Time horizon

Ideally, the time interval between the data points used to compute the covariance
matrix should agree with the time horizon used in the VaR calculation. However,
this is often impractical, particularly for longer time horizons. An alternative approach
involves scaling the covariance matrix obtained for daily time intervals. Assuming
relative returns are statistically stationary, the standard deviation of changes in
portfolio value over n days is n times that over 1 day.
The choice of time horizon depends on both the nature of the portfolio under
consideration and the perspective of the user. To obtain a realistic estimate of
potential losses in a portfolio, the time horizon should be at least on the order of the
unwinding period of the portfolio. The time horizon of interest in an investment
environment is generally longer than that in a trading environment.