Maybe a Long/Short Approach Is the Answer...

A day does not go by that I get one of two questions over email regarding the paper I published:

1. Have you considered going all-in to the positions on a buy signal?
2. Have you considered a long-short version of the model?

I will attempt to answer both of these questions below. First, here is a table of returns for variations of the strategy I published. As usual, no fees are accounted for (including taxes, bid ask spreads, commissions, and management fees). Interest is paid at 90-Day Commercial Paper Rates, margin is deducted at the broker call rate, and short rebates are also paid at the 90-Day Commercial Paper Rate (in reality this overstates returns - more later).

AA = Asset allocation (20% each in US Stocks, Commodities, Foreign Stocks, US Govt Bonds, and REITs since 1972)
LS = Timing model always in either long or short.
EW = Equal weight portfolio for whatever is on a buy signal. If none, 100% in cash.

Have you considered going all-in to the positions on a buy signal?

The basic take-away is that it increases returns by about 2% per annum, while increasing volatility to near buy and hold vol. It also increases drawdown, and in my opinion, exposes the portfolio to far too much event risk. From the table below, about 15% of the time the portfolio is only invested in one or two asset classes. On top of that, you would be rebalancing every month with significant brokerage commissions and bid/ask spreads if you use ETFs. There is serious event risk in having your portfolio in only one or two funds, and adding in the headache and expenses of excess trading, this model holds little appeal. Maybe one could get by with a FolioFn type account, but I would not feel comfortable with this setup (although you would be 50% bonds and 50% commodities right now). I would much rather leverage the whole portfolio 2:1...

You also lose the tax benefits of the timing model - all the losses were short term losses while most all of the gains were long term gains.

Have you considered a long-short version of the model?

Yes, but I didn't publish the results for a number of reasons.

1. Most retail investors have no experience shorting, and don't understand it whatsoever.

2. Most brokerages don't pay any interest on short sales. (This also applies to the leverage model and ridiculous margin rates.)

Now, there are certainly ways around this. The first is to obviously switch to a better broker like Interactive Brokers. Second would be to use options or futures, but that goes back to the same problem as #1.

However, here are the results. If you assume that you get paid on your shorts with 90-Day Commercial Paper rates (in reality there is a small (or large!) spread to cash rates, but depends on the broker, see below.) Let's assume these returns should come down by about .5% per annum. (And if you don't get paid on short balances the results drop to <10%.)

So basically you get in the same ballpark with returns as long/flat and long/only. Long/short ends up having buy and hold like volatility, with buy and hold like drawdowns. The big difference comes with correlation. The long/flat is basically a long/only substitute, while long/short has no correlation to the long/only portfolio.

Assuming you have a $1m account, below are the rates of some retail brokerages. I saw margin rates as high as 10% - so the same applies to running the leveraged version of the model!

Comments

[systemseeker]

For someone with a 401(k) where taxes and commissions aren't an issue, here's an idea -- A hybrid of "EW" and "2X": for a given month when you're in only 1 or 2 assets, go in 40% each (2X). If you're in 3+ assets, divide the whole portfolio evenly (33% for 3, 25% for 4, 20% for 5, etc). I wonder how that would fare?