Evaluating The Efficacy Of The Model That’s Flashing Green On Many 2040-FANMAG Themed ETFs

My last post discussed the Thematic Global ETFs and their potential to point today’s investors toward stocks that may become the FANMAG of 2040. I provided access to a spreadsheet identifying 97 of them and in the text of the article, I named 16 ETFs as being rated Bullish or Very Bullish under the Chaikin-Analytics-PortfolioWise Power Ranks we use to evaluate the relative future performance potential of ETFs. (The list of favorably-rated ETFs will be reproduced below.) My goal today is to explore the nature and current efficacy of the model we use to rate these ETFs.



Ranking ETFs For Probable Future Performance

For starters, here is what we do not do: We do not take various measures of historical price changes, sort, and say the largest changes are best. We, like many or most others, display such information, but it is strictly informational. We agree wholeheartedly with the well-known mantra that past performance does not assure future outcomes.

But we, like everyone else, want to predict the future even though we and the rest of humanity lack the gift of extra-sensory psychic vision of tomorrow. (Actually, I do have a crystal ball that is forecasting blowback from the 1-800-whatever psychic community! Whatever.) The goal here is to work with probabilities; to identify factors we believe, based on the combination of theory and empirical observation (with a particularly heavy dose of the former — we do not practice “curve fitting” or “data mining”).

Click here for a “white paper” explaining the Power Rank model in detail.

Starting With The Stocks

Where an ETF consists of US equities, it makes sense to start with the Chaikin Power Gauge ranks we compute for stocks. It seems reasonable to assume that the Power Rank for an ETF like the SPDR S&P 500 ETF (SPY), which is a portfolio of stocks, will be the average of the ranks of the individual stocks weighted as they are in the SPY portfolio.

Actually, it’s not so simple. If you’ve taken a finance class, chances are you are already primed to accept the notion that the whole may not equal the sum (or even the weighted sum) of the parts. If you don’t believe me, answer on an exam that Portfolio Variance is equal to the weighted average of the Variances of the individual securities and see how the instructor reacts. (If you’d rather not walk the plank to find out, click here to see the correct formulation, which requires more information.). 

Beyond that, when we get into advanced statistical methods of analysis, there is something known as “residual error,” which refers to the impact of unknown factors not accounted for in the model. When quants do what quants do, they try to get residual error as close to zero as possible and often assume it out of the picture by treating. it as a random thing that over the long term averages to zero. That’s all well and good — until you use a model to invest real money in which case you find that residual error is neither random nor zero and can cause a lot of financial pain.

We Need More

So, while we use stock-specific Power Gauge ranks in our ETF model, we go a lot — a heck of a lot — further. Under ideal conditions (i.e., the ETF owns only US stocks that are rankled by us), the portfolio weighted average Power Gauge rank is only 20% of the overall ETF rank. Another 20% of the rank is based on a different sort of analysis of the distribution of ranks within the ETF portfolio. And 60% is based on our ETF Technical model. (Yes, even for an ETF like SPY, 60% of its rank is based on technical indicators). 

1 2 3 4
View single page >> |

Disclosure: I am long ARKF, PAWZ.

How did you like this article? Let us know so we can better customize your reading experience.


Leave a comment to automatically be entered into our contest to win a free Echo Show.