CREDIT RISK CAPITAL DIAGNOSTICS UNDER LIMITED DEFAULT DATA IN UZBEKISTAN

Abstract

Publicly available bank disclosures report stocks of non-performing loans (NPLs) but rarely provide the default transition data required to calibrate forward-looking credit risk models. This study examines the interpretability of capital diagnostics derived from an asymptotic single-risk-factor (ASRF) model, using 56 bank-year observations covering 14 Uzbek banks over the 2022–2025 period. Official disclosures are reconciled and recalculated, and the resulting estimates are evaluated through parameter sensitivity analysis and nonlinear aggregation. Assuming a loss given default (LGD) of 45%, an asset correlation of 12%, and a confidence level of 99.9%, the system-level capital diagnostic declines from 8.20% to 7.36%, while the corresponding amount, scaled by gross loans, increases from UZS 31.99 trillion to UZS 44.44 trillion. Sensitivity analysis across alternative correlation scenarios produces substantially different estimates. Furthermore, an illustrative distressed-bank case demonstrates that unexpected loss may decrease even as modeled total tail loss increases. The study contributes a reproducible analytical framework that distinguishes observed asset quality indicators, conditional model-based estimates, and validated capital adequacy assessments. The findings support exploratory credit risk analysis; however, translating these model outputs into regulatory or economic capital requirements requires additional data on default transitions, recoveries, and exposures.