Perceptual Determinants of Nonprofit Giving Behavior in the United States John B. Ford, Professor of Marketing CBPA Dean’s Research Seminar, Old Dominion University Friday, November 11, 2005 The study will be forthcoming in Journal of Business Research. Introduction page 1 How and why individuals elect to offer their help to others is a topic that has puzzled philosophers and economists since the “dawn of antiquity” (Wispe 1978). Comparatively recently marketing’s contribution to “giving” has been recognized and a succession of authors have demonstrated its utility (Bendapudi et al. 1996; Kotler and Andreason 1987; Lovelock and Weinberg 1984; Varadarajan and Menon 1988). Introduction page 2 In the U.S. the total value of gifts from individuals was $159.32 Billion in 1998/9 (Giving USA 2000). In the U.K. through the period 1986-1996, aggregate charity income from individuals rose 38% in real terms from $4.2 Billion to $5.2 Billion (U.S. equivalent) outpacing growth in GDP of 20% during the same period (Pharoah 2000). Since 82% of the US and 80% of the UK population give either time or monetary donations to nonprofits, the economic and social significance of the sector becomes readily apparent. Introduction page 3 Studies concerned with monetary donations have tended to regard the decision to donate (or not) as the primary output from any model of giving behavior (Sargeant 1999). While some studies have addressed the factors driving the value of gifts, much of the existing literature has focused on distinguishing donors from non-donors (Schlegelmilch et al. 1992). Fundraisers have become increasingly interested in 1) the level of the gift, 2) the likely lifetime value of the donor, and 3) the extent to which the donor may be persuaded to support the organization for extended periods of time. Introduction page 4 Guy and Patton (1989) and Burnett and Wood (1988) represent the first real attempts to develop composite models of giving behavior, but neither study focused on donor perceptions of various aspects of the recipient organization or the charity sector as a whole. It is reasonable to expect that donations may be driven by such factors as the perceived reputation/professionalism of the organization, and, in the cases where a relationship already exists, the perceptions of the quality of service provided in the past. Introduction page 5 Our goal was to focus on the individual determinants of giving as a subset of factors related to the individual within the variety of comprehensive (theoretical) models of giving behavior presented in the marketing literature (e.g., Bendapudi et al. 1996). Perceptual Determinants page 1 Phase I was designed to elicit a list of relevant perceptual factors that could be tested quantitatively using a preliminary sample in Phase II to assess psychometric properties before moving in Phase III to a larger survey sample to use for structural equations modeling purposes. Phase I began with a review of the literature to elicit items for the pool which was followed by a series of eight focus groups each with 8-10 members and stratified by socioeconomic group and age. Perceptual Determinants page 2 The focus groups allowed for the exploration of public perceptions of charities and key variables that would impact on the decision to donate. Each group ran for around 90 minutes and discussions were taped, transcribed and subjected to content analysis using QSR N5 (Nud*ist). A series of potential scale items were identified and commonalities were assessed across the transcriptions. Perceptual Determinants page 3 The content analysis suggested that there were two distinct categories of constructs which should be included in any reasonable modeling: 1) Perceptions of the benefits (if any) that might accrue from support and 2) Perceptions of the behavior of the specific (i.e., supported) organization. Perceptual Determinants page 4 The content analysis also suggested that the impact of these perceptual constructs could be mediated by the additional constructs of trust and commitment. It should be noted that a different set of relationships among the constructs was identified from the focus groups than was suggested by the literature. This resulted in our development and testing of two different models. Trust, Commitment and Giving Behavior page 1 Few empirical studies have addressed the role of trust in relation to nonprofit-donor relationships (Sargeant and Lee 2004). There is now a wealth of empirical evidence to support the critical role that trust can play in influencing relationships between companies and customers (e.g., Bensaou and Venkatraman 1995, Gounans 2005). These studies concluded that higher levels of trust improve the likelihood that a relationship will be entered into. Trust, Commitment and Giving Behavior page 2 It has also been found that where a relationship already exists, higher levels of commitment will be generated by virtue of the presence of trust and that higher levels of both sales and loyalty will accrue as a consequence (Anderson and Weitz 1989, Dwyer et al. 1987). In such relationships, trust is viewed as particularly important where intangible services are provided because consumers often lack objective criteria to assess the performance of a relationship (Coleman 1990). Trust, Commitment and Giving Behavior page 3 This has relevance to the voluntary sector context where not only is the service provided to donors often highly intangible (Polonsky and MacDonald 2000) but the service provided to the beneficiary group (as a consequence of the donation) can frequently not be assessed by donors at all. Donors must rely on the nonprofit to deliver benefits to society that have either been explicitly or implicitly promised (Hansmann 1980). Trust, Commitment and Giving Behavior page 4 Trust, commitment and giving behavior are related sequentially. As a result, the following hypotheses are posited: H1: There is a positive causal link between the degree of commitment and donor giving behavior. H2: There is a positive causal link between the degree of trust and the degree of commitment. Perception of Benefits page 1 Social exchange theory suggests that donors will often be motivated to give because they perceive some benefits will accrue to them as a consequence of their gift (e.g., Amos 1982). These benefits may be categorized as “demonstrable” in the sense that some tangible benefit might accrue to the donor, “emotional,” where a donor derives a sense of self worth from making the gift or “familial” is the sense that these benefits might accrue on behalf of (or because of) a family member or loved one. Perception of Benefits page 2 Social exchange theory suggests that donors will often be motivated to give because they perceive some benefits will accrue to them as a consequence of their gift (e.g., Amos 1982). These benefits may be categorized as “demonstrable” in the sense that some tangible benefit might accrue to the donor, “emotional,” where a donor derives a sense of self worth from making the gift or, “familial,” in the sense that these benefits might accrue on behalf of (or because of) a family member or loved one. Perception of Benefits page 3 In conceptualizing these three categories of benefits and their connection to giving, the literature suggests that the ability of the nonprofit to supply a particular package of benefits would tend to build trust in and commitment to the organization. Since this study is concerned with regular givers to charities, receiving and responding to ongoing communications with the nonprofits, rather than donors responding to street collections (Burnett 1992 suggests that these would be considered impulse forms of giving), the theory suggests that trust and commitment would play sequential and mediating roles between donor perceptions and their recorded giving behavior. Perception of Benefits page 4 As a result, the following hypotheses are offered: H3a: There is a positive causal link between the degree of perceived demonstrable utility and donor giving behavior mediated by trust and commitment. H4a: There is a positive causal link between the degree of perceived emotional utility and donor giving behavior mediated by trust and commitment. H5a: There is a positive causal link between the degree of perceived familial utility and donor giving behavior mediated by trust and commitment. Perception of Benefits page 5 During the focus groups it became clear to us that donors felt that the perceived benefits of their association with the nonprofit were directly influenced by the level of commitment that they felt to the organization. Given the disparity between the conclusions of the focus group and existing theory, the following alternative hypotheses were suggested: H3b: There is a positive causal link between the degree of perceived demonstrable utility and commitment. H4b: There is a positive causal link between the degree of perceived emotional utility and commitment. H5b: There is a positive causal link between the degree of perceived familial utility and commitment. Organizational Factors Performance of the Organization: page 1 Tonkiss and Passey (1999) argue that the extent to which a potential donor has trust in an organization will be driven by the extent to which they believe that it has demonstrated that it will use the funds wisely. Trust has been viewed as developing over time as donors receive feedback on how their previous donations have been used (Burnett 1992). It is therefore suggested that: • H6: There is a positive causal link between the perceived performance of a nonprofit and trust in the organization. Organizational Factors Responsiveness: page 2 Focus group data further suggested that the way in which the organization was perceived as treating its donors would drive trust in the organization and as a result giving. Burnett (1992) was the first to advocate that the tools and techniques of relationship marketing should be applied to fundraising with nonprofits taking account of the needs of their donors. It has been suggested that higher value donors place greater emphasis on an organization’s perceived responsiveness to their needs (Burlingame 1997, Sargeant and MacKenzie (1998). It is therefore suggested that: • H7: There is a positive causal link between the degree of perceived responsiveness of a nonprofit organization and donor trust in the organization. Organizational Factors Communication: page 3 Work by Schlegelmilch et al. (1992) and Greenfield (1996) found that the perceived quality of communication can impact giving, but the data suggests that this may again be mediated through the construct of trust. It is therefore suggested that: • H8: There is a positive causal link between the perception a donor might have of fundraising communications from a nonprofit organization and donor trust in that organization. Construct Scale Development Page 1 Phase II of the research comprised identifying appropriate scales to measure the various study constructs. Commitment was straightforward as after considerable review it was decided that an adaptation of the scale developed by Morgan and Hunt (1994) was appropriate for this particular context. During the pretest an acceptable Alpha Coefficient of 0.89 was obtained for the scale. Similarly, it was possible to employ the trust scale developed by Sargeant and Lee (2004) specifically in the fundraising context, and here an acceptable Alpha Coeffiecient was obtained in pretest of 0.96. Construct Scale Development Page 2 In measuring the remaining six perceptual constructs noted in the models, the protocols suggested by Churchill (1979) were followed. A review of the pertinent literature suggested an appropriate pool of 50 items to measure the constructs. This item pool was then subject to scrutiny by a panel of judges consisting of faculty and senior nonprofit professionals. Each judge was provided with a definition of each construct and asked to categorize each item according to these definitions. Scale Development Protocols source: Churchill JMR 1979 Construct Scale Development Page 3 Following Pritchard et al. (1999), panel members were also required to appraise each item for its appropriateness and clarity on a measurement scale from 1-5. A consensus in respect to categorization, fit and clarity ratings (i.e., 4 or above) was sufficient to admit items to the final item pool in each case. This process concluded with the retention of 42 items. Five point numeric bipolar scales ranging from 1=strongly disagree to 5=strongly agree were then attached to each statement. Construct Scale Development Page 4 A convenience sample was chosen for the purposes of scale purification made up of 1000 individuals working in three public institutions in a large mid-Western city who had all given to nonprofits. Questionnaires were distributed by managers and returned to a central point for collecting and processing. Subsequent analysis revealed that 25 questionnaires were incomplete or unuseable, which left a final sample of 975. Scale purification procedures were then applied to the scales. Survey responses were split into two halves to cross validate any decisions that might be made in respect to item reduction (Pritchard et al. 1992). Construct Scale Development Page 5 The procedure began with an analysis of Alpha Coefficients for each of the attitudinal statements. In accordance with Zaichowsky (1985), it was decided to eliminate items which improved corresponding Alphas to the point where all retained items had corrected item to total correlations greater than 0.40. This process was cross validated between samples and items common to both split samples were retained, which resulted in a pool of 30 items. Each set of items was then subjected to an exploratory factor analysis. Cattel’s scree test indicated that a 6-factor solution was preferable for each of the samples. Construct Scale Development Page 6 Factor structures appeared to be consistent when compared, but there were some mixed item themes which appeared. To eliminate this inconsistency, a further reduction of scale items was undertaken and the analysis repeated. A common core of 26 items was then generated. Finally, both samples were combined to re-examine internal homogeneity. It proved necessary to remove two additional items. The remaining 24 statements were then subjected to a further exploratory factor analysis with a 6-factor solution being specified. Each factor had an eigenvalue greater than 1.00, and there were no cross-loadings at higher than .200. Construct Scale Development Page 7 In terms of reliability, all the scales exhibited Cronbach Alpha Coefficients of near or greater than .70. It should be noted that Emotional Utility and Performance of the Organization, both included scales of only two items. These two constructs are therefore underidentified, and even though the inter-item correlations were high, it is worth noting that construct scales should contain at least 3 or 4 items to be properly identified. We were now at a point to continue with Phase III. Construct Scale Development Page 8 Phase III tested the models depicted in the earlier figures. Data were obtained from a six-page mail questionnaire sent to 4,000 donors. A sample of 500 individuals was selected from the databases of 8 national, well-known and leading nonprofits, representing a diverse range of different causes. Considerable negotiation was required to gain this access. After one mailing 1,355 replies were returned, providing a useable response rate of 33.9%. It was unfortunately not possible to assess the extent of nonresponse bias due to the negotiation requirements, but a use of Armstrong and Overton (1977) indicated that there were no differences (p>.10) between early and late responses. Construct Scale Development Page 9 Giving histories (drawn from the nonprofit databases) were matched post hoc to questionnaire responses (using zipcode data) making it possible to add the following variables to the dataset: total amount given (to the nonprofit in question) and the number of gifts given. These were both utilized to calculate the mean gift offered to the organization in question. All construct scales used in the modeling were then subjected to Confirmatory Factor Analysis using AMOS 5.0. The six perceptual scales that were developed and refined were made up of the 24 items previously discussed, and these were analyzed along with the two previously validates scales for Trust and Commitment. This produced eight constructs with 32 total items for assessment. Construct Scale Development Page 10 These eight constructs were all assumed to be intercorrelated for CFA purposes. The CFA results were as follows: Chi-Square 679.5 with 437 degrees of freedom, IFI=.972, CFI=.971 and RMSEA=.035. An examination of the modification indices and standardized residuals suggested that there were no major sources of error in the final model. This then paved the way for the structural equations modeling. The Structural Equations Models Page 1 The two models were compared to determine which had the better fit and explanatory power. Once the better model is chosen, then the hypotheses can be tested by assessing the various linkages in the SEM. The central giving measure was the mean donation offered since both total amount given to the charity and the total number of gifts offered are both likely to be influenced by the date the donor was acquired rather than the constructs used in this study. Future Research page 1 The next step is to test the linkages involved using Structural Equations Modeling. The UK models have already been run and the fit was found to be reasonable at from .911 to .934. The ability to explain the dependent giving constructs (squared correlation) was found to be .15 for active givers, and this was improved to .24 when the amount given to charities during the previous year was logged due to skewness in the responses – reflecting a small number of large givers). Hypothesis Testing The results indicated the following: H1 was supported (a positive link between the degree of commitment and donor giving behavior: .36, p<.01). H2 was supported (a positive link between trust and commitment:.33, p<.01). H3b was not supported (no link between demonstrable utility and commitment). H4b was supported (positive link between emotional utility and commitment: .17, p<.01). H5b was supported (positive link between familial utility and commitment: .57, P<.01). H6 was supported (link between performance of the organization and trust: .26, P<.01). H7 was not supported (no link between responsiveness of the organization and trust). H8 was supported (positive link between communications from the organization and trust: .15, p<.01). Discussion Trust appears to be significantly affected by performance of the charitable organization and communication from the charity but not by responsiveness of the charity. Trust is therefore created by a nonprofit perceived by a donor to have had an impact on the cause and by maintaining appropriate communications with him or her rather than by being perceived as responding quickly to that donor’s individual issues or concerns. Trust is an important driver of commitment. But also, emotional utility and familial utility are similarly important drivers of commitment. The final relationship tested was between commitment and giving behavior as measured by the average gift, which was supported, although the proportion of giving behavior explained by commitment is rather modest. Future Research Specific causes should be assessed to determine whether the perceptual determinants or the role on specific determinants might vary by context. There is also a need to replicate this work in other geographical contexts as the manner in which the voluntary sector or the organizations that comprise it may vary from one country setting to another.