Read on Twitter
@alphaarchitect asked what happened to style Premia in 2017.
Let's start with the fact that something happened.
Like...something very material.
Let's start with the fact that something happened.
Like...something very material.
These are CUMULATIVE ALPHAS (compounded residuals from regressions of fund returns on IVV ETF returns setting intercept to 0) From Jan 2018 through end of October. I scaled all lines to have the same volatility (10% annualized) for easy comparison.
My analysis suggests there are 8 independent factors among these 13 funds based on the diversification ratio.
Here is what I get when I create a portfolio with equal risk exposure across these 8 independent factors
Here is what I get when I create a portfolio with equal risk exposure across these 8 independent factors
You might be thinking, “Ok fine, it’s been quite a string of bad luck but there’s no reason to believe ‘things are different this time’."
Which prompted me to wonder, “What is the distribution of expected portfolio trajectories for a portfolio of 8 uncorrelated time-series...
Which prompted me to wonder, “What is the distribution of expected portfolio trajectories for a portfolio of 8 uncorrelated time-series...
cont/...(i.e. factors) with similar volatility to the 8 uncorrelated factors under investigation, and where we expect each factor to have at least 0 compound mean over the long-term?”
So I performed the following experiment:
Randomly generate 8 uncorrelated “factor” return series with 0 compound returns and the same standard deviation as the factor components of the index above
Equally weight the 8 factors in a portfolio
Randomly generate 8 uncorrelated “factor” return series with 0 compound returns and the same standard deviation as the factor components of the index above
Equally weight the 8 factors in a portfolio
Simulate the trajectory of the portfolio over the same 709 daily periods as our actual factor portfolio above
Repeat 100,000 times.
In the following chart I plot a cone that describes the 1st percentile, median, 99th percentile and worst result from 100,000 samples.
Repeat 100,000 times.
In the following chart I plot a cone that describes the 1st percentile, median, 99th percentile and worst result from 100,000 samples.
You can see that, not only does our actual portfolio of 8 factors (purple line) project below the 1st percentile (light blue) line of our cone. It also projects far below the *worst* draw out of 100,000 simulations.
In other words, it is statistically inconceivable that 8 independent factors with an average positive population mean would produce the actual return trajectory that we’ve witnessed over the almost 34 months since January 2018.
Therefore, we must assume our independent factors have a negative average population mean.
Once you’ve eliminated from consideration the possibility that this is just a statistical fluke, are we not forced to accept the alternative that “things are different this time?”
Once you’ve eliminated from consideration the possibility that this is just a statistical fluke, are we not forced to accept the alternative that “things are different this time?”
With this analysis in mind, I invite you to explore the thesis presented in this video: https://investresolve.com/are-we-living-in-a-post-factor-world/
