Valuation-Informed Indexing #69: Is Valuation-Informed Indexing the Product of Backtesting?

Published on

The academic research shows that investors who change their stock allocations in response to big swings in valuations obtain fair higher returns at greatly reduced risk. Valuation-Informed Indexers can realistically expect to be able to finance a middle-class retirement five to ten years sooner than their Buy-and-Hold counterparts. But can we trust the research?

There’s certainly no guaranty. The research is based on the historical stock-return data. I think it’s reasonable to believe that stocks will continue to perform in the future at least somewhat as they have always performed in the past. But we cannot say with certainty that that will prove to be so.

If that were the only criticism of Valuation-Informed Indexing, it would obviously be the best choice. There’s no guaranty attached to any investing strategy. So pointing out that there is no guaranty attached to Valuation-Informed Indexing is not saying much. However, there’s a more targeted criticism that can be directed at the new investing model.

The more meaningful criticism is the charge that Valuation-Informed Indexing is the product of backtesting. Yes, there’s 140 years of data supporting the concept. But it is possible to find data supporting all sorts of crazy ideas if you set out to find it. Statistical support for a strategy becomes more powerful when the strategy survives both backtesting (seeing that it has always worked in the past) and front-testing (seeing that it continues to work on a  going-forward basis after the strategy has been publicly described).

Valuation-Informed has survived front-testing. The front-testing support is not as compelling as the backtesting support (because there are not yet as many years of front-testing support). But it is significant and it becomes more significant each time this bear market takes prices lower.

The toughest part of convincing people to become Valuation-Informed Indexers is that it takes so long for this strategy to produce good results. Short-term timing strategies can be easily checked. The person advocating the strategy announces his predictions and those interested look in a year or so to see whether the predictions have proven out or not.

Valuation-Informed Indexing is not promoted as something likely to work in a year or two. It can take 10 years or even a bit longer for the investor emotions that cause insanely high or insanely low valuations to weaken and thereby permit prices to revert to the mean. Many people are not willing to wait 10 years to make a decision whether an investing strategy is for them or not. They lose interest and go on to something else.

I can report to you what the data identifies as the most likely annualized 10-year return for a purchase made at today’s price levels — it’s 2.7 real. But I cannot realistically expect you to be willing to wait 10 years to see whether the prediction ends up being on the mark. I can explain to you why the predictions should always work and point you to the 140 years of data showing that they in fact always have worked. But that’s not the same thing as letting you front-test the concept and thereby see for yourself that it really does always work.

However, I can do something close.

Valuation-Informed Indexing is not yet the dominant model. There are still lots of people promoting Buy-and-Hold. However, Shiller’s research showing that this works dates back to 1981. Shiller’s research is publicly available. So, if you look at what Shiller said should work in 1981 and then look forward to see how well it really did work in the following three decades, that’s front-testing. That should be sufficient to persuade any remaining skeptics.

Guess what? This front-tests as well as it backtests!

Shiller’s research shows that, when stocks are priced as they were in 1981, they should provide amazing returns for the next 15 years. And they did.

Shiller’s research shows that, when stocks are priced as they were in 1996, they should provide poor returns for the next 15 years. And they did.

Stocks provided amazing returns from 1996 through 1999 even though they were priced to provide poor returns. But Shiller’s research shows that that can happen. If investor emotion is the dominant influence on stock prices in the short term, as Shiller’s work suggests, short-term prices cannot be predicted.

Shiller’s 1981 research showed how stock investing worked from 1870 through 1980. From 1981 through 2011, stocks continued to perform just as they always have in the past. Shiller’s model is now supported not only by extensive backtesting but by 30 years of front-testing as well.

Should we want to see even more proof? Yes. We should want to see all the proof we can get our hands on. And it is indeed possible that starting today stock prices will begin to play out in a manner in which stock prices have never played out in the past. That could happen.

I think it’s fair to say, though, that that is very much the long-odds bet.

Rob Bennett has explored the question of why retirements fail. His bio is here. 



Leave a Comment