Jan

31

After writing two pieces for DailySpec (Link1. Link2.) I've become a little bit obsessed with the topic of low beta vs high beta stocks.

I reported there and confirm here that:

1) high beta, high risk stocks tend to look bad in studies of their compound growth, i.e. what would happen if you kept 100% of your portfolio there, but that

2) they actually look fine in terms of their "excess return" or "alpha", which is the proper (according to the "Capital Asset Pricing Model") way to measure returns on a risk-adjusted basis

Here's the expression used for "alpha" or "excess return":

alpha = R - Rf - beta * ( Rspy - Rf) ,

where R is the % return of the stock, Rf the percent return over the same period of a "risk free" asset (read "T-bills"), and Rspy the return of the S&P 500 index ETF ticker SPY.

The universe used here is the 500 largest market cap US-domiciled stocks, trading on either NYSE or Nasdaq, selected at the start of each calendar year in the study. The list includes tickers that have since vanished one way or another, whether by bankruptcy, merger, or other event.

In order to characterize the dependence of forward alpha on trailing beta, I measure, for each month, the correlation (sometimes called the "information coefficient") between each stock's alpha/excess return and its trailing beta, measured over the trailing 250 trading days. The results are shown in the attached table. With measurements for a bit less than 500 stocks each month (I exclude stocks with share prices less than $5 and stocks without a trailing 250 trading day history), the statistical uncertainty in the measured correlation is approximately 500^(-1/2), or 4%. Data ranges from May, 2001 through December, 2011 — 128 months.

A word about signs and magnitudes: the table shows correlation between forward alpha and trailing beta, so a negative sign for a given month would indicate that low beta stocks had outperformed. Regarding magnitude, I'd say that the one would require a magnitude of at least something like 10% for the effect to be meaningful and "actionable".

The table shows that the average of the 128 monthly correlation measurements is -0.5%, with a t-score of -0.3. This is most definitely a null result. So over this period, high and low beta stocks did, on average, roughly what the capital asset pricing model / "efficient market" theory would predict.


 

—–Also attached is a plot of the cumulative sum of the monthly correlations. When the cumulative sum is decreasing (increasing), that means that low (high) beta stocks are outperforming. The plot snakes around but in the end doesn't go very far. Low beta stocks do very well over the ~2004-2005 window, but then give it all back and more from 2008-2009. (It may be surprising that high beta outperformed low beta over the very bearish 2008-09, but first, remember that we're talking about risk-adjusted "alpha" here, and second, during 08-09 saw 50-ish percent declines even among not-so-risky stocks.)

So I don't think that the enthusiasm for low-risk, low volatility equities, as seen in the popularity of ETFs like SPLV, is deserved. It makes more sense to buy the whole market, SPY, and be tax-efficient. If you want lower volatility, then just buy less of it.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


Comments

Name

Email

Website

Speak your mind

1 Comment so far

  1. Gary Phillips on February 2, 2012 7:20 pm

    With markets so inextricably correlated it appears like average beta = average returns - transaction costs. Nash equilibrium comes to mind; damned w/low b, damned w/high b, not that much better with average b.

Archives

Resources & Links

Search