Second Quarter 2016 Newsletter Commentary

Second Quarter 2016 Newsletter Commentary

Click for a printable version

Valuation spreads are wide, approaching those last seen during the internet bubble. History strongly suggests this is a significant opportunity for value.


Throughout our history we have consistently highlighted the phenomenon of valuation spreads. Spreads measure the valuation difference between cheap stocks and "the market" and so naturally are of interest to value investors. To put it simply, when spreads are wide, the opportunity for value investing increases. Figure 1 illustrates this idea for a global universe comprising the largest 1600 stocks in the developed world1 over a period extending from December 1974 through June 2016.

Figure 1 presents the theoretical upside to a first quintile stock upon getting re-rated to a book-to-price valuation commensurate with that for the equally-weighted universe (Please see the Methodology section at the bottom of the page). A value of 150%, for example, would imply a fair value (upon re-rating) of 2.5x the current stock price. Wide spreads are characterized by a high implied upside for the cheapest quintile, with narrow spread environments by low values, as the figure indicates.

A visual inspection of Figure 1 suggests a cyclical pattern for spreads, albeit one where the cycles are neither regular nor predictable. Periods of wide spreads represent times when investors come to believe in "one thing," shunning one set of stocks in favor of another. The most extreme example of this was the 1998-1999 period, where investors came to believe that "old economy" stocks tied to the negative events of the Asian currency crisis were to be sold in favor of the limitless possibilities available in the dot-com boom. As is evident from the chart, the resulting spread in valuations grew to be the largest in the history, and an investment in Q1 stocks at that time implied a theoretical upside of more than 150%, when spreads blew out to a level more than 4 standard deviations2 above their long-term average level.

Looking at spreads by region

While the global spreads of Figure 1 are certainly of interest, they are also affected by differences in valuation across regions, making interpretation of the data more difficult. Figures 2-4 separately examine spreads for the United States, Europe, and Japan. As is the case with Figure 1, each chart shows three horizontal lines representing the long-term mean and ±1 standard deviation relative to the mean. Each region shows a cyclical pattern of spreads tied to fear and greed, periods when investors came to believe in the "one thing." The United States chart, like the global one, is dominated by the 1998/1999 spike. However, other periods of wide spreads are visible as well, such as the "Nifty Fifty" era of the early 1970's (Polaroid at a 400 P/E!), the onset of the 1980 recession, and the commercial real estate crisis of the early 1990's. In Japan, the 1987 asset bubble stands out. In Europe, the mid-90's recession (Germany's post- reunification blues) created worries visible in the lowest rated stocks, and the period since 2008 has been dominated by first the Global Financial Crisis and more recently the Sovereign crisis affecting the Eurozone in particular.

While the above observations look at the spread of cheapest quintile relative to the midpoint of the universe, it is also instructive to examine the spread between the lowest and highest quintiles of valuation. In doing so, one gets a better representation of true valuation dispersion. Figure 5 plots this spread by region, now measured in terms of the number of standard deviations vs historical averages rather than in terms of the percentage upside.

The figure shows that in the United States, 1Q vs 5Q spreads at the end of June were close to 4 standard deviations wide versus historical average, with Europe and Japan about 2 standard deviations wide. These high levels suggest that part of the wideness we observe in spreads today is due to the high valuations for fifth quintile stocks in comparison with history. Investors are clearly awarding premium valuations to certain segments of the market. Interestingly (though we do not show the data here), those segments are largely composed of two types of stocks. As one would expect, there are the high growers (e.g., parts of technology and biotechnology), but there are also many stocks in health care and consumer staples, areas more associated with stability of earnings than with growth. In the uncertain macro environment that has prevailed since the Global Financial Crisis, so-called bond proxies have essentially become momentum stocks. The data suggest that one possible path for spreads to narrow toward normal levels is for confidence to rebound and for this group to under-perform.

Spreads and performance

Theoretical upside as described by spreads is fine, but the more interesting question is whether wide spreads have historically been associated with subsequent outperformance for value strategies. To put it another way, when spreads get to an interesting level (say, one standard deviation wider than average), what does that signal for future relative performance? To explore the issue, we examined all incidents where spreads crossed the one-standard deviation threshold, then looked at subsequent 3- and 5-year performance for Q1 versus the market capitalization weighted performance of the defined universe.3

The results are summarized in the table of Figure 6. There are a total of 14 times in the observation period where we could observe subsequent 3-year performance (6 for the United States, 6 for Europe and 2 for Japan) and 12 such times with subsequent 5-year performance observable 3-year relative performance (alpha) for Q1 was positive for 13 out of 14 observations across the three regions (6 out of 6 for United States, 5 of 6 for Europe, and 2 of 2 for Japan). The average 3-year alpha ranged from 7.6% to 12.6% across the three regions. The picture for 5-year relative performance is similar, with positive alpha in 11 out of 12 observations.

These numbers strongly suggest that periods of wide spreads signal opportunity. Investors willing to be patient for 3-5 years following a widening of spreads are overwhelmingly likely to be well-rewarded. This is not to suggest that precise timing strategies are readily available: the findings do not apply to shorter time frames, and even when we restrict attention to 3- and 5-year horizons, history is not without its painful exceptions, including the very recent period. Indeed, the data highlight the misery of value-based strategies over the past eight years. In October of 2008, spreads in Europe crossed one standard deviation, but subsequent 3-year and 5-year relative performance was poor. Post-2011 performance for low P/B has been similarly disappointing in the United States and the five year record is likely to be negative.

Nevertheless, the historical numbers remain extraordinarily compelling. And with wide spreads across all developed world regions today, a phenomenon not observed over the historical period, the environment suggests a bullish outlook for value investors today.


We express the spread S as the ratio of two book-to-price ratios minus 1. The numerator is the book-to-price for the cheapest quintile of stocks, while the denominator is the book-to-price for an equally weighted average of the universe.

In computing spreads for this analysis, we first defined a relevant universe for each region to roughly resemble the corresponding performance benchmark:

  1. For the United States we took the 1000 largest stocks (approximating Russell 1000).
  2. For Europe we used the 500 largest stocks (including stocks from Switzerland and the U.K.), thus approximating MSCI Europe.
  3. For Japan we used the largest 300 stocks (approximating MSCI Japan). For each universe U we defined the longest possible time

period over which data were available (going back to the 1960's for the United States universe). We obtained B/P at the end of every month m over the range for the universe as a whole (simple average across stocks in the universe) and for each of five quintiles ranked on B/P (Q1 having the highest ratio, i.e., the cheapest valuation for that month-end). On the basis of these data, simple spreads for each month-end m are expressed as:

Figures 1-4 use this statistic, which represents the percentage upside for Q1 upon revaluation to the universe average. Across all month-ends we also computed longitudinal means and standard deviations for S, denoted by µ (S) and s(S). The horizontal lines in Figures 1-4 represent this mean ±1 standard deviation.

To produce Figure 5 we used z-scores based on the longitudinal means and standard deviations. At month-end, the z-score is computed as

It expresses that month's spread in terms of the number of standard deviations away from the mean. Figure 5 plots these z-scores for the different regions.

For relative performance following 1- s spreads:
We define incidents where spreads cross through one standard deviation as meeting two criteria:

1.) The standard deviation for at least the preceding 9 months is less than 1; and 2.) The standard deviation of the spreads at the incident point is equal to or greater than one.

We then computed relative performance (alpha) for 3-year and 5-year periods subsequent to the defined crossing points.

  1. 1 Emerging markets are not part of this analysis due to the limited histotry available.
  2. 2Standard deviation is a measure of the dispersion of a sample around its arithmetic average. Because we need to compute average and standard deviation, we use B/P rather than P/B; the latter can assume values approaching infinity since B can approach 0 or even negative values.
  3. 3These universes roughly approximate the Russell 1000 for United States, MSCI Europe, and MSCI Japan. We follow standard practice in measuring performance relative to (cap-weighted) benchmarks. This means that measured alpha technically captures both a value effect (based on spreads) and a market cap effect. We explore this further in an upcoming white paper, but note here that while commingling the two effects does increase the magnitude of the observed alpha, it does not change the basic finding that wide spreads do signal opportunity for value-based strategies.


Past performance is no guarantee of future results. The historical returns of the specific portfolio securities mentioned in this commentary are not necessarily indicative of their future performance or the performance of any of our current or future investment strategies. The investment return and principal value of an investment will fluctuate over time.

The specific portfolio securities discussed in this commentary were selected for inclusion based on their ability to help you understand our investment process. They do not represent all of the securities purchased, sold or recommended for our client accounts during any particular period, and it should not be assumed that investments in such securities were, or will be, profitable.