Asset Quality Across Indian Public-Sector Banks: Persistence, RegimeDependence, and a Forward Stress Test

Abstract

This study examines the dynamics of asset quality across all twelve Indian public-sector banks (PSBs) over fiscal years 2011–2026, using a merger-consistent (pro-forma) panel of 192 bankyears constructed from Reserve Bank of India (RBI) data and bank disclosures. The empirical workhorse is deliberately simple: the canonical non-performing-loan (NPL) determinants panel, in which the gross non-performing-asset (GNPA) ratio is regressed on its own lag, real GDP growth, the ten-year government-bond yield, and lagged credit growth, with bank fixed effects and standard errors clustered by bank. Three findings emerge. First, asset quality is highly persistent: the autoregressive coefficient is about 0.71 with the full set of controls—implying a half-life of roughly two years—and is robust to adding year fixed effects and to excluding the pandemic years. Second, the relationship between GNPAs and the macroeconomy is not structurally stable: interest rates matter, but the coefficient on GDP growth is weak and wrongsigned in the pooled panel, and when the identical regression is estimated separately before and after the AQR the GDP coefficient collapses from +1.20 to zero—evidence that the 2016 surge was a supervisory recognition event rather than a business-cycle effect. Third, a transparent, calibrated stress test—reconciled to the RBI June 2026 Financial Stability Report—finds that all twelve banks remain above the eight-per-cent CET1 prompt-corrective-action floor even under a severe five-percentage-point GDP contraction, with the asset-weighted GNPA ratio rising from 1.94 to 3.93 per cent. Cross-sectionally, the stressed-capital ranking is governed by starting capital and bank size rather than current asset quality.