📈 economics

Where does corruption risk sit? Core–periphery structure and clustering of single bidding in Ecuador’s subnational public procurement networks

This study analyzes Ecuador's subnational public procurement networks to reveal that corruption risk, measured by single bidding, is significantly clustered and concentrated in market cores rather than being uniformly distributed, suggesting that audit strategies should prioritize central actors in provinces where this specific concentration pattern is detected.

Original authors: Lissete Reyes-Vanegas, Jessica Matute-Petroche, Carlos Redroban-Ortiz

Published 2026-11-06

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Lissete Reyes-Vanegas, Jessica Matute-Petroche, Carlos Redroban-Ortiz

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Public procurement is the massive engine that keeps governments running, from building roads and schools to buying medical supplies and office equipment. It is the point where public money meets private business, a process that involves billions of dollars changing hands every year. Because so much value is at stake, this system is a frequent target for corruption, where officials might steer contracts to specific friends or companies in exchange for bribes. Detecting this wrongdoing is notoriously difficult because the people involved have every reason to hide their tracks. In recent years, data scientists have found a clever way to look for trouble without needing to find a smoking gun in a single contract. They noticed that when a government asks for bids on a project and only one company shows up to compete, it is a warning sign. While a single offer can happen for innocent reasons, such as a very specialized product or a lack of local suppliers, a pattern of single offers across a region suggests that the competition has been rigged or that the process was designed to favor a specific winner. By mapping out who buys what from whom, researchers can turn these millions of transactions into a network, revealing hidden patterns of favoritism that are invisible when looking at contracts one by one.

A team of researchers in Ecuador decided to test whether these patterns held true not just across entire countries, but within the smaller, local markets of a single nation. They turned their attention to the public procurement records of Ecuador, a country where the government publishes detailed data on every contract awarded through its electronic platform. The team gathered nearly 46,000 competitive contracts awarded between 2015 and 2021, linking thousands of public buyers to tens of thousands of suppliers. Their goal was to see if the risk of corruption, measured by the frequency of single bids, clustered in specific groups of buyers and sellers, and whether this risk was concentrated in the most central, powerful parts of the market or pushed out to the edges. To do this, they built a map of relationships for each of Ecuador's twenty-one provinces, treating every contract as a connection between a buyer and a supplier. They then used a method to identify the "core" of each market—the most active and connected buyers and suppliers—and checked if the suspicious single bids were happening mostly in that central group or in the less connected periphery.

The results revealed a landscape of corruption risk that was far more varied than expected. Across the country, about thirty percent of competitive procedures received only one bid, a figure that varied significantly from province to province. In some coastal regions, nearly half of all procedures attracted a single offer, while in the central highlands, the number was closer to twenty percent. The researchers found that these risky contracts were not scattered randomly; instead, they were tightly clustered in specific communities of buyers and suppliers in almost every province. This means that if an auditor found a suspicious contract in one part of a local market, they would be wise to look closely at the other contracts made by the same group of buyers and sellers, as the risk tended to travel in packs. However, the location of this risk within the market structure was not uniform. In some provinces, the single bids were heavily concentrated in the core, meaning the most powerful and frequent players were the ones involved in the uncontested deals. In other provinces, the risk was found in the periphery, involving smaller, less connected actors.

This finding challenged a previous theory derived from studies in Europe, which suggested that the more centralized a market was, the higher the risk of corruption would be. The researchers found that in Ecuador, the size of a province's central market had no connection to how much corruption risk it held. A large, central market could be relatively safe, while a smaller one could be rife with risk. What actually predicted a high level of risk was not the size of the core, but where the risk sat within it. Provinces where the single bids were concentrated in the core had significantly higher overall rates of single bidding. This suggests that in these areas, corruption was organized around a stable set of relationships between powerful buyers and favored firms, effectively capturing the market. In contrast, in provinces where the risk was in the periphery, the problem was more diffuse, spread across many small buyers and occasional suppliers.

The study suggests that there is no single strategy for fighting corruption that works everywhere. In provinces where the risk is concentrated in the core, like El Oro, a small number of high-volume relationships account for a large share of the exposure, making targeted audits of central actors a highly efficient use of resources. In provinces where the risk is scattered in the periphery, broader preventive measures, such as better training or standardized documents, might be more effective. The researchers emphasized that while single bidding is a strong indicator, it is not proof of wrongdoing in any single case, and their analysis relied on patterns across thousands of contracts. By using open data to map these networks, authorities can move beyond guessing and start looking for trouble exactly where the data says it is hiding, tailoring their approach to the specific structure of each local market. This approach offers a low-cost way to understand the hidden geography of public spending, turning raw numbers into a clear picture of where the system is most vulnerable.

Technical Summary: Corruption Risk in Ecuador's Subnational Public Procurement Networks

Problem Statement
Public procurement is a primary channel for the diversion of public resources, yet measuring corruption within this domain remains challenging. Traditional methods, such as perception surveys and conviction data, often fail to capture actual behavior or are confounded by institutional capacity. While "single bidding" (awards to a single offer in a competitive tender) has emerged as a robust objective proxy for corruption risk, existing literature—specifically Wachs et al. (2021)—has established patterns of risk distribution primarily within high-income European Union markets. These patterns include the clustering of risk within specific buyer-supplier communities and a variable relationship between market centralization (core-periphery structure) and risk levels. It remains an open question whether these regularities hold within the subnational context of a middle-income Latin American country, where markets are smaller, more localized, and governed by a unified national regulatory framework.

Methodology
The study adapts the network analysis framework of Wachs et al. (2021) to Ecuador's subnational procurement landscape.

Data Source: The analysis utilizes 46,279 contracts awarded through competitive procedures between 2015 and 2021, sourced from the National Public Procurement Service (SERCOP) via the Open Contracting Data Standard (OCDS). The dataset links 2,838 public buyers and 15,856 suppliers across 21 of Ecuador's 24 provinces.
Network Construction: Weighted bipartite networks were constructed where nodes represent buyers and suppliers, and edges represent contracts. Two sets of networks were analyzed: national networks for each year and provincial networks pooling seven years of data to ensure statistical stability.
Core-Periphery Analysis: Market cores were identified using a strength-based k-shell decomposition. Recognizing that standard thresholds (based on average strength) often left provincial cores empty or negligible, the authors defined the core as the top decile of the core-number distribution for buyers and suppliers separately.
Community Detection: To identify clustering, bipartite networks were transformed into line graphs (where edges become nodes), and the Louvain algorithm was applied to maximize modularity.
Statistical Validation: Both core concentration and community clustering were tested against null models. Single-bid labels were randomly permuted within two-digit Central Product Classification (CPC) codes (and in stricter models, within CPC-by-procedure strata) 1,000 times to establish expected baselines. Ratios of observed to expected values were calculated to determine over- or under-representation.
Sensitivity Analysis: The study employed sensitivity checks regarding core definition thresholds (ranging from top 5% to top 20%) and procedure types (specifically electronic reverse auctions) to ensure robustness.

Key Results

Prevalence of Single Bidding: Single bidding affected 30.2% of competitive procedures nationally, with significant variation across provinces (ranging from 20.2% in Cotopaxi to 44.3% in Los Ríos).
Clustering: Single bidding was significantly clustered in every national network and in 17 of the 21 provinces. The clustering ratios ranged from 1.2 to 2.0 times the null expectation, though these magnitudes were generally lower than those observed in European markets.
Core-Periphery Dynamics: Unlike the European cross-country pattern where larger market cores correlated with higher risk, the size of the core in Ecuadorian provinces was unrelated to the overall single-bid rate ( $r = -0.19$ ).
Risk Location: The critical predictor of provincial risk was the location of the risk within the network structure. Provinces with higher overall single-bid rates were those where single bidding was significantly over-represented in the core (e.g., El Oro, Manabí, Guayas). Conversely, in provinces like Loja and Pichincha, risk was more prevalent in the periphery or showed no significant core concentration.
Heterogeneity: The study identified distinct regional profiles: some provinces exhibited risk concentrated in the core and clustered (El Oro), others had clustered risk without core concentration (Pichincha), and some showed risk located mainly in the periphery (Loja).

Key Contributions

Empirical Extension: The paper provides the first systematic description of the network distribution of corruption risk across Ecuadorian provinces, extending network-based procurement analysis to a subnational Latin American context.
Methodological Adaptation: The authors propose a quantile-based rule (top decile) for defining market cores in sparse regional networks, addressing the limitation where standard strength-based thresholds fail to identify cores in smaller markets.
Nuanced Risk Mapping: The study demonstrates that within a single country under uniform regulation, the structural organization of corruption risk varies significantly. It challenges the assumption that market centralization alone is a universal indicator of risk, showing instead that the concentration of risk among central actors is the distinguishing factor.

Significance and Claims
The authors claim that their findings offer practical implications for audit strategy. They argue that anti-corruption agencies should tailor their approaches based on local network structures:

In provinces where risk is concentrated in the core (e.g., El Oro), audits should prioritize a small number of high-volume relationships between central buyers and suppliers.
In provinces where risk is dispersed in the periphery, broader preventive measures (e.g., training, standardized documents) may be more effective than targeting specific central actors.

The paper concludes that network measures derived from open contracting data provide a low-cost mechanism to differentiate these risk profiles. However, the authors remain modest in their claims, noting that single bidding is a proxy rather than proof of wrongdoing, that cross-provincial correlations are descriptive rather than causal, and that the analysis is limited by data sparsity in certain years (notably 2018) and the pooling of multi-year data for provincial networks. They suggest future work should validate these network signals against actual audit findings and judicial cases.

Technical Summary: Corruption Risk in Ecuador's Subnational Public Procurement Networks

More like this