We chose NPOs in the environmental sector, with operations in Brazil, as units of analyses for this study. By delimiting the work to an area of activity and a geographic field of action, in an operational cutting, we contribute to the homogeneity of the units of analysis in all selected NPO. At the same time, aspects related to the relevance of the social object promoted by environmental NPO strengthen this option.

Considering the significance level of 0.05, statistical power of 0.8, effect size of 0.35, and six predictors, the minimum sample size were 46 observations. As we carried out the moderation evaluation through multigroup analysis (two groups), we considered as minimum size 92 observations, or two groups of 46. Then we developed the sampling plan (Table 1).

Table 1.

Elements	Description
Environmental NPO	2.242
NPO registered at CNEA	654
Minimum sample size	92
Sample size	108
Geographic range	Brazil
Temporal range	2015

Source: Elaborated by the authors.

We built the sample in a simple random way, from the list of NPOs recorded at the National Register of Environmental Entities in Brazil (CNEA). There was a need for new draws, without repetition, because the necessary data for all the NPOs initially drawn were not identified.

We collected data predominantly from Financial Statements, Annual Reports, and Statutes, which were available at the websites of NPOs, or found in the repository “Map of Civil Society Organizations”. In addition, we contacted NPOs (telephone and email) for which we were unable to access the documents directly, using records at the Brazilian National Environment Council (CONAMA).

Data were collected during the first two months of 2017, and they refer to the fiscal year of 2015, in order to avoid bias due to unavailable information. Since the responsible body only discloses all documents of a specific year in the subsequent year, we did not want to risk having incomplete data, if we worked with information regarding 2016 or 2017.

This study aimed to analyze the relationship between the governance structure of NPOs and donations. Therefore, donations with donor restrictions were the endogenous variable.

NPOs oftentimes develop their activities through projects (Diallo & Thuillier, 2004). To do so, NPOs submit proposals to national and international bodies to raise funds for projects that will carry out the activities defined by their institutional mission (Lacruz et al., 2019). The most recent yearbook “Environmental Management Analysis 2013/2014” (Análise Gestão Editorial, 2015) shows that 97% of the resources of environmental Brazilian NPO were donations from partners, and national and international organizations. Therefore, the volume of other resources (revenues from services rendered and sale of goods) is minimal.

As a measure of governance (explanatory factor), we used the presence of a set of governance factors (proxies) identified in the literature, following other authors (for example Bromley & Orchard, 2016; Feng et al., 2016).

In order to evaluate the possible moderation of NPO Certification in the relationship between Governance and Donations we included with a moderator variable the public certifications provided to NPO by Federal Government of Brazil.

Furthermore, we examined the need to include the control variable ‘NPO age’ in the model, assuming that NPO need time to operate and implement governance practices (Saxton et al., 2014).

Regrettably, we could not use NPO’s size as a co-variable, under the argument that larger NPO tend to have more resources to implement governance practices (Feng et al., 2016; Haski-Leventhal & Foot, 2016). Table 2 shows the operationalization of the variables.

Table 2.

Variable		Scale	Description
Endogenous	Donations	Continuous	Natural log of the Donations with donor restrictions in the Financial Statements for the period ended at December 31, 2015
Manifest	Board	Nominal (dichotomous)	Existence of Board
	Election-Board		Formal rules for the election of the members of the Board
	Mandate- Board		Members of the Board with Formal Mandate
	CEO-Remuneration		Remuneration of the CEO approved by the Board
	Independence-Board		CEO does not participate, with voting rights, in the meetings of the Board
	Fiscal_Council		Existence of Fiscal Council
	Election-Fiscal_Council		Formal rules for election of the members of the Fiscal Council
	Mandate-Fiscal_Council		Members of the Fiscal Council with Formal Mandate
	Advisory_Committee		Existence of at least one advisory committee
	Corporate_Identity		Formally established institutional mission, vision and values
	Annual_planning		Annual Plan of Activities approved by the Board
	Policies_codes		Policies and/or codes approved by the Board
	Independent_audit		Existence of independent audit
	Annual_Report-General_Assembly		Annual Report approved by the General Assembly
	Financial_ Statements-Internet		Financial Statements on public websites
	Financial_Statements-Fiscal_Council		Financial Statements appreciated by the Fiscal Council
	Annual_Report-Internet		Annual Report on public websites
	Annual_Report-Donnors		Annual Report sent to donors
	Annual_Report-Board		Annual Report assessed by the Board
	Team_Qualification		Qualification of council members and CEO on public websites
	CEO_Evaluation		Formal evaluation of the CEO by the Board
Moderator	NPO_ Certification	Nominal (dichotomous)	Certified as OS or OSCIP or CEBAS
Control	Age	Continuous	Years since NPO was founded, until December 31, 2015

Source: Elaborated by the authors.

Table 3 shows the descriptive statistics of the variables concerning donations and the age of NPO.

Table 3.

Statistics	Donations with donor restrictions (In thousands of dollars)a	Ageb
Mean	1.317	16,7
Standard deviation	2.585	7,7
Minimum	26	5
Maximum	16.667	29
1st quartile	110	10
2nd quartile	308	16
3rd quartile	1.282	24

Source: Elaborated by the authors.

There is a relevant discrepancy in the donations with restrictions and the age of NPO in the sample, which we expected, given the heterogeneous NPO’s profile in Brazil (Instituto de Pesquisas Econômicas Aplicadas, 2018). We also present the frequency statistics of the manifest variables in Table 4.

Table 4.

Manifest variables	Frequency
Fiscal_Council	92%
Financial_Statements-Fiscal_Council	92%
Election-Fiscal_Council	90%
Mandate-Fiscal_Council	89%
Annual_Report-Donnors	86%
Board	82%
Annual_planning	82%
Election-Board	81%
Mandate- Board	81%
Annual_Report-Board	80%
Annual_Report-General_Assembly	77%
Corporate_Identity	72%
Independence-Board	64%
CEO-Remuneration	59%
CEO_Evaluation	53%
Policies_codes	50%
Advisory_Committee	44%
Team_Qualification	43%
Annual_Report-Internet	42%
Independent_audit	38%
Financial_ Statements-Internet	31%

Source: Elaborated by the authors.

Since we wanted to investigate relationships between latent variables, we chose the technique Partial Least Squares Structural Equation Modeling (PLS-SEM). As the variables related to manifest variables are qualitative, assuming dichotomous values, and this technique is not appropriate for measuring constructs under this condition (Hair et al., 2016), we initially conducted the Multiple Correspondence Analysis (MCA) to identify the underlying governance structure in NPO.

We used the dimensions detected in MCA as constructs to specify the structural model. Just like Donations, a single-item construct, we used as observations’ value the standardized score of the sum of the observations of each dimension’s variables.

Regarding the potential moderator effect (H₂), as data assumed dichotomous values (zero and one), we evaluated it through a multigroup analysis, as suggested by Hair et al. (2016) and Sanchez (2013).

In the data processing, we used the software packages R (R Core Team, 2017) and psych, for correlation; FactoMineR and nFactors, for MCA; and plspm for PLS-SEM.

In order to identify the underlying governance structure in NPO, we conducted a MCA. Using the scree plot criteria and parallel analysis, we retained five dimensions, which together accounted for 74% of the variables’ variance.

Next, by evaluating the discrimination measures of the variables, we decided to add a sixth dimension, since we did not get discrimination measures for three dimensions above the dimensions’ inertia. The six dimensions explained 78% of the variance, and allowed, at least, one value of the discrimination measure to be above the inertia of its corresponding dimension. This improved the adjustment of the model.

Due to the theoretical domain defined for each dimension, we decided to group some variables with the dimensions for which they showed higher discrimination measures. Two referred to dimensions for which their discrimination measures were not the highest. However, they were superior to the inertia of the dimension to which they were associated. Thus, we arranged the underlying governance structure according to the following dimensions, as Table 5 shows.

Table 5.

Board (Inertia = 0.30)	Fiscal Council (Inertia = 0.18)	Transparency (Inertia = 0.12)
Board Election-Board Mandate-Board CEO-Remuneration Annual_Report-Board Evaluation-CEO Independence-Board Advisory_Committees	Fiscal_Council Election-Fiscal_Council Mandate-Fiscal_Council Financial_Statements- Fiscal_Council	Financial_Statements- Internet Annual_Report- Internet Team-Qualification
Management (Inertia = 0.08)	Accountability (Inertia = 0.06)	Audit (Inertia = 0.05)
Annual_Planning Corporate_Identity Policies_Codes	Annual_Report-General_Assembly Annual_Report-Donors	Independent_Audit

Source: Elaborated by the authors.

We understand, therefore, that Governance comprises a set of dimensions, composed of mechanisms of incentive and control, in order to mitigate the agency problem arising from the contractual relationship between principal and agent, thus minimizing agency costs (Jensen & Meckling, 1976).

Hence, from this theoretical domain, we assumed that governance is an abstraction of a higher order that operates through a set of dimensions. Harris et al. (2015), among others, characterized the theoretical attributes that are potential determinants of governance as latent variables, assuming the reflective character for the constructs. Thus, in the context of this research, we considered the observable variables (proxies) as indicators that reflect the underlying theoretical constructs.

Operationally, governance is a second-order construct reflected by the dimensions that emerged from MCA: Board, Management, Fiscal Council, Transparency, Accountability and Independent Audit. In turn, these dimensions are composed of governance mechanisms that arose from the theoretical background.

The Board represents the collective decision-making body responsible for keeping the strategic direction of the NPO. In other terms, it is responsible for the definition of guidelines that support NPO's practices and business, and whose main function is to establish a link between cause and management - according to Fama and Jensen (1983), Jensen (1993) and Jensen and Meckling (1995). Operationally, the construct Board involves elements that relate to its constitutive aspects (election, mandate, independence, and committees) and attributions (the approval of CEO’s compensation and the Annual Report).

A reasonable explanation for the association of the Board with governance is the perception of donors that it is an important body for the separation between control and management (Fama & Jensen, 1983), which oversees management actions, and reduces the risk of the agent operating against their own interests (principal).

However, the Board, in this study, does not include deliverables arising from its actions of agents’ monitoring, or agents’ actions whose behavior is not harmful to the principal (Jensen & Meckling, 1976; Harris et al., 2015). In this study, these aspects refer to the Management dimension, which involves the elements that guide the activities of the executive team, the Board, the Fiscal Council, and the General Assembly or the Board of Trustees (Corporate Identity, Annual Activity Plan, Codes and Policies).

The presence of formal policies, in a broad sense, can encourage employees to confidentially report unethical behavior, avoid conflicts of interest (for example, code of conduct, anti-corruption policy), and serve as a guide for management practices. As these policies can improve NPO’s governance, by reducing information asymmetry, they will increase donors’ confidence that their resources contribute to advance the NPO’s institutional mission.

The Fiscal Council is the supervisory body of NPO’s accounting and financial management, involving its constituent aspects (election and mandate) and attributions (assessment of the financial statements). Thus, it is different from the Board, whose theoretical domain refers to keeping NPO’s purposes, while the Fiscal Council deals with the supervision of administration acts, providing opinions on the organization's financial statements (Lamb, 2002). The presence of a Fiscal Council, duly constituted, can be an additional line of defense for the principal's interest.

Transparency, in this study, is similar to OECD disclosure principle (2004), and means the public dissemination of relevant information. In our study, it consisted of disclosing information through the Internet (websites, fanpages etc.) regarding financial statements (audited or non-audited), the Annual Report, and the professional qualification of NPO’s team (executive and board members).

Accountability, in turn, also similar to the principle of OECD (2004), involves an administrative body that reports to a higher authority; in the present case, it is the assessment, by the General Assembly or Board of Trustees, of the Annual Report presented by NPO’s chief executive; and also accountability to donors, by sending them the Annual Report.

Higher authorities and donors exercise control through the regular reception of follow-up information by NPO executives, who make them aware of the appropriate use of resources.

Finally, in relation to first-order constructs that reflect Governance, there is Audit, a single item construct: the auditing of the financial statements by an independent audit company. This independent company issues an opinion on NPO’s financial statements, whether they do not show relevant distortions and meet current standards (Verbruggen et al., 2015).

Thus, the audit report is a measure of an NPO’s reputation (Watts & Zimmerman, 1978), in addition to providing an external overview, thereby reducing agency costs (Harris et al., 2015; Watts & Zimmerman, 1978). This makes donors more confident about the reliability of an NPO's financial-accounting information, and safe about the protection of the donated resources.

We next examined the need to include the variable ‘NPO age’ in the model, assuming that NPOs need time to operate and implement governance practices (for example Haski-Leventhal & Foot, 2016). Since we did not identify any statistically significant correlation between age and governance, measured by the sum of the dimensions’ scores identified in MCA (r = 0.093; p-value = 0.339), we did not include this variable in the model.

Figure 1 shows the initial research model.

Figure 1. Source: Elaborated by the authors

We operationalized the variable Donations, the main beneficiary of the research results, through the amount of donations with donor restrictions, reported in the Financial Statements.

In addition, the single item construct ‘NPO Certification’ had as proxy the public certifications provided to NPO in Brazil as OS (Social Organizations), OSCIP, or a certificate of CEBAS (Charitable Organization for Social Assistance). We expected that such certifications, some of them related to governance mechanisms (statutory provision of independent audit, or requirement of superior deliberation body), as they increase the list of obligations, would positively moderate the relationship between Governance and Donations. We assumed that public certifications provided to NPO would be similar to quality standards (for example, ISO). Companies traditionally seek certification to gain competitive advantage (Rao, 1994), and many NPO are adopting such practices as means to improve their competitive position in the donations market (Slatten et al., 2011).

In this research, the governance construct only comprised the internal mechanisms. Regarding the external mechanisms, the Market of Products and Factors and the Political, Legal and Regulatory System, we used them to delimit the object of empirical investigation (similar to a control variable); on the other hand, the main effects of this research fall on the Donations Market, that is, the impact of Governance on the Donations with donor restrictions. We also used a specificity related to the political, legal and regulatory system (NPO Certification) to check a potential moderating effect on the relationship between Governance and Donations. Thus, we explored the set of governance dimensions.

We conducted the validation of the structural model using the plspm package by checking the statistical significance (α = 0.05) and relevance of path coefficients, through the bootstrapping procedure; and the assessment of the coefficient of determination (R²) as a measure of the model’s accuracy.

Figure 2 shows the values of R², the path coefficients, and their statistical significance.

Figure 2. Note. * Significant at the 0.01 level. Sample size = 108. Bootstrapping = 1,000 subsamples. Source: Elaborated by the authors.

In step with Sarstedt et al. (2020) and Schamberger et al. (2020), a further analysis was conducted so that this paper can confirm that the data does not contain outliers with the potential to distort the results of structural equation models (Cook’s distance), the relationships of model are linear (Linktest) and the model has no omitted constructs bias are omitted (RESET for omitted variables). At first, the model was estimated in order to implement the tests, using the resulting construct scores as input for it. Outliers were not identified (D < 1), evidencing the linear effect’s robustness (p-value = 0.09) and that omitted variables bias does not occur (p-value = 0.31).

Second, according to Sarstedt et al. (2019), we assessed the measurement model of the higher-order construct, represented by the relationships between the higher-order component and its lower-order components: unidimensionality (Cronbach’s alpha between 0.7 and 0.9, Dillon-Goldstein’s rho between 0.7 and 0.9, and the dominance of the first eigenvalue), convergent validity (outer-loadings higher than 0.7, and the average variance extracted higher than 0.5) and discriminant validity (i.e. cross-loadings and Fornell-Larcker criterion).

Figure 2 shows that the direct relationships of the structural model were significant and caused a positive impact, since path coefficients assumed positive values. Within the second-order hierarchical model, it was possible to observe that The Fiscal Council mentions Governance in a less pronounced way, with a path coefficient of 0.425. Governance appears in the other dimensions in a similar way (between 0.631 and 0.727). This may occur because the Fiscal Council, within the legal environment of the sample NPO (a Brazilian NPO), is not a mandatory body for NPOs or for the companies.

We observe, in Figure 2, that Governance affects Donations with a statistically significant path coefficient of 0.713.

This result strengthens the field of agency theory, which underlaid and inspired this study, because it indicates that Governance positively affected Donations. This confirms Singh and Ingdal (2007), who discussed best practices for NPO donors in Nepal, and showed that donors require a range of NPO’s governance practices for deciding to donate. Thus, governance practices allow NPO to have an easier access to the donations market, since they ensure more efficient and effective operations (Greiling & Stötzer, 2015). This contributes to the improvement of the organization's reputation in the market of product and factors, and reduces the information asymmetry and consequent agency costs; therefore, it follows that agents apply the resources according to the principal’s interests (mitigating agency problems).

Therefore, the absence of or poor quality governance practices can result in donors’ reluctance to contribute to NPO. According to Fisman and Hubbard (2005), a poor governance leads to a poor management monitoring, thus giving rise to agency costs.

Figure 2 presents the R² of each endogenous latent variable, which is a measure of the model’s accuracy. It is important to mention that R² of the latent variable of the first order reflective construct indicates how much this latent variable is a sign of the latent variable of the second-order construct and, therefore, we analyzed it, since it also generates paths in the structural model.

We found relevant R² values, especially considering the set of variables absent in the model.

A R² of 0.508 indicates that the relationships established account for 51% of the variance of Donations, which provides the model a reasonable adjustment, given its economical character; through the gradation exhibited by Sanchez (2013), R² values lower than 0.2, between 0.2 and 0.5, and higher than 0.5 showed a low, moderate and high explanatory power, respectively.

According to Wetzels et al. (2009), in the reflective-reflective type of models, first order constructs reflect the second-order constructs, which leads to the conclusion that R² values of the latent variables of the first order reflective constructs indicate how much they reflect the latent variables of the second-order constructs. R² of the latent variable of the first order reflective construct indicates how much this latent variable is affected by the latent variable variance of the second-order construct. Using Wetzels et al. (2009) terms, we highlight that in the Governance construct, the sub-dimension that best reflects it is Management, R² = 0.528, while for the Fiscal Council sub-dimension we achieved R² = 0.181.

In order to evaluate the possible moderating role of NPO Certification in the relationship between Governance and Donations, we conducted a multigroup analysis, with 54 observations for each group.

Contrary to intuition, we observed that NPO Certification did not moderate the relationship between Governance and Donations. That is, we did not confirm the hypothesis that NPO Certification, with at least one Brazilian Federal Government certification, would increase Governance relationship with Donations (p-value = 0.1461).

This result disagrees with Feng et al. (2016), who identified that the certification Standards for Excellence® was associated with the increase of donations, compared to a control group of NPO that did not receive the same certification. However, this certification is granted by the Maryland Association of Nonprofit Organizations, and the requirements for getting and keeping it are more rigorous than for the other certifications considered in this study (OS, OSCIP and CEBAS). It does not confirm either the results of Desai and Yetman (2015), who found that the legal and report requirements demanded from NPOs moderated the relationship between the percentage of changes in charitable spending and the percentage of changes in program revenues. These findings support the notion that state regulation, through additional requirements for certification, helps NPOs to play their social role. On the other hand, results by Lee (2016), with data from the National Center for Charitable Statistics, suggest that state regulation does not affect the adoption of good governance policies by NPOs. NPOs that operate in states that demand reporting and registration were no more susceptible to adopting accountability policies than organizations that act in states that have no such requirements.

On the other hand, the result of our research indicates that, for the sample composed of Brazilian organizations, public certifications provided to NPOs (as OS, OSCIP or CEBAS) does not contribute to increase Donations with donor restrictions. Thus, Brazilian donors did not respond favorably to NPO certification, regarding the volume of donations. We suggest that donors did not perceive the benefits (tangible or intangible) of these certifications, for several reasons: (i) a possible lack of credit of the issuer (the Brazilian State), since one expects that certifications will lead to legitimacy and favorable reputation, as the issuer lends its credibility to the recipient - in this case, it would be a ‘discredited accreditation’; (ii) because NPO without certification have voluntarily incorporated non-mandatory mechanisms, due to donor pressures (Lacruz et al., 2019), so that donors cannot distinguish NPO with and without the seal of distinction, only by observing their governance practices; or (iii) because the presence of governance mechanisms is sufficient, which makes the certification unnecessary, from donors’ point of view.

As a general measure of model adjustment, we got the Goodness-of-Fit index (GoF) of 0.4177. Wetzels et al.(2009) suggest for gradation a low GoF = 0.1, average = 0.25 and high = 0.36. By this criterion, the general fit of this study’s model can be considered high.

Through the analysis of confidence intervals, we evaluated the accuracy of the PLS parameter estimates for path coefficients and R², assuming, for all of them, statistical significance at 0.05 level.

The analysis allowed us to infer the positive impact on donations with donor restrictions, as well as the absence of moderation of an NPO’s certification in this relationship (Governance -> Donations).

It is possible to associate the identified relationships to the process of marketization (Salamon, 1997), according to the findings of Lacruz et al. (2019). To deal with the donations market constraints, NPOs have adopted private for-profit market approaches. This lead to what Salamon (1997) named as ‘the non-profit private sector marketization’. Thus, NPO extended their external responsibility, and had to show results in terms of efficiency and effectiveness (Arvidson & Lyon, 2014), besides adopting management models for this purpose (Smith, 2010).

In this context, an important element is the temporality of governance in the third sector. NPO, in general, develop their actions to attain their institutional missions through projects (Diallo & Thuillier, 2004). Thus, the project, understood as a temporary effort to generate a specific delivery (Project Management Institute, 2017), according to Lacruz et al. (2019) refers to the object of the contractual relationship between NPO and the donor organization. Hence, it is a business venture executed within a specific time limit, with a defined cost, scope and quality and cost, in exchange for donation, and is subject to a contract (or related term) between the parties. Therefore, the project is the object of the contract, which, in turn, is the link of the agency relationship between the NPO (agent) and the donor organization (principal) (Lacruz et al., 2019). Specifically, in this research, we delimitated the donations with donor restrictions, according to this understanding.

The temporality of the project leads to the understanding that governance in NPOs is largely due to the influence of the project; each project converges to practices of governance at the organization level (Lacruz et al., 2019).

Temporality leads to the reflection that NPO, because of the process of marketization (Salamon, 1997), can incorporate in their governance necessary elements and those imposed by donors over the duration of the project. However, at the end of a project, these mechanisms are interrupted, and a future project incorporates or modifies them, according to the new donors’ demands (Lacruz et al., 2019). Although the results of the study do not allow it, we consider, from the whole discussion, that it is possible to make such a proposition.

This reflection has a strong association with the assumption of different planning horizon of the agency theory, which implies the transience of the principal in relation to the NPO, whose efforts are limited to the time of the contractual relationship, that is, to the project (Lacruz et al., 2019).

Lacruz et al. (2019) show that the donation market acts as a complement of the legal and regulatory political system, such as an external mechanism of governance for NPO in the delimitation of legal security, given the general legal context and low regulation environment in Brazil.

The analysis of external forces (for example Desai & Yetman, 2015), in particular the donation market, for the configuration of governance in the third sector, still lacks operational and empirical evidence for a stronger support (Lacruz et al., 2019). The underlying governance structure identified in this research is a reflection of internal and external forces (including the donation market) in NPO governance.

Renz (2007) suggest that project governance be the response to the gap between project management (operation) and governance (strategy) in the third sector context. Lacruz et al. (2019) advance this understanding, proposing that projects bring reflections to governance through temporary governance mechanisms; that is, governance would be (re)configured through the projects, in a continuum.