Bear with me as I set this up.
Back in 1996, Alan Greenspan coined the term "irrational exuberance" in a speech, after the market had been up double digits many years in a row. In 2000, Robert Shiller of Yale used the term as a title for a book, making an extended argument that the stock market bubble had to burst because it was, in historical terms and using a metric he devised, too expensive. Shiller had been making that argument and using that metric for almost five years previous in academic circles and the market appeared to prove him wrong. But Shiller's book timing was dead on. The book came out, the Nasdaq cracked, and then everything else cracked with it. Two years later, the Naz was down 80% and the S&P 500 about 50%.
Shiller's primary metric was PE10, which is market price divided by an average of the previous 10 years earnings. Investors commonly use a price earnings ratio to determine whether a stock is cheap or expensive, but that tends to be a pretty noisy time series. Shiller was looking economy-wide, so used the S&P 500 as a measure, and wanted a less noisy series, so used the 10 year average. (He never explained to anyone's satisfaction why earnings from 10 years ago, whether it be for an individual business or an economy spanning index, should be an important contributor to rational decision making about a correct level of stock prices. Nor did he explain why PE10 and not PE5 or PE8 or ...) Being a careful academic, Shiller was from the start careful to provide caveats and warnings about data mining. Until 2000, those caveats came in handy because the the market said his thesis was bunk. By 2000, PE10 was above 40, an unprecedented level. Two years later, Shiller looked a prescient genius.
However, if you look back over periods longer than a few years, it's not clear that Shiller's forecasting is worthwhile. Shiller looked very wrong for years before the market cracked, and the cumulative effect of being out of the market in the late 1990s would have put an investor so far behind that it's not clear that even the 2000-2002 bear market would have allowed catching up to a very stoic buy-and-hold investor.
In my professional life, I counsel buy-and-hold, albeit with a much greater variety of asset classes than Shiller used in his argument. I think and hope that diversification protects my clients. But there's never any reason to close your mind to the possibilities before you're dead, so let's explore a little. I'm not advocating market timing, nor bearishness because the market is still "expensive" today. I'm just exploring.
For a few years now on various US investing discussion forums, there's been a succession of registrations by an annoying internet troll going by the name of hocus. (In real life, he's a guy named Rob Bennett who lives in Purcellville, VA.) Hocus advocates market timing based on Shiller's PE10. It wouldn't be that annoying except that, once he starts, he won't stop talking about it, inserting his favourite topic into any thread that looks plausible. As a result, he's been banned from many forums, is arguably responsible for the collapse of a few forums where the moderators didn't stop the nonsense, and is now pretty much confined to one forum which has as its sole purpose lampooning him (google hocomania if you must know).
In any case, I went through some of hocus' market timing strategy (named Valuation Informed Indexing by him, VII from now on), most of which is built on Shiller's argument and all of which ignores Shiller's caveats and ran some backtests on US return data. I compared it to buy-and-hold (B&H).
The monthly return data I used is from Ken French's website, which goes back to 1926. French's data is not for the S&P 500 but for all US stocks because it was collected to measure the factors in the Fama-French 3 factor model. French also kindly provides the monthly return on T-bills as the risk-free asset. I amalgamated this return data with Shiller's monthly record of PE10 back to 1926 and ran the tests.
B&H was really simple: 60% in stocks, 40% in bills, rebalance if a category goes more than 5% away from the target.
VII was more complex. Hocus is a bit scattered, so it's hard to distill an algorithm from his verbiage. In the end, I settled for as good an approximation as I could manage for what he was suggesting. (He's revised the strategy many times, he contradicts himself over and over, it's very data mined, I take it all with a grain of salt.) VII goes pretty much like this.
1. A higher PE10 means less stocks, lower PE10 means more. Hocus provided a few anchors, so I ended up with a scale that is 100% stocks at PE10=5 and 0% stocks at PE10=35.
2. Small variations in PE10 are unimportant, so the stock allocation steps by 10% at intervals of 3 in PE10. For example, 60% stocks at PE10=17 is reduced to 50% at PE10=20, or increased to 70% at PE10=14.
3. Rebalance if you're more than 5% off the target, which may have changed because of a PE10 change.
Over the 80 years of data, B&H rebalances 58 times, VII 150 times. Years ago, it would have been hard to do this and keep transaction costs reasonable. Nowadays, it's dead cheap and easy to rebalance twice a year.
Jeez, that's a lotta foreplay. How about some meat, Norbert?
Here's a graph of 30 year annualized returns. B&H is black, VII is magenta.
VII is pretty clearly superior everywhere, although there are a few occasions ending in the last half of the 90s, i.e. starting in the last half of the 60s, where B&H has a small advantage. Recent differences appear to be quite minor, although I would point out that an extra 1/2% a year over 30 years is not chicken feed.
Below 20 year annualized returns.
Again VII looks better but the advantage has not been spectacular since at least the mid-70s (which means periods starting in the mid-50s).
Finally 10 year annualized returns.
Now this is an interesting picture. If you ignore the right hand side and imagine you are in 1996, just as Shiller was when he began to think and argue and publish about this, the evidence is pretty incontrovertible. VII, i.e. market timing, is everywhere superior to B&H over ten year periods. An occasional time, sometimes for decades, B&H is close, but there are periods of serious underperformance relative to VII. In 1996, you could not make an argument that the historical data ever showed a period where B&H was better. Timing was sometimes no better but it was never worse.
But then the late 1990s happened and it all went to hell. B&H turned out to be a huge winner in the late 1990s. Tie it together with widespread acceptance of modern portfolio theory and efficient markets - and no looking too far back because really old history is probably less relevant - and B&H is obviously the way to go. The 2000-2002 bear market might make you question that, but maybe not.
We have a few more years under our belt now. Is Shiller a hero or a fool? Is buy-and-hold appropriate because the old days aren't coming back or are we fooling ourselves with recency bias?
3-2-1.
Go.
