-State Presence in Brazilian Social Assistance Services: Effects on the Creation of Nonprofit Private Providers ^*

Abstract

A growing body of literature addresses Brazil’s National Policy of Social Assistance, but little is known about the factors that affect the creation of nonprofit, private social assistance providers (PSAPs) in the country. This paper analyzes the historical patterns of PSAP creation in Brazil. We argue that the place of PSAPs within the social protection system changed during the 2000s and that this change stems from a reassessment of the state’s role in this area. We also contend that a switch in incentives and increased state provision slowed down the rate of PSAP creation. We conducted a document analysis to create a panorama of the institutional landscape (1930s-2000s); using a unique dataset, we also estimated the association between state presence and PSAP creation (2000-2017). This mixed-method research strategy supports our claim that the direct provision of social services by the state has contributed to the decline in PSAP creation, as the mixed-method approach reveals both a mechanism for institutional change and evidence of its implications.

Social assistance; social protection; nonprofit service providers; state-run facilities; Brazil

A growing body of research has addressed Brazil’s National Policy of Social Assistance (Política Nacional de Assistência Social – PNAS), but little is known about the factors conditioning the creation and operation of nonprofit, private social assistance providers (PSAPs) in the country’s Unified System of Social Assistance (Sistema Único de Assistência Social – SUAS). In fact, only a few recent studies have directly examined this theme ( AMÂNCIO, 2008 ; BRETTAS, 2016 ; GABRICH JUNIOR, CASTRO, and MOREIRA, 2015 ; MESTRINER, 2008 ; STUCHI, PAULA and PAZ, 2012 ). Our goal is to contribute to filling this gap.

In the last fifteen years, as the state’s role regarding the provision of social assistance services changed, the institutional landscape in which PSAPs operates has also been altered. Brazil has historically had a system in which the state plays a residual role, characterized by subsidiary and indirect state provision in which social assistance services are predominately offered through philanthropic action. Currently, however, the Brazilian model is hybrid, with social assistance services being provided both directly (mainly by municipalities) and indirectly (by PSAPs) in the same territory. The state is at the same time the regulator and provider of social assistance, while PSAPs operate as complementary providers.

The state’s responsibility for social assistance, although established by the 1988 Federal Constitution (FC88), was only effectively undertaken in 2004, when the PNAS was passed. The PNAS defines two levels of social protection: basic and special (the latter can be of medium or high complexity). Basic social protection prevents risk conditions; it supports the population living in situations of social vulnerability due to poverty, deprivation, or weakening of affective ties – be it personal ties or those associated with sources of social belonging (e.g., age group, ethnic group, gender identification, or disability discrimination). Special protection of medium complexity assists people facing personal or social risks because their rights are being threatened or violated – people whose family and community ties have not yet been broken. As for the special social protection of high complexity, it guarantees full protection – housing, food, hygiene, and sheltered employment – for people without references or in life-threatening situations who need to be removed from family or community life. Private providers are expected to supply both levels of protection. In practice, the state is more actively involved in basic protection and special protection of medium complexity, whereas high-complexity protection is predominantly a niche of PSAPs.

In the 1990s, there was a boom in PSAP creation, reaching more than five hundred new providers per year. After 2010, the number of new PSAPs dropped substantially, to almost zero in 2016 ( Figure 01 ). Although this trajectory may also be explained by other factors (e.g., resource availability), we highlight the changes in the ideational basis of Brazilian presidents’ agenda in the recent democratic period. Specifically, we are interested in assessing the effects of these different agendas on the legal framework under which the PSAP operate and on the level of service directly provided by the state.

Figure 01
Number of PSAPs created, by year (1939-2017)

Notes: Data retrieved from the ‘year of creation’ field in the National Registry of Legal Entities (Cadastro Nacional de Pessoa Jurídica – CNPJ) through a batch query in January 2018 via ReceitaWS: https://www.receitaws.com.br. Information on the year of creation was obtained for 16,632 (90%) of the 18,562 CNPJs listed in the National Registry of Social Assistance Providers (Cadastro Nacional de Entidades de Assistência Social – CNEAS) as of June 1^st, 2017 (date of data extraction). To get a random sample of 10% of the 1,930 CNPJs for which the batch query did not return a year of creation, we collected that information manually at http://www.redesim.gov.br/consultas-cnpj. In this random sample, 72% of the CNPJs were created up to 2000; such percentage is similar to that found in the set of 16,632 CNPJs listed at ReceitaWS (67%). For our purposes, a PSAP corresponds to a combination of a CNPJ and a municipality; 19,159 PSAPs were registered in the CNEAS. For 17,226 of these (16,632 single CNPJs), the year of creation is available; this is the set of PSAPs shown in Figure 01. Any PSAP extinguished before June 2017 could not be observed. The earliest year of creation is 1939, and the most recent, 2016.

This paper analyzes the historical patterns of PSAP creation in Brazil. We argue that PSAPs’ role within the social protection system has changed and that this change stems from the state’s reassessment of its own role in this area. Furthermore, we contend that a switch in incentives (Hypothesis 01 – H1) and increased state provision (Hypothesis 02 – H2) have decelerated the creation of new PSAPs.

This study is based on a mixed-method approach. We employ two analytical strategies. First, we examine the institutional parameters (norms and incentives) for PSAP operation (from the 1930s to the mid-2010s, with special attention to the last three decades). To that end, we analyze laws and regulations from the perspective of the historical institutionalism ( MAHONEY and THELEN, 2010 ; THELEN and STEINMO, 1992 ), which is done in the next section. We argue that changes in rules reduced the incentives for PSAP creation (H1).

In addition, we contend that increased state presence in service provision starting in the 2000s contributed to plunging PSAP creation (H2). To examine this element of our twofold argument and identify other potential sources of variation in PSAP creation, we used a second strategy: we estimated the association between state presence and the creation of PSAPs (2000-2017). We propose an explanatory model for the creation of PSAPs in a given municipality, one that combines four accounts found in the literature (explained below). The model is estimated via count data regression. Information on PSAPs was collected from the National Registry of Social Assistance Providers (Cadastro Nacional de Entidades de Assistência Social – CNEAS) and supplemented by socioeconomic indicators from various sources.

Combined, these strategies support our claim that direct state provision has contributed to the decline in PSAP creation. Admittedly, we should not take our regression results as causal, since we work with observational data and lack a design feature that could emulate a counterfactual scenario. Nonetheless, considering the array of alternative explanations we control for, we believe that the fitted models help us make a strong case for H2.

The international literature has sought to explain the location of nonprofit civil society organizations (NPCSOs) in different regions of a country or across countries. However, there is still only a handful of studies in this field (among which, COSTA, 2016 ; JEONG and CUI, 2020 ; LECY and VAN SLYKE, 2013 ; MATSUNAGA, YAMAUCHI, and OKUYAMA, 2010 ), and they do not specifically address the context in which PSAPs operate. Thus, in theoretical terms, we consider this class of phenomena (NPCSO location) broader than the object of this study.

The remainder of this article is organized around our hypotheses and divided into eight sections. The first section presents the theoretical basis for H1 (institutional change); the subsequent three sections offer an analysis of the relevant legislation. The fifth section presents H2 (state provision and new PSAPs), followed by a description of PSAP creation since 2001 and the multivariate analysis thereof. The conclusion presents our final considerations.

Hybrid systems and institutional change

The Brazilian social protection system has a hybrid nature. Rodrigues-Silveira (2010) and Sátyro and P. Cunha (2018b) have mapped the different welfare regimes that coexist in the country. Post, Bronsoler, and Salman (2017) have devised a useful analytical framework for understanding such systems. The authors propose a typology of hybrid systems where public services are provided through state and non-state organizations. The typology is based on two dimensions: type of state involvement – direct or indirect provision – and prevalence of private providers – marginal or extensive. Post et al. (2017) classify hybrid systems into four categories: state-dominant (direct state provision and marginal prevalence of private providers), supplemented state (direct state involvement and extensive private prevalence), regulated provision (indirect state provision and marginal private penetration), and ‘free’ market (indirect state provision and extensive private prevalence).

The typology seems appropriate for the analysis of PSAPs. Our interpretation is that, in recent years, Brazil went from being a state with indirect involvement in the provision of social assistance services (resembling the ‘free’ market type) to a supplemented state system, which, we argue, has slowed the creation of PSAPs. We understand that the state’s new leading role in social assistance – assuming more responsibilities and directly providing services – and the regulation of the sector have removed some of the incentives for creating new PSAPs. We believe that the resulting institutional framework, stemming from the central government’s programmatic vision, has transformed the role of the state in social assistance, resulting in fewer incentives and less potential demand for new PSAPs. We argue that there has been an institutional change, advanced by the central government, which led to a shift in the role of the state.

The contemporary literature identified with historical institutionalism addresses institutional change by focusing both on the pace of change (incremental or radical) and on explanations for its occurrence ( MAHONEY and THELEN, 2010 ). Among the causal mechanisms that generate incremental institutional change, Mahoney and Thelen (2010) highlight agents’ compliance with the rules – obedience generates stability while disobedience brings about resistance and conflicts, which in turn can lead to change. Therefore, agents – who are especially sensitive to distributional implications of the rules – sustain or withdraw their support for an institution according to the political context and characteristics of the institution itself. Mahoney and Thelen (2010) describe four types of rule-related change: 01. displacement, when new rules are introduced, in opposition to previous ones; 02. layering, when new rules overlap old ones, changing the original stability; 03. drift, when rules are deliberately kept constant, regardless of contextual shifts that alter their effects; and 04. conversion, when pre-existing rules are strategically employed in a different sense.

Therefore, in public policy analysis, it is important to consider the contexts and the role of leading decision-makers in the promotion of change. Such agents mobilize cultural and ideological repertoires that are reflected in the formulation and implementation of policies ( THELEN and STEINMO, 1992 ). For that matter, the power of the agenda ( KINGDON, 1995 ) emerges as an important mechanism in the study of change, since it materializes in choices and decisions on rules, which creates incentives and constraints. In Brazil, given that many public policies are national policies and that the legislative agenda is historically set by the leadership of the federal executive ( FIGUEIREDO and LIMONGI, 1999 ), the presidential agenda power must be considered.

In the realm of non-state providers, there has been growing regulation – for NPCSOs in general and those in social assistance specifically. In this sector, regulation was the product of a presidential agenda committed to the structuring of a national policy (the PNAS). The fact that the state’s role has changed has also transformed the institutional context for private providers. These changes are discussed in the next three sections.

First regulations and proliferation of PSAPs

The process in which the legal framework for NPCSO operation in Brazil was developed has been studied by some authors; however, only a few have focused on social assistance providers, such as Mestriner (2008) and Brettas (2016) , who have examined this issue comprehensively. The following discussion is based on their work, supplemented by others who have helped make sense of the regulatory transition of NPCSOs ( ALVES and KOGA, 2006 ; SILVA, 2010 ; STUCHI, PAULA, and PAZ, 2012 ).

The first regulation on budgetary transfers to NPCSOs was issued in 1931 (Decree Nº 20,351). Despite introducing an auditing process, the decree was silent on the use of such resources, which could be freely employed by NPCSOs. Law 91/1935 instituted the Title of Federal Public Utility (Título de Utilidade Pública Federal), which could be granted to NPCSOs. Organizations possessing the Title were obligated to submit activity reports to state departments if demanded. In exchange, NPCSOs could mention the Title in promotional material (e.g., emblem, flag). Later, in 1938, the National Council of Social Service (Conselho Nacional de Serviço Social) was created, an important step in the process of regulating federal transfers; the Council became responsible for releasing such resources based on the evaluation of proposals.

In 1942, Decree-Law Nº 4,830 recognized the Brazilian Legion of Assistance (Legião Brasileira de Assistência – LBA) as a nonprofit organization working in close collaboration with the state. The LBA provided social assistance services, especially maternal and child protection, either directly or in partnership with specialized providers. Brettas (2016) describes the creation of the LBA as the cornerstone of a nationwide system of public assistance.

Although the 1946 Federal Constitution had exempted social assistance nonprofits from taxation, regulation of such provision was introduced only a decade later (Law Nº 3,193/1957). In 1959, Law Nº 3,577 provided economic benefits to organizations possessing the Title of Federal Public Utility by freeing them from paying the employer’s pension and retirement contributions; in 1961, Decree Nº 50,517 finally regulated Law Nº 91/1935, which had created the Title ( BRETTAS, 2016 ; SILVA, 2010 ). We take these tax and contribution exemptions as a plausible, yet not sole, explanation for the substantial rise in the rate of PSAP creation starting in the mid-1960s ( Figure 01 ). Another relevant factor was the creation of a contractual instrument in the 1960s, the agreement (convênio); this type of covenant was vastly used by the state to commission NPCSOs until Law Nº 13,019/2014 was passed.

Note that during the military dictatorship (1964-1985) no specific regulation was aimed at PSAPs. In this period, three events stand out: the creation of the National Foundation for the Well-Being of Minors (Funabem) in 1964; the issuing of Decree-Law Nº 593/1969, which transformed the LBA from a civil association into a public foundation; and the creation of the Program for the Assistance of the Rural Worker (Prorural) in 1971. The transformation of the LBA potentially explains the increased number of PSAPs created in the period, but we have not found any evidence in the literature to support that specific claim. However, several studies argue that the military tried to gain legitimacy and increase their support base by delivering social assistance benefits and implementing more comprehensive social programs ( MIOTO and NOGUEIRA, 2013 ).

Social assistance was forged through the benevolence of organized civil society – initially, through the Catholic Church, then by other religious denominations and secular philanthropic organizations ( SÁTYRO and CUNHA, 2014 ). For a long period, the Brazilian state chose to transfer the responsibility for social assistance to civil society. Legislation preceding the FC88 is consistent with a system in which the state plays a subsidiary role in the social protection of the most vulnerable. A national policy of social assistance was only established after the redemocratization.

From the 1988 Federal Constitution to the FHC administrations

The FC/88 assigned social assistance responsibilities to the state – more than that, it attributed to the state the primary responsibility for protecting the needy. This happened during a period of surge in PSAP creation, when PSAPs were routinely operating without much oversight over the services they provided or the public they served.

Sátyro and Cunha (2014) demonstrate that the FC/88 alone, however, did not drive governments to develop a plan capable of breaking out of this legacy of PSAP autonomy. On the contrary, the Organic Law of Social Assistance (Lei Orgânica da Assistência Social, Law Nº 8,742/1993) faced substantial resistance before it was finally passed, five years after the enactment of the FC88.

During the presidency of Fernando Henrique Cardoso (FHC, 1995-2002), a set of isolated actions were taken, with the state participating only in some niches of social assistance, and for specific target populations. The FHC government advanced a strong state reform agenda and even created a ministry for this purpose, the Ministry of Federal Administration and State Reform (Ministério de Administração Federal e Reforma do Estado), under the command of Luiz Carlos Bresser-Pereira ( BRESSER-PEREIRA, 1998 ; BRESSER-PEREIRA and GRAU, 1999 ; FISCHER and FALCONER, 1998 ). This rather conservative agenda, aimed at reducing state responsibilities, was divided into four basic components – among them the “delimitation of state functions, reducing their size through privatization, outsourcing, and ‘ publicização ’ (the latter implying the transfer of the social and scientific services that are currently provided by the State to the non-state public sector)” ( BRESSER-PEREIRA, 1998 , p. 60; our translation from Portuguese).

Despite this project of setting up a residual participation of the state in social assistance, a modest expansion of public responsibility began in the 1990s, especially after the LBA was extinguished and the decentralization process began to advance ( JACCOUD, LICIO and LEANDRO, 2018 ). The LBA was dismantled exactly on the first day of the FHC administration, altering the relationship between the federal government and the social assistance entities. The closing of LBA also affected the interactions between different levels of government, which until then had been characterized by cronyism and bargains concerning the public resources allocated by the LBA. In addition, Law Nº 8,742/1993 had assigned municipalities the responsibility of establishing partnerships with social assistance entities in their territories, thus reducing the federal government’s role in directly financing these organizations. Once a municipality had taken charge of provision, the central government would not sign new agreements with PSAPs in that locality ( ARRETCHE, 1999 )² .

An important strategy used to develop this new type of relationship between the levels of the federation and to discuss the municipalities’ responsibilities regarding social assistance entities was to create the Tripartite Interagency Committee (Comissão Intergestores Tripartite – CIT), composed of social assistance managers from the federal, state, and local levels.

Despite the regulations, the nature and objectives of the direct and indirect provisions were not clear ( JACCOUD, LICIO, and LEANDRO, 2018 ), and there was no room for the state to take primary responsibility for social assistance. Conversely, the Solidary Community Program (Programa Comunidade Solidária), created in 1995, became the flagship initiative in the area. Consistent with the existing vision for the state, the Solidary Community Council was shaped like a quasi-nongovernmental organization (quango). It should operate as an interface between government and civil society, with the clear goal of encouraging the creation and strengthening of non-state public arenas, as well as promoting innovative forms of mobilization and partnerships to fight poverty and social exclusion ( ALVES and KOGA, 2006 ; PERES, 2005 ). One of the Solidary Community Council’s tasks was to develop a legal framework for the third sector. The proposal emphasized the strategic role of NPCSOs and designed partnership models between state and NPCSOs, fostering transparency and accountability ( ALVES and KOGA, 2006 ).

The federal government took other actions still. In 1996, the Continuous Cash Benefit (Benefício de Prestação Continuada – BPC) was implemented to assist the low-income disabled and the low-income elderly (sixty-five years or more) with a monthly minimum wage; beneficiaries had to prove that they did not have the means for their survival, nor could they rely on family for that end. This benefit had been created by the FC88 itself. In that same year, the Program for Eradication of Child Labor (Programa de Erradicação do Trabalho Infantil – PETI) was launched to tackle the worst forms of child labor by transferring cash to families with working children. In addition, in 2001 and 2002, three other conditional cash transfer initiatives were implemented, all targeted to low-income families with children: Auxílio Gás (Gas Aid), Bolsa Escola (School Grant), and the Cartão Alimentação (Food Card).

Nonetheless, actions designed to effectively institutionalize the social assistance policy did not have a place on the agenda of the FHC government. One indicative evidence of that was the lack of institutional space for this policy: social assistance did not have a ministry of its own but shared one with Social Security (Ministério da Previdência e Assistência Social) – it was allocated in the shadow of a much more prominent policy area ( SÁTYRO and CUNHA, 2014 ).

At that time, new regulation kept the responsibility for social assistance provision in civil society and reinforced the vision that state intervention should be minimal. The legislation issued addressed the question of third sector professionalization: the Volunteerism Law (Law Nº 9,608/1998); the Social Organizations Law (Law Nº 9,637/1998); and the Third Sector Law (Law Nº 9,790/1999) ( ALVES and KOGA, 2006 ; SILVA, 2010 ).

As a result, the residual role of the state in the provision of social assistance was sustained. The central government only implemented targeted, low-coverage cash transfers; social assistance lacked institutional space, which suggests that it was not central to the government’s agenda. The focus was on downsizing the state, a view that pushes for service outsourcing.

It is our understanding that the increase in the number of annual PSAP registrations in the period reflects the presidential project, manifested in the constant use of terms such as social responsibility and ‘publicização’. The tone had been set for strengthening the role of civil society in providing services of public interest. The governmental agenda preserved the legacy of indirect provision and PSAP autonomy, reinforcing the importance of PSAPs without emphasizing state responsibility as a guideline or principle to be followed in social assistance.

Change of paradigm: the primacy of the state and the place of PSAPs in SUAS

PSAP creation continued at a high rate through most of the 2000s ( Figure 01 ). However, this trend was now unfolding in a rather different political landscape. With the election of Luiz Inácio Lula da Silva (Lula) in 2002, a new vision of the role of the state in social protection was presented; especially, the social assistance policy was effectively integrated into the political arena. Having a left-wing government with a vision that was substantially different from that of previous governments in two areas – the size of the state in general and its role in social assistance in particular – is seen as a key factor driving the paradigmatic change in the area ( SÁTYRO and CUNHA, 2014 ).

In 2003, existing cash transfer initiatives were consolidated into the Programa Bolsa Família (Family Grant Program). In 2004, the Ministry of Social Development (Ministério do Desenvolvimento Social e Combate à Fome – MDS) was created and the SUAS started to be structured by bringing together a network of state and non-state actors that provides transfers, benefits, and social assistance services. Since the formulation of the PNAS (CNAS Resolution 145/2004), the MDS had been constantly producing more operating rules for this policy area, building state capacity at the local level and creating financial incentives for expanding the provision of public service in the municipalities ( MESQUITA, PAIVA, and JACCOUD, 2020 )³ . With these measures in place, many municipalities chose to join SUAS. In doing so, they took responsibility for managing and providing social assistance services, but they also became subject to centrally-defined guidelines – a requirement for receiving federal resources ( SÁTYRO and CUNHA, 2014 ; SÁTYRO and CUNHA, 2018a ).

According to Stuchi, Paula, and Paz (2012) , by the end of 2007 social assistance had practically been structured as recommended by the FC88: a citizen’s right and a responsibility of the state. In fact, social assistance had transitioned from a ‘free’ market system – indirect state involvement and dominance of non-state provision of services ( POST, BRONSOLER, and SALMAN, 2017 ) – to a system where direct state provision and extensive private provision coexist.

The institutional arrangement that since 2004 was being built for the implementation of SUAS started to operate as government facilities were being inaugurated to provide services for the three levels of social protection ( Figure 02 ). Social Assistance Reference Centers (Centros de Referência de Assistência Social – CRAS) and Community Centers (Centros de Convivência) offer basic protection. Specialized Reference Centers for Social Assistance (Centros de Referência Especializados de Assistência Social – CREAS), Centers for the Homeless Population (known as Centros POP), and Day Centers (Centros-Dia, for people with disabilities and their families) provide social protection of medium complexity. Finally, Foster Units (Unidades de Acolhimento, which admit children and adults) and Foster Family Units (Unidades Executoras do Serviço de Acolhimento em Família Acolhedora) work with special protection of high complexity.

Figure 02
Public facilities providing social assistance services (2003-2017)*

Note: *Only government facilities created since 2003 are considered, totaling 15,649 (92%) of the 16,987 government facilities identified in the 2017 SUAS Census.

According to Colin and Jaccoud (2013 , pp. 49-50), when the first CRAS and CREAS were inaugurated (2005), there were 2,292 municipal social assistance facilities, which were then converted into 1,978 CRAS and 314 CREAS. In 2010, 4,823 (87%) of the 5,570 municipalities in the country had at least one CRAS. Notably, the fall in new PSAP registrations starting in 2006 coincides with the creation of SUAS public facilities. However, the expansion of public facilities alone does not explain Figure 01 . In addition to considering the implementation of SUAS – more specifically, the start of social assistance provision through public facilities –, one must acknowledge that at that moment the state had significant regulatory power and transformative capacity. We argue that the state becoming a provider of social assistance services ‘and’ a regulator of the private network discouraged new PSAPs from opening. The second explanatory factor is suggested by the analysis of the relevant legislation, as shown below.

In the mid-2000s, the National Council of Social Assistance (Conselho Nacional de Assistência Social – CNAS) discussed the role of PSAPs in SUAS and the relationship of nonprofit providers with the National Secretariat for Social Assistance (Secretaria Nacional de Assistência Social – SNAS). A salient topic of that discussion was the Certificate of Charitable Social Assistance Provider (Certificado de Entidade Beneficente de Assistência Social – CEBAS), which exempts providers from paying social security contributions and allows nonprofits (initially only those working in social assistance, health, and education) to establish partnerships with public authorities. For a CEBAS to be issued, it had to be approved by the CNAS, composed of representatives of the MDS and of the social assistance entities; these entities had an active role in defining the exemptions, even though the exemptions were effectively granted by the ministries of Finance and Social Security ( STUCHI, PAULA, and PAZ, 2012 ). The debate was marked by a dispute about how to define what is a charitable organization working in social assistance, health, and education.

Law Nº 12,101/2009 transferred the responsibility of issuing CEBAS to the ministries – in the case of social assistance, to the MDS. This change affected negatively and significantly the entities’ expectations regarding CEBAS and the possibility of benefiting from CEBAS-based exemptions. Stuchi et al. (2012) identify two lines of thought overlooked by this law: first, that which advocated an encompassing definition of charitable organizations in social assistance, emphasizing the affinities between health, education, and social assistance services; second, the one that supported nonprofits’ liberty to act without state regulation, as they had historically done. At the time, a heated controversy over the role of PSAPs took place at the MDS. Although there was strong resistance to their existence at first, they were later not only recognized as part of SUAS but also seen as strategic providers of services that the state was not yet able to offer directly, despite the implementation of CREAS and the Foster Units.

Other important steps were taken to regulate PSAPs, an indication that the state was moving on to act as a regulator, exercising its transformative capacity in a way that was consistent with the new vision for its role. In 2007, Decree Nº 6,308 defined the criteria for characterizing social assistance providers and organizations. In 2009, CNAS Resolution 109 created the National Typification of Social Assistance Services (Tipificação Nacional de Serviços Socioassistenciais), a catalog of services. This inventory was the result of a dialogue between the MDS, CNAS, and CIT.

Brettas (2016) highlights the creation, by Decree Nº 7,079/2010, of a department to oversee the network of SUAS private social assistance (Departamento da Rede Socioassistencial Privada do SUAS) – a testimony of how relevant these organizations were for the effectiveness of the PNAS, as well as of how interested the state was in regulating and monitoring the performance of PSAPs in the system. In the following year, Law Nº 12,435/2011 defined the criteria for recognizing NPCSOs as part of the PNAS and classified PSAPs into three categories: care (atendimento); advising (assessoramento), focused on strengthening social movements, service users’ associations, and leadership); and advocacy for social assistance rights (defesa e garantia de direitos sociais). This classification was further specified by CNAS Resolutions Nº 27/2011 and 14/2014.

The necessary measures for state and local governments to assume the responsibility for monitoring and inspecting the entities were discussed and negotiated at the CIT. In 2014, the National Program for the Improvement of the Network of SUAS Private Social Assistance was initiated (Resolution CNAS 4/2014). The Program aimed at monitoring and upgrading the services offered by PSAPs; this Resolution also devised the CNEAS registry. In addition, the Legal Framework for the Civil Society Organizations (Marco Regulatório das Organizações da Sociedade Civil – MROSC, Law Nº 13.019/2014) was passed. The MROSC abolished the direct agreements (convênios) between NPCSOs and the state, which were replaced by competitive calls for proposals (editais públicos). Therefore, all interested NPCSOs could apply. The MROSC entered into force only in 2016 for the federal and state governments, and in 2017 for those at the municipal level.

Above we listed in chronological order the regulations that affected the performance of nonprofit, private entities in SUAS. However, these legal rules have different natures, having been issued by different authorities. Chart 01 provides an alternative presentation, as it lists the 25 legal rules cited by type and issuer. Understanding where the changes came from is important. For example, the CNAS is formed by representatives of entities and government, while the CIT comprises federal, state, and municipal managers. It is worth stressing that draft versions of CNAS Resolutions 109/2009, 27/2011, 4/2014, and 14/2014 were discussed both at the CNAS and at the CIT, which favors, at least in principle, their formulation and implementation within the SUAS.

Chart 01
Legal rules cited in this article, by type and issuer

Type of rule	Issuer	Number	Year	Brief description
Constitution	Constituent Assembly	-	1946	Exempted social assistance nonprofits from taxation.
		-	1988	Established the state’s responsibility for social assistance.
Law	Parliament; president	91	1935	Instituted the Title of Federal Public Utility (Título de Utilidade Pública Federal).
		3,193	1957	Regulated social assistance nonprofits' exemption from taxation.
	Parliament (vetoed by the president)	3,577	1959	Freed organizations possessing the Title of Federal Public Utility from paying the employer’s pension and retirement contributions.
	Parliament; president	8,742	1993	Organic Law of Social Assistance (Lei Orgânica da Assistência Social).
		9,608	1998	Volunteerism Law.
		9,637	1998	Social Organizations Law.
		9,790	1999	Third Sector Law.
		12,101	2009	Transferred the responsibility for issuing CEBAS to the ministries – in the case of social assistance, to the MDS.
		12,435	2011	Defined the criteria for recognizing NPCSOs as part of the PNAS and classified PSAPs into three types: care; advising; and advocacy for social assistance rights.
		13,019	2014	Establilshed the Legal Framework for the Civil Society Organizations (Marco Regulatório das Organizações da Sociedade Civil – MROSC).
Decree-Law	President	4,830	1942	Recognized the LBA as a nonprofit working in close collaboration with the state.
		593	1969	Transformed the LBA from a civil association into a public foundation.
Decree	Chief of the Provisional Government	20,351	1931	Regulated budgetary transfers to NPCSOs.
	President	50,517	1961	Regulated Law 91/1935 (which had created the Title of Federal Public Utility).
		6,308	2007	Defined the criteria for characterizing social assistance providers and organizations.
		7,079	2010	Created a department to oversee the network of SUAS private social assistance (Departamento da Rede Socioassistencial Privada do SUAS).
Resolution	CNAS	145	2004	Instituted the National Policy of Social Assistance (Política Nacional de Assistência Social – PNAS).
	CNAS**	109	2009	Launched the National Typification of Social Assistance Services (Tipificação Nacional de Serviços Socioassistenciais).
		27	2011	Specified the criteria for the classification of PSAPs.
		4	2014	Created the National Program for the Improvement of the Network of SUAS Private Social Assistance.
		14	2014	Specified the criteria for classifying PSAPs.
	CNAS	21	2016	Established that complete registration in the CNEAS is required for granting SUAS funds to a PSAP.
Ordinance*	Minister of Social and Agrarian Development	130	2017	Established that complete registration in the CNEAS is required for granting SUAS funds to a PSAP.

Source: Created by the authors with the raw data from legal rules cited in this article.
• Notes: *Portaria, in Portuguese. **Drafts of resolutions were also discussed by the CIT. Abbreviations and their meaning are listed below: CEBAS - Certificate of Charitable Social Assistance Provider (Certificado de Entidade Beneficente de Assistência Social); CIT - Tripartite Interagency Committee (Comissão Intergestores Tripartite); CNAS – National Council of Social Assistance (Conselho Nacional de Assistência Social); CNEAS - National Registry of Social Assistance Providers (Cadastro Nacional de Entidades de Assistência Social); LBA - Brazilian Legion of Assistance (Legião Brasileira de Assistência); MDS - Ministry of Social Development (Ministério do Desenvolvimento Social e Combate à Fome); NPCSOs - Nonprofit civil society organizations; PNAS - National Policy of Social Assistance (Política Nacional de Assistência Social); PSAPs - Private social assistance providers; SUAS – Unified System of Social Assistance (Sistema Único de Assistência Social).

So far, we have argued that the change in government brought in a different vision of social assistance, which led to the institutionalization of the social assistance policy according to the principles embedded in the FC88. The resulting system is characterized by state responsibility and provision in coexistence with private provision; however, the new arrangement removes private entities’ autonomy and regulates their integration into the system and the way in which the service is provided. In this transformation process, it became imperative to review PSAPs’ position within this policy area. As the state adopted a coordinating role, it became necessary to regulate PSAPs.

We believe that state regulation is an explanatory factor for the decline in the PSAP registration rate after 2007 ( Figure 01 ). The government’s use of its regulatory power in a series of initiatives (e.g., issuing of rules for PSAP recognition and certification, typification of services, abolishment of agreements for direct financing), coupled with the establishment of public facilities, has arguably affected the rate of PSAP creation. Although state action did not erase the legacy of indirect provision and PSAP autonomy, it reshaped the system by combining regulated provision with the provision of governmental services of high penetration.

In these last three sections, we retraced the evolution of Brazil’s social assistance system. It is our view that the recent institutional change occurred through displacement, that is the issuing of new rules that modified previous ones. We highlight that the regulatory changes discussed above have national reach; however, the fact that the installation of public facilities varies across the territory offers an opportunity to examine the effects the presence of such facilities. In the next sections, we address theoretically and empirically the question of whether direct state provision of services has inhibited the emergence of new PSAPs.

Accounts of NPCSO creation

There are two main approaches for explaining the location of NPCSOs. For the Government Failure Theory ( WEISBROD, 1977 ), the existence of NPCSOs is explained by deficiencies in government action, which leads to the appearance of services that replace or even compete with those provided by the state. On the other hand, the Interdependence Theory ( YOUNG, 2000 ) sees private provision as complementary, since neither the state nor the market has the expertise and funds to meet the various demands by themselves ( SALAMON, 1987 ). As public appropriations are an important revenue source for nonprofits’, the Interdependence. Theory predicts a positive relationship between government size and nonprofit density. Analysts argue that these visions are not mutually exclusive and that together such perspectives characterize the complexity of the phenomenon at hand ( LECY and VAN SLYKE, 2013 ; LIU, 2017 ; YOUNG, 2000 ). Jeong and Cui (2020) point out that for-profit service organizations should also be listed as explanatory factors for the spatial distribution of nonprofits; however, there are no for-profit service providers of social assistance in Brazil.

Based on the Government Failure Theory, as well as on some of its empirical applications and expansions, our regression models (below) consider four different potential accounts for PSAP creation in Brazil. The first is government failure. In studies about NPCSOs (e.g., COSTA, 2016 ; McKEEVER, 2015 ), state participation is more often measured by the amount of expenses than by service provision. This tendency possibly derives from the very object under analysis: NPCSOs in the United States – a country with a residual welfare state system, which provides minimal social protection through the state, typically via cash transfers. Regarding PSAPs, H2 states that PSAPs are more likely to be created in places (municipalities) where the state is less active in social assistance; H2 is thus aligned with the Government Failure Theory. H2 is examined not only at the aggregate level, but also by the type of service provided – by a PSAP or the state – in each municipality. That allows for a fine-grained analysis in which evidence of government failure (i.e., a negative association between state presence and PSAPs) or, conversely, interdependence (i.e., a positive association between state presence and PSAPs) can be identified by the type of service offered.

The second account discusses potential demand, that is the level of poverty not tackled by the state; it is taken as a conditioning factor for the creation of PSAPs because the vulnerable (economically or otherwise) constitute the target population of NPCSOs ( FRUTTERO and GAURI, 2005 ; LECY and VAN SLYKE, 2013 ). Thus, the higher the unmet demand is, the greater the expectation will be for new PSAPs to be located in a given municipality, all else constant. To assess potential demand, however, one must consider poverty from a multidimensional standpoint. In fact, the national policy (the PNAS) seeks to address different manifestations of vulnerability.

While the account of the Government Failure Theory and the potential demand emphasizes unmet needs as the main motivation for the creation of NPCSOs, they ignore other determinants of the location of non-state organizations. Critics ( BEN-NER and VAN HOOMISSEN, 1991 ; LU, 2020 ) argue that a high degree of poverty ‘per se’ does not lead to NPCSO provision of services; nonprofits will exist when and where resources are available (financial, human, service expertise). Therefore, the third account posits that the more available local resources are – whether they originate in the municipality’s treasury, in private companies, or in the skills of a given population – the more likely it is that PSAPs will be created. Moreover, covariates such as local government investment and capacity to raise its own revenue shed some light on municipalities’ capacity to fund nonprofits, allowing for a consideration of the Interdependence Theory’s predictions.

A more specific view of existing resources points to the previous presence (or density) of NPCSOs as a predictor of the creation rate and spatial distribution of these providers ( COSTA, 2016 ; FRUTTERO and GAURI, 2005 ). A high density of NPCSOs may indicate that demand and resources are present. Furthermore, Costa (2016) suggests that agglomeration might bring organizational benefits to these entities, such as access to specialized resources, knowledge sharing, and lower costs to identify the target population. According to this author’s analyses, the density of NPCSOs in Brazilian municipalities is the most important factor in defining the location of new NPCSOs. Thus, in line with the fourth account, it is expected that the higher the PSAP density is, the greater the chances are that other PSAPs will be created in a given locality, ‘ceteris paribus’.

PSAP creation from 2001 to 2017

Data on PSAPs were collected from the National Registry of Social Assistance Providers (Cadastro Nacional de Entidades de Assistência Social – CNEAS). Municipal Councils of Social Assistance feed the CNEAS with data on nonprofit providers registered in these councils. On June 1^st, 2017 (date of extraction), 19,159 PSAPs⁴ were registered in the CNEAS, at the time maintained by the Ministry of Social Development (Ministério do Desenvolvimento Social e Combate à Fome – MDS), with 18,562 different CNPJs (see notes on Figure 01 )⁵ . Near 18 thousand PSAPs (94%)⁶ informed the level of social protection provided. For 16,218 of these, year of creation was also available; the forthcoming analyses ( Figures 03 and A01 plus all tables) consider this set of PSAPs.

Figure 03
PSAPs created, per ten thousand inhabitants (2001-2010 and 2011-2017)

Notes: Only PSAPs with information available for both year of creation and level of social protection provided are considered (n = 16,218). Moran’s I for 2001-2010 = 0.08 (statistically significant at 1%); Moran's I for 2011-2017 = 0.01 (not statistically significant at 10%).

In our sample, out of the 5,570 Brazilian municipalities, 4,113 (74%) had no PSAP created between 2001 and 2010; between 2011 and 2017, that percentage rose to 95% (5,267 municipalities without new PSAPs). Figure 03 shows municipalities according to the number of PSAPs created, per ten thousand inhabitants. It depicts the drastic drop in new PSAPs, all across the country.

Nearly 30% of the existing PSAPs were created between 2001 and 2010 ( Table 01 ), while less than 03% were created between 2011 and 2017. The bulk of the PSAPs created since 2001 are distributed similarly to those created previously. About half of the PSAPs in the country and approximately half of the PSAPs created since 2001 are located in the Southeast. Notably, the percentage of PSAPs created in the Northeast region increased more than 50% (from 15.3 to 23.2%), while the South and Southeast lost about five percentage points.

Table 01
PSAPs created, by region, level of social protection offered, and year of creation

	Created up until 2017*			Created up until 2000		Created between 2001 and 2010			Created between 2011 and 2017*
	Frequency	%		Frequency	%	Frequency	%		Frequency	%
Region
Brazil	16,218	100.0		11,006	100.0	4,781	100.0		431	100.0
Central-West	1,078	6.6		722	6.6	326	6.8		30	7.0
North	623	3.8		320	2.9	281	5.9		22	5.1
Northeast	2,900	17.9		1,679	15.3	1,121	23.4		100	23.2
South	3,115	19.2		2,319	21.1	725	15.2		71	16.5
Southeast	8,502	52.4		5,966	54.2	2,328	48.7		208	48.3
Level of protection: aggregated (one PSAP may be included in more than one level)
Basic	12,084	74.5		8,144	74.0	3,588	75.1		352	81.7
Medium complexity	10,410	64.2		6,969	63.3	3,162	66.1		279	64.7
High complexity	7,560	46.6		5,210	47.3	2,133	44.6		217	50.3
Level of protection: detailed (each PSAP is included in only one level)
Basic only	4,351	26.8		2,911	26.5	1,310	27.4		130	30.2
High complexity only	830	5.1		647	5.9	168	3.5		15	3.5
Basic + High complexity	627	3.9		479	4.4	141	3.0		7	1.6
Basic + Medium complexity + High complexity	5,599	34.5		3,691	33.5	1,723	36.0		185	42.9
Medium complexity only	2,800	17.3		1,822	16.6	924	19.3		54	12.5
Medium complexity + High complexity	504	3.1		393	3.6	101	2.1		10	2.3
Basic + Medium complexity	1,507	9.3		1,063	9.7	414	8.7		30	7.0

^st
• Notes: *By the time the data were collected (June 1^st, 2017), no PSAP with information for year of creation had been registered in 2017. This table considers only PSAPs with information for year of creation and level of protection offered (n = 16,218).

Among the PSAPs created since 2001, over three-quarters offer basic protection, about two-thirds offer medium-complexity services, and nearly 45% offer high-complexity services, as shown in Table 01 . The majority (34.5%) of the existing PSAPs offers all three levels of protection, a type of PSAP that is more prevalent among PSAPs created since 2001, reaching 42.9% among those created between 2011 and 2017. The second and third most common types of PSAP offer only basic protection (26.8%) and only medium-complexity services (17.3%) respectively. Frequently, PSAPs in the high-complexity segment offer other levels of social protection as well.

The account and figures above demonstrate that a large number of new PSAPs were being created each year until very recently. Even though the number of entrants has drastically reduced, a considerable stock of PSAPs has accumulated, ranging across all levels of social protection. In the next section, we propose a multivariate model for estimating the association between state presence and PSAP creation.

Direct state provision and PSAP location across municipalities

Unconditional associations in the sample defy the expectation that increased state presence helps explain the fall in the rate of PSAP creation since the mid-2000s (H2), therefore favoring, at face value, the Interdependence Theory (to the detriment of the government failure account). Pearson correlation coefficients between presence of public facilities (specifically, CRAS, Community Centers, CREAS, and Foster Units)⁷ and the count of PSAPs created, considering the panel data comprised of the two periods analyzed (2001-2010 and 2011-2017)⁸ , are all positive and statistically significant (at α = 2%). In fairness, these are not substantial, except for the CRAS coefficient, of 0.19. For well-known reasons, mere correlations would not be an adequate strategy for testing H2. We propose instead an empirical model structured around the four possible accounts of PSAP creation:

${\mathrm{P}\mathrm{S}\mathrm{A}\mathrm{P}}_{\mathrm{i}\mathrm{t}}-\mathrm{P}\mathrm{S}\mathrm{A}\mathrm{P}{\mathrm{P}}_{\mathrm{i}(\mathrm{t}-1)}=f\left(\beta {\mathrm{S}}_{\mathrm{i}(\mathrm{t}-1),\mathrm{t},}\gamma {\mathrm{N}}_{\mathrm{i}(\mathrm{t}-1)},\delta {\mathrm{R}}_{\mathrm{i}(\mathrm{t}-1)},\zeta {\mathrm{D}}_{\mathrm{i}(\mathrm{t}-1)},{\epsilon }_{\mathrm{i}(\mathrm{t}-1),\mathrm{t}}\right)$

The dependent variable is the number of PSAPs created between t-1 and t per municipality ( i ) as a whole and by level of social protection; t is either 2010 or 2017; t-1 is 2000 or 2010 (this is so because the last two population censuses took place in 2000 and 2010, and they are the source of several explanatory variables). To conduct a sensitivity analysis, we estimated the model separately by Brazil’s five macro-regions (Central-West, North, Northeast, South, and Southeast).

The dependent variable captures the sector dynamics, since it is a measure of the flow – as opposed to the stock – of organizations. As other indicators used to depict the size and growth of the nonprofit sector, the number of newly-founded PSAPs has its flaws ( PENNERSTORFER and RUTHERFORD, 2019 ). In particular, it is blind to one organization’s size (e.g., as measured by operational capacity) and growth. Notwithstanding, the number of PSAPs created is our selected measure because we believe it is more reliable than alternative indicators available at the CNEAS (e.g., number of staff) and it is comparable to the dependent variable in Costa (2016) – to our knowledge, the only empirical work that similarly worked on the creation of NPCSOs in Brazil.

S encompasses governmental social assistance facilities in the municipality. For each CRAS, we counted the number of months from January 2003 (Lula’s inauguration) or the beginning of its operation, whichever came first, until year t (more specifically, December 2010 or May 2017). Then, we calculated the age in months of all CRAS in the municipality and divided this sum by twelve. The resulting value is the number of CRAS-years in the locality in t . The same calculations were made for the other types of state facilities (Community Centers, CREAS, and Foster Units). These are our explanatory variables of interest, since they speak to H2. Additionally, S includes the coverage of the Continuous Cash Benefit (BPC) and the Family Grant Program (Programa Bolsa Família); thus, the model represents state presence both in services and in the form of cash transfers (to the elderly, people with disabilities, and poor families)⁹ .

N contains indicators of multidimensional poverty, representing the potential demand account: inequality, poverty, unemployment, child poverty, percentage of elderly, and percentage of people with disabilities. R reports resources available in the municipality, addressing the third account of nonprofit location; it includes population density, percentage of rural population, local government investment capacity, local government capacity to raise its own revenue, and a dummy tagging state capitals¹⁰ . D presents the number of existing PSAPs in the municipality, in reference to the fourth account.

The sample used in the multivariate analysis comprises the 5,527 municipalities (99% of the total 5,570) with complete information for the regression model, in either t = 2010 (5,274) or t = 2017 (5,319). Sources and descriptive statistics of covariates (and other selected variables) are shown in ^{Tables A01} and ^A02 (both in the ^Appendix).

The explanatory variables were chosen based on Costa (2016) . However, our analysis innovates in several aspects: it is specific to social assistance providers; it is the first to explore CNEAS data; it includes state provision of social assistance services and benefits; it uses panel (instead of cross-sectional) data; and it applies econometric models suitable to the nature of the dependent variable (count).

Number of PSAPs created (between t-1 and t ) is a count outcome: it indicates how many times something happened, and it takes discrete, non-negative values. If applied to counts, linear regressions can generate “inefficient, inconsistent, and biased estimates” ( LONG and FREESE, 2001 , p. 223). Specific models are recommended for count data; these models assume for the dependent variable a Poisson distribution, a negative binomial distribution, or variations of such distributions ( AGRESTI, 2007 ; LONG and FREESE, 2001 )¹¹ .

Counts can be calculated for different lengths of time, geographical areas, and other delimitations. The number of probable events is greater when exposure (across time, territory, or else) is increased. Count models deal with this source of variation by including an exposure variable, presented as a logarithm whose regression coefficient is forced to be one – this way, the model adequately controls for exposure, without mistaking it for an explanatory variable of the event per se. In our models, population size was taken as the exposure variable.

Count models are estimated using the maximum likelihood approach, and their fit is assessed by the Bayesian Information Criterion (BIC). Smaller BIC values indicate that the model is a better fit for the data. We tried Poisson and negative binomial specifications, always with random intercepts for municipalities; coefficients reported come from the specification with the smallest BIC.

Estimated coefficients (^{Table A03} and ^{Figure A01}, both in the ^Appendix) are presented in the original log count unit, though commonly count models are reported as incidence rate ratios (IRR), for ease of interpretation. IRR interpretation is similar to that of odds ratio in logit models: the IRR corresponds to the ratio of the incidence (count) of the event for the group exposed to an increment of the explanatory variable to the incidence of the event among units unexposed to such increment. The IRR is obtained by raising the constant e to the power of the estimated coefficient.

Table 02 synthesizes the various specifications estimated. It focuses on the explanatory variables relevant for testing the hypothesis that increased state presence in service provision starting in the 2000s contributed to the plunge in PSAP creation (H2). Instead of regression coefficients, this table reports estimated percentage changes in the incidence rate of PSAPs created, considering that the respective explanatory variable is at its mean level. We illustrate the calculation using the estimate for the association between CRAS presence and PSAP creation (any level of protection) in the countrywide specification. The regression coefficient, -0.007 (^{Table A03}, in the ^Appendix), is statistically significant at α = 1%. Its IRR of 0.993 indicates that a decrease of 0.7% in the incidence rate of PSAPs (i.e., the count of PSAPs created between t-1 and t ) is expected for each one-unit increase in the variable CRAS-years, ceteris paribus. This seemingly immaterial result is better gauged in the context of the regression sample, where the mean value for this explanatory variable is 8.49 (^{Tables A01} and ^A02, in the ^Appendix): in municipalities with CRAS presence at the sample mean level, we expect a 5.943% lower (-0.007 * 8.49) count of new PSAPs than in municipalities with no CRAS, all else constant. If anything, we believe that this calculation underestimates the drop in PSAP creation associated with the establishment of state-run facilities, since the means are calculated over all municipalities within a region, including the ones with no such facilities. In other words, those means are considerably higher in the subsample of municipalities that actually have state-run facilities¹² .

Table 02
Estimated percentage change in the incidence rate of PSAPs created, at X = mean(X) relative to X = 0, by level of PSAP social protection and region

	Brazil	Central-West	North	North-east	South	South-east
X: CRAS-years in municipality, in t
Any level of social protection	-5.9	-35.9	-25.9	-14.1
Basic social protection	-5.9	-40.3
Medium complexity				-51.5	-41.2
High complexity	-7.6		-30.8	-21.5
X: Community Center-years in municipality, in t
Any level of social protection
Basic social protection
Medium complexity
High complexity
X: CREAS-years in municipality, in t
Any level of social protection
Basic social protection
Medium complexity
High complexity
X: Foster Unit-years in municipality, in t
Any level of social protection		-11.3
Basic social protection
Medium complexity
High complexity	-2.7					-5.5

Source: Created by the authors. Refer to Table A01 (Appendix) for sources of raw data.
• Notes: This table shows estimates only for regression coefficients that are statistically significant (at α = 5%) and relevant for testing H2; shading means that the regression coefficient is not statistically significant. For regression coefficients, see Table A03 and Figure A01 (both in the Appendix).

The regression analysis provided four main pieces of evidence. First, in line with H2, all statistically significant coefficients are negative (and so is the vast majority of those that are not significant, as reported in ^{Table A03} and ^{Figure A01}). Therefore, conditional associations have returned the opposite result to the unconditional ones, supporting our claim that the Government Failure Theory sheds light on the consequences of direct state provision in social assistance.

Second, detected associations concentrated on CRAS presence; only one other explanatory variable, concerning Foster Units, returned significant coefficients. Notably, CRAS is the most frequent among state facilities: in 2017, 99% (5,249 out of 5,319) of municipalities in the regression sample had at least one CRAS, 27% (1,424) at least one Community Center, 42% (2,234) at least one CREAS, and 19% (1,034) at least one Foster Unit.

Third, estimated associations seem to vary considerably across the territory. Almost all significant associations were found in the less-developed parts of the country (Central-West, North, and Northeast).

Lastly, although not our main interest here, we should mention that PSAP density initially showed a positive and significant association with the creation of PSAPs (^{Appendix Table A03}). For each pre-existing private provider, the incidence rate for the creation of a PSAP (any level of protection) increases by 2.1%, ceteris paribus. Such association appears to be somewhat stronger for the creation of PSAPs offering high-complexity services; this suggests that economies of agglomeration underlying the fourth account ( FRUTTERO and GAURI, 2005 ; COSTA, 2016 ) may be more relevant to this level of protection.

Conclusion

This paper provides an explanation for the downward trend in PSAP creation in Brazil. Our argument is twofold: We contend that a switch of incentives (H1), combined with increased state provision of services (H2), decreased the PSAP creation rate.

In the 2000s, the central government took effective responsibility for social assistance. The government’s agenda shifted, and a renewed understanding of the state’s role emerged. Consequently, legislation has changed – previously, it encouraged provision by civil society, now it regulates private providers. The Brazilian state went from having almost 2,300 municipal social assistance facilities in 2005, with low penetration and indirect involvement (mainly through subsidies to PSAPs), to exerting regulation capacities ‘and’ directly providing social assistance services throughout the country, at about 17,000 public social assistance facilities in 2017.

For decades, PSAPs had been the only option available for those in need of social assistance services, and these organizations had great latitude. Since Lula’s administration (2003-2010), PSAPs have been submitted to state regulation. As a part of SUAS, PSAPs must adapt to the typified services; therefore, they no longer have the same level of autonomy and are now under stricter oversight. Our document analysis of the relevant legislation reveals that this institutional change occurred by displacement. It resulted from the issuing of new rules, which modified previous ones and advanced the government’s agenda of strengthening social assistance. In fact, a different vision of the state’s role regarding regulation and direct provision emerged in the 2000s. We contend that this new context discouraged the creation of PSAPs.

The argument that increased state provision explains, at least in part, the drop in the PSAP creation rate was analyzed in conjunction with other potential determinants of PSAP location: potential demand, resource availability, and pre-existence of nonprofit providers. Thereby, we were able to measure the association between direct state provision and the location of new PSAPs, controlling for alternative explanatory accounts.

Between 2001 and 2017, not only did PSAP creation lose steam, but the slow-down was also more dramatic in municipalities that operated more facilities (CRAS). Altogether, our findings lend support to the Government Failure Theory.

PSAPs continue to be an integral part of the social assistance system, which is now a hybrid system. Although recent legislation seems to undermine – or, at least, discourage – the creation of PSAPs, nineteen thousand of them are operating across the country, nearly half of which are providing a type of service – of high complexity – with which the state has had little firsthand experience. The complementary nature of the services offered is a striking feature of the current model.

To advance the discussion and allow for our main argument to be further tested, future studies should explore how the implementation schemes for state-level facilities, especially those offering high-complexity services, may affect the location of new PSAPs. In such schemes, coordinated by the respective state governments, municipalities are grouped into intra-state regions ( JACCOUD et al., 2020 ). Research on this topic may shed light on how the associations we found between state presence and new PSAPs vary regionally.

In addition, it would be interesting to compare the trend in PSAP creation with that of NPCSO creation in other areas, especially in those with no significant regulatory changes. Furthermore, it is relevant to hear what PSAP managers have to say about how the regulation has affected the operation and expansion of these organizations. Findings from Mendonça, Medeiros, and Araújo (2019) suggest that regulation in social assistance has created growing tension. According to the authors, the detailed standards generated confusion and imposed bureaucratic burdens on nonprofits, which were not highly professionalized; they have historically operated under religious or charitable – rather than policy-oriented – goals. Such an organizational perspective is yet to be explored, and it would certainly improve our understanding of PSAP creation.

Another approach to consider is to compare the volume of services provided by state and private providers over time. Such analysis would look beyond the number of public facilities and PSAPs created (what we did here), allowing for inferences to be made about how PSAPs’ role in delivering services has evolved. Finally, changes in the federal government’s agenda since 2016 have substantially altered the implementation of the PNAS, with public funding being cut back and philanthropy and volunteering being encouraged as a means of addressing social issues ( IPEA, 2020 ). Future studies will be able to assess the effects of these changes on the creation and maintenance of social assistance entities.

References

• AGRESTI, Alan (2007), An introduction to categorical data analysis . Hoboken: John Wiley & Sons. 394 pp..
• ALVES, Mário Aquino and KOGA, Natália Massaco (2006), Brazilian nonprofit organizations and the new legal framework: an institutional perspective. Revista de Administração Contemporânea. Vol. 10, Nº esp., pp. 213–234.
• AMÂNCIO, Júlia Moretto (2008), Os desafios da gestão de políticas públicas a partir das parcerias entre Estado e sociedade civil: o caso da assistência social em São Paulo. e-cadernos CES. Vol. 02, pp. 01–18.
• ARRETCHE, Marta Teresa da Silva (1999), Políticas sociais no Brasil: descentralização em um Estado federativo. Revista Brasileira de Ciências Sociais . Vol. 14, Nº 40, pp. 111-141.
• BEN-NER, Avner and VAN HOOMISSEN, Theresa (1991), Nonprofit organizations in the mixed economy: a demand and supply analysis. Annals of Public and Cooperative Economics. Vol. 62, N° 04, pp. 519–550.
• BRESSER-PEREIRA, Luiz Carlos (1998), A reforma do Estado dos anos 90: lógica e mecanismos de controle. Lua Nova. Vol. 45, pp. 49–95.
• BRESSER-PEREIRA, Luiz Carlos and GRAU, Nuria Cunill (ed) (1999), O público não-estatal na reforma do Estado . Rio de Janeiro: Editora FGV. 500 pp..
• BRETTAS, Gabriela Horesh (2016), O papel das organizações da sociedade civil na política pública de assistência social no Brasil: dilemas e tensões na provisão de serviços. Master’s dissertation . Escola de Artes, Ciências e Humanidades. Universidade de São Paulo.
• COLIN, Denise and JACCOUD, Luciana (2013). Assistência social e construção do SUAS – balanço e perspectivas: o percurso da assistência social como política de direitos e a trajetória necessária. In: 20 anos da Lei Orgânica da Assistência Social . Edited by COLIN, Denise Ratmann Arruda; CRUS, José Ferreira da; TAPAJÓS, Luciele Maria de Souza, and ALBUQUERQUE, Simone Aparecida. Brasília: MDS. pp. 42-65.
• COSTA, Marcelo Marchesini da (2016), What influences the location of nonprofit organizations? A spatial analysis in Brazil. VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations . Vol. 27, N° 03, pp. 1064–1090.
• FIGUEIREDO, Argelina Cheibub and LIMONGI, Fernando de Magalhães P. (1999), Executivo e Legislativo na nova ordem constitucional . Rio de Janeiro: Editora FGV. 231 pp..
• FISCHER, Rosa Maria and FALCONER, Andres Pablo (1998), Desafios da parceria governo e terceiro setor. Revista de Administração da USP – RAUSP . Vol. 33, N° 01, pp. 12–19.
• FRUTTERO, Anna and GAURI, Varun (2005), The strategic choices of NGOs: location decisions in rural Bangladesh. The Journal of Development Studies . Vol. 41, N° 05, pp. 759–787.
• GABRICH JUNIOR, Elder Carlos; CASTRO, Luiza Moreira Arantes de, and MOREIRA, Maíra dos Santos (2015), O SUAS atua em rede? Uma análise das unidades de acolhimento. Paper presented at I Encontro Nacional de Ensino e Pesquisa no Campo de Públicas. Brasília. Available at ˂http://www.anepcp.org.br/anaisenepcp/20161128174710_st_10_elder_carlos_g.pdf˃ Accessed on October, 18, 2018.
  » http://www.anepcp.org.br/anaisenepcp/20161128174710_st_10_elder_carlos_g.pdf˃
• IPEA – Instituto de Pesquisa Econômica Aplicada (2020), Assistência social. In: Políticas sociais: acompanhamento e análise . Brasília: IPEA. Nº 27, pp. 57-90.
• JACCOUD, Luciana; LICIO, Elaine Cristina; LEANDRO, José Geraldo (2018), Implementação e coordenação de políticas públicas em âmbito federativo: o caso da política nacional de assistência social. In: Implementação de políticas públicas: questões sistêmicas, federativas e intersetoriais. Edited by XIMENES, Daniel de Aquino. Brasília: ENAP. pp. 23-61.
• JACCOUD, Luciana; MESQUITA, Ana Cleusa; LICIO, Elaine Cristina, and LEANDRO, José Geraldo (2020), Implementação e coordenação intergovernamental na política nacional de assistência social. In: Coordenação e relações intergovernamentais nas políticas sociais brasileiras . Edited by JACCOUD, Luciana. Brasília: IPEA. pp. 113-145.
• JEONG, Joowon and CUI, Tracy Shicun (2020), The density of nonprofit organizations: beyond community diversity and resource availability. VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations. Vol. 31, pp 213–226.
• KINGDON, John W. (1995), Agendas, alternatives, and public policies . Nova York: Harper Collins College Publishers. 254 pp..
• LECY, Jesse D. and VAN SLYKE, David M. (2013), Nonprofit sector growth and density: testing theories of government support. Journal of Public Administration Research and Theory. Vol. 23, N° 01, pp. 189–214.
• LIU, Gao (2017), Government decentralization and the size of the nonprofit sector: revisiting the government failure theory. American Review of Public Administration. Vol. 47, N° 06, pp. 619–633.
• LONG, J. Scott and FREESE, Jeremy (2001), Regression models for categorical dependent variables using stata . College Station: Stata Press. 311 pp..
• LU, Jiahuan (2020), Does population heterogeneity really matter to nonprofit sector size? Revisiting Weisbrod’s demand heterogeneity hypothesis. VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations . Vol. 31, N° 05, pp. 1077-1092.
• MAHONEY, James and THELEN, Kathleen (eds)(2010), A theory of gradual institutional change. In: Explaining institutional change: ambiguity, agency, and power. Cambridge/New York: Cambridge University Press. pp. 01–37.
• MATSUNAGA, Yoshiho; YAMAUCHI Naoto, and OKUYAMA, Naoko (2010), What determines the size of the nonprofit sector?: a cross-country analysis of the government failure theory. VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations. Vol. 21, Nº 02, pp. 180–201.
• McKEEVER, Brice S. (2015), The nonprofit sector in brief 2015: public charities, giving, and volunteering. Washington, DC: The Urban Institute. 16 pp..
• MENDONÇA, Patrícia Maria Emerenciano; MEDEIROS, Anny Karine de, and ARAÚJO, Edgilson Tavares de (2019), Models for government-nonprofits partnerships: a comparative analysis of policies for AIDS, social assistance and culture in Brazil. Revista de Administração Pública. Vol. 53, N° 05, pp. 802–820.
• MESQUITA, Ana Cleusa; PAIVA, Andrea Barreto de; JACCOUD, Luciana (2020), Instrumentos financeiros de coordenação no SUAS. In: Coordenação e relações intergovernamentais nas políticas sociais brasileiras . Edited by JACCOUD, Luciana. Brasília: IPEA. pp. 183-213.
• MESTRINER, Maria Luiza (2008), O Estado entre a filantropia e a assistência social . São Paulo: Cortez Editora. 320 pp..
• MIOTO, Regina Celia Tamaso and NOGUEIRA, Vera Maria Ribeiro (2013), Política social e serviço social: os desafios da intervenção profissional. Revista Katálysis . Vol. 16, Nº esp., pp. 61-71.
• PENNERSTORFER, Astrid and RUTHERFORD, Alasdair C. (2019), Measuring growth of the nonprofit sector: the choice of indicators matters. Nonprofit and Voluntary Sector Quarterly. Vol. 48, N° 02, pp. 440–456.
• PERES, Thais Helena de Alcântara (2005), Comunidade solidária: a proposta de um outro modelo para as políticas sociais. Civitas-Revista de Ciências Sociais. Vol. 05, N° 01, pp. 109–126.
• POST, Alison E.; BRONSOLER, Vivian, and SALMAN, Lana (2017), Hybrid regimes for local public goods provision: a framework for analysis. Perspectives on Politics. Vol. 15, N° 04, pp. 952–966.
• RODRIGUES-SILVEIRA, Rodrigo (2010), Gobierno local y Estado de bienestar: regímenes y resultados de la política social en Brasil. Doctoral thesis . Instituto de Iberoamérica. Universidad de Salamanca.
• SALAMON, Lester M. (1987), Of market failure, voluntary failure, and third-party government: toward a theory of government-nonprofit relations in the modern welfare state. Nonprofit and Voluntary Sector Quarterly. Vol. 16, N° 01–02, pp. 29–49.
• SÁTYRO, Natália Guimarães Duarte; CUNHA, Eleonora Schettini Martins (2018a), The transformative capacity of the Brazilian federal government in building a social welfare bureaucracy in the municipalities. Revista de Administração Pública. Vol. 52, N° 03, pp. 363–385.
• SÁTYRO, Natália Guimarães Duarte and CUNHA, Eleonora Schettini Martins (2014), The path of Brazilian social assistance policy post-1988: the significance of institutions and ideas. Brazilian Political Science Review. Vol. 08, N° 01, pp. 80–108.
• SÁTYRO, Natália Guimarães Duarte and CUNHA, Pedro Schettini (2018b), The coexistence of different welfare regimes in the same country: a comparative analysis of the Brazilian municipalities heterogeneity. Journal of Comparative Policy Analysis: Research and Practice. Vol. 21, N° 01, pp. 65–89.
• SILVA, Carlos Eduardo Guerra (2010), Gestão, legislação e fontes de recursos no terceiro setor brasileiro: uma perspectiva histórica. Revista de Administração Pública. Vol. 44, N° 06, pp. 1301–1325.
• STUCHI, Carolina Gabas; PAULA, Renato Francisco dos Santos, and PAZ, Rosangela Dias Oliveira da (eds)(2012), Assistência social e filantropia: cenários contemporâneos . São Paulo: Veras Editora. 352 pp..
• THELEN, Kathleen and STEINMO, Sven (1992), Historical institutionalism in comparative politics. In: Structuring politics: historical institutionalism in comparative analysis. Edited by STEINMO, Sven; THELEN, Kathleen, and LONGSTRETH, Frank. Cambridge/New York: Cambridge University Press. pp. 01–32.
• WEISBROD, Burton Allen (1977), The voluntary nonprofit sector: an economic analysis. Lexington: Lexington Books. 179 pp..
• YOUNG, Dennis R. (2000), Alternative models of government-nonprofit sector relations: theoretical and international perspectives. Nonprofit and Voluntary Sector Quarterly. Vol. 29, N° 01, pp. 149–172.

• *
  Data from the National Registry of Social Assistance Providers (Cadastro Nacional de Entidades de Assistência Social – CNEAS) – this article’s main data source – were made available to the authors through a statement of responsibility signed with the Ministry of Social and Agrarian Development (Ministério do Desenvolvimento Social e Agrário, Secretaria Nacional da Assistência Social, Departamento de Gestão do SUAS, Coordenação-Geral de Rede e Sistemas de Informações do SUAS).
• 2
  By 1997, only 33% of the municipalities met the minimum criteria for assuming social assistance responsibilities ( ARRETCHE, 1999: , p. 120); the requirements were to establish a local council and a local fund for social assistance, and to draft a social assistance plan (Law Nº 8,742/1993, article 30).
• 3
  “[…] federal funding started to operate through regular and automatic transfers, but it was subject to the implementation of nationally typified offers. The regularity of fund-to-fund transfers signaled security and financial predictability to the municipalities, encouraging their engagement in the provision of services within Suas” (MESQUITA, PAIVA and JACCOUD, 2020, p. 207; our translation from Portuguese).
• 4
  A PSAP corresponds to a combination of a CNPJ and municipality.
• 5
  Only 9,763 (51%) PSAPs had completed registration in the CNEAS. Complete registration in the CNEAS is required for granting SUAS funds to a PSAP (Resolution CNAS 21/2016; Ordinance of the Ministry of Social and Agrarian Development 130/2017). We believe that the incomplete registration status does not affect our analyses. For our purposes, the main variables drawn from the CNEAS dataset are CNPJ and level of social protection offered. No organization in CNEAS is missing its CNPJ; the percentages of missing data for level of social protection are similar across subsamples defined by registration status.
• 6
  17,952 out of 19,159.
• 7
  POP Centers, Day Centers, and Foster Family Units are far less frequent than other public facilities (see Figure 02 and ^{Table A01}). For this reason, they are not included in the correlations or in the regression analyses.
• 8
  This dataset was also used for estimating the regressions, with n = 10,593. Strictly speaking, these are pooled cross-section data; for the sake of simplicity, we will use the term “panel” data.
• 9
  States and municipalities may develop their own conditional cash transfer programs; due to the lack of a systematic data source, the model cannot incorporate those transfers.
• 10
  Admittedly, some covariates in N and R (e.g., percentage of poor and percentage of unemployed) could be interpreted as both a measure of needs and of resources, as those two concepts negatively correlate and, to a certain extent, are the inverse expression of one another. Here, we grouped individual and family characteristics in N (needs) and municipality characteristics in R (resources); like all other variables, these two groups are presented for the municipality level. As neither N nor R is the focus of our analyses – both were included as controls –, no efforts to further define these groups of variables seemed warranted.
• 11
  The Poisson distribution is defined by a single parameter, μ , corresponding to both the mean and the variance – a property known as equidispersion. Many count variables, however, exhibit variance greater than their average (overdispersion). Models based on the negative binomial distribution relax the premise of equidispersion.
• 12
  Of the 5,527 municipalities in the regression sample, 5,388 had at least one CRAS in May 2017. Mean CRAS-years in the latter group is 9.51.
• Data replication: BPSR is truly committed to open science and makes available the databases on the journal’s website and the Harvard Dataverse. An important part of our data sharing policy includes the checking of metadata and databases, with the aim of ensuring transparency and the replicability of the results presented in articles. Concerning to this article in particular, the databases were not checked due to a clause in a statement of responsibility signed by the authors with the Ministry of Social and Agrarian Development. The statement of responsibility can be found at the end of the article.

Appendix

Table A01 Mean of selected variables in the regression sample

	Regression sample	t = 2010	t = 2017
PSAPs created (between t-1 and t )
PSAPs created: any level of protectiona	0.48	0.89	0.08
PSAPs created: basic protectiona	0.32	0.59	0.05
PSAPs created: medium complexitya	0.08	0.15	0.01
PSAPs created: high complexitya	0.28	0.51	0.04
State presence in social assistance
CRAS-years in municipality, in t b	8.49	3.61	13.33
Community Center-years in municipality, in t b	1.44	0.37	2.50
CREAS-years in municipality, in t b	1.93	0.52	3.32
POP Center-years in municipality, in t b	0.09	0.00	0.18
Day Center-years in municipality, in t b	0.04	0.01	0.07
Foster Unit-years in municipality, in t b	1.06	0.38	1.74
Foster Famility Unit-years in municipality, in t b	0.11	0.03	0.19
Elderly (65 or older) receiving BPC-Elderly, in t-1 (%)c, d	6.73	5.49	7.95
People with disabilities receiving BPC-Disabilities, in t-1 (%)d,e	3.94	4.32	3.55
Programa Bolsa Família's coverage: poor households in the 2000 census, in t-1 (%)d	82.52	63.13	101.73
Needs (in t-1 )
Gini index for household income p.c. inequality (0-100)c	52.05	54.64	49.48
Poor (% of population)c	31.65	40.38	22.98
Unemployed (%)c	7.92	9.61	6.24
Children vulnerable to poverty: 14 or younger (%)c	66.63	74.51	58.82
Elderly: 65 or older (%)c	7.49	6.55	8.41
People with disabilities (%)e	20.44	16.32	24.52
Resources (in t-1 )
Population (thousand)c	33.10	31.16	35.03
Ln Population (thousand)c	2.50	2.46	2.55
Population density (population/km2)c, f	105.79	99.86	111.66
Ln Population density (population/km2)c, f	3.21	3.19	3.24
Rural population (%)c	38.14	40.88	35.41
Mean investment p.c. by local government (BRL)g	240.03	186.62	292.99
Ln Mean investment p.c. by local government (BRL)g	5.19	4.95	5.44
Own revenue (%)h	11.65	10.32	12.97
State capitalf	0.00	0.00	0.00
PSAP density (i.e., stock in t-1 )
PSAPs: any level of protectiona	2.49	2.05	2.93
PSAPs: basic protectiona	1.59	1.30	1.88
PSAPs: medium complexitya	0.56	0.48	0.63
PSAPs: high complexitya	1.42	1.16	1.68
t = 2017	0.50	0.00	1.00
Observations	10,593	5,274	5,319

Source: Created by the authors, with the raw data from the following sources:^aMDS - Cadastro Nacional de Entidades de Assistência Social (CNEAS); extracted on June 1^st, 2017.^bMDS - Censo SUAS 2017: <https://aplicacoes.mds.gov.br/sagirmps/portal-censo/>.^cAtlas do Desenvolvimento Humano no Brasil: <http://www.atlasbrasil.org.br>; poverty line = BRL 140.00 p.c.; vulnerability to poverty line = BRL 255.00 p.c.; monetary values as of August 2010.^dMDS - Matriz de Informação Social. For t = 2010, t-1 = 2004; for t = 2017, t-1 = 2010. Because the measure of PBF’s reach suffered methodological modifications throughout the years, its face value comparability is not guaranteed. Still, cross-municipality comparisons are warranted. <http://aplicacoes.mds.gov.br/sagi-data/misocial/tabelas/mi_social.php>.^eIBGE - Censo Demográfico 2000 (Resultados da Amostra): Table 2112; Censo Demográfico 2010 (Resultados da Amostra): Table 1495. Because this item of the census suffered modifications between 2000 and 2010, its face value comparability is not guaranteed. Still, cross-municipality comparisons are warranted.^fDatasus - Tabela de Municípios: <http://datasus.saude.gov.br/noticias/atualizacoes/59-sistemas-e-aplicativos/cadastros-nacionais/313-municipio>.^gIpeadata: <http://www.ipeadata.gov.br>; monetary values in BRL of December 2017. For t = 2010, mean over 1996 and 2000; for t = 2017, mean over 2006 and 2010.^hFinbra - Finanças do Brasil: <http://www.tesouro.fazenda.gov.br/pt_PT/contas-anuais>.

Note: Unless otherwise noted, for t = 2010, t-1 = 2000; for t = 2017, t-1 = 2010.

Table A02 Mean of selected variables in the regression sample, by region

	Brazil	Central-West	North	North-east	South	South-east
PSAPs created (between t-1 and t )
PSAPs created: any level of protection	0.48	0.39	0.35	0.35	0.35	0.76
PSAPs created: basic protection	0.32	0.30	0.26	0.23	0.24	0.49
PSAPs created: medium complexity	0.08	0.10	0.11	0.06	0.05	0.12
PSAPs created: high complexity	0.28	0.21	0.24	0.25	0.18	0.40
State presence in social assistance
CRAS-years in municipality, in t	8.49	7.81	8.49	9.46	6.36	9.16
Community Center-years in municipality, in t	1.44	1.87	0.77	1.61	1.21	1.46
CREAS-years in municipality, in t	1.93	2.52	2.34	2.29	1.35	1.69
POP Center-years in municipality, in t	0.09	0.05	0.05	0.07	0.08	0.14
Day Center-years in municipality, in t	0.04	0.00	0.01	0.02	0.03	0.09
Foster Unit-years in municipality, in t	1.06	1.51	1.16	0.53	1.22	1.37
Foster Famility Unit-years in municipality, in t	0.11	0.04	0.06	0.02	0.26	0.14
Elderly (65 or older) receiving BPC-Elderly, in t-1 (%)	6.73	14.75	11.38	6.22	3.48	6.25
People with disabilities receiving BPC-Disabilities, in t-1 (%)	3.94	4.62	4.14	4.34	2.80	4.08
Programa Bolsa Família's coverage: poor households in the 2000 census, in t-1 (%)	82.52	76.47	69.69	83.50	88.64	81.99
Needs (in t-1 )
Gini index for household income p.c. inequality (0-100)	52.05	52.77	58.22	54.41	49.34	49.79
Poor (% of population)	31.65	21.15	45.75	52.75	16.49	19.73
Unemployed (%)	7.92	7.55	8.85	9.40	5.09	8.21
Children vulnerable to poverty: 14 or younger (%)	66.63	59.13	77.89	86.25	49.76	57.32
Elderly: 65 or older (%)	7.49	6.15	4.65	7.36	8.30	8.10
People with disabilities (%)	20.44	19.66	18.55	22.59	20.13	19.09
Resources (in t-1 )
Population (thousand)	33.10	23.93	33.97	29.19	22.94	46.41
Ln Population (thousand)	2.50	2.30	2.63	2.67	2.22	2.56
Population density (population/km2)	105.79	26.96	21.81	88.52	76.66	185.14
Ln Population density (population/km2)	3.21	1.82	1.47	3.46	3.42	3.61
Rural population (%)	38.14	29.01	45.41	47.16	41.38	27.22
Mean investment p.c. by local government (BRL)	240.03	281.58	243.88	157.06	295.32	275.36
Ln Mean investment p.c. by local government (BRL)	5.19	5.42	5.20	4.79	5.42	5.39
Own revenue (%)	11.65	11.85	7.79	5.84	16.32	15.30
State capital	0.00	0.01	0.02	0.01	0.00	0.00
PSAP density (i.e., stock in t-1 )
PSAPs: any level of protection	2.49	1.76	1.13	1.32	2.34	4.33
PSAPs: basic protection	1.59	1.24	0.87	0.90	1.42	2.70
PSAPs: medium complexity	0.56	0.51	0.31	0.22	0.63	0.93
PSAPs: high complexity	1.42	1.06	0.73	0.88	1.30	2.33
t = 2017	0.50	0.51	0.53	0.50	0.49	0.50
Observations	10,593	868	798	3,380	2,271	3,276

Source: Created by the authors. Refer to Table A01 (Appendix) for sources of raw data. Table A03 Estimated coefficients for regression models of PSAPs created in a municipality

Dependent variable	PSAPs created between ( t-1 and t )
Independent variable	Any level of social protection	Basic social protection	Medium complexity	High complexity
CRAS-years in municipality, in t	-0.007***	-0.007**	-0.009	-0.009**
	(0.002)	(0.003)	(0.006)	(0.004)
Community Center-years in municipality, in t	0.002	0.002	-0.017	-0.002
	(0.003)	(0.003)	(0.011)	(0.004)
CREAS-years in municipality, in t	-0.002	-0.004	0.010	-0.004
	(0.006)	(0.007)	(0.013)	(0.009)
Foster Unit-years in municipality, in t	-0.004	0.000	0.005	-0.026***
	(0.003)	(0.004)	(0.006)	(0.009)
Elderly (65 or older) receiving BPC-Elderly, in t-1 (%)	0.009**	0.011**	0.011	0.018***
	(0.004)	(0.005)	(0.008)	(0.005)
People with disabilities receiving BPC-Disabilities, in t-1 (%)	0.021**	0.012	0.012	0.018
	(0.010)	(0.012)	(0.017)	(0.012)
Programa Bolsa Família's coverage: poor households in the 2000 census, in t-1 (%)	0.003**	0.002	-0.004**	0.002*
	(0.001)	(0.001)	(0.002)	(0.001)
Needs (in t-1 )
Gini index for household income p.c. inequality (0-100)	0.002	0.004	0.007	0.001
	(0.005)	(0.006)	(0.009)	(0.006)
Poor (% of population)	-0.029***	-0.031***	-0.023***	-0.026***
	(0.004)	(0.004)	(0.006)	(0.004)
Unemployed (%)	-0.009	-0.007	-0.019	-0.011
	(0.007)	(0.008)	(0.012)	(0.008)
Children vulnerable to poverty: 14 or younger (%)	0.021***	0.021***	0.021***	0.025***
	(0.004)	(0.005)	(0.007)	(0.005)
Elderly: 65 or older (%)	0.005	-0.030	0.029	0.006
	(0.017)	(0.020)	(0.031)	(0.021)
People with disabilities (%)	0.003	0.010	-0.005	0.011
	(0.008)	(0.009)	(0.014)	(0.010)
Resources (in t-1 )
Ln Population (thousands)
Ln Population density (population/km2)	-0.045*	-0.032	-0.102**	-0.050*
	(0.024)	(0.027)	(0.042)	(0.029)
Rural population (%)	-0.004*	-0.003	-0.004	-0.003
	(0.002)	(0.002)	(0.004)	(0.003)
Ln Mean investment p.c. by local government (BRL)	0.100**	0.102**	0.150**	0.052
	(0.040)	(0.046)	(0.074)	(0.050)
Own revenue (%)	0.003	-0.000	-0.004	0.001
	(0.003)	(0.003)	(0.006)	(0.004)
State capital	-1.196***	-0.936***	-1.464***	-1.584***
	(0.244)	(0.263)	(0.459)	(0.329)
PSAP density (i.e., stock in t-1 )
PSAPs: any level of protection	0.021***	0.018***	0.011***	0.027***
	(0.004)	(0.004)	(0.003)	(0.005)
Year of t
t = 2017	-2.626***	-2.658***	-2.657***	-2.663***
	(0.117)	(0.138)	(0.225)	(0.146)
Constant	-3.556***	-3.292***	-6.690***	-3.832***
	(0.477)	(0.561)	(0.835)	(0.590)
Observations	10,593	10,593	10,593	10,593
Number of groups (municipalities)	5,527	5,527	5,527	5,527
Reported model (lowest BIC)	Negative binomial	Negative binomial	Poisson	Negative binomial

Source: Created by the authors. Refer to Table A01 (Appendix) for sources of raw data.

Notes: Mean Variance Inflation Factor (VIF) = 2.72; variable percentage of children vulnerable to poverty (14 or younger) presents the largest VIF: 7.76. All specifications have random intercepts for municipality. Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10.

Figure A01
Estimated coefficients for regression models of PSAPs created in a municipality

Notes: Markers indicate point estimates with lines for 95% confidence intervals. All 24 specifications have random intercepts for municipality and the complete set of covariates. They are Poisson models, except for the following six, which are negative binomial: Brazil any level of protection, Brazil basic protection, Brazil high complexity, North any level of protection, North high complexity, Southeast any level of protection.

Edited by

Revised by Karin Blikstad

Publication Dates

• Publication in this collection
  23 Sept 2022
• Date of issue
  2022

History

• Received
  30 Nov 2020
• Accepted
  31 Jan 2022