An empirical investigation of the quality of value-at-risk disclosure in Australia

We study the level and quality of value-at-risk (VaR) disclosure at Australian banks. We find that Australian banks have increased disclosure about their VaR recently, reaching a level post-crisis that is similar to other regulatory jurisdictions. We find that the actual VaR estimates produced by banks are generally rejected by standard backtesting procedures. During quiet periods bank VaRs are too high, while during high volatility stress periods bank VaRs are too low. We are able to reject the null hypothesis that the daily VaRs for two banks are the 1st percentile using a quantile regression-based test.