cancel
Showing results for 
Search instead for 
Did you mean: 
     
Participant ○○

CEF volatility

Historical volatility (the standard deviation of returns) is commonly used to predict future volatility. For PDI, and probably for other CEFs, a reasonable predictor (using maximum likelihood) of 1-day volatility is the 20-trading-day historical volatility, with a decay factor of 0.9 that gives progressively less weight to older squared returns (an exponential moving average similar to what RiskMetrics does).

When I apply this volatility model to PDI, using data from 5/12/2012 to 3/24/2020, statistics on daily predicted unannualized volatility of percent returns are as follows:

       count     median       mean         sd        min        max       last
        1968     0.5877     0.7058     0.5747     0.1275     8.2598     7.9128

So your median prediction of the standard deviation of PDI 1-day returns would be 0.6%, but at the close of 3/24 you would have predicted a 1-day volatility of 7.9%. In the past few days, 1-day moves of 10% or more in either direction have been common in many CEFs, so such a prediction is plausible.

I am thinking about what role CEFs should have in a portfolio when their volatility can change by a factor of 10 or more, moreover at the worst time (when stock and bond market volatility has also spiked). Here is how predicted daily volatilities evolved in the recent past.

  Date       %volatility
02/03/2020 0.25 02/04/2020 0.32 02/05/2020 0.31 02/06/2020 0.31 02/07/2020 0.29 02/10/2020 0.28 02/11/2020 0.28 02/12/2020 0.29 02/13/2020 0.33 02/14/2020 0.31 02/18/2020 0.31 02/19/2020 0.30 02/20/2020 0.28 02/21/2020 0.27 02/24/2020 0.25 02/25/2020 0.36 02/26/2020 1.55 02/27/2020 1.81 02/28/2020 2.44 03/02/2020 2.62 03/03/2020 3.33 03/04/2020 3.32 03/05/2020 3.39 03/06/2020 3.31 03/09/2020 3.14 03/10/2020 3.75 03/11/2020 3.60 03/12/2020 4.07 03/13/2020 5.61 03/16/2020 6.00 03/17/2020 7.06 03/18/2020 6.72 03/19/2020 8.26 03/20/2020 7.95 03/23/2020 7.61 03/24/2020 7.91

In early February, predicted volatilities were in the 0.2%-0.3% range, about 30 times smaller than the recent values. Even at the beginning of March, before most of the CEF carnage, the volatility model would have warned you that risk was much higher than usual (2.6% vs. the historical mean of 0.7%). Considering the elevated volatility, I had positions that were too big last week.

 

1 Reply
Highlighted
Participant ○○

Re: CEF volatility

Interesting work. Although we get a day-to-day 'feel' for the changes in volatility, your approach makes it possible to observe the trend of growing volatility. It could be a useful early warning indicator that would be easier to act on, compared with just having an uneasy feeling because volatility is up.

Great stuff - thanks. I don't know it it would be realistic to turn this into a an indicator to go along with your recent CEF trading charts. Something along the lines of a slope of volatility change marker to go along with the median would be very useful. 

0 Kudos
Announcements

Morningstar is here to help you respond to the Coronavirus crisis.