Quantitative Analysis: It’s (Not) All About The Numbers

May 7, 2018

I’m a big fan of the work of Cliff Asness and AQR (Applied Quantitative Research).  Cliff was recently on Bloomberg Radio with Barry Ritholtz. You can listen here to his Feb 21st interview.

About twelve minutes in Barry asked Cliff what it means to be a quantitative investor in which Cliff replied that “quantitative investment managers are about two things – averages and diversification.” Diversification is broadly accepted in finance, often called “the only free lunch” in investing.

Quant managers care about what works in trading and investing and what has statistically significant historical evidence to prove it. Value, momentum, and trend to name a few. Many in finance hold their beliefs about one particular style of investing (for example “active vs passive”) so closely that it’s like debating politics. To paraphrase Meb Faber, “many people have a hard time having two different points of view on investing in their head at the same time without it exploding.”

But quantitative analysis is not only about performance data. When done properly only modest and general conclusions should be drawn about the likelihood of the future to be like the past. Let’s look at an example:

Portfolio visualizer is a nice web site for backtesting passive asset allocations as far back as the early 1970’s. I took a few minutes to create a simple asset allocation of the following:

15% US equity

15% Foreign equity

15% TIP’s

25% Treasuries

15% Corporate bonds

5% REIT’s

5% Gold

5% Commodities

Cliff also talked about how quant managers like to do in sample and out of sample testing to confirm the validity of an investment strategy on both past data that you have seen as well as data you haven’t. This eliminates or at least reduces hindsight. Let’s go back in time a few years and assume for a moment it’s 12/31/2007. From 1973 to 2007 this simple asset allocation, rebalanced annually, produced some very impressive results (gross of fees and taxes as this is simply index data. Past performance doesn’t guarantee future results, and that’s one of the major points in this post):

Annualized return: 11.34%

Annualized volatility: 7.82%

Sharpe Ratio: .65

Worst Year: -3.65%

Here is the wrong way to review this information: “Wow, this portfolio makes over 11% per year, 32 out of the last 35 years, and can only lose -3.65%? I can easily stomach that kind of risk! I know my advisor told me not to focus too much on the past or something, but I wasn’t really listening because I was focused on the numbers and this looks as good as guaranteed to me!” 

Now let’s look at 2008-2014 and consider it “out of sample” results since at the time we didn’t know how it would work out:

Annualized return: 6.06% (47% lower)

Annualized volatility: 10.54% (35% higher)

Sharpe: .57 (12% lower)

Worst Year: -12.21% (234% higher)

Since 2008, every performance measurement has been worse than 1973-2007. But is there actually anything “wrong” with this portfolio? Not at all. 2008 was a historic global sell-off in asset classes and 7 years is a relatively small sample size. The problem that investors run into is focusing WAY too much on past performance without taking into consideration market conditions from a much higher level.  Humans tend to think the recent past will continue on forever, which is known as recency bias. Now analyzing this simple portfolio over the entire period of 1973-2014 gives an annualized return of 10.44%, and it will continue to change every year! Just like reviewing it only from 2009-2014 produces an annualized return of 9.46%. So is this portfolio going to make 6%, 9%, 10%, or 11% annually going forward? Probably all of them, it will just depend when you look at it. The only thing I’d be pretty confident in is that a portfolio like this will continue to provide real returns over the long term. Quantitative analysis can be an extremely powerful process for building evidence based investment strategies and portfolios. But there is a right way and a wrong way to do it.

“Thus timing, and in particular the selection of the beginning point and end point for studying a performance record – plays an incredibly important role in perceptions of success or failure” -Howard Marks

“No strategy is so good that it can’t have a bad year or more. You’ve got to guess at worst cases: No model will tell you that. My rule of thumb is double the worst that you have ever seen.” -Cliff Asness, AQR

“When investing over the long run, all you can have confidence in is that…holding assets should provide a return above cash…That’s it. Anything else (asset class returns, correlations, or even precise volatilities) is an attempt to predict the future.” -Ray Dalio, Bridgewater Associates

“You must be rigid in your rules and flexible in your expectations. Most traders (and investors) are flexible in their rules and rigid in their expectations.” -Mark Douglas