4 – M2L5 06 Rolling Windows V3

So, we know how to calculate volatility for a given period. This gives us one number to capture return variability for this period, but we also know that the market changes over time. We know the market goes up and down, and while sometimes it seems highly variable, other times it seems relatively placid. For example, during the last financial crisis in 2008, stock prices were extremely volatile. Whereas 2016 and 2017 have marked a relatively low volatility period. So, what do you do if you want to understand how volatility is changing over time? One option is to use a rolling window. So, if you wanted to estimate today’s volatility, you would calculate the standard deviation of a set of log returns from sometime in the past up to yesterday. This is a practical method that is commonly implemented, but how long should the time window be? When you take a look at the volatility of previous days to estimate today’s volatility, how important is the volatility from one day ago? How about from one week ago? Or how about from one month ago? A long window will mean that the value you compute may not react to changes in market conditions quickly enough. To short window may mean that the computed value becomes two variable over different window periods to be used reliably. Your choice of time window is going to depend on your application. In general, if your signal or strategy involves holding onto purchases for long periods, you can afford to use a longer time window for computation and vice versa for short holding periods. You can also use windows of different sizes to gain insight into how volatility evolves.

%d 블로거가 이것을 좋아합니다: