Long run peso/dollar exchange rates and extreme value behavior: Value at Risk modeling

摘要

We apply an extended VaR integrating a generalized extreme value distribution to estimate potential losses from investing in the peso/dollar exchange market using daily data for the period 1970-2007; the block maxima approach is used to minimize impact from dependency in prices due to the presence of heteroscedastic-ity. Estimations are presented for short and long positions. Our evidence confirms the potential of the GEVD to explain the extreme behavior from exchange rates. It also supports the hypothesis that EVT is a more precise and conservative approach estimation than conventional VaR. Backtesting is used to gauge robustness of the results.