Party Expertise, Campaign Donation and Government Contracts: Evidence from an Electoral Quasi-Experiment

Transcription

1 Party Expertise, Campaign Donation and Government Contracts: Evidence from an Electoral Quasi-Experiment Paulo Arvate Klenio Barbosa Eric Fuzitani Insper Working Paper WPE: 377/2018

2 Party Expertise, Campaign Donation and Government Contracts: Evidence from an Electoral Quasi-Experiment Paulo Arvate Klenio Barbosa Eric Fuzitani March, 2018 Abstract This paper establishes that party expertise defined as experience gained from time in congress is an essential element for firms that are politically connected to legislators. By employing a regression discontinuity design that exploits the quasi-random assignment of seats in Brazilian state legislatures, we estimate the additional public contracts that companies obtain for having contributed to legislators electoral campaigns. Our estimations show that an electoral victory of a candidate from an expert party substantially raises the value of contracts for its donors at least four times the value of contracts of the defeated candidates donors. No effect was found on the contracts for donors of candidates from non-expert parties. The extra contracts obtained through campaign donation are explained by experience of the party staff members of state legislatures, but not by reelection of individual politicians or by other party characteristics like size and closeness to the executive government. Expertise built by party staff members is crucial for donors benefits from campaign contributions. Keywords: Party Expertise, Campaign Donation, Government Contracts, Political Connection. JEL classification: D21, D72, D73, H57, H72. We would like to thank Audinga Baltrunaite, Rodrigo Bandeira de Mello, Decio Coviello, Leandro De Magalhaes, Dakshina de Silva, Salomé Drouard, Sergio Firpo, F. Daniel Hidalgo, Georgia Kosmopoulou, Margaret Kyle, Luis Meloni, Lucas Novaes, Hector Perez-Saiz, Bruno Rocha, Sergio Sakurai, Aggey Semenov, Rodrigo Soares, Stephane Straub and Cezar Zucco for insightful suggestions. We are also grateful to all participants at the EEA 2014, EARIE 2014, LAMES-LACEA 2014, MPSA 2014 Annual Conference, Lancaster Management School, EBAPE-FGV, INSPER, Federal University of Pelotas and Federal University of Sergipe seminars for very helpful comments. We also thank Andre da Silva for his research assistance. Paulo Arvate and Klenio Barbosa gratefully acknowledge, respectively, CNPq and CAPES for financial support. This paper was previously circulated under the title Campaign Donations and Government Contracts in Brazilian States. All remaining errors are our responsibility. São Paulo School of Business Administration - Getulio Vargas Foundation, and C-Micro. paulo.arvate@fgv.br Insper Institute of Education and Research. kleniosb@insper.edu.br BrUsed Corporation. eric.fuzitani@brused.com.br

3 1 Introduction Recent empirical studies have shown that firms with personal and financial connections to politicians enjoy higher stock valuation (Fisman, 2001; Johnson et al., 2003; Ferguson et al., 2008; Goldman et al., 2009), are awarded more public procurement contracts (Goldman et al., 2013; Cingano and Pinotti, 2013), and gain favorable access to loans from government banks (Khwaja et al., 2005). Campaign contribution has been documented as an important way firms get connected to politicians explaining why certain firms obtain favorable political decisions. Particularly, firms that have contributed to elect politicians increase their bank financing (Claessens et al., 2008), secure more public contracts (Boas et al., 2014; Baltrunaite, 2016), and have more chance of meeting with legislators and high-level staffers about matters of donor concern (Kalla et al., 2016). 1 While the corporate benefits of political connections and the ways that firms tie themselves to politicians are well-understood by the profession, little is known about the role of political parties in such firm-politician associations. By contrast, parties have been shown to be important institutions in democratic countries, especially by aggregating collective interest (Snyder and Ting, 2002), framing the relationship between legislative and executive members (McCarty and Poole, 1995; Martin and Vanberg, 2004), and mobilizing potential voters (Larreguy et al., 2016). Besides this, parties are also seen as tools used by politicians to accomplish their goals (Cox and McCubbins, 1993), and to restrain politicians from opportunistic behavior (Aldrich, 1995; Caillaud and Tirole, 2002). Given the different functions played by parties in society, it seems natural to ask what the party s attributes are, if any, that impede or facilitate firm-politician relationships, thereby shaping the value of politically connected firms. In this paper we empirically investigate the importance of parties for firms that are politically connected to legislators. Firm-legislator associations provide fertile grounds to study the role of parties in political connections given the mutual influence that legislators and parties exert on each other in legislatures (Snyder and Groseclose, 2000; Cox and McCubbins, 2005). Specifically, we estimate the extra public contracts that companies obtain for having contributed to statelegislators electoral campaigns in Brazil, and then analyze how those additional contracts change with the parties characteristics. We particularly focus on a singular characteristic of parties which is key for the allocation of government contracts to donors of legislative candidates: experience over time of a party in the legislative branch, namely party expertise. While political scientists have highlighted the effects of parties persistence in power on individual and party incumbency advantage (Fowler and Hall, 2012; Gelman and King, 1990), government size and accountability (Rogers and Rogers, 2000; Elkins, 1974), none have shown that this persistence can give parties expertise that makes it easier for their politicians to secure public contracts for their respective donors. A party with long 1 Ansolabehere et al. (2003) and Stratmann (2005) provide a summary of the literature on campaign donation. 1

4 experience in the legislature an expert party accumulates through time, via past representation in the assembly, knowledge of the functioning of government. Therefore, expert party members know and have access to key bureaucrats who make the primary expenditure approvals, specify the goods and works needed by the government, and have the discretion to set the budget and the scope of public contracts, potentially tailoring them to certain companies. In such context, legislators that belong to expert parties may use their party knowledge and connections to link their donor-firms to those bureaucrats. Consequently, they are more likely to secure government contracts for their campaign finance supporters than other legislators. Accordingly, we hypothesize that if expertise is an essential party attribute for legislators to secure government contracts for their donors, then the procurement contracts allocated to companies that donate to legislators from expert parties will be higher than those awarded to the donors of other legislators. We investigate that situation by identifying how party expertise changes the allocation of procurement contracts for donor-firms that make campaign contributions to legislators. To assess whether expert party membership helps lawmakers to secure more contracts for their donors is not an easy task since it requires two essential ingredients. The first is a meaningful estimate of the extra contracts that a firm obtains for having contributed to campaigns that does not suffer from endogeneity bias caused by the correlation between politicians characteristics, campaign revenue and the value of public contracts of their donors (Ansolabehere et al., 2003). The second is a sensitive measure of party expertise that captures the experience over time of a party in the legislature. Employing a regression discontinuity design (Lee, 2008), we exploit the quasi-random assignment of state legislative seats in very close races in the 2006 Brazilian elections to infer the additional amount of public contracts that a firm obtains when its recipient of campaign donation gets elected. By comparing the post-election value of procurement contracts allocated to donors of state legislative candidates who won close races (treatment-group) to the ones awarded to donors of candidates who barely lost (control-group) within the same political party, we estimate the extra contracts for companies that contributed to elected legislators. As one may note, our gross return of campaign donations is unlikely to be affected by the above-mentioned endogeneity bias since the electoral victory of a legislative candidate is mainly a random event in close elections, so it is not related to the candidate s characteristics. To develop an adequate measure of party expertise, which captures the party s past representation in the legislature, we look at all past electoral races in Brazil to identify how often every party elects legislators for each state assembly in the country. Based on that, we classify as expert any party that in the three consecutive elections before 2006 (1994, 1998 and 2002) gained at least one seat in the state legislature. Note that our expert party classification is party-state-specific, and not party-specific: a certain party, that commonly elects legislators in one state may not 2

5 often elect legislators in other states. That is key in our empirical strategy by allowing us to make sure we are indeed estimating a clean effect of party expertise on the campaign donation benefits, instead of some other party-specific effect. From those two building elements, we empirically investigate how the public contracts obtained through campaign donation changes along with party expertise. Based on our estimations, we find that, on average, an electoral victory of a candidate from an expert party substantially increases the value of government contracts for a corporate donor - at least four times the value of contracts of defeated candidates donors. That corresponds to an impressive increment of 38 thousand Brazilian reais (equivalent to 17 thousand U.S. dollars in 2006) in public contracts for donor firms during the three years after elections four times the average contribution in our close electoral races. Surprisingly, we find no effect among candidates elected from non-expert parties. Those results indicate that party expertise is a key ingredient in the donor-legislator associations by making it more likely for legislators to secure benefits for their campaign contributors. Intuitively, individual legislator expertise, built by incumbent state legislators, rather than party expertise, could also shape the benefits of campaign donation. 2 To examine this potential threat to our empirical strategy, we test whether reelected legislators bring more contracts to their donors than defeated legislators that were running for reelection. Interestingly, we do not find that a reelection of a candidate raises the value of government contracts for his/her donors. This indicates that party expertise instead of individual expertise is a determinant element in the connection between donors and legislators. In particular, it suggests that knowledge/experience accumulated through party-staff members can be the primary source of expertise for parties. In order to formally assess whether expert parties build expertise on the experience of their party-staff members, we build a novel dataset containing information about the career path of all individual party members (staff of parties) that worked for the eight Brazilian state legislatures in the period between 2007 and 2010 (i.e., the political term after the 2006 election). This database was constructed by combining two comprehensive datasets: RAIS and FiliaWeb. RAIS is a longitudinal matched employer-employee administrative dataset, which is assembled and managed by the Brazilian Ministry of Labor. It takes the form of an annual census of all formal workers in Brazil, thus including all employees of all state legislatures in Brazil. FiliaWeb is an official party affiliation database from the Superior Electoral Tribunal (TSE, the highest electoral court) that contains information about every party member at any point in time in Brazil. This new dataset contains rich information about party-members careers, including the number of years in office of every individual party member employed in the state legislature in Based on that information, we classify the expert parties in two groups: the ones with and 2 Esterling (2007) finds evidence that expert legislators in the U.S. have high capacity to develop effective policies and legislation, and, for this reason, they receive more hard money contributions. Lester et at. (2008) show that expertise (tenure in office) of former public officials makes them more attractive to become company board members. 3

6 without an experienced/expert staff. Following our definition of expert party, we define an expert party as a party with an expert staff if it had at least one party member working as an appointee for a state legislature from 1998 to (Note that 1998 is the last year of term for the elected deputies (state legislators) in This makes our definitions of expert party and expert staff compatible.) We then investigate the effect of an electoral victory on the value of donors contracts in two different sub-samples: one containing only candidates from expert parties with an expert staff, and another which contains only candidates from expert parties without an expert staff. We find that only the elected candidates from expert parties with an expert staff raise the contracts obtained by donors. It provides evidence that the experience accumulated by party-staff members in state legislatures is the main driver of party expertise. In principle, other party characteristics could likewise affect the allocation of contracts to campaign contributors. For this reason we also investigate whether important features of parties, such as size in state legislatures (Jones et al., 2002) and closeness to state and federal executive governments (Goldman et al., 2013; Boas et al., 2014), also determine the allocation of contracts to legislators contributors. Strikingly, we do not find evidence suggesting that those other party characteristics affect the contracts obtained by donors. Those results suggest that party expertise seems to be the only party attribute that matters for allocation of contracts across donors. Our findings are interesting not only because they shed light on how parties determine the private returns from campaign contributors, but also because they are consistent with the theoretical predictions of some classic lobbying and campaign donation models. Notably, our results are consonant with Grossman and Helpman s (2001) and Helpman and Persson s (2001) predictions about the rents accruing to interest groups aligned with key decision makers. In our context, key decision makers are the ones that belong to expert parties. 3 Additionally, our paper is related to studies that analyze the relation between campaign donation and political outcomes. It innovates by combining a regression discontinuity design to electoral races with party characteristics to estimate the return from donations that tends to be unaffected from the typical endogeneity bias that is caused by unobservable characteristics of parties and politicians. 4 There are a number of reasons that make Brazil a perfect testing ground to examine whether party expertise does influence the allocation of public contracts to campaign contributors. First, Brazil is known for being a country with a high perceived degree of corruption, where the limited level of institutional development is one of the main explanations for the distortions of public policies toward certain interest groups. 5 According to Faccio (2006), those are the typical institu- 3 Our paper is also related to studies in political science that have previously documented that parties in Brazil are weak institutions, thereby playing a limited role in the political scene (Mainwaring, 1991; Pereira and Mueller, 2002). Differently, our results show that parties are keys for legislators to secure private benefits for their donors. 4 It is worth mentioning that, together with Lundqvist (2011), Boas et al. (2014), De Magalhaes (2014) and Firpo et al. (2015), our paper is one of the pioneers in applying RD design to proportional electoral rules. 5 The semi-annual survey conducted by Datafolha in Brazil since 1996 shows that corruption has been consistently cited by Brazilians as one of the top 10 problems of the country. 4

7 tional aspects of countries with high benefits of political connections, making Brazil a good case to study the returns of political connections and their determinants. 6 Second, public procurement accounts for a relevant fraction of government expenses in Brazil (around 15% of GDP), and given the high stakes involved in those transactions, they attract the attention of rent-seekers who ally with politicians and key decision-makers to obtain favorable government contracts. Consistently, Boas et al. (2014) document that firms that make campaign contributions to elected legislators from the president s party (Workers Party - PT) enjoyed a substantial increase in the value of their public contracts during the period studied. This suggests that public contracts in Brazil are allocated favorably to politically connected companies. Third, companies were allowed to make campaign contributions directly to candidates in Brazil until the 2014 elections, and their donations constituted the predominant source of candidates campaign revenue. 7 Besides this, Brazil is one of the few countries in the world where political candidates are legally required to make full disclosure of all their individual and corporate donors, the amount contributed, and their campaign expenses (Claessens et al., 2008). 8 Such detailed information on candidates donors and their contributions, combined with the information on government contracts awarded to every private firm, which has been publicly available at the national and subnational levels since 2006, allows us to trace a link between candidates, donor firms and public contracts, which is a relevant ingredient in our empirical investigation. Last, Brazil has a multiparty political system, which comprises a large number of parties. There are currently 35 officially registered ones, and, as Mainwaring and Scully (1995), Mainwaring (1999), Ames (2001), and Desposato (2006) point out, they are dissimilar in the following dimensions: ideology (left, center and right), closeness to the executive (a government s coalition partner, or an opposition party), political representation (number of state governors, city mayors, seats in the legislature), and tradition (experience in government). Interestingly, due to the presence of subnational governments in Brazil, the same characteristic of a party can vary across Brazilian states. It allows making cross-state comparison of party characteristics which are not party-specific, and drawing conclusions on the role of party attributes in the political landscape. That is the reason why we examine the function played by parties in corporate donor-legislator associations at the state level. The remainder of this paper is organized as follows. Section 2 describes the institutional aspects of the Brazilian electoral and political system, the sources and amounts of campaign contributions, and the key elements concerning campaign donations and public procurement in 6 Claessens et al. (2008), Bandeira-de-Mello et al. (2012), Boas et al. (2014) and Lazzarini et al. (2015), respectively, provide evidence that political connection does matter for the value of firms, the allocation of public contracts from the federal government, and for obtaining credit from government-owned banks in Brazil. 7 Corporate donations in Brazil represent about 40% of all legislative candidates total campaign revenue. 8 Although many contributions are not officially declared and similarly hidden by politicians, a considerable portion of them are registered by candidates, which implies that legal campaign contribution is a powerful element in electoral races (Samuels, 2001). 5

8 the country. Section 3 describes how the allocation of public contracts in Brazil can be related to campaign donation, and the role of party expertise. Section 4 describes the dataset, and Section 5 presents our empirical strategy and estimation method. In Section 6 we present our main findings. Section 7 shows robustness tests and Section 8 concludes. Figures and tables are in the appendix. A supplementary material section contains additional tables and figures used in the paper. 2 Institutional Background This section provides background information on the Brazilian electoral and political system. It also presents the key institutional aspects of campaign contributions and allocation of public contracts in Brazil. 2.1 The Brazilian Electoral and Political System State Organization and Electoral System. system organized at three levels: federal, state and municipal. Brazil is a federal republic with a presidential The president is the head of the federal government and also the head of State, leading the executive branch of the country. There are 26 states, a Federal District (DF), and approximately 5,600 municipalities. Each level of government includes an executive and a legislative branch, and has its own budget, and the freedom to implement its public policies and to make spending choices. Every state has a governor and a state assembly. The governor is the head of the executive branch and of the state government. Each state assembly is composed of legislators (known also as state deputies), who are responsible for making state laws, overseeing the corresponding executive branch, proposing and approving the annual state budget. The number of seats in each state assembly depends on the population size. The smallest states have 24 seats in their legislative branch, while the biggest one, São Paulo, has 94 seats. Elections for governors and state legislators are held every four years at the same time as the national elections. 9,10 As this paper looks at elections for state legislative seats, we next provide detailed information on the institutional aspects surrounding state legislator electoral races in Brazil. 9 Governors can be elected for no more than two consecutive terms, and they are elected by a simple, popular majority vote for a four-year term in a two-round system. If no candidate receives an absolute majority in the first round, a second round between the two leading candidates is held. 10 Presidential elections are held every four years in a national race, and rules for the presidential races are the same as those for gubernatorial elections. The federal legislature, the National Congress, is bicameral, and consists of the Senate, and the Chamber of Deputies. The Senate has 81 members (three per state and for the DF), elected on a majority basis for a term of eight years. Representation in the Senate is renewed every four years, with one-third or two-thirds of the seats being up for grabs in alternative national elections. The Chamber of Deputies has 513 federal legislators, elected in statewide elections by proportional representation according to the total votes cast for the party slate, for a four-year term. Those seats are distributed to states according to their population, from a minimum of eight seats (for the eight smallest states and the DF), to a maximum of 70 seats (state of São Paulo). 6

9 State legislators are elected in an open-list proportional election for a four-year term, where the states and the DF serve as electoral districts. Elected legislators can always run for reelection. Each voter from an electoral district casts a single vote for one legislator candidate, and this vote cannot be transferred to other candidates. Parties can form state electoral coalitions (alliances) with other parties, which can differ from one state to another, and from the coalition at the federal level. Each coalition can present a large number of candidates typically three times the number of seats to be filled at the state legislature which makes the election for legislators highly competitive. 11 Since coalitions do not rank-order their candidates on the ballots, campaigners compete for votes in order to be ranked highly in their coalition list: The number of votes obtained by an individual candidate in his/her coalition determines election or not. The election rules dictate that seats in state legislatures are first distributed to coalitions according to the total number of votes their candidates get, based on D Hondt s method, and then within coalition according to the number of individual votes. Some coalitions may not reach the minimum number of votes to win a seat in a state legislature. Mandates are distributed to candidates within a coalition with the highest number of votes until the last seat obtained by the coalition is assigned. So, in order to be elected, candidates must compete against their list-mates and candidates from other lists in their states. As a result, electoral campaigns for state legislative seats are notably individualized, especially because politicians power is enhanced by the number of votes obtained in elections (Mainwaring, 1991). Political System. Brazil has a multiparty political system, composed of a large number of parties. Currently there are 35 officially registered ones. All parties are national, and any political candidate must be member of a legal party to run for any elected public office. To be legally recognized, a party must have a minimum number of members, which corresponds to 0.5% of the votes cast in the previous national election homogenously distributed across the 26 Brazilian states. All eligible voters can join a party. 12 According to size, political power and representation, the most important parties are the Workers Party (PT), the Brazilian Democratic Movement Party (PMDB), the Brazilian Social Democratic Party (PSDB), and Democrat Party (DEM). They are also the parties with the longest 11 A party that does not make alliances with others (single-party coalition) can present one and a half times the number of seats to be filled. A two-party coalition can present twice the number of seats to be filled (Law 9,504/1997, called the Electoral Law). Coalitions effectively present the maximum number of candidates allowed (Mainwaring, 1991). 12 Multiple affiliations are not allowed. The Superior Electoral Tribunal (TSE) actively tracks party affiliation of all voters, canceling the oldest membership if a voter is affiliated with more than one party. Every party has its own membership rules, such as registration fees and selection process. Each party sets its own period for membership and ensures that members, if they wish, can be removed from the party. A party member may change the party affiliation any time after they register; except elected politicians and candidates for whom specific period for changes and cancellations of affiliation apply. The right and duties of members are described in the bylaws of the party. Parties may request financial contributions from their members for internal party maintenance, support to party candidates in electoral campaigns, and physical presence in certain party activities. 7

10 tradition in the country, having elected numerous and important legislators, senators and governors since redemocratization in the Other parties, like the Republican Party (PR) and the Progressive Party (PP), play significant roles in the National Congress; and others, like the Brazilian Socialist Party (PSB) and the Democratic Labor Party (PDT), have political power in strategic states. No party in Brazil has political power without making alliances with other parties. As such, parties create political coalitions in order to field candidates for elections and govern. The federal governing coalition from 2003 to the beginning of 2016, for instance, included the PT, PMDB, PP, PR, PDT and other smaller, center-left parties. The main opposition parties were the PSDB, DEM and PSB. Parties also make political alliances for state and municipal elections, but these do not always coincide with the alliances made at the federal level. In terms of ideology, political scientists categorize parties in Brazil in a wide range. Coppedge (1997), for instance, based on the perception of experts, assigns Brazilian parties as left, center and right-wingers. Specifically, Coppedge classifies PT and PSB as left-wing, the PSBD and PDT as center-left, PMDB as center, PP as center-right, and PR (former PL) and DEM (former PFL) as right-wing. In the same vein, Power and Zucco (2009, 2012) develop an alternative measure of party ideological position. Based on survey responses of federal legislators from 1990 to 2005, they construct a party ideological index from zero to ten, with smaller numbers being associated with left-wing ideologies. By this measure, there is a great dispersion across Brazilian parties. 14 Despite the large number of parties and their differences in ideology, party switching is frequent in Brazil exceeding one-third of legislators during the average legislative session, one of highest switching rates in the world (Desposato, 2006). Because of the presence of subnational governments in Brazil, parties accumulate contrasting experiences in different local political arenas. Particularly, some parties have a long tradition, strong political representation, and participation in the government coalition of certain Brazilian states, but have minor relevance in others. That is the case, for instance, of the PSB, which is a key party in the state of Pernambuco, but has weak political power in the other states. That is interesting because it enables making cross-state comparisons of political party attributes, and studying the role of those aspects in the political scenery. A detailed description of those different party dimensions at the state level is provided in Section PT was the leading party of a coalition that governed at the federal level from 2003 until the recent impeachment of President Rousseff in It holds the largest number of seats in the Chamber of Deputies (14%). PMDB has consistently maintained the largest representation in at least one of the two houses of Congress (currently the Senate, 22%), and a significant number of governors (26% in 2014 elections). PSDB was the main opposition party since 2003 (and now is part of the new governing coalition), with a large number of legislators and governors. 14 Coppedge (1997) s classification and Power and Zucco (2009, 2012) s rank ordering of parties are in general compatible, except for the PSDB which, according to Power and Zucco, has be classified as a centrist since it is to the right of the PMDB, which is a party considered as a center one by Coppedge. 8

11 2.2 Campaign Donations and Public Procurement in Brazil Campaign Contributions. In Brazil, electoral campaigns are financed by public and private funds. The matter is regulated by the Electoral Law (Law 9504/1997), which established a comprehensive system of disclosure and contribution limits for individuals and corporations to finance campaigns of all political candidates, also requiring candidates to declare their own funds spent in their campaign. 15 Companies were allowed to make direct contributions to candidates running for all office until the 2014 elections, but the candidates had to provide detailed information on their financing source and amounts received, at the donor level, to the electoral courts. Noncompliance with the regulations could bring severe penalties, including fines, disqualification of candidacy, and removal from office. The annual campaign contribution of a company to all candidates and parties could not exceed 2 percent of its gross annual revenue, and individuals could donate up to 10 percent of their gross annual income. However, there was no limit for donations to a candidate. Individuals contributions to finance any political candidate s campaign are permitted. Labor unions, trade associations and nonprofit organizations are forbidden to make campaign contributions. Public funding is also provided, and mainly comes from the federal government. It is only allocated to parties and corresponds to a small fraction of the total official resources in electoral campaigns (Bourdoukan, 2009). Parties are allowed to transfer recourses to candidates, but this mainly happens in executive races (Boas et al., 2014). 16 The limited amount of public and party funds available for legislative races, along with the intense competition for legislative seats, obliges candidates in Brazil to engage in extensive fundraising from private donors. Since Brazilian legislation allowed corporate donations until 2014, candidates developed capabilities for raising money from companies, making corporate contributions the most important source of funding for their campaigns (Samuels, 2001). In 2006 state-legislator elections, for instance, corporate donations accounted for 36 percent of the total fundraising by 15 In 2012, the Superior Electoral Tribunal (TSE) created complementary rules on campaign finance, and more recently the TSE banned any kind of corporate donations to any candidate for all elections starting from Such change in the electoral policy was triggered by recent scandals in Brazil, where campaign contributors of candidates from the Workers Party (PT), the Brazilian Democratic Movement Party (PMDB), and the Progressive Party (PP), the three main parties of the federal governing coalition until the beginning of 2016, were accused of paying kickbacks to obtain lucrative contracts with Petrobras (the giant state-controlled oil company). 16 The rules governing campaign contributions vary from one country to another. In the United Kingdom, for instance, the Representation of the People Act (RPA) and the Political Parties Elections and Referendums Act (PPERA) regulate campaign financing in all elections in the country, establishing restrictions on expenditures of candidates and parties, without imposing any limits on donations that they can receive. In the U.S., by contrast, the Federal Election Campaign Act (FECA) specifies limits for individuals, organizations and party committees to contribute to federal candidates. FECA also determines that organizations wishing to contribute must create political action committees (PACs), which raise voluntary donations from individuals. Unions, trade associations and corporations can indirectly subsidize their PACs by paying for overhead, but individual contributions are the main source of all PAC funds. Interestingly, U.S. states have different laws and disclosure practices for campaign contributions to candidates and parties. For instance, 25 states have limits on the amounts corporations may contribute, and another 21 states ban corporate contributions. 9

12 state legislators, followed by individual donations and own-resources, which amounted to 23 percent each. Party transfers represented 12 percent, and others 6 percent. Public Procurement. Brazilian state governments have budget and spending autonomy from the federal government, but they must follow national laws when awarding public contracts. The allocation of procurement contracts by all public entities (including government-controlled companies) is regulated by the Public Procurement Act (Law 8,666/1993), which determines the procedures for the acquisition of goods and services, as well as disposal of government assets. The legislation establishes that, before searching for a supplier, public entities have to provide clear and detailed description of their needs. Every public entity has its own contracting department or office, which formulates calls for proposals, published in the respective federal, state or municipal official gazette. Companies subsequently submit their bids, which are evaluated by a public official or committee. The latter then makes the final decision based on the award procedure approved for the corresponding procurement. The contracting procedure varies according to the monetary values involved. High value contracts must be awarded through open competitive bidding, while those of lower values can be acquired though invitation to bids to selected pre-approved suppliers. Electronic auctions can also be used for purchasing standard goods. Besides the lowest price, other selection criteria can be applied, such as financial capacity, technical qualification or a non-objective combination of price and technique. Also, in certain situations a tender procedure can be waived, such as when only one company can supply the needed product or service or in cases of urgent public need. While this possibility has the advantage of allowing more flexibility, it can also lead to abuse. 3 Allocation of State Government Contracts and Corporate Donations: The Role of Party Expertise In theory, the process of awarding contracts in Brazil is fully based on the merits of the proposals. However, in practice, the relationship between politicians and their respective donor firms can play an important role in the allocation of government contracts, potentially increasing donors likelihood of winning a public tender. Brazil, as other countries with limited level of institutional development, is a country where candidates and donors make commitments to one to another, through arrangements about the exchange of benefits for campaign donations (Samuels, 2001; Faccio, 2006). Accordingly, politicians from all government branches can use their political power and connections to secure government contracts for their campaign contributors. There are a number of ways politicians can influence the allocation of contracts to their donors. The possibility of only inviting certain bidders or of using other criteria than prices to award 10

13 contracts gives a margin of discretion to public buyers who, in benefit of politicians, may lessen competition and favor particular companies in the procurement market. Additionally, overbilling, not publicizing calls for proposals, and allowing fewer bidders than the minimum required are the typical fraudulent and irregular mechanisms in Brazilian procurement transactions, as documented by Ferraz and Finan (2008), related to corruption schemes used by politicians to appropriate public resources for their allies. 17 Corruption and weakening of competition are not the only ways politicians can help their donor firms to boost revenues from government contracts. State-government contracting is an administrative process that requires the approval of funds by bureaucrats, who are often willing to authorize disbursements and payments to winning firms connected to key politicians. Besides this, large-scale infrastructure projects are often renegotiated due to changes required after initial execution (Bajari et al., 2014), and monetary compensation for change orders is also subject to political influence. Likewise, firms that were hired in the first place are frequently rehired to make adaptations and extensions of the initial project, and payment to a donor-firm that has already been selected can easily be padded. Although elected politicians holding offices in the executive branch (president, governors and mayors) are in better positions to interfere with the allocation of public contracts, legislators have substantial political influence and a wide number of connections that can increase public payments to their donors. Executive politicians, for instance, can grant access of legislators and of their respective staff to key bureaucrats in the process of awarding government contracts, aiming to obtain political support in the legislature. 18 Legislators can then arrange one-to-one meetings and promote interactions between their donor firms and members of contracting offices, which potentially affect a donor s chance of being contracted. Given the role played by parties in the political arena, as documented by previous studies (McCarty and Poole, 1995; Snyder and Ting, 2002; Martin and Vanberg, 2004), one may wonder whether Brazilian political parties help their legislators to secure more public contracts for their respective donors. Political scientists argue that Brazil has an electoral system that generates weak parties with limited role in distribution of individual budget amendments (the typical sources of funds for pork-barrel projects). The country is also noted for a high party switching (Mainwaring, 1991; Pereira and Mueller, 2002; Desposato, 2006). Nevertheless, there is a particular feature of parties that has not been investigated by previous studies, and potentially shapes the value of public contracts awarded to donors of legislators: the experience over time of parties in legislature, 17 Several countries have recently made important institutional reforms and developed innovative strategies to curb corruption in public administration (see Rose-Ackerman, 2006; Rose-Ackerman and Soreide, 2011; Rose- Ackerman and Truex, 2013). Regular staff rotation in federal government is, for instance, a precautionary measure adopted in Germany to reduce or preempt corruption in public procurement (Abbink, 2004). 18 Although there are no studies at the state level on candidate-legislator-governor connections, Pereira and Mueller (2002) investigated the executive-legislative relationship at the federal level by looking at how Brazilian presidents reward those legislators who vote for their interests. 11

14 what we call as party expertise. A party that frequently holds seats in the legislature (an expert party) accumulates through time, via past elected legislators and staff and knowledge of the functioning of the government. This allows an expert party s legislators to access public officials who have a major role in the allocation of contracts. Hence, legislators from expert parties are presumably more likely to secure government contracts for their donors than other legislators. Newspaper articles often report corruption scandals in Brazil that involve legislators of expert parties who help their donors to obtain public contracts. For instance, in 2006 Correio Brasiliense, a leading newspaper, reported that Egesa Engenharia, an important construction company in the country, made substantial campaign contributions to legislative candidates in the state of Minas Gerais from the PT (an expert party in the Congress), who after being elected, voted in favor of bills that culminated in the approval of budget appropriations and disbursements for road construction in that state (Highway BR-265), contracted to Egesa (Vaz and Krieger, 2006). In another scandal, Folha de Sao Paulo (2011), also a leading newspaper, documented that Valdemar Costa Neto, a federal legislator and general secretary of the PR (another expert party), was accused of having used his political influence with the Ministry of Transportation to obtain campaign donations from construction firms. Such ability of legislators from expert parties to secure public contracts for their donors is the particular topic that we aim to document in this paper. 4 Database This section presents the dataset that we constructed to estimate the return in terms of government contracts for donors of state legislators, and to analyze how those returns change with party expertise. Since it is a novel and unique database, we start by describing the different sources of information for construction of our dataset, highlighting the most important elements available for our investigation. We then provide descriptive statistics of the main variables. 4.1 Sources and Available Information We collected detailed data on state-government contracts, campaign contributions, state legislative elections and individual party members working for state legislatures (staff of parties) for eight of the 27 Brazilian states (counting the Federal District as a state), and we match them in order to link donor firms, legislative candidates and public contracts. The eight states in our dataset are heterogeneous, and represent fairly well the diversity of the country in its various dimensions. They range from states in the poorest states (Alagoas, Pernambuco, and Rio Grande do Norte), to the richest ones (Espírito Santo, Rio de Janeiro, and São Paulo), including states in the central region of the country (the Federal District and Goiás). The last one is mainly agricultural, whereas the more urban states, such as Rio de Janeiro and São Paulo, are close to the coast. Our analysis relies 12

15 only on the information from those eight states since we could not find data on state-government contracts at the company level in the other ones. The geographic location of each state is displayed in Figure S1 in the supplementary material. Data on state government contracts at the company level is available from the Transparencia database of each Brazilian state, where information on procurement contracts awarded by each state is recorded. Using the company taxpayer identification number, we hand-collected data on the value of government contracts granted to every private firm from 2007 to 2010, a period in which state public expenditures were under political influence of the state legislators elected in 2006, so these could not have resulted from interference of politicians elected in the 2010 elections. 19 In addition, we removed from the sample the state-government disbursements made in 2007, since these expenditures were mainly approved in 2006 by the legislators elected in the previous election. To conclude, we aggregated all expenses by receiver firm, and then assigned the total disbursements received for every company to its respective recipient legislative candidates. Information on campaign donations and elections for seats in state legislatures was obtained from the Superior Electoral Tribunal (TSE). That court makes available information on characteristics of candidates, political affiliation and electoral results for all political offices in Brazil. The campaign donation dataset contains detailed information about donors and their contributions to candidates in all elections since For every state-legislative candidate in the 2006 elections, we identify the donors and the sum of their contributions using the company s taxpayer identification number (CNPJ). 20 We removed from the sample the candidates who received no corporate donations since the contracts received by their donors cannot be defined. In order to have clear treatment and control groups, we also dropped from the sample candidates (355 ones from our initial sample) that received donations from companies that contributed to both winning and losing candidates in the 2006 elections. Our final sample includes 982 state-legislative candidates and comprehensive information about them. The electoral dataset contains information on the 2006 state legislative elections, comprising data on pre-election characteristics of all candidates (gender, educational background, marital status, wealth, and occupation), their respective parties and electoral coalitions, and the number of votes that each candidate received. We dropped from our data politicians whose candidacy was withdraw or canceled. As political parties are in the center of our investigation, we collected information on a number 19 The 2006 Transparency Act (Decree 5,482) compels all levels of government in Brazil to make full disclosure of all their public accounts. Although it has increased transparency of government transactions in the country, it does not precisely specify the level of detail by which governments must reveal their expenditures and revenues, giving a margin of discretion for public bodies to decide what to make public. For the eight states analyzed in this paper we obtained data on state-government contracts at the firm level. However, the remaining ones do not provide such detailed information on the state government disbursements. 20 CNPJ is a 14-digit number, where the first eight digits identify the company, and the others the branch or subsidiary. We aggregated all donations and contracts by the first eight digits of the company s CNPJ. 13

16 of different characteristics of parties. Subsequently, we assigned them to their respective legislative candidates in our sample. For that, we relied on the information provided by the TSE for the 2006 and previous elections. It includes data on all candidates elected to political offices in Brazil (president, senators, federal and state legislators, governors, mayors and city councilors), and also information on the party affiliation and the electoral coalition that each elected and non-elected politician belonged to. This additional data allows us to look at how the party characteristics of the state legislative candidates in the 2006 elections (such as ideology, political representation in the state legislature, closeness to the state and federal executive governments, and particularly party expertise) determine the return of campaign donations. To investigate in particular whether the experience accumulated by party-staff members in state legislatures is one of the main drivers of party expertise, we built a dataset containing information about the career path of all individual party members that worked for the eight Brazilian state legislatures in the period between 2007 and 2010 (i.e., the political term after the 2006 election). To construct that database, we combined two comprehensive datasets: RAIS and FiliaWeb. RAIS (Relação Annual de Informações Sociais) is a longitudinal matched employer-employee administrative dataset which takes the form of an annual census of all formal workers in Brazil (Menezes-Filho et al., 2008). It is a database from mandatory reporting assembled and managed by the Brazilian Ministry of Labor every year since It contains above 40 million observations per year and includes information on earnings, demographic characteristics (level of education, age, gender), occupations, current and past employers, along with their identification numbers and locations (municipality and metropolitan area). In the RAIS dataset all formal job entries are identified by worker identification number (PIS), worker full name, and firm-plant taxpayer identification number (CNPJ). These identifiers are unique and do not change over time, which allows us to track individuals over time and across formal employers. It is widely recognized as a high-quality census of the Brazilian formal labor market (Dix-Carneiro, 2014; Dix-Carneiro and Kovak, 2017). Apart from the informal sector workers, RAIS covers almost all public and private sector jobs, except for very few categories of workers (a subset of self-employed individuals and elected politicians) who are not required to report information to the Ministry of Labor. This is an important feature of the dataset as it contains information about all employees in every state legislature in Brazil. RAIS also comprises information about employment contract details, displaying detailed data on hiring and firing reasons and dates, type of work contract (regular, temporary, short-term, apprenticeship), and, more importantly, information on how a worker was hired by the public sector: as a permanent civil servant or an appointee. Civil servant hiring is based on a set of objective selection criteria, demonstrated by a public examination (concurso público), while appointees are selected by politicians to occupy high-level public positions (including directors, managers, supervisors and advisors). Appointees working for state legislatures are the typical 14