Proper Procedure for Calculating Sortino Ratio

[ttubson]

I'm writing VBA code for calculating the Sortino ratio and I ran into a speedbump. There are a ton of different sites all saying different things about how to properly calculate the Sortino ratio. So, when I calculate the downside standard deviation, should I divide the sum of the squares by the total number of observations in the series or just the observations below the minimum acceptable return? My gut wants to just take the observations below the minimum acceptable rate or return and divide by (n-1), since it seems like that would give me a more accurate estimate of the downside volatility. But I want to get an educated opinion....

 

[Brad Warren]

You should use the total number of observations. The downside risk should be less if there is a smaller number of observations below the target return.

 

[aaronhotchner]

Numerator is real return less target return. Denominator is (average squared downside difference)^(1/2), where returns above the target are considered to have 0 difference -- you are only considering downside standard deviation. I'm not even sure you'd divide by (n-1) instead of n since you are (presumably) not sampling within your time horizon.

 

[Sim Con]

If you're still looking for VBA for the Sortino Ratio, then you can find some here