Download A Theory of Family Business Groups and

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Corporate venture capital wikipedia , lookup

Private equity in the 1980s wikipedia , lookup

Early history of private equity wikipedia , lookup

Transcript
A Theory of Family Business Groups
and Pyramidal Ownership*
Heitor Almeida
New York University
[email protected]
Daniel Wolfenzon
New York University
[email protected]
(This Draft: August 18, 2003 )
Abstract
Existing theories of pyramidal business groups cannot explain recent empirical findings about the
prevalence of these groups and their characteristics. In this paper, we propose a simple theory of
pyramids that helps us understand the empirical evidence. Our theory explicitly recognizes that a
full understanding of pyramidal business groups necessitates answering two different questions in an
integrated framework: why business groups arise, and what determines their ownership structure,
that is, why firms are owned through pyramidal chains. In our model business groups arise to
substitute for missing financial markets, and families choose pyramidal ownership structures when
the benefit of using retained earnings that belong to existing firms in the group is large. Our model
makes a number of empirical predictions that are consistent with available evidence and suggests
other implications that can be tested in future research.
Key words: pyramids, business groups, investor protection, ownership structure.
JEL classification:...
* We thank... for valuable comments.
1
Introduction
Many firms in the world have a controlling shareholder, usually a family or the State (La Porta,
Lopez-de-Silanes and Shleifer, 1999). In fact, in several countries a few wealthy families control a
large number of firms and consequently a vast amount of capital through the creation of business
groups.1 This concentration of control is usually accomplished with the use of pyramidal ownership
structures. A typical pyramidal business group consists of a collection of legally independent firms,
controlled by the family through a chain of ownership stakes. Because the family does not need
to retain 100% of the shares to retain control at each chain of the pyramid, pyramidal ownership
structures allow for separation of cash flow and control rights, and can facilitate family control.
There is a large economics and finance literature that attempts to explain the existence of
pyramidal business groups, and the role they play in the economy. While some authors argue that
pyramidal business groups arise to substitute for missing or poorly functioning markets (Leff, 1978,
Khanna and Palepu, 1997),2 others believe that pyramids are used by families to retain control over
a country’s corporate sector with a low cash flow stake.3 In particular, because of the potential
separation of cash flow from control rights engendered by pyramids, their existence can exacerbate
agency problems between controlling and minority shareholders.
The fact that pyramidal business groups are more common in countries with poor investor
protection (La Porta, Lopez-de-Silanes and Shleifer, 1999), appears to be consistent with a role
of business groups as substitutes for poorly developed financial markets. However, there is also
recent evidence that a firm’s group membership increases the scope for expropriation of the firm’s
minority shareholders by the controlling one (Bertrand, Mehta and Mullanaithan, 2002, Bae, Jang
and Kim, 2002). This evidence appears to be more consistent with the view that pyramids are a
mechanism to retain control and extract private benefits. Proponents of this view argue that the
reason why pyramids are more common in countries with poorly developed financial markets is
that in those countries private benefits of control are large (Bebchuk, 1999). Thus, while families
might benefit from the concentrated control achieved through pyramids, the resulting allocation of
1
See, among others, Claessens, Djankov, and Lang (2002) for the evidence on East Asia, Faccio and Lang (2002)
for Western Europe, Khanna (2000) for emerging markets, and Morck, Strangeland and Yeung (2000) for Canada.
2
Khanna and Palepu’s value-enhancing argument is really about all types of business groups, not only about
pyramidal business groups. We come back to this important issue below.
3
This argument goes back, at least to Berle and Means (1932). In their classic book they describe the use of
pyramids in the railroad industry in early 20th century America.
1
corporate control may become sub-optimal for the country (Morck, Strangeland and Yeung, 2000).
At first glance, the agency view of pyramids appears to be broadly consistent with the data, and
one can argue that it has become the traditional view of pyramids in the academic literature.4
Nevertheless, a more detailed examination of the available data on pyramidal ownership structures reveals some facts that cannot be adequately explained by this traditional view. Notice that
the traditional story revolves around the notion of control of voting rights. If pyramids are simply
a way of minimizing the monetary cost of achieving control, one should never observe cash flow
stakes in individual firms that are larger than the treshold that is required to achieve control. A
major problem for the conventional story is the substantial empirical evidence that in many countries families do hold large cash flow stakes in the firms that they control through pyramids. For
example, Franks and Mayer (2002) find that in Germany, for a large majority of the firms controlled through pyramids the controlling shareholder could have achieved the same level of control
by simply holding shares directly in the firm. The authors conclude from the evidence they compile
that, in Germany, pyramids are simply not used as a device to achieve control. Similar findings
have been reported for other countries such as Brazil (Valadares and Leal, 2001), Chile (Lefort and
Walker, 1999), and Canada (Attig, Fischer and Gadhoum, 2003).
Another problem with the traditional story is that pyramids are not always the only available
method to separate cash flow rights from voting rights. For example, if there are no restrictions
to the use of dual-class shares, any degree of separation is possible by the use of dual-class shares.
In such a case, why would a family choose to control a firm through a pyramid instead of issuing
dual-class shares? Nevertheless, the evidence compiled by La Porta, Lopez-de-Silanes and Shleifer
(1999) suggests that pyramids are much more common throughout the world than the direct use
of shares with superior voting rights. Of course, in many countries there are restrictions to the use
of dual class shares. If these restrictions exist, there will be an upper bound to the deviation that
can be achieved by a combination of direct holdings and dual-class shares, and the conventional
story would predict that pyramids would appear when the desired deviation from one-share-onevote is above the maximum allowed by the law. Empirically, however, we do observe the use of
pyramids to separate cash flows and votes below the upper bound achieved by a combination of
4
However, this traditional story does not explain why shareholders agree to buy into business groups that they
know will expropriate them (see, e.g., Bertrand, Mehta and Mullanaithan, 2002).
2
direct holding and dual-class shares (see Bianchi, Bianco, and Enriques, 2001, and Volpin, 2002, for
Italian evidence). All this evidence suggests that considerations other than control rights motivate
the creation of pyramidal groups.
Our goal in this paper is to propose a simple theory of business groups and pyramidal ownership
that is consistent with the empirical evidence. Our theory blends elements of the value-enhancing
view with the agency view to build a positive theory that explains when pyramids should arise,
and the properties that pyramidal ownership structures should have.
We use Khanna and Palepu (1997)’s argument that managerial talent is a scarce commodity,
and thus ownership of several firms by the same family may become a viable allocation of control
in some countries. However, it is unlikely that families are always the best possible managers for
all the firms they end up controlling. In order to explain how families can end up controlling a
large number of firms in countries with poor investor protection, we embed Khanna and Palepu
(1997)’s argument in a framework that builds upon Shleifer and Wolfenzon (2002)’s model of the
relationship between investor protection, equity markets and the allocation of capital. In that
model poor investor protection limits the amount of capital that talented entrepreneurs can raise
in financial markets. This limitation allows wealthy families to own firms even when they are not
the best possible managers for these firms, because they own another scarce resource - their wealth.
These considerations imply that families should own more firms in countries with poor investor
protection, and thereby explains why business groups are more likely to arise in those countries.
This part of our model can be thought of as a simple formalization of the value-enhancing view
of business groups - they arise to substitute for missing markets, in our particular case financial
markets that are endogenously curtailed by poor investor protection.
However, this argument on its own cannot explain why pyramids arise. Even when families
have scarce resources that make them the constrained-efficient owners of a firm, this does not mean
that the ownership structure has to be pyramidal. In fact, the family has a choice of owning the
firm through a horizontal structure where the firm will be controlled directly, or placing the firm
in a pyramid where it will be controlled by the family but legally owned by another firm that is
also controlled by the family. This simple observation underscores an important point that has
been neglected in previous literature. The questions of why business groups arise and their optimal
3
ownership structure are two different questions.5 Clearly, a model that explains pyramidal business
groups must tackle both questions in an integrated framework.
We argue that there are three main differences between the pyramidal and the horizontal structure. In the pyramidal structure, the family shares the security benefits (that is, the fraction of the
benefits that is not diverted by the family) of new projects with existing shareholders. Furthermore,
because pyramidal firms are owned through a chain of control diversion of cash flows tend to be
higher in the pyramid. However, the pyramidal structure also generates a potential benefit for the
controlling family. Suppose the family already owns a firm A. If firm A has been successful in the
past, it will have accumulated a stock of retained earnings that can be used to make investments
in new firms. Crucially, the family will not be able to use the whole cash that belongs to firm A
(and thus to the existing shareholders) to set up a new firm using a horizontal structure, because
the new firm in this case would belong only to the family (and to new shareholders who make fresh
investments in the new firm). However, the new project can be undertaken using the whole cash
when the family forms a pyramid, because in this case the new project belongs to all the existing
shareholders.
We allow the family to choose freely whether it will undertake new projects in a pyramid (with
access to the entire stock of cash but sharing the NPV with the existing shareholders), or using a
horizontal structure. Because the family must share the security benefits of the new project when
it uses a pyramid, it will have greater incentives to divert the cash flows of the new project from
security holders to its own pockets. This effect tends to create more diversion in pyramids than
in horizontal structures. Because the family internalizes the costs of diversion every time it must
issue new shares, it has an incentive to keep its ownership of cash flow rights in the new project
as high as possible and minimize the issuance of new shares. This intuition might explain why we
observe high cash flow stakes in firms that are owned through pyramids.
Still, the potential for high levels of diversion in the pyramid remains. We show that pyramids
are chosen over horizontal structures only if the family expects to divert enough cash flows of new
projects away from existing shareholders. Effectively, the new investments that are undertaken
through pyramids always become negative NPV investments for existing shareholders, in an ex5
Again, the exception is Khanna and Palepu (1997), who say explicitly that their theory is about all types of
business groups, irrespective of their ownership structures.
4
post sense. In other words, existing shareholders would prefer that the cash stock was paid out
as a dividend, but the family may choose to use the cash to invest in new projects. Importantly,
this does not mean that the pyramidal structure is inefficient, because using the whole cash (and
investing through pyramids) might be the only way to undertake positive NPV investments when
the project cannot be financed using only part of the cash.6 Thus, the excessive diversion that
occurs in pyramids may be a “necessary evil” when the availability of cash is important enough,
and when the family has some talent to manage the project. We believe these observations help
conciliate the two seemingly disparate views of pyramids that have been proposed in the literature.
In addition, our model generates a number of more specific empirical implications. Some of them
are consistent with available data and anecdotal evidence, while others have not been systematically
tested. We discuss our implications in detail in section 6. For example, our theory suggests that
pyramids are more likely to arise following successful performance of the existing firms in the group
(because of the crucial role of the stock of cash). Firms that are owned through pyramids will also
tend to be more capital intensive, and less profitable (because these firms are harder to finance
as stand-alones or in horizontal structures). Finally, the model predicts that the prevalence of
pyramids will be higher when investor protection is lower. There are two reasons for this. First,
the family is more likely to own several firms in countries with poor investor protection, because
high investor protection will decrease the comparative advantage that families have in undertaking
new projects. Second, families are more likely to choose pyramids over horizontal structures when
investor protection is lower, because pyramids are only chosen when there is enough diversion and
the costs of diversion are lower in countries with poor investor protection. In fact, we show that
the same conditions that are conducive to the appearance of business groups are also conducive
to the appearance of pyramidal structures. This result might explain why many business groups
appear to be organized as pyramids.
In the next section (section 2), we discuss the related literature on pyramids and business
groups. We introduce our model in section 3. In this section we consider a family who already
owns and controls an established firm and can set up a new firm. We assume that the second firm
can only be set up by the family, effectively forming a business group. In section 4 we relax this last
6
Notice also that the negative NPV of ex-post investments for existing shareholders does not mean that they
cannot break even ex-ante.
5
assumption and explore the issue of when a business group is formed. Section 5 analyzes the origin
of the business group and discusses the optimality of committing to a particular organizational
structure. Section 6 discusses the empirical implications of the model, and section 7 concludes the
paper.
2
Related Literature
The term ‘business group’ is used in the literature to describe many different types of organizations.
It is used for a collection of legally independent firms controlled by a single family, such as the
family groups in Western Europe, Latin America, and East Asia. However, the term has also
been used for organizations with a lower degree of central control achieved through interlocking
directorates, common ethnicity, school ties, etc. Example of these groups are the Japanese keiretsu
in which individual managers have a great degree of autonomy in their firms but coordinate their
activities through the President Council and a common Main Bank (Hoshi and Kashyap, 2003).
Another example of these loose associations are the horizontal financial-industrial groups in Russia,
‘which are more properly industry alliances’ (Perotti and Gelfer, 2001, p. 1604). In this paper we
concentrate on family business groups.7
There is a vast economics’ literature on family business groups (see Khanna, 1999, for a survey).
The literature has suggested a number of arguments about why business groups exist, and why they
use different ownership structures. To achieve control of the various member firms, a family can
hold shares directly in each one (‘horizontal structure’) or, alternatively, it can control some of
the firms through a chain of ownership relations (‘pyramidal ownership’). We consider first the
literature on the first issue (why business groups exist), and then we discuss the arguments about
ownership structure in business groups. Our goal is not only to review the existing theoretical and
empirical literature, but also to point out to some important questions that we believe are not
well explained by existing theories of pyramidal business groups. We conclude this section with a
brief discussion of the relationship of our approach to other related strands of theory in corporate
finance, especially the literature on internal capital markets.
7
Khanna (1999) also argues that the degree of central control in business groups varies across countries.
6
2.1
Why business groups?
Most theories explain the presence of business groups as an efficient organizational form that adds
value to member firms. Leff (1978) argued that business groups arise to substitute for missing
markets (e.g. labor and financial markets).8 The modern reincarnation of this argument is in
Khanna and Palepu (1997, 1999), who argue that business groups provide an important substitute
for active external capital markets in developing countries. Some of the empirical evidence in
La Porta, Lopez-de-Silanes and Shleifer (1999) can be interpreted as being consistent with this
argument. Indeed, La Porta, Lopez-de-Silanes and Shleifer show that pyramids (one type of business
groups) are more common in countries with low investor protection, which are also countries with
less developed financial markets (La Porta et al., 1997).9
An alternative value-enhancing argument suggests that business groups add value to member
firms because they are better positioned to lobby the governments for favors (Pagano and Volpin,
2001). These political economy theories suggest that business groups should be more prevalent
in countries where the government has a stronger role in allocating resources, or in more corrupt
countries.
There is mixed empirical evidence on the direct valuation effects of group membership. Khanna
and Rivking (1999) study 15 countries,and find that in only 3 of them affiliation with a business
group adds value to member firms. They find however, that, affiliation does not decrease value
in any of the cases. Khanna and Palepu (2000) analyze the performance of business groups in
India and find that only the members of the largest groups have valuations higher than those
of independent firms. Member firms of medium-sized Indian groups have valuations below their
independent counterparts. Claessens et al. (2002) find evidence that firms in business groups
organized as pyramids (especially those in which the divergence between votes and cash flows is
the greatest) have lower Tobin’s Q. Similar results are found by Lemmons and Lins (2003), Lins
(2003), Mitton (2002) and Joh (2001). Finally, for a sample of East Asian countries, Claessens,
Djankov, and Lang (2002) find that group affiliation has on average no effect on firm valuation.
8
For example, business groups might act as a risk sharing mechanism (Aoki, 1984). Khanna and Yafeh (2001)
test the risk-sharing hypothesis and find no evidence for it.
9
However, this is not necessarily evidence that business groups are more common in countries with low investor
protection, because La Porta, Lopez-de-Silanes and Shleifer (1999) measure the proportion of pyramids as a fraction
of the total number of firms in their sample, some of which could belong to non-pyramidal business groups.
7
They find, however, that there are benefits and costs for different types of member firms: The
oldest and slowest-growing firms in the group benefit at the expense of younger, fastest-growing
firms.
In contrast to the value-enhancing view, a recent string of papers argues that business groups
exacerbate agency conflicts and are thus detrimental to firm value. The main argument is that when
firms belong to a group, the scope for expropriation of minority shareholders by the controlling
shareholder might increase. Bertrand, Mehta, and Mullanaithan (2002) find evidence consistent
with expropriation in Indian business groups, from firms in which the controlling shareholder holds a
small equity stake to firms in which the controlling shareholder holds a large stake. Bae, Kang, and
Kim (2002) find that Korean chaebols use M&A transactions between member firms to expropriate
shareholders of the bidder firm and benefit the controlling family. These findings raise important
questions. Why are business groups so common if they exacerbate agency conflicts? And if business
groups do increase agency costs, why is it that they arise precisely in countries with low investor
protection, where these costs are high to start with?
2.2
Why pyramidal ownership?
Pyramidal ownership structures are very prevalent around the world, especially in countries with
low investor protection (La Porta, Lopez-de-Silanes and Shleifer, 1999, Claessens, Djankov, and
Lang, 2000, Faccio and Lang, 2002). While these empirical findings might appear to be consistent
with the view that groups arise to replace missing markets, they also raise a puzzling question.
One of the key differences between a pyramidal structure and alternative organizational structures
such as a horizontal one is that the pyramid exacerbates the potential for agency conflicts between
minority shareholders and the family who retains control. One example is suggested by the evidence
in Bertrand, Mehta, and Mullanaithan (2002). Since controllers have incentives to tunnel cash flows
from firms where they have low cash flow stakes to firms where they have higher stakes, pyramids
will automatically create an incentive for the controlling family to expropriate shareholders of firms
that are at the bottom of the pyramid. But if this is the case, then why is the pyramid created
in the first place? Even if there is some value-enhancing reason for the firm to be controlled by
the family, why can’t the family control the firm directly through a horizontal structure that will
presumably reduce the incentives to steal?
8
The conventional wisdom for the existence of pyramids gives up value-enhancing explanations
altogether. The traditional view starts from the assumption that families value control because
of the large private benefits associated with it. With a pyramidal structure, a family can obtain
control of a firm with a small equity stake. Furthermore, a family might want to reduce its equity
holdings while retaining control, in order to maximize the amount of capital under its control
(Hilferding, 1910, and Berle and Means, 1932). An alternative, but similar argument is that when
private benefits of control are large families might elect to maintain a lock on control by separating
cash flow from voting rights (Bebchuk, 1999, and Bebchuk, Kraakman, and Triantis, 2000). The
common element in these stories is that pyramids are cost-efficient mechanisms used by families
to retain control of voting rights. In particular, pyramids could be more efficient than horizontal
structures because a group of firms can be controlled in a pyramid with a lower aggregate cash flow
stake.
This traditional view of pyramids can potentially explain why pyramids are observed in countries with low investor protection, if these are also countries where private benefits of control are
particularly large (Bebchuk, Kraakman, and Triantis, 2000). The evidence in Bertrand, Mehta,
and Mullanaithan (2002) can also be consistent with this view of pyramids. If they are set up to
extract private benefits of control, the fact that there is tunneling in pyramids is not particularly
surprising.
However, there are several reasons why a more complete theory of pyramids is needed. Notice
that the traditional story revolves around the notion of control of voting rights. If pyramids are
simply a way of minimizing the monetary cost of achieving control, one should never observe cash
flow stakes in individual firms that are larger than the treshold that is required to achieve control.
This treshold is at most 50%, but will in general be lower than that.10 Thus, one can only argue
that pyramids enhance family control above their cash flow stake when the family’s cash flow stake
in individual firms is sufficiently low.
A major problem for the conventional story is the substantial empirical evidence that in many
countries families do hold large cash flow stakes in the firms that they control through pyramids.
For example, Franks and Mayer (2002) find that in 2/3 of the firms controlled through a pyramid,
10
For example, La Porta, Lopez-de-Silanes and Shleifer.(1999) use empirical control thresholds of 10% and 20%.
Deviations from one-share-one-vote reduce these thresholds even further.
9
the cash flow and voting stake of the controlling shareholder does not straddle a control treshold.
That is, the controlling shareholder could have achieved the same level of control by simply holding
shares directly in the firm (see also Emmons and Schimd 1998). Valadares and Leal (2001) conclude
that, in Brazil, pyramids are not used to deviate from one-share-one-vote. In a study of ownership
and control of Chilean firms, Lefort and Walker (1999) find that the controlling shareholder owns
more cash flow than necessary to achieve control. They compute the ultimate cash flow ownership
of the controlling shareholder in all the members of a pyramidal group and find this ‘integrated’
ownership to be 57%. Thus, the separation of ownership and control achieved through pyramids
is minimal. Attig, Fischer and Gadhoum (2003) find that, in Canada, the cash flow stake of the
controlling shareholder in a pyramid is, on average, 31.78% while the controlling stake is only a bit
higher, 41.68%.
Another problem for the traditional story is that pyramids are not always the only available
method to separate cash flow rights from voting rights. For example, in the extreme case in which
there are no restrictions to the use of dual-class shares, any degree of separation is possible by a
combination of direct holding and dual-class shares. In such a case, why would a family choose
to control a firm through a pyramid instead of issuing dual-class shares? In fact, if the issue is to
control the maximum possible number of firms with the minimum amount of capital, why waste
valuable capital in the form of high capital investments in individual firms instead of issuing nonvoting shares? Nevertheless, the evidence compiled by La Porta, Lopez-de-Silanes and Shleifer
(1999) suggests that pyramids are much more common throughout the world than the direct use of
shares with superior voting rights. This evidence suggests that considerations other than control
rights motivate the creation of pyramidal groups.
In many countries there are restrictions to the use of dual class shares. If these restrictions
exist, there will be an upper bound to the deviation that can be achieved by a combination of
direct holdings and dual-class shares, and the conventional story would predict that pyramids
would appear when the desired deviation from one-share-one-vote is above the maximum allowed
by the law. Empirically, however, we do observe the use of pyramids to separate cash flows and
votes below the upper bound achieved by a combination of direct holding and dual-class shares.
For example, Bianchi, Bianco, and Enriques (2001) measure the ultimate ownership in each firm
of a pyramid and compute the number of units of capital that the controlling shareholder controls
10
with one unit of his own capital. Then they average this ratio for all the firms in a pyramid. As
a benchmark, consider a family who holds directly 50% of the cash flows and votes in a firm. In
this case the ratio is 2. The family can increase this ratio because Italian law allows the issuance of
50% of the firm’s capital in non-voting shares (savings shares or azioni di risparmio). If the family
uses the maximum fraction of dual class shares and retains 50% of the voting shares (i.e., 25% of
the total capital), it can achieve a ratio of 4. Bianchi, Bianco and Enriques find that, while some
pyramids allow the controlling shareholder to control a large amount of capital (e.g., the ratio for
the De Benedetti group is 10.33 and that for the Agnelli group is 8.86), the ratio for other groups
is below 4, and sometimes even below 2 (e.g., for the Berlusconi group, it is 3.66, and for the Pirelli
group it is only 1.95). Volpin (2002) also finds that, although in the level 3 and above in a pyramid
the separation of ownership and control is very large (CF=6%, V=42.6%), in the second level of a
pyramid the separation (CF=22.5%, V=52.5%) is not high relative to what it could be achieved by
a combination of direct holding and dual class shares. Again, this evidence suggests that there must
other reasons for the existence of pyramids, which are not captured by the traditional argument.11
A partial explanation based on regulatory or tax considerations might help explain some of this
empirical evidence.12 Indeed, regulatory and tax considerations such as taxes on inter-company
dividends do seem to affect the incidence of pyramidal structures (Morck, 2003). Other countries
impose explicit restrictions to prevent some type of firms (e.g. banks, pension funds, etc.) from
obtaining a controlling interest in others (see Roe, 1991, for the case of the US). These are not
complete theories as the benefit of pyramids is not spelled out. A related argument suggested in
the literature is that pyramidal structures can be used as an elusive tool to hide the identity of the
ultimate owner from either the market or the state (Bianchi, Bianco, and Enriques, 2001). Our
theory in this paper will not rely on tax and regulatory considerations, even though we believe
these to be important in the real world.
11
In addition, Claessens, Djankov and Lang (2000) and Faccio and Lang (2002) find that pyramids are used more
often than dual class shares even in countries (such as Italy, Brazil and Sweden) where the restrictions on dual class
shares do not always bind.
12
Gomes (2000) also suggests a rationale for why separation of cash flow and voting rights might appear in countries
with poor investor protection in a model where reputation effects are an important governance mechanism. However,
and similarly to the traditional view of pyramids, his model cannot explain why pyramids appear to be more common
than direct issuance of dual-class shares.
11
2.3
Related theory
The argument that business groups arise as a substitute for missing financial markets is directly
related to the idea that internal capital markets in conglomerates mitigate the effect of external
financing constraints on the conglomerate’s divisions. There is a vast literature in corporate finance
which models and tests this idea (see Stein, 2001, for a survey).13 While some aspects of our theory
can be thought of as applications of these ideas to the question of pyramidal business groups, we
believe that there are also important differences in the arguments. The most important difference is
that the literature on internal capital markets does not focus on the question of optimal ownership
structure, which is so crucial for the analysis of pyramidal business groups. In other words, the
literature on internal capital markets considers only the question of whether a division should be
a part of a conglomerate with other divisions, or a stand-alone firm. This corresponds to one of
the questions that we need to answer when building a model of pyramidal business groups - why
business groups arise. However, the existing literature on internal capital markets has little to say
about the second question that we analyze in this paper - why business groups are organized as
pyramids, as opposed to a horizontal structure.
This second question may seem to be more directly related to the literature on the optimal
design of organizations (see Harris and Raviv, 2002, and Stein, 2002, for recent contributions
that include surveys of this vast literature). One might conjecture, for example, that a horizontal
structure is a “flatter” organizational structure than a pyramid, which looks more hierarchical.14
In this literature it might matter whether organizations are flatter or more hierarchical because,
for example, of frictions in the flow of information among decision-makers. While we do not deny
the possibility that these considerations might have some explanatory power for the structure of
pyramidal business groups, our approach in this paper follows a very different route. In fact, we
assume that firm productivity (gross of diversion costs) is independent of whether business groups
are organized as pyramids or horizontal structures. We believe this is a reasonable assumption,
especially when the controlling family effectively controls and manages all the different firms in
13
Stein (1997), Fluck and Lynch (1999), and Inderst and Muller (2003) are a few theoretical papers that are most
closely related to our approach.
14
In particular, a pyramid might represent a hierarchy in which managers at the lower levels of the pyramid are
subordinates of managers at upper levels.
12
pyramids and horizontal structures.15 In such a case, it is not clear whether there are any significant
differences in the flow of information (for example) between the pyramidal and the horizontal cases.
Furthermore, because the empirical evidence indicates very strongly that new firms are always set
up at the bottom of the pyramid, in order to explain pyramidal ownership as a hierarchy one would
require an organization theory that predicts that managers of new firms are always subordinated
to the previous managers.
3
Pyramidal and horizontal structures
Consider the following set up, which has three dates. At date 0 a family (F) sets up a firm A,
keeping a fraction α of its shares. At date 1, firm A generates a free cash flow equal to c, and the
opportunity to set up another firm (firm B) arises. In this section we assume that α and c are given,
and that the family is the only possible owner of firm B. The latter assumption effectively means
that firm B must belong to the business group that is controlled by family F. Later we consider the
effect of a potential competitor that can own firm B, and discuss properties of the optimal contract
at date 0, especially whether the family wishes to rule out pyramids or not. We also assume that
the family has all its wealth invested in firm A (that is, there is no outside wealth other than the
cash c). We analyze the optimal ownership choice of firm B, that is, whether B will be owned by
the shareholders of firm A (the pyramidal case), or whether it will be owned directly by the family
(the horizontal structure). In both cases the family can issue new shares in B if it chooses to.
Firm B requires an investment of i and generates revenue of r at date 2. We assume that r > i,
that is, the project is positive NPV. F decides whether to set up firm B and which structure to use
(pyramidal or horizontal). In a pyramidal structure, F sets up firm B as a subsidiary of firm A and
thus can use the cash c in firm A to set up firm B. To raise more funds, F can sell additional shares
of firm B. In an horizontal structure, F sets up firm B himself. He has access only to his personal
wealth of αc. In this case, he can also sell a stake in firm B to raise additional funds.
We assume that F always retains control of firm B, irrespective of its cash flow stake in B.
This would be the case, for example, if there is no legal limit to the deviation from one-shareone-vote that can be achieved with dual class shares. This assumption is meant to rule out the
15
La Porta, Lopez-de-Silanes and Shleifer. (1999) have direct evidence that the controlling family participates
directly in the management of most firms controlled through pyramids.
13
advantage that the conventional argument have attributed to pyramids, namely, that they allow a
larger deviation from one-share-one-vote than dual class shares and hence allow a family to control
a firm with a smaller cash flow stake. Without legal restrictions on dual-class shares, F can control
firm B in a horizontal structure holding as little a stake as he wants. Thus, in our setting, the
conventional argument is mute as to the choice of structure. While counterfactual in some countries,
this assumption allows us to concentrate on the trade-off that we want to emphasize. We introduce
legal limits to dual class shares in section ?? (to be done).
Control allows F to divert cash from firm B into his pocket. At date 2, and regardless of the
structure, F can divert funds from firm B at a cost. We assume that this cost depends on the
amount diverted and on the level of investor protection. In particular, when F diverts a fraction d
of the cash flows he pays a cost (one can think of this as waste involved in the diversion process)
that equals a fraction kd2 /2 of cash flow of the firm. Along the lines of Shleifer and Wolfenzon
(2002), we interpret the parameter k as the level of investor protection: a high cost of diversion
(high k) corresponds to good investor protection.16
The crucial trade-off that we analyze is the following. In the pyramidal structure, F has more
internal cash (the entire c) to invest. However, even though the potential NPV of the project is the
same in the two structures (r − i), the actual NPV that is captured by the family can be different
because of two reasons. First, the existing shareholders of firm A will be entitled to a share (1 − α)
of the non-diverted fraction of the payoff (fraction 1 − d). If there is no diversion, for example, the
existing shareholders will get (1 − α)(r − i) out of the new project. However, in general there will
16
We assume that the diversion opportunities are the same regardless of the structure F uses. In a more complete
model, this assumption could be derived from the assumption that F retains control. Control confers F the right to
take a number of actions not ruled out by a previous contract or the law. If F retains control regardless of structure,
his set of feasible actions and hence his diversion opportunities should be the same. A different assumption, implicit in
discussion about pyramids (perhaps inspired by the structures’ ‘geometry’) is that, while in the horizontal structure
diversion from firm B goes to the F’s pocket, in the pyramidal structure the diverted must go to firm A. First,
empirically, it is not always the case that diversion goes to the firms at the top of the pyramids. In some instances,
controlling families divert resources from firms in the pyramid to themselves (in the form of high salaries, for example)
. It is also the case that families divert from firms in the pyramid to other firms they hold that are not part of the
pyramid (quote see FT on Turkey). In any case, empirical evidence on actual diversion (an equilibrium outcome) can
not shed light on the diversion opportunities, which is what our assumption is about. As an example assume that
control confers F the right to sell the output of firm B to any firm at any price. Our assumption simply states that,
since F controls firm B regardless of the structure he chooses, he also has this right in both structures. Consider now
actual diversion level. Since, in a pyramid, the stake of F in firm B is smaller, F will have an incentive to sell the
output of B to firm A at below market prices (a form of diversion). In an horizontal structure, however, the relative
size of F’s stake in firm A and firm B is not mechanically determined. If for example, the stakes are of roughly the
same size, F will have no incentives to price the output of firm B either above or below the market price. Thus, the
fact that we observe actual diversion from firm B to firm A in a pyramid but not in an horizontal structure does not
tell us anything about diversion opportunities.
14
be diversion in equilibrium, and the amount of diversion will depend on the ownership structure.
In particular, for a given direct ownership level that the family has in firm B diversion will be
greater in the pyramid, because of the chain of control - if direct ownership in firm B is equal to β,
the ultimate ownership in the pyramidal case is αβ. Thus, diversion will tend to be greater in the
pyramid.
To sum up, the pyramidal structure increases the availability of internal funds but it may
decrease the NPV of the project and the fraction of the security benefits that is captured by the
family. The horizontal structure decreases financing capacity,17 but it also decreases diversion
(increasing the NPV net of diversion costs), and ensures that the family will capture the entire
NPV of the project.
We solve the model by working backwards.
3.1
Date 2
At date 2, F can divert from firm B to its pocket. Assuming the “ultimate” (taking into account
the chain of control) cash flow stake of F is x, then at date 2 F maximizes:
max x(1 − d)r + dr −
d∈[0,1]
kd2
r,
2
The FOC is: −x + 1 − kd ≥ 0 (with strict inequality only if x = 1) and so:
½
d(x) =
1−x
k
1
if k ≥ 1 − x
.
k <1−x
(1)
The level of diversion decreases with ultimate ownership concentration and with the level of investor
protection. For future reference, we define N P V (x) = r − i −
kd(x)2
2 r
as the total value generated
by firm B when the ultimate ownership concentration is x. Note that N P V 0 (x) > 0 because higher
ownership concentration reduces diversion and consequently reduces the total cost of diversion.
17
If F could costlessly divert the cash of firm A and use the entire c to set up his own firm, the horizontal structure
would clearly dominate. However, at date 0, it would be impossible for F to sell any shares of firm A as investor would
expect to get nothing in return. As a result, all firms would be financed entirely by F with no outside shareholders.
We could consider an intermediate case in which F could divert at a cost some of the cash of firm A. In this case,
however, the results would be similar to those we have, since the same trade-off would appear. In a pyramidal
structure, F would have access to the entire cash c, and in an horizontal structure, F would have access to less cash
(possibly, however, more than αc due to diversion).
15
3.2
Date 1
We next analyze the choice of ownership structure at date 1. For each structure (horizontal or
pyramidal), we calculate F’s payoff and also whether the structure is feasible (that is, whether F
can raise sufficient funds to set up firm B). We then characterize the region of the parameter space
over which each structure is chosen.
3.2.1
Horizontal
F has cash of αc available from his ownership in firm A. He sells 1 − βH of firm B to the public and
keeps a fraction βH (his ultimate ownership). Investors expect F to divert a fraction d(βH ) of the
cash flows. The total cash available to F for investment is:
RH = αc + (1 − βH )(1 − d(βH ))r.
(2)
There are two opposing effects of increasing βH . On the one hand, the direct effect is that fewer
shares are sold, reducing the amount collected. On the other hand, with higher ownership concentration, diversion is lower and as a result the price per share is higher, increasing the amount
collected. We show in the appendix that RH is concave, first increasing with βH and decreasing.
The maximum firm size that can be financed in an horizontal structure is given by
½
RH = max RH =
βH ∈[0,1]
αc + k4 r
if k ≤ 2
.
1
αc + (1 − k )r if k > 2
The function RH attains its maximum when
½
βH =
1−
0
k
2
if k ≤ 2
.
if k > 2
Conditional on RH ≥ i, F’s payoff at the end of date 2 is given by:
UH = αc + r − i −
∗ )2
kd(βH
r
2
(3)
∗
= αc + N P V (βH
)
∗ is the optimal ownership concentration.18 F gets the entire NPV of the project and incurs
where βH
the costs associated with diversion. To minimize diversion (i.e., maximize his payoff), he chooses
18
We use an upper bar for variables associated with maximizing the amount raised, and an asterix to denote
variables associated with maximizing F’s payoff.
16
∗ is the maximum
the highest level of βH that satisfies RH (βH ) ≥ i. By continuity of RH (), βH
solution to
∗
RH (βH
)=i
3.2.2
(4)
Pyramidal
There is a stock of cash equal to c available in firm A. Firm A sells βP of firm B to raise more cash
and retains 1 − βP . The ultimate ownership of F in firm B is then αβP and so diversion is given by
d(αβP ). The total cash available to F for investment is:
RP = c + (1 − βP )(1 − d(αβP ))r.
(5)
As in the case of RH , RP first increases and then decreases with βP .
The maximum amount that can be raised in a pyramidal structure is given by

c
if k < 1 − α

(α−1+k)2
RP = max RP =
c + 4αk r if 1 − α ≤ k ≤ 1 + α

βP ∈[0,1]
c + (1 − k1 )r if k > 1 + α
The corresponding values of βP that maximize RP are:

1
if k < 1 − α

1+α−k
βP =
if
1−α≤k ≤1+α
 2α
0
if k > 1 + α
Finally, conditional on RP > i, F’s payoff is
kd(αβP∗ )2
r
UP = αβP∗ (1 − d(αβP∗ ))r + d(αβP∗ )r −
2
·
¸
kd(αβP∗ )2
∗
∗
= αc + α [(1 − d(αβP ))r − i] + d(αβP ) −
i
2
(6)
(7)
= αc + N P V (αβP∗ ) − (1 − α)[(1 − d(αβP∗ ))r − i]
where βP∗ is the optimal direct ownership concentration when the pyramidal structure is chosen.
As in the case of the horizontal structure, βP∗ is the maximum solution to:
RP (βP∗ ) = i
(8)
The second line of the equation 6 follows by using the above equation.
The expression in equation 6 makes it clear what are the differences in payoff between the
pyramid and the horizontal structures. In a pyramidal structure the family gets the NPV of the
17
Pyramidal structure
Contribution
αβP
αc
(1 − α)βP
(1 − α)c
1 − βP
(1 − βP )M V
Family
Non-family
shareholders of A
Horizontal structure
Stake
New direct
shareholders
Family
New direct
shareholders
Stake
Contribution
αβP
αc
(1 − α)βP
(1 − α)βP M V
+ 1 − βP
+ (1 − βP )M V
1 − αβP
(1 − αβP )M V
Table 1: Types of shareholders in firm B according to structure
project but, since firm B is set up as a subsidiary of firm A, F shares part of the value with the
shareholders of firm A. Since shareholders of firm A do not participate in the diversion proceeds,
they just get their share (1 − α) of the security benefits (i.e, benefits after expropriation) created by
firm B. Notice that these benefits are not necessarily positive when diversion occurs in equilibrium.
Furthermore, the NPV of the project net of diversion costs is endogenous, and depends on the
ultimate ownership stake held by the family. Of course, the family’s payoff in equation 6, as
well as in equation 3, are conditional on the project being taken. The choice between pyramidal
and horizontal structures will also be driven by differences in the financing capacity in the two
structures, RH and RP .
3.2.3
Choice of structure
We find the optimal structure by comparing F’s payoff for all feasible structures. We first state
a number of results and then fully characterize the region of the parameter space for which each
structure is chosen.
Result 1 When the pyramid is chosen, shareholders of firm A realize a negative return. That
is, when the pyramid is chosen, the market value of firm B, (1 − d(αβP∗ ))r, is smaller than the
investment, i.
The proof of this result, as well as all other proofs, is in the appendix.
The intuition behind this result lies in the shareholder distribution in a horizontal and a pyramidal structure. Table 1 shows this distribution assuming that the ultimate ownership of the family
in firm B is the same in both structures, that is, βH = αβP . This assumption makes the comparison
relevant, as it keeps the value of firm B constant across structures.
18
When firm B is a pyramid, it has three types of shareholders: the family, non-family shareholders
of firm A, and new direct shareholders. New direct shareholders hold a fraction 1 − βP of firm B
and pay a fair price for their shares, (1 − βP )M V , where M V = (1 − d(αβP ))r is the market value
of firm B. Non-family shareholders of firm A hold a stake (1 − α)βP of firm B. Since F uses the cash
c in firm A to set up the firm, these shareholders, in effect, pay (1 − α)c for their stake. Similarly,
the family holds a fraction αβP of firm B and pays αc for it.
In an horizontal structure there are only two types of shareholders in firm B: the family and
new direct shareholders. New shareholders buy a stake of 1 − βH = 1 − αβP in firm B and pay
a fair price for it, (1 − αβP )M V. To allow easy comparison, table 1 decomposes the stake of new
direct investors in two stakes of (1 − α)βP and 1 − βP . The family retains a fraction βH = αβP of
firm B and contributes all his wealth to this firm, i.e., αc.
Comparing line by line in the table, it can be seen that the difference between structures is
that while in horizontal structure the stake (1 − α)βP is sold at market prices, in the pyramidal
structure, this stake is effectively sold to non-family shareholders of firm A at a price of (1 − α)c.
The result above can now be explained. Suppose the price paid by non-family shareholders of
firm A is below the market price (that is, c < βP M V ). These will be the only cases when the
existing shareholders will end up with a positive gain in the pyramidal structure. We will show
here that in such a case, the family always chooses the horizontal structure. First, notice that the
family prefers the horizontal structure in this case because it does not need to share the positive
non-diverted NPV with non-family shareholders of firm A. In addition, the family can also invest
more with the horizontal structure, because new shareholders will pay more for the stake (1 − α)βP
than do non-family shareholders of firm A. Thus, every time the non-family shareholders of A gain
in the pyramidal structure, the horizontal structure is both feasible and better for the controlling
shareholder.19
Next we describe the effect of the parameters in the choice of structure.
Result 2 The pyramidal structure is less likely to be optimal when
• Firm B generates high revenues
19
This intuitive proof implicitly assumes that the family issues new shares in both cases. However, the result holds
more generally even when the family does not need to issue new shares.
19
• Firm B requires a small investment
• Firm A generates higher cash flows
• Investor protection is good
The intuition for this result can be derived from result 1. If, when set up in a pyramid, the
market value of firm B is above the necessary investment (that is, (1 − d(αβP∗ ))r ≥ i), then our
previous result shows that the pyramidal structure will not be chosen because the family prefers
to keep the entire NPV of the project by using a horizontal structure. Now, if the revenue of firm
B is high, the inequality 1 − d(αβP∗ ))r ≥ i will be relaxed for two reasons. First, keeping diversion
constant higher revenues translate into higher cash flows for firm B’s shareholders. Second, when
revenues are high the family needs to sell a smaller cash flow stake in firm B. This increases
ownership concentration and reduces diversion also contributing to the increase in market value.
When the required investment in firm B decreases, it is more likely that the market value of
firm B is above the investment required for two reasons. The first is the obvious decrease in the
investment. The second is that when the investment required goes down the family needs to sell
a smaller cash flow stake to raise the necessary funding. As a result, ownership concentration
increases and so does the market value.
The effect of higher c on the availability of internal funds is stronger in the pyramidal case. As
a result the pyramidal structure becomes more attractive.
Finally, when investor protection increases diversion is smaller for any level of ownership concentration. As a result, the market value of B is higher and so it is less likely that the value is
below its investment requirements.
Figure 1 shows the regions of the parameter space over which each structure is chosen. The
pyramidal structure is chosen for low levels of r and high levels of i. When investor protection
increases (not shown in the figure), the curve that separates the regions where horizontal and
pyramidal structures are chosen shifts down, effectively reducing the area over which the pyramidal
structure is chosen.
This figure also shows a region over which the pyramidal structure is the only feasible structure.
This region exists when the stock of cash c is high enough. Still in this range existing shareholders
of firm A must lose when the family makes the new investment. However, the existence of such a
20
region makes it possible that the pyramid is efficient for the “coalition” of the controlling family
and the minority shareholders. This happens because the full NPV of the project, net of diversion
costs, might be positive and so the total value generated by the pyramidal investment is positive.
(we haven’t proved yet that the total NPV, net of diversion costs, can really be positive in this
region but we think it will hold).
In addition to the choice of structure, our model has implications for the ultimate ownership in
each structure, when it is chosen in equilibrium. Of course, whether a structure is chosen or not
is a function of the underlying parameters of the model. Here we focus on variation in ownership
choices induced by the revenue of firm B (r).
Result 3 Suppose there are a number of families that differ only with respect to the parameter
measuring firm B’s revenue, r, and investment required. Variation in r and i leads to corresponding
variation in ownership structures, with different ownership concentrations. Ultimate ownership
concentration in any of the resulting pyramidal structures is lower than that of any of the resulting
horizontal structures.
Note that this result is stronger than simply comparing the ultimate ownership for given parameter values. It says that the observed ultimate ownership should be lower in a pyramidal structure.
Note also that this result is not trivial. In the pyramidal structure the chain of control mechanically reduces the ultimate ownership. However, the fact that the pyramid has more funds to invest
(compared to the horizontal structure) allows for a higher direct ownership in firm B. The above
results show that the first effect dominates.
It should also be the case that direct ownership will be larger in the pyramid, but we haven’t
proved it yet. In both structures you need to raise i. The horizontal structure has greater financing
needs, and thus from this effect we get that direct ownership in the horizontal structure is lower.
However, there is a countervailing effect. The pyramidal needs less external money but it is worse
at raising money than the horizontal (since the chain of control reduces share prices). When c is
large, it is very likely that the direct ownership in the pyramidal is higher.
21
4
Business groups
We have defined a business group as an organization in which a family owns and controls more
than one firm. In the last section we assumed that F is the only party with the ability to set up
firm B. This effectively means that we assumed the existence of a business group. In this section
we investigate the conditions under which a business groups arises.
We modify the model by assuming that at date 1 there is an entrepreneur with the ability to set
up firm B. The set up cost of firm B for the entrepreneur is also i. In order to capture the possibility
that the family may not be the best owner of firm B, we assume that the entrepreneur is a better
manager than the family, so that under his control revenues of firm B are (1 + t)r with t > 0.20
Assume also that the entrepreneur’s wealth is zero. This assumption intends to capture the idea
that the family has higher financing capacity than the entrepreneur, because of the accumulation
of internal funds in existing firms owned by the family. Recall that we also assumed that the
family’s outside wealth is equal to zero, so this assumption really means that the entrepreneur is
less wealthy than the family. We believe this to be the empirically relevant case.
We assume that the market in question only allows for one firm, and that if the entrepreneur can
raise sufficient funds he will be the only one to enter the market because of his higher productivity.
If he cannot raise the necessary funds, then the family sets up firm B using any of the two structures
as described in the last section.
Result 4 Business groups are less likely when investor protection is good and when the profitability
of firm B is high.
The comparative advantage of the family is that they have accumulated wealth, and thus need
to rely less on external capital markets. As investor protection improves, the comparative advantage
of the family eventually disappears and the entrepreneur is able to set up his firm. The entrepreneur
is also better able to set up the firm when the firm is very profitable, since the higher the revenues
the more the entrepreneur can raise in the external market.
Recall that in section 3 we argued that the horizontal structure is chosen when firm B is
profitable and when investor protection is higher. However, it is precisely in these situations that
20
The case when the family is the best owner is trivial, because the family also has more financing capacity and so
will always own firm B.
22
the entrepreneur can set up the firm. As a result, competition from an entrepreneur reduces the
region of the parameter space over which a business group uses an horizontal structure.
Result 5 Whenever they arise, business groups are more likely to use pyramidal structures than
horizontal ones.
Figure 2 shows the region of the parameter space over which the entrepreneur sets up a firm,
and the region over which the business group is formed. As the talent t of the entrepreneur,
investor protection or the revenue r increase, the region over which the entrepreneur arises grows
and that over which the business group is formed is reduced. One can also see from the figure that
the area that used to correspond to the horizontal structure starts to disappear first. Thus, the
model suggests that the conditions that are conducive to the formation of business groups are also
conducive to the formation of pyramids.
5
Ruling out pyramidal structures
Section 3 shows that whenever the pyramidal structure is chosen, the shareholders of firm A lose.
In addition, the ultimate ownership structure in a pyramidal structure is lower than that in an
horizontal structure. This leads to higher diversion levels in a pyramid. These results are consistent
with a growing empirical literature showing significant expropriation of investors in firms that
belong to pyramidal structures. On the other hand, these results beg the question of why pyramids
are chosen in the first place. After all, rational investors who anticipate expropriation discount the
price of the shares they buy. Even more problematic is the fact that pyramidal structures arise in
countries with poor investor protection. It is precisely in these countries that the family would like
to reduce expropriation.
The first response that we have to these questions is that families do not set up all their firms
at one point in time, but rather do it over time.21 Thus, they face a time inconsistency problem.
Once their first firm is set up and they have sold some of the shares, they will be facing the
type of incentives described in section 3. When projects with low profitability or high financing
requirements arise, they will prefer to set them up in a pyramidal structure. If their existing
firms have accumulated enough internal funds, they will be able to finance them. Absent a good
21
This also seems to be the case in the data, as we discuss in section 6.
23
commitment mechanism the family will end up using pyramids. In addition, as shown in section 3,
they will be more likely to do so in poor investor protection countries.
The family can solve the time inconsistency problem by committing not to set up a pyramid.
There is, however, a cost of doing this. In particular, as shown in figure 1, there is a region of
the parameter space over which the only feasible way to set up the firm is by using the pyramidal
structure (need to prove NPV can be positive in this range). Thus, committing not to use a pyramid
will raise the price of shares of firm A, but at the expense of a reduced probability of undertaking
firm B.
6
Empirical Implications
Our theory generates a number of empirical implications, which we list and discuss here. We also
mention empirical and anecdotal evidence that is consistent with our theory, and suggest some
additional implications that can be tested in the future.
1. Family business groups will be more prevalent in countries with poor investor
protection
This implication is a direct consequence of the “value-enhancing” part of our theory (business
groups arise to substitute for missing financial markets). We believe there is no systematic evidence
about this, but there is some scattered and anecdotal evidence. For example, La Porta, Lopez-deSilanes and Shleifer (1999) show that pyramids (though not necessarily all types of business groups)
are more prevalent in countries with poor investor protection. Claessens et al. (2002) show that in
several developing Asian countries more than 50% of the firms in their sample are affiliated with
business groups.
2. Business groups are more likely to be organized as pyramids
In our model, the conditions that are conducive for the choice of pyramids over horizontal
structures (such as low investor protection), are also the same conditions that make business groups
more likely. Thus when business groups tend to arise, they are likely to be organized as pyramids.
Again, there is no systematic evidence on this point. Perhaps the best evidence that groups are
organized as pyramids is the fact that researchers have treated these two terms as synonymous.
24
3. It is possible to observe pyramids where the controlling family has high cash
flow stakes in member firms, and thus the separation between ownership and control
is not large
In our model the controlling family has an incentive to issue as few new shares as possible when
undertaking new investments, in order to minimize diversion. This is true both when horizontal
structures or pyramids are chosen. It is also true for unaffiliated firms. However, direct ownership
in pyramidal firms tend to be particularly high because the pyramid needs to raise the least amount
of capital, and because diversion is higher in pyramids making it harder for the pyramid to raise
new equity.
Thus, our model is consistent with the substantial empirical evidence that in many countries
families hold large cash flow stakes in the firms that they control through pyramids (see also the
discussion in sections 1 and 2). Furthermore, it is not surprising to find that pyramids arise even
when families did not exhaust the possibility of issuing dual-class shares to separate ownership and
control. In our model, pyramids are not used to separate ownership and control.
4. Diversion is higher in firms placed at the bottom levels of the pyramid, than in
firms controlled directly by the family through horizontal structure and firms at the
top of the pyramid
In our model, the amount of diversion is determined by ultimate ownership of cash flow rights
(as opposed to direct ownership), taking into account the effect of the pyramid’s chain of control.
We show that every time the pyramid is chosen the ultimate ownership of the family in new firms
will be lower than in horizontal structures, and thus diversion is higher. Despite the fact that
pyramids replace missing markets in our model, they are associated with high levels of diversion.
This prediction is consistent with empirical findings that pyramids increase incentives to expropriate
(Bertrand, Mehta, and Mullanaithan 2002).
One caveat is that the comparison with stand-alone firms is not entirely clear cut, because these
firms need to raise more funds in order to invest, and thus may be pushed towards lower ultimate
ownership.
5. Performance will tend to be lower in firms that are owned through pyramids,
than in unaffiliated firms or horizontal structures
25
Our model predicts that projects of lower profitability will be undertaken inside pyramids.
Thus, even when the pyramid does not have a direct negative effect on performance, one could
observe a negative relationship between measures of firm value such as Tobin’s Q and pyramidal
membership. As we discussed in section 2, there is evidence in the literature that firms in business
groups organized as pyramids have lower Tobin’s Q (Claessens et al., 2002). Similar results are
found by Lemmons and Lins (2003), Lins (2003), Mitton (2002) and Joh (2003). There is also
evidence that low Q predicts pyramidal membership (Attig, Fischer and Gadhoum, 2003), which
is consistent with the idea that pyramids undertake lower profitability projects.
However, the evidence on the relationship between pyramidal ownership and value is not unambiguous. Some papers find little evidence of a correlation between membership in business groups
and value (Claessens et al., 2002), while others find a positive effect for some firms (Khanna and
Palepu, 2000). Our model suggests that the difference in performance between pyramids and unaffiliated firms should become more negative when the divergence between votes and cash flows
is greater in pyramids (the chain of control effect). This prediction is consistent with findings in
Claessens et al. (2002).
6. Firms in pyramids are larger or they are in capital intensive industries
As the required investment increases, projects are more likely to be undertaken inside pyramids
(as opposed to horizontal structures and unaffiliated firms). Attig, Fischer and Gadhoum (2003)
find evidence consistent with this implication, using Canadian data. Claessens et al. (2002) also
find that in East Asia group firms tend to be larger than unaffiliated firms. Bianchi, Bianco and
Enriques (2001) find similar evidence for Italy.
7. Pyramids are created dynamically
Our model suggests that pyramids are created as families use retained earnings from existing
firms to invest in new ones. Thus, pyramids evolve over time, as a function of the performance of
the existing firms in the pyramid. There is some anecdotal evidence for this. Khanna and Palepu
(2000) claim that one of the most important role of groups is to set up new firms in which the
family and the member firms acquire equity stakes (p. 869). Aganin and Volpin (2003) describe
the evolution of the Pesenti group in Italy, and show that this group was created by adding new
subsidiaries to the firms the Pesenti family already owned. Claessens et al. (2002) find that firms
26
with the highest separation of votes and ownership (those at the bottom of the pyramid) are younger
than those with less separation (those at the top).
8. Pyramids are more likely to be created following good performance of existing
family firms
In our model previous good performance of the existing group increases the chance that pyramids
arise, because of the crucial role of internal funds in the model. Again, there is some anecdotal
evidence for this. Aganin and Volpin (2003) suggest that, in Italy, business groups expand through
acquisitions when they are big and have significant cash resources.
7
Concluding Remarks
(to be written...)
27
References
Aganin, A. and P. Volpin, 2003, History of Corporate Ownership in Italy, mimeo.
Aoki, M. (editor), 1984, The economic analysis of the Japanese firm, North Holland.
Attig, N., K. Fischer and Y. Gadhoum, 2003, On the determinants of pyramidal ownership:
Evidence on expropriation of minority interests, working paper, Laval University.
Bae, K. H., J. Kang, and J. M. Kim, 2002, Tunneling or value added? Evidence from mergers by
Korean business groups, Journal of Finance, 2695-2740.
Bebchuk, L., 1999, A rent protection theory of corporate ownership and control, NBER working
paper 7203.
Bebchuk, L., R. Kraakman, and G. Triantis, 2000, Stock pyramids, cross-ownership, and dual class
equity, in Randall K. Morck, ed.: Concentrated Corporate Ownership (University of Chicago
Press, Chicago, IL).
Berle, A. and G. Means, 1932, The modern corporation and private property, McMillan.
Bertrand, M., P. Mehta and S. Mullanaithan, 2002, Ferreting out tunneling: An application to
Indian business groups, Quarterly Journal of Economics, 121-148.
Bianchi, M., M. Bianco, and L. Enriques, 2001, The separation between ownership and control in
Italy, mimeo, Bank of Italy.
Claessens, S., S. Djankov, J. Fan, and L. Lang, 2002, Disentangling the incentive and entrenchment
effects of large shareholdings, Journal of Finance 57(6), 2741-2771.
Claessens, S., S. Djankov, and L. Lang, 2000, The separation of ownership and control in East
Asian Corporations, Journal of Financial Economics 58, 81–112.
Claessens, S., S. Djankov, and L. Lang, 2002, The benefits of group affiliation: Evidence from
East Asia Working paper University of Amsterdam.
Emmons, W. and F. Schmid, 1998, Universal banking, control rights and corporate finance in
Germany, Federal Reserve Bank of St. Louis Review July/August, 19-42
Faccio, M., and L. Lang, 2002, The ultimate ownership of Western European Corporations, Journal
of Financial Economics 65, 365-395.
Fluck, Z., and A. Lynch, 1999, Why do firms merge and then divest? A theory of financial synergy,
Journal of Business 72, 319-46.
Franks, J., and C. Mayer, 2001, Ownership and control of German corporations, Review of Financial Studies 14(4), 943-977.
Gomes, A., 2000, Going public without governance: managerial reputation effects, Journal of
Finance 55, p. 615-646.
Harris, M. and A. Raviv, 2002, Organization design, Management Science 48, p. 852-865.
Inderst, R., and H. Mueller, 2003, Internal vs. external financing: An optimal contracting approach, forthcoming, Journal of Finance.
28
Joh, S. W., 2001, Corporate governance and profitability: evidence from Korea before the economic
crisis, forthcoming, Journal of Financial Economics.
Khanna, T., 1999, Business groups and social welfare in emerging markets: existing evidence and
unanswered questions, forthcoming, European Economic Review.
Khanna, T., and K. Palepu, 1997, Why focused strategies may be wrong for emerging countries,
Harvard Business Review 75, 41-51.
Khanna, T., and K. Palepu, 1999, The right way to restructure conglomerates in emerging markets,
Harvard Business Review 77, 125-134.
Khanna, T., and K. Palepu, 2000, Is group affiliation profitable in emerging markets? An analysis
of diversified Indian business groups, Journal of Finance 55(2), 867-891.
Khanna, T., and J. Rivking, 1999, Estimating the performance effects of groups in emerging
markets, Harvard Business School working paper.
Khanna, T., and Y. Yafeh, 2001, Business groups and risk sharing around the world, Harvard
Business School Working Paper.
La Porta, R., F. Lopez-De-Silanes, and A. Shleifer, and R. Vishny, 1997, Legal determinants of
external finance, Journal of Finance 52, 1131-1150.
La Porta, Rafael, Florencio Lopez-de-Silanes, and Andrei Shleifer, 1999, Corporate Ownership
around the world, Journal of Finance 54(2), 471-517.
Lefort, Fernando, and Eduardo Walker, 1999, Ownership and capital structure of Chilean conglomerates: Facts and hypotheses for governance, Working paper Pontificia Universidad Catolica
de Chile.
Lemmon, Michael L., and Karl V. Lins, 2003, Ownership structure, corporate governance and
firm value: Evidence from the East Asian financial crisis, Journal of Finance xx, xx-xx.
Lins, Karl V., 2003, Equity ownership and firm value in emerging markets, Journal of Financial
and Quantitative Analysis xx, xx-xx.
Mitton, Todd, 2002, A cross-firm analysis of the impact of corporate governance on the East Asian
financial crisis, Journal of Financial Economics 64, 215-241.
Morck, Randall K., 2003, Why some double taxation might make sense: The special case of
inter-corporate dividends, Working paper University of Alberta.
Morck, Randal K., David A. Strangeland, and Bernard Yeung, 2000, Inherited wealth, corporate
control and economic growth: The Canadian disease, in Randall Morck ed.: Concentrated
Corporate Ownership (University of Chicago Press, Chicago, IL).
Pagano, M. and P. Volpin, 2001, The Political Economy of Finance, Oxford Review of Economic
Policy.
Perotti, Enrico C., and Stanislav Gelfer, 2001, Red barons or robber barons? Governance and
investment in Russian financial-industrial groups, European Economic Review 45, 1601-1617
Stein, J., 1997, Internal capital markets and the competition for corporate resources, Journal of
Finance 52, 111-133.
29
Stein, J., 2001, Agency, information and corporate investment, forthcoming in the Handbook of
Economics and Finance, edited by G. Constantinides, M. Harris and R. Stulz.
Stein, J., 2002, Information production and capital allocation. Decentralized versus hierarchical
firms, Journal of Finance 57, p. 1891-1821.
Valadares, Silvia, and Ricardo P. C. Leal, Ownership and control structure of Brazilian companies,
Working paper Universidade Federal do Rio de Janeiro.
Volpin, Paolo, 2002, Governance with poor investor protection: Evidence from top executive
turnover in Italy, Journal of Financial Economics 64, 61-90.
30
Appendix
Proof of Result 1
Suppose not, that is, the pyramid is chosen and
(1 − d(αβP∗ ))r ≥ i.
∗
βH
Recall that
last equality
and
βP∗
are defined (see equations 4 and 8) by
∗
RH (βH
)
(9)
= i and
RP (βP∗ )
= i, respectively. From the
RP (βP∗ ) = c + (1 − βP∗ )(1 − d(αβP∗ ))r = i
(10)
βP∗ (1 − d(αβP∗ ))r = c + (1 − d(αβP∗ ))r − i
βP∗ (1 − d(αβP∗ ))r ≥ c
Where the second line is obtained by rearranging terms and the last line follows from equation 9. Now,
RH ≥ RH (αβP∗ ) = αc + (1 − αβP∗ )(1 − d(αβP∗ ))r
= αc + (1 − βP∗ )(1 − d(αβP∗ ))r + (1 − α)βP∗ (1 − d(αβP∗ ))r
≥ αc + (1 − βP∗ )(1 − d(αβP∗ ))r + (1 − α)c
= RP (βP∗ ) ≥ i
where the first inequality follows because RH is the maximum value of RH (βH ), the second line follows from algebraic
manipulations, the third line uses equation 10, and the last line is the definition of RP . Two conclusions can be
drawn from the above. First, RH ≥ i, that is the horizontal structure is feasible. Second, since RH (αβP∗ ) ≥ i and
∗
∗
RH (βH
) = i, then it follows from the fact that RH is decreasing that βH
> αβP∗ .
Finally,
∗
∗
U H (βH
) − U P (βP∗ ) = N P V (βH
) − N P V (αβP∗ ) + (1 − α)[(1 − d(αβP∗ ))r − i] > 0.
∗
∗
Since βH
> αβP∗ then N P V (βH
) > N P V (αβP∗ )and (1 − d(αβP∗ ))r − i > 0 by equation 9. This is a contradiction
since the horizontal structure is feasible and yields a higher payoff to F.
Proof of Result 2
In the text.
Proof of Result 3
We show the result for variation in r. The proof for i is almost identical. Suppose that r is the treshold level
such that for r ≥ r the horizontal structure is chosen and for r ≤ r the pyramidal structure is chosen (this is shown
∗
(r) and βP∗ (r) be the direct ownership chosen when revenue of firm B is r in the horizontal
in Result 2). We let βH
∗
and in the pyramidal structure, respectively. Note that when r increases so do βH
(r) and βP∗ (r) since fewer shares
need to be sold to finance a given investment.
∗
We first show that βH
(r) ≥ αβP∗ (r). By definition of r, F is indifferent between the two structures, that is
∗
U H ≡ ac + N P V (βH
(r)) = αc + N P V (αβP∗ (r)) − (1 − α)[(1 − d(αβP∗ (r))r − i] ≡ U P .
∗
By Result 1, it must that [(1 − d(αβP∗ (r))r − i] ≤ 0 which implies that N P V (βH
(r)) ≥ N P V (αβP∗ (r)). Since N P V ()
∗
∗
is increasing, then βH (r) ≥ αβP (r).
Take any set of parameters values (indexed by 1) for which the horizontal structure is chosen, and a different
set of parameter values (indexed by 2) for which the pyramidal structure is chosen. By Result 2, it must be that
∗
∗
r1 ≥ r ≥ r2 . Finally, βH
(r1 ) ≥ βH
(r) ≥ αβP∗ (r) ≥ αβP∗ (r2 ),where the first and third inequalities follow because both
∗
∗
βH (r) and βP (r) are increasing and the second inequality was proved above.
31