Diversification

Corporate Diversification
** Contributed by Belén Villalonga **

Diversification is one of the oldest and most widely researched topics within the field of strategy, as it speaks directly to the core corporate strategy question of what business(es) a firm should be in. More recently, the specific issue of diversification’s effect on firm value has also generated substantial debate in the finance field. As a result, this reader cannot possibly do justice to the literature on diversification. Fortunately, there are several excellent review papers on the subject that provide much broader coverage, and I would encourage Ph.D. students to read at least a couple of them. I have organized this reader around what I see as the three most important questions about corporate diversification: its causes, its consequences for firm performance, and the relatedness among a firm’s businesses. Relatedness is the most important moderator of the relation between diversification and performance, and its measurement––and more broadly, the measurement of diversification––is critical to the observed empirical relation. Ramanujam and Varadarajan’s (1989) review covers some additional issues related to diversification such as the strategy’s implementation, or diversification mode. Montgomery (1994) provides a concise and insightful review of three leading theories of diversification and their predictions about its effect on firm performance: market power, agency, and firm resources. The most important of these theories for the field of strategy is the resource-based view of diversification––the notion that firms diversify to exploit their excess capacity in resources that are imperfectly tradable outside of the firm yet susceptible of being used productively across different businesses within the firm. This view was originally developed by Penrose (1959), which is a “must-read” book for any strategy scholar. Penrose’s ideas were ahead of their time by several decades, and the emergence of the resource-based view as the leading paradigm in the strategy field during the 1980s and until this date attests to this fact. The resource-based rationale for diversification is also at the heart of the concept of synergies, or economies of scope, as articulated by Panzar and Willig (1981). As Teece (1980, 1982) points out, however, for a firm’s resources, or for economies of scope, to justify diversification, it has to be the case that those resources cannot be exchanged through a market mechanism––an extension of Coase and Williamson’s transaction cost economics rationale for horizontal and vertical integration into the realm of corporate diversification.

Hoskisson and Hitt’s (1990) review paper covers some additional theories of diversification, as does Villalonga (2003). Montgomery and Hariharan (1994) and Merino and Rodríguez (1997) provide nice empirical tests of the motives behind firms’ diversification decisions. The latter is less well-known, but is a remarkably strong paper on its empirical methods. So is Silverman (1999), which is also a test of diversification theories but I have included it among the relatedness papers because of the exemplary job it does at measuring relatedness in a technological sense.

The effect of diversification on firm performance has received an enormous amount of attention, probably more than any other empirical relation in the field of strategy. Rumelt (1974) was the first to find that effect is contingent on the degree and type of relatedness across a firm’s businesses. In essence, related diversification is good for firm profits while unrelated or conglomerate diversification is not. This finding has been widely replicated by many subsequent studies using many different samples and measures, as any of the review papers cited above shows. Barney (1996) in particular contains a useful summary table including 30 of these studies.

Given the importance of the role played by relatedness in the diversification-performance relation, its measurement––and more broadly, the measurement of diversification––has proven to be critical to the observed empirical relation. Wrigley (1970) and Rumelt (1974) pioneered the use of categorical measures of diversification based on the relatedness among the firm’s businesses. Jacquemin and Berry (1979) devised a continuous measure of diversification labeled “entropy” that allows a more objective separation between related and unrelated diversification and has been very widely used. A common critique to most measures of relatedness including the entropy one, however, is that they are anchored on the Standard Industrial Classification (SIC) system, which is arguably not ideal for this purpose. Several studies have come up with clever and objective measures of relatedness that do not rely on SIC codes: Lemelin (1982) uses input-output tables to classify relationships between pairs of industries as vertical or complementary in either inputs or outputs. Farjoun (1998) uses data from the Bureau of Labor Statistics about the distribution of job categories across industries to construct measures of skill-based relatedness. Robins and Wiersema (1995) use a technology input-output matrix developed by Scherer (1982) to create measures of relatedness at both the industry and firm levels. Silverman (1999) uses patent data to create an alternative measure of technological relatedness. In an unpublished working paper that eventually got subsumed into Teece, Rumelt, Dosi and Winter (1994), Rumelt devised a measure of relatedness based on the relative frequency with which pairs of activities are combined within the same corporation. In the mid- and late 1990s, the center of gravity of the literature about diversification and performance shifted from the strategy field to the finance field—where it opened a can of worms. The conventional wisdom about the value of corporate diversification in U.S. stock markets had evolved toward the view that diversified firms as a whole traded at an discount relative to the sum of their parts––the so-called “diversification discount” or “conglomerate discount.” Two academic studies came to provide empirical support for this view: Lang and Stulz (1994) and Berger and Ofek (1995). In 1999 two working papers took issue with the interpretation of these findings as evidence that diversification was value-destroying: Campa and Kedia (ultimately published in 2002), and Villalonga (ultimately published in 2004a). The two papers raise essentially the same point, but use different empirical methods to tackle it. The point is that, as the vast strategy literature on diversification motives suggests, firms do not diversify randomly, and once this self-selection is controlled for econometrically, the alleged diversification discount disappears or even turns into a premium. In this reader, I have only included the latter paper (admittedly self-servingly) because, in addition to its substantive findings about diversification and firm value (for which I fully share credit with Campa and Kedia), it made another significant contribution: It was the first paper to introduce the propensity score matching method to the field of finance, where it has now become fairly standard. Several other studies have further contributed to the demise of the conventional wisdom about the existence of a diversification discount. Villalonga (2003) and Martin and Sayrak (2005) provide reviews of this literature. Villalonga (2003) also includes a series of comments on the issue by a number of leading finance scholars. Villalonga (2004b) makes an even simpler but even more fundamental critique of the diversification discount literature. The literature is entirely based on Compustat’s segment database, which presents a number of aggregation issues (very well described by Davis and Duhaime (1992)), and is potentially subject to reporting biases. Using a new Census database that covers the whole U.S. economy at the establishment level, Villalonga (2004b) finds a diversification premium (on a sample that yields a discount according to segment data). One possible explanation for this contrast is that segment data measure purely unrelated diversification, whereas establishment data also measure related diversification. Hence, the findings in Villalonga (2004b) can be interpreted as evidence that there is a discount to conglomerate diversification, but a premium to related diversification––thus providing further support for the findings of Rumelt (1974) and subsequent strategy studies, but using market value as a performance measure instead of accounting profitability.