AN INTEGRATED APPROACH TO A SEGMENTED MARKET

While one might think that U.S. equity markets are fluid and fully integrated, in reality there are barriers to the free flow of capital. Some of these barriers are self-imposed by investors. Others are imposed by regulatory and tax authorities or by client guidelines.

Some funds, for example, are prohibited by regulation or internal policy guidelines from buying certain types of stock—nondividend-paying stock, or stock below a given capitalization level. Tax laws, too, may effectively lock investors into positions they would otherwise trade. Such barriers to the free flow of capital foster market segmentation.

Other barriers are self-imposed. Traditionally, for example, managers have focused (whether by design or default) on distinct approaches to stock selection. Value managers have concentrated on buying stocks selling at prices perceived to be low relative to the company's assets or earnings. Growth managers have sought stocks with above-average earnings growth not fully reflected in price. Small-capitalization managers have searched for opportunity in stocks that have been overlooked by most investors. The stocks that constitute the natural selection pools for these managers tend to group into distinct market segments.

Client preferences encourage this balkanization of the market. Some investors, for example, prefer to buy value stocks, while others seek growth stocks; some invest in both, but hire separate managers for each segment. Both institutional and individual investors generally demonstrate a reluctance to upset the apple cart by changing allocations to previously selected style managers. Several periods of underperformance, however, may undermine this loyalty and motivate a flow of capital from one segment of the market to another (often just as the out-of-favor segment begins to benefit from a reversion of returns back up to their historical mean).

The actions of investment consultants have formalized a market segmented into style groupings. Consultants design style indexes that define the constituent stocks of these segments and define managers in terms of their proclivity for one segment or another. As a manager's performance is measured against the given style index, managers who stray too far from index territory are taking on extra risk. Consequently, managers tend to stick close to their style homes, reinforcing market segmentation.

An investment approach that focuses on individual market segments can have its advantages. Such an approach recognizes, for example, that the U.S. equity market is neither entirely homogeneous nor entirely heterogeneous. All stocks do not react alike to a given impetus, but nor does each stock exhibit its own, totally idiosyncratic price behavior. Rather, stocks within a given style, or sector, or industry tend to behave similarly to each other and somewhat differently from stocks outside their group.

An approach to stock selection that specializes in one market segment can optimize the application of talent and maximize the potential for outperformance. This is most likely true for traditional, fundamental analysis. The in-depth, labor-intensive research undertaken by traditional analysts can become positively ungainly without some focusing lens.

An investment approach that focuses on the individual segments of the market, however, presents some theoretical and practical problems. Such an approach may be especially disadvantaged when it ignores the many forces that work to integrate, rather than segment, the market.

Many managers, for example, do not specialize in a particular market segment but are free to choose the most attractive securities from a broad universe of stocks. Others, such as style rotators, may focus on a particular type of stock, given current economic conditions, but be poised to change their focus should conditions change. Such managers make for capital flows and price arbitrage across the boundaries of particular segments.

Furthermore, all stocks can be defined by the same fundamental parameters—by market capitalization, P/E, dividend discount model ranking, and so on. All stocks can be found at some level on the continuum of values for each parameter. Thus, growth and value stocks inhabit the opposite ends of the continuums of P/E and dividend yield, and small and large stocks the opposite ends of the continuums of firm capitalization and analyst coverage.

As the values of the parameters for any individual stock change, so too does the stock's position on the continuum. An out-of-favor growth stock may slip into value territory. A small-cap company may grow into the large-cap range.

Finally, while the values of these parameters vary across stocks belonging to different market segments—different styles, sectors, and industries—and while investors may favor certain values—low P/E, say, in preference to high P/E—arbitrage tends to counterbalance too pronounced a predilection on the part of investors for any one set of values. In equilibrium, all stocks must be owned. If too many investors want low P/E, low-P/E stocks will be bid up to higher P/E levels, and some investors will step in to sell them and buy other stocks deserving of higher P/Es. Arbitrage works toward market integration and a single pricing mechanism.

A market that is neither completely segmented nor completely integrated is a complex market. A complex market calls for an investment approach that is 180 degrees removed from the narrow, segment-oriented focus of traditional management. It requires a complex, unified approach that takes into account the behavior of stocks across the broadest possible selection universe, without losing sight of the significant differences in price behavior that distinguish particular market segments.

Such an approach offers three major advantages. First, it provides a coherent evaluation framework. Second, it can benefit from all the insights to be garnered from a wide and diverse range of securities. Third, because it has both breadth of coverage and depth of analysis, it is poised to take advantage of more profit opportunities than a more narrowly defined, segmented approach proffers.

A Coherent Framework

To the extent that the market is integrated, an investment approach that models each industry or style segment as if it were a universe unto itself is not the best approach. Consider, for example, a firm that offers both core and value strategies. Suppose the firm runs a model on its total universe of, say, 3,000 stocks. It then runs the same model or a different, segment-specific model on a 500-stock subset of large-cap value stocks.

If different models are used for each strategy, the results will differ. Even if the same model is estimated separately for each strategy, its results will differ because the model coefficients are bound to differ between the broader universe and the narrower segment. What if the core model predicts GM will outperform Ford, while the value model shows the reverse? Should the investor start the day with multiple estimates of one stock's alpha? This would violate what we call the law of one alpha.³

Of course, the firm could ensure coherence by using separate models for each market segment—growth, value, small-cap, linking the results via a single, overarching model that relates all the subsets. But the firm then runs into a second problem with segmented investment approaches: To the extent that the market is integrated, the pricing of securities in one segment may contain information relevant to pricing in other segments.

For example, within a generally well-integrated national economy, labor market conditions in the United States differ region by region. An economist attempting to model employment in the Northeast would probably consider economic expansion in the Southeast. Similarly, the investor who wants to model growth stocks should not ignore value stocks. The effects of inflation, say, on value stocks may have repercussions for growth stocks; after all, the two segments represent opposite ends of the same P/E continuum.

An investment approach that concentrates on a single market segment does not make use of all available information. A complex, unified approach considers all the stocks in the universe, value and growth, large and small. It thus benefits from all the information to be gleaned from a broad range of stock price behavior.

Of course, an increase in breadth of inquiry will not benefit the investor if it comes at the sacrifice of depth of inquiry. A complex approach does not ignore the significant differences across different types of stock, differences exploitable by specialized investing. What's more, in examining similarities and differences across market segments, it considers numerous variables that may be considered to be defining.

For value, say, a complex approach does not confine itself to a dividend discount model measure of value, but examines also earnings, cash flow, sales, and yield value, among other attributes. Growth measurements to be considered include historical, expected, and sustainable growth, as well as the momentum and stability of earnings. Share price, volatility, and analyst coverage are among the elements to be considered along with market capitalization as measures of size.

At a deeper level of analysis, one must also consider alternative ways of specifying such fundamental variables as earnings or cash flow. Over what period does one measure earnings? If using analyst earnings expectations, which measure provides the best estimate of future real earnings? The consensus of all available estimates made over the past six months, or only the very latest earnings estimates? Are some analysts more accurate or more influential? What if a recent estimate is not available for a given company?⁴

Predictor variables are often closely correlated with each other. Small-cap stocks, for example, tend to have low P/Es; low P/E is correlated with high yield; both low P/E and high yield are correlated with dividend discount model (DDM) estimates of value. Furthermore, they may be correlated with a stock's industry affiliation. A simple low-P/E screen, for example, will tend to select a large number of bank and utility stocks. Such correlations can distort naive attempts to relate returns to potentially relevant predictors. A true picture of the return-predictor relationship emerges only after disentangling the predictors.