CHAPTER 4 BUSINESS GROUPS AND FINANCIAL MARKETS AS EMERGENT PHENOMENA - BRUCE KOGUT

THE study of the business group was once lamented for its paucity of research. The research of the past two decades has remedied this deficit quantitatively and qualitatively. However, the search for robust findings has proved often elusive, a challenge for comparative economic sociology in general due to the complexity of the subject and the uneven access to data. This chapter proposes an alternative perspective on business groups, and their relation to finance, to respond to this challenge.

This perspective draws upon the sciences of complexity. The sociology of financial markets can mean the confrontation of the macrostructural patterns in economic markets by the detection of the micro-behaviors These patterns are signatures in the market infrastructure by which foreign exchange trading operating or in the pricing dynamics of securities and derivatives that provide candidate mechanisms which generate these larger patterns. While these signatures are statistical, their origins and dynamics are governed by the social relations among traders, investors, and entrepreneurs. Financial and economic markets are complex systems, and yet considerable progress has been made toward understanding their structural patterns and dynamics, and the micro-behaviors of traders and the rules by which they interact.

Business groups are sociologically important because they are invariably associated with economic and political power through their control over sizeable business enterprises. They pose the interesting question of why this form is so pervasive across countries despite large institutional differences. Their common occurrence across many countries suggests that they are emergent phenomena arising from similar underlying dynamics. While pervasive, they come in many varieties and flavors. Therefore, a simple definition is valuable. Granovetter (2005) offers such a definition in the following: “Business groups are sets of legally separate firms bound together in persistent formal and/or informal ways.”

  

76 BRUCE KO GUT

Granovetter's definition is useful for distinguishing business groups from other kinds of financial markets and organizations, such as venture capital or private equity. Venture capital and private equity share a common structure: a financial firm creates a fund in which private individuals or financial institutions invest as limited partners and this fund then invests in companies in return for equity. Venture capital investments target new and private firms; private equity is investment in established firms, some of which are already public with stocks that may be traded on formal exchanges. In addition, in both cases, funds can co-invest through syndications, though syndicates are more common in venture capital. These investments are intended to be of limited duration, with the goal to “exit” the investment in a few years by selling shares to the public through an initial public offering (IPO) or to private investors. Because they are of limited duration, venture capital and private equity funds do not meet the criterion of “persistent” in Granovetter's definition.

The activities of a business group are often similar to venture capital and private equity funds insofar that they are all types of entrepreneurial investments in new technologies or business models or in “turning around” existing businesses. In most countries, venture capital and private equity markets do not exist or are very new and small. Business groups provide a source of entrepreneurial finance. However, business groups are also very different from venture capital and private equity firms. Through favored access to capital (either due to relationships with external banks or to access to a “house” bank that belongs to the group), business groups can reinvest capital in entrepreneurial opportunities by redirecting capital from one firm to the other. To the contrary, venture capital and private equity firms treat each of their investments as stand-alone entities. The investment horizons of business groups are long-term and there is rarely any planning for a quick exit — often there are no public equity markets in any event.

In addition to persistence, the Granovetter definition implies financial and business coordination through formal or informal ways. The definition appropriately leaves, as an empirical issue, the question of whether this coordination is primarily economic and financial (i.e., equity ties) or social, for example, through the ties of family, ethnicity, and identity. It is also sufficiently general to permit business groups to be horizontally or vertically organized. Horizontal means the group enterprises are specialized to different businesses, such as electronics, telecommunications, chemicals; vertical means that the enterprises are organized in a supply chain, such as in car production which includes suppliers of electronics, drive trains, and engines to a final assembler. This structural dimension is separate from the way in which financial control is legally exercised, such as through a single entity (such as a main bank), which may just hold portfolio shares, or through pyramidal ownership chains.

There are then three dimensions to business groups: social type of the group; activity structure, for example, horizontal and vertical; and financial organization. For example, overseas Chinese businesses in East Asia are frequently family groups, horizontally diversified, and organized through pyramidal ties often lacking transparency. The classic German Konzern is frequently a portfolio company owned either by families, trusts

BUSINESS GROUPS AND FINANCIAL MARKETS 77

(e.g., Zeiss, Bosch), or dominant owners; the portfolio holds controlling shares of equity in diversified businesses. The sociological distance between a Chinese family business group and a German Konzern is large, and yet the term business group applies to both. Even within a country, business groups can differ. For example, the Japanese keiretsu refers to the classic economic grouping of horizontally diversified firms with equity cross-holdings and debt financed by a main bank—but it can also mean a vertically organized tiered supply chain to a final assembler, for example, Toyota, who will hold equity in the first-tier suppliers but will not directly own equity in the second- or third tier suppliers.” This distinction between horizontal and vertical keiretsu was briefly important to a somewhat fragile economic literature in the early 1990s that sought to attribute efficiency to the latter form, and inefficiency to the horizontal type (Lawrence 1993).

The attraction of business groups as a subject of study to economics and sociology, and therefore to economic sociology, is not entirely mysterious. Since the entity is so variable in definition and the empirical heterogeneity is so vast, the subject provides multiple doors of analytical entry into this palatial domain. Thus, it is possible to have a lively discussion in many rooms in this palace, fairly untroubled by global consistency of argument or fact.

This quip is, as all members of this rhetorical genre are, grossly unfair to the many excellent studies. Nevertheless, I will stand by it, for it suggests the importance of a healthy reliance on seeking common invariants across contexts rather than rushing too quickly into the heterogeneity. Both economics and sociology are guilty, especially in regard to examining the function-form question of the efficiency of business groups while neglecting larger structural questions. It may be profitable, as argued below, to analyze business groups as emergent from the social and cultural rules that guide entrepreneurial formation and ownership. Let's begin, however, with the classic argument over the efficiency of the business group form before turning to a structural understanding of the business group as emergent.

FUNCTION-FORM EFFICIENCY: THE LEFF HYPOTHESIS

As the literature on function-form efficiency is well developed in economics, it is instructive to summarize the polar views on this question. One side emphasizes the capture of private benefits by dominant owners. The other side sees the business group as providing an entrepreneurial response to the institutional deficiencies of the business environment. In either case, this argument is primarily about the functional efficiency of different forms.

78 BRUCE KO GUT

Morck, Wolfenzon, and Yeung (2005) summarize the arguments for the first side of this debate by starting with the position that business groups are most often pyramidal devices designed to extract cash from minority shareholders—who nevertheless volunteer to buy the equity or to make the loan. The size of business groups also leads to political power. This literature focuses heavily on the pyramid structure, whereby a root firm owns a controlling percentage in tier one firms, which own in turn a controlling percentage in tier two firms, and so on. This structure permits the root firm to extract cash or services (e.g., corporate jets) from lower tier firms, though ownership control is progressively weaker at each tier. A pyramid in which each focal firm at each tier owns 50 percent of the firms at the lower tier means that control of the root firm falls at a rate of a where a is the percentage of ownership and tis the index for the tier level. Thus, if a root firm diverts $1 million from a tier firm to private benefits for owners (called “tunneling” in the literature), these owners only incur the cost of less than $80,000 (since a 3 o.o.8). This focus on the potential abuse by the pyramid form was observed by Berle and Means for the United States in the 1930s." This literature holds the following then: the business group form is a pyramid; its function is to divert private benefits to the ultimate owners.

Predictably, the counter position is that these groups are efficient, because they solve market and institutional failures, or they possess specialized managerial capabilities. This position can be called the “Leff Hypothesis” in deference to Nathaniel Leff’s (1978) seminal paper that argued that groups consist of entrepreneurial and managerial capabilities to redress the institutional market failures in developing countries that impede entrepreneurship. “The institution of the group is thus,” he concludes, “an intrafirm mechanism for dealing with deficiencies in the markets for primary factors, risk, and intermediate products in the developing countries” (Leff 1978: 667). This hypothesis that business groups compensate for market and institutional failures is wider than Leff stated, and applies generally to the argument that corporate form compensates for failures in markets and institutions.

When empirical questions are complicated, a good strategy is to look for simpler cases. The above contradictory explanations for the existence of business groups overlaps with the empirical debate in the field of strategy concerning the question of whether diversification, even for the unitary firm, adds value. The nomenclature for this question is whether there is a diversification discount. Consider first the statistical studies, largely based on US data, which ask if corporate headquarters have any effect. These studies show that the variance of profitability is determined largely by factors specific to an individual business or to the industry, with evidence indicating that headquarters account only for a modest portion of the variance (see, e.g., Rumelt 1991).

Consider next if the mean effect of this contribution to profit variability is positive or not, that is, is there a diversification discount. Most studies find that diversification on average decreases the value of the firm. However, empirical studies on firms located outside the United States do not find this discount, and often find a premium (e.g., Lin and Servaes 1999).

BUSINESS GROUPS AND FINANCIAL MARKETS 79

Context matters to the efficiency of the function-form relationship.

The relevance of these studies to business groups should be self-evident—does it help to be a member of a large diversified business group? It is not surprising that many studies test if business groups earn a higher return to capital.’ Take, for example, the study of Khanna and Rivkin (2001), which found that for 14 countries, six showed evidence that business groups add value, three that the value was negative, and five evidenced no difference in added value. If we view each country as an experimental subject and the treatment is business groups, then we have an experiment of size 14, with only 42 percent of the subjects showing a positive benefit. On this basis, the null of no relationship is not rejected. There are surely moderating institutional factors, but no study has been able to identify this relationship. While many within-country studies show that belonging to a business group helps relative to independent firms in the same country (e.g., Keister 1997, for China), many studies do not."

 Less concerned by the function-form debate, the sociology literature ventures from the institutional perspective that countries are each led by particular logics and these logics are templates for the organization of business enterprises. This thesis has been usefully applied to the national design of the railroad policy (Dobbin 1994) that differed in France, the UK, and the United States. Hamilton and Biggart (1988) developed a similar argument in pronouncing that capitalism in Korea, Taiwan, and Japan followed different institutional logics. They conclude in noting that “organizational structure is situationally determined, and, therefore, the most appropriate form of analysis is one that taps the historical dimension” (Hamilton and Biggart 1988: S87). In less careful hands than those of Hamilton and Biggart, this otherwise appealing argument has the tendency to slip toward a “vive la difference” claim, whereby countries are unique, following their own logics, and must be understood sui generis. The sociological literature is full of examples along these lines.

This extreme institutional methodology has the important benefit of fostering deep country case knowledge. Still, the comparative analysis has rather modest goals, namely to assign countries to different (unique?) logics without isolating causal statements or providing analysis of micro-mechanisms. A classic example is the Reinhard Bendix (1956) methodological legacy of using two-by-two comparisons (that is, two causes that generate four causal or descriptive outcomes) of four countries; with two causes and four countries chosen to maximize variance, the design is perfectly saturated with no error. This design makes sense in the case of stratified randomized experiments; it is a purely descriptive model when four countries are assigned a priori to exactly four possible outcomes. Arguing from outcome to the selection of cause is a weak methodology to support the conclusion that (two) institutional factors unique to each country's logic determine the existence of business groups. Charles Ragin (1987) proposed a qualitative comparative analysis in order to isolate feasible causal configurations to which countries can be assigned, thus allowing for an informed reduction of country cases to categories that are causally meaningful; these categories are more than just descriptive boxes.

8O BRUCE KO GUT

STRUCTURAL SIGNATURES AND THE EMERGENCE OF BIG DIVERSIFIED FIRMS

It is not surprising, as discussed above, that the business group literature is marked by profound disagreements. It is probably a good idea to step back and try a different avenue of attack. An alternative approach focuses less on the function-form comparison across contexts (i.e., countries) or on broad macro-institutional comparisons. Instead, this alternative focuses on the emergent processes by which large and diversified firms are produced. History matters to this analysis insofar as values and opportunity structures preserve the past. Feedback from the environment provides current information to local actors. As is characteristic of this kind of approach, the goal is to isolate viable mechanisms that relate micro-motives and macrostructures.

There is, though, a more intriguing theoretical edge to this analysis in the following sense. Random behaviors of local actors do not aggregate to random macro-outcomes. There is invariably more structure and order in the macro-organization than suggested simply by observing random micro-behaviors. Random micro-behaviors are a baseline benchmark. Micro-motives such as business group relationships, which are often described as “cronyism,” are a type of “friend-of-friends” dynamic which leads to triangular closure and clustering; when this is coupled with male gender preference, the structural outcome is a “boys' club.” These patterns are structural signatures in data that can usefully distinguish between bad and good candidates for mechanisms that give rise to business groups. These signatures can be thought of as “patterns of relatedness and coordination” that are “microsocial” (Knorr Cetina and Bruegger 2002: 907); patterns of relatedness and coordination are an excellent description of business groups, with the proviso that, to see these patterns, we need to look at the macro-aggregation to detect the regularity of these microsocial structures.

Let's focus on two structural signatures, which are almost law-like in their regularity, in regard to the distribution of firm sizes and diversification. Both of these are useful baseline models for understanding the dynamics that lead to the emergence of big diversified organizational entities labeled business groups.

Why are baseline models useful? If the empirical data across different contexts (e.g., countries) reveal similar patterns, the suggestion is that there is a common process that generates the distribution. For institutional sociologists, this finding is a challenging result, for it says, “don’t tell me the institutional differences and I will still predict the outcome.” However, this challenge can provoke insightful responses, which we develop later.

Business groups are characterized by at least two properties: they are collectively large—often with big individual companies—and they are diversified; it is undesirable to have competing companies within the same group. As one of the great early students of business groups noted, business groups are often the entrepreneurs in emerging markets and they expand in areas that suit their skills and advantages (Leff 1978). Since capital markets are weak, business groups frequently populate capital-intensive industries, such as transportation, heavy industries, and, of course, finance.

BUSINESS GROUPS AND FINANCIAL MARKETS 81

The size distribution of firms

Any student of business organizations comes to appreciate quickly that no matter the initial sizes of new firms, firms tend to quickly become power-law distributed. There are notable exceptions, for example Italy has not generated large firms to populate the extreme value tails that otherwise might be expected. However, the finding of a power law distribution of sizes is nevertheless so common that an appropriate criticism is to accept this property as law-like as we are able to get in social science but then to ask, why is there is such variety in the exponent? This is prosaically asking why a few large firms tend to dominate industrial sectors and national economies—a precondition to the properties of business groups that make them important objects of study. Of course, the size distributions of countries differ in scale. There is a fractal quality to power laws due to their scale-free property. By scale free, it is meant that a scalar (lambda) increase in size proportionally scales the probability, such that

P(Ax) = K(Ax) = \"Kx* = \"P(x).

where Kis a constant, X is the size of the entity, and a governs the skewness of the distribution. Consequently, the ratio of the scaled distribution to the original distribution is simply

P(Ax)/P(x)=\".

This relationship means that the shape of the distribution is not affected by the level of the sizes; large and small countries will differ in scale but not in the shape of the distribution.’ In other words, the global size distribution of all firms in the world can be power-law distributed, as can the sizes of firms within countries, within national sectors, and within business groups (if the population of components is sufficiently large). This repetition of a power-law distribution at every level of the hierarchy (global, national, sectoral, firm) is consistent with a fractal organization and it implies far more order than might otherwise be expected by purely institutional accounts of size distributions.

The relation of a firm's size and the variance of its growth rate

In addition to their size, a second property of horizontal business groups is their diversification. Again, to keep things simple, we start by considering simply the relation of diversification to size distributions. From an economic perspective, the motivations for businesses, to paraphrase our above quote from Granovetter, to persist in long-term relations is that there is a gain to this relationship. This gain may be due to shared resources, markets, or bargaining power via the state. If businesses persist in being bounded to each other but the gains to this persistence vary by firm, what should be the signature in the macrostructure?

 This structural signature is found in a relatively obscure pattern whereby the variance of growth of firm sizes is also power-law distributed (Stanley et al. 1998), such that  o°(o/y) = Ky  with the exponent estimated to be in the vicinity of -o.15. Borrowing from statistical physics, Stanley et al. propose that mild “firm level” correlation among business units can explain this exponent, for which they find some evidence.”

82 BRUCE KO GUT

These results on the power-law distribution of firm sizes and the variance of growth of firms point to persistent binds among interdependent entities through formal ways, namely the unitary diversified firm. These statistical results suggest the existence of heterogeneous capabilities among firms who differ in their abilities to manage their internal diversification. We have noted above the studies in strategic management that indicate that corporate headquarters matter (though modestly) to the variance in profitability; that there is a diversification discount in the United States; and that some countries show a diversification premium.

However, we have not said anything about whether it matters what firms diversify into. Teece et al. (1994) refer to the correspondence between the industry interrelation and the business portfolio of a firm as defining its “coherence.” This definition of coherence is a strong form of technological epiphenomenalism. This leads to the following research strategy. If successful firms benefit from combining businesses with overlapping technologies or markets, then the following pattern should be true at the macrostructural level: pairs of industries that frequently appear in the business portfolios offirms indicate that these industries are technologically or market related. Firms whose diversification portfolios reflect this relatedness are coherent. To test this proposition, Teece et al. analyze the industrial activities of a large sample of diversified firms to see if their portfolios “cohere.” The random baseline is to assume that every industry pair is equally likely to be conditioned on the marginal, that is, conditioned on the overall frequency of firms active in that industry. Coherence then is a type of Student-t statistic indicating the statistical departure from this random expectation. The authors subsequently construct a measure of coherence for each diversified firm. This measure represents the extent to which a firm is diversified into industries that mimic the interindustry correlations.

This study is a nice example of using macrostructural data to investigate micro-behaviors and to apply a random baseline to generate the measure of the nonrandom departures. However, this top-down approach inverts our bottom-up proposal and comes at the cost of a circularity derived from their assumption of a survival bias—that the industry coherence is the aggregation of firm-level choices. A naïve test is simply to ask if the same patterns of industry coherence are observed in other countries. Kogut, Walker, and Anand (2002) provided this test, rejecting a common coherence across countries. Coherence, to the extent that it exists, is not determined by technologies, but most likely by market and institutional factors as well. If cross-national coherence varies for the diversified corporation under a unitary ownership structure, it is surely unlikely to hold for the business group either.

COUNTING BUSINESS GROUPS

Having confronted two statistical regularities with empirical data regarding the unitary firm, let's turn now to the more prosaic task of identifying business groups. Harry Strachan observed in his study of business groups in Nicaragua that,

BUSINESS GROUPS AND FINANCIAL MARKETS 83

in response to the question “to what business group do you belong?” a businessman might say none, or name a group, but the question would not evoke the response of asking what a group is (Strachan 1976: 26–9; cited in Granovetter 1995: 99). In fact, many directories of business groups have been drawn up based on insider knowledge of the group structures.

Another approach is focus on the ownership data published by investment houses, business press, or stock exchanges in reference to companies whose stock or debt is traded. Of course, pure private entities are not captured by these listings, thus Granovetter's firms bounded by “informal ways” are excluded. Still, the advantage of this approach is to mine the data for groupings that do not rely upon judgment so much as statistical methods.

Glattfelder and Battiston (2oo9) analyzed data on 48 countries using Bureau van Dijk's ORBIS database. In all, the data consisted of 24,877 companies and 1o6,141 shareholders. The structure of an ownership-firm network is bipartite (sometimes called an affiliation network). Since the data include the percentage of the equity owned by the shareholder, they were able to construct a control index that identifies the ownership backbone of the global network. This backbone consists of the largest shareholders by their direct and indirect equity holdings.

While the backbone analysis is helpful for identifying big shareholders, it does not directly count the number of business groups. Counting business groups is complicated by loops, since firms may have cross-holdings directly or indirectly via third parties. Since a pyramid is a root firm owning other firms, which may own other firms, one solution for counting pyramids is to identify the directed acyclic graphs (DAGs) in the networks. The procedure begins by removing the largest DAG, or pyramid, then turning to the next largest, until there are no DAGs left that meet our threshold criteria. If two root firms have cross-holdings or common investments, the largest DAG is first counted, then collapsed to a single node; then the second root is checked to see if it is a pyramid and, if so, is counted. This procedure was applied by Kogut et al. (2012). For the period around the year 2000, Chile was clearly the pyramidal champion, followed by Korea and Brazil. Denmark had many pyramids but they were relatively small. Nevertheless, these countries have more pyramids (uncorrected for network size) than other countries. As most studies show, pyramids are largely absent in the United States.

LOCAL TOPOLOGIES AND COMPARATIVE ANALYSIs”

Counting the average branches to trees is a good first attempt at comparing the importance of business groups by country. A more ambitious program of research is to parse the institutional logics into micro-motives into clear rules. After all, a logic should be expressible by rules, or else why call it a logic?

84 BRUCE KOGUT

An emergent approach welds both together: it analyzes the correspondence of a country macrostructure and the micro-rules that could generate the structural signatures. An ideal answer to the question of what it means to be a country entails in my view an understanding of the micro-rules that guide the individual choice whether to own or to sell a firm or whether to be a member of a business group. If the way business groups are formally bound is through boards of directors, then the corresponding micro-question is what rules, or what logics, promote the appointment of directors. Multiple studies on directors show that the choice of a director is strongly conditioned on the social network, such as by homophily; in other words, “birds of a feather flock together”; men choose men, graduates of an elite school choose other graduates of the same university, and so on.

Homophily is an excellent micro-rule candidate for explaining the sociological formation of business groups. Business groups are often associated with family, ethnic, and status affiliations. In any cross-section, the ties between the firms will reveal a tendency of triangular closure due to these affiliations. Triangle closure captures the social phenomena that “friends of friends are also friends.” By applying a statistical method called exponential random graph models (or ERGMs), we could estimate the coefficient to triangular closure in two countries called North and South. This estimation might, for sake of illustration, show that the coefficients to triangular closure differ by a factor of 2; everything else is the same. Then, by tuning this coefficient, it would be possible to grow the North from the South.

Below, we develop further this idea of growing business groups. First, though, it might be useful simply to know if business groups are more common in one country than another. Earlier, we simply offered a count of the pyramids in the country normalized by the size of the network. Business groups, however, are broader than just pyramids. Moreover, a simple count does not reveal much about the micro-foundations.

There is, however, a simpler way to post this question: Is the number of business groups in a country greater than would be randomly expected? If business groups can be detected by triangular closure, then is the number of triangles greater than would be expected if the ownership or social ties among firms were simply randomized? Maybe a given country has 20 business groups. Is this a lot for a country with N number of firms and Kownership ties between these firms? It is quite possible that a random graph consisting of N nodes (firms) and K edges (ties) will on average show the same triangular frequency.

A random graph is the appropriate representation for the pure market economy and a baseline model: social ties should not matter. We will focus on one network statistic, clustering, as it captures well local differences, where local means the formation of neighborhoods and the rules by which neighbors are chosen. (Neighborhood has a technical graph definition: the nodes that are linked directly to any arbitrarily chosen node j belong to the neighborhood of j.) Clustering is an operationalization of Coleman's definition of social capital, whereby a more connected neighborhood has the social properties of high trust as well as monitoring. Thus, a business group should be marked by clustering, indicating a high level of social capital and trust.

BUSINESS GROUPS AND FINANCIAL MARKETS 85

Having chosen now the ideal type of a market economy as a random graph and defined clustering as the appropriate network statistic, we are prepared to analyze how a country differs from the random polar case. The measure we propose relies upon this proposition: the nodal-expected clustering coefficient for a random graph is the density of the graph.

We exploit this property to measure how much a network differs from a typical random network as the difference between the empirical average clustering coefficient and the density for any given country. We call this difference “distance to random.” By applying this algorithm to data from over 20 countries, the findings showed that the ownership clustering found in Anglo-Saxon countries are relatively close to being random (Kogut et al. 2012). This approach can also be applied to board interlocks. Here, the observed clustering in countries is considerably different than their random graphs prediction. Boards indeed look clubby.

A basic requirement for the comparative study of business groups should be evidence that their structural form is expressive of a social and cultural rule. The earlier section proposed counting business groups by looking for tree-like relationships in the ownership data; this section proposes that for many kinds of business groups, we should expect to see the social rules develop into relational patterns that reveal clustering. Some business groups, such as the vertical keiretsu in Japan, are better measured as (acyclic) trees; others, such as Chinese family enterprises, are best measured through clustering. This difference makes the larger point: it is by understanding the generative rules behind business groups that we will start to understand their sociological properties and origins.

BUSINESS GROUPS AND GENE ALOGICAL RULES OF CULTURE

This above illustration points to the utility of understanding the growth and prevalence of business groups across countries from the lens of an agent-based model (ABM). An ABM treats agents as pursuing particular strategies, which may be social or economic. A social rule might be to befriend friends of friends (to continue with the above example), or it could be economic in forming ties with economically powerful firms or people. Of course, social and economic motivations are often mixed together.  

The goal of an ABM is to explain the macrostructural patterns as generated by the behavior of rule-driven agents. Epstein (2007) has pithily summarized this goal in the following adage: if you didn't grow it, you didn't explain it. The astute observation is that the converse (if you grow it, you have explained it) is clearly not true since there may be, and often are, multiple models to grow a given topological structure. Still, the rule is useful as a deduction by elimination in order to disprove a causal claim. An ABM analysis can be useful to understand the sociological origins of business groups. By ABM type analysis, I mean that simply framing business groups as the product of a bottom-up generation from social rules can be useful.

86 BRUCE KO GUT

Here is the proposed framing. As noted above, the literature on business groups is contradictory. One camp to this debate claims that pyramidal groups violate good governance and permit powerful people, often families, to reap benefits to the economic detriment of the underlying businesses. As an example, here is a quote from a review article by Morck, Wolfenzon, and Yeung (2005:55), published in the Journal of Economic Literature:

In many countries, large pyramidal groups effectively entrust the corporate governance of substantial parts of their corporate sectors to a few extremely wealthy families. This can potentially magnify the poor governance of a few family patriarchs into inefficient economy-wide capital allocation, reduced investment in innovation, and retarded economic growth. Moreover, to preserve the status quo, these elite families sometimes appear able to influence public policies so as to curtail private property rights development, capital market development and economic openness. We dub this situation economic entrenchment. We argue that much existing work points to economic entrenchment as a significant issue in many countries.

The second camp in this debate argues for the capacity of groups to compensate for institutional weaknesses and to outperform alternative organizational forms. Since good data are notoriously difficult to acquire, neither camp has declared victory.

A generative analytical approach does not resolve the above debate, but shifts attention to explaining the interaction of the rules by which such groups, especially family owned, evolve. In the relation of social rules to structure — such as business groups — the study by Bertrand et al. (2008) presents one of the rare analyses of the generative rules of social structure (intergenerational family firms) and their economic outcomes. In their statistical description of the population parameters to business groups, they note that sons inherit the ownership of the firms inside the family group. Based upon their statistical description, they proceed to show that the performance of the family groups worsens, the higher the number of sons and daughters involved when the founder dies. Thus, they have delineated a social rule regarding a male patrilineal pattern and, from this, determined the economic consequences.

A full agent-based model would consider not only the demographics, but the social rules by which property is inherited and by which firms are bound together; agents would be assigned utility functions which would guide their choice of linkages across firms. For our purposes, it is instructive to consider just the implications of fecundity on the predicted size of business groups.

In other words, we would like to derive the size distribution of business groups for Thai family businesses using statistics we can estimate from the data." These statistics are simply the average number of children () and the probability that a child will create a subsidiary (); for simplicity, we do not distinguish between sons and daughters, though, in fact, this probability is very much a function of the number of sons. A lineage is a DAG, which is a tree that has a root parent from which branches a generation, which then branches into a second generation, and so on. We begin by generating the tree starting from the root (founder).

BUSINESS GROUPS AND FINANCIAL MARKETS 87

To create the second generation, we choose the number of children randomly by using a Poisson distribution with a specified A. We assume that the decision to start a new business is only made if the father owns a business; otherwise the entrepreneurial line dies. We repeat the process for each child to generate the third generation. This process goes on until a specified number of generations is reached. We can then count the size of the business group by the number of affiliated businesses.

Figure 4.1 shows the prediction for the average size of Thai groups by running the simulation several thousands of times; for each instance, we calculate the size of the business tree and calculate the average given the size of the family. The \value (6.8) used is the one found in empirical data; these are big families. We compared the simulation results with the numerical solution and they show that they are in close agreement. The interesting observation is that for low probabilities of alpha (probability of a child starting a business), there is little difference between the predicted size of the groups. Even in this deterministic model, the key factor driving the size of the pyramid is the entrepreneurial probability. We have assumed that this alpha is fixed across generations.

The above exercise illustrates the idea of genealogy in the very strict sense: the lineage dynamics of the family generates the structure of the business group. Stark and Vedres (2010) propose a related idea in the context of the “intercohesion” within business groups in Hungary whose boundaries change through a process of splintering and folding. Business groups are defined through the sharing of directors on their boards (this is called “folding”); the more directors shared, the more cohesion. Stark and Vedres (2010) analyze the lineage of these groups by treating the member firms as genes in a sequence. When these sequences are unstable the dynamics are nevertheless coherent and preserve a genetic heritage through recombination.

FIGURE 4.1 Effects of Entrepreneurial Ability of a Family and the Growth of the Business Group (from Kogut 2012).

88 BRUCE KO GUT

Sociology has an important contribution to make here. The Stark and Vedres analysis returns to the primary interest of Leff: business groups are a form of entrepreneurship. In other words, this poorly explained phenomenon of entrepreneurship probably owes a lot to familial cultures. In a wider welfare analysis, this dynamic of the innovation function can compensate well for the inefficiency of the form. It is not obvious that if patrilineal practices were to be removed in the Thai case, the social incentives for innovation would be the same. In a universe of second bests, the optimal policy intervention is often not the one that appears to move the current state “toward” the best of a market, with no family firms but just atomistic competitors. Economics has a hard time separating this ideal type from an ideological predilection for competitive market solutions, even when such solutions are rarely available. A sociology of finance would have much to contribute to the understanding of development by analyzing the rules of lineage (be they familial or corporate) in financing entrepreneurship and innovation.

THE STATE AND BUSINESS GROUPS: LINEAR VERSUS EXPONENTIAL TIME

The sociological literature on business groups has commented extensively on the role of the state in promoting business groups (see the excellent discussion in Granovetter 2005.) When the state is viewed as a static unitary actor, an agent-based model for understanding the emergence of the business group is problematic. However, in conditions when neither the state nor business is unitary, an agent approach can provide useful insights.

Consider two related applications. The first is the observation from Gershenkron's famous thesis of late industrialization that the state can speed up development, especially through an alliance with capital. Amsden (1989) adopts this idea when analyzing the role of the state in post-World War II Korea. Because of the Japanese occupation, Korea had an experienced factory workforce but a repressed indigenous entrepreneurial class. Buoyed in part by American foreign aid, the regime of Syngman Rhee created a state of patronage while relying on the business groups, or chaebols, that dated to, and sometimes before, Japanese occupation. The military coup of 1960 did not so much change the role of the state as switch allegiance to a new class of entrepreneurs who founded business groups under the tutelage of the state and its state-owned financial institutions. For Amsden, this political change transformed the modality of the economy from rent-seeking to investing (Amsden 1989: 20). Under this new regime, the state facilitated the financing of, and manipulated prices to steer, investment; the chaebols proved to have a strong capacity for learning and absorbing foreign technology given their already well-educated and hardworking workforce.

Granovetter (2005) notes that the social distance between the military state and business eroded, particularly through intermarriages. Like the concept of network “folding” discussed by Stark and Vedres, Granovetter stresses the “network overlaps” between business, society, and the state.

BUSINESS GROUPS AND FINANCIAL MARKETS 89

This similarity points to the potential of seeing these overlaps as the outcome of a genealogical sequence, as Stark and Vedres applied to Hungarian governance ties. But for many countries, the principal form of governance is not the board of director but the family. Indeed, the chaebols have been family-owned businesses and an important element in the expansion of the businesses has been the number of sons; not uncommonly, the details of familial rivalries are sometimes subjects in the press. Thus, the genealogical dynamics driven by fecundity and by social class are pertinent to an analysis of the sustainability of the Korean chaebol as a familyowned business group.

The second observation couples the earlier discussion of the tendency offirm sizes to be distributed by a power law and the time it takes a state to consolidate. The usual narrative for the Korean case is that the state fostered the business sector. Without questioning this particular history, consider instead the relations of the state and business in Russia after the fall of communism. The decision to distribute ownership through mass privatization led to a strikingly rapid rise of powerful financial industrial groups. By the late 1990s, only six years after the first wave of privatization, the size distribution of Russian banks showed the usual power-law distribution of firm sizes (Kogut 2012)— large and powerful financial actors had quickly emerged. Meanwhile, the central government faced substantial problems in the collection of tax revenues and curbing the rampant corruption not only in the provinces, but in Moscow.

One way to understand the Russian case is to note that the power-law distribution suggests that the banks were growing in exponential time, but the state revenues were only growing linearly; in the 1990s, this growth was negative (Treisman 1999). The Russian government faced an essential divergence, by which their inability to consolidate their powers to tax was increasingly dominated by the growing fortunes of oligarchs who had seized ownership of vast swaths of Russia's natural resources. The first decade of the twenty-first century in Russia has seen a struggle between a weak but dangerous central state and powerful oligarchs, with signs of a new political alignment.

A simple point is suggested by the comparison of the Korean and Russian cases: it matters to the development of business groups and the state which one consolidates first. The development plan of the business group is not always the product of the state; sometimes the state is the hostage. An insightful way to understand how time matters to this struggle is to meter the generation of large firms against the consolidation of the state and its ability to tax and to spend.

BUSINESS GROUPS AND THE MORAL ECONOMY

One of the most intriguing dimensions by which Granovetter proposed to study business groups is the “extent of the moral economy.” “Groups may but need not,” he observed, “be coherent social systems in which participants have a strong sense of moral obligation to other members and a well-defined conception of what is proper behavior.

90 BRUCE KO GUT

Such conceptions are almost invariably accompanied by a strong sense of group identity, which confers a normative and extraeconomic meaning on economic action” (Granovetter 2005:433).

The notion of a moral economy dovetails with organizational theories that treat firms not simply as a “nexus of contracts” but as social communities in which members hold identities and common values. Surely, there are many business groups fabled for their public and philanthropic interests, such as the Tata business group in India. The empirical questions are twofold: Do business groups on average evoke more of a sense of moral economy than non-group firms, and does this economy indeed extend across the independent firms? The first question returns to the ambiguous results on the functional superiority of the business group form; the second is related to the notion of “relatedness” or “shared resources” that has occupied the attention of the economic and business literatures.

A moral economy may have another kind of utility, though, namely in attracting and selecting a high-quality workforce. In many countries, the firms that constitute the business groups are among the most reputable enterprises in a country. The precocious study by Sakakibara and Westney (1985), comparing the rankings by college graduates of prospective employers, indicated almost exclusively a preference for firms belonging to business groups. Similar results no doubt could be found for many other countries, with the caveat that globalization (much like colonization before it) provides competition for the best graduates.

IMPLICATIONS FOR THE SOCIOLOGY OF FINANCIAL MARKETS

Business groups are large, consisting of diversified firms that are persistently bound. Seen as isolated phenomena in very distinct institutional environments, their dominance is often inexplicable. The trick suggested by this chapter is to step away from the cross-section and to study the longitudinal bottom-up emergence of business groups, thus coupling statistical models with historical context.

This approach is similar to the argument used in the now-canonical study for the sociology of financial markets by MacKenzie and Millo, “Constructing a Market, Performing Theory: The Historical Sociology of a Financial Derivatives Exchange.” They note first that the Black-Scholes-Merton option pricing model, which assumes constant volatility as a model parameter, predicts that the implied volatility should be invariant to price levels of the stock or more exactly to stock price relative to the call price (for a European call option). However, subsequent to the Black Monday crash of 1987, empirical studies found that this relation was U-shaped, poetically named “a smile.” This smile has remained a feature in the empirical data.

BUSINESS GROUPS AND FINANCIAL MARKETS 91

MacKenzie and Millo proposed a narrative that iterated between particular macro-patterns in trading (e.g., the signature of  “smiles”) and the micro-behaviors of traders (e.g., they followed the same script, they responded similarly to sudden downward jumps in prices). This methodological approach was applied in this chapter to propose that business groups are a type of signature. The interplay between law-like properties of firm-size distributions and the micro-motives of economic agents lead to the emergence of business groups. An approach that uses baseline models that describe the macro-patterns, such as the presence or absence of power-law distributions, is useful to bounding the number of candidates for micro-behavioral explanations. In the end, the analysis of business groups as emergent may reveal new, if not greater, commonalities across countries than are currently available through comparative institutional analysis.