Defining a One-Month Horizon for Historical VaR

Summary

The discussion explains how to interpret a one-month horizon when estimating historical Value at Risk from a year of equity portfolio history. The proposed approach is to calculate rolling portfolio returns over a fixed interval at each historical date, rather than treating each calendar month as one observation.

The accepted reply endorses using 22 trading days as a practical approximation for one month, based on the rough conversion from annual business days to months. This avoids variation in the number of working days across calendar months. The response is brief and offers no comparison of alternative horizon definitions, sample-size implications, or empirical validation, so the convention should be understood as a standard approximation rather than a universal rule.

Key ideas

• Historical VaR with a one-month horizon uses rolling returns measured across the chosen holding period.
• A fixed 22-trading-day window is presented as a common approximation for one month.
• Using a fixed trading-day horizon avoids differences in calendar-month business-day counts.
• The response provides a convention but no empirical comparison with other horizon choices.

Tags

# What does "1-month horizon" mean in VaR calculations


# What does "1-month horizon" mean in VaR calculations












For a simple equity portfolio, I need to compute historical 1-year simulation VaR with a "1-month horizon".

I imagine this means I need to compute the 1-month return of the portfolio at each point in my 1-year history, eg. what is the 1-month return on the 1st of January, on the 2nd of January, on the 3rd, 4th, etc. Is that correct?

And if yes, does it make sense to use a 22-day horizon to eliminate any silly biases such as some months having more work days than others? (I get the 22-day window because there are roughly 260 business days in a year, divide that by 12 to get roughly 22)

## Answer by Alex C (score 1, accepted)

https://quant.stackexchange.com/a/33352

yes it is correct and yes it makes sense and is fairly standard to use 22 days for "1 month".

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.