Why historical vola for short periods is almost worthless

[nitro]

Compute the annualized HV of this data:

AdjClose 38.24
AdjClose 40.29
AdjClose 41.46
AdjClose 41.52
AdjClose 42.71
AdjClose 39.15
AdjClose 40.32
AdjClose 42.4
AdjClose 41.23
AdjClose 41.43
AdjClose 41.28
AdjClose 42.62
AdjClose 43.21
AdjClose 40.88
AdjClose 37.42
AdjClose 38.44
AdjClose 39.16
AdjClose 40.66
AdjClose 41.16
AdjClose 40.99
AdjClose 42.64
AdjClose 42.18
AdjClose 41.67
AdjClose 42.72
AdjClose 40.9
AdjClose 39.55
AdjClose 40

Now compute the hv of this data, windowed one day ahead:

AdjClose 40.29
AdjClose 41.46
AdjClose 41.52
AdjClose 42.71
AdjClose 39.15
AdjClose 40.32
AdjClose 42.4
AdjClose 41.23
AdjClose 41.43
AdjClose 41.28
AdjClose 42.62
AdjClose 43.21
AdjClose 40.88
AdjClose 37.42
AdjClose 38.44
AdjClose 39.16
AdjClose 40.66
AdjClose 41.16
AdjClose 40.99
AdjClose 42.64
AdjClose 42.18
AdjClose 41.67
AdjClose 42.72
AdjClose 40.9
AdjClose 39.55
AdjClose 40
AdjClose 41.58

 

[spindr0]

No need to do the math. If you're doing short term calculations (volatility, simple moving averages, etc.), any time you add a small amount with the new day while taking away a larger amount via the removed day (or vice versa), there's going to be a decent amount of variance in the daily results. If the results are annualized, that further magnifies the differences. That's just the way it is.

 

[nitro]

Quote from spindr0:

No need to do the math. If you're doing short term calculations (volatility, simple moving averages, etc.), any time you add a small amount with the new day while taking away a larger amount via the removed day (or vice versa), there's going to be a decent amount of variance in the daily results. If the results are annualized, that further magnifies the differences. That's just the way it is.

Do yourself a favor and compute the HVs above. The variance is absurd.

 

[asiaprop]

because you probably calculated the standard deviation on the prices themselves. Calculate the vols on the returns (log returns) and you will see that the difference boils down to about 0.06% between those 2 different time series.

Quote from nitro:

Do yourself a favor and compute the HVs above. The variance is absurd.

 

[nitro]

If I did I'm an idiot because that is certainly what I meant to do.

Lemme check...

Quote from asiaprop:

because you probably calculated the standard deviation on the prices themselves. Calculate the vols on the returns (log returns) and you will see that the difference boils down to about 0.06% between those 2 different time series.

 

[timbo]

If not for the AdjClose, I'd would've done the math. But statistics is useless against unknown parms.

 

[Occam]

I got an annualized vola of .607 for the first and .597 for the second.

 

[asiaprop]

yes, about, depending on how you calculate the return measure. Not sure what the problem really is...

Quote from occam:

I got an annualized vola of .607 for the first and .597 for the second.

 

[MasterAtWork]

Quote from occam:

I got an annualized vola of .607 for the first and .597 for the second.

Back of the envelope calculation seems to valid your numbers.

What matters with those datas ?

 

[asiaprop]

for the record, the 0.06% spread was on the daily vol.

Quote from asiaprop:

because you probably calculated the standard deviation on the prices themselves. Calculate the vols on the returns (log returns) and you will see that the difference boils down to about 0.06% between those 2 different time series.