One of the most dangerous things for an investor to do is to make big changes to his asset allocation based on valuations. Valuation-based tactical asset allocation has proven very hard to execute over time, for a simple reason: Asset-class valuations do not exhibit much mean reversion. In fact, they often exhibit prolonged regime shifts. A dramatic example of this can be seen in Treasury yields.

Source: Robert Shiller's online data website.

Exhibit 1 shows that long-term Treasury yields have moved in decades-long regimes of either falling or rising rates. A hypothetical investor in the 50s who observed that yields were at 20-year highs and unlikely to rise further would have been wrong for over 30 years. On the flip side, an investor in the mid 80s who stayed out of the bond market on the expectation that yields would spike again would have been wrong for 30 years (and counting!). These long-lasting changes to bond yields can’t be adequately explained by secular changes in inflation, as investors persistently underestimated future inflation during the 60s to 80s and persistently overestimated future inflation from the 80s and on. (To finance nerds, the market being wrong on inflation for decades on end is a puzzle; in a reasonably efficient market, inflation surprises should be random.)

The stock market has also exhibited big, long-lasting shifts in valuations that have frustrated value-conscious investors who have tried to get out when the market looked “expensive” and buy when it was “cheap.” Exhibit 2 shows the cyclically adjusted earnings yield of the S&P 500, calculated by taking 10 years of inflation-adjusted per-share earnings and dividing it by price (the inverse of the famous Shiller price-to-earnings ratio). There are decades where valuations can remain elevated or depressed relative to history. The poor value investor who got out of the stock market in the mid-90s as the earnings yield hit hit lows unseen since the late 60s—almost 25 years prior—would have sat out much of the fantastic returns generated by the dot-com bubble. Some very smart investors did just that, with disastrous consequences for their careers.

The tendency for valuations to remain at extreme levels relative to history for years or even decades makes contrarian asset allocation a frustrating and dangerous exercise. Even with the benefit of hindsight, it is hard to come up with a simple valuation timing model that doesn’t suffer from decade-long dry spells of underperformance. The most successful practitioners of market-timing strategies (George Soros and Stanley Druckenmiller) are largely trend-followers.

Does this mean we should ignore valuations when determining our asset allocations? No. There is still some hope for the valuation-conscious tactical asset allocator. Warren Buffett famously made two market calls: In July 1999, he gave a talk at the Allen Sun Valley Conference where he predicted that stock returns would be much lower than widely expected. The market peaked within a year and collapsed. In October 2008, he wrote an editorial in the New York Times urging investors to “Buy American.” The market bottomed five months later. While Buffett grounded his analyses in valuations, valuations alone can’t account for his uncanny timing. In both cases, he noted the acute, widespread certainty on part of market participants that the market would be heading in only one direction (up in 1999, down in 2008). This suggests that to the extent investors must market time based on valuations, they should not just focus on valuation, but also bet against extreme sentiment, the kind that appears at most once in a decade.