Articles
Index providers continue to deliver new ways to access equity markets in response to investor demands in a highly uncertain climate. Low-volatility and equity-income indexes are just a few examples. Yet two basic mainstays—economic sector indexes and style indexes—remain hugely popular with investors as passive investment vehicles, analytical tools and benchmarks. The sector pie chart and the style box are the first tools most investors reach for—or at least the first tools they see—when reviewing any equity market. The Sector Perspective Evaluating Sector Classification Methodologies
Take basic materials, for example. Coal companies are considered by some to be in the basic materials sector, and by others to be in the energy sector, and the choice of where coal goes can make all the difference in comparing these funds. Look no further than the divergence in performance over the past year through the end of June between the Vanguard Materials ETF (NYSE Arca: VAW) and the iShares Dow Jones U.S. Basic Materials ETF (NYSE Arca: IYM). Both funds cover the U.S. basic materials segment and yet VAW outperformed IYM by 8.8 percent. The difference? IYM's index provider uses the ICB system, which considers coal to be a basic material; VAW's GICS-based index considers firms mining coal to be in the energy sector. Coal got clobbered in the past year, as mounting political pressure, falling natural gas prices (an easy substitute) and the bankruptcy of Patriot Coal weighed on the industry. Other examples abound. Is Amazon a technology company or a retailer? Is Berkshire Hathaway an insurance firm or an investment management company? Is Visa a consumer stock or a financial company? Things get more complicated as you move into the consumer space. ICB splits consumer markets into goods and services, while GICS and TRBC break it down by cyclicality: consumer non-cyclicals or staples; and consumer cyclicals or discretionary. Figure 2 lists the top 10 holdings in the Dow Jones Consumer Goods and Consumer Services (ICB) indexes and the TRBC consumer cyclicals and non-cyclicals indexes. The difference in approach between the two systems plays out as you might expect. The Dow Jones indexes, which break the market out by goods and services, overlap with both the TRBC consumer cyclical and non-cyclicals indexes. By breaking out the market by the type of consumer product as opposed to whether the good or service is cyclical or non-cyclical, the index provider creates economic overlap between the two sectors.
Any investor looking to employ a sector strategy must understand these differences to ensure the exposure they are getting is exactly what they want. For our performance analysis, we used the widely followed suite of GICS-based indexes published by Standard & Poor's.
Sector Performance
While some might argue this highlights the case for long-term buy-and-hold sector strategies, it actually forms the basis of a case for more tactical sector use. That is supported by the widely variable year-by-year performances; the leader in one year can and does become the laggard in the next. Figure 4 shows just how much these results can vary year to year. Here we have the annual performance of each sector from 2002 to 2012 with each year's best performer highlighted in green and the worst performer highlighted in red. Over the past 10 years, each sector, with the exception of utilities and industrials, has represented either the best or worst performer at least once. This further underscores the distinctive performances offered by sectors.
Of course, the performance of each sector provides little information without proper context, which is where correlation comes in (see Figure 5).
Over all time frames studied, there have been consistent intersector correlation patterns. For example, over the past year, utilities firms have shown very low correlation to basic materials and technology firms, and muted correlations to consumer discretionary stocks. As the time frame expands, utilities show decreasing correlations to all sectors. In fact, over the past five years, the average correlation between sectors is just 0.68. At the same time, the elevated correlations between some sectors also highlight how properly defined sectors will show logical economic relationships. Consumer staples and health care firms—two defensive sectors—have shown high correlations to each other, and are the only two sectors that have shown average or better correlations to utilities. These sectors are all less dependent on high rates of economic growth than, say, the industrials or consumer cyclicals sectors. The convergence of returns among the more cyclical sectors of the market—technology, materials, consumer discretionary and industrials firms—are much more dramatic. Not only are correlation pairs between these sectors above average, they reach as high as 0.911 (industrials and consumer discretionary). Of course, every sector is affected by different economic indicators, but there is also significant overlap. Properly defined sectors should therefore show high intersector correlation between segments of the economy whose economic exposures are similar. For example, energy and basic materials are highly dependent on global GDP rates, construction spending, and mining activity. We would therefore expect the two sectors to behave similarly. This is exactly what we have seen, as the intercorrelation between the two sectors did not drop below 70 percent in any of the periods of study and was as high as 91 percent over the past year. These correlation groupings provide an outlet for investors to express their opinion about the economy and the markets. But these relationships are dynamic. One way to show this is by charting the correlations on a rolling basis, as shown in Figures 6a and 6b.
As the market and economy have ebbed and flowed, so has the correlation between technology and the rest of the market. This tells a story about the market and the economy. Negative correlations do not persist over time, but on a rolling basis, different sectors will show negative correlations with each other. These occurrences, while fleeting, are immensely valuable to investors as they allow for true risk diversification. At the height of the tech bubble, technology actually had negative correlations to energy and materials firms, and during the subsequent market sell-off, its correlation to all sectors normalized. This underscores a common theme in the period of study: During times of market stress, correlations all converge to 1. While this impairs an investor's ability to diversify away market risk during these periods, it is also a predictive piece of information for investors. Further, these correlation convergences do not persist over time, moving away from 1 as a crisis abates. Looking at Figure 7, we see that during the 2008-2009 financial crisis, the performance correlation of all sectors spiked to 1 but normalized as the economy moved out of the recession.
Even when we measure sectors individually against the market—in this case, the S&P 500—we see a wide range of correlations. And just as with intersector correlations, each sector's relationship with the market changes over time. Over long horizons, each sector has a lower correlation to the market, which is logical. Since the S&P 500 is a roll-up of the firms in each sector, the changes in the importance and influence of each sector over time is represented by changes in each sector's weighting in the index. When a sector like energy becomes an increasingly significant portion of the market, it follows that it will have an increasing correlation to the broad market. Between 1998 and 1999, technology went from 17.7 percent of the S&P 500 all the way up to 29 percent before falling to 14.3 percent two years later. This coincided with a steep rise and fall in its correlation to the market. The financial crisis provides a good reference point for sector correlations as well, and the data seems to support the commonly held belief that correlations across all assets classes—including between sectors of the equity universe—have spiked. The data also show that this may be changing. Over the past 15 years, sectors have shown extremely low correlations between each other, with an average of 0.50 and a high of 0.85 (industrials and consumer discretionary). Over the past 10 years, it is also low, but the average has crept up to 0.61, with the highest correlation at 0.88. Over the past five years, that average correlation among sectors climbs all the way to 0.68, with eight different sector correlations above 0.85. The past three years had an even higher average correlation among sectors, but in the past year, that figure has dropped back down to 0.62. Clearly, the ability to effectively use low correlation pairs is compromised in times of significant financial stress, but it seems the further removed we get from the financial crisis, the more pronounced the divergence among returns is. How the various sectors move in relation to each other and to the market is just part of the story. We must also analyze how much volatility each sector has shown historically and whether that has changed over time. As we would expect, the defensive sectors—consumer staples, utilities and healthcare—all showed 20 percent or less annualized volatility over all time frames. On the opposite end are financial and energy firms, which showed volatility in excess of 25 percent over all periods of study. Recent history bears this out. Financials were the hardest-hit sector through the financial crisis of 2008 and the recent European debt crisis. Meanwhile, energy prices spiked in 2007, only to come crashing down to earth before spiking again. This distribution of risk among the sectors offers even more information to investors attempting to fine-tune their risk exposure in various market environments. The recent wave of high- and low-beta and volatility index strategies is a logical extension of these sector risk profiles. Whereas index providers and ETF issuers are looking to provide new ways to slice the market, for investors focused on risk as opposed to exposure, sectors already allow them to do this. In all, the intuitive nature of sectors lines up with statistical evidence. Over the past 15 years, with the exception of times of financial stress, sectors have shown all of the necessary characteristics to prove how valuable they are in asset allocation. Their returns vary greatly, their individual performances show economically logical relationships, and they exhibit distinctly different volatility patterns. The Style Perspective Investors have long used a value and growth perspective to parse the market in a different way than sectors. The problem is that the line between growth and value is blurry, and the overlap in exposure between high- and low-correlated sectors diminishes the efficacy of this strategy as a way of segmenting the market. Investors of all sizes have made style-based investing hugely popular over the years. The stereotypical value investor wants to buy stocks that are lower in price or are otherwise out of favor. Value investors might also want to see consistent positive earnings, especially when regularly paid out as dividends. A growth investor is willing to pay a higher relative stock price with the goal of latching on to the next Apple Computer, i.e., a stock with huge price appreciation driven by earnings growth beyond expectations. Style indexes use a variety of fundamental measures to describe stocks as value or growth oriented. These include price/book, as mentioned above, as well as price/earnings, price/sales and price/cash flow. In addition, indexes often include stock price momentum and various growth rates, using historical and forward-looking estimates. When stocks are screened by these metrics, results don't always fall into neat buckets that clearly indicate value or growth. Index designers need to decide what to do with muddled results from firms that sit in the middle gray zone between the two extremes. Some indexes choose to split a stock's weight across the value and growth buckets, with the benefit being growth and value indexes that roll up into a complete picture of the market. Other indexes assign the stocks that don't show strong style biases into a separate core bucket, leaving the growth and value indexes with more "pure" components. These core stocks—those firms that do not show a pronounced growth or value bias based on the aforementioned metrics—muddy the growth and value picture. Because they show characteristics of both growth and value, they end up detracting from the ultimate goal, which is to separate the market into two distinct exposure groups. The removal of these firms should exaggerate the difference between growth and value. The problem is that there is still too much ambiguity and disagreement over what makes a company a growth or value firm. A growth firm? It is just not an intuitive concept. You could ask 10 different people what Chevron is, you will get 10 identical answers: It's an energy firm. Ask the same 10 people if Chevron is a growth or value firm and you may get 10 different answers, along with some quizzical looks. For our purposes, we chose to look at the pure style approach—the S&P 500 Pure Growth and Pure Value Indexes—in an attempt to measure the performance of indexes pulled from the same universe as our sector indexes and that emphasize style.
Style Performance
These summary figures don't capture the full story, however. Correlations vary significantly over time, as the graph of one-year rolling correlations shows in Figure 9. In the recent past, we see correlations approaching 1 during the financial crisis, then decreasing in 2010 as markets recovered and shifted away from the risk-on/risk-off mentality, and most recently increasing again as economic recovery falters in the U.S. and debt worries dominate in Europe. Regardless, the trend is clear: a sharp spike in average correlations, with fewer periods of significant dispersion. Pure growth, in particular, has enormously high correlations to the S&P 500, over virtually any period studied, suggesting that any diversification benefit lies in the value portion of the equation.
Significantly lower correlations exist in brief periods in late 2007 and summer of 2006, but clearly the most dramatic divergences between the two style indexes occur in the 1999 to 2002 period, which saw tech boom and bust. Investors looking for a clearly different pattern of returns between the two style indexes going forward have little to hang their hat on here. Recent history shows that return patterns have diverged occasionally; sometimes by a great amount. But accessing this divergence seems more akin to tactical rather than strategic allocation—timing the market, in other words. As with the sectors, the style correlations tend to converge during periods of high volatility and market stress, which makes intuitive sense. If the companies in each sector are moving increasingly in lock step with each other, it stands to reason that portfolios cutting across these sectors would as well, regardless of their exclusion of a portion of the market (core). Correlations are only part of the picture. Risk/return measurements provide more insight into recent history of the pure style indexes. The low correlation over the 15-year period, driven in part by the low and negative correlations seen in the 1999 to 2002 range, might lead one to expect that returns over the 15 years would differ greatly from value to growth. In fact, the compound annualized returns differ by only 13 bps, with value at 8.0 percent and growth at 8.1 percent. The cumulative return chart for the period highlights the different paths the two indexes took to the same end (see Figure 10).
Relative to the market itself, returns for growth as well as value exceeded those of the S&P 500 in three of five periods, lagged it in one and split in one (see Figure 11). The split was in the five-year period where the S&P 500 was essentially flat, but growth beat both the market and value handily mostly due to its underexposure to financials as described above. The style indexes lagged the market in the one-year period, with growth dragged down by materials and industrials and value by financials and energy.
Growth returns showed less volatility than value over four of the five periods too. The difference in annualized standard deviation of weekly returns was greatest in the five-year period, where the volatility of the financial sector (43.4 percent) drove value volatility higher than that of growth. More interesting perhaps is that volatility of both the style indexes exceeds that of the market in every time period we looked at. Volatility in growth-oriented sectors came from consumer cyclicals and tech, while volatility on the value side was driven by financials and energy. We ran regressions of growth and value on the S&P 500, but the fit of the estimates to the data, especially for the longer time periods, was insufficient for us to feel comfortable with them. Still, the regressions generally hinted at higher beta for both pure style indexes, consistent with the higher return and higher volatility shown above.
Investors who come to index investing from a risk perspective rather than an economic one might reach for style more on reputation than reality. Perhaps they'd think a growth fund suits a younger investor with a long time horizon, while a value index works better for a pensioner's equity exposure. Our data show, however, that both value and growth have been more volatile than the market in each of the five time periods we looked at (see Figure 11). Yet as mentioned above, certain sectors have shown consistently low volatility (consumer staples, health care, utilities) just as other sectors have shown consistently higher volatility (financials, energy, materials). This suggests that sector exposure may be a more effective way for investors to find the desired spot on the risk/reward spectrum (see Figures 12a and 12b). The S&P 500 Low Volatility Index behind the wildly popular PowerShares SPLV ETF shows strong biases to these low-volatility sectors, while exhibiting mixed style messages (growthlike high P/E and P/B mixed with a valuelike high-dividend yield).
Can style indexes be viewed as aggregates of sector exposure? Exhaustive style pairs hold the entire market between them, and so must hold all the sector stocks. Style indexes certainly show bias toward certain sectors, as we've highlighted. Yet the aggregation of sectors into style indexes is far from neat. Industrials and basic materials, for example, are split about equally between the pure growth and value indexes. Pure style indexes leave out about one-third of the market, which undercuts their ability to reflect groups of sectors, even if they were perfectly defined. In the end, style indexes are engineered to select and weight stocks by fundamental and momentum factors, not by industry exposure. A value fund captures financial stocks by their low P/Bs (from high book values) but also snares energy stocks with low P/Es (from high earnings). However, some firms from these two sectors exhibit different ratios that park them in the pure growth index (which has roughly 10 percent combined financials and energy). Style funds work with discrete buckets sizes. Exhaustive style pairs split the market 50/50. Style-pure indexes split the market in thirds. A stock migrates over boundaries because its attributes change (a higher P/B ratio forces it out of value and into core perhaps) or because the attributes of other stocks get stronger and force it out of the bucket. In this sense, the indexes remain true to their style mandate while the names inside it change. A pure growth fund will always represent the "growthiest" one-third of the market, even if the nature of the constituents changes (e.g., less tech, more health care). A sector fund stays true to its mandate as well, but its constituents tend to stay put. A tech firm, for example, stays in a tech index as it journeys from brash startup to cash-cow dividend payer. In contrast, the same firm might migrate out of a growth index and into a core index (or have its weight partly assigned to value). The tech fund's footprint in the market—the relative size of each sector's bucket, in other words—can vary dramatically over the long run, as its aggregate market value of equity ebbs and flows.
Takeaways |
Previous Page
