🔍 Puzzle
Why do judges sometimes rule against autocrats’ preferences even when judicial independence and secure tenure are absent? Common explanations point to strategic defection under weak governments. This study shows that a different dynamic can produce similar behavior under strong autocrats who monitor selectively: when oversight focuses on the most severe political cases, judges can gain room for maneuver by downplaying case severity to avoid triggering scrutiny.
🧾 Evidence From Taiwan’s Military Trials
New data comes from military trials during authoritarian rule in Taiwan, including the institutional introduction of a formal review threshold that drew executive oversight. Key pieces of the empirical strategy include:
- A dataset of sentencing outcomes in military political cases before and after the review threshold was introduced
- Analysis of how often judges issued sentences above the review threshold once oversight was possible
- Attention to cases where judges’ discretionary sentencing could place outcomes just above or just below the threshold
📌 Key Findings
- Judges systematically avoid issuing sentences above the review threshold after it is introduced, even without formal tenure protections.
- Sentence reductions are generally moderate rather than dramatic. They are concentrated in borderline cases where discretion allows outcomes to fall slightly above or below the threshold.
- These reductions appear calibrated to evade executive scrutiny while not being so large as to arouse autocratic suspicion and risk sanctions against judges.
đź’ˇ Why It Matters
This evidence reframes judicial agency in nondemocracies: selective monitoring by strong autocrats can create incentives for judges to behave strategically—downplaying severity to expand autonomy and protect the rule of law in limited ways. The findings illuminate how monitoring institutions and case-level discretion shape patterns of political repression and the modest margins where judicial independence can emerge even under authoritarian control.