Revealed and Concealed Repression: Theory and Measurement

1. Problem Statement

The paper addresses a critical measurement problem in political science and human rights research: observed repression is a distorted proxy for total repression.

• The Core Issue: Regimes strategically choose whether to conceal acts of repression (e.g., secret torture, covert disappearances, censorship) or reveal them (e.g., public executions, arrests).
• The Distortion: Because regimes substitute between revealed and concealed repression based on costs and incentives, the trend of observed (revealed) repression can move in the opposite direction of total repression.
  • Example: If a regime becomes more efficient at hiding repression (lower concealment costs), observed repression may drop while total repression rises.
  • Consequence: Policy interventions (sanctions, aid) and academic inferences based solely on observed data can be perverse, rewarding regimes that increase total repression and penalizing those that decrease it.
• Secondary Issue (Deterrence): Standard measures fail to account for the deterrent effect of repression. If the threat of repression prevents activists from organizing, no repression occurs, but the "repression index" remains low despite high underlying coercive capacity.

2. Methodology

The authors employ a game-theoretic approach combined with structural estimation and empirical application.

A. Theoretical Model

The authors develop a signaling game involving three agents:

1. Activists: Decide whether to organize (challenge the regime). They face uncertainty about whether their demands will be viewed as "good" (beneficial to the public) or "bad" (harmful) by the public.
2. The Regime: Observes if activists organize and their type (Good/Bad). The regime chooses to:
   • Concede.
   • Repress-and-Reveal (observable).
   • Repress-and-Conceal (unobservable, incurring a cost $c$).
3. The Public: Observes the outcome (Concession, Revealed Repression, or "No News"). The public updates beliefs about the presence of activists and the nature of their demands, then decides whether to protest.

Key Equilibrium Insights:

• Strategic Substitution: The regime balances the cost of concealment against the political cost of revealed repression. If concealment is cheap, the regime represses more but reveals less.
• Belief Updating:
  • No News: The public updates beliefs that activists are less likely to have organized (due to the possibility of concealed repression), but does not update beliefs about whether activists are "good" or "bad."
  • Revealed Repression: The public infers activists are organized. Crucially, because the regime is more likely to reveal repression against "bad" activists (who are harder to justify conceding to), the public updates negatively regarding the quality of the activists ($q' < q$).

B. Measurement Indices

The authors derive two indices of Total Repression ($R$) that account for both concealment and deterrence:

1. $R$ (Total Repression): Measures the probability of repression against all activists.
2. $R_G$ (Legitimate/Good Activist Repression): Measures repression against activists the public views as "good" (whose success benefits the public).

Theoretical Breakthrough (Proposition 3):
The authors prove that these unobservable indices can be recovered using observable variables via equilibrium relationships:
$$R = 1 - \frac{q - q'}{1 - q} \times P(\text{Revealed Repression})$$
Where:

• $q$: Public's prior belief that activists are "good" (before seeing repression).
• $q'$: Public's posterior belief that activists are "good" (after seeing repression).
• $P(\text{Revealed Repression})$: The observable frequency of repression events.

This formula allows researchers to estimate the "hidden" component of repression by observing how the public's opinion shifts in response to visible repression events.

C. Empirical Strategy

The authors propose a multi-step estimation strategy:

1. Survey Data: Use public opinion polls to estimate $q$ (pre-repression support) and $q'$ (post-repression support).
2. Event Data: Use high-frequency event data (e.g., daily counts) to estimate $P(\text{Revealed Repression})$.
3. Calculation: Plug these estimates into the derived formula to calculate the monthly/annual indices $R$ and $R_G$.

3. Key Results

A. Theoretical Findings

• Negative Correlation: The model demonstrates that revealed and concealed repression can be negatively correlated. A decrease in observed repression does not imply a decrease in total repression; it may signal a shift to more effective concealment.
• Deterrence: The indices capture the "missing" repression caused by activists being deterred from organizing.
• Robustness: The results hold even when introducing direct costs of revealed repression (e.g., international sanctions) or imperfect concealment (leakage).

B. Empirical Application: Russia (2020–2025)

The authors apply their method to Russia using:

• Data: VCIOM public opinion surveys (for $q$ and $q'$) and ACLED event data (for revealed repression).
• Findings:
  • Revealed Repression: Shows an upward trend, consistent with existing indices (Freedom House, V-Dem), suggesting the regime is becoming more visible in its crackdown.
  • Total Repression ($R$): Remains near its maximum throughout the period. This suggests that in earlier years (pre-2022 invasion), the regime was engaging in high levels of concealed repression that standard indices missed.
  • Legitimate Repression ($R_G$): Shows a decline over time. This indicates that the public increasingly views state actions against activists as legitimate (i.e., the public perceives the activists as "bad" or the state's coercion as justified), leading to a lower index of "illegitimate" repression.

C. Validation with Brazil (1964–1979)

Using data from the Brazilian National Truth Commission, the authors document a negative correlation (-0.45) between revealed repression (book censorship) and concealed repression (political killings). This empirically validates the model's prediction that regimes substitute between these strategies.

4. Key Contributions

1. Theoretical Innovation: Introduces a "disclosure game" framework where the sender (regime) has a type-dependent outside option (concession) and a choice to conceal information. This fills a gap in the literature on signaling and information manipulation.
2. Measurement Solution: Provides a rigorous method to recover unobserved variables (concealed repression and deterrence) using only observable data (event counts and public opinion surveys), bypassing the need for direct detection of hidden acts.
3. Policy Implications: Warns that relying on standard indices (like Freedom House or Political Terror Scale) can lead to perverse policy outcomes. For instance, a regime might appear "more tolerant" as it improves its ability to hide repression, potentially reducing international pressure while increasing actual human rights abuses.
4. Normative Distinction: Distinguishes between repression of "good" vs. "bad" activists, allowing for a nuanced analysis of how public perception of legitimacy evolves alongside state coercion.

5. Significance

This paper fundamentally shifts the approach to measuring state repression. It moves beyond the assumption that "what is seen is what exists" or that "what is seen is a lower bound." Instead, it treats repression as a strategic equilibrium outcome where the absence of visible violence can be a sign of increased total violence.

By linking public opinion dynamics to repression measurement, the authors offer a tool for scholars and policymakers to detect "silent" crackdowns in authoritarian regimes. The application to Russia demonstrates that the regime's repression was likely at maximum capacity even before the visible surge in 2022, challenging narratives that suggest a linear increase in repression over time. This framework is applicable to any context where regimes strategically manage information and where public opinion data is available.