Quantile relationships between standard, diffusion and jump betas across Japanese banks

The ability of the banking sector to absorb unexpected news is critical to its ability to disseminate relevant information to the financial markets and real economy. Using high frequency financial data and quantile regression techniques we characterise some stylised facts about standard betas, diffusion betas and jump betas and the relationships between them for Japanese bank stocks and portfolios. Data are for the period 2001–2012 at 5-min frequency. We find that jump betas, which relate to the arrival of unexpected news, are on average higher and more dispersed than the diffusion betas across Japanese banks. While the standard beta is a weighted average of the diffusion and jump betas, the magnitudes of the weights differ significantly across quantiles (the 5th, 25th, 50th, 75th, 95th), indicating a non-linearity in how jump information is incorporated across the quantiles. On average, small bank stocks and portfolios have smaller diffusion betas and smaller jump betas than large bank stocks and portfolios. While there are no significant differences between the jump-diffusion beta ratios when conditioned by market capitalisation, during times of financial crisis small banks have significantly higher jump beta-diffusion beta ratios than large banks on average, indicating that during time of financial crisis smaller Japanese banks face much higher relative jump risks than larger Japanese banks.