How Congressmen Connect: Analyzing Voting and Donation Networks in the Brazilian Congress

Victor S. Bursztyn, Marcelo Granja Nunes, Daniel R. Figueiredo2016 · DOI

Abstract The goal of this work is to analyze two of the most central activities in the life of a congressman: raising funds and voting bills. We investigate the Brazilian Congress to shed light on the relationships between the donations received by congressmen elected in 2014 and their voting behaviors during the year of 2015. We merged publicly available data obtained from the Brazilian House of Representatives and the Superior Electoral Court (TSE) in order to create a tripartite network containing campaign donors, elected congressmen, and legal bills. Using this data, we create two projected networks having congressmen as nodes and links given as follows: 1) congressmen who received donations from the same donors (donation network); and 2) congressmen who voted in accordance to each other on legal bills (voting network). After characterizing these networks, we propose three fundamental questions on the behavior of congressmen that could benefit from the methods and concepts provided by Network Science. Finally, we analyze the results and compare them to general domain knowledge.

Introduction

Political science has long debated the influence of money on politics, particularly regarding corporate donations. In 2015, Brazil’s Supreme Court banned corporate donations, marking a historic shift in campaign funding. However, prior to this ban, there was a lack of unified, quantitative data connecting campaign "inputs" (funds) to legislative "outputs" (voting). This research bridges that gap by applying Network Science to the Brazilian Congress, mapping how campaign donations and voting behaviors create distinct social structures.

The study addresses three core questions:

  1. Does belonging to the same party or the same region more strongly influence shared donors and shared votes?
  2. How cohesive are Brazilian political parties during floor votes?
  3. Can network topology identify the most influential "heads" of Congress?

Methodology and Datasets

The researchers integrated public data from the Superior Electoral Court (TSE) and the House of Representatives API, covering 52,245 donors, 456 congressmen, and 234 bills from 2014–2015. The authors have made this consolidated dataset publicly available on GitHub.

The core of the analysis is a tripartite network model. This graph consists of three types of nodes: campaign donors, congressmen, and legal bills. To facilitate analysis, this complex structure is "projected" into two separate unipartite networks where nodes represent only congressmen:

  • Donation Network: A link exists between two congressmen if they received money from the same donor.
  • Voting Network: A link exists if they voted the same way (Yes, No, Abstain, or Obstruction) on the same bill.
Figure 1: Visualization of the tripartite network structure connecting donors, congressmen, and voting events.

To measure connection strength, the study utilizes the Jaccard Index. Furthermore, a "Similarity Gain" metric was introduced to quantify homophily—the tendency of individuals to associate with similar others based on party or region. Finally, centrality metrics like PageRank and Strength were tested to correlate network position with political influence, using the DIAP "Heads of Congress" report as a ground truth.

Key Insights: The Wallet vs. The Vote

The analysis reveals a striking "decoupling" of political influence drivers. The factors that drive campaign funding are fundamentally different from those that drive legislative behavior.

Drivers of Donation Similarity

In the donation network, geography is the primary driver. Being from the same state (co-regionality) provides a 51% similarity gain, while being in the same party only provides a 25% gain. This suggests that donors, particularly before the corporate ban, focused heavily on regional interests.

Figure 3: Jaccard gain for campaign donors associated with co-partisanship (top, green) and co-regionality (bottom, red).

Drivers of Voting Similarity

In the voting network, the priorities flip. Co-partisanship provides a 46% similarity gain, while co-regionality's influence drops to a negligible 5%. Political parties are nearly 10 times more effective at channeling voting behavior than regional interests. Despite perceptions of political fragmentation, Brazilian parties maintain strong, expressive voting identities.

Party Cohesion and Influence

Cohesion does not strictly depend on party size. While the small left-wing party PSOL was the most cohesive, the small PV (Green Party) showed very low cohesion. Large parties like PSDB and PT maintained high cohesion, whereas the large centrist PMDB showed only mild internal unity.

Regarding influence prediction, PageRank and Strength centrality identified influential congressmen with 27% precision—2.5 times better than random selection (11%). This suggests that while voting patterns reveal leadership, topology alone does not capture the full picture of political power.

Conclusion and Future Work

The study highlights that the voting network is extremely dense (0.88), making it difficult to distinguish nodes based solely on topology. Additionally, Betweenness Centrality proved problematic as high abstention rates created artificial "bridges" in the network.

Future research will compare these 2015 networks with 2018 data to assess how the ban on corporate donations affected the "homophily of the wallet." Incorporating more granular data, such as participation in special commissions, may also improve the precision of influence prediction models.

Summary updated

Abstract

The goal of this work is to analyze two of the most central activities in the life of a congressman: raising funds and voting bills. We investigate the Brazilian Congress to shed light on the relationships between the donations received by congressmen elected in 2014 and their voting behaviors during the year of 2015. We merged publicly available data obtained from the Brazilian House of Representatives and the Superior Electoral Court (TSE) in order to create a tripartite network containing campaign donors, elected congressmen, and legal bills. Using this data, we create two projected networks having congressmen as nodes and links given as follows: 1) congressmen who received donations from the same donors (donation network); and 2) congressmen who voted in accordance to each other on legal bills (voting network). After characterizing these networks, we propose three fundamental questions on the behavior of congressmen that could benefit from the methods and concepts provided by Network Science. Finally, we analyze the results and compare them to general domain knowledge.
BibTeX
@misc{bursztyn2016how,
  author = {Victor S. Bursztyn and Marcelo Granja Nunes and Daniel R. Figueiredo},
  title = {{How Congressmen Connect: Analyzing Voting and Donation Networks in the Brazilian Congress}},
  year = {2016},
  doi = {10.5753/brasnam.2016.6451},
}
Figure