This paper investigates how and when banks underreport market risk. We study hand-collected data
on modelling and disclosure choices and examine how they relate to the level and accuracy of predicted
Value-at-Risk (VaR). We find that more elaborate modelling and more transparent reporting can correspond
to more accurate and/or conservative VaR-predictions. Using longer-than-required historical data
and relying on external scaling to approximate required holding periods, however, also enables distressed
banks to report lower and/or less accurate VaR. Monte Carlo (MC) simulation methods, instead, which
are generally more accurate, seem to be abandoned over time.