Corruption Studies: An Interdisciplinary Overview

• Corruption studies is an interdisciplinary field that analyzes the misuse of entrusted power through complex, networked interactions among public officials, corporate agents, and institutions.
• Recent research applies methodologies like legal analysis, network science, and agent-based modeling to uncover hidden, multidimensional corruption risks.
• Empirical work integrates macro indicators with micro-level proxies, using computational models to reveal regularities and inform targeted anti-corruption strategies.

Corruption studies is the interdisciplinary analysis of corruption as the “misuse of entrusted power for private gains,” the “abuse of entrusted power or public authority for private benefit,” and, in a more legal formulation, “the use by an official of his authority and the rights entrusted to him for personal gain, which is contrary to the rules established by law” (Gnaldi et al., 2023, P et al., 19 Mar 2026, Antonov et al., 2021). Contemporary work increasingly treats corruption as a latent, hidden, multidimensional, and networked phenomenon that unfolds across individuals, organizations, markets, and institutions, and that must therefore be studied through legal analysis, perception measures, transaction-level indicators, network science, agent-based modeling, and dynamic macro-political models (Nicolás-Carlock et al., 2021, Villamil et al., 2022).

1. Conceptual scope and analytical objects

Corruption studies centers on public officials, corporate agents, intermediaries, government agencies, firms, NGOs, criminal organizations, and regulatory bodies, but it does not stop at actor lists. A central shift in recent work is from isolated acts to “the complex networks in which corruption takes place,” including the interactions, incentives, and flows of power, resources, and information through which corruption emerges, spreads, stabilizes, or destabilizes within a “system of systems” (Nicolás-Carlock et al., 2021). This move places corruption simultaneously at micro, meso, and macro levels.

The field also distinguishes among multiple types and levels. One standard taxonomy includes petty administrative corruption, grand corruption, and state capture; another distinguishes transactional corruption from systemic corruption; and legal-institutional typologies distinguish bribery, extortion, embezzlement, procurement collusion, favoritism, conflict of interest, influence peddling, misuse of information, obstruction, and abuse of discretion (P et al., 19 Mar 2026). In a historical-legal register, corruption has also been classified as large-scale versus small-scale, systemic versus systematic, political versus “civilized” corruption, and “Western” versus “Eastern” forms, with the latter emphasizing long-term relations rooted in family, professional, and corporate ties (Antonov et al., 2021).

A related conceptual distinction is between corruption and corruption risk. In procurement research, red flags are “proxy measures for corruption signalling risks of corruption, rather than actual corruption,” and are “expected to be correlated with corrupt practices, rather than perfectly matching them” (Gnaldi et al., 2023). This distinction has become foundational because actual corruption is often hidden, selectively prosecuted, and only partially observable through convictions or investigations.

2. Definitional and measurement problems

A programmatic diagnosis of the field identifies three intertwined obstacles: the complexity of the phenomenon and its context, the complexity of the analytical description, and the complexity of multiple disciplinary perspectives (Nicolás-Carlock et al., 2021). Corruption is behaviorally complex, “hidden,” and “interwoven” with legitimate actions; it is embedded in the “intricate structure and dynamics of the social, economic and political systems of society”; and it is studied through legal, economic, sociological, anthropological, and political-science vocabularies that do not always align.

Measurement problems follow directly from this hiddenness. Aggregate indices remain central because corruption is hard to quantify through “objective” indicators. One influential cross-national study uses the Heritage Foundation’s “Freedom from Corruption” index, derived directly from Transparency International’s CPI as

$\text{COR} = 10 \times \text{CPI},$

with 134 countries observed from 1996 to 2014 (Paulus et al., 2015). That work treats corruption perception as a time series and shows a “well-defined hierarchy” of countries using average linkage clustering, but it also exemplifies the field’s dependence on perception-based measures rather than directly observed corrupt transactions.

At the same time, corruption studies has moved toward more specific indicators and micro-level proxies. A procurement-based validation study computes fifteen red flag indicators on the Italian National Database of Public Contracts and then evaluates them within a multidimensional Item Response Theory framework (Gnaldi et al., 2023). The indicators include non_open_count, single_bid_count, MEAT_count, excluded_bids, modifications, amount_deviation, time_deviation, and winners_homog, among others. The results show a multidimensional structure rather than a single latent corruption-risk scale, with dimensions linked to non-open procedures, single bidding, discretionary award criteria, bid exclusion, and contract modification.

The measurement literature therefore supports a dual conclusion. First, perception indices remain analytically useful because they provide broad coverage and temporal depth. Second, corruption is not empirically exhausted by scalar national scores. This suggests that cumulative corruption science requires both macro indicators and context-specific, transaction-level measures.

3. The computational and network turn

Recent research argues that corruption is fruitfully understood as a collective action problem involving “embedded people and organizations,” and that computational social science is particularly well suited to studying this embeddedness (Villamil et al., 2022). In this view, interactions and dependencies matter more than the nature of isolated parts. Network science and complex systems approaches are therefore not auxiliary tools but a candidate common language for corruption studies.

Networks appear in several canonical forms. In procurement markets, issuers and winners are modeled as weighted bipartite networks; in political scandals, nodes are individuals and edges join co-participants in the same scandal; in local social-capital studies, nodes are residents and edges are social ties; in offshore and shell-company research, nodes may be firms, intermediaries, and jurisdictions linked by ownership, representation, or control (Villamil et al., 2022). This permits the use of centrality, community detection, clustering, assortativity, and core–periphery analysis to study corruption-related structure.

The substantive payoff of this turn is evident in comparative procurement and scandal research. In European Union contracting markets, “highly centralized markets tend to have higher corruption risk,” corruption risk is “significantly clustered,” and the same aggregate level of corruption risk can be concentrated either in the core or in the periphery of the market depending on the country (Wachs et al., 2019). In political scandal data from Spain and Brazil, corruption networks share “universal structural and dynamical properties,” including similar degree distributions, clustering and assortativity coefficients, modular structure, and a growth process marked by the coalescence of components through a few recidivist criminals (Martins et al., 2022).

The network turn also extends to the social foundations of corruption. At the town level, settlements with “fragmented social networks,” indicating an excess of bonding social capital, have higher corruption risk, while settlements with “more diverse external connectivity,” suggesting a surplus of bridging social capital, are less exposed to corruption (Wachs et al., 2018). This finding connects network topology to impartiality, conformity, favoritism, and local accountability.

4. Empirical regularities across scales

Cross-national work shows that corruption is not randomly distributed across states. Using Euclidean distances on 1996–2014 COR time series and average linkage hierarchical clustering, one study identifies four main clusters: a low-corruption, high-income cluster; a moderately low-corruption “catching-up” cluster; a high-corruption, low-income cluster; and an intermediate-corruption cluster of transitional regimes (Paulus et al., 2015). At the cluster level, “the ranking of countries according to their corruption perfectly copies the ranking according to the economic performance measured by the gross domestic product per capita of the member states,” although within-cluster heterogeneity remains substantial.

Administrative microdata reveal more differentiated structures. In Italian public procurement, a five-dimensional solution is preferred for corruption-risk indicators, with dimensions centered on non-open procedures, single bidding, discretionary use of MEAT criteria, bid exclusions, and contract modifications (Gnaldi et al., 2023). The dimensions are “generally non-superimposable,” and several popular indicators—such as advertisement, evaluation, and winners_homog—do not load strongly on a distinct latent dimension. This finding is methodologically important because it cautions against collapsing heterogeneous procurement anomalies into a single unvalidated scalar index.

Organizational studies reveal threshold behavior at the intra-organizational level. In a stochastic contagion model on hierarchical networks, the critical corruption threshold depends on organizational structure: for the flattest structure $(k=10, L=4)$, $\lambda_c \approx 0.32$; for the moderately tall structure $(k=4, L=6)$, $\lambda_c \approx 0.38$; and for the tallest structure $(k=3, L=7)$, $\lambda_c \approx 0.42$ (Nekovee et al., 2017). Flatter organizations are thus more susceptible at onset, while taller or moderately tall structures can sustain higher final corruption prevalence once the threshold is crossed. The same model reports that, for a 1,000-strong organization, 5 percent of the workforce is a critical threshold of whistle-blowers needed to constrain the spread of corruption, and if this number is around 25 percent, the corruption contagion is negligible.

A plausible implication is that corruption studies increasingly treats “corruption” not as one observable object but as a family of scale-dependent empirical regularities: clustered country trajectories, multidimensional procurement risks, and threshold-driven organizational contagion.

5. Dynamic and formal models

Formal modeling in corruption studies now spans macroeconomic interdependence, public-goods enforcement, bribery networks, and public-contract evolution. A recent macro-level model uses a coupled vector autoregressive system for GDP and CPI across 13 countries: $\left\{ \begin{array}{rl} \mathbf{x}_t &= \mathbf{b} + \sum_{s=1}^{p}\Phi(s)\mathbf{x}_{t-s} + \sum_{s=1}^{p}\Pi(s)\mathbf{y}_{t-s} + \boldsymbol{\xi}_t,\[0.4em] \mathbf{y}_t &= \mathbf{c} + \sum_{s=1}^{p}\Psi(s)\mathbf{y}_{t-s} + \sum_{s=1}^{p}\Gamma(s)\mathbf{x}_{t-s} + \boldsymbol{\zeta}_t, \end{array} \right.$ with $p=1$ selected by AIC/BIC and directed adjacency matrices built from conditional Granger-causality tests (Bartolome et al., 2024). The resulting networks separate GDP→GDP, CPI→GDP, CPI→CPI, and GDP→CPI effects, and support the claim that corruption and growth behave as “interdependent, cross-border processes,” not isolated national attributes.

Evolutionary game theory provides a second family of models. In an institutional punishment public goods game combined with a bribery game, cooperation evolves according to

$\dot{x} = x(1-x)(f_C-f_D),$

and the effects of corruption depend on who is more inclined to bribe (Liu et al., 2022). Theoretical and numerical results show that corruption reduces cooperation when cooperators are more inclined to provide bribes; stronger leader and richer economic potential are both important to enhance cooperation; and, when defectors are more inclined to provide bribes, stronger leaders can sustain the contributions of public goods from cooperators if the economic potential is weak.

A third strand models corruption as a hierarchical bribery network. In a police bribery model focused on harassment bribery, the citizen’s expected utility is

$(k=10, L=4)$0

with the reward function specified as

$(k=10, L=4)$1

The paper argues that the probability of detection of harassment bribery $(k=10, L=4)$2 can be enhanced by asymmetric punishment and an award equivalent to the amount of punishment to the network, thereby increasing the probability of detection of overall bribery (Pramanik, 2022). The model’s emphasis is explicitly on network collusion and on the distinction between overall bribery detection $(k=10, L=4)$3 and harassment-bribery detection $(k=10, L=4)$4.

Agent-based models of public contracts add structural variables such as group size and salary dispersion. In one such model, there are two types of agents—business people and public servants—both of whom may adopt corrupt or honest strategies, and the asymptotic corruption level is summarized in three phases: a phase where corruption dominates; a phase where corruption remains in less than 50% of the agents; and a phase where corruption disappears (Valverde et al., 2023). The paper concludes that a combination of large group sizes of interacting servants and business people and small dispersion of the salaries of public servants contributes to the decrease of systemic corruption in public contracts.

6. Governance, monitoring, and emerging directions

Institutional determinants remain central even in computationally enriched corruption studies. A spatial Bayesian Model Averaging analysis of 115 countries over 1985–2015 concludes that, among 39 predictors of corruption, Rule of Law is “the most persistent determinant of corruption” once spatial spillovers and model uncertainty are integrated out (Rahimian, 2021). The same study shows that corruption is spatially clustered and that accounting for spatial dependence sharply reduces the apparent robustness of many familiar correlates.

Policy and monitoring agendas increasingly combine this institutional emphasis with multidimensional measurement. Procurement-based indicators are explicitly linked to SDG 16, including Target 16.5 on substantially reducing corruption and bribery in all their forms, Target 16.6 on effective, accountable and transparent institutions, and Target 16.10 on public access to information (Gnaldi et al., 2023). At the same time, a complexity-oriented perspective insists that “complex corruption networks are tackled with complex anti-corruption networks,” requiring coordinated cooperation among governments, private institutions, and civil society rather than isolated interventions (Nicolás-Carlock et al., 2021).

An emerging extension applies standard corruption taxonomies to institutional AI. In multi-agent governance simulations, corruption is scored as the abuse of entrusted power or public authority for private benefit, and, among models operating below saturation, governance structure is reported to be a stronger driver of corruption-related outcomes than model identity (P et al., 19 Mar 2026). Lightweight safeguards can reduce risk in some settings but do not consistently prevent severe failures. This does not replace traditional corruption studies, but it indicates that the field’s concepts—abuse of authority, favoritism, procurement collusion, state capture, accountability sabotage—are being generalized to autonomous agents occupying formal governmental roles.

Across these strands, the future research agenda is convergent. Corruption studies is moving toward multi-layer and multi-scale models, broader use of registries and transparency portals, standardized procurement data, stronger integration of network structure with dynamic modeling, and more systematic linkage between micro behavior and macro outcomes (Nicolás-Carlock et al., 2021, Villamil et al., 2022). This suggests a field increasingly oriented toward exacting empirical validation, explicit structural representation, and context-sensitive institutional design rather than solely toward aggregate perception scores or single-cause explanations.