Lending Intensity, Funding Composition, and Liquidity Risk Persistence in Sri Lankan Commercial Banks: A Dynamic ARDL Analysis

Abstract

This study examines the dynamic effects of lending intensity and funding composition on liquidity risk persistence in Sri Lankan licensed commercial banks, using monthly balance sheet data spanning December 1995 to November 2024 and an autoregressive distributed lag framework. Employing an autoregressive distributed lag (ARDL (2,2,1,2) framework, the study captures short-run adjustment dynamics of lending intensity, borrowing dependence, and loan growth on liquidity risk. Unit root tests confirm all variables are integrated of order one I(1), validating the ARDL approach. The bounds cointegration F-test (F = 0.923) does not confirm a long-run equilibrium relationship over the full sample, a finding attributed to parameter instability identified by the CUSUM test (Figure 1, S = 5.560, p < 0.001), reflecting structural changes in Sri Lanka's banking and regulatory environment over the sample period. Short-run results reveal that liquidity risk is highly persistent, with AR coefficients summing to 0.989. Lending intensity exerts a significant negative contemporaneous effect, followed by a delayed positive reversal, indicating maturity transformation risk. HAC-robust standard errors confirm the significance of lending intensity and liquidity persistence under heteroskedasticity. Breusch-Godfrey tests confirm the absence of serial correlation at any lag. These findings underscore the importance of disciplined asset–liability management and stable funding structures for enhancing liquidity resilience in Sri Lanka's banking sector.