A Tolerable Kind of Error: ETF Pullback Choices

by: Marc Gerstein

Economists are useless; they can’t forecast economic activity. Analysts are useless; they can’t estimate earnings or forecast stock prices. Stock-picking is useless; nobody can beat the market. You’ve heard all this, and more, plenty of times. Investors hate mistakes, but are confronted with them time after time.

Interestingly, though, investment commentary is skimpy at best when it comes to meaningful discussion of the kinds of mistakes that occur and how we might incorporate those possibilities into our strategies. In this era of rising correlations and volatile markets, even such old standbys as diversification and cushion of safety often wind up working better on paper than in your brokerage account.

Reviewing the weekly selections of the ETF Pullback model I introduced on Seeking Alpha in late July, including this week’s choices, has been bringing my mind back to some concepts some of us learned in statistics classes but haven’t thought much about since then: Type I and Type II errors.

A Type I error occurs where we reject an idea we ought to have accepted. A Type II error occurs where we accept an idea we should have rejected. Perhaps we should think about these more openly and better understand the nature of the damage to which we’re exposed if we do the wrong thing.

Today, the big question is whether we’re heading for a double-dip recession. One day, it look like we’re going down again. Another day, it seems that all is well and that economic recovery will be sustained. We’ve pretty much been in this on-again off-again pattern since the spring and it’s really hard to have much confidence in anything we do with our money.

What we can do, however, is make choices about the kind of error to which we are willing to be exposed. Working with the idea of sustainable economic recovery, we can accept this and invest in cyclical stocks. If we’re wrong, we will have committed a Type II error. In the alternative, we can reject the idea and invest defensively. If we’re wrong in this regard (i.e. the economy turns out fine), we will have committed a Type I error.

The ETF Pullback model has, in a sense, been rejecting the idea of sustained recovery and keeping me primarily in fixed income for several weeks, including now.

Here’s this week’s list.

  • Vanguard Long-Term Govt. Bond (NASDAQ:VGLT)
  • Nuveen New York Quality Income (NUN)
  • PIMCO 25+ Strips (NYSEARCA:ZROZ)
  • Global X/InterBolsa FTSE Colombia 20 (NYSEARCA:GXG)
  • SPDR Barclays Capital Long Treasuries (NYSEARCA:TLO)

This is last week’s list.

  • Vanguard Long-Term Govt. Bond (VGLT)
  • Vanguard Long-Term Bond (NYSEARCA:BLV)
  • iShares IBoxx $ Invest Grade Bonds (NYSEARCA:LQD)
  • Vanguard Intermediate-Term Bonds (NYSEARCA:BIV)
  • SPDR Barclay

Last week’s performance wasn’t so good. For the most part, the markets acted on the basis of an assumed economic expansion and I wound up underperforming based on a Type I error, and that’s the kind of mistake to which I remain exposed.

One reason why I’ve been continuing to work with this model is its long-term backtest record (see Appendix below) and the fact that my real life (out of sample) results have been consistent with what the tests led me to expect. No model beats the market every time out but the cold snaps experienced by this one have been very tolerable. That may have something to do with the nature of the mistakes it’s prone to make.

The model, looking for presumably fleeting pullbacks in the context of a longer-term uptrend, uses the bullish scenario as its main idea. When it’s right, it’s right. When it’s wrong, it will have more often than not have inappropriately gone defensive. You never want to be wrong, but if you have to make a mistake, being inappropriately defensive is often likely to be a heck of a lot less painful than being inappropriately aggressive.

A lot of this may be coming from the fact that it’s an ETF model. I’ve been working on an equity version of this strategy but am not yet ready to invest real money based on where it stands thus far. ETFs cover a full range of geographies and asset classes thereby giving the model an opportunity to take a much more substantial defensive stance when it so wishes. This is harder with stocks. There are some groups that are considered less cyclical than others, but even the best of defensive stocks (including dividend plays) find it hard to measure up to fixed income when times are truly tough. But I haven’t given up on an equity version of the model; I’m still plugging away.


To create this model, I started with a very broad-based ETF screen I created in StockScreen123.com.

  • Eliminate ETFs for which volume averaged less than 10,000 shares over the past five trading days
  • Eliminate HOLDRs (I don't want to be bothered with the need to trade in multiples of 100 shares)
  • Eliminate leveraged and short ETFs (I think of these as hedging tools rather than standard ETF investments of even trading vehicles)

Then I sorted the results and select the top 5 ETFs based on the StockScreen123 ETF Rotation - Basic ranking system, which is based on the following factors:

  • 120-day share price percent change - higher is better (15%)
  • 1-Year Sharpe Ratio - higher is better (15%)
  • 5-day share price percent change - lower is better (70%)

The idea of using weakness as a bullish indicator is certainly not new. But often, it's an add-on to other factors that, on the whole, emphasize strength. Here, the weakness factor is dominant, with a 70 percent weighting.

This model is designed to be re-run every week with the list being refreshed accordingly. I trade through FolioInvesting.com, where I pay a flat annual fee rather than a per-trade commission, so I don't care about the fact that turnover form week to week is often 80%-100%. If you want to follow an approach like this but do have to worry about commissions, the strategy tests reasonably well with three ETFs, or even with one. (Cutting the number of ETFs is far preferable to extending the holding period.)

Figure 1 shows the result of a StockScreen123 backtest of the strategy from 3/31/01 through 7/22/10.

Figure 1

(Click to enlarge)

Figure 3 covers the past five years, a very challenging market environment that witnessed the fizzling of many strategies that had succeeded for a long time.

Figure 2

(Click to enlarge)

Disclosure: Author long VGLT, NUN, ZROZ, GXG, TLO