Bypassing the Enemy: Distributive Politics,Credit Claiming,and Nonstate Organizations in Brazil

Introduction

Presidents who spend money on government programs in the hopes of building and cultivating political support face a dilemma. Presidents need to rely on political foes to distribute resources to voters in places ruled by the opposition. Thus, presidents find themselves in the following predicament: Should they deprive voters of services to hurt their political opponents, or should they dispense resources to the opposition at the risk of benefiting their political foes?

The typical answer gleaned from studies of distributive politics in multitiered systems is that presidents ¹ favor copartisans over opposition mayors. Voters often have difficulty distinguishing between levels of government, and they do not accurately identify the federal government as the source of grants, which creates the risk of “credit hijacking.” By strategically allocating federal money to copartisans at lower levels, incumbents minimize the potential costs that come with spending resources in places ruled by an opposition incumbent (Arulampalam, Dasgupta, Dhillon, & Dutta, 2009; Brollo & Nannicini, 2012; Pop-Eleches & Pop-Eleches, 2012; Stokes, Dunning, Nazareno, & Brusco, 2013).

Yet shifting resources away from opposition-governed municipalities may backfire. Incumbent politicians may wish to extract political advantage in opposition municipalities—and they may care about the welfare of voters. How then can incumbents avoid the risk of credit hijacking when they expend resources in localities controlled by the opposition?

One answer, which I study in this article, is to channel funding through nonstate welfare providers. I argue that incumbents can strategically allocate resources to nonstate organizations to minimize the risk of credit hijacking by political foes and reduce the political costs of allotting goods to voters who live in towns and cities under opposition control. In clear contrast to mayors, providers are not organs of the state, and they do not need to amass voter support; they have fewer incentives to claim electoral credit for themselves. In this way, nonstate organizations provide a path for the central incumbent to provide goods to voters in opposition-ruled districts while bypassing the opposition incumbents.

In this article, I test these claims, using novel data sets as well as extensive quantitative and qualitative data. I present evidence from Brazil including fine-grained information about the allocation of transfer payments to nonstate organizations and mayors. I use a close-race regression-discontinuity design (RDD) to identify the effect of partisan mayoral alignment on transfers. This design yields two sets of municipalities that plausibly differ only in the presence or absence of an aligned mayor, that is, whether the mayor and the president are copartisans or not. My argument implies that the election of a nonaligned mayor in a locality causes the center to spend more on nonstate providers (NSPs) in that municipality, thereby bypassing the mayor.

My findings are consistent with the bypassing hypothesis: Among municipalities that elected mayors by a narrow margin, the central government channels more resources through nonstate welfare organizations in municipalities ruled by opposition mayors—those who do not belong to the president’s party—than through NSPs in municipalities governed by copartisan (aligned) mayors. Substantively, the effect of alignment is considerable for transfers to nonstate service providers. The average effect of alignment on transfers to nonstate organizations is similar in size to the mean per capita expenditure of the flagship social program Bolsa Família. I also find that partisan alignment causes an increase in intergovernmental transfers to mayors, which is consistent with previous studies (Berry, Burden, & Howell, 2010; Bracco, Lockwood, Porcelli, & Redoano, 2015; Brollo & Nannicini, 2012; Fouirnaies & Mutlu-Eren, 2015). Finally, using the same research design, I also show that both strategies take place at the level of the state: Governors reward copartisan mayors and circumvent opponent mayors by funding NSPs.

In addition to measuring the effect of alignment on transfers, I investigate the mechanisms that help to explain the political usefulness of nonstate organizations in circumventing opposition. I contend that channeling money through NSPs hinders local opponents’ ability to steal credit. I also show evidence of behavioral and institutional mechanisms that make it difficult for opponents to hijack credit. First, I use original survey data to illustrate that survey respondents are less supportive of mayors who govern municipalities in which nonstate organizations provide public services with federal money compared with mayors who administer the same services with federal funds. Second, I use qualitative evidence from over 70 interviews with managers of nonstate organizations and bureaucrats to establish that mayors do not control the flow of money between the federal government and NSPs. This means that institutional features stymie opposition mayors trying to use these subsidies to NSPs to their own advantage. Third, I put forward that providers with partisan leaders may offer complementary benefits related to party building: mobilizing organized civil groups, creating a presence in localities, and recruiting candidates.

Finally, I also address additional interpretations to the bypassing strategy. Instead of avoiding credit hijacking, central incumbents could be using corrupt nonstate welfare providers to funnel public funds to local copartisans or to target potential supporters in a clientelist fashion. I present evidence from a data set on contract irregularities between NSPs and the federal government, as well as qualitative data from visits and interviews to make the point that corruption and clientelism do not play a major part in this case.

This article makes several contributions to the scholarship on distributive politics. It identifies a little-noticed strategy that central incumbents implement to avoid credit seeping over to the opposition, especially in scenarios of low clarity of responsibility such as federal systems. It neatly uncovers electoral motives behind the distribution of public money to nonstate organizations that are crucial service providers in both developing and developed countries. Moreover, scholars of distributive politics have extensively examined allocation of grants, whereby many have discussed to what extent low clarity of responsibility may influence attribution of credit (Pop-Eleches & Pop-Eleches, 2012). I examine both incumbents’ strategies in distributing resources and voters’ attribution of credit in a single article. My findings also suggest that political authorities have electoral reasons for relinquishing control of service provision: Incumbents cede service provision as a way to cripple opposition. In addition to that, I underscore the ways in which NSPs are part of a political and partisan strategy, thereby contributing to recent literature on the political logic of nonstate welfare provision (Cammett, 2014).

In the rest of the article, I discuss the reasons why I expect central incumbents to sidestep local opponents through NSPs and the institutional setting in Brazil (“The Distributive Politics of Nonstate Welfare Provision” section). I present a natural experiment testing the effects of political alignment on the central incumbent’s allocation of public money to mayors and NSPs (“Testing the Bypassing Strategy” section). Finally, I examine the mechanisms through which nonstate welfare provision is useful for the central incumbent and explore additional explanations (“Are Opposition Mayors Unable to Hijack Credit?” and “Additional Interpretations” sections).

The Distributive Politics of Nonstate Welfare Provision

In many countries, responsibilities for provision of public goods are often split among different levels of government. But, elected and nonelected officials are not the only providers of public goods. Nonstate organizations are responsible for a significant share of public service provision. Most countries rely on some form of nonstate, often nonprofit, providers of public goods: religious and charitable organizations that make up part of the welfare system, private and mixed public–private companies that run key services such as water and electricity, military contractors who help to fight wars, among others. In developing countries, more than half of physicians work in clinics run by charitable organizations, which accounts for 28% of all hospital beds (Hanson & Berman, 1995).

Yet, far from being autonomous, and self-financed entities, these organizations often rely on public funds and subsidies. A comparative study by Salamon, Sokolowski, and List (2003) of 35 countries found that about 35% of all nonstate organizations’ revenue comes from domestic public sources. If only health and social service organizations are considered, the proportion of revenue that comes from the governments’ purses is even higher. Why do governments hand over authority and resources to organizations that do not belong, by definition, to the state?

Nonstate organizations are an alternative channel through which political elites distribute benefits to voters. I argue that passing on public money to nonstate organizations allows incumbents to undercut the opposition’s claims of credit. Thus, central incumbents have electoral incentives to use NSPs as part of their distributive strategies, using these organizations as a strategy for bypassing opposition.

I argue that funding NSPs indirectly yields a political payoff, by hurting opposition mayors, and thus strengthening the electoral chances of the president’s allies. This argument shares many of the assumptions found in studies of intergovernmental transfers. I assume that the central incumbent cares not only about staying in office, but also about voters’ welfare—a “quasi-benevolent” ruler (Arulampalam et al., 2009; Bracco et al., 2015). This central ruler also cares about “the electoral fortunes of politically aligned incumbents” Bracco et al. (2015, p. 78). Based on these assumptions, the key prediction in many of these studies is that central authorities favor aligned subnational incumbents over opposition officials because intergovernmental grants benefit copartisan mayoral candidates who, in turn, help the president (Brollo & Nannicini, 2012). This strategy maximizes both the central incumbents’ own chances of reelection as well as their party’s electoral fortunes in subnational elections because the central incumbent “loses” some electoral credit when she transfers resources to mayors (Arulampalam et al., 2009; Brollo & Nannicini, 2012).

Yet, there is no consensus in literature on exactly how grants to mayors help the president: Ames (1994), Novaes (2014), and Avelino, Biderman, and Barone (2012) argue that mayors operate as brokers who trade votes in exchange for future benefits. Fouirnaies and Mutlu-Eren (2015) argue that grants to local authorities improve these authorities’ performance and thus reinforce the “party brands” at the national level: “. . . voters’ assessment of the party’s performance at one level spills over and affects the assessment of the party’s performance at other levels of governance” (p. 806). Bracco et al. (2015) view grants as “signals” that boost the aligned local incumbents’ chances of reelection and thus decreasing grants to unaligned incumbents work in reverse: They weaken the quality signal of the opposition incumbent, increasing the aligned candidate’s chances.

My argument is closer to those developed by Bracco et al. (2015) and Brollo and Nannicini (2012): Voters reward mayors who provide public goods and, if mayors are not providing these goods because they receive fewer resources, voters will penalize mayors thereby increasing challengers’ chances of winning. Also, I argue that voters give much less credit to mayors for services administered by NSPs than for services administered by city agencies. Thus, voters consider nonstate welfare provision a weaker signal of quality about the incumbent.

Note that this argument does not imply that voters observe grants directly (neither to mayors nor to NSPs). Indeed, the fact that the spending is partially unobserved opens the door to credit hijacking in the first place (Bracco et al., 2015). It does, however, imply that voters use levels of service provision as a signal of the incumbent’s quality, and that voters attribute credit to the mayors (when government agencies administer services) and attribute it less (or not at all) to the mayors when NSPs administer these services.

The types of organizations I investigate differ from mayors in key aspects. Cammett and Maclean (2014) define them as “all providers outside of the public sector, including charitable and for-profit institutions, as well as domestic and international actors” (p. 6). I focus here on domestic nonstate, usually not-for-profit welfare providers that receive public funding to deliver social welfare to citizens. These include church-run health clinics, community schools, day centers, and after-school programs run by neighborhood associations, artisan cooperatives, and other service providers. The most important distinction between these organizations and mayors is that the former have fewer electoral incentives to expropriate credit and especially to turn credit to their electoral advantage than mayoral candidates do. Moreover, these providers are insulated from interference by the local opposition when it comes to receiving funds from the central government, as they typically do not rely on local incumbents’ mediation or approval to receive federal funds. ²

Furthermore, entrusting third-party private organizations with public service provision generally hampers voters from connecting incumbents to these services that are subsidized by the government. I argue that the Mayhewian conceptualization of credit-claiming activity—as a governmental benefit given out in an apparently ad hoc fashion to a recipient unit in which the incumbent could be perceived as the claimant for the benefit, whereas others are perceived as indifferent and hostile (Mayhew, 1974, pp. 53-54)—difficultly holds true for transfers to NSPs. Public subsidies to private organizations often go undetected by voters, creating what studies call “subterranean” or “submerged politics,” with low levels of visibility (Allard, 2009; Fox & Jordan, 2011; Mettler, 2011; Post, 2014). The implication is as follows: Nonstate welfare provision largely reduces incumbents’ ability to claim credit.

This means that it may be difficult for central incumbents to claim credit for themselves when they fund NSPs—at least as challenging as claiming credit when the federal government awards grants to mayors. However, channeling public monies through NSPs thwarts opposition incumbents’ electoral fortunes, making this strategy worthwhile. Yet, why do central incumbents shift resources to NSPs as opposed to simply withdrawing when unable to take credit for the reallocation?

This circles back to the initial question: Do presidents deprive voters of welfare to hurt opposition? I assume that voters’ welfare considerations play a role. Forgoing grants in places governed by opponents will not only harm the opposition, but it will also negatively affect voters’ welfare. Nonstate welfare provision is attractive because it mitigates such welfare losses while reducing the political costs. In addition, there are other mechanisms that also provide positive incentives for the central incumbents to shift funds to NSPs.

These nonstate welfare providers may have ties to political parties. I expect the bypassing logic to be stronger when central incumbents channel money through aligned NSPs whose members or leaders belong to the same party as the president. These are even more helpful in preventing opponents from hijacking credit because aligned organizations prevent the risk of “enemies” infiltrating said providers, thereby undermining the bypassing strategy. More importantly, these organizations may support the president’s party in party building by mobilizing civil society groups or recruiting candidates for political office (Zucco & Samuels, 2015).

Finally, my claim is that partisan alignment is one of the reasons for the central government’s spending decisions. Incumbents use federal transfers to accomplish different and competing goals: They may wish to use these transfers to gain legislative support (Pereira & Mueller, 2002), to mobilize swing and core voters (Cox & McCubbins, 1986), or to trade transfers for campaign contributions (Samuels, 2002). These strategies are not mutually exclusive.

Nonstate Organizations and Transfers: Evidence From Brazil

I construct new data sets to test my predictions about transfers to NSPs and to replicate findings on transfers to Brazilian mayors with different data ⁴ from that used in earlier studies. Brazil provides an excellent setting in which to test the bypassing hypothesis. On one hand, intergovernmental transfers are an important part of Brazil’s federal system, as both presidents and governors send resources to mayors and NSPs. And on the other hand, subnational office holders are heavily dependent on revenues from transfers.

In Brazil, many public policies are bound to the “federal pact” that imposes a division of labor between levels of government. Most basic welfare, health, and educational services are the responsibility of lower tiers of government. NSPs are not a constitutional part of this pact, but the 53 interviews with federal bureaucrats clearly make the case that NSPs share de facto responsibilities for major policies. One interviewed federal bureaucrat stated, “Although we tried, these organizations have not been acknowledged in the law as in charge of policies (executoras) . . . Yet, to us, they [NSPs] effectively are [in charge of policies].” ⁵

The central government then faces a choice when it come to making a discretionary transfer: It may transfer resources to state governments (about 30% of federal transfers), municipal governments (46%), or NSPs (21%). ⁶ I focus on discretionary transfers to mayors and NSPs. Note that the same federal program may be administered by mayors, governors, and NSPs, depending on the municipality, state, and even point in time. In fact, mayors and NSPs share responsibilities for a substantial proportion of all federal programs. ⁷

These transfers are disbursed at the executive’s discretion during the fiscal year. The executive sends the budget bill to congress. And, while legislators have several opportunities to modify and amend it, the executive has exclusive decision-making power over it once it is signed into law (Boas, Hidalgo, & Richardson, 2014; Limongi & Figueiredo, 2008).

These transfers are usually for collective goods—locally nonexcludable, nonrival, and not targeted to individuals. ⁸ This type of non-formula-based allocation is usually defined as pork, where benefits are not targeted at individuals, but their allocation has a partisan bias (Stokes et al., 2013, p. 7). Nonstate organizations in Brazil usually receive funds to organize events, provide social welfare (such as day care centers, soup kitchens, care for special needs children, or rehabilitation for drug users) or training to workers, or to build and expand organizations’ facilities. Mayors and NSPs have different roles in the provision of goods, but the type of goods they provide often overlaps. Mayors themselves contract NSPs to supply public goods—and they do so at increasing rates. ⁹ Still, there are important differences between the types of goods administered through NSP and mayors.

In general terms, NSP transfers tend to focus on welfare services while mayoral transfers target urban improvements. Most resources spent on mayors come from the Ministries of Cities, Tourism, Sports, Agriculture, and National Integration, while most resources spent on NSPs come from Ministries of Health, Social Assistance, Sports, Labor, and Tourism. ¹⁰ This is suggestive of the types of expenditures as Ministries of Health and Social Assistance tend to fund more service-provision programs, whereas the Ministries of Cities and National Integration usually fund more urban construction. Yet, it should not be taken as definitive evidence of different types of expenditures—for example, the health ministry may also finance investment expenditures, such as hospital construction and renovation. ¹¹

This may have important implications for credit claiming. On one hand, mayors receive sizable “discretionary allocations of capital construction, particularly those made out of the public eye with partisan goals” (Golden & Min, 2013, p. 77). On the other hand, NSPs benefit from funds to service provision. As these are more difficult to detect, and thus assign responsibility, mayors find it more difficult to convincingly claim credit. Thus, this type of service-provision contracts may be less visible, which would aid a bypassing strategy that limits opposition mayors’ ability to hijack credit.

Furthermore, these discretionary transfers to nonstate organizations are detailed: They specify the district in which an NSP will administer a project and also include details about the project. Thus, NSPs do not typically have a say in the location of a future project after the budget law has been approved.

The evidence in Table 1 suggests that mayoral alignment plays a role in how the president decides to target her transfers. NSPs in municipalities governed by opposition parties received about 1.54 times more resources, on average, than municipalities whose mayor belongs to the Workers’ Party. Aligned mayors receive about 1.16 times more mayoral transfers, on average, than unaligned mayors. The pattern is similar for transfers to municipalities governed by coalition parties and parties outside of the governing federal coalition, but the differences are smaller. Overall, however, these are relevant differences, and they suggest a stark contrast between the tactical logic of intergovernmental transfers and allocation to nonstate organizations.

OPEN IN VIEWER

Table 1.

Partisan Alignment and Transfers, in Reais per Capita (2003-2015).

	Mean NSP transfers	Mean mayoral transfers
Workers’ Party mayor	10.1	136.8
Non–Workers’ Party mayor	15.6	117.4
Federal coalition party mayor	9.3	125.7
Nonfederal coalition party mayor	10.2	120.1

See Supplemental Material Table S.C.9 for list of parties in coalition. NSP = nonstate providers.

Of course, the evidence in Table 1 clearly does not identify a causal effect of partisan alignment. Municipalities governed by aligned mayors may differ in many ways that influence the mix of transfers to nonstate organizations and mayors, such that confounding factors are likely. For example, aligned and unaligned municipalities are different in terms of their state capacity and the availability of service-providing organizations, which is directly related to their ability to provide services as well as the role NSPs take in providing these services (Zucco & Samuels, 2015). I therefore use a research strategy designed to identify the effect of mayoral alignment.

Testing the Bypassing Strategy

I expect partisan alignment between the mayor and the president to cause both a decrease in money funneled through NSPs and an increase in public funds to mayors. However, pinning down the effect of mayoral alignment on transfers is challenging. Chief among these challenges is that of distinguishing the effect of mayoral partisanship from observable and unobservable factors associated with both transfers and mayoral alignment.

My strategy for identifying the effect of mayoral alignment is to compare winners and losers of close mayoral elections, conditional on whether the winner or loser is a member of the president’s party. This type of RDD is also particularly well suited for investigating the effect of political alignment on transfers. Following previous theoretical and empirical work on tactical allocation of transfers places (Arulampalam et al., 2009; Bracco et al., 2015; Brollo & Nannicini, 2012), I expect that politically motivated grants are more likely to take place in competitive elections, such that the proportion of politically motivated grants are a decreasing function of the margin of victory as incumbents target resources to places where resources are more likely to influence local election outcomes. ¹²

I estimate the local average treatment of alignment using different approaches. ¹³ I follow Imbens and Kalyanaraman (2012) and Calonico, Cattaneo, and Titiunik (2014) and identify the causal effect of alignment at the cutoff using local linear regressions at the optimal bandwidths for each estimator. I also compare the averages of transfers between aligned and unaligned municipalities in municipalities in which the president’s party barely won or barely lost in the previous election, using difference of means (Dunning, 2012). ¹⁴

A Natural Experiment on Partisan Alignment

Close-race RDDs typically assume that fair and close races have an element of unpredictability: Near-winners and near-losers are determined by chance and should thus be very similar on average. This is an assumption that should not be taken for granted (Caughey & Sekhon, 2011). I evaluate the claim of “as-if randomness” of alignment for near-winners and near-losers in Brazilian mayoral races. Evidence suggests that municipalities in which the president’s party barely won or lost do not differ on observed pretreatment characteristics, just as they would (in expectation) if the treatment were truly randomly assigned.

Figure 1 shows that, in the smallest window of 0.5% of the vote margin, the difference between treatment and control groups reaches levels of statistical significance smaller for only one of the covariates (Mayor’s Age). Reassuringly, I do not observe statistically significant differences for this covariate for other windows or specifications. ¹⁵ I also evaluate balance using rank-sum tests, local linear regressions, and an F-test using the entire set of covariates and lagged outcomes. I find no systematic differences between treated and control groups. ¹⁶

OPEN IN VIEWER

Figure 1.

The p value plot, covariate balance between aligned and unaligned municipalities for 20 covariates across different windows.

The Effect of Alignment on Transfers

The results are consistent with the bypassing hypothesis. Figure 2 visually represents the effect of alignment on transfers at the discontinuity. Figure 2(a) and 2(b) show that, at the discontinuity (election of an aligned mayor), there is a decrease in transfers to NSPs and an increase in transfers to mayors. ¹⁷

OPEN IN VIEWER

Figure 2.

NSP and mayoral transfers (2003-2015).

Figure 3 and Table 2 show alternative forms to estimate and present the effects of alignment. In Figure 3, the solid black circles represent a simple average difference in federal transfers between a municipality where the president’s party won and where the president’s party lost by different vote margins, ranging from 0.2% up to 5%. For instance, in Figures 3(a), the first solid black circle illustrates that municipalities in which the president’s party won by a 0.2% vote margin or less receive about 16 reais per capita less of federal money to NSPs, on average, than municipalities in which the president’s party lost by a 0.2% vote margin or less. In Figure 3(b), the first black circle shows that aligned mayors who won by a 0.2% vote margin or less receive 50 reais per capita more federal funds, on average, than mayors who lost by a 0.2% vote margin or less.

OPEN IN VIEWER

Figure 3.

NSP and mayoral transfers, RDD estimates (2003-2015).

OPEN IN VIEWER

Table 2.

RDD Estimates, NSP and Mayoral Federal Transfers, 2003 to 2015 (reais per capita).

Estimator	Difference of means	Difference of means	Difference of means	Difference of means	Difference of means	Difference of means	Local linear regression	Local linear regression (bias-corr.)	Local linear regression (robust)
Vote margin (%)	0.5	1	2	3	4	5	Optimal bandwidth	Optimal bandwidth	Optimal bandwidth
Nonstate transfers
Estimates	−17.21	−5.64	−8.02	−6.39	−6.08	−9.72	−8.51	−12.57	−12.57
SE a	16.94	7.84	5.81	3.95	3.28	5.20	4.59	5.38	6.29
p value b	.15	.24	.08	.05	.03	.03	.03	.01	.02
Mean dep. var.	13.06	11.78	12.24	10.10	9.47	12.03	10.41	11.75	11.75
SD dep. var.	102.86	69.52	67.65	56.06	53.28	101.30	67.46	81.15	81.15
Non zero	25	59	98	152	190	240	610	397	397
n	92	194	358	534	680	828	2,188	1,397	1,397
Clusters	88	183	324	470	583	699	1,437	1,063	1,063
Mayoral transfers
Estimates	63.13	34.72	28.30	27.45	19.13	20.58	13.17	21.28	21.28
SE a	40.42	28.45	20.05	16.71	16.83	15.05	15.18	16.90	19.21
p value b	.06	.11	.08	.05	.13	.09	.19	.10	.13
Mean dep. var.	129.19	127.63	122.84	124.74	129.39	129.44	131.53	133.14	133.14
SD dep. var.	99.47	178.87	172.60	168.30	219.06	212.85	217.66	211.54	211.54
Non zero	80	172	316	467	599	730	2,614	2,035	2035
n	92	194	358	534	680	828	2,949	2,294	2,294
Clusters	88	183	324	470	583	699	1,746	1,480	1,480

Units are total transfers at each municipality, during the mayors’ term, pooled across election cycles. RDD = regression-discontinuity design; NSP = nonstate providers.

a.

Robust standard errors for difference of means and local linear regression, columns (8) and (9) use bias-corrected and robust standard errors for local linear regressions using approach by Calonico, Cattaneo, and Titiunik (2014); all standard errors are clustered at the municipality level. b. p values were calculated using a normal approximation and a one-tailed test statistic. Local linear regression at column (7) uses a triangular kernel and Imbens and Kalyanaraman’s (2012) method for estimating the optimal bandwidth (15.9% for NSP Transfers and 29.8% for Mayoral Transfers). Local linear regression at columns (8) and (9) uses uniform kernel (NSP transfers) and triangular kernel (Mayoral Transfers), bias estimated with quadratic polynomial, and optimal bandwidth selector developed by Calonico et al. (2014; main bandwidths of 8.5% for NSP Transfers and 16.2% for Mayoral Transfers). I report the standard deviation of the outcome variable in the control group as a reference to calculate standardized effect estimates.

Table 2 corroborates that this patterns holds across different estimators and bandwidths. At the optimal bandwidth, ¹⁸ the causal effect of partisan alignment is a decrease of about 13 reais per capita on federal funds to NSPs and an increase of 21 reais per capita to mayors. In terms of standardized effect sizes, the election of an opposition mayor causes an increase in transfers to NSPs of approximately 15% of the standard deviation of outcome for the control group and the standardized effect electing an aligned mayor on mayoral transfers is of approximately 10%. In other words, municipalities in which the mayor belongs to the opposition receive, on average, more transfers to nonstate organizations than do municipalities in which the mayor and the president are copartisans. The opposite pattern holds when I analyze money given to mayors: Copartisan mayors receive more resources than opposition mayors. ¹⁹

What do these estimates mean in context? Are funds given to NSPs an important resource in these municipalities? As scholars of Brazilian politics know, these discretionary transfers represent only a small part of the federal budget. Yet, for most municipalities, this is a large amount of resources, both relative to politically relevant expenditures and in comparison with welfare expenditures.

The effect size of alignment is as large as the per capita expenditures of the conditional cash transfers program Bolsa Família and larger than expenditures through legislative amendments, which are known to have substantial political relevance (Barone, 2014; Mesquita, 2009; Zucco, 2013). For example, the average expenditure with Bolsa Família between 2008 and 2012 was about 14 reais per capita (median was 11 reais per capita). ²⁰

Effect sizes are also relevant when compared with sources of municipal revenue that do not rely on federal transfers at all. In Brazil, property tax is one of the few sources of tax revenue for municipalities available to the mayor, and it represents 7% of the entire local budget (despesa corrente; de Carvalho, 2006). The average municipal revenue from property taxes (IPTU) is approximately 45 reais per capita (median was 14 reais per capita). ²¹ Both property taxes and NSP transfers are revenues to be invested in public goods for the population; the former goes to the municipal treasury while the latter goes to NSPs. NSPs in municipalities run by the opposition mayors receive, on average, an additional 28% public resources allocated through NSPs compared with average funds available to the mayor from the main local property tax. In terms of expenditures, the amount of additional money given to nonstate organizations in municipalities run by the opposition equals about 9% of the average spending by Brazilian municipalities on social welfare (assistência social e previdência social). ²²

Despite sizable effects of alignment on nonstate and mayoral transfers, the effect of alignment on transfers to mayors and NSPs does not always reach conventional levels of statistical significance, even though it does so in several windows and specifications. In fact, reaching statistical significance is striking because both the very large variances of the outcome variable and the small number of observations make it less probable for treatment effects to be detected.

The distribution of the dependent variables has many zero values and a right-skewed distribution, as commonly found in transfers data across different countries (Nunes, 2013). Thus, I also analyze using different specifications of the dependent variable by logging the dependent variable (Supplemental Material Tables S.C.21 and S.C.22). ²³ The results are stable: The direction of the estimated effects holds consistently across different specifications and across an extensive set of the president’s party margins of victory. For example, at the optimal bandwidth, the election of an aligned mayor causes a decrease in about 27% in transfers to NSPs and an increase in about 58% in transfers to mayors (Supplemental Material Table S.C.22).

In addition to the results from the RDD, other tests further corroborate that mayors’ alignment influences the amount and type of transfers a municipality receives. I examine the consequences of switches in mayors’ alignment within the same municipality: Do transfers to mayors increase after the election of an aligned mayor who succeeds an opposition mayor and do transfers to NSP decrease? In other words, does the central government sign new, or boost, contracts with the city when an aligned mayor is elected? And, does the central government revoke or let old contracts to NSPs wither within the same municipalities?

I focus on municipalities where an aligned mayor is elected after losing by a small margin. ²⁴ As expected, the election of aligned mayors leads to a substantial drop in transfers to NSPs (80 reais per capita) and a large boost in transfers to mayors (185 reais per capita). This does not mean that the central government completely revokes contracts to either mayors or NSPs; there is some degree of path dependence in both transfers to mayors and NSPs.

Still, one might worry that my alignment measure picks up party ideology rather than or in addition to partisan alignment as alignment with the federal government and having a Workers’ Party mayor overlap perfectly. To address this concern, I examine state-level transfers from São Paulo, ruled since 1995 by the social democrats (PSDB, a center-right wing party). Governors also transfer resources to mayors and to NSPs. Table 3 shows the results using the same research design described earlier and data from state-level transfers from governors to NSPs and to mayors between 2007 and 2015. ²⁵

OPEN IN VIEWER

Table 3.

RDD Estimates, NSP and Mayoral State Transfers, 2003 to 2015 (reais per capita).

Estimator	Difference of means	Difference of means	Difference of means	Difference of means	Difference of means	Difference of means	Local linear regression	Local linear regression (bias-corr.)	Local linear regression (robust)
Vote margin (%)	0.5	1	2	3	4	5	Optimal bandwidth	Optimal bandwidth	Optimal bandwidth
Nonstate transfers
Estimates	−5.84	−5.79	−7.70	−2.09	−1.74	−2.11	−5.03	−6.26	−6.26
SE a	4.21	5.78	5.58	4.96	3.89	3.28	4.26	4.45	5.12
p value b	.08	.16	.08	.34	.33	.26	.12	.08	.11
Mean dep. var.	5.78	8.39	9.83	11.01	10.92	10.00	11.34	10.58	10.58
SD dep. var.	13.97	24.08	31.40	27.54	24.57	23.10	20.17	19.57	19.57
Non zero	12	20	37	57	81	98	258	216	216
n	23	42	76	112	155	191	466	401	401
Clusters	22	41	72	104	142	169	335	301	301
Mayoral transfers
Estimates	81.50	80.61	88.94	124.05	121.75	89.66	131.55	101.36	101.36
SE a	45.79	49.78	61.71	66.26	53.64	45.92	50.24	61.68	68.96
p value b	.04	.05	.07	.03	.01	.03	.00	.05	.07
Mean dep. var.	213.06	225.32	283.09	323.34	308.86	321.49	330.73	344.67	344.67
SD dep. var.	91.02	157.04	289.88	339.04	305.89	326.78	353.22	377.60	377.60
Non zero	23	41	75	111	154	190	753	495	495
n	23	42	76	112	155	191	756	497	497
Clusters	22	41	72	104	142	169	444	355	355

Units are total transfers at each municipality, during the mayors’ term, pooled across election cycles. RDD = regression-discontinuity design; NSP = nonstate providers.

a.

Robust standard errors for difference of means and local linear regression, columns (8) and (9) use bias-corrected and robust standard errors for local linear regressions using approach by Calonico, Cattaneo, and Titiunik (2014); all standard errors are clustered at the municipality level. b. p values were calculated using a normal approximation and a one-tailed test statistic. Local linear regression at column (7) uses a triangular kernel and Imbens and Kalyanaraman’s (2012) method for estimating the optimal bandwidth (13.4% for NSP Transfers and 29.7% for Mayoral Transfers). Local linear regression at columns (8) and (9) uses uniform kernel (NSP transfers) and triangular kernel (Mayoral Transfers), bias estimated with quadratic polynomial, and optimal bandwidth selector developed by Calonico et al. (2014; main bandwidths of 10.3% for NSP Transfers and 12.7% for Mayoral Transfers). I report the standard deviation of the outcome variable in the control group as a reference to calculate standardized effect estimates.

I find the same pattern: São Paulo’s governors sidestep local opposition by funding NSPs and reward their copartisan mayors. ²⁶ These results are reassuring because they show that the bypassing strategy is contingent neither upon a particular political party nor a specific type of party. ²⁷ Moreover, these findings also suggest that this strategy is not exclusive to presidents but may in fact apply to any upper tier incumbent. ²⁸

Finally, it is not surprising that the effects of alignment on transfers to mayors are on average larger than the effects of alignment on funds sent to NSPs in both state and federal transfers: Rewarding an aligned mayor may yield higher electoral payoffs than circumventing opposition through NSPs for both presidents and governors. Besides, contracting out to NSPs is not feasible in every municipality. As a result, aligned municipalities appear to be, on average, consistently rewarded as they seem to receive more resources than they lose with NSP transfers. Yet, tests fail to reject the null that the sum of coefficients is statistically different from zero, even though the sum of coefficients is statistically significant for state transfers (Supplemental Material Tables S.C.8 and S.C.29). Thus, these results do not clearly determine if aligned municipalities are better off than places ruled by the president and governor’s party.

My main measure of alignment is mayor and president’s copartisanship, but the president’s party in Brazil governs in a coalition. Hence, other parties have the opportunity to influence the targeting of resources to municipalities and NSPs because they control many ministries. However, I expect the president to prioritize her own party’s electoral chances over other parties’ in the coalition (Boas et al., 2014). In Brazil’s multiparty system, federal allies are often local foes, resulting in alignment measures that account for both federal and local allegiances that are far from perfect. Thus, ministers who belong to coalition parties may act in their own party’s interest—not the president’s.

This way, I examine the effect of electing of mayors who belong to parties in the federal coalition on transfers to mayors and NSPs. The election of mayors who belong to parties in the federal coalition tends to increase the transfers to those mayors and decrease public funds sent to NSPs in those municipalities. However, effects are smaller and noisier: The estimated effect at the optimal bandwidth for NSP transfers is −3.74 reais per capita (p value = .18) and 2.17 reais per capita (p value = .39) for mayoral transfers—see Supplemental Material Table S.C.10. These smaller and noisier effects likely stem from the difficulty in mapping federal coalitions to local coalitions in Brazil. ²⁹

I also investigate the distribution of transfers in ministries run by the Workers’ Party and by coalition parties: Do ministries run by the president’s party and ministries run by coalition parties follow the same strategy in distributing transfers to mayors and NSPs? According to Supplemental Material Table S.C.11, Workers’ Party ministers clearly sidestep opposition and reward copartisan mayors, whereas coalition ministers do not (Supplemental Material Table S.C.12). Importantly, ministers from the Workers’ Party administer most transfers to NSP and to mayors, which could partly explain why results pooling all ministers follow the same distribution strategy as Workers’ Party ministries.

Are Opposition Mayors Unable to Hijack Credit?

When an opposition candidate defeats an aligned candidate, transfers to NSPs grow. My explanation for this effect is that transfers to NSPs reduce local (opposition) incumbents’ ability to steal credit for federally funded services relative to mayoral transfers. This implies that the type of service provider (either a city government or an NSP) can influence how voters attribute credit and, by consequence, evaluate mayors.

However, efforts to identify how federal grants and types of service providers affect how voters attribute credit and assess incumbents face several challenges. Mayors who receive transfers and those who rule municipalities that receive NSP transfers may differ in key aspects. Mayors who receive transfers may be more competent than mayors governing municipalities in which NSPs receive federal money; or these municipalities may have different needs for services. Or, as my evidence suggests, mayors who receive transfers tend to belong to the president’s party, whereas those who rule places in which NSPs receive funds usually belong to other political parties. These differences confound comparisons between mayors who receive federal monies and those who rule places in which NSPs receive federal funds.

My new survey evidence overcomes some of these challenges by using an experimental design that allows for more credible estimates of the causal effect of federal grants and type of service providers. This design offers evidence on how different forms of welfare provision (administered by the city government in comparison with an NSP) affect how voters rate mayors’ performance. These are the underlying mechanisms explaining why central incumbents bypass opposition.

Survey Data

I conducted an original survey with Brazilian voters to assess the effect of information about welfare provision on voters’ attribution of credit and incumbents’ evaluation. Respondents were recruited by Qualtrics in August 2015. The survey contained 1,103 completed interviews. ³⁰

Conjoint design: Voters’ assessment of mayors

This experimental design is motivated by trying to gain an understanding of how voters evaluate mayors based on provider types and funding sources for welfare services. The main objective is to measure whether subjects’ evaluation of mayors varies by type of service provision (either administered by the city government or an NSP), while assuming the same levels of service provision, which is often hard to find in real settings. My argument relies on opposition mayors meeting limited success in hijacking credit, making the inclusion of mayors’ partisanship necessary to test that mayors’ alignment with the federal government did not change voters’ perception of nonstate welfare provision.

Each respondent is presented with a comparison between two profiles of municipalities. These profiles contain information about the type of welfare provider (city government or nongovernmental organization [NGO]), the source of funds (federal government or no mention of the federal government), the mayor’s partisanship (PT or PSDB), and whether or not the city and the federal governments jointly provide a public good. In summary, there are three sets of treatments, for a total of eight treatment conditions, as shown in Table 4. After seeing each comparison of two profiles, respondents are asked to choose one mayor over the other and to rate their performances. ³¹ In total, each respondent sees four comparisons between two profiles of municipalities. Treatment conditions are selected at random within each set so that every profile includes one treatment from each of the three sets.

OPEN IN VIEWER

Table 4.

Conjoint Design.

Sets	Treatments
Welfare provision	The city government runs a homeless shelter.
	An NGO runs the homeless shelter.
	An NGO, with funds from the federal government, runs a homeless shelter.
	The city government, with funds from the federal government, runs the homeless shelter.
Partisanship	The mayor belongs to the PSDB.
	The mayor belongs to the PT.
Joint provision	The municipality does not have a new day care facility because federal and city governments do not have a partnership.
	The municipality has a new day care facility because federal and city governments have a partnership.
Outcomes
“Which mayor do you prefer?” (forced choice)
“In your personal opinion, how well do you think each mayor is doing?” (ordinal 7-point scale)

Outcome question order is randomly determined. NGO = nongovernmental organization; PSDB = Brazilian Social Democratic Party; PT = Workers’ party.

I follow the approach developed by Hainmueller, Hopkins, and Yamamoto (2014) to estimate treatment effects in conjoint designs. I use an ordinary least squares (OLS) regression model, with standard errors clustered by respondent, for both choice and ordinal outcomes. I test whether the causal effects of welfare and joint provision vary by mayor’s partisanship, which is the Average Component Interaction Effect (Hainmueller et al., 2014). Thus, I estimate the effect of types of welfare provider, source of funding, and joint provision of public goods on respondents’ assessments of mayors conditional on the mayor belonging to the aligned party (PT) or to the unaligned party (PSDB). ³²

Results from Figure 4 show that respondents systematically prefer mayors, both aligned and unaligned, of municipalities in which the city government administers a homeless shelter over mayors of municipalities in which an NGO administers the shelter. This evidence bolsters the argument that giving federal funds to a nonstate organization hurts unaligned mayors. When an NGO administers social welfare services with federal funds, respondents were, on average, less likely to choose the mayor from that municipality. Also, respondents still punish mayors for nonstate welfare provision when they are informed that the federal government is the source of funds.

OPEN IN VIEWER

Figure 4.

Effects of welfare and joint provision on probability of mayor being preferred.

Nonstate welfare provision reduces incumbents’ ability to claim credit in general, not only opposition incumbents—results are very similar for aligned (PT) and unaligned (PSDB) mayors. This helps to explain why central incumbents favor aligned incumbents over giving money to NSPs in places governed by copartisans: Transfers to mayors help central incumbents by boosting their own popularity, whereas transfers to NSPs depress opposition’s electoral support. But, in general, it may be harder for any incumbent to claim credit for services administered by NSPs compared with services administered directly by governments, which might reduce incumbents’ incentives to hijack credit.

Descriptive evidence from the survey further corroborates my claim that respondents attribute less credit for nonstate welfare provision to mayors compared with services provided by the city. Perceptions of responsibility for state and NSPs are very different: The majority (56%) of respondents considered the mayor responsible for the good or bad performance of the local state-run hospitals, whereas only 14% of respondents considered the mayor responsible for the religious hospital’s good or bad performances (Supplemental Table S.D.8 and S.D.6). I find a similar pattern when I ask respondents about sources of funds to NSPs: Most respondents do not believe most funds for these NSPs come from the city government. ³³

In sum, this evidence bolsters my explanation for why central incumbents subsidize NSPs in localities run by the opposition: Survey respondents do not reward opposition mayors for welfare services administered by NSPs. Thus, the survey evidence supports the claim that voters reward incumbents differently depending on the type of service provision: Relative to services administered by the mayor, nonstate welfare provision is a weaker signal of quality.

In the next subsection, I turn to yet another mechanism that helps explain the bypassing effect: the institutional features that hamper opposition mayors from hijacking credit for services administered by NSPs.

Complementary Mechanism: Providers’ Partisan Alignment

In addition to limiting opposition mayors’ ability to hijack credit, these organizations may offer additional benefits to the presidents’ party: Organizations may be helpful in party-building services, such as mobilizing organized civil society groups, creating a local presence in localities (Zucco & Samuels, 2015), recruiting and endorsing candidates or even socializing partisans (Palmer-Rubin, 2016). These activities and party-building services could operate as complementary benefits to the central incumbent in channeling resources to NSPs in aligned and unaligned localities.

After all, these organizations are not merely pawns caught in the middle of a distributive game between mayors and upper tiered incumbents; they have connections to both politicians and voters. To measure organizations ties to political parties, I collected information on leaders (directors and presidents) for all organizations that received funds in places with close races. ³⁷ I found that 38% of organizations in close races had a director who belonged to a political party at some point since 1985. ³⁸ In addition, most organizations (74%) had at least one board member who belonged to a political party at some point and, on average, 32% of board members in each organization belonged to a political party. ³⁹

Yet, no single political party had control over any one organization. None of the political parties had more than 15% of all board members overall. When considering organizations with at least two members from a political party, 92% of boards had members from more than one political party—and 66% of these board members with political affiliations had switched parties at least once.

The evidence in Table 5 suggests that directors’ partisan alignment influences the president’s allocation of funds. Central incumbents have even stronger incentives to transfer resources to “aligned” nonstate organizations in opposition-held districts because these could further reduce the opponents’ ability to steal credit. In municipalities governed by opposition parties, NSPs whose directors belong to the Workers’ Party received about 1.8 and 2.5 times more resources, on average, than NSPs whose directors who belong to other political parties or to no party at all.

OPEN IN VIEWER

Table 5.

Average Federal Transfers to NSPs in reais per Aligned and Unaligned Mayors and NSPs’ Directors’ Partisan Affiliation (2008-2015).

	Aligned mayor	Unaligned mayor
	(number of organizations)
Workers’ Party (aligned) director	205,130.95 (6)	1,055,123.78 (9)
Non–Workers’ Party (unaligned) director	277,289.52 (35)	572,752.88 (36)
Nonpartisan director	413,136.44 (48)	418,873.39 (42)

Transfers to NSPs in municipalities within 5% vote margin. NSP = nonstate providers.

In aligned municipalities, organizations whose directors are not affiliated with political parties receive, on average, more resources than both Workers’ Party and non–Workers’ Party directors. This may suggest that allocations of funds in aligned places are based on voters’ welfare rather than other electoral considerations—yet, the available evidence on this is far from conclusive.

Still, it is puzzling that unaligned directors receive more resources than nonpartisan directors in places ruled by opposition and more than Workers’ Party directors in aligned places. There are a few possible interpretations. Organizations have board members from several parties as well as former members of the Workers’ Party and, while we consider director’s partisanship as the organization’s main and primary connection to political parties, these board members could also influence the allocation of resources. Moreover, these leaders’ could be members of parties in the federal coalition, and in this way, receive funds from ministries run by these parties. These results do admit different interpretations as I cannot pin down the causal effect of leaders’ partisanship with this type of descriptive analysis and the data available. Yet this evidence is consistent with an interpretation that the central government favors “aligned” organizations to further decrease opponents’ capacity to reap credit.

Yet, despite these connections, organizations are still mostly organizationally independent of political parties and the state. Many of these organizations were founded prior to the establishment of the main parties in Brazil. Organizations are essentially autonomous, even though an important share of them has political connections to parties through their leadership (and not necessarily through beneficiaries or staff).

In comparative perspective, most organizations I examine appear to be distinct from organizations that are “party affiliates,” “associated service institutions,” or “party wings” of identity-based parties in Lebanon and India, described by Cammett (2014) and Thachil (2014). These authors describe organizations whose foundations are inescapably related to larger political movements (such as ikhwan movements and parties inspired by the Muslim Brotherhood in Egypt, Yemen, and Indonesia or the Future Movement in Lebanon, and the Hindu-nationalist Bharatiya Janata Party [BJP] in India). The web that ties political parties and organizations in Brazil seems to be much looser and less sectarian than those found in party systems with deeper religious and ethnic cleavages—after all, many organizations have multipartisan boards and high rates of party switching. The set of organizations I examine appear to combine different types of entities: the “extremely political, yet non-partisan” small business organizations that Palmer-Rubin (2016, p. 17) examines in Mexico, to the heterogeneous, yet politically active, NGOs and grassroots organizations in Bolivia that Boulding (2014) describes, and even to third-sector organizations in developed countries, such as the United States (Chaves, Stephens, & Galaskiewics, 2004; Salamon & Anheier, 1997; Salamon et al., 2003).

Additional Interpretations

Why does the election of an opposition mayor cause an increase in transfers to nonstate organizations? Besides avoiding credit hijacking to improve her party’s electoral success, the central incumbent could have other objectives in mind—most notably, the incumbent could be channeling money through these nonstate organizations to defeated copartisans, which actually means funding her own party while it is out of local office. This competing possibility is consistent with my findings that more transfers go to NSPs in unaligned municipalities, even though the consequences for redistribution would be different under this alternative interpretation.

I assessed the possibility of these organizations funneling money to the president’s party while out of local office. I obtained the tax identification code for recipient organizations in close races ⁴⁰ and cross-checked them against a data set ⁴¹ of all nonstate organizations that had contracts considered to be irregular by the general or state monitoring office. Twenty-six out of 432 (or about 6%) organizations were on this list of organizations with irregular contracts.

Out of the 31 organizations that received transfers and were listed with irregular contracts, I found information tying corruption to the mayor in only one case. The mayor involved in this investigation is an aligned mayor, which does not challenge my explanation, because, if corruption is indeed an alternative explanation to the bypassing strategy, we would expect to see corruption money in unaligned municipalities.

There is little evidence to suggest that organizations are façade party organizations that arbitrarily funnel resources to potential supporters. During interviews and visits to 20 organizations, as well as conversations with beneficiaries, I could verify that access to most of these services is based on public criteria, such as residency, income, and number of children—in contrast to the receipt of benefits contingent on individual support. Yet previous studies found evidence of some service-provision programs that are, in principle, excludable and open to clientelistic manipulation (Nichter, 2011; Weitz-Shapiro, 2012).

Thus, how might the logic of circumventing opposition that I have studied here apply to overtly clientelistic organizations? We could still observe the same bypassing strategy. If the president’s party is out of local office, it could take advantage of organizations’ provision of service to sway elections in their favor through the exchange of votes for benefits. For example, Holland and Palmer-Rubin (2015) present evidence on the association between organizational membership and vote buying, which could suggest that politicians turn to these organizations to buy their members’ votes. This additional path is compatible with the interpretation that nonstate welfare provision reduces opposition’s ability to hijack credit: If these organizations are another bolt in the party machine, they are likely to hinder opposition mayors from reaping credit.

Still, there are other plausible explanations for the strategic allocation of public monies to NSPs. Samuels (2002) argues that legislators in Brazil do not use pork barrel projects for credit claiming, but instead as a way to obtain campaign contributions from private contractors—and then trade that money for votes. However, these NSPs are not allowed to contribute to electoral campaigns—they cannot repay the federal government’s largesse with campaign contributions. ⁴² Rich (2013) describes the protagonism of reform-minded federal bureaucrats in creating alliances with nonstate organizations to pressure subnational incumbents who hesitate to implement federal policies. These mechanisms illustrate that there do exist different ways in which these nonstate organizations are involved in party politics. More research is necessary to determine under which conditions and to what extent each of these situations take place.

Conclusion

This article provides evidence of various political strategies that underlie the channeling of public money to NSPs, a pattern of distribution that is ordinarily difficult to uncover. Central governments have compelling reasons to employ the bypassing strategy. Presidents and governors wish to provide services to voters who live in places governed by the opposition, but they do not want to boost their opponents’ electoral prospects. Nonstate organizations are attractive channels for distributing benefits as they reduce opposition mayors’ ability to hijack electoral credit. In this way, transfers to NSPs minimize the president’s concerns that spending on voters in opposition-held municipalities will inadvertently lead the opposition party to reelection. My findings suggest that, by excluding these organizations, analysts risk failing to fully discern the strategic dynamics of distribution in multilevel systems.

I estimated the causal effect of mayoral partisan alignment on transfers to nonstate organizations and mayors using new federal data from Brazil between 2003 and 2015 and state-level data from 2007 to 2015. I find that federal and state authorities disproportionately fund aligned mayors and circumvent unaligned mayors by making use of NSPs. Central incumbents use these two transfers in tandem. Their strategy is aimed at boosting support for themselves and their political party, while reducing the opposition mayors’ ability to claim credit. In addition to documenting these effects, I probe the mechanisms that explain why NSPs help to avoid credit hijacking by the opposition and may provide party-building services. Finally, I contrast my explanation with additional accounts that revolve around corruption and clientelism.

This type of strategy is neither exclusive to the Workers’ Party nor to left wing parties. Moreover, the institutional set up is not necessarily particular to Brazil and not even to federal countries. In fact, previous studies from the United Kingdom suggest that similar logic may be at play: “. . . political factors were clearly at work here [United Kingdom] also, with the [nonprofit] sector being used as a tool for centralizing power by bypassing local democratic control.” (Salamon & Anheier, 1997, p. 267). My argument requires that the upper tier authorities are able to make grant allocations to NSPs without interference from lower level political actors.

Several questions remain about the types of connections between NSPs and political actors, and the type of services these organizations may provide to political parties. My article suggests that political targeting of resources through NSPs may exist even in places where the connections between these organizations and political parties are neither stable nor necessarily rooted in deep political cleavages, demarcated by religious and ethnic identities. In places where nonstate actors are more polarized and overtly partisan, I would expect that the bypassing logic would still apply, even though in these cases incumbents might be more interested in using these providers to target supporters or mobilize constituencies rather than avoiding credit hijacking. This bears the following questions: How do political parties build and keep ties to nonstate organizations? How do “political business cycles” and changes in office affect these organizations’ relationships to political parties? Finally, one possible limitation to the argument lies on the type of grantor: National or subnational governments decide how to allocate these funds, instead of international agencies. Even though previous studies suggest that international aid agencies may use nonstate organizations to bypass national governments, the motivations and capacity for circumventing a foreign government and an opponent ruler are different (Dietrich, 2014; Jablonski, 2014).

My findings also underscore the complex ramifications of delegation on accountability, in particular the claim that delegation undermines voters’ ability to hold politicians accountable (Fox & Jordan, 2011). This argument rests on the premise that voters will not, to some extent, hold politicians accountable for decisions taken by their agents. I share the assumption that delegation to both subnational incumbents and to NSPs engenders additional hurdles for voters as they try to hold central incumbents accountable. Subsidies to private organizations may obfuscate the role of government funding in nonstate welfare provision, but different types of institutional arrangements may influence the extent to which there is clarity of responsibility and the consequences for accountability are ambiguous.

Finally, in standard accounts of the provision of local public goods in developing countries, nonstate welfare providers exist where the state (or the market) fails to provide services. My findings counter this view. I show that partisan alignment—independent of state-capacity constraints—causes the central government to sidestep the opposition by channeling resources through NSPs instead. However, the estimates discussed in this article come from competitive elections and discretionary transfers, in which I expect partisan targeted transfers to be largest. Other logics, based on need, expertise, and redistribution, may also determine which places receive resources and through which type of provider. An important next step for future research is assessing to what extent each of these different motivations and factors make up government-funded nonstate welfare provision.

The implications for voters’ welfare are distinct depending on which political motivations drive the bypassing strategy: If most organizations are involved in some form of vote buying and patronage, voters in unaligned places may not only have lower levels of welfare provision, but the distribution of services is also biased toward clients and not based on need or welfare criteria. Part of the welfare consequences may simply arise from the instability in funds available to either the city government and welfare providers because of changes in mayors’ political alignment. Moreover, the amount of funds is a rather crude measure for quality of service provision. There are important differences about the potential mismanagement of these funds, risk factors for corruption, and the type of goods provided by state organizations and NSPs that also influence the quality of services offered to citizens.

There is a large body of literature that examines the consequences of partisan bias for spoils of office, but many questions regarding the repercussions of government use of these NSPs remain. These distributive decisions affect citizens’ well-being, in particular those whose welfare relies on services provided by these organizations. Documenting the political and electoral reasons for these distributive choices would contribute to understanding the welfare consequences of these forms of public service provision.