20220224-IMF-Watch_What_They_Do,_Not_What_They_Say_Estimating_Regulatory_Costs_from_Revealed_Preferences_89页_1mb

Summary: Estimating Regulatory Costs from Revealed Preferences

Context:

• Uses changes in bank size distribution to infer regulatory costs, particularly under the Dodd-Frank Act, following regulatory thresholds (e.g., $10 billion and $50 billion).
• Banks avoid high regulations by clustering assets below thresholds, creating distortions in size distribution.

Methodology:

• Structural model where regulatory costs act as discrete taxes at thresholds.
• Revealed preference approach analyzes bank behavior (size adjustments) rather than self-reported costs.
• Maximum likelihood estimation using power-law distribution of bank sizes and identification via local density distortions.
• Includes extensions to test robustness, regulatory relief effects, alternative distributions, and indirect costs on the broader economy.

Key Findings:

• Regulatory costs at the $10 billion threshold: ~0.41% of average bank profits.
• Regulatory costs at the $50 billion threshold: ~0.11% of average bank profits.
• Total regulatory burden on banks above $50 billion: ~$4.16 million annually.
• Findings significantly lower than many self-reported estimates from banks.
• Regulatory costs have quantifiable impacts but are modest in aggregate terms, with limited evidence supporting claims of substantial harm to bank values or market structure.

Implications:

• Revealed preference approach provides an objective measure to inform cost-benefit analyses (CBA) of regulations.
• Method complements survey-based or reduced-form approaches, especially in cases of regulatory avoidance.
• Useful for policymakers to assess the true costs of regulations without relying on potentially inflated self-reports.

Robustness:

• Results hold across different data definitions of bank size, time horizons (including regulatory relief in 2018), and distributional assumptions.
• Extensions include analysis of diffuse regulatory costs, indirect economic impacts, and no regulatory effects at non-threshold round numbers.

Citation:
WP/22/41 by Adrien Alvero, Sakai Ando, and Kairong Xiao