Low Turnover Strategies as Alternative to High Turnover

Models
1

If you are like me, you may have been initially attracted to the eye-popping theoretical simulated returns of high-turnover, low liquidity, low number of stocks models available. But in practice the results are mixed. In some of these models, I made lots of money. In others, I lost lots of money. In the end you could say it’s a wash. There is simply no way of knowing ahead of time which will perform. Even within models from a single designer, you can see highly variable performance. It’s a “lottery ticket”. Shoot to the moon or crash miserably.

In additional to uncertainty in real-time performance, I saw a major deviation between the model results P123 was reporting and my real-money brokerage account. One model (Model A) was performing flat, but my real-money account was up a whopping 50%. Another model (Model B) showed it was up 7%, but my account was only up 1.5%. The designer of Model B suggested that I make all trades within 2 hours of market open, but I was making all trades within the first 15 minutes.

Due to these challenges, and others such as transaction costs and hassle of trading large numbers of stocks, I started developing low-turnover strategies. Some of these models are nearing the one-year out-of-sample real-time performance.

Little to no optimization is used, and the stocks it picks are checked to see if they make sense, and the historic simulated performance is checked to see if it makes sense and fits in a pretty wide range of expected values.

These models have one or more of the following characteristics:
-Stable dividends
-Growing dividends
-Safe dividends
-Defensive sectors that resist cycles (consumer staples, utilities, healthcare)
-Low turnover “investing”, not high turnover “trading”
-Well known factors such as value, quality, growth, financial safety
-Reasonable expectation of earnings, sales, cash flow
-Low voliatilty in stock price or business fundamentals
-Conservative accounting
-Influenced by Marc Gerstein’s forum posts and online course

So far, these model appear to perform well. I would judge these models not on whether they beat the market index, but whether they continue to pick stocks with the intended characteristics.

“Ultra Low Turnover” means the holding period is significatnly longer than one year. “Long Term Capital Gains” means the model will try to hold the stock at least for 1 year, to reduce taxes.

Some of these models are free, the others are low priced.

Defensive Dividends - Long Term Capital Gains - 10 stocks
https://www.portfolio123.com/app/r2g/summary?id=1372671

Healthcare Investor - Low Turnover - 10 stocks
https://www.portfolio123.com/app/r2g/summary?id=1376455

Hedged Investor - 3 month rebalance - TLT Constant Hedge - 20 stocks (free)
https://www.portfolio123.com/app/r2g/summary?id=1383766

Large Cap Consumer Sectors - Ultra Low Turnover - 10 stocks
https://www.portfolio123.com/app/r2g/summary?id=1417539

Large Cap Defensive Dividends - Long Term Capital Gains - 10 stocks
https://www.portfolio123.com/app/r2g/summary?id=1386764

SP500 Defensive Investor - Long Term capital Gains
https://www.portfolio123.com/app/r2g/summary?id=1410449

Staples Investor - Ultra Low Turnover - 10 stocks (free)
https://www.portfolio123.com/app/r2g/summary?id=1392722

As usual, stocks are risky, I’m not a RIA, and I make no gaurantees.

2

I’m glad to have helped. But this post makes a very important point beyond the material I posted.

While it’s always dangerous to hold on to stocks that, based on investment merit, really should be sold, we need to be careful about allowing ourselves to be lulled by the Hollywood (traders screaming, gesticulating, and moving rapidly) and technocrat (cut nanoseconds wherever they can be cut and then speed up even more) views of the stock market: It is not necessarily true that faster is better.

No amount of technology can change the fact that we’re still investing in businesses that however fast they may move internally, still move at the speed of humans when it comes to interacting with the world, which is what they do to get revenues and profits, those being what ultimately drive stock prices. Investment cases need tome to develop and the length of time is determined by human providers and consumers, not by state-of-the-art data collection, computation and transmission speeds.

Some human endeavors are faster than others and our strategies need to take account of that. For example, one based on analyst sentiment could more justifiably rebalance more quickly than one based on company quality or accelerated growth. The more you’re willing to slow down, the more possible approaches you can consider as you open yourselves to a wider range of business developments. And that’s pretty much what MisterChang has been experiencing.

This doesn’t mean you should take your one-week strategies to 4 weeks, three months or a year. It does mean you should recognize that one-week or one-day type strategies confine you to a very narrow slice of the total pie and that you can accomplish much, for your real-money accounts if not for your sims, if you start thinking in terms of strategies that play out at the speed of business and not necessarily at the speed of Hollywood or IT folks.

And by the way, it doesn’t even have to mean that you need to use longer re-balancing. I notice that MisterChang is still using one-week. But the low turnover tells us that although his models look for change every week, they often find no reason to make moves. That suggests he’s using criteria with a large degree of stability. (In other words, a 1-year refresh period is the same as a 1-week refresh period that produced the same answer 52 times in a row.)

3

This brings up a disturbing trend in Smart Alpha. Developers have reacted to the big recent success of defensive sectors. There are a few designers that have released 5 or more of these models each. Some of them are offered for free. This is saturating SA with many models based on Defensive Sectors: Staples / Health Care / Utilities. These will take over SA at the expense of other strategies which may temporarily underperform. Similar to what happened with microcap value previously, these models may not look so great in the future, and has the possibility of killing off SA. Developers: we need to be providing a variety of strategies for all market conditions.

Steve

4

I think this also raises the question about good Portfolio construction strategies.

Marc, do you have any thoughts on that? Maybe for a separate thread?

5

I can’t see the appeal of high turnover strategies–isn’t your money just feeding the brokers and market makers? Anyway, since most of our rules and rankings are based on quarterly financial reports, it doesn’t make sense to me to have a turnover higher than 4X.

The Smart Alpha menu should really allow users to sort by turnover. As it stands, you have to click on every model in order to find out what the turnover is.

6

I agree with Steve: diversity is the key. Moreover, low-turnover models are statistically not very robust.

As far as I’m concerned, it is better to spread risk by combining several models with very low correlation (i.e. different ranking systems, universes etc).

Not sure if I will ever reach a low drawdown as in the attached book example, but I’ll give it a try.

low_DD.jpg
low_DD.jpg921×358 116 KB

7

I’d rather phrase it this way: Defensive models (what Steve talked about) are usually built around the theme of stability makes for fewer occasions to shift gears and trade – low turnover isn’t the goal, but it is dragged along for a free ride. And yes, they are definitely not robust. But . . .

This is a perfect example of the difference between investing and other disciplines that make use of statistical analysis.

If you want to moderate risk, you DO NOT want robustness; robustness would be a bad thing. You’re looking to moderate risk for a specific reason, often because you have fears regarding the market. Therefore, you want a model that addresses those. A truly robust model, one that is just as good for raging bull markets, would probably cause you to suffer the big losses you fear. Another reason for wanting to moderate risk may relate to personal life/financial circumstances and have nothing to do with market assessment.

Robustness is not a legitimate goal here. The only genuine target is meeting the investor’s needs; your own needs or the needs of Smart Alpha subscribers you’ve chosen to target (which may or may not be the same thing). I’m OK with more defensive models on Smart Alpha, and that’s a market thing. I think the gunslinger market is small and hyper-fickle (i.e. extremely market sensitive) and I’m sure many who’ve seen SA subscriptions vaporize would now agree. Defensive models have larger and more stable customer bases, and income models are even better from a marketing standpoint. I would not want to see gunslinger models vanish; there is a time and customer group for it. But those who offer them do need to appropriately manage their expectations as to what, commercially, they can expect.

If you want to go on autopilot, then yes, I think a book of different well-constructed models is the best way to go. And none of them should be robust individually (if you get robustness along the way, send a thank-you bouquet to Lady Luck). They should each do well at whatever goal they pursue. Personally, I don’t use the book feature but that’s only because I’m to lazy to pull together the diversified portfolio of models I’ve been running for years. I would never dream of putting 100% into any model, even my own (especially my own because I know different models are good for different purposes). And as I’ve said in the past, I think all SA designers would do well to think book and price their models appropriately (assuming all subs will hold multiple models by multiple designers and will be sensitive to how much they pay in total for everything).

P.S. To get that sort of low drawdown in the real world, or anything even remotely close, that book is going to need to hold a heck of a lot of cash and/or very short-term fixed income, or be heavy in magnificently-timed hedges.

8

This was a tough question for me because I had never explicitly contemplate it. But after a good night’s sleep, or at least a sleep infused by artsy cocktails (for which I have a soft spot), here goes . . .

I think good portfolio construction strategies could just as easily lead to high turnover or low turnover.

The conversation, here, lately has had the appearance of equating good modeling with low turnover because it’s come in the context of MisterChang’s initial post, and the strategic factors he considered, and the defensive issues posed by Steve. In both cases, we’re talking about “investors.”

But the financial markets are an intriguing place, fodder, if you will, for scifi or fantasy writers because it violates the laws of physics that suggest only one object can be in particular placer at a particular time. (Real physicists here – think metaphor, not real science.) The stock market is a “place” where multiple “objects” share the same space at all times. The community of investors is one “object.” The community of “traders” (those who look to profit from short-term behavioral tendencies of others) is another “object.” The community of “market makers” (a sub-set of traders who not only need to profit but whose reason for being is to provide liquidity to others) is another “object.” The community of “speculators” (those who taker big risks in the hope of disproportionately large returns and whose time horizons can just as easily mimic those of traders or investors) is another “object.”

Lower turnover is most likely associated with good modeling by investors. Good modeling by market makers and traders probably requires very high turnover, especially for market makers – if their turnover falls, we’re back to 2008 or worse. Good modeling for speculators can go either way in terms of turnover. And the existence of FolioInvesting may muddle everything – personally, I’m willing to take high turnover in a tax deferred investment account because my variable cost of trading is zero. Whatever low turnover modes I offer are dictated by the strategy or pushed that way to accommodate my sense of what SA investors who pay commissions will need. Some of my most successful personal models, however, have been high turnover (I’m not a trader, but I’m willing to don the speculator hat – but not in SA; I’d rather keep that personal.)

We have to remember, too, that I talk about good modeling for investors because that’s what I see as the best match for many, perhaps most, Portfolio123 users and my own skill-set. But I had a good close look at the other side when I was at Reuters, which owns MetaStock. I had a really good relationship with the folks there and got to know their operation very well. If you think I bombard you folks with tutorials and the like, you should see what they do. They are EXTREMELY education heavy and offer a massive body of written knowledge and verbal support on good modeling practices in the technical-trading arena.

So in my opinion, there is no relationship between the level of skill in modeling and turnover. It’s about the goal, and any goal can be pursued well or not effectively.

9

I just have to say Marc’s comments are excellent.

Some people have such a bias against high turnover ports or high turnover trading in general. To them it seems to be moral. Some literally think they are ethically superior for holding stocks long term.

Marc shows no bias either way in his post. Very perceptive, IMHO.

There are obviously things one should consider. The return for high turnover ports is more sensitive to slippage, for example. You have to be more accurate in your slippage calculations or leave a margin of error. But the only bias should be toward your bottom line.

If you are not biased either way you should look at Hoyt’s ports without bias. Hoyt has been around for a while. I think he is a good, smart developer.

10

About defensive models: When I look at a developers offerings of defensive models, I have a great deal of difficulty distinguishing between one model and another. And that is just for one developer. Collectively, I don’t think Smart Alpha is capable of supporting that many defensive models. Well P123 is, but developers aren’t going to make any money on them because there are just too many models, and not enough subscribers, and there are not enough distinguishing features between one model and another for anyone to make a decision. We are all on the Titanic and all trying to jump into the same lifeboat. There will be so many defensive models that they will drown out other models, which will starve from lack of subscribers, and won’t be there when the markets change.

This is also a reason why Book technology needs to be improved, so that subscribers can easily diversify among focused SA models. Robustness and stability are important at the account level, not at the individual port level. If Book functionality is short changed then Smart Alpha will continue to suffer.

Steve

11

That’s a great point, completely agree.

12

Maybe P123 could extend the number of subscriber Book assets if those assets are SA models.

Walter

13

Yes. And that’s how the marketplace works, for SA models, for widgets, for everything. Nobody ever said market economics was easy or pain free in the real world.

Excess supply relative to demand is as old as human interaction itself, as is the reverse. It sucks to be the last or high cost of worst marketed etc. supplier of a defensive SA model into an overcrowded marketplace. It sucks to be the last buyer to settle a trade in the context of excess demand for internet stocks or TLT – I couldn’t resist). It sucks to be the last seller of a barrel of oil into a saturated energy market.It sucks to be the last provider of a mobile operating system into a market that has as many as it needs or wants (probably why Steve Ballmer moved from microsoft to the nba). Annd a ,lot of folks here are engineers. Either you, or others at your firms, know it sucks to be the last to start selling, say, power-plant construction services into a saturated market, or something like that. Any cardiologists out there? Even neighborhoods that favor sirloin with a side of deep-fried cheesecake generates business for only so many practitioners.

So yes, it sucks for you that there are so many defensive models (though it’s good for buyers). And for those who are empirical in their approaches, it sucks even more because the past is fixed and there are just so many rank factors out there to be discovered and deemed “robust” and combined, making that marketplace much more prone to becoming saturated much more quickly.

14

Well Marc - I’m all for capitalism. And being one of the first in with a defensive model, it doesn’t suck for me. But it will suck for P123 as subscribers try to jump from one lifeboat to another and they eventually drown.

Steve

15

Not if the lieboats are all seaworthy. It’s unrealistic to expect perfection in this regard, but we did change the platform to make it harder for the non-seaworthy lifeboats (great sims, bad live results) to get noticed.

16

To bad we can’t see all the non seaworthy lifeboats but that ship has already sailed.

MV