Partisan alignment and requests for federal transfers in Brazil

Abstract

This article analyzes how subnational governments request transfers from the federal government in Brazil. Using microdata on applications for discretionary transfers from the Brazilian federal government to municipalities between 2009 and 2016 and a regression discontinuity design (RDD), I show that mayors affiliated with the president’s party demand substantially more resources than opposition mayors, meaning that the partisan alignment is an important channel to request transfers. On the other hand, the effect varies among mayors from other coalition parties. Thus, the results show that partisan alignment between different levels of government is a key factor in explaining the requests for discretionary resources in Brazil.

Keywords:
distributive politics; government coalitions; discretionary transfers

Resumo

Este artigo investiga como governos subnacionais solicitam transferências federais junto ao governo central no Brasil. Usando microdados sobre os pedidos de transferências voluntárias dos municípios à União entre 2009 e 2016 e um desenho de regressão descontínua, mostra-se que prefeitos do partido que ocupa a Presidência da República requerem muito mais recursos do que prefeitos oposicionistas - indicando que o alinhamento partidário constitui canal importante para solicitá-los. Por outro lado, esse efeito é heterogêneo entre prefeitos de partidos que integram a coalizão de governo federal. Desse modo, os resultados mostram que o alinhamento partidário entre diferentes níveis de governo importa para explicar a demanda por recursos discricionários no Brasil.

Palavras-chave:
política distributiva; governo de coalizão; transferências voluntárias

Resumen

Este artículo investiga cómo gobiernos subnacionales en Brasil solicitan transferencias federales al gobierno central. A partir de microdatos sobre las solicitudes de transferencias discrecionales de los municipios brasileños al gobierno central (2009-2016) y de un diseño de regresión discontinua (RD) se demuestra que los alcaldes del partido que ocupa la presidencia requieren sustancialmente más recursos que los alcaldes opositores, lo que indica que la alineación partidaria es un canal importante para solicitarlos. Por otro lado, este efecto es heterogéneo entre los alcaldes de partidos que integran la coalición de gobierno a nivel federal. Los resultados demuestran que la alineación partidaria entre diferentes niveles de gobierno es importante para explicar la demanda de recursos discrecionales en Brasil.

Palabras clave:
política distributiva; gobierno de coalición; transferencias discrecionales

1. Introduction

I have the impression that being the opposition makes it a little bit more difficult [to be successful in obtaining intergovernmental transfers]. The dialog between people from the same party is easier, you ask more, you get more.

(Fuad Gabriel Chucre [Former mayor of the city Carapicuíba, in São Paulo, Brazil, from the party PSDB] as quoted in Bragon & Nublat, 2007)

The statement of the former mayor above summarizes one of the most recurrent findings observed in the literature on distributive politics in Brazil: that the federal government distributes more resources to regions governed by allied parties, and less to regions governed by the opposition at the national level (Amorim & Simonassi, 2013; Bandeira-de-Mello, 2016; Brollo & Nannicini, 2012; Ferreira & Bugarin, 2007; Sakurai & Theodoro, 2014). And while the literature offers lots of documentation on this allocative bias, research on the logistics of the political distribution of resources is scarce.

Much of the discretionary resources in Brazil are distributed through Transferências Voluntárias da União (TVU) (federal government’s discretionary transfers), which are released under request by subnational governments through detailed proposals and investment plans. This condition limits the federal government’s distributive strategies because if an allied subnational government does not submit proposals, it will not be able to receive transfers. Despite being a strong element in intergovernmental transfers, few studies investigate the strategies that subnational governments adopt to request discretionary transfers - which are essential sources of revenue for the Brazilian municipalities (Brollo & Nannicini, 2012; Sakurai & Theodoro, 2014).

This article investigates whether the partisan alignment influences the intergovernmental transfers between federal and municipalities governments in Brazil. The study makes use of microdata of TVU requests (primarily agreements and contracts) submitted by all Brazilian municipalities to the federal government between 2009 and 2016. These data allow analyzing the hypothesis that the requests of the municipal governments are influenced by the local governments’ alignment with the federal government, a circumstance that would encourage municipalities to submit more or fewer proposals.

Taking advantage of cases where municipal governments were elected in coinflip elections, I employ a regression discontinuity design (RDD) to identify the causal effect of the partisan alignment between local and federal governments on the number of proposals requesting intergovernmental discretionary transfers, and on the amount requested in these proposals (Lee, 2008; Imbens & Lemieux, 2008). The main strength of this design is to allow the use of few assumptions to identify the causal effect of such alignment with the federal government, bypassing potential sources of endogeneity.

The results of this empirical exercise show that the partisan alignment between levels of government has a significant influence on the Brazilian municipalities’ request for federal transfers. In particular, partisan alignment between the mayor and the president causes an increase of about 16 grant proposals submitted by municipalities to the federal government, and this effect is also observed regarding the volume of resources requested. On the other hand, among all the parties that form the political coalition ruling the federal government, it was not possible to see the same effect, although a complementary analysis shows that there is variation in the municipalities’ fundraising strategies. These findings are not sensitive to the use of RDD and the removal of transfers related to agreements with the federal government oriented under the discretion of the members of the Congress instead of the executive branch.

Thus, my results stress the importance of party alignment when it comes to requesting federal transfers. Moreover, my findings also suggest that part of the allocative bias detected by the existing literature may result from a selection bias in the submission of proposals by the municipalities, and not entirely by interference of the executive branch. Thus, if mayors aligned with the federal coalition in government - in terms of partisan belonging - submit more requests, they increase their chances of having their proposals approved even if the executive carried out a random selection. Investigating the reasons that induce the behavior of submitting more proposals, therefore, is central to advancing the understanding of discretionary intergovernmental transfers in Brazil.

2. Distributive politics and partisan alignment

The majority of studies on distributive politics in countries with decentralized fiscal federalism assume that the national government has limited resources and that these resources are valuable electoral assets because they can be used in the implementation of local public policies serving their constituents (Bracco, Lockwood, Porcelli, & Redoano, 2015; Bueno, 2017). In this sense, the decision on resource allocation turns into an electoral decision for the central government, considering that subnational authorities may obtain electoral advantages when they channel resources and are at risk of not being re-elected otherwise (Arulampalam, Dasgupta, Dhillon, & Dutta, 2009). As a consequence, national executives have incentives to adopt allocative strategies that maximize the future electoral performance of their party allies or coalition throughout the territory.

In a simple case, where the choice to distribute resources is limited to two possibilities - to allocate resources to politically aligned subnational governments or to other subnational governments that are aligned to the national government’s opposition - the assumption is that the best strategy for the central government is to channel more resources to its allies (Arulampalam et al., 2009; Bracco et al., 2015; Brollo & Nannicini, 2012). First, because this prevents the local opposition from misappropriating credit from the application of these resources (Bueno, 2017). Second, because in doing so the central government provides resources for securing and expanding its constituents, promoting the electoral performance of the party or coalition (Arulampalam et al., 2009).

In this way, the models predict that central governments would differentiate the regions governed by allies and by oppositionists in the transfer of resources, producing an allocative bias. In fact, two decades of studies on the subject in various contexts are consistent with this prediction (Golden & Min, 2013). In Brazil, the literature points towards the same direction. According to Brollo and Nannicini (2012) and Ferreira and Bugarin (2007), to mention two examples, in the distribution of discretionary resources, the federal government favors municipal allies of the same party or parties that form the national coalition in power, and this is evident on the eve of elections, corroborating the model of political-budget cycles (see also Soares & Melo, 2016). The same occurs in the transfers to state governments (Amorim & Simonassi, 2013).

Much less investigated, on the other hand, is the role of subnational governments in the distribution of discretionary resources. If the results above are correct, to belong to the party aligned to the federal government means to access more resources, better ways to pressure for disbursement, to have a higher capacity for investment, and credibility with constituents (Brollo & Nannicini, 2012; Bueno, 2017; Ferreira & Bugarin, 2007). Also, if subnational governments anticipate this, it is possible to hypothesize that they will have more or fewer incentives to apply for federal transfers according to their degree of alignment. On the one hand, if a mayor anticipates that they will not have resources released by the federal government, the job of submitting applications is worthless; on the contrary, if the same mayor knows that they will have their demands met, they have the incentive to request more resources.

This logic is not new in the literature. Goldstein and You (2017), for example, show that partisan misalignment between the municipal and state levels encourages US municipal governments to lobby the federal government, as this would be a natural way out of having to submit their requests to the states that, being opposition, would be more likely to deny resources. A similar argument can be seen in the literature on fiscal federalism, which suggests that partisan alignment between different levels of government may encourage subnational governments to preferentially look for support from one governmental entity in detriment of others (Cammisa, 1995).

The central point of the argument lies in the unequal distribution of preferences and bargaining power between entities at different levels of government (Rodden, 2010). While subnational governments that have more chances to obtain concession from the federal government (because of a significant representation at the national Congress (Amorim & Simonassi, 2013), or capacity to threaten the economy and the territorial unity (Treisman, 1996), the degree of political alignment between levels (as mentioned by the former mayor in the epigraph that illustrates this article), can facilitate negotiations and access to bureaucracy via intraparty channels. Thus, the central government’s electoral incentives and factors related to the demand for intergovernmental transfer, are part of the strategies available to subnational governments to raise funds. Some of these strategies are: lobbying the central government (Goldstein & You, 2017), requesting support from members of Congress to intervene with the federal bureaucracy (Bueno, 2017), and investing more or fewer resources and time to send more proposals, or to request more substantial amounts from the federal government, which is the specific strategy analyzed in this work.

The literature on the subject in Brazil, on the other hand, takes the budgetary prerogatives of the presidents as sufficient factors to explain the allocation bias. As is well known, during the process of elaboration (a task carried out in the Federal Budget Office (SOF)) to the approval (by Congress) of the Annual Budget Law, the federal government has means to impose its preferences in the commitment and execution of federal resources (Figueiredo & Limongi, 2008). In any case, this context does not imply that subnational governments will be passive. It is reasonable to assume that a mayor who knows that their request will be accepted will invest more in fundraising, while another who knows that their requests are likely to be refused will not waste their energy in submitting many proposals.

3. Discretionary transfers in Brazil

A considerable portion of the intergovernmental discretionary transfers in Brazil are made through contract agreements (called convênios), which are celebrated around proposals submitted by the subnational governments and evaluated and accepted/refused by the federal bureaucracy. In this type of agreement, a federal government agency signs the agreement and it is responsible for transferring the grant, whereas the other part responsible to execute the object of the agreement is a subnational government or non-profit organization. In addition, the subnational government or non-profit usually participates with some resources as a counterpart, (financial or logistics) according to provisions of federal legislation, especially Decree N^o 6170/2007 (2007) and Ordinance n^o 507 (2011).¹

Mainly because of the constitutional obligation that municipalities have to spend part of their budget on health and education, these transfers are one of the primary sources of non-earmarked resources available to them (Brollo & Nannicini, 2012; Soares & Melo, 2016). More importantly, because these transfers’ disbursements are neither mandatory nor constitutional, the federal government can decide whether or not to invest, so these resources come and go, over time, according to the context or political factors such as the proximity of elections (Ferreira & Bugarin, 2007).

As for the practicalities, the process of signing these agreements follows three main steps. In the first, municipalities submit a proposal with a working plan describing what would be the object of the agreement, the resources needed, the schedule for the use of the resources, as well as other details. This process may be induced by a call for proposals launched by agencies of the federal government (Portaria Interministerial CGU/MP/MF n. 127, 2008), or by amendment to the budget which makes resources available for the object of the agreement (Brasil, 2009). In the second stage, the federal government agencies technically and legally receive and evaluate the project submitted and decide whether or not to sign the agreement - a step where pressures from members of Congress may occur to accelerate the commitment of the resources to transfer (Bueno, 2017). Finally, once signed, the release of resources from the federal government may begin, according to the schedule, followed by the delivery of monitoring reports from the subnational government. In case of irregularities, the federal bodies can start a parallel monitoring process, conducted by the Brazilian Audit Courts (Decree n. 6.170, 2007). Even with variations among the government agencies, which can implement different systems for receiving and evaluating proposals (especially when it comes to specific programs launched by the federal government), the process is similar for any agreement.

Graph 1
Proposals submitted by municipal to federal governments between 2009 and 2016, per level of alignment between the levels of government

Although the description provided above suggests that the requests for discretionary transfers are eminently bureaucratic processes, there are differences in the number of proposals submitted when comparing municipalities ruled by governments aligned with the federal government with municipalities ruled by parties of the opposition (see also Brollo & Nannicini, 2012). As observed in Graph 1, mayors at the beginning of their term in office between 2009 and 2014 submitted a higher number of proposals, and the same is shown for mayors of parties that are in the coalition that governs at the federal level. Of course, technical capabilities, such as the qualification of the municipal bureaucracy, influence the ability to write proposals to request resources through agreements (Soares & Melo, 2016). However, a pattern is visible of partisan alignment between subnational governments and the federal government in the submission of proposals.

When considering the number of inhabitants of each municipality, as shown in the right graph in Graph1, this relationship is no longer salient. At the end of the article, I show that the finding of the two graphs depends on the classification of the party alignment with the federal government and that the effect of the alignment is robust only for the party in the presidency.

4. Research design

Examining whether mayors’ party alignment with the federal government causes changes in their strategies of requesting proposals to obtain intergovernmental discretionary transfers implies the use of an appropriate identification strategy. In addition, information on the submission of proposals of agreements, including those not approved, is required to carry out this empirical exercise. The procedures used to comply with these requirements are as follows.

4.1 DATA

In response to a request from the Brazilian Federal Audit Court to avoid pending accountability and misuse of resources, as of 2007, the Decree N^o 6170/2007 (2007), provides that the entire process of managing agreements with the federal government has to be made using the Management System of Transfers Agreements and Contracts (Siconv). The system provides open data (Ministério do Planejamento, Desenvolvimento e Gestão, 2016) and makes it possible to identify all proposals of agreements submitted by municipal governments through individual registers, and also discloses the non-approved ones. This shows how many proposals each municipality has submitted, by mandate, in addition to the requested transfer in current values, discounting the counterparts. Also, the Siconv data distinguish proposals receiving transfers from the federal government in response to amendments to the budget requested by members of Congress. This differentiation made in the system is useful for this study, as it is possible to assess whether the results of the research are maintained when removing from the analysis this specific type of transfer that shows a protagonism of the Congress instead of the initiative of the subnational governments (Figueiredo & Limongi, 2008). Since Siconv was not fully adopted in 2008, the year of municipal elections in Brazil, the research sample was formed by proposals submitted from 2009 until 2016, resulting in 508,261 proposals (444,088, when excluding those intermediated by members of Congress as mentioned before) from 5,539 municipalities (an average of about 92 proposals per municipality).

The research analyzes information at the municipal level, and the most relevant are those related to the elections, which are necessary to identify the party that ruled each municipality in the two municipal mandates that took place during the years studied. The votes of the candidates for mayor in the 2008 and 2012 elections, as well as their party affiliations and the number of voters in each municipality, were extracted from the Electoral Data Repository of the Superior Electoral Court (Tribunal Superior Eleitoral[TSE], [n.d.]).

Table 1
Parties in the federal government coalition in the period, per number of chairs in the Chamber of Deputies

Coalition	Party	# of Chairs (at the beginning)	# of Chairs (at the end)
Lula (2nd mandate)	PMDB	91	89
	PT	82	77
	PR	41	43
	PP	42	40
	PSB	29	27
	PDT	23	23
	PTB	26	23
	PCdoB	13	12
	PRB	1	8
Rousseff (1st mandate)	PT	77	87
	PMDB	87	79
	PP	40	38
	PR	43	35
	PSB	27	28
	PDT	23	27
	PCdoB	12	14

• Source: Elaborated by the author, based on the Database of the Legislative branch - Cebrap.

Also, tax information from municipalities was compiled from the National Treasury Secretariat (STN) (Tesouro Nacional, [n.d.]), such as the collection from Urban Land and Building Tax (IPTU) and the mandatory intergovernmental transfers from the Municipal Participation Fund (FPM). The Brazilian Institute of Geography and Statistics (IBGE, [n.d.]) collected information on the number of inhabitants of the municipalities in the period. Finally, the party composition of government coalitions in the period was taken from the Database of the Legislative branch - Cebrap (Núcleo de Estudos Comparados e Internacionais, [n.d.]) and are used to indicate whether or not the candidates for mayor belonged to parties that integrated the coalitions of government at the national level during their mandates. Specifically, the Workers’ Party (PT) was in the presidency throughout the analyzed period, with the Party of the Brazilian Democratic Movement (PMDB), the Progressive Party (PP) and the Party of the Republic (PR) as the major coalition partners. Table 1 shows the composition of coalition governments in the period studied.

The dependent variables used in the analysis were operationalized as (1) the number of proposals of agreements registered per municipality during the municipal mandate (2009-2012 and 2012-2016); and (2) total amount requested, which is equal to the total requested by the municipalities during the municipal mandate, deflated by the National Consumer Price Index (IPCA) of December 2016. Thus, the final research base counts 10,784 observations, corresponding to each municipality i observed in the election t; all information were merged using the IBGE and TSE muncipalities’ identification codes. In some of the estimated models, these variables were transformed (a) by weighting them by the number of inhabitants of each municipality; (b) calculating their logarithms; and (c) subtracting from them proposals related to amendments to the budget proposed by members of Congress, in order to test the robustness of the results. Below, I will show further that the findings are independent of these choices.

4.2. IDENTIFICATION STRATEGY

In order to identify the causal effect of the alignment between mayors and the federal government on the number of proposals submitted, I use a regression discontinuity design (RDD) adapted for close races (Brollo & Nannicini, 2012; Lee, 2008). In general, the purpose of this application is to compare municipalities that elected (treatment group) or not (control group) a mayor politically aligned with the federal government, by a few votes. If the difference in number of votes is small enough to be considered random, it is possible to consider that the differences between these municipalities are causal.

Taking the model of potential outcomes (Rubin, 2005), I am interested in knowing what would have happened to a municipality that elects a mayor belonging to a party that integrates the national government coalition, if the reality was different, i.e., if this elected mayor was not from the same coalition. Thus, I considered municipalities that had, in the mayoral elections of 2008 and 2012, two best-voted candidates, where one was part of the national coalition, and the other was aligned with the opposition parties at the national level. Operationally, it was calculated a margin of votes between these two candidates in an election for mayor in municipality i in election t with the following equation:

${margin}_{it}=\frac{{Votesforalignedcandidate}_{it}-{alignedcandidate}_{it}}{{Votesforalignedcandidate}_{it}+{Votesfornon-alignedcandidate}_{it}}$ (1)

The variable ‘margin of votes’, therefore, indicates the percentage of votes that a mayor who is part of the national coalition obtained more than the other candidate (or would need to obtain to be elected). If the margin of votes in a municipality is positive, a candidate aligned with the federal government was elected mayor having, as the closest rival, an opposition candidate; conversely, if the margin of votes is negative, the elected mayor belonged to an opposition party at the national level. Undetermined cases, when the voting margin is equal to 0 (i.e., the vote of the first two candidates is equal and, according to the TSE, the oldest candidate takes office), were removed from the sample, without prejudice to the design; where there was a second round, only the results of the second round were used.

Comparing only municipalities where this absolute margin is small, we can estimate the exogenous effect of partisan alignment on their strategies for requesting intergovernmental transfers. This means that in these municipalities, the election of a mayor aligned with the federal government was as similar as possible to a random result. Thus, it is possible to estimate the effect of the local average treatment of the alignment on a dependent variable according to the equation:

$ETML\mathrm{}=\mathrm{}E\left[{\gamma }_{it}\left(1\right)|{margin}_{it}=0\right]-E\left[{\gamma }_{it}\left(0\right)|{margin}_{it=}0\right]$ (2)

Where Y _it is the value that assumes the dependent variable in municipality i in the election t for the treatment and control groups. As the equation shows, a simple comparison of the difference between means of these groups when the margin of votes equals zero (or tends to be zero, with the RDD estimate) identifies the causal effect of the alignment between mayors and the government coalition at the federal level. In this case, the treatment group is composed of municipalities where the elected mayor was part of the national coalition, and their closest rival was from an oppositional party; and, for the control group, the opposite. Thus, when a municipality goes from having almost elected an oppositionist at the national level to almost electing a mayor aligned with the central government, this creates discontinuity that can be used to explore the exogenous effect of alignment with the federal government.

Building on other studies that applied this design in municipal elections in Brazil (Brollo & Nannicini, 2012), elections with two or three candidates were analyzed because, with more candidates, the design tends to lose validity. In the end, the sample used contains 4,234 municipalities in the 2008 and 2012 mayoral elections, disputed between the two candidates with more votes, from parties aligned or not with the federal government. Also, another sample of 742 municipalities was analyzed, which included only those where the elections had competing candidates from the Workers’ Party (PT) - party in the presidency in the period researched - and from another party that was the opposition at the national level. Table 2 shows the descriptive statistics of the municipalities in these samples, comparing them with the complete sample.

Table 2
Descriptive statistics of the samples

Variable	OLS	RDD
	Complete Sample	Sample PT	Sample Coalition
Quota FPM (mean)	14,258,565.00	12,637,570.00	10,813,870.00
Quota FPM (deviation)	25,685,818.00	10,060,550.00	7,538,176.00
HH (mean)	0.48	0.50	0.51
HH (deviation)	0.10	0.10	0.09
IPTU (mean)	4,051,923.00	1,628,878.00	707,565.90
IPTU (deviation)	79,756,475.00	6,653,695.00	3,899,222.00
Margin of votes (mean)	0.16	0	0.02
Margin of votes (deviation)	0.17	0.26	0.24
No voters (mean)	24,300.61	17,353.34	12,324.29
No voters (deviation)	148,299.50	34,708.62	20,058.90
Submitted requests (mean)	38.76	38.05	39.63
Submitted requests (deviation)	39.90	35.71	38.45
PIB (mean)	895,976.50	598,432.90	313,717.70
PIB (deviation)	9,440,101.00	2,513,548.00	1,015,794.00
No inhabitants (mean)	34,386.62	24,319.73	16,915.80
No inhabitants (deviation)	203,141.70	51,855.26	29,497.54
Amount requested (mean)	45,607,453.00	33,276,014.00	38,070,997.00
Amount requested (deviation)	213,357,309.00	54,012,779.00	89,615,379.00
Observation	10,784	742	3,492

• Source: Elaborated by the author, based on data from Siconv, TSE, IBGE, and Cebrap.

The conventional practice of the literature was followed to estimate the effect of the alignment (Calonico, Cattaneo, & Titiunik, 2014; Imbens & Lemieux, 2008), using ordinary least squares (OLS) models:

$Y_{it}=\alpha +{\beta T}_{it}+{\theta margin}_{it}+{\delta T}_{it}{margin}_{it}+{\in }_{it}$ (3)

where Y _it is the dependent variable, T _it is a dummy that indicates whether or not the elected mayor belongs to parties in the national government coalition, and margin _it is the variable with the margin of votes. Specifically, the coefficient of T _it , β, is the estimator of the mean causal effect in the discontinuity, indicating the exogenous effect of the alignment with the federal government on the municipalities’ strategy of request for intergovernmental transfer. Also following the current practice in the literature, the study adopts a scheme of triangular weights in the models, which emphasizes the observations closer to the discontinuity and that decreases symmetrically as this proximity decreases. All models were estimated using the method proposed by Calonico et al. (2014), which calculates robust confidence intervals for the estimates, as well as automatically selecting the bandwidth (i.e., selecting only candidates whose margins are small enough to be considered as almost random) automatically.

4.3 VALIDATION AND ROBUSTENESS

In order for a RDD to be valid, it must meet the following assumption: that mayors cannot determine their votes. If that happens, the result of the election, and also the degree of alignment between the mayor and central government, cannot be considered almost random, indicating that some mayors manipulate the results of the elections. Eggers, Fowler, Hainmuelle, Hall e Snyder (2015) examine this possibility in the mayoral elections in Brazil and find no evidence to support that there is manipulation. The procedure proposed by McCrary (2008) to test the alternative hypothesis that there is a difference in the density of the variable ‘margin of votes’ around the discontinuity was applied. The results of this test, shown in Graph 2, support the validity of the design - in the same way as observed in other studies that focus on municipal elections in Brazil (Brollo & Nannicini, 2012; Bueno, 2017).

Graph 2
Results of McCrary test

As in experimental designs, the expectation is that treatment and control groups are similar in aspects prior to the application of the treatment. Using the specification described above, I show in the annexes that this applies to the municipalities of the sample: they are similar in several socioeconomic characteristics, such as in an experimental design.²

5. Results

This section presents the results of the effect of party alignment between municipal and federal governments on the number of requests and the volume of discretionary transfers proposed by municipalities. As observed in Table 3, this effect is substantive and significant for mayors who belong to the party in the presidency, although not for the sample that includes mayors of parties that are part of the coalition. The coefficients in Table 3 evidence the local mean causal effect, estimated with RDD (i.e., the difference between treatment and control groups), accompanied by the estimates standard errors and the absolute margin of votes used.

Table 3
Effect of the party alignment with the federal government on the municipalities’ proposals of agreements

	Dependent variable		Standard Errors	Bandwidth	P-value	N
PT	Number of proposals	16.261*	5.510	0.14	0	386
	Proposals per capita	0.002*	0	0.14	0.04	415
	Amount requested (log)	0.451*	0.220	0.15	0.04	422
	Amount per capita (log)	0.611*	0.230	0.15	0.01	420
Coalition	Number of proposals	-0.815	3.100	0.15	0.79	2,515
	Proposals per capita	0	0	0.13	0.52	2,328
	Amount requested (log)	0.129	0.100	0.15	0.18	2,583
	Amount per capita (log)	0.206*	0.100	0.15	0.04	2,608

• Source: Elaborated by the author. Note: * p-value < 0.05. Robust standard errors. Windows were selected according to the algorithm proposed by Calonico et al. (2014).

In all cases where the municipal government is formed after an election where the two leading candidates were one from the Worker’s Party (PT) and the other from any party that was opposed to the federal government, the effect of alignment is significant, causing an increase of about 16 proposals submitted, or about 0.002 proposals per capita. Although it seems little, in this sample the standard deviation for submitted proposals is 40.5, which means an increase of 40% of the standard deviations. For example, in a municipality of 10 thousand inhabitants, the alignment would cause an increase of 20 proposals (0.002 × 10.000) - an effect of magnitude similar to that of the additional amount of resources that municipalities governed by the party in the presidency effectively receive more than others (Brollo & Nannicini, 2012; Soares & Melo, 2016). As for the governments resulting from elections disputed between candidates from parties that form the coalition ruling the federal government and another candidate from opposition parties, the effect is practically null in three of the four models, being significant only in the last one.

Graph 3
Alignment with the party in the presidency and proposals of agreement with the federal government submitted by municipalities

These results can also be seen in Graph 3and 4. On the graphs on the right, the black dots represent the average of the dependent variable in intervals of 1.3% in the margin of votes, while the white smaller dots indicate the average of the same variable in intervals of 0.5% in the margin of votes. This is a procedure recommended in the literature that facilitates the visualization of discontinuities in the dependent variable (Imbens & Lemieux, 2008). Substantively, it is possible to see that the difference between electing a mayor of the opposition and one of PT, in Graph 3, is significant for the dependent variable: the vertical difference when the two curves touch the dashed line is accentuated and indicates that the causal effect of the alignment is positive on both graphs on the right. In Graph 4 that examines the sample with mayors from parties that are part of the federal coalition, the effect disappears, as indicated by the curves in the dashed line.

The left graphs, on the other hand, evidence the estimated causal effect and the 95% confidence interval in different bandwidth in the margin of votes (i.e. when the candidates have won or lost by margins of 1 to 20% of votes), which gives greater assurance that the result reported in the tables is not sensitive to this choice. In this sense, it is clear that belonging to the party in the presidency or parties that occupy ministries brings different incentives to request intergovernmental transfers, just as this affects the authorization of these transfers (Brollo & Nannicini, 2012; Ferreira & Bugarin, 2007).

Graph 4 Alignment
with parties that are part of the coalition ruling the federal government and proposals of agreement with the federal government submitted by municipalities

Being aligned to a party that integrates the governing coalition at the federal level, therefore, does not seem to incentivize a mayor to submit more proposals of agreements. This observation, however, does not eliminate the possibility that even submitting a similar number of proposals in comparison to mayors from parties of the opposition, the mayors from parties of the coalition may be requesting a more significant amount of resources. Also, it is worthwhile observing among the mayors from PT if they are indeed requesting greater amounts in their proposals. To test this possibility directly, Table 3 and Graph 3 and 4 register the results of the effect of alignment on the amount requested by municipal governments.

As with the number of proposals, the effect of alignment on the amount requested is only positive and significant for mayors from PT. Because the distribution of this variable is quite asymmetric, converting it into a logarithm makes the substantive interpretation of the effects more difficult, but some indicators help to assess their magnitude. The standard deviation of the logarithm of the total amount requested from the federal government by municipality is 1.17 for the PT sample, which shows that the effect of the alignment, in this case, is about 38% of a deviation; in the case of the logarithm of the total amount requested per capita, the effect reaches about 52% of a deviation.³

It is also necessary to understand whether the amendments that members of Congress are entitled to carry out when preparing the Federal Budget influences in the findings. As these amendments may result in agreements between the federal government and the municipalities, it is possible that this influences the results observed so far, which is something not specifically studied in previous works (e.g., Brollo & Nannicini, 2012; Soares & Melo, 2016). To examine this possibility, Table 4 replicates the results of Table 3 excluding the proposed dependent variables related to amendments. Despite the increased uncertainty in the estimates in two models in the PT sample (one due to a p-value of 0.09), estimates do not change substantively - which is particularly revealing considering the reduced ‘N’ and the use of different operations of the dependent variable.

Table 4
Effectof the party alignment with the federal government on the submission of proposals by the municipalities (excluded those related to amendments from members of Congress)

	Dependent variable		Standard-Error	Bandwidth	P-value	N
PT	# of proposals	16.321*	5.120	0.14	0	389
	Proposals per capita	0.001	0	0.13	0.09	378
	Amount requested (log)	0.179	0.360	0.13	0.62	382
	Amount per capita (log)	0.565*	0.270	0.15	0.04	427
Coalition	# of proposals	-4.917	3.350	0.15	0.14	2,192
	Proposals per capita	-0.001	0	0.13	0.37	1,937
	Amount requested (log)	0.132	0.150	0.15	0.37	2,170
	Amount per capita (log)	0.163	0.120	0.18	0.16	2,380

• Source: Elaborated by the author. Note: * p-value < 0.05. Robust standard errors. Windows were selected according to the algorithm proposed by Calonico et al. (2014).

In general, the results show that the effect of alignment with the party in the presidency on the number of proposals, and on the amount requested is robust, independent of the operation adopted in RDD or the choice of the dependent variable. This, on the one hand, corroborates other studies that suggest that partisan alignment (by encouraging the sharing of credit by public investments), affects the intergovernmental transfer of resources (Arulampalam et al., 2009). On the other hand, as discussed earlier, this also suggests that intraparty mechanisms can explain how this process is carried out, with an active part of subnational governments.

Because the result is different from the ones found for mayors belonging to other parties in the federal coalition, it is possible to conclude that there are intraparty mechanisms that are triggered only through the presidency, facilitating the articulation between the local and national party that rules the government. In order to explore this possibility, two complementary models were estimated by least squares using all the observations available on the database. These models include, as controls, other variables previously described, such as the collection of the municipal tax (IPTU) and the population size, as well as fixed-effects for years and states and robust cluster errors by states. Although this practical strategy is not adequate to identify the causal effect of the alignment, it is possible to examine whether there is a relationship between this alignment and the requests for intergovernmental transfers, using a broader base of variation.

An alignment indicator was created especially for these models. It consists of a categorical variable with the following categories: PT, PMDB, Partido Progressista (PP) (Progressive Party), PR, and PCdoB (Communist Party of Brazil), other coalition parties and ‘Opposition Parties’, which is the category reference of the variable. With this categorical variable in the models, the predictive effect of alignment can be disaggregated. The model used follows the formula:

$Y_{ist}\mathrm{}=\mathrm{}\alpha \mathrm{}+\mathrm{}{\beta Alignment}_{ist}\mathrm{}+\mathrm{}{X´}_{ist\gamma }\mathrm{}+\mathrm{}{\delta }_t\mathrm{}+\mathrm{}{\theta }_s\mathrm{}+{\in }_{ist}$ (4)

where δ_t represents fixed-effects for the years (the two elections), θ_s represents fixed-effects for states, X’ _ist is the set of control variables, and finally β returns the effect of party alignment of municipal with federal government. Table 5 shows the results of these models.

Table 5
Party alignment with the federal government and proposals submitted between 2009 and 2016

Dependent variable	#of agreements	Amount requested (log)
	(1)	(2)
Coalition	1.094	0.050
	(1.185)	(0.038)
PCdoB	1.683	0.233*
	(5.051)	(0.137)
PMDB	3.614***	0.136***
	(1.209)	(0.043)
PP	2.089**	0.011
	(1.024)	(0.042)
PR	1.331	0.087*
	(0.863)	(0.044)
PT	7.009***	0.210***
	(1.303)	(0.050)
GDP (log)	-0.070	0.055***
	(0.830)	(0.019)
IPTU (log)	0.710***	0.032***
	(0.246)	(0.010)
Population (log)	8.708***	0.423***
	(0.895)	(0.045)
Constant	-36.915***	12.093***
	(7.837)	(0.309)
Fixed-Effects, year	Yes	Yes
Fixed-Effects, state	Yes	Yes
N	9,511	9,511
R2 Adjusted	0.373	0.418
Residual Standard Error (df = 9475)	31.406	0.991

• Source: Elaborated by the author. Note: *p<0.1; **p<0.05; ***p<0.01. The coefficients were estimated by least squares with fixed effects for years (mandates) and states, not reported. Standard errors with cluster for states. The reference category for the indicator of alignment is Opposition Parties.

Although the number of observations falls due to missing data, the results are consistent with the previous ones: when the mayor belongs to PT, it is expected the submission of seven proposals of agreements more than when the mayor belongs to an opposition party (model 1), as well as a difference of around 21% in the amount requested in favor of the mayor from PT (model 2). However, it is important to note that these estimates cannot be interpreted as causal and therefore it is possible that the effect in the case of PT is being underestimated because when using RDD with a sample of tight electoral race, the effect is more than twice bigger.

These results help to understand the differences in intergovernmental transfer request between coalition members. It is observed that being mayor of the PMDB and, in some cases, of the PP, PR, and PC do B predicts an increase in the proposals and the amount requested by the municipalities, but the magnitude of this effect is considerably smaller. It may be the case that parties with greater bargaining power in the coalition, such as PMDB and PP (which occupied relevant ministries with great capillarity in the period, such as Ministry of National Integration, Ministry of Agriculture and Ministry of Cities), had mayors submitting more proposals of agreements, or requesting more transfers. This suggests that one of the channels of intraparty articulation to carry out demands goes via the ministries (Batista, 2015). In any case, it is possible to say that there is substantial variation in the fundraising strategy between parties that are part of the executive branch, and that, as estimates with RDD confirm, municipal governments ruled by the party in the presidency during the period studied (Workers’ Party, PT), were those who most actively requested for discretionary intergovernmental transfers.

Conclusion

There is vast literature showing that national governments have incentives to use discretionary resources to reward or punish specific regions, which can either be used to garner electoral or legislative support (Arulampalam et al., 2009; Bracco et al., 2015). How subnational governments request such resources, on the other hand, is much less studied. The objective of this article is to show the importance of this perspective. The results of the analysis of a period covering two municipal mandates show that mayors belonging to the party in the presidency submit more proposals of agreements involving intergovernmental transfer, also requesting more resources, when compared to mayors from parties that are opposition at the national level. Moreover, although it does not extend to all parties in the federal coalition, there is substantial variation between them.

Therefore, the evidence presented shows that the demand for public resources, in Brazil, varies according to political factors (Brollo & Nannicini, 2012; Ferreira & Bugarin, 2007; Sakurai & Theodoro, 2014). This phenomenon by itself has important implications for our understanding of how discretionary resources are distributed in the country.

Firstly, the findings suggest the existence of a selection bias in the submission of proposals, emphasizing the importance of analyzing not only the offer of discretionary resources by the central government, but also the proposals submitted by subnational governments, which represent the demand, and how these governments adapt strategically to the federal government’s distributive policy to expand their budget.

Second, the results raise issues on how central governments distribute resources in federalist countries. If there is a variation in the demand for resources, it is necessary to investigate how different levels of government coordinate the political distribution of resources, either by intraparty strategies, as suggested by the findings of this article, or institutionally (Batista, 2015; Golden & Min, 2013). An example of strategy is the protagonism of members of Congress, who can take information to mayors on how to access central government programs, or establish communication channels with ministers from the same party to prioritize proposals from mayors who support their mandates.

As the results obtained in this study speak only to the governments of the Brazilian Workers’ Party, it is essential to be cautious in making generalizations: the strategies of subnational governments to request resources may vary over time. Notwithstanding, the findings presented in this study may promote further research on fundraising strategies adopted by subnational entities to attract discretionary resources, whether replicating this analysis to proposals submitted by municipalities to state governments for previous periods based on other sources of data; or to clarify the factors that encourage local politicians to migrate to another party in order to be better positioned to attract intergovernmental transfers to their municipalities or regions (Novaes, 2018).

REFERÊNCIAS

• Amorim, O Neto; & Simonassi, A. G. (2013). Bases políticas das transferências intergovernamentais no Brasil (1985-2004). Revista de Economia Política, 33(4), 704-725.
• Arulampalam, W; Dasgupta, S; Dhillon, A; & Dutta, B. (2009). Electoral goals and center-State transfers: a theoretical model and empirical evidence from India. Journal of Development Economics, 88(1), 103-119.
• Bandeira-de-Mello, R. (2016). Betting on winner: the effect of local elections on corporate political activity outcomes (FGV EAESP - Textos para Discussão n. 4). São Paulo, SP: Fundação Getulio Vargas. Recuperado de http://bibliotecadigital.fgv.br/dspace/handle/10438/15122
  » http://bibliotecadigital.fgv.br/dspace/handle/10438/15122
• Batista, M. (2015). A conexão ministerial: governo de coalizão e viés partidário na alocação de emendas parlamentares ao orçamento (2004-2010). Revista Ibero-Americana de Estudos Legislativos, 4(1), 3-16.
• Bracco, E; Lockwood, B; Porcelli, F; & Redoano, M. (2015). Intergovernmental grants as signals and the alignment effect: theory and evidence. Journal of Public Economics, 123(1), 78-91.
• Brasil. (2009). Convênios e outros repasses(3a ed.). Brasília, DF: Tribunal de Contas da União.
• Brollo, F; & Nannicini, T. (2012). Tying your enemy’s hands in close races: the politics of federal transfers in Brazil. American Political Science Review, 106(4), 742-761.
• Bragon, R; & Nublat, J. (2007, 19 de novembro). Governo federal privilegia prefeituras da base aliada Recuperado dehttps://www1.folha.uol.com.br/fsp/brasil/fc1911200702.htm
  » https://www1.folha.uol.com.br/fsp/brasil/fc1911200702.htm
• Bueno, N. S. (2017). Bypassing your enemy: distributive politics, credit claiming, and non-State organizations in Brazil. Recuperado de http://journals.sagepub.com/doi/full/10.1177/0010414017710255
  » http://journals.sagepub.com/doi/full/10.1177/0010414017710255
• Calonico, S; Cattaneo, M. D; & Titiunik, R. (2014). Robust nonparametric confidence intervals for regression‐discontinuity designs. Econometrica, 82(6), 2295-2326.
• Cammisa, A. M. (1995). Governments as interest groups: intergovernmental lobbying and the federal system Westport, CT: Greenwood.
• DecretoDecreto n. 6.170, de 25 de julho de 2007 (2007). Dispõe sobre as normas relativas às transferências de recursos da União mediante convênios e contratos de repasse, e dá outras providências. Brasília, DF.
• Eggers, A. C; Fowler, A; Hainmueller, J; Hall, A. B; & Snyder, J. M Jr. (2015). On the validity of the regression discontinuity design for estimating electoral effects: new evidence from over 40,000 close races. American Journal of Political Science, 59(1), 259-274.
• Ferreira, I. F. S; & Bugarin, M. S. (2007). Transferências voluntárias e ciclo político-orçamentário no federalismo fiscal brasileiro. Revista Brasileira de Economia, 61(3), 271-300.
• Figueiredo, A; & Limongi, F. (2008). Política orçamentária no presidencialismo de coalizão Rio de Janeiro, RJ: Ed. FGV.
• Golden, M; & Min, B. (2013). Distributive politics around the world. Annual Review of Political Science, 16(1), 73-99.
• Goldstein, R; & You, H. Y. (2017). Cities as lobbyists. American Journal of Political Science, 61(4), 864-876.
• Imbens, G. W; & Lemieux, T. (2008). Regression discontinuity designs: a guide to practice. Journal of Econometrics, 142(2), 615-635.
• Instituto Brasileiro de Geografia e Estatística. [n.d.];. Estimativas de População Recuperado de http://www.ibge.gov.br/home/estatistica/populacao/12/2/2018
  » http://www.ibge.gov.br/home/estatistica/populacao/12/2/2018
• Lee, D. S. (2008). Randomized experiments from non-random selection in US House elections. Journal of Econometrics, 142(2), 675-697.
• McCrary, J. (2008). Manipulation of the running variable in the regression discontinuity design: a density test. Journal of Econometrics, 142(2), 698-714.
• Ministério do Planejamento, Desenvolvimento e Gestão. (2016, 9 de março). Portal de Convênios Recuperado de http://portal.convenios.gov.br/
  » http://portal.convenios.gov.br/
• Novaes, L. M. (2018). Disloyal brokers and weak parties. American Journal of Political Science, 62(1), 84-98.
• Núcleo de Estudos Comparados e Internacionais. [n.d.]. Banco de Dados Legislativos do Cebrap Recuperado dehttp://neci.fflch.usp.br/
  » http://neci.fflch.usp.br/
• Portaria Portaria Interministerial CGU/MP/MF n. 127, de 29 de maio de 2008 (2008). Estabelece normas para execução do disposto no Decreto n. 6.170, de 25 de julho de 2007, que dispõe sobre as normas relativas às transferências de recursos da União mediante convênios e contratos de repasse, e dá outras providências. Brasília, DF.
• Portaria Portaria Interministerial CGU/MP/MF n. 507, de 24 de novembro de 2011 (2011). Estabelece normas para execução do disposto no Decreto n. 6.170, que dispõe sobre as normas relativas às transferências de recursos da União mediante convênios e contratos de repasse. Brasília, DF.
• Rodden, J. (2010). Federalism and inter-regional redistribution. In N. Bosch, M. Espasa, & A. S. Olle (Eds.), The political economy of inter-regional fiscal flows: measurement, determinants and effects on country stability (pp. 191-219). Cheltenham, England: Edward Elgar.
• Rubin, D. B. (2005). Causal inference using potential outcomes: design, modeling, decisions. Journal of the American Statistical Association, 100(469), 322-331.
• Sakurai, S., & Theodoro, M. I. A. (2014Dezembro). Alinhamento político e transferências governamentais: novas evidências para os municípios brasileiros. In Anais do 42 ^o Encontro Nacional de Economia Natal, RN.
• Soares, M., & Melo, B. (2016). Condicionantes políticos e técnicos das transferências voluntárias da União aos municípios brasileiros. Revista de Administração Pública, 50(4), 539-562.
• Tesouro Nacional. [n.d.]. Contas Anuais Recuperado dehttp://www.tesouro.fazenda.gov.br/pt_PT/contas-anuais
  » http://www.tesouro.fazenda.gov.br/pt_PT/contas-anuais
• Treisman, D. (1996). The politics of intergovernmental transfers in post-Soviet Russia. British Journal of Political Science, 26(3), 299-335.
• Tribunal Superior Eleitoral. [n.d.]. Repositório de Dados Eleitorais Recuperado de http://www.tse.jus.br/eleicoes/estatisticas/repositorio-de-dados-eleitorais
  » http://www.tse.jus.br/eleicoes/estatisticas/repositorio-de-dados-eleitorais

• 1
  For a detailed description of the discretionary transfers, see Soares and Melo (2016). It is worth mentioning that regulation on the topic varied with time, although the use of Siconv - discussed later on in this article - has offered a standard way to submit proposals, which allows to better analyze the submission process.
• 2
  I also estimated some of the tested models using polynomials in the margin of votes (Imbens & Lemieux, 2008). The results may be provided if requested, and they do not differ substantially from the results shown in this article.
• 3
  As an example, the estimated effect of the alignment on the total amount requested without adopting logarithm is about R$ 17 million, with a minimum confidence interval of 95% of R$ 15 million.

• [Translated version] Note: All quotes in English translated by this article’s translator.

ANNEXES

Graph A
Causal effect of the alignment with parties of the coalition in pre-treatment variables

Graph B
Effect of the alignment with the party in the presidency on pre-treatment variables

Data availability

Data citations

Núcleo de Estudos Comparados e Internacionais. [n.d.]. Banco de Dados Legislativos do Cebrap Recuperado dehttp://neci.fflch.usp.br/

Publication Dates

• Publication in this collection
  Jan-Feb 2019

History

• Received
  06 Sept 2017
• Accepted
  07 May 2018