Analyzing the size, diffusion, and spillover of loans risk

Abstract We analyze the diffusion and spillover effects of credit risk among banks within a banking system, using the Mexican financial system as case study. Our proxy to measure credit risk is the non-performing loans ratio (NPL). For this purpose we construct a VAR model to identify the composition of the variance of NPL's ratios dividing it into two parts: one that is explained by the VAR coefficients, and the other attributed to the contemporary "error" or "shocks" on other banks in the system. The error in the structural model represents the "news" that disturbs the stable risk in each period. Our work builds on the spillover index proposed by Diebold and Yilmaz (2009) that indicates the degree on which the overall risk in the system is explained by the spillover effects. The method allows us to measure the long-run contributions of each bank's risk on the rest of the banking system through the diffusion of risk between intermediaries. Moreover, we are able to gauge the relative importance of spillover by increasing the length of prediction periods for each bank's NPL. Our estimations for the Mexican banking system between 2002 and 2013 suggest that the overall spillover effect index accounts for 15 percent of the aggregate risk's observed variation in the short term and almost 40 percent in the long term. The spillover effect explains 32 percent of total risk in the short term and 78 percent in the long term when we control for individual bank's characteristics, even though the total size of risk originated by news in the banks decreases relative to the model without control variables.