Searching for recipes to compete in the charitable contributions market: A configurational approach

Correspondence Gina Rossi, Department of Economics and Statistics, University of Udine, Via Tomadini 30/A, 33100, Udine, Italy. Email: gina.rossi@uniud.it Abstract Nonprofits that compete for charitable contributions often question which are the most effective factors that lead to high levels of donations. To date, the research has been dominated by linear models mainly based on the economic model of giving, and has reported mixed and sometimes conflictual findings about the net effect of certain individual organization-specific factors on donations. In this study, we introduce a configurational approach to explore how the factors considered by the economic model of giving may be combined with each other in multiple configurations with the goal of obtaining high levels of donations. Applying fuzzy set qualitative comparative analysis, we focus on a sample of British community foundations and identify four combinations that lead these organizations to collect large amounts of charitable contributions. The results show that, whereas young foundations should rely on high levels of program spending and large amounts of online disclosure combined with, alternatively, efficiency or aggressive fundraising, old foundations should contain administrative costs and strengthen fundraising efforts while, alternatively, spending the most part of their resources on programs or disseminating large amounts of information through their public websites. Received: 13 November 2020 Revised: 27 September 2021 Accepted: 17 November 2021

nomic model of giving, and has reported mixed and sometimes conflictual findings about the net effect of certain individual organization-specific factors on donations. In this study, we introduce a configurational approach to explore how the factors considered by the economic model of giving may be combined with each other in multiple configurations with the goal of obtaining high levels of donations. Applying fuzzy set qualitative comparative analysis, we focus on a sample of British community foundations and identify four combinations that lead these organizations to collect large amounts of charitable contributions. The results show that, whereas young foundations should rely on high levels of program spending and large amounts of online disclosure combined with, alternatively, efficiency or aggressive fundraising, old foundations should contain administrative costs and strengthen fundraising efforts while, alternatively, spending the most part of their resources on programs or disseminating large amounts of information through their public websites.

| INTRODUCTION
In an increasingly competitive environment, many nonprofit organizations face challenging fundraising issues to raise the money needed to accomplish their goals (Garcia-Rodriguez & Romero-Merino, 2020). Considerable attention has been paid to the organizational factors that can increase the amount of grants awarded to nonprofits. To date, one of the best consolidated models used to explain what affects the ability of an organization to attract donations is the economic model of giving (Weisbrod & Dominguez, 1986), according to which donations are a linear function of the price of giving, the age of an organization, and its fundraising efforts. Several studies have applied this model across different industries and geographical contexts within the nonprofit arena, proposing extensions to the original model to include the role of both the efficiency of the use of the money donated and online disclosure of financial and performance information. However, the findings of these studies are often conflicting and inconclusive, suggesting that the phenomenon under scrutiny is intricate and that one best solution might not exist. Therefore, moving away from a linear approach and considering a configurational one, this study aims to identify how the organization-specific factors considered by the economic model of giving can be combined with each other in multiple configurations to generate high levels of donations (Misangyi et al., 2017;Urry, 2005;Wu et al., 2014). In this way, managers are offered alternatives when taking spending allocation decisions across programs, administration, fundraising, and disclosure to increase the collection of donors' contributions.
We use fuzzy set qualitative comparative analysis (fsQCA), a complementary methodological technique that-going beyond the net effects analysis-facilitates the examination of the multiple configurations of interconnected factors that lead to the outcome (Fiss, 2011;Ragin, 2008). The analysis was conducted on a sample of British community foundations (CFs) that represent a particularly meaningful context for understanding the antecedents of the ability of an organization to attract donations.
The findings capture four configurations of organization-specific factors leading British CFs to reach high levels of donations and contribute to the debate on the competition for charitable contributions by highlighting that donations do not depend on individual factors but, rather, on specific combinations of them.

| THE ECONOMIC MODEL OF GIVING
In the nonprofit arena, the examination of charitable contributions in both the accounting and economics literature has been dominated by the standard donations demand model derived from Weisbrod's widely accepted theory of nonprofit organizations (Harris & Neely, 2016Kingma, 1997;, known as the economic model of giving (Weisbrod & Dominguez, 1986). Weisbrod (1975) postulated that nonprofits are private organizations that satisfy a demand for public goods left unfilled by governments, which are usually focused on the needs and desires of the median voter. Since nonprofits lack the power to tax and the public good nature of their output limits direct sales, they depend financially on the altruism of citizens who want to support the provision of public goods. Therefore, nonprofits must look to voluntary contributions from donors who demand nonprofits' outputs and are willing to pay in return for them.
The economic model of giving proposed by Weisbrod and Dominguez (1986) is a linear model that estimates the demand function for a nonprofit organization's output and some organization-specific factors influencing it. As in the case of purely private goods, donors' demand function for nonprofits' goods is supposed to depend on the effect of conventional market variables such as the price of the goods, their quality, and advertising or information about the value of those goods. In this model, donations serve as the proxy for demand, whereas the price of goods, quality, and advertising are proxied by the price of donations, the age of the organization, and its fundraising expenses, respectively.
The price of donations is a measure that approximates the cost to the donor to buy one dollar of the nonprofit's output and depends on the efficiency of the organization in turning charitable contributions into final output (Jacobs & Marudas, 2009;Weisbrod & Dominguez, 1986). According to the assumptions of the economic model of giving, donors are interested in contributing a dollar of output rather than a dollar's worth of money. Thus, the larger the amount of resources that a nonprofit devotes to administration and fundraising costs, the greater the price of donations. Since the donor is supposed to be price-sensitive, he or she is expected to dislike giving to nonprofits with a high price as it means the donation is less able to purchase charitable outputs (Wong & Ortmann, 2016).
Donors' willingness to pay for a public good output is also influenced by the quality of the output. However, quality is difficult to observe because of information asymmetries that make donors uncertain about it. Therefore, nonprofit organizations need to build up their reputation and a stock of trust among donors, which affect their perception of output quality (Weisbrod & Dominguez, 1986). In the economic model of giving, organizational reputation is proxied by the age of a nonprofit; that is, the number of years the organization has existed. Older nonprofits, thus, are expected to attract more donations because they have achieved greater recognition with donors (Tinkelman & Mankaney, 2007).
Finally, in the market of private goods, demand is affected by the transfer of information on outputs through advertising, since consumers cannot buy products unless they know what they are and where to get them. In the economic model of giving, fundraising serves as a proxy for information dissemination and advertising, as fundraising activities are aimed to promote the organization and its outputs. Therefore, all the other conditions being equal, fundraising expenses should increase charitable contributions (Frumkin & Kim, 2001;Haski-Leventhal & Foot, 2016;Trussel & Parsons, 2008). Nevertheless, scholars have warned that fundraising costs could also produce a lagged negative effect on donations because donors may consider them as nonimpact creating expenses that divert resources from programs and reduce the perceived efficiency of nonprofits' operations, thus reducing the propensity of donors to donate (Burkart et al., 2018;Hager & Flack, 2004;Weisbrod & Dominguez, 1986).
Over the past few years, several dozen studies have tested the Weisbrod and Dominguez price-age-fundraising model (Jacobs & Marudas, 2009). Variants and extensions of this parsimonious linear model have also been proposed by researchers, always confirming its robustness. Administrative expenses were included among the organization-specific factors that can affect donations, since charitable contributions are assumed to be sensitive to the amount of resources allocated to administration, management, and general expenses (Jacobs & Marudas, 2009). Several studies have supposed that administrative expenses are a measure of inefficiency because they divert resources from mission programs and thus they are expected to negatively affect donations because donors prefer their money to be spent on projects that create impact for the society (Ashley & Faulk, 2010;Burkart et al., 2018;Gneezy et al., 2014;Jacobs & Marudas, 2009;Lecy & Searing, 2015;Tinkelman & Mankaney, 2007;Weisbrod & Dominguez, 1986). Conversely, a few studies have assumed that administrative expenses allow nonprofits to conduct their work and build capacity to deliver on promises (Bowman, 2006;Gneezy et al., 2014). Thus, shortages in these expenses can cause underinvestment in key assets, technology systems, and staff training. This threatens organizations' long-term productive capacity, and ultimately undermines their efforts to serve charitable causes effectively (Coupet & Berrett, 2019;Garven et al., 2016;Hager & Flack, 2004;Lecy & Searing, 2015).
Moving from signal theory (Connelly et al., 2011;Spence, 2002), other studies have further extended the economic model of giving-which relies on fundraising as the sole informational channel-by including online disclosure among the organization-specific factors that are thought to affect donations (Blouin et al., 2018;Gandía, 2011;Harris & Neely, 2018;Haski-Leventhal & Foot, 2016;Rossi et al., 2020;. They measured online disclosure as the level of information that nonprofits disseminate through their public websites and predicted that the use of this medium could recompense an organization in terms of future increased charitable contributions. Despite its central informational role in the Weisbrod and Dominguez's (1986) model, fundraising is an information channel with restricted capabilities because it can reach a limited number of donors at limited times of the year . Additionally, high levels of fundraising expenses decrease the amount of resources available for spending on programs, and thus could have a negative impact on donations because of the higher price of giving. Online disclosure has the function of reducing the information asymmetries between a nonprofit and its actual and potential donors at a limited cost. Hence, nonprofits competing for resources are assumed to have an incentive to signal their virtue and their superiority over other nonprofits by providing web-based information about their finances and performance (Blouin et al., 2018;Gandía, 2011;Gugerty, 2009;Harris & Neely, 2018;Haski-Leventhal & Foot, 2016;Prakash & Gugerty, 2010;Rossi et al., 2020;. Research that has tested the donations demand model according to the linear approach proposed by Weisbrod and Dominguez (1986) has provided mixed findings regarding the effect of each organization-specific factor on donations. As Table 1 shows, for the same variable, some studies have highlighted a positive effect whereas others found a negative impact or no statistically significant correlation with charitable contributions.
In light of the results of prior empirical research, one optimal combination of independent variables (a single path) capable of influencing the level of donations does not seem to exist. A possible explanation is that these researchers have adopted a net effect approach in their investigations, which isolates each individual variable from the others and estimates its unique (nonoverlapping and analytically separate) contribution to donations, holding everything else constant (Fiss et al., 2013;Ragin, 2006Ragin, , 2008. Our study breaks with this "net effects thinking" (Ragin, 2008) and employs a configurational approach, which enables a richer understanding of how the factors considered by the economic model of giving can influence the level of donations to a nonprofit organization. The complexity involved in charitable giving, indeed, is still high despite the attempts to overcome information asymmetries through the dissemination of online information and urges the adoption of a complementary approach that shows how several intertwined factors combine with each other in multiple configurations to generate high amounts of donations (Fiss, 2011;Ragin, 2008).

| A CONFIGURATIONAL APPROACH TO THE STUDY OF DONATIONS
The idea that some organization-specific factors have an independent net effect on donations is challenged by the causal complexity of the charitable giving phenomenon, as donations derive from multiple factors that combine in complex, and at times contradictory, ways (Furnari et al., 2020). The limitations of the net effect (linear) approach that dominated the previous studies can be counterbalanced with a configurational approach that treats variables as combining, rather than competing, to create an outcome (Fiss, 2007;Furnari et al., 2020;Ragin, 2008). Configurational approaches assume that causally complex phenomena are characterized by three features: conjunction, equifinality, and asymmetry (Furnari et al., 2020;Misangyi et al., 2017;Ragin, 2008). Organizational outcomes are rarely explained by a single cause; but rather, are the result of a constellation of multiple interconnected factors synergistically combined in a "recipe" (conjunction). As noted by Ragin (2008, p. 109), "to think in terms of recipes is to think holistically and to understand causally relevant conditions as intersections of forces and events." There are multiple ways in which these multiple interconnected factors can combine to produce a given outcome, as different equifinal configurations exist that lead to the same outcome of interest (e.g., donations). In other words, the same outcome may be generated by amalgamating ingredients according to different recipes (equifinality). As all the ingredients in a given recipe have to be present for the outcome to occur, recipes-that is, how ingredients combine in each case-"are more important than the ingredients" (Ordanini et al., 2014, p. 134). Finally, the factors (ingredients) leading to the presence of a particular outcome do not need to be the inverse of the factors that lead to the absence of this same outcome (asymmetry). That is, when a recipe leads to an outcome (e.g., high amounts of donations), this does not mean that its reverse leads to the reverse outcome (e.g., low amounts of donations).
In this study, we introduce a configurational approach to explore how the organizationspecific factors considered by the economic model of giving can be combined in different recipes, so that they jointly bring about the same outcome in terms of the amounts of donations. Specifically, we are interested in high levels of donations because we are investigating the factors that allow a nonprofit organization to compete successfully in the market of charitable contributions from private donors (Gandía, 2011;Harris & Neely, 2018;Rossi et al., 2020;. Thus, our analysis is only geared toward this positive outcome and not the negative one (low levels of donations).
Prior research has suggested that organization-specific factors could better explain the amount of donations collected by nonprofits when considered jointly, rather than in isolation. For example, some researchers explored the interaction between organizational age and fundraising and found that fundraising is less effective in attracting donations when nonprofits are mature (Gandía, 2011;Okten & Weisbrod, 2000;Saxton & Wang, 2014;Tinkelman, 2004). Other studies suggested that large amounts of online information boost the positive impact of fundraising , and that nonprofits accrue more in contributions when disclosure is combined with large amounts of program expenses (Harris & Neely, 2018). Despite their valuable contribution, these attempts are still limited to analyzing the net effect of pairs of individual variables as if they were one. As such, the current study posits that the capacity of attracting charitable contributions cannot be captured by examining the basic relationships between isolated antecedents. Rather, it requires identifying different combinations of factors that might work together in different ways to influence the capacity of a nonprofit to attract donations. In this way, we could better understand how to obtain charitable contributions even in the absence of certain organization-specific factors. Thus, we formulate the following: Configurational hypothesis 1: Program spending, fundraising, organizational age, administrative spending, and online disclosure can contribute positively or negatively-or not contribute at all-to the capacity of a nonprofit to collect donations, depending on the presence or absence of the other factors in the combination.
A configurational approach highlights that any single antecedent could be necessary, though not sufficient, to predict the outcome of interest. For example, while fundraising is a key driver of the capability to attract donations, it might not be sufficient on its own to achieve this goal. Fundraising activities play an advertising function, but they are limited to discrete time periods and cannot convey all the information valuable to donors. Thus, fundraising could need to be integrated and complemented by online disclosure, which continuously disseminates a broader array of financial and performance information on an organization's website (Gandía, 2011;Harris & Neely, 2018;. Similarly, age-because of its reputational effectcould be necessary but insufficient for obtaining large amounts of charitable contributions. Efficiency in the use of resources and the ability of an organization to signal its virtue to current and potential donors through online disclosure might translate into an improved reputation and contribute to high donations (Gandía, 2011). Thus, we consider the following: Configurational hypothesis 2: Program spending, fundraising, organizational age, administrative spending, and online disclosure can be necessary but insufficient to reach high levels of donations.
Finally, considering the complexity of the donation phenomenon, we posit that there is not a single best configuration of organization-specific factors capable to lead to high amounts of charitable contributions. Rather, different configurations of factors can equally predict the ability of a nonprofit to collect donations. Thus, we explore the following: Configurational hypothesis 3: Different configurations of program spending, fundraising, organizational age, administrative spending, and online disclosure could successfully lead to the collection of high amounts of donations.
In QCA, the examination of multiple conjunctural causations is made possible by the use of the Boolean algebra and the operator and to capture the intersection of sets of variables. Conversely, general linear regression models are based on linear algebra and are unable to investigate the interactions of more than two variables (Vis, 2012). Moreover, unlike conventional correlational approaches that view attributes of cases as separate, independent, and discrete (and thereby examine the relations between attributes across cases), QCA is based on comparative analyses between cases. This is because cases have the integrity of attributes and different cases demonstrate different patterns of combination of attributes (Fiss, 2011). Thus, QCA is intended to determine the different causal models that exist among comparable cases, rather than to identify the single causal model that fits the data best (Rihoux, 2003).

| Sample and data collection
The nonprofit sector has a rich morphology of organizations with specific features and different needs for charitable contributions. In this study, we focus on CFs as they are a particular type of nonprofit organization highly dependent on private donations, that fiercely compete in the market of charitable giving for collecting from individual donors the money needed to accomplish their goals. Although differences characterize CFs around the world, scholars agree in defining them as independent, publicly accountable grant-making bodies that are controlled by community members, derive their funds mainly from the community and, through their grant-making and leadership, seek to enhance the quality of life in a specific geographic locale (Graddy & Morgan, 2006;Harrow et al., 2016;Ostrower, 2007). Specifically, we have focused on the UK, a country where CFs significantly took place between 1985 and 2006 in response to cuts that led local governments to withdraw financial support to nonprofit organizations. Although British CFs tend to be smaller in size, staff and boards when compared to US CFs, they have been a major feature of the UK nonprofit sector since the 1990s and created tailored programs of grant making based on local needs mainly related to health and well-being, education, employability, and community cohesion. By 2019, British CFs held assets of about £800 million, including endowed funds of near £700 million, and awarded grants of approximately £100 million (Yang et al., 2021). Over the last dozen years, they are experiencing an increasing competition for charitable contributions because of the recession of 2008-2009 and the following increase of inequalities in income and wealth that have augmented both the supply and demand for philanthropic funds (Hoolwerf & Schuyt, 2017;Walker, 2020). Therefore, building resources through private donor engagement is a key priority for ensuring their future sustainability.
As suggested by Greckhamer et al. (2018), our sample of cases is constituted by the entire population of British CFs surveyed by the Community Foundations Atlas in 2015 (N = 59). According to the classification of the National Council for Voluntary Organizations (NCVO) based on income, they are all medium and large organizations. After we removed four organizations with missing financial data that we have been not able to retrieve by asking for them via email or telephone, our final sample comprised 55 organizations.
Financial data useful for the analysis (donations, program and administrative spending, and fundraising) were retrieved from the audited financial reports for 2015, and 2016, which were acquired directly from the CFs' websites or by request via mail or telephone. Organizational age was obtained from the CFs Atlas website. Data on disclosure were derived from a content analysis (Krippendorff, 2013), which considered the information disclosed on CFs websites.

| Operationalization of variables
High levels of subsequent-year donations were the outcome variable for this study. Donations were measured as the amount of charitable contributions received by the organization in the year t + 1 and include contributions, gifts, and grants from private donors. Following prior literature (Gandía, 2011;Harris & Neely, 2018;Tinkelman & Mankaney, 2007;Trussel & Parsons, 2008), this variable was measured with a 1-year lag to allow donors time to review and react to the information available to them because donors would not have access to current year financial and performance reports when making a current year contribution decision.
Five causal conditions-all measured in year t-were considered: program spending, administrative spending, fundraising, organizational age, and online disclosure. Program spending was calculated as expenses spent on programs and services out of total functional expenses (Blouin et al., 2018;Buchheit & Parsons, 2006;Harris & Neely, 2016Ryan & Irvine, 2012;Valencia et al., 2015). This is an alternative to the price of donations (Haski-Leventhal & Foot, 2016;Trussel & Parsons, 2008) and informs donors how much of their contributions have been used to deliver the programs and services they aimed to support through their grants. Administrative spending was measured as the ratio between the amount of administrative expenses and total functional expenses (Ashley & Faulk, 2010;Ashley & Van Slyke, 2012;Frumkin & Kim, 2001). Administrative expenses refer to management and general expenses and include, for example, overheads, administrative staff and associated costs, and organizational meetings. Fundraising is reflective of nonprofits' expenditure to attract donations from the public and was calculated as fundraising expenses out of the total functional expenses (Frumkin & Kim, 2001;Haski-Leventhal & Foot, 2016;Ryan & Irvine, 2012;Trussel & Parsons, 2008). Fundraising expenses include, for example, campaign printing, publicity, mailing and staffing, and costs incurred in soliciting donations and grants. Age represents the number of years that an organization has been operational and was measured as the number of years since the reported date of founding (Gandía, 2011;Kirk & Beth Nolan, 2010;Tremblay-Boire & Prakash, 2015). Finally, disclosure was measured as the total number of disclosure items retrieved from each CF's website (Harris & Neely, 2018;. Specifically, we checked for the presence of nine items derived from previous studies on the disclosure practices of foundations (Rossi et al., 2018;Saxton & Guo, 2011). Two main dimensions have been considered: (1) financial information (annual reports, investment policies, funds held for grant-making purposes, and calls for grants) and (2) performance information (mission, history, strategic priorities, programs, and impact reports). We treated each item as a distinct dummy variable that equaled one when the websites provided the information, and 0 otherwise. Accordingly, the maximum value for the disclosure variable was nine.

| Configurational analysis
In viewing cases under study as constellations of multiple interconnected factors, the configurational approach emphasizes that causality is complex and requires attention to contrarian cases. Indeed, in the same dataset, causal conditions might be positively related, not related or even negatively related to the outcome in a significant number of cases. Accordingly, we first conducted a contrarian case analysis to search beyond the main effect of our causal conditions on the outcome and verify the existence of non-symmetric relationships (Woodside, 2014). To achieve this aim, we performed a quintile analysis  and used contingency tables to test if the organization-specific factors considered in this study were asymmetrically related to donations.
The results (see Appendix A) revealed that the main effect on donations is largely derived from program spending (4 negative contrarian cases and 5 positive contrarian case), administrative spending (6 negative contrarian cases and 3 positive contrarian cases), and disclosure (2 negative contrarian cases and 2 positive contrarian cases), indicating that these conditions have a symmetric relationship with donations. Conversely, non-negligible contrarian cases exist for age (10 negative contrarian cases and 9 positive contrarian cases), and fundraising (7 negative contrarian cases and 6 positive contrarian cases). The rather high number of contrarian cases suggests that an individual factor in some cases can contribute to a high level of donations, whereas in others it leads to a low amount of charitable contributions, depending on the presence or absence of the other factors in the combination. The results from the contrarian analysis, thus, support Configurational hypothesis 1.
Supported by the contrarian analysis results, we performed a configurational analysis to gain a richer perspective of the data adopting a fuzzy set QCA (fsQCA) that allows "for degrees of membership, thus differentiating between different levels of belonging anchored by two extreme membership scores at 1 and 0" (Schneider & Wagemann, 2012, p. 37). Following the four-step procedure recommended by Fiss (2011) and using a fsQCA 3.0 software, first we defined the property space and the resulting truth table that highlights all the theoretically possible combinations of factors considered by prior studies (Figure 1).
Second, we transformed our variables into fuzzy sets through calibration at three thresholds: full membership in a set (value 1), full non-membership (value 0), and the crossover point (value 0.5), which is "the point of maximum ambiguity in the assessment of whether a case is more in or out of a set" (Ragin, 2008, p. 30).
For calibrating the outcome and the causal conditions, we used benchmarks retrieved from Community Foundations Atlas-a directory of CFs and community philanthropy organizations around the world-and some of the most utilized watchdog agencies (e.g., Charity Navigator, Charity Watch) that monitor nonprofit incomes and expenditures and help donors make giving decisions (Blouin et al., 2018). For disclosure, the calibration was based on points of cumulative distribution (Misangyi et al., 2017;, because of the lack of coordinated and comparable data on philanthropy in Europe (European Research Network on Philanthropy, 2019). As noted by Greckhamer et al. (2018, p. 488): "When criteria external to the study's sample and theoretical knowledge to guide calibration are lacking, researchers may rely on expert panels or, at times as a last resort, use properties of the study's sample (e.g., its cumulative data distribution or its frequency or density distribution) to determine thresholds that capture differences in kind and in degree among cases." All measurements, calibration methods, and anchors are summarized in Table 2.
Donations were calibrated according to Community Foundation Atlas. We anchored full membership, full non-membership and the crossover point to the 75th, 25th, and 50th percentile, respectively. All the CFs below the threshold of full nonmembership have a membership score of 0, while those above the 75th percentile have a membership score of 1. According to Charity Watch, 75% of program spending was used as benchmark for full membership in the category "high program expenses", while the threshold for full nonmembership was set at 35% and the crossover point at 57%. All CFs below the threshold of 35% have a membership score of 0, while those above 75% have a membership score of 1. According to Charity Navigator data, we assumed that CFs that spend more than 25% of their total budget on administrative spending are organizations that divert resources from programs. Thus, we used 25% of administrative expenses on total expenses to determine full membership in the category "high administrative expenses". The threshold for full non-membership and the crossover point were set at 10% and 17.5%, respectively. According to these thresholds, CFs below the 10% threshold have a membership score of 0, while those above the 25% threshold have a membership score of 1. fundraising was calibrated according to Charity Navigator. Thus, a ratio of 20% was used as threshold for full membership. The threshold for full non-membership was set at 5% and the crossover point at 12.5%. We calibrated age using 50 years as the benchmark to identify mature CFs and determine full membership in the "high age" category, while the threshold for full nonmembership and the crossover point were set at 7 and 19.5 years, respectively (Atan et al., 2012). Hence, we assigned a membership score of 0 to CFs younger than 7 years, and a score of 1 to those older than 50 years. Since no prior studies have suggested calibration thresholds, we used points from the cumulative data distribution to calibrate disclosure. We coded a CF as fully in if the website disclosed eight of nine items listed in the measurement of variables section (80th percentile), and as fully out for four disclosure items (20th percentile). The crossover point was six disclosure items (50th percentile).
After calibrating the data, we used frequency and consistency to reduce the number of combinations of causal conditions that lead to high levels of donations (Ragin, 2008). Moving from a number of 32 theoretically possible configurations, we refined the truth table through specification of a minimum frequency of one observation, meaning that at least one CF had to present a certain configuration in order to be considered in the analysis (Figure 2).
In a second stage, we considered only those CF for which the outcome of high donations was present. Among these, we maintained only configurations showing a consistency measure exceeding a threshold of 0.85 (Ragin, 2008).
Finally, we identified configurations that could be considered relevant because they covered a sufficient number of cases. To do this, we considered solutions with a coverage greater than 0.01. We also performed a robustness check by varying the full membership, full nonmembership and crossover point, and the frequency and consistency cut-offs. The analysis confirmed that the results are not overly sensitive to specific design choice (Fiss, 2011). 1 In line with Ragin (2008, p. 175), we discussed only the intermediate solution, as it allows to "strike a balance between parsimony and complexity."  Table 3 summarizes the alternative configurations identified by the fsQCA for achieving high levels of donations. 2 Following the notation system suggested by Ragin and Fiss (2008), each column represents a configuration of conditions linked to the outcome. Black circles (•) indicate that a condition must be present to reach the outcome (full membership score), while crossed circles () indicate its absence (full non-membership score). Further, a blank cell indicates that a specific condition is not contained in a solution. The overall consistency (0.85) and solution coverage (0.58) indicate that the configurations can explain a substantial portion of the outcome. Specifically, a CF is likely to achieve high donations in four situations that, unlike prior studies adopting linear approaches, reveal the complex amalgam of organization-specific factors that can affect charitable contributions. The identification of multiple solutions suggests equifinality (Fiss, 2011;Misangyi et al., 2017;Woodside, 2015), thus supporting that many configurations of program spending, fundraising, organizational age, administrative spending, and online disclosure could successfully lead to the collection of high amounts of donations (Configurational hypothesis 3). Additionally, as there are no single-condition configurations, individual factors are not sufficient and need to be integrated and complemented by the other factors for reaching high levels of charitable contributions (Configurational hypothesis 2).

| RESULTS AND DISCUSSION
The four solutions in Table 3 can be analyzed as recipes, where age appears as an ingredient in all the configurations and assumes both high and low scores depending on the combination with other organization-specific factors. This evidence contributes to shed light on the apparently confounding results highlighted by prior linear approach studies, (see Table 1) and shows that both mature and young CFs can reach high amounts of donations by an appropriate combination of age-which represents a structural condition that is given-with the other organization-specific factors that, instead, can be influenced by managers' decisions and strategically directed toward the outcome. Looking at the results, two configurations work for mature CFs (Configurations 1 and 2), and two configurations work for young CFs (Configurations 3 and 4).
Solution 1 combines the presence of age, program spending, and fundraising with the absence of administrative expenses. Web disclosure is a "do not care" condition not included in F I G U R E 2 Refined truth table the solution, meaning that this configuration works for both large and small amounts of online information. According to this recipe, mature CFs can reach high amounts of donations when they are efficient in the use of resources; that is, they maintain low costs of administration, management, and general activities and spend the most part of their funds on mission programs. Additionally, this solution requires, contextually, that CFs base their advertising and information efforts on fundraising activities, that is engaging interactions with specific donors-for example those on fundraising lists-during fundraising campaigns. Solution 1 seems to be particularly suitable for mature CFs that turn to selected categories of donors because of the specificity of the cause they serve, or when they have to solicit donors relatively passive in gathering organizational information (Blouin et al., 2018). For example, CF #30 supports local charities with training or advice for raising their profile and improving their ability to support those in need in the community. This organization spends the most part of its funds in grants for charities and voluntary groups and maintains low levels of administrative expenses. Moreover, it pays great attention to fundraising activities, also encouraging individuals or other organizations to raise money for the CF. To this aim, detailed advice on how to organize a fundraising event to support the CF's cause is provided on the institutional website.
Compared with Solution 1, Solution 2 somehow substitutes the presence of online disclosure for the role of program spending. That is, with other conditions remaining the same, high transparency in financial sustainability and mission-related goals and achievements can offset high expenses for programs. Solution 2 combines the presence of age, fundraising, and online disclosure with the absence of administrative expenses. Program spending is a "do not care" condition indicating that both high and low amounts of expenses on programs lead to high levels of donations. While prior linear approach studies stressed the importance of using contributions to deliver mission-related programs and services (Gandía, 2011;Gordon et al., 2009;Harris & Neely, 2016;Jacobs & Marudas, 2009;Trussel & Parsons, 2008;Wong & Ortmann, 2016), this combination of ingredients suggests that mature CFs can obtain high donations regardless of the resources they devote to programs, as long as they engage in major fundraising activities and simultaneously signal efficiency through low administrative expenses ( Lecy & Searing, 2015;Tinkelman & Mankaney, 2007), and transparency by disseminating large amounts of online information (Atan et al., 2012;Blouin et al., 2018;Gandía, 2011;. The presence of high levels of both fundraising and web disclosure suggests that online disclosure plays an additional informational role beyond that of fundraising and conveys information in a manner distinct from fundraising. Web disclosure, indeed, is largely and continuously available to the public, and offers a wider type and quantity of evidence that adds to the limited advertising-like information contained in typical fundraising campaign materials addressed to specific lists of actual and potential donors . This solution is implemented, for example, by CF #41, which according to its strategic plan achieves its strategic goals through a set of diversified fundraising activities (e.g., raffles, marketing, and communications) and relies on transparency through high web disclosure (8 out of 9 items) for building a sense of community ownership and belonging. Solutions 3 and 4 work for young CFs. Specifically, Solution 3 combines the presence of program spending and online disclosure with the absence of age and administrative expenses. This means that young CFs that cannot benefit from reputational effects deriving from their age can receive large contributions when they provide several outcome-and finance-related online information and spend significant resources to deliver services, while containing administrative expenses. Fundraising is not included in the solution, thus suggesting that online disclosure is a key donorrelevant information channel that could be satisfactory for providing donors with effective information about the organization and its operations. In this situation, a young CF could contain fundraising costs without missing out on the positive advertising effect usually associated with them (Blouin et al., 2018;Burkart et al., 2018;Wong & Ortmann, 2016). This is the case of CF #50, which devotes more than 95% of its functional costs to programs. Information about the organization is entrusted to a website that discloses complete and detailed financial-and outcome-related information, while fundraising costs are limited to approximately 3% of total expenses.
Finally, Solution 4 combines the presence of program spending, fundraising, and online disclosure with the absence of age. Administrative expenses are not included in the solution, indicating that donors do not care about the efficiency of the organization when they donate. When a CF has recently entered the charitable market, it should pay attention to devoting resources to programs and to provide large amounts of information through both specific fundraising activities and disclosure on their institutional website. This means that young CFs could invest in key assets, technology systems, and staff training to ensure their long-term capacity to serve charitable causes effectively, without worrying that the increase in administrative expenses could negatively affect donations (Bowman, 2006;Coupet & Berrett, 2019;Garven et al., 2016;Gneezy et al., 2014;Hager & Flack, 2004;Lecy & Searing, 2015). The combination of ingredients described in Solution 4 characterizes CF #29. This young organization devotes nearly three quarters of its resources to grants that tackle issues of disadvantage and exclusion within the local community. To achieve this goal, the CF adopts an aggressive fundraising policy aimed to expand its endowment funds through increased legacies and donations, and discloses on the institutional website a complete set of financial and performance information, paying particular attention to provide a detailed description of grants awarded and programs supported.

| CONCLUSION
Nonprofits that compete for charitable contributions in contemporary settings often question how to collect large amounts of donations for supporting their activities. The extant research has investigated the independent net effects of certain organization-specific factors included in the economic model of giving on donations, producing inconclusive and often conflicting results. Linking the economic model of giving and signal theory, our study recognizes the causal complexity of the donation phenomenon and, breaking with previous literature, introduces a configurational approach to advance the knowledge on the organizational factors that lead to high amounts of charitable contributions. Unlike prior research that sought a linear best-fit solution, this study, focused on British CFs, suggests that there is no "gold standard" to achieve high levels of donations but, rather, four different equifinal recipes that combine decisions on fundraising, administrative expenses, sums allocated in programs and decisions on the extent of online disclosure. Therefore, any single factor may be necessary but insufficient for high donations and its effect cannot be predetermined, but rather, depends on the presence or absence of other complementary ingredients.
Being aware of these aspects could help British CFs' managers, who have different resources available to them, to achieve high donations by appropriately combining several strategically influenceable conditions. Managers continually face spending allocation decisions across programs, administration, fundraising, and disclosure. By identifying four different combinations of factors that lead to high levels of donations, they are offered alternatives when exploring ways to increase the collection of donors' contributions. The four configurations highlighted by this study always include organizational age, meaning that solutions differ for young and old CFs. Whereas young CFs should rely on high levels of program spending and large amounts of online disclosure combined with, alternatively, efficiency or aggressive fundraising, old CFs should focus on containing administrative costs and strengthening fundraising efforts while, alternatively, spending the most part of their resources on programs or disseminating large amounts of information about their outcomes and financials through their public websites. As these solutions lead to the same outcome, managers can select the pathway most appropriate for their organization depending on the age of the CF-a structural condition that they cannot modify-and the main strategic levers they can handle according to the organizational strategy.
Under a methodological perspective, this study shows that fsQCA has the capacity to analyze small-and medium-sized samples. This might be useful for researchers in the European nonprofit context who are often faced with a lack of coordinated and comparable data, which represents a strong limitation in implementing research designs with traditional linear regression models (European Research Network on Philanthropy, 2019). Additionally, unlike conventional statistical testing, fsQCA can address causal complexity and support developing multiple and more holistic solutions for CFs to reach high levels of charitable contributions.
This study is not without limitations. First, it focuses on the case of success in collecting high amounts of donations. Further research could consider applying QCA to identify those recipes that determine low levels of charitable contributions, thus investigating which combinations of factors might lead CFs to fail in attracting donations. Second, this study explored the influence of five key factors identified by the economic model of giving on donations. Future research could consider additional causal conditions-such as outcomes achieved through program spending, organizational governance, and ratings from watchdog agencies-to gain a more holistic, nuanced, and comprehensive explanation of the outcome. Additionally, future studies could measure some factors of the economic model of giving in a different way. For example, disclosure might be considered in terms of depth or criticality of the information provided rather than its magnitude. Third, the current study focused on a sample of British CFs. Further research could test what configurations of organizational factors are more effective in attracting donations in a broader sample to verify the extent to which the results for British CFs could be generalized or identify more solutions across countries and types of nonprofit organizations. Finally, future studies could consider alternatives to the economic model of giving (e.g., neoinstitutionalism, resource dependency theory, contingency theory) to see the phenomenon from a different angle.
Despite these limitations, this study sheds new light on the causal complexity underlying the capacity of nonprofits to attract donations, and suggests ways to combine multiple factors to effectively compete in the market for charitable contributions.

ACKNOWLEDGEMENT
Open Access Funding provided by Universita degli Studi di Udine within the CRUI-CARE Agreement.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.

ORCID
Gina Rossi https://orcid.org/0000-0003-1062-0070 Chiara Leardini https://orcid.org/0000-0003-1015-6095 ENDNOTES 1 We performed a further sensitivity analysis varying the outcome. We operationalized DONATIONS on a relative scale (donations/size). The calibration anchors, lacking external standards, used 75th, 50th, 25th percentiles of the distribution as thresholds. The other conditions did not change. The analysis showed three solutions; these solutions are identical to configurations 1, 3, and 4 in the main model (see Table 3), thus confirming that the results are robust. 2 We also performed an analysis for the absence of the outcome, to explore the asymmetry assumption. The results ensured that the inverse combinations of conditions resulting in high levels of donations are not related to low levels of donations. Chiara Leardini is full professor of business administration in the Department of Business Administration at the University of Verona (Italy). Her research interests include governance and accountability in nonprofit organizations, performance management in public utilities and accounting history.
Stefano Landi is researcher in the Department of Management at the University of Verona (Italy). His primary research focus is on accountability in nonprofit organizations, healthcare management and waste management.
Luca Piubello Orsini is a PhD student at the University of Verona (Italy). His research interests center around accountability in nonprofit organizations and healthcare management.