In this post I want to briefly return to putting together quantitative strategies into a an overall portfolio. I wrote about this in 2014 but I have better tools and more data now. Basically let’s build a portfolio of quant strategies that reflects a typical 60/40 US stock US bond benchmark and compare portfolio statistics to the SP500 and to the 60/40 benchmark.
First things first. Picking the quant strategies (you can find the background to all the strategies in the Portfolios section of the blog). You can definitely spend a ton of time here and go way off into the weeds. Lets keep it relatively simple. I will choose 4 strategies. Why 4? Well, the tool only allows me 5 assets and one of them has to be the bond allocation in the portfolio. And 4 is plenty. I also want some value and some momentum. I’ll choose the two most conservative quant strategies I’ve presented, the Utilities Value Strategy and the Consumer Staples Strategy. Then I will add the Trending Value Strategy and the Pure Momentum Strategy. The strategies will buy and hold a portfolio of stocks for one year, then re-balance. The strategies will hold 10 stocks each. While the strategies on a stand alone basis normally hold 25 stocks in a portfolio it would lead to too many stocks for no benefit. Actually holding fewer stocks leads to some benefits – higher performance for the individual strategies and some diversification benefit when value and momentum strategies are combined (see this great post from Alpha Architect). We are targeting a 60/40 allocation, so we’ll hold 40% of the portfolio in US gov’t bonds (IEF) and the 60% equally divided among the 4 quant strategies (15% each). Finally, we’ll re-blance the entire portfolio once a year.
Below are the results for the backtest of the 60/40 quant strategy US bond portfolio from the beginning of 1999 through about mid November of 2016 compared against the SPY.
Returns are 13.7% a year vs about 6% for SPY but with less than half the drawdowns which leads to much a higher Sharpe ratio. In the same time frame the 60/40 US stock bond portfolio has returned about 7.4% a year with -29.5% in drawdowns. So the diversified quant bond portfolio delivers better than equity returns at drawdown levels of a classic 60/40 portfolio. Impressive to say the least. Two more things I want to touch on before I conclude; correlations and implementation costs.
If you analyze each individual quant strategy you may have notice the large drawdowns, especially in the more concentrated 10 stock portfolios. What’s great when you put them in a portfolio is the diversification benefit you get. Below is the correlation table from the portfolio simulation.
Bonds are the best diversifier by far and are what keeps the portfolio drawdown at the 60/40 type level. But the quant strategies help also. The highest correlation among the strategies is at a pretty low level of 0.59. Basically, it’s important to keep your focus on the portfolio level and not too much at the individual strategy level.
Finally, let’s talk about implementation costs. In this portfolio we are talking about 40 stocks re-balanced once a year. If you assume the entire portfolio turns over once a year, you’re looking at 80 trades a year. At a high trade cost of $10/trade you’re looking at $800 in commissions. That’s 0.8% for a $100K portfolio and lower for larger portfolios. If you do a bit of searching on trade costs you can get this much lower, even for free in some cases. But even $5/trade is not too hard to find these days. Slippage can also be kept to a minimum. The portfolio simulation buys at the close of trading. You can use MOC or last minute conditional orders to minimize the difference between the buy/sell price and the actual closing price. You also need to be slightly careful to choose a broker with good execution.
In summary, it is relatively straight forward to build a portfolio of individual quant strategies and US bonds that has significantly better performance characteristics than the classic 60/40 portfolio and even 100% equity portfolios.
12 Comments
Don Thompson · November 22, 2016 at 9:12 am
Paul, that’s fantastic. I just love your posts. Like you, I’m in the withdrawal phase. So, minimizing drawdowns is important to stick with the program. This seems even more critical today with stock prices so high. We’ve got a great start with having only annualized rebalancing. Do you plan to research how to minimize the drawdowns, possibly on a monthly basis? Recently, you’ve explored being in or out of the market, unique to each quant, based on employment and other factors like 200 day MAs . Drawdown limits per quant could also be applied. Do you plan to add this type research? And what about the effects of changing from 60/40 to other ratios?
paul.novell@gmail.com · November 23, 2016 at 5:38 am
More frequent rebalancing of the entire portfolio, i.e. between strategies and bonds, doesn’t improve performance or drawdowns. But using other techniques like the SPY-UI indicator within the quant strategies does, by quite a bit. I’ve talked about this already among individual quant strategies. It’s a similar effect for the others. Finally, changing from 60/40 to other ratios has the impact you would expect. More bonds, less performance lower drawdowns, and vice versa for less bonds. So, an investor can dial in whatever level of risk they could tolerate.
Paul
Damian · November 22, 2016 at 10:30 am
Paul, thanks — appreciate this and all your other fine work. You probably didn’t want to do it for fear of it appearing like you’re promoting the tool, but what is the referenced “tool” that allowed this? Did P123 allow you to run that correlation matrix?
Also, is it a misinterpretation of your prior writings to say that one could achieve even better risk-adjusted returns with this portfolio by (i) using a trend-following strategy with bonds and/or (ii) using the SPY-UI timing indicator (or something else)? Is there a way to easily backtest that?
Basically, if it makes sense to use those additional optimizations, wouldn’t one want to, assuming it’s as easily as getting weekly emails from P123 to implement the quant model changes once per week and keeping an eye on whether the UI-SPY indicator is triggered (which would require minimal checks as well)?
Basically, I’m (i) trying to understand why anyone wouldn’t make those relatively simple optimizations and (ii) trying to understand the cheapest, most direct way to go from theory to practice here. In this latter regard, it seems like P123, if offering relevant quant strategies as “ready to go” models, would require only a relatively inexpensive Investor-level subscription, but if it didn’t, you’d have to step up to the Screener or Designer levels.
paul.novell@gmail.com · November 23, 2016 at 5:50 am
Hey Damian, great questions. I didn’t mention P123 because I’ve mentioned it so many times in the past. It’s the tool I use for all this stuff. Yes, I did everything for this analysis in P123.
No, it’s not a mis-representation to say better risk-adjusted returns are possible with some ‘tweaks’. The biggest bang for the buck would come from using the SPY-UI indicator within each quant strategy. I can and have backtested that. It works great. Don’t know about using the trend following bond strategy for the bond component. As you allude to I can’t backtest some of this stuff. The bond ETFs don’t go back that far which always makes me wary.
As to your last two questions; 1. adding those simple optimizations makes things more complicated and increases the work load to implement these strategies. It may be worth it for some. What I presented was the most basic and simple, takes little work, and has incredible results. And yet 90%+ of investors won’t even do that. 2. Cheapest way is to do it yourself. As for other ways, I am considering either Ready to Go models and a separate business to offer these strategies.
Paul
Douglas Nashif · November 22, 2016 at 11:03 am
Hi Paul,
Of course I remain grateful for your work and sharing of knowledge. I have been a great benefactor for some time now.
Knowing full well that my weakest trait as an investor is to second guess strategies of all kinds, doesn’t the current route of the bond market (30 year bull run) as well as being at historic highs in the stock market (very late for a 7 year BIG credit cycle selloff) cause concern for both the stock and bond markets?
Yes historically these strategies would not have worked for people that pull out and enter back on their own judgement. But when I look around now I honestly see perhaps the worst time to history to invest in stocks or bonds.
This comment and entire post to your excellent blog is really is off topic but as a person who would like to get into the markets and stop trading them I can’t seem to do it. Like so many I have become convinced that world leaders and other ultra rich and very powerful concerns have ruined the markets and rigged them to secure their own power pray on the masses.
hate to say it but when I see a 20% correction that will be a good time to dip my toe in with 10%-20% of my cash. Until then I’ll trade futures, covet my gold and silver, and wait for this market to recognize its true (lack of solid) underpinnings. 🙂
Best regards and happy Thanksgiving sir!
Doug
paul.novell@gmail.com · November 23, 2016 at 5:55 am
Hey Doug, that is one path. And I totally understand the reasoning. But I would encourage you to at least try a more quant/rules based approach. I chose to start walking away from the pure discretionary path almost 5 years ago. Now, I am primarily a rules based investor. As of this summer when I did my last re-balance I am less than 15% discretionary or even buy and hold. I focus my attention on choosing strategies that have withstood the test of time over all kinds of economic and political conditions. My results have been markedly better an more importantly the stress of investing has gone way way down. It works for me. And I think it would good for more investors.
Good luck with your path. Happy Thanksgiving.
Paul
Scott · November 22, 2016 at 11:38 pm
Hi Paul,
Thank you for the excellent article. Two questions:
1) I’m curious about using absolute momentum instead of the bond component to manage drawdown and SD.
2) Are any of your P123 strategy implementations marked as publicly visible? I’m also a subscriber.
Thanks again,
Scott
paul.novell@gmail.com · November 23, 2016 at 6:02 am
Hey Scott,
1. Removing the bond component would increase drawdowns, they increase to -47% with no bonds in the portfolio. Of course, returns go up. To 18% annualized.
2. No. They are private. I have plans for them.
Paul
Victor · November 23, 2016 at 9:09 pm
Hi Paul, a quick question… what do you mean by applying SPY-UI to each strategy individually? Wouldn’t you sell all the quant portfolios if SPY-UI triggered a sell? Or are you saying that UI enables 200MA exit on each quant portfolio?
Thanks, as always, for the great research.
Sidebar: could you comment on how the quant portfolios have performed in recent years compared to the long term results? I ask this because it appears that most quant models (particularly factor & momentum) have been deteriorating in recent years. Diversifying by strategy is a good idea, for sure. Just wondering if “alpha” is disappearing on these methods. I bit of a philosophical question, would appreciate your comments.
Thanks,
Victor
paul.novell@gmail.com · November 24, 2016 at 5:31 am
Victor, yes, I’m saying that when the SPY-UI triggered a self you’d sell the quant portfolios. Since the signal is the same for each portfolio you’d sell each portfolio.
On your sidebar: quant portfolios are holding up very well. For example, since the start of the bull market in 2009 through the end of 2015, TV2 has returned 16% a year (from 1964 to 2009 it returned 18% a year), and a combined Consumer Staples/Utilities portfolio has returned 18.5% a year (from 1964 to 2009 it returned 16.5% a year).
Paul
Paul
Victor · November 24, 2016 at 7:22 pm
Thanks for the quick reply, appreciate it. Wonder if it’s worth pursuing the idea that UI triggers monitoring each of the portfolios against their own 200MA. Might be “safer” than relying on the SPX to signal the exit. Guessing it would get you out of smaller cap, higher vol stuff sooner; Utes & CS further down the road. Hardest part would be paper monitoring and possibly trading the portfolio to check for re-entry conditions.
Cheers,
Victor
paul.novell@gmail.com · November 25, 2016 at 10:40 am
Sounds intuitive but doesn’t work as well as using SPX.
Paul
Comments are closed.