychology mailing lists. These searches were further supplemented by an additional search run on the ProQuest Theses & Dissertation Global database in April 2020, as well as articles identified from alerts based on the original searches up to April 2, 2020. All search efforts yielded 1,482 unique articles.

Initial screening was done in three rounds: first on title, second on abstract, and third on the full text. To be included, articles needed to be written in English and to have quantitatively measured both trust and charitable giving, specifically donations of money to organizations. A total of 1,419 articles and theses were excluded: 1,127 based on title, 204 based on abstract, 36 for being an inappropriate type of publication (e.g., book review, legal decision summary, and correction), 36 for not measuring both trust and charitable giving, 9 for not being published in English, and 7 for not having a full text available.

A final sample of 63 articles met the inclusion criteria, and these articles were incorporated into the quantitative literature review. Of these, however, 59% (n = 37) did not report the bivariate associations necessary for meta-analysis. To capture these missing data, the first author emailed authors of affected articles to request the necessary information. In all cases, at least three emails were sent to solicit this information. Any article whose authors either responded advising that data were no longer available or could not be provided (n = 7) or failed to respond to any of three or more email contacts (n = 11) were excluded from analyses. In instances where two or more articles reported identical effects drawn from the same data and analyses (e.g., Alhidari, 2014; Alhidari et al., 2018; Sargeant & Lee, 2004a, 2004b), we retained both articles for the quantitative literature review but only included the effects reported in the first peer-reviewed publication in the meta-analysis. This left 41 articles reporting 69 effect sizes to be included in the meta-analysis.

Sample sizes ranged from 36 to 33,062 participants (M = 1,942, SD = 5,162), with an average gender representation of 53% female and 47% male. Participants’ average age ranged from 20 to 64 years (M_age = 40.61, SD = 11.91). As illustrated in Figure 2, much of the research that has investigated the relationship between trust and giving has used community or nationally representative samples. Indeed, only 10 studies (16%) conducted their research in student populations. This is important to note as younger people traditionally give less (e.g., Wiepking & James, 2013) and are therefore unlikely to be representative of donor psychology. Twenty-two studies (35%) used methods to ensure their samples were close to nationally representative. Thus, we can have some confidence that results discussed in this article are generalizable to typical community and donor samples.

Almost half (44%) of the studies were conducted in either the United States (k = 16) or the Netherlands (k = 12), both hubs for research on charitable giving. Considering the region in which research was conducted (see Figure 3), our evidence of the relationships between trust and charitable giving primarily comes from Europe (k = 27) and North America (k = 18), although it is heartening to see a handful of studies being reported from each of Asia (k = 9), the Middle East (k = 3), Latin America (k = 3), and Africa (k = 1). No results were returned for studies looking at trust and charitable giving in Oceania (i.e., Australia, New Zealand, and the Pacific).

Of the 63 articles identified by the systematic review, 58 (92%) used survey methods, four (6%) used experimental methods, and one (2%) used archival methods. Mode of delivery was not always noted but included online (k = 17), face-to-face (k = 9), telephone (k = 5) and paper-based (k = 5) survey methods. Although several studies used experimental approaches, none manipulated trust experimentally. Instead, other factors were manipulated, and trust was captured as a measured control variable. Therefore, all evidence for the relationship between trust and giving is correlational.

Various referents of trust were assessed in relation to charitable giving, and in some cases more than one type of trust was measured within the same study. Figure 4 depicts the number of times each type of trust—generalized, institutional, sectoral, and organizational—was assessed in relation to giving outcomes.

Nearly half of the studies (k = 27) measured generalized trust. This was usually measured by a variant of the generalized trust scale (e.g., “Most people can be trusted”) or an aggregated scale of general social trust items, where participants reported their degree of trust in various targets. Only five studies measured institutional trust. This was generally measured by asking participants how much trust they had in a range of social institutions (e.g., parliament, hospitals, religious organizations, police, and politicians). Twenty studies measured sectoral trust—confidence or trust in charities, nonprofits, or nongovernment organizations. Measures included single items (e.g., “How much confidence do you have in charities?”), multi-item scales measuring trust in charities (e.g., “Most charitable organizations are honest and ethical in their use of donated funds”), and scales relating to Mayer et al. (1995) multidimensional conceptualization of trustworthiness (i.e., ability, benevolence, integrity). Organizational trust—in which participants rated trust in a specific organization—was assessed in 23 studies. Diverse measures assessed this construct, including Sargeant et al.’s (2006) trust in organizations scale (5 items, e.g., “I would trust this nonprofit to conduct their operations ethically”), novel or substantially adapted multi-item scales (e.g., “In my opinion the management of this charitable organization is trustworthy”), and measures based on Mayer and colleagues’ (1995) conceptualization of organizational trustworthiness.

Across the 63 articles, 41 measured the likelihood of being a donor and 33 measured the value of donations. Some studies measured both likelihood and value. Likelihood was assessed in various ways. Some studies asked whether or not participants had donated any amount to charity in the previous 12 months and coded responses as 1 = donor, 0 = nondonor. In contrast, some studies used post hoc database matching to assess whether or not the participant donated to the organization’s last fundraising campaign (Waters, 2008). These measures captured the likelihood that they had donated in the past. Other studies assessed willingness to make donations or donation intentions, capturing the likelihood that they would donate in the future. Combining future and past likelihood is not a problem in this instance because the meta-analysis considers the bivariate association and makes no comment on causality.

Value was typically assessed by asking participants to recall how much they had donated to all charities in the previous year, often log-transformed to address skewness. Alternatively, value was assessed by (a) asking how much of a hypothetical $100 they would donate to a particular charity, (b) providing payment or a bonus for participation and assessing how much, if any, the individual donated to charity, or (c) database matching to assess the value of the participant’s last gift to a particular organization. In all but six cases, charitable giving was measured with self-reports, with participants either recalling their past behavior or advising their current willingness or future giving intentions.

Traditionally, it has been easier for researchers to publish studies reporting statistically significant effects. This creates the possibility that meta-analyses based on published findings will over-estimate the true effects. To address the potential for publication bias, we created Funnel Plots of the observed correlations and standard errors for all the studies. Any asymmetry around the aggregate effect would suggest publication bias is present in the sample. In this case, visual inspection of the funnel plot of just the published articles in the corpus (Figure 6A) suggested that studies with smaller sample sizes (based on standard errors) reported larger effect sizes than studies with larger samples. The funnel plot indicated large degrees of heterogeneity in observed effects (perhaps attributable to moderating factors) but no systematic bias toward larger-than-expected effects in the published literature. Follow-up tests of asymmetry provided contradictory conclusions: Egger’s regression test was not significant (z = 0.61, p = .544) but the Rank correlation test was (Kendall’s τ = 0.29, p = .017), suggesting asymmetry in the funnel plot. Given this ambiguity, and the potential for publication bias impacting conclusions, we actively sought to include unpublished data in the meta-analysis. The aggregate effect reported above includes 17 effect sizes from 9 unpublished studies and doctoral theses. Tests indicate no evidence of publication bias when these unpublished works are incorporated: neither Egger’s regression test (z = −0.42, p = .677) nor the Rank correlation test (Kendall’s τ = 0.18, p = .104) were significant. We also report publication status as a potential moderator below.

Significant levels of heterogeneity between studies was observed, Q(68) = 3,968.78, p < .001, τ² = .03, I² = 98.76%, leading us to conclude that the true effects significantly vary across studies. The meta-regressions reported below attempt to identify likely causes of this observed heterogeneity.

A meta-regression revealed significant differences in the relationship between trust and giving dependent on the type of trust, QM(3) = 26.06, p < .001. Comparing aggregate effect sizes by subgroup, the weakest relationship was observed between charitable giving and generalized trust, r = .11, CI [.09, .14], and this was statistically equivalent to the relationship with institutional trust, r = .14, CI [.11, .18], p = .613. Compared to generalized trust, a stronger relationship with giving was observed for sectoral trust, r = .27, CI [.19, .34], p = .001; and especially for trust in the particular organization, r = .35, CI [.22, .48], p < .001.

Comparing aggregate effect sizes by subgroup, there was a trend such that a stronger relationship with trust was found for studies measuring likelihood of giving, r = .26, CI [.19, .32], than giving value, r = .17, CI [.12, .22]. However, this difference was not statistically significant, QM(1) = 3.48, p = .062.

A meta-regression revealed sample type to be a significant moderator of the trust-giving relationship, QM(2) = 7.30, p = .026. Compared to studies using nationally representative samples, r = .16, CI [.10, 21], larger effects were found in studies using nonrepresentative community samples, r = .29, CI [.20, .37], p = .009. No statistical differences were found, however, between nationally representative samples and student samples, r = .25, CI [.15, .34], p = .120, perhaps because of the small number of student samples available for comparison.

Western countries were operationalized as countries in Europe or North America, and compared to non-western countries (i.e., in South America, Asia, Middle East, and Africa). Compared to studies conducted in western countries, r = .19, CI [.14, 23], studies conducted in non-western countries, r = .30, CI [.19, .41], reported significantly larger relationships between trust and giving, QM(1) = 4.73, p = .030.

Comparing the aggregate effect sizes by subgroup confirmed that published studies reported stronger relationships, r = .24, CI [.19, .30], than unpublished studies, r = .15, CI [.11, .19], but the difference did not reach statistical significance, QM(1) = 3.34, p = .068.³

The year of data collection did not significantly moderate the effect, QM(1) = 0.74, B = .00, SE = .00, CI [.00, .01], p = .389, indicating that the relationship between trust and giving has not changed over time.

We set out to test whether or not trust was a prerequisite for giving. Theories of trust have differed in their perspectives on this issue. For example, the trust-commitment-giving model (elaborated below) assumes a causal process whereby trust promotes giving (Sargeant & Lee, 2004a). However, other theorists have posited that giving—and other forms of community contribution—can promote trust (Putnam, 2000; Uslaner, 2002).

Although we found a significant relationship between trust and giving, it is important to highlight that the available evidence does not address the question of causality: it remains unclear whether trust is a prerequisite for or consequence of charitable giving. None of the studies in our systematic review experimentally manipulated trust or analyzed trust longitudinally to assess its causal relationship with giving. Thus, despite a strong presumption that trust is important for promoting giving, evidence only supports a positive association, not a causal relationship. For now, it remains equally possible that the observed relationship goes the other way, such that charitable giving leads to increased trust (see also Putnam, 2000; Uslaner, 2002). Some donors may give to charity for reasons other than trust—to alleviate guilt, for example, or because they wanted to be seen to be generous (e.g., Basil et al., 2006; Bénabou & Tirole, 2006). In such cases, trust in the organization may not be a key factor in their giving. However, the experience of giving may lead them to trust the organization or nonprofit sector more if, for example, they receive detailed reports about how donations are distributed. To address this critical gap in the evidence base, we call for future research designs that address the issue of causality, such as large-scale experimental and longitudinal studies.

In the economics literature, giving is broadly conceptualized as being motivated by altruism versus “warm glow,” or the emotional utility donors derive from the act of giving (see Andreoni, 1990). Although our results do not speak directly to this division, they invite speculation. First, the finding that trust and giving are fairly consistently associated suggests trust is a factor in donor decisions. Because the relationship is stronger as the target becomes more precise, this aligns with the altruistic explanation for the relationship between trust and giving: if donors care about outcomes for beneficiaries, they must trust that the fundraising organizations that they deliver their aid through will actually deliver it to the targets of their concern (see also Bekkers, 2003). However, the relatively modest size of the relationship—and the fact that it varies substantially across contexts and samples—suggests the importance of other motives, such as being asked to donate, or the desire to experience the “warm glow” of giving, act in accordance with a particular identity, or enhance one’s reputation (see Bekkers & Wiepking, 2011 for other drivers of charitable giving).

Sargeant and Lee (2004a, 2004b) propose that the importance of trust is mediated (or explained) by relationship commitment (see also Shang et al., 2019). In other words, it is argued that trust in a particular nonprofit cultivates a sense of commitment—a desire to maintain the relationship with that organization—which in turn promotes giving. It would have been ideal to test this relationship meta-analytically. However, commitment (and other potential mediators) were not included in sufficient studies and therefore could not be analyzed. Nevertheless, we speculate that the focal finding here—that organizational trust is more important for giving than other types—lends weight to the trust-commitment-giving model’s approach. Those authors explicitly say that the model applies best to the question of trust in a particular organization (vs trust in the wider sector; Sargeant & Lee, 2004a).

Relatedly, prior research has often focused on generalized trust as a driver of giving (Bekkers, 2003; Evers & Gesthuizen, 2011; Glanville et al., 2016; Uslaner, 2002). Indeed, generalized trust was measured in almost half of the articles reviewed here. Yet our results demonstrate that generalized and institutional trust are only weakly associated with giving. This highlights a gap between researcher assumptions and the evidence base.

Rather than discounting the role that generalized trust plays, we suggest future studies actively compare and contrast different forms of trust within the same study to understand how they inter-relate. Seminal theories of trust (e.g., Mayer et al., 1995) highlight that dispositional and referent-specific forms of trust combine to influence outcomes, and this has been supported by multiple empirical studies on trust (see Fulmer & Gelfand, 2012, for a review). Management research has also shown that dispositional forms of trust are more influential at the start of a relationship, before trust in the specific organization has been established (Mayer et al., 1995; van der Werff & Buckley, 2017). Given that our study shows that dispositional, organizational, and sectoral trust are all associated with giving, it will be important to consider how they may combine and interact to impact giving over the tenure of a donor’s relationship with a particular nonprofit. Longitudinal studies can be employed to examine the importance of trust dynamically over time: for example, does trust matter more at the point of acquisition or later in the supporter journey when relationship commitment has been established?

Our results indicate that dispositional forms of trust (i.e., generalized and institutional) are less strongly associated with giving than trust in charity-specific targets (i.e., sectoral and organizational trust). The triadic approach to understanding giving that was mentioned in the introduction can shed light on these findings. Chapman (2019) argues that it is not enough to understand the dispositions and characteristics of donors—such as their general propensity to trust—but instead one must consider the relationships that exist between donors, beneficiaries, and fundraising organizations and how these influence donor decisions (see also Chapman et al., 2021). This study shows that the unique relationship of trust between a donor and a fundraising organization is a much stronger predictor of giving than the donor’s general disposition to trust others.

Extrapolating from this triadic theory of giving, it is also possible that the type of beneficiary may further influence the importance of a relationship of trust existing between the donor and the fundraising organization. For example, when the beneficiary is highly valued (e.g., children), the importance of this relationship of trust may be accentuated because donors are more invested in positive outcomes for beneficiaries. Or when beneficiaries are distant (i.e., overseas), the relationship of trust may be more important as donors cannot see the work of the organization directly. Indeed, previous research supports the notion that trust is more important for giving to international charities (e.g., Wiepking, 2010). The current data do not allow us to test these ideas, but future research should consider how the beneficiary in question may also influence the role that trust plays in donor decisions (see also Chapman et al., 2018; Chapman, Masser, & Louis, 2020).