In finance there is a bias towards more data. The idea being with more data we can make ever finer distinctions and more accurate forecasts. The challenge is that more data is not necessarily better. Things change in financial markets over time and old data may be more of a liability than an asset in trying to answer today’s questions.
In my new book, Abnormal Returns: Winning Strategies from the Frontlines of the Investment Blogosphere, I spend a devote a section of the book to a discussion of the equity risk premium. One of the things that frustrates me most is when analysts naively extrapolate the historical US equity risk premium into the future. There are two reasons for this.
First, the historical data gathered from the pre-electronic era is suspect. Laszlo Birnyi writing recently at the Financial Times notes the problem with relying on historical data. It is often wrong:
Given the availability of computers and information, investors too often consider data as irrefutable information which can be sorted, stacked, analysed and processed. Estimates become hard-wired, often to several decimal places, and one or two data points are projected as long-term estimates.
Unfortunately, relatively little analysis or thought is undertaken as to the validity or source of many of the numbers on which we make investment decisions.
Second, it is naive to think that the equity risk premium generated in a period before the onset of modern society has much relevance for investing today. The world (and markets) have changed so dramatically as to make this earlier experience practically mute. Today we have access to nearly limitless information data and essentially free trading. An equity investor a century ago was in a very different situation. It took a far more intrepid investor to brave the equity market then.
That is why it heartened me to read some comments by Robert Shiller in Money recently. From the interview by Penelope Wang:
Why buy and hold doesn’t work anymore
My old friend Jeremy Siegel [Wharton professor and author of “Stocks for the Long Run”] makes the strongest claim about this. He has data going back 200 years showing that the market has had a real 7% return over that period.
But there’s no solid reason it should do so well. Things can go for 200 years and then change. I even worry about the 10-year P/E — even that relationship could break down. But I believe I’m on better ground thinking that the P/E forecasts returns than thinking one asset just always outperforms.
Are you saying that there’s no reason stocks should outperform bonds at all?
Oh, no. If you go back to textbook finance and make some assumptions about investors’ risk aversion and that assets should be priced to pay investors for added risk, you would get some outperformance for stocks. But not as much as it’s been; it looks as if past high returns were a historical anomaly.
So when I said the 10-year P/E predicts a 4% return, that’s conditional on past returns being a guide to future returns, and the truth is, we don’t know. Maybe it will be only 2%.
What does it all mean? It seems like many pension funds have gotten tired of equity market volatility in favor of bonds. Peter Rudegair at Reuters notes how both institutional and individual investors have recently eschewed the once popular equity-centric Yale Model of investing. This at a time when prospective real returns from presumably safe assets like Treasuries are at near historic lows.
The bottom line is that the equity market is going to do whatever it does based on future earnings and changes in valuation parameters. The distant history of the stock market has no role in how that will play out over time. How much (or little) you allocate to equities in your portfolio should be based on a thoughtful process based on analysis not the naive extrapolation of data from a long gone era.
Items mentioned above:
It’s time to take stock of historical analysis. (FT)
More readings on the equity risk premium. (Capital Spectator)