From solicitation to search: a study of monitoring costs as a driver of donor giving behavior in online portal websites


Shena R. Ashley, Department of Public Administration, Maxwell School Of Citizenship and Public Affairs, Syracuse University, 215 Eggers Hall, Suite 215 F, Syracuse, New York, USA.



This study focuses on market contexts, in particular online giving portals, where donors are able to search and select organizations to contribute to instead of responding to solicitations. To understand donor behavior in this context, we construct and test a two-stage donor search model. A relationship between the behavior of the donor and that of the charitable organization is sought along two dimensions: how organizational behavior involving the reduction of monitoring costs affects the amount that a donor gives and the likelihood that donors will make a subsequent contribution. Copyright © 2011 John Wiley & Sons, Ltd.

The increasing trend towards donating to nonprofit organizations through online giving portals1 turns the prevailing way of understanding the philanthropic transaction on its head. Online giving marketplaces provide a searchable database of organizations or projects that allow donors to “shop” for donation recipients on the basis of their interests. In such settings, where donors are in a position to search and choose among a large set of competing organizations, solicitation-based theoretical perspectives may be limited. The challenge, therefore, is to adjust and expand models of giving behavior to fit the emerging online giving portal market.

Although online giving portals offer more opportunity for self-direction by the donors, donation-seeking organizations nonetheless remain actively engaged and can influence donor behavior through the information they share and the way they share it. It is important that theory development in philanthropy capture these dynamics. We propose the application of economic search theory and develop a two-stage donor search model to understand how organizational actions, chiefly through the reduction of monitoring costs, influence donor behavior.

We are concerned with how organizations can undertake activities to decrease their associated monitoring costs. Donors want assurance that they can trust the organization to use the donated money for its stated purpose and are motivated by the perception that their donation will make a difference (Sargeant et al., 2001) in the outcome of a project. To gain this assurance, they incur costs associated with monitoring the extent to which the organization provides the socially valuable benefits that they expect. The main contention of our donor search model is that certain types of information provided by organizations can signal efficacy and influence donor behavior through a reduction of monitoring costs. This is an issue of considerable practical importance because—given the ways in which online giving portals operate outside the traditional model of donor solicitation—monitoring costs are among the few factors over which organizations raising funds can exert some control. Because of both the practical and theoretical significance of these emerging donor-driven market contexts, it is our aim to highlight the import of monitoring costs and their influence on donor giving behavior.

Using an observational data set provided by the online charitable marketplace GlobalGiving,2 we investigate the effects of monitoring cost manipulation through project updates posted to the site, on both the initial amount of the donor's gift and his likelihood of repeat giving. Because of GlobalGiving's market structure and the requirement that its participating organizations post project updates, we are able to formalize the abstract notion of monitoring cost by using the number and timing of project updates as proxies. We investigate the impact of differences in the number of project update reports on both the amount of the donor's gift and their likelihood of repeat giving.

The results are complicated by discussed validity issues and in general show that within the context of GlobalGiving proxied differences in monitoring costs do not appear to have a significant practical effect on the amount of either the initial gift or repeat giving by a given donor. Some significance is shown for various constructions, although the results are not shown to be robust. Potential explanations are offered for the lack of significance, including the unique structure of GlobalGiving's marketplace or the possibility that solicitation outside of the portal still has a significant impact on donor decision making. Although the analysis of the GlobalGiving data is an important first step in the verification of the donor search model, future research is needed to yield more conclusive results.

Application of search theory

Search theory has been influential in many areas of economics (Stigler, 1961; Rothschild, 1973; Lippman and McCall, 1979) and applied to diverse areas such as marriage (Becker, 1973) and labor markets (Jovanovic, 1979) where, like philanthropy, there are dynamic market exchange and conditions of information asymmetry. Search theory, at its core, recognizes that individuals make decisions on the basis of the information they have at a given point in time, and that in general, better decisions are made when more information is available. Furthermore, the theory recognizes that the acquisition of additional information is costly (Williamson, 1979). As such, how much information an individual gathers during the course of the decision-making process depends on the costs associated with information gathering. These costs can be inhibitors to effective search, forcing one to make a sub-optimal decision.

Monitoring costs

The primary goal of a search model is to understand the effect of information-related costs on actor behavior. In charitable giving, costs related to providing a donation to a given organization can be viewed as being composed of two components. The first component can be considered to consist of standard structural costs associated with making a contribution to a charitable organization, such as the time it takes to write a check to the organization. The second component of the cost is the perceived and real costs to the donor to stay informed about organizational activities and outcomes (or monitoring costs). In this sense, the donor has to pay the cost to find out how well the organization matches with his interests or altruistic preferences. Given the explanation of what the cost represents, it is also important to consider what the manipulation of the cost means (i.e., why might the cost for one organization be smaller than for another?). Making the assumption that all organizations start out with the same cost, an organization could reduce its cost through increased transparency or providing updates to donors about mission success, effectively reducing the cost of finding this information on their own.

The notion that organizations can alter monitoring costs to influence giving behavior integrates and builds on two categories of recognized drivers of donor behavior detailed by Bekkers and Wiepking (2011). The first driver is related to costs. Both absolute and perceived costs can serve as barriers to giving, and donors are more willing to give when the costs of making a donation are lowered. In this vein, we expect that donors will respond to monitoring costs —the costs associated with gathering information to monitor and determine satisfaction with the organization—as they do to other costs. The second driver is efficacy, which stresses that donors are motivated by the perception that their donations make a difference. This is related to how donors perceive the facets of the organization and its services (Sargeant et al., 2001). Concern with efficacy motivates the need to monitor organizations. Services provided through nonprofit organizations are frequently not assessed easily by donors. Consequently, donors must trust organizations to deliver the expected services (Steinberg, 2006). Organizations, through the provision of information that signals that the donation will be used to generate the socially valuable service the donor intends, can reduce the need for donors to monitor for efficacy. Together, these two drivers elucidate the mechanisms through which monitoring costs are expected to affect giving behavior.

Two-stage donor search model

We proposed a two-stage donor search model3 (presented in Figure 1) to understand giving behavior in contexts, such as online giving portals, where donors are self-directed. The basic economic search model has to be extended to capture some of the unique attributes of the nonprofit context. To do this, we drew from the varying theoretical perspectives of philanthropic behavior and the aforementioned assumptions about monitoring costs, which are outlined below:

  • There are altruistic donors who want to give money to charitable causes (Andreoni, 1990; Wilhelm and Bekkers, 2010).
  • Individual donors have preferences over which of the causes are the most important to them (Wagner and Wheeler, 1969; Duncan, 2004).
  • Within the causes, the donor has preferences over which organizations he or she would like to contribute to (Hibbert and Horne, 1996; Sargeant et al., 2001).
  • The donor knows his or her preferences but must take time to find the organization that matches them best (i.e., information about best matches is not readily available and may require a commitment of the donor's resources to obtain it.) (Glazer and Konrad, 1996).
  • Nonprofit organizations will act strategically to obtain donor funds (Frumkin and Kim, 2001).
  • A donor must pay some cost to invest in the organization and to subsequently monitor the behavior of the organization. Monitoring is necessary in that the donor needs to acquire information, ex post, to come to a better understanding of how well he or she matches with the organization and its outcomes.
Figure 1.

Two-stage donor search model

The model consists of an introductory stage in which donors discover a particular project or organization. The donors, once introduced, must make a decision about whether to contribute to the project or take a pass and continue looking for a worthy project (i.e., accept or defer). In this context, the term “match” is used to denote a donor and organization pairing that is stable, in the sense that the donor does not desire to break the pairing by withholding future donations. The donor's first stage decision of whether to pass or not is driven by two factors: (1) the expected benefit from making the contribution, of which there are both material and psychological components; and (2) the donor's view of the cost to monitor the organization.

If the perceived costs of donating to a specific organization relative to others are low enough, then the donor will choose to make a first stage contribution. In this sense, the donor's benefit from donating in this initial period is the experience of the expression of altruism, and real and psychological benefits less the cost of monitoring. If the donor chooses not to commit to the current organization, he or she goes back to the initial phase and encounters another organization via the same introduction mechanism. However, if the donor chooses to pay the cost and donate, in the next period they are shown the true value of the match and then presented with the option of either continuing with this match in the future and leaving the search process or discarding their current match and continuing the search process (Wardell, 2009). This decision differs from the first stage decision because presumably the donor has more information about the quality of the project or organization and his level of satisfaction with the internalized benefits. At both stages, the model assumes that donor behavior is motivated by real and perceived monitoring costs, and that the costs can be managed through information sharing from the organization to the donor.

Two testable hypotheses emerge from the model with regard to donor and organizational interaction within giving markets.

Hypothesis 1. Controlling for all other attributes of both the donor and the organization, organizations with lower monitoring costs will attract greater donation support. Hypothesis 1 (H1) says that a donor who is already predisposed to give to an organization will increase his donation level in relation to monitoring costs associated with that organization.4

Hypothesis 2. Controlling for all other attributes of both the donor and the organization, organizations with lower monitoring costs will be more likely to attract subsequent donations from previous donors.

A test of the donor search model using GlobalGiving

The online giving portal used for this analysis, GlobalGiving, is a website designed to allow project leaders, once their project has been accepted, to post information about their project and solicit donations for their work via a dedicated page on the site. Leaders provide a summary of their project along with photos, information on the problem it is attempting to address, and the potential for long-term impact. Included alongside this information is a donation form that enumerates various donation levels and the tangible impact that a particular level can have (e.g., $95 will send one poor child to school for a year).

The success of GlobalGiving as a fund-raising portal for grassroots projects can perhaps be understood along the dimensions of variety and credibility. Allowing the donors to capture a wide range of altruistic preferences, GlobalGiving provides a large variety of options for potential donors, with over 1500 projects listed on it. More specifically, the site contains projects from close to 100 countries. In addition to geographical variation, the projects also vary by topic. GlobalGiving provides the donor with projects from 17 topic groupings that, almost the same as the geographical variation, allow the site to retain donors over a wide range of giving interests.5

Donors can browse the website, filter projects by several criteria, and pick the one that matches their interests most. Donors make a tax-deductible donation, and their gift is combined with the gift of other donors doing the same thing. GlobalGiving ensures that 85–90% of the donation is on the ground within 60 days. Subsequently, donors obtain project updates that provide insight into the work of the organization and the presumed impact of their contribution posted by the project leaders.

Every project leader must write project updates that are posted on the website and e-mailed to past donors.6 GlobalGiving recommends that project leaders post updates at least quarterly. However, project leaders decide for themselves how often they post and what they will include in their updates. Painting the organization in a positive light, updates typically include photos, project activity data and stories, financial accountability measures, and output data.

As a kind of clearinghouse for grassroots organizations, GlobalGiving fosters credibility and transparency through several mechanisms. One of the primary mechanisms in providing credibility to projects listed on the site is the due diligence process that GlobalGiving adheres to when vetting projects to appear on the website.7 This process provides the donor with upfront knowledge that, at minimum, each of the listed projects meets the requisite qualifications.

This due diligence is supplemented by a money-back guarantee, in the form of a voucher, for any donor who is not satisfied with the giving experience. Almost the same as in the for-profit sector, this guarantee helps to underscore a level of confidence in the quality of the product that GlobalGiving is offering, which in this case is the set of charitable projects (Moorthy and Srinivasan, 1995). Collectively, these mechanisms help to provide both GlobalGiving and projects posted on the site with a strong brand as it concerns credibility.

Data and measurements


For this empirical study of the two-stage donor search model, GlobalGiving provided an anonymized data set of all US-originated donations spanning an 11-month period from 1 February 2008 to 31 December 2008.8 The analysis of H1 draws on a set of over 11 000 donations, and the second hypothesis (H2), which focuses on repeat giving by a donor, includes more than 7000 donations.9 Of particular interest in the data, we generated two variables that track the number of project updates that are posted by each project and the number of days between project updates. We utilized these variables as proxies for the perceived monitoring costs associated with projects. Organizations that post a greater number of updates relative to other organizations may be perceived to be more transparent to donors and therefore, require less monitoring. As described in the two-stage donor search model and in our hypotheses, the provision of information to reduce monitoring costs will influence donors to give more and to provide subsequent donations.

Econometric specification

Panel regression model with fixed effects

If Yijt is the amount donated by donor i to project j in period t, then the regression function below defines the relationship between project attributes, donor characteristics, and Y.

display math

The number of project updates posted at the time the donation was made along with the days since the last project update are used as proxies for monitoring costs and are the primary regressors of interest. However, in addition to the variables of interest, it is also necessary to control other relevant project characteristics, which might contribute to the decision of how much to donate to a particular project. In particular, the number of previous donors to a project at the time of donation along with the amount of funding the project has received to date as a percentage of its funding goal are considered to be influential in shaping a donor's donation level decision. A control for the project's age, in days, is also included. Lastly, a control is included via a dummy to distinguish between the projects that have at least one project update and those that have none.

In addition to project attributes, several observable donor characteristics are also considered. Although most donor characteristics remain unobservable, the existence of the characteristics such as whether or not the donor subscribes to the GlobalGiving newsletter, or whether or not their donation was made via a guest checkout allows one to segment the donor population at a very coarse level. We provided descriptive statistics for relevant variables in Table 1.

Table 1. Summary statistics for fixed effects panel model regressors
VariableMeanStd. Dev.
Project updates1.5522.716
Days since the last project update23.96237.213
Previous donors456.018782.060
Amount of funding (%)18.3318.771
Project age (days)39.65753.807
Subscribe (0/1)0.3650.481
Guest checkout (0/1)0.1130.316
At least one project update (0/1)0.4680.499

Project characteristics vary across several dimensions, most notably the theme of the project and the country in which it takes place. It is assumed that individual donors have preferences over the type of projects they like to fund. Although the project theme is observable, one can imagine other project attributes that are not observable within the set that may bias a donor's decision process. One such unobservable attribute might be the public opinion as it concerns the perceived value of the project. A project may be generally perceived by the public to be of high value, and so its higher donation levels are not the result of more or less relative monitoring costs but simply a function of this perception bias.10 Consequently, unobservable project-related variables might be correlated with both the dependent variable and an independent regressor. To control for these project specific unobservable attributes, a fixed effects regression model is used.

Although the model has been specified as one in which fixed effects are accounted for, there are also some concerns that unobservable time effects11 may play a role in shaping behavior as it regards a donor's gift amount. An attempt is made to control for time effects on the level of giving by adding dummy variables for each of the time periods for which the data set spans. The presented specification corresponds to specification (10) in Table 2.12

Table 2. Fixed effects panel regression model of the donation account ($)
  • a

    Dummies for each month (February–December) are used for specifications (9) and (10), but they are suppressed.

  • b

    This variable is a dummy for the period surrounding the Myanmar cyclone and Chinese earthquake. The dates for which it is true are: May 6–23.

  • c

    This variable is a dummy for the month of May, which includes two significant disaster events.

  • *

    p < 0.1;

  • **

    p < 0.05;

  • ***

    p < 0.01, standard errors in the parenthesis.

Days since the last project update−0.0285** (0.0120)−0.0228* (0.0120)−0.0267** (0.0121)−0.0272** (0.0123)−0.00715 (0.0168)−0.00874 (0.0175)−0.00412 (0.0176)0.00875 (0.0176)0.00654 (0.0181)0.000878 (0.0194)
Previous donors 0.00243 (0.00199)0.00334* (0.00202)0.00321 (0.00208) 0.00210 (0.00220)0.00178 (0.00220)0.000283 (0.00220)−0.000469 (0.00222)−0.000537 (0.00223)
At least one project update (0/1)   −0.262 (1.030)     −1.242 (1.144)
Project updates−1.314*** (0.136)−1.069*** (0.413)−1.184*** (0.416)−1.158*** (0.429)−1.227*** (0.146)−0.901* (0.461)−0.612 (0.480)−0.0765 (0.472)0.0506 (0.475)0.0788 (0.476)
Amount of funding (%) −0.148*** (0.0388)−0.169*** (0.0392)−0.164*** (0.0441) −0.156*** (0.0402)−0.155*** (0.0402)−0.143*** (0.0402)−0.135*** (0.0433)−0.118** (0.0470)
Subscribe (0/1)  −1.739*** (0.587)−1.739*** (0.587)−1.674*** (0.588)−1.704*** (0.588)−1.687*** (0.588)−1.583*** (0.586)−1.798*** (0.581)−1.544*** (0.586)
Guest checkout (0/1)  3.717*** (1.092)3.759*** (1.105)3.728*** (1.113)4.086*** (1.122)3.938*** (1.124)3.356*** (1.123) 4.076*** (1.250)
Project age (days)    −0.0288** (0.0135)−0.0210 (0.0147)−0.0217 (0.0147)−0.0119 (0.0148)0.217*** (0.0489)0.127** (0.0495)
Disaster window (0/1)b      2.567** (1.191)   
May (0/1)c       9.637*** (0.895)  
Constant43.41*** (0.425)44.51*** (0.540)44.97*** (0.587)45.02*** (0.625)44.10*** (0.497)45.21*** (0.610)43.96*** (0.839)40.05*** (0.895)60.53*** (14.92)60.40*** (14.93)
Observations11 42611 42611 42611 42611 42611 42611 42611 42611 42611 426
Number of projid251251251251251251251251251251

Multinomial logit model

A multinomial logit model is developed to test H2 with respect to the donor search model. We estimated the likelihood of selection among the donor choice set after the initial contribution to either make a subsequent contribution to the same project, make a contribution to a different project or not to contribute again. The set of regressors defined here are the same as in the previous section; however, because of the temporal nature of the dependent variable, between-period variables are also included within the model.13 For x containing the previously described regressors, a logit model over the donor choice set can be described as14 follows:

display math
Table 3. Relative risk ratios for the multinomial logit model to estimate selection into repeat giving (Do not return to GlobalGiving is the reference group)
VariablesaChoice equations
Give to different GG projectGive to same GG project
  • DF, degrees of freedom; GG, GlobalGiving.

  • a

    Dummies for projects and periods are suppressed.

  • *

    p < 0.1;

  • **

    p < 0.05;

  • ***

    p < 0.01, standard errors in the parenthesis.

Days since the last project update1.008 (0.00648)1.000 (0.00681)
Project updates1.175 (0.350)0.753 (0.250)
Previous donors0.990 (0.0127)0.997 (0.0190)
Amount of funding (%)1.138 (0.116)1.503 (0.544)
Donors since last project update0.989 (0.0127)0.996 (0.0189)
Updates since last project update1.118 (0.321)1.000 (0.316)
Funding increase since last project update (%)1.167 (0.117)1.487 (0.535)
Subscribe (0/1)1.318* (0.198)0.835 (0.142)
Guest checkout (0/1)1.189 (0.810)0.786 (0.606)
Project age (days)0.964***(0.0127)0.978 (0.0139)
Number of donor's previous donations4.198***(0.358)3.090***(0.278)
Log likelihood−1673−1673


Fixed effects panel regression results

The results from the fixed effects panel regression shown in Table 2 imply that neither of the monitoring cost proxy variables, days since the last update and the number of project updates, have a statistically significant impact on the amount of a donor's gift.15 Considering specification (10), which includes a full complement of regressors and controls for time effects, it can be seen that only two project-related attributes, the amount of funds raised by a project as the percentage of the fund-raising goal and the project's age, were shown to be significant in adjusting a donor's gift level. Although statistically significant, it is not clear that either regressor is economically significant in the sense that they make any practical difference in the level of gifts.

The project's age has an associated estimated coefficient of 0.13, which would imply that as a project's age increases by one day, the amount of the expected donation would also increase by approximately 13 cts. Considering project's range in age from 1 to 327 days, this would imply a range of $0.13–$42.51, which is economically significant within the context of a donation range of $10–$100. While at the lower end of the spectrum, the range is not very significant; as a project hits the 39-day mark, there is an increase of approximately $5 in the expected donation amount.

There are several factors that could explain the significance of the project's age on the expected donation amount. In particular, there may be a process of attrition taking place in which only high quality projects survive beyond a certain point, and lower quality projects are dropped from the available listing. Consequently, those left in the set are those projects that receive higher funds because of their quality, but the effect is seen via the project age regressor.16

The interpretation of the amount of funds raised variable is motivated by the literature on impact philanthropy. Impact philanthropists (Duncan, 2004) are donors who seek to maximize the impact of the last dollar of their gift. In this sense, a donor who goes to GlobalGiving ready to make a $20 donation may give only $10 to a project because it already has a high level of funding relative to its goal. The negative coefficient on the amount of funds raised variable suggests that, on average, the donors on the GlobalGiving site are impact philanthropists and will adjust their donation level downward as the percentage of the fund-raising goal completed is increased.17

The variables for both donors who subscribe to the newsletter and those who made a donation using guest checkout remain significant throughout the various specifications. Donors who subscribe to the GlobalGiving newsletter contribute $1.54 lesser, on average, than those who do not subscribe. In the same vein, donors who create GlobalGiving profiles (i.e., they do not use the guest checkout option) contribute $4.08 lesser, on average, than those who do not create a profile. Somewhat counterintuitive, a possible explanation for the results is that donors who subscribe to the newsletter or create profiles are more likely to contribute on a sustained basis but at lower amounts.

Contrary to H2, the results of the multinomial logit do not show any significant effects for project update variables (see Table 3). Whether or not the donor has given previously to any project on GlobalGiving is shown to be the primary driver in determining the likelihood of a donor returning to the site to make a subsequent donation after an initial contribution. For each additional donation, the donor becomes 4.20 times more likely to give to a different GlobalGiving project in the future than they are not to give again at all, and 3.09 times more likely to give to the same project in the future than not to give again at all. The project's age is shown to be significant in the comparison between the choice of giving to a different project and not giving at all, but it is not so in considering whether the donor gives again to the same project. In the instance of giving to a different project, it is shown that with each additional day that the project has existed, the less likely the donor will opt to give again in the future to a GlobalGiving project. Specifically, an increase of one day in a project's age will cause the donor to be 0.96 times more likely to be in the group of individuals that contribute again via GlobalGiving.

Conclusions and discussion

It was assumed that monitoring costs, manipulated by organizations through the provision of information, would have a positive effect on both the level of the gift and the likelihood of repeat giving as described in the two-stage donor search model. An initial reading of the results would lead one to reject both H1 and H2, as these regressors were shown to be either not statistically significant or statistically significant but not practically meaningful. However, these results must be contextualized within the framework of the population of study.

The two-stage donor search model begins from the premise that there is a significant subset of the population that wanted to give charitably, and it will do so in greater volume and amounts if information is provided to decrease monitoring costs. Although this analysis lacks the power to make that conclusion definitively, nonprofit practitioners looking to attract donations in contexts where donors may search for organizations that match their interests should not draw from this study that monitoring costs and the organization's ability to influence perceptions of these costs will not matter in all online giving portal markets. These results may be unique to portal websites that operate with a model similar to GlobalGiving. The criteria for projects to be posted on the GlobalGiving website and their money-back guarantee could provide donors with a sense of trust that may have already lowered the monitoring cost to a level where donors feel comfortable giving to any project posted on the site. Consequently, additional reductions in monitoring costs through the posting of project updates do not seem to affect donor behavior in a significant way. A practical implication is that nonprofit organizations may find online giving portals that promote some form of donor assurance to be advantageous in providing a measure of legitimacy to organizations, particularly for those with a less established reputation. It also suggests that nonprofit practitioners looking to fundraise through online giving portals need to be aware of the differences in markets and to appropriately develop strategies for engaging with donors in each market.18

Validity issues and future research

As with most observational data, problems arise when one attempts to establish causal relationships from the provided data set. In contrast to data collected through experiments, observational data require that the researcher attempts to control for the effect of outside influences on the relationship of interest at the time of analysis, as opposed to that at the time of collection (Rosenbaum, 2005). The GlobalGiving data set is not without exception, and some of the issues with the data, particularly as it relates to both internal and external validity, are explored.

The results suffer from external validity threats in that the population under consideration is a specific subset of the general population that has self-selected into using the GlobalGiving website. While self-selection limits the conclusions that can be made via the data set, an issue that may threaten the internal validity of the analysis is the exclusion of a variable to capture project ranking. The ranking, which GlobalGiving determines and uses to drive project exposure, was not available in the data set. The exposure provided by a project's ranking could subsequently influence the number of donors that a project will receive. As a consequence, without having the project's ranking readily accessible, it is difficult to distinguish exposure effects on the gift amount and the likelihood of return from the effects of monitoring costs and other project-related variables.19

Another vulnerability of the analysis is the absence of information about how a donor came to discover a particular project on the online portal. The model does not account for who introduces the project to the donor or the social context in which the donation takes place. If the donor discovers the project on his own, through a pure search process, then the assumptions of the presented model still hold. However, if another individual refers the donor to the project, then the model should be extended to account for the social context of that solicitation, or referral, on the giving decision. It may be the case that traditional solicitation-based fund-raising techniques are still dominant even when donations are made through online portals and are further enhanced through the use of portals. Portals make the donation process simpler and more diffuse than it has been traditionally. As a consequence, the opportunity for a given individual to solicit donations from his network on an organization's behalf is increased. The extent to which online portals induce this type of social solicitation and the effect that the type of solicitation has on altering other donor considerations merit further research.

The donor search model we developed, although it was not supported in this empirical test, has the potential to advance the theoretical understanding on how the cost and efficacy drivers of donation behavior (Bekkers and Wiepking, 2011) combine through monitoring costs to impact giving behavior in search-based giving transactions such as online giving portals. This is a step of adjusting from what we have learned from solicitation-based contexts of fund-raising toward understanding giving behavior in search and other donor-driven giving contexts. The substantial conclusion from our analysis is that search costs did not influence giving in the GlobalGiving portal. One interpretation of this finding is that donors who give through GlobalGiving have a level of confidence in the projects and/or website's quality assurance procedures that allows them not to be responsive to project updates. Future tests of the donor search model would be informative in cases wherein donors have lower or more variable levels of confidence. This would be the situation in the absence of quality assurance procedures (i.e., projects are not checked before they can be advertised).


The authors would like to thank Marco Castillo, PhD for his invaluable suggestions in helping to shape the analysis and to acknowledge the financial support of the Georgia Tech Tennenbaum Institute for helping to make the work possible.


  • 1

    Online giving portals are search-based websites that host a large number of nonprofit organization profiles and process donations, often for a fee. Examples of portal websites include GlobalGiving (, (, and Causes (

  • 2

    Founded on the February of 2002, GlobalGiving ( is an online portal that brings donors together with grassroots charity projects from around the world. Since 2002, GlobalGiving has been a conduit for over $50 million in donations from over 213 000 unique donors. GlobalGiving takes a 15% fee from each donation for its work and passes along the remaining contributions to the over 4400 projects that it has helped to provide with funding since its founding.

  • 3

    This model is largely based on Jovanovic's (1979) two-stage extension of the one-stage foundational search model.

  • 4

    Although H1 is stated from the donor perspective, one could alternatively consider a hypothesis from the organization perspective. In this sense, we assume that the organization will try to maximize the level of donations by strategically choosing when to submit project updates. Thus, controlling for all other attributes of the organization, organizations with lower contributions over a given period of time will be more likely to submit a project update, effectively lowering its perceived associated monitoring costs.

  • 5

    Project topics include the following: Animals, Children, Climate Change (GG Green), Democracy and Governance, Disaster Recovery, Economic Development, Education, Environment, HIV-AIDS, Health, Human Rights, Malaria, Microfinance, Peace and Security, Sport, Technology, Women, and Girls

  • 6

    Donors are automatically subscribed to the mailing list for project updates after they donate to a project. Once a donor receives the initial project update after their donation, they can opt out of future mailings.

  • 7

    The due diligence process is characterized, in addition to several other factors, by projects being required to meet four criteria. First, their work has to have the potential for significant social impact. Second, the organization must have a track record for delivering on promises. Third, the organization cannot be listed in any terrorist databases. Fourth, the organization must be eligible for international philanthropic donations so donors in the US receive full tax benefits.

  • 8

    The month of January was excluded because of a special promotion that was running on the site at that time, which highly skewed the amount and type of donations made from normal activity.

  • 9

    We restricted analysis to " small" gifts in the range of $10–$100, as those gifts accounted for over 90% of the donations in the initial data set.

  • 10

    This can occur when there are campaigns or natural occurring events, external to the GlobalGiving portal, which focus attention on the charitable needs of a particular cause or geographic area.

  • 11

    On 2 May 2008, Cyclone Nargis hits Myanmar causing over 146 000 fatalities and thousands of injuries. Shortly thereafter, on May 12, an earthquake hit Sichuan province of China resulting in over 68 000 fatalities. The occurrence of both of these significant disasters within days of each other caused May to be the highest giving month of 2008 on the GlobalGiving website. The average number of donations per day over the entire data set is 59.07 donations per day. May averaged 169.9 donations per day.

  • 12

    Here, we present the main fixed effects specification. Table 2 includes several alternative specifications, which were used to test the robustness of the model.

  • 13

    Between-period regressors are defined over a 100-day period, post-donation. In this sense, the regressors include the number of new project donors, the number of new project updates, and the increase in project funding, all relative to the donor's most recent donation to the project. We also include a control for the number of donations made by a particular donor.

  • 14

    The dependent variable is defined for the 100-day period subsequent to the initial donation. In other words, a donor is only credited for giving a subsequent donation to the project if he makes the donation within 100 days. This limitation on the period in which a donor could be considered to have provided a subsequent donation to the project was made to mitigate issues related to truncated data. Consequently, donations made within the last 100 days of the provided data set were excluded, leaving the multinomial logit analysis to be conducted with a set of 7292 observations, a subset of the 11 426 observations used for the fixed effects analysis.

  • 15

    Correlation between the amount donated to a project and the number of updates at the time of donation was 0.1174.

  • 16

    A project can be removed from the site if it fails to post a project update within the requisite 3–4 months from the most recent update.

  • 17

    Alternatively, a positive sign on the regressor may have led us to the conclusion that donors view GlobalGiving projects as threshold goods. Public goods theory introduces the notion of threshold public goods, which are goods that can only be provided once a certain funding threshold is met. Because the projects on GlobalGiving have an associated funding goal donors may view these projects as threshold goods that will not be provided unless the funding goal is met. Presumably, both types of donors, those who view projects as threshold goods and those who view projects as continuous goods, are found in the data set. Our analysis implies that the majority of donors are impact philanthropists who view projects as continuous goods.

  • 18

    The initial hypotheses and analyses did not consider that organizations may strategically use project updates to increase donations when they are lagging behind. In this way, a reverse causality may exist whereby a higher frequency of project updates may be the result of lower amounts donated to a project. If true, this organization–donor dynamic would work to counteract any significance we might expect on the project update regressors in our initial analyses. To test this, we constructed a probit regression model where the number of donations and the amount donated in a defined period after a project update was issued were assumed to be predictors of subsequent project updates. If true, a decrease in either predictor should cause an increase in the probability of an additional project update during some period after the initial update. For an observational set of 287 project updates, probit regressions were run using the amounts donated in periods of 10 and 20 days after an initial project update was posted to the site as the predictors of subsequent updates within 30 days of the initial update. In no case did the amount donated or the number of donations regressor attain statistical significance.

  • 19

    The absence of the ranking variable is more pronounced in its effect on the level of the gift than on the donor's likelihood of giving to a project again. Analysis of a donor's likelihood of giving again is largely not affected, as the donor already knows the project when the analysis begins.


  • Clarence Wardell III received his PhD from the School of Industrial and Systems Engineering at the Georgia Institute of Technology in 2009. His current research focuses on understanding the effects of social information, transaction costs, and transparency on donor and nonprofit behavior, along with work related to understanding the role of social media in disaster response and preparedness.

  • Shena Ashley is an assistant professor of Public Administration and International Affairs at the Maxwell School of Public Affairs at Syracuse University. Her research is currently focused on broadening the understanding of the mechanisms through which donor (individuals, foundations, and government agencies) funding decisions and selection criteria both impact and are influenced by nonprofit market structure.