Nonperforming Loans and Output Growth in a 140-Economy Panel, 2001–2024: Dynamic Fixed Effects, the Nickell Bias, and an Indirect-Inference Correction, with Notes on India

I estimate how the ratio of bank nonperforming loans (NPLs) to gross loans responds to real GDP growth in an unbalanced panel of 140 economies (2,041 observations, 2001–2024) built from World Bank WDI data (FB.AST.NPER.ZS and NY.GDP.MKTP.KD.ZG; source update 8 October 2026). The two-way fixed-effects estimate is −0.10 percentage points of NPL ratio per percentage point of contemporaneous growth (clustered s.e. 0.04), with a similar coefficient on lagged growth. I derive the Nickell bias in the persistence coefficient and confirm by Monte Carlo that the exact formula predicts the simulated bias at these panel dimensions to within 0.004–0.006. An indirect-inference correction moves the persistence estimate from 0.73 to about 0.90, which raises the implied long-run effect from −0.78 to about −2.2. India's NPL path, by contrast, is not explained by growth: the WDI ratio rose to 9.98% in 2017 while growth exceeded 6%. The RBI's June 2026 Financial Stability Report puts the scheduled commercial banks' gross NPA ratio at 1.8% in March 2026.