Volatility - what period?

Doing some volatility calculations, correlations and sharpe ratios. What do you usually use for time period, I have been taught to use 36 or 60 months data but sometimes it seems a little bit long time period since volatility and correlations often change over time? Should I reduce the time period and use weeks instead?

There’s no such thing as “volatility”, “correlation” or “Sharpe”. You have to define a time range and a sampling period. So, there is such a thing as “30-day daily correlation”, however. Naturally the 360-day daily correlation may be different than the 180-day daily correlation and the 360-day weekly correlation.

What are you talking about, “there is no such thing…” ? You obvisouly have no clue so why do you bother? Historical relationships are often used to estimate future parameters for instance when building an efficient portfolio. I was looking for some insight from people that do this on a regular basis. The time interval and sampling period is what I am actually asking about. Naturally the standard deviation would be annualised, the sharpe ratio would be for a certain time period.

I don’t think there is really a right answer to this question. Maybe you could graph your data and try to see if there is a perceptible pattern. For instance, if you see that there is a 5 year cycle that tends to repeat, using 60 month data is the obvious choice. Personally, I would say to err on the side of taking a longer sampling period to include more information. You could also take a long time period but weight older periods progressively less. What kind of asset prices are you looking at?