Tobin’s Q, a market-valuation metric, is back in the news, in part thanks to a widely read Bloomberg article. The story, published earlier this week, offered the provocative title: “Nobel Winner’s Math Is Showing S&P 500 Unhinged From Reality.” Oh, my, that sounds serious. Or maybe not. As several observers have pointed out (see here, here, and here, for instance), the practical connection between real-world money management and Tobin’s Q is a dicey proposition. It does, however, provide the raw material for dramatic headlines.

Bloomberg’s article brings up the issue of valuation metrics generally, which should be used with caution. The problem is two-fold. First, accounting-based measures of stock market valuation aren’t timely–the underlying data arrives with a considerable lag. Corporate earnings for a given quarter, for instance, are published months after the fact. Second, the link between valuation and return is relatively weak for time horizons below three years. That’s not an issue if you’re truly investing with a very long horizon. But that’s a very small minority of investors (despite the claims to the contrary).

That doesn’t mean we should ignore valuation. But in a world with many possibilities for estimating value (or the lack thereof), Tobin’s Q certainly wouldn’t be my first choice (Professor Shiller’s CAPE, for example, is a more practical albeit still-flawed alternative). In any case, “high” and “low” market-valuation definitions come in many flavors. But the idea that any one—even a relatively robust measure–is a silver bullet for making real-time decisions about expected return for the near-term horizon is just asking for trouble.

The bigger lesson is that if you’re modeling the market in an effort to develop some relatively objective measures on the potential for severe corrections (aka bear markets and crashes), it’s wise to look at a diversified set of gauges. Market valuation can and probably should be on the short list, but only with the understanding of its limitations. Its one factor and arguably an important factor, but its application as a timing tool waxes and wanes. Unfortunately, it’s rarely clear when it’s useful, or not, without the benefit of hindsight. Actually, that’s true for every other metric, which is why it’s crucial to think long and hard about building a robust benchmark that draws on an array of signals for monitoring what we might call crash-risk potential. But that’s a subject for another day.

As for market valuation, we can’t do much about its limited use for anticipating near-term returns. On the other hand, perhaps there’s a slightly better definition of valuation in the day-to-day business of monitoring risk. One possibility is using rolling five-year return as a proxy for valuation, a time frame that Asness, et al. consider in a 2013 study in the Journal of Finance (“Value and Momentum Everywhere”). The advantage is that the data is available in real time and is immune to revisions and various errors that have been known to bedevil estimates of earnings, book value, etc.

With that in mind, let’s review the history of the S&P 500’s rolling five-year annualized performance over the past half century on a daily basis. As you can see, Mr. Market has a history of going to extremes, albeit infrequently, based on trailing five-year return.


For another perspective, let’s transform the rolling return into percentile rank.


Sometimes extreme valuation is a useful signal for anticipating expected return in the near term, but not always. For instance, the market’s trailing five-year performance was in nose-bleed territory for several weeks in 2014’s first quarter–i.e., return was above the 95th percentile. As it turned out, that was a false alarm in the sense that the market has rallied further since those comparatively extreme returns of more than a year ago.

It was a different story when the 95th percentile was breached in 2000, which turned out to be a timely warning ahead of the hefty 2001-2003 correction. Then again, the market’s five-year return had been at or above the 95th percentile since 1998.

Valuation, in other words, is useful but we shouldn’t use it in isolation. Developing broader context is essential via the business cycle, the financial climate, and for the market across multiple dimensions. Valuation deserves to be part of that mix. But like every other quantitative measure of risk, it’s not perfect.

The good news is that we can build market-risk indexes that are superior to any one component by intelligently blending several metrics. The improvement, in the best of circumstances, will be modest, but that’s the nature of the money game–progress and enlightenment arrive in incremental doses. But no one knows which risk measures will be relevant on any given day. Developing a diversified measure of stock market risk is the worst possible solution… except when compared to everything else (apologies to Winston).