Does anyone use longer term holding periods for taxable accounts. The tax drag from short term vs. long term (1 year +) are not insignificant. Here in MA, the ST rate is 12%, which means the total tax burden for someone in the 28% bracket is a total of 40% vs. 20% (15% fed and 5% state) for long term gains. This means a short term model would need to produce average gains significantly higher than a longer term model. And while the higher returns of short term models are guaranteed, the tax burden they impose is guaranteed.
I have have produced some promising sims using mid and large cap stocks (25 to 40 positions), with the sell rule: rank<90 and (gainpct<0 or nodays>365). This allows for tax loss harvesting and forces you to hold winners a year +. Even in the screener, using a rebalance of 1 year vs. 3 months, produces comparable returns.
Curious on people’s thoughts, as low turnover systems and tax drag issues don’t get much attention here.
Yes, I think I’ve learned the hard way on taxable accounts that it’s generally well advised to have a holding period of 1 year, for the reasons you described. It’s difficult enough to just consistently beat the indexes, much less when you raise the bar to include the tax differences
For 1 year holding patterns, I’ve found that the vice universe models discussed on this board generally do well (liquor, beverages, tobacco, etc.). And I also think Value + Quality combination models do well with 20+ holdings. With such low turnover, I can take on higher stock ports with negligent slippage costs.
Unfortunately, I don’t think there is much demand for low turnover models, because people gravitate towards the eye-popping returns of high turnover models.
Personally, I only trade high turn models in my tax advantaged Roth IRA, and trade models that target long term capital gains in my regular account.
In addition to simply holding for over a year, you can also try to target “hold forever” companies to delay paying taxes for as long as possible. I have models with average holding period of 700 or 800 days. I think this could be appropriate for stable, predictable companies, such as consumer staples or the aforementioned vice stocks. Ideally, you would also screen for companies that can grow consistently year after year, but “growth” is one of those elusive things that everyone wants, but only a few people like Buffett can find.
Be careful about allowing taxes to become the tail that wags the dog.
Different models require different holding-rebalancing periods because different kinds of investment-trading stories take different amounts of time to pan out. My goal typically is to pay as much in tax as possible (because that means I made more money). If you have a situation where you really do need to watch your tax liability, the most important thing to do in terms of the stock market is to limit yourself to the very small number of possible strategies that can withstand very long holding periods of a year or more
Be careful about interpreting sims of such models. They can look good but easily collapse with real money. Here’s why:
The lower the turnover, the more your success is hostage to how effective your model was on day one, A successful sim is one that may have been terrific given the state of the world on, say, 1/2/99. But if you buy the portfolio today, you may be getting into stocks that have already run as much as they will and whose only virtue is that they don;t produce tax liabilities for those who take big paper profits – for them, sideways future movement may work when viewed over a multi-year perspective. But for you, buying in at today’s price, you may be looking at sideways only since you didn’t actually get the big sim gains.
Al alternative is that you re-start the model with an inception date of the present. But now, you don;t have any reason to assume a model that worked great in 1999 or 2002 or whatever will give you a good list of stock if you run it in 2017. Here’s how I manually examine sustainability of a very low turnover sim. (This is a more granular look at what the optimizer does.)
Run it with a 1/2/99 start date (or whatever other inception date out like).
Look only at the performance for calendar 2016.
Run it again with a 1/2/00 start date.
Look again only at cal. 2016
Run again on 1/2/01.
Look again only at 2016.
etc. etc. etc. until you re-run with a 1/2/15 start date and look only at 2016.
With a very low turnover model, you might be amazed at how the various 2016 tallies change. If they change a lot, abandon the model not matter how good your original sim was.
Higher turnover definitely has tax and trading cost implications and you need to know these. But you have to balance it against the good aspect of high turnover, which is the fact that if you wind up with a bad list of stocks, as can happen any time to anyone given that we’re always dealing with probabilities, not certainties, there’s something to be said for being able to blow out the portfolio and reboot every 4 or 13 weeks. So I make sure my brokerage costs are such that I can tolerate 400% to about 1000% annual turnover; that’s how it usually works out for the models with which I’ve been most successful. (1 week could be good for some folks, but my style is rarely compatible with the idea of stories that succeed or fail in that short a time frame, though I do have some).
Tax will be a moot point if Trump’s tax proposals are adobted. All one has to do is to trade as a LLC or sole proprietorship. Here are the key elements of Trump’s proposed tax reform:
The tax rate would decrease from 35% to 15% for C corporations;
For pass-through entities (partnerships, LLCs and sole proprietorships), tax rates go from 43.4% to 15%;
It would repeal the 3.8% net-income investment tax (NII), which would reduce the tax on S-corporation income and on S-corporation capital gains if a company is sold;
There would be a one-time “repatriation” tax of 10% on off-shore earnings; and
The worldwide taxation system would change so that U.S. taxpayers would pay taxes in the country where income is earned instead of in the U.S.
You’ll see from my sell rules that losing positions can get kicked out before a year. Appreciate your suggestions for robustness testing. I have found with the ranking systems I’m using that rolling back tests, using a 1 month 3 month or 1 year holding period, as well as odd/even id restrictions produce pretty consistent results. Focusing mainly in the large and midcap space. This goes back to a previous post of mine, suggesting that this universe is more forgiving than the small cap space in terms of poorly constructed systems, which gives another layer of comfort that a poor model will more likely produce market results as opposed to gross underperformance, compared to a small/micro model where a poor model and underestimating transactions cost in a high turnover model can lead to catastrophic results.
I totally agree with Marc's comments. In addition, I would like to point out that a lot of P123's users left in 2008, due to the sell-off. Holding almost any equity during that period was bad for financial health.
Charles, As I go over the P123 database it seems like an illiquidity focused approach offers potential for longer hold periods. In a very simple screen 6 month hold worked much better than 12m hold, but I haven’t really looked at turnover or tried to add further selection criteria beyond liquidity. It shows promise and is something I plan to work on going forward.
When I first got on P123, I built a bunch of annual models. But they are so dependent on start date, as Marc points out. I do think that value models can work on an annual holding period since by definition you are buying cheap stocks and giving the market time to recognize the value. December is a decent month to re-balance such models for a variety of reasons.