Last month, we looked at the relationship between dividend yield and the subsequent 5-year return for the U.S. stock market. The motivation: searching for a liaison that binds relatively high yields with relatively high returns. If it holds, then low yields lead to low/negative returns.
In our short sample of history, the relationship held up rather well. For the period January 1995 through February 2003, higher yields were linked with higher returns over the next five years. Indeed, running a regression analysis on the data produced a persuasive 0.95 R-squared. (A quick digression: R-squared ranges from 0 to 100. A reading of 100 means that the variance of one factor fully explains the movement of the other. A reading of 0 tells us that there’s no relationship between two variables. In short, the higher the R-squared, the greater the influence of one factor on the other.)
We also warned in that post that the strong relationship documented in 1995-2003 wasn’t absolute. We cautioned readers to “be careful about thinking the relationship offers easy and sure profits.” It doesn’t. Bottom line, we counseled last month to remain “skeptical” whenever someone shows you a relationship that purports to bring easy profits. No such animal exists, even if the evidence suggests otherwise in the short run.
Our essay inspired others to run a deeper analysis of dividend yield and stock returns. The Aleph Blog, for instance, correctly points out that over a longer sweep of history, the 0.95 R-squared we quoted falls dramatically. Using data from Professor Robert Shiller, Aleph reports that the R-squared for dividend yield and subsequent stock market returns is a mere 0.07 for 1871 to 2003. Crunching Shiller’s numbers on our own, we come up with a similar reading.
Does that mean that looking at the market’s dividend yield is worthless? No, not at all. Although a simple analysis of one holistic view of 100-plus years of stock market history seems to tell us otherwise, there’s a compelling reason to look at dividend yields as one of several metrics for judging prospective return for equities generally. No, we shouldn’t drive blindly down this road, or think we’ve stumbled upon a short-term trading strategy. But with eyes wide open, we should consider the possibilities of using dividend yield as one of several metrics for judging long-term return opportunities.
To explain why, let’s begin by pointing out that there are no absolutes in investing. If it were otherwise, we’d all be rich with minimal effort. Economics, in that case, would cease to exist as we know it. In fact, risk is always and forever present in the money game, which is to say that we don’t know if (or when) an investment strategy will pay off. That’s the nature of risk, and so we must proceed cautiously, gathering whatever clues history sends our way in an effort to manage risk as intelligently as possible.
That said, let’s recognize that a number of studies over the years have found a relationship between dividend yield and return, i.e., higher yield sometimes leads to higher returns. Again, it’s not an absolute relationship, but it’s sufficiently robust to compel us to monitor the trend and, at times, act on it when the odds appear in our favor.
We cited one study in our post from last month. We could cite many others, including one from Professor Burton Malkiel, who reviews the evidence for using valuation to forecast performance. One test: comparing dividend yields for the U.S. stock market (measured by the S&P 500) for each quarter from 1926 through 2001 relative to subsequent 10-year total returns. The result, he writes, “shows that investors have earned a higher rate of return from the stock market when they purchased a market basket of equities with an initial dividend yield that was relatively high and [earned] relatively low future rates of return when stocks were purchased at low dividend yields.” Similar results are found using price-earnings (p/e) ratios: buying the stock market when p/e ratios are relatively low leads to higher returns in the next 10 years, and lower returns when stocks are purchased at relatively high p/es.
Valuation, it seems, matters. Not always, but enough of the time to make the subject a worthwhile area of study. But dividend yields are also quite nuanced when we look over long periods of time and so we must be careful of reading too much into one massive overview that attempts to summarize decades of market data in search of one enduring rule of thumb. Consider again the low R-squared for dividend yield and subsequent return for the 1871-2003 period. True enough, but when we look more closely we find that the R-squared waxes and wanes through time.
Consider our chart below, which tracks a rolling 10-year R-squared for current dividend yield and subsequent 5-year return for 1881 through 2003. Note the recurring cycles. During some periods, the R-squared is quite high; sometimes it’s quite low. The implication: using dividend yield as a window for estimating future returns is more valuable at certain times, but not always.
In a perfect world, the R-squared would remain at 1.0. In that case, dividend yield would be a perfect forecasting tool for returns, in which case everyone would always have perfect foresight about prospective equity performance. But we live in an imperfect world and so dividend yield is less than perfect, just like every other metric and strategy in the investment game. If we wait for the perfect methodology for managing money, we’ll wait in vain.
Nonetheless, we must recognize the limitations of our analysis and act accordingly. For instance, dividend yield’s limits as a forecasting tool compels us to look at other metrics, such as p/e ratio and earnings yield. We should also monitor interest rates for additional context. Ditto for knowing where we are in the economic cycle. Putting dividend yield in financial and economic context, in other words, gives us more information for deciding if the yields we’re looking at in real time are relevant at the moment, or not.
Indeed, if we only consider those points in time when dividend yield is relatively high (based on, say, the previous 20 years), this metric fares better as a tool for estimating future equity returns. The same is true if we look at yield when it’s relatively low. Extremes are, by definition, rare, and so by this standard dividend yield will be used infrequently. Using it each and every month, by contrast, and thereby ignoring the larger context, will likely lead to mediocre results.
Don’t misunderstand: it’s not easy. There are no turnkey software packages that spit out the right answer on a regular basis. It’s always hazardous to make decisions in real time. But strategic-minded investors ignore dividends at great risk. Over time, dividends (including their reinvestment) have proven to be a crucial part of total return in the U.S. market. One quick example: for the 20 years through 2005, $1 invested in the S&P 500 grew to $1.45 in price-only terms. The same dollar grew to $9.54 if we include dividends and their reinvested return. (The source for this data is How to Select Investment Managers & Evaluate Performance, p. 115.)
Dividend yields can tell us a lot. It’s not a crystal ball, and if we’re not careful it can even lead us astray. But that shouldn’t keep us from studying dividend yields. As valuation metrics go, this one’s too important to ignore.