Putting Charity to the Test

Theories of Giving Time and Money

Over the last two decades, scholars—primarily economists, social psychologists and political scientists—have devoted significant attention to the behavioral mechanisms of giving time and money, with attempts to test these theories using observational data and both lab and field experiments. What has emerged is a suggestive theoretical framework for understanding time and money gifts, both separately and taken together. Although there have been some attempts to knit these literatures together, with some success, much work can still be done (Andreoni, 1989; Becker, 1976; Coleman, 1990; Fehr & Gintis, 2007; Monroe, 1994). For example, the economic approach to charitable giving relies heavily on a rational-actor, self-interested model of individual behavior, while sociologists and psychologists focus on the effects of norms, values and compliance (Fehr & Gintis, 2007). Early economic modeling predicted a “crowding out” effect in the provision of public goods, where individuals’ contributions are crowded out dollar-for-dollar with public provision of the good (Warr, 1982). However, after repeated observations of only partial crowding out, later models incorporated the concept of “impure altruism” or the benefit of a “warm glow” feeling (Andreoni, 1989, 1990) as another form of self-interest, entering utility functions through self-gratification (see, for example, Becker, 1976, 1978). In one study, Crumpler and Grossman (2008) found that even when a donation to a charity is pre-set, and each donation is crowded out by a proctor dollar-for-dollar, 57% of participants still made a donation. Others found various levels of crowding out (Andreoni, 1993; Eckel, Grossman, & Johnston, 2005). Matching subsidies and tax rebates may also effect charitable giving, as Eckel and Grossman (2008) found in a field experiment of public radio donors. Alternatively, Andreoni’s (2006) signaling model suggests individuals determine quality through a lead gift, and some argue that individuals give to a charity to signal their own wealth and gain prestige (Glazer & Konrad, 1996; Harbaugh, 1998).

Importantly, the economic literature on volunteering theorizes a dynamic and interactive relationship with giving money, when much nonprofit research treats gifts of time and money as mutually exclusive. This need not be the case as intuition suggests that individuals may choose to give a mix of both. Specifically, economists have sought to determine if time and money are perfect substitutes (one crowds out the other dollar-for-dollar), or if they complement each other (both increase and decrease together). Similar to giving money, individuals receive a “warm glow” for their donations of time to an organization. However, Andreoni et al. (2004) argue that individuals prefer to give the highest value gift, and since money is perceived as more productive, it is therefore of higher value to the organization than time. On the other hand, other research indicates that this relationship is more complementary (Apinunmahakul, Barham, & Devlin, 2009; Brown & Lankford, 1992; Freeman, 1997). Other expected benefits, including prestige, skill building, job training, and access to information have also been hypothesized as motives in volunteering (Handy et al., 2010; Schiff, 1990; Steinberg, 1997; Wilson, 2012).

Much of this later work has departed from the limits of a pure rational choice approach as it grappled with the observed complexity of charitable giving. As Fehr and Gintis (2007, p. 60) write, “Indeed, a major thrust of experimental economics has been to convince economists of the need for a far richer notion of human preferences than those given in traditional economic theory.” To address these “human preferences” sociologists and social psychologists have also sought to address many of the same questions of motivation in charitable giving and altruism, and move beyond the single individual to the effect of the interactions of that person with their environment. This approach “tends to explain altruism as an outcome of a decision-making process in which the internal characteristics of the actors join with the external environment in a pattern of mutual influence” (Monroe, 1994, p. 881). Individuals may be incentivized to behave prosocially in an attempt to avoid feelings of guilt or create feelings of pride (Bator & Cialdini, 2000; Batson & Powell, 2003; Benabou & Tirole, 2006), especially if their behavior is made public (Panagopoulos, 2010; Smith, Webster, Parrott, & Eyre, 2002). Some have found that making a pledge to act increases the likelihood of carrying it out as individuals feel accountable to their promises (Bator & Cialdini, 2000; Cialdini & Goldstein, 2004; Cotterill, John, & Richardson, 2010; Freedman & Fraser, 1966). People also appear to be sensitive to public perception through social monitoring (Posner & Rasmusen, 1999; Rind & Benjamin, 1994), or public acknowledgement (Panagopoulos, 2010). Additionally, the possibility of making friends can also be a determinant of volunteering (Prouteau & Wolff, 2008; Wilson & Musick, 1997).

A Comparative Analysis of the Research Toolbox

Unfortunately, most of this research has not been published in nonprofit journals, making it difficult to spread this knowledge across the sector. Considering the multidisciplinary nature of research on the sector and the crucial role charity plays in nonprofit organizations, this is surprising. Why have many scholars not included field experiments in their research? A lack of training or experience in field experiments, or perceived cost, may be preventing many researchers from choosing this method. However, experiments can provide important insights to the relevant behavioral mechanisms and can provide a unique bridge between the literatures, as well as between lab experiments and observational data (List, 2008). Instead of embracing field experiments, however, most empirical research in the nonprofit sector is conducted with the use of either observational data available from a few discrete sources or responses to independent surveys constructed by scholars in the field. Unfortunately, problems remain in relying solely on observational data and surveys to understand causal relationships.

First, while all statistical research designs using observational data depend on a set of assumptions regarding the quality and accuracy of the data, these problems are often exacerbated in nonprofit research. Grønbjerg and Clerkin (2005, p. 233) state that “almost without exception, the existing nonprofit information systems are limited in scope” and do not include many variables that might be of interest to the researchers, such as organizations structure, staffing, or use of volunteers. Tinkelman and Neely (2011) discussed the challenges they faced with data accuracy and heteroskedasticity while utilizing IRS filings of nonprofits (Form 990) available through the National Center for Charitable Statistics (NCCS). In addition, much extant data sources provide data only at the organization level, making it impossible to study individuals as the unit of analysis.

Field experiments, on the other hand, can provide a powerful methodology to measure and analyze the relationships between variables. Importantly, random assignment of participants in an experiment allows for unobserved characteristics across the sample to be equally applied to both control and treatment groups. Proper assignment to the treatment and control groups assures that each observation will have no systematic relationship to any other variable, whether observed or unobserved, and will support unbiased estimates. In one experiment concerning charitable giving, Shang and Croson (2009) test “social information” on charitable contributions—that is, whether or not an individual will give, or give more if they know how much others are giving compared with those receiving no information. Setting their experiment in a public radio station (a recognized public good) allowed them to operate in an environment where free-riding is easy and likely. As individuals called in during a pledge drive, the experimental team posing as volunteers responded with either no information about the contributions of other donors or stated that a previous caller had made a US$75, US$180, or US$300 gift. The experimenters were blind to the treatment on each call until they removed a post-it note with information about the specific manipulation to be used (or not). The authors randomized the treatments between each experimenter and each hour to ensure that the callers were distributed randomly, and evenly, between the treatment options. Their results suggested that “social information” did indeed have an effect on giving for new supporters, and especially at the higher levels, and also increased the donors’ future giving. By using effective research designs such as this, scholars are able to manipulate variables and begin to eliminate alternative explanations. Indeed, Nobel Memorial Prize winner Elinor Ostrom stated, “Careful experimental research designs frequently help sort out competing hypotheses more effectively than does trying to find the precise combination of variables in the field” (Ostrom, 1998, p. 17).

Many researchers also use surveys, and a significant body of literature has been developed to address the challenges inherent in survey methodology, including an entire issue of Nonprofit and Voluntary Sector Quarterly (2001, Vol. 30, Issue 3) and Voluntas (1993 Vol. 4, Issue 2). Hall (2001, p. 515) wrote, “Surveys are frequently used to collect data about giving and volunteering; however, the quality of the data is seldom known, and the measurement challenges inherent in such surveys are not well recognized.” Different sampling techniques across data sets, sampling errors, lack of randomization, response rates and changing technologies such as web surveys make the use of surveys difficult and can result in biased estimates (Cnaan et al., 2011; Hager, Wilson, Pollak, & Rooney, 2003; Hall, 2001; Kennedy & Vargus, 2001; Lin & Van Ryzin, 2011). Non-responders, specifically, may not only weaken generalizability, but may also lead to missing critical, yet unobserved, characteristics. For example, Smith (1997) found behavioral differences between organizations that responded to a survey of peace groups from those that did not, while Rooney et al. (2004), Hall (2001) and Steinberg, et al. (2002) found that the length of survey questions regarding volunteering and giving had an impact on the quality of responses. Field experiments can help circumnavigate these problems by allowing researchers to study actual behavior in real time, rather than a respondent’s memory, perception or opinion of events.

In addition, Gerber and Green (2003) argue that the generalizability is less affected in experiments than in survey or observational data. They write, “Field experiments may be imperfect, but if uncertainty about bias is substantially lower for this type of research, they will effectively trump observational and lab studies” (p. 100). The generalizability of a study can also be strengthened and confirmed with additional experiments and other quantitative methods, such as using observational or survey data.

An Experimenting Sector

The research community, especially nonprofit research, often negotiates the dual interests of theory and real world applicability (Bushouse & Sowa, 2011) and field experiments can provide one bridge linking theory and practice. In addition, nonprofit scholars are often well suited to work closely with practitioners, providing opportunities for effective partnerships. The results of research may not only benefit the partnering organization (by realizing increased revenue or volunteer participation), but can benefit the sector as a whole. While cost may be a factor in implementing an effective experiment, it is likely that many organizations would be amenable to working with researchers through shared desire for efficiency and effectiveness. In addition, scholars can also adjust the experimental design to the needs or context of the organization and still be able to leverage the strengths of random sampling techniques. Although Loewen et al. (2010) focus on political elites and campaigns, many of their recommendations are relevant to working with other third-party organizations. They acknowledge that while real world experiments are fraught with chaos, they recommend that researchers embrace ambitious projects that both scholars and practitioners are interested in exploring. They encourage all scholars to be flexible, maintain good documentation, and build relationships with organizations, lessons that many nonprofit scholars already know from working with organizations on a diverse research agenda. It is important to note, however, that samples selected for experimentation may only be representative of those particular populations in that place and time. In one field experiment on pledging and publicity in charitable giving, Cotterill et al. (2010) mailed a notice about a book drive to a randomized sample in two distinct neighborhood, one higher income than the other. The neighbourhoods had different giving rates, with the higher income population donating books at approximately twice the rate as the lower-income neighbourhood. This demonstrates that different populations and samples, based on observed or unobserved characteristics, may behave quite differently in a similar situation. However, this challenge can be mitigated with additional replication of the study.

An experimental research agenda on charitable giving offers many opportunities to address as yet unanswered questions. What is the relationship between time and money? Are donors more likely to volunteer or contribute money if given the choice between the two? Are the underlying mechanisms of time and money gifts similar, or divergent? It stands to reason that although they may interact with each other, there may be different motives for people in selecting one over the other. In addition, what is the extent of “warm glow” giving, and can it be crowded out with extrinsic rewards such as trinkets or other benefits? These and other questions demonstrate that there are opportunities for an aggressive experimental research agenda to make a lasting contribution to the field.

Conclusion

Scholars are just beginning to scratch the surface in understanding the complex environments of nonprofit organizations and the dynamic motivations of donors and volunteers. This research note has provided a roadmap of the leading theories on both giving and volunteering, and suggests opportunities to study this phenomenon using the strengths of field experiments. Ultimately, scholars should include experimentation as one option in an agressive research agenda that includes other quantitative and qualitative methods. Nonprofit studies would benefit from expanding existing methodologies and embracing experiments as a way to learn more about the individuals that so many nonprofits depend on for survival. Experiments may help not only add to these literatures, but support and strengthen the nonprofit sector generally by providing new tools and strategies to increase efficiency and efficacy.