Debt Financing and Limited Liability in Business Groups

Debt Financing and Limited Liability in Business Groups.



Magda Bianco* Giovanna Nicodano**

Bank of Italy, Ufficio Diritto dellEconomia Universitك di Torino.
bianco.magda@insedia.interbusiness.it giovanna.nicodano@unito.it.


First Draft, December 1999

This Draft, April 2001





















* The views expressed in this paper are those of the authors and do not involve the responsibility of
the Bank of Italy.

We are grateful to Arnoud Boot, Luigi Buzzacchi, Gabriella Chiesa, Szusanna Fluck, Bruno Parigi,
Enrico Perotti, Rafael Repullo and Bruce Smith for helpful conversations, and to participants in the
ESSFM 2000 at Studienzentrum Gerzensee, the EEA 2000 and EARIE 2000, Tor Vergata VIII
Financial Conference, and in seminars at CEMFI, Ente Einaudi, University of Freiburg, of Padova
and of Torino for useful comments. Giovanna Nicodano is also indebted to CEMFI, the Tinbergen
Institute and the University of Haifa for hospitality, and to MURST Cofin98 for funding. The usual
disclaimer applies.

---

Debt Financing, Limited Liability and the Internal Capital Market
in Business Groups.



ABSTRACT

This paper highlights that the holding company in a pyramidal group is not responsible for
its subsidiaries" debt obligations. This implies that there can be limited liability for holding
companies, beside the usual form of limited liability for individual shareholders that is associated to
independent companies. This in turn suggests that the allocation of debt across companies within a
pyramid is not irrelevant when bankruptcy is costly, since limited liability could not be used if debt
is entirely raised by the holding. We propose a model of group capital structure, showing how
external debt is allocated to holding and subsidiaries and how it is affected by the Controlling
Shareholder’s (CS) cash-flow share in each company.
The empirical section examines the correlation between company characteristics and
external debt in a sample of Italian groups. External debt over assets is smaller in operating than in
holding companies, and is also smaller in listed companies after controlling for usual determinants.
The correlation between external debt and the CS cash flow share - when statistically significant - is
found to be positive. These surprising results are consistent with one equilibrium configuration of
our model, in which the Controlling Shareholder commits not to increase risk in subsidiaries by
effectively giving up limited liability of the group holding company.


Keywords: capital structure, insolvency risk, pyramidal groups, internal capital market, protection
of minority shareholders, asset partitioning.
JEL Classification Number: G32; K22; P59; D82



2

---

1. Introduction.

Business groups are a widespread corporate organizational form across developed (ECGN,
1997; La Porta et al., 1999) and developing countries (Khanna, 2000). Groups often have a
pyramidal structure, in which a controlling shareholder (CS) owns the majority of voting rights in
each company, either directly -- as in the holding at the top of the pyramid of companies -- or
indirectly -- as in subsidiaries, that are separate legal entities. Each company can both raise debt
from external financiers and transfer resources to other companies within the group. This paper
studies how debt is allocated inside business groups. A simple model suggests one reason why this
choice is relevant, while the empirical section provides evidence on debt financing in Italian groups
by studying the correlation between company characteristics and the portion of bank debt they
raise. Our empirical study relies on a unique data base, combining balance sheet figures with
information on intra-group loans and the controlling shareholder’s cash flow share.
The allocation of debt across companies within a group should affect its value because no
obligation is imposed on the holding or sub-holding companies for the insolvency of a subsidiary,
unless it can be proved that it resulted from some fault on their part.1 This implies that in groups
there is limited liability for holding and sub-holding companies, beside the usual form of limited
liability for individual shareholders that can be found in independent companies. Given this peculiar
form of limited liability, raising external debt from subsidiaries with potentially large losses should
be value maximizing when bankruptcy is costly. This opens up the option of partial liquidation,
which is exercised when realized losses to one of such indebted companies threaten the solvency of
other companies in the group. This allows to continue operations in the profitable subset of the
pyramid2.
In the model below the first best group capital structure allocates debt to operating
companies, thus protecting private benefits from control (which represent bankruptcy costs) -
thanks also to the implementation of safer investment projects. This may not be the equilibrium
allocation under asymmetric information between borrowers and lenders due to the entrepreneur’s
incentive to increase risk of subsidiaries’ investment projects. Indeed, it may pay the entrepreneur to
choose riskier projects in operating companies when lenders charge the lower interest rate
associated to safer investment. This is because expected profits from the riskier project - conditional
on the subsidiary being successful - are larger and the higher probability that debt will not be paid

1 This is true only if that the subsidiary is not fully owned by the group. See Blumberg (1989) for the US, Hadden
(1996) on Britain, France, Germany and the US, and Libonati (1996) on Italy.
2 Limited liability coexists with the advantages of diversification and inside information. If an insolvent subsidiary can
be rescued thanks to other firms’ cash flows and if - conditional on the CS information – the investment is expected to
pay off in the future, then the option of partial liquidation is not used.

3

---

back is not accounted for by lenders in the interest rate. Anticipating this moral hazard problem
lenders charge a higher interest rate, and the equilibrium is sub-optimal because bankruptcy costs
are incurred in too often relative to the first best. In this equilibrium, as well as in the first best one,
the correlation between the cash flow share and debt is negative and external debt is raised by
subsidiaries. Alternatively, the entrepreneur may choose to tie her own hands by raising a
sufficiently high amount of debt in the holding company which prevents it from enjoying limited
liability. With this amount of debt the holding goes bankrupt when the subsidiary is insolvent, and
the entrepreneur consequently pursues the safer investment projects. In this equilibrium, the
correlation of debt with the cash flow share may turn positive and debt could be higher in holding
companies than in subsidiaries.
Agency theory has long recognized the conflict of interest between managers and
shareholders. The former own a small share of cash-flows and extract private benefits from control
of company operations, while the latter are the less-informed residual claimants. In groups there is a
similar agency problem between the CS, who is typically involved in management and has a small
share of cash-flows, and minority shareholders. Recent analyses point to incentive problems in the
choice of investment projects in both partial (Bebchuk et al, 1998, and Marseguerra, 1997) and
general equilibrium (Wolfenzon, 1998). Our model also shows that the entrepreneur may have an
incentive to raise more bank debt from companies where his cash-flow share is lower, thus shifting
onto outside shareholders a larger portion of the burden of debt service.

In our sample parent companies raise a larger portion of external debt over assets and turn
out to be net lenders with respect to subsidiaries. The correlation between net external debt over
assets and the cash flow share - when statistically significant - is positive, and this aligns also with
the fact that external debt is lower for listed companies. These results suggest that the equilibrium
with entrepreneur’s commitment not to increase risk is prevailing. We do control for standard
variables used in the capital structure literature (size, profitability, intangibles…) since external
financing should also response to those factors applying to independent companies (incentive
mechanism for managers; method to raise financing without diluting voting rights; bargaining tool
vis-à-vis workers…). Hence our results on the differences between holding and operating
companies support the specificity of capital structure in business groups.
Our analysis compares external debt of companies with different roles (operating versus
holding) within each group since we are interested in exploring group capital structure, i.e. the
choice to allocate external debt across group companies. The design of our experiment differs from
that by Faccio et al (2000) who compare independent to group-affiliated listed companies without

4

---

controlling for their role within the group. They also suggest that incentive theories of debt predict a
negative correlation between the cash flow share and debt and they find the opposite result in
European groups – as we do.

This paper is organized as follows. Section 2 relates this paper to the literature. Section 3
describes the implications of the model and other trade-offs arising from allocating external debt in
lower rather than in higher levels of the pyramid. In section 4 we present the empirical analysis
before the concluding comments. Appendix A presents our model.

2. Neighboring Literature.
In this section we relate our paper to the literature on multidivisional conglomerates, capital
structure and limited liability.
In a group the CS allocates external resources and cash-flows coming from both the holding
and the subsidiaries to investment projects. The headquarter in a conglomerate (which is a typical
corporate organization in Anglo-Saxon countries) also shifts funds across divisions, and such flows
of funds have already been studied both theoretically and empirically. The literature on internal
capital markets shows that in conglomerates the allocation of capital among divisions improves,
relative to the case of stand-alone companies, because monitoring incentives are stronger (Gertner,
Scharfstein and Stein, 1994) and the well informed headquarter is able to allocate scarce financial
resources to the best projects (Stein, 1997). There are however influence costs that may distort
budget choices (Rajan et al., 2000). These arguments carry over to the context of groups, but are not
the focus of this paper. We instead highlight that the group’s Controlling Shareholder typically
holds different equity stakes in each of the companies, whereas the shareholders of a conglomerate
hold the same stake in all the divisions. Another neglected difference between groups and
conglomerates lies in the relationship with external financiers. Divisions of a conglomerate do not
have autonomous access to the credit market. Hence empirical research on conglomerates studies
how division cash-flows are redistributed within the internal capital market to investment
opportunities (Lamont, 1997; Shin and Stulz, 1998). In a group each company can instead ask for a
loan from a bank, in which case it becomes responsible for its service; however, the CS might still
prefer centralizing such operations in the holding company, which is the group counterpart of the
conglomerate headquarter. Our analysis focuses on the allocation of debt across group companies,
which are characterized by different equity stakes of the CS and are responsible for their debt
obligations, bypassing the issue of the allocation of funds to the most profitable investment
opportunities.

5

---

Research on investment and liquidity constraints highlights differences between the
financing needs of group-affiliated and stand-alone companies (Hoshi et al, 1991). No previous
paper studies the allocation of debt within groups, to our knowledge.
This is the case also in the empirical literature on capital structure, that typically examines
the relationship between different measures of leverage and some explanatory variables, such as
size, profitability, the portion of intangible assets and the standard deviation of stock returns as a
proxy for risk. Comparative studies of capital structure (Rajan and Zingales, 1995; Faccio et al,
2000; Booth et al. 2001) use consolidated accounts when studying companies with subsidiaries,
thus by-passing the issue of the allocation of debt among them. We use unconsolidated accounts,
and therefore we also use a measure of “external” leverage accounting for loans obtained from
outside-the-group financiers.
We address limited liability in groups with respect to debt obligations, and recognize that
lenders and minority shareholders may alter financing conditions ex-ante in response to anticipated
increases in subsidiaries’ riskiness. MacMinn and Brockett (1995) deal instead with potential
liability claimants that cannot ex-ante charge the Controlling Shareholders. They recognize that a
carve out, in those conditions, redistributes value from future liability claimants to the shareholders
in the holding company.

3. Debt Financing within the Group.
3.1 Cash flow share, limited liability and external debt.
This section explains why limited liability makes group capital structure relevant for the
entrepreneur’s payoff and summarizes the implications of the model which is presented in the
Appendix. If bankruptcy is costly, then raising external debt from subsidiaries as opposed to the
holding company at the top gives to the group’s Controlling Shareholder an option with non-
negative value even in a risk-neutral world. If a subsidiary becomes insolvent, the CS is able to
either (a) rescue it when profits from the rest of the group are sufficiently large; or (b) let the firm
go bankrupt, when profits are insufficient for its rescue, thus saving bankruptcy costs for the rest of
the group. Gains from co-insurance, (a), can be exploited also if the holding company alone raises
external debt and makes loans to subsidiaries. Gains from limited liability cannot instead be
exploited if the subsidiaries are financed by the holding company raising all the external debt for
them: the whole group goes bankrupt if a company incurs into sufficiently large losses. But this
line of reasoning can be replicated as we proceed down the pyramid of companies. It follows that
external debt should be raised by operating companies at the bottom- other things being equal.
This argument holds, provided that risk-taking is observable, despite the fact that the interest rate
charged to the subsidiaries is higher. Indeed external lenders require a higher interest rate from

6

---

subsidiaries because they anticipate their higher bankruptcy probability.
In the model the group is composed just of a holding and of an operating company, and the
gains from avoiding the bankruptcy of the holding company thanks to limited liability are
represented as private benefits from control. The operating company may start a safer or a riskier
project.3 Figure 1 summarizes the equilibrium configurations. In the first best equilibrium the
operating company adopts the safer project while debt is allocated to the subsidiary, so as to save
the holding from bankruptcy when the subsidiary is insolvent. This may not be an equilibrium
under asymmetric information, however, when lenders cannot observe the risk of investment
projects. It may pay the entrepreneur to choose a riskier project in the subsidiary given that lenders
charge the lower interest rate associated to the safer project4. This happens because expected profits
from the riskier investment - conditional on the subsidiary being successful - are larger and the
higher probability that debt will not be paid back is not accounted for in the interest rate.
Anticipating this moral hazard problem lenders charge a higher interest rate, and the equilibrium is
sub-optimal because too much risk is incurred in relative to the first best.
Alternatively, the entrepreneur may choose to tie her own hands by raising a sufficiently
high amount of debt in the holding company. With this debt the holding goes bankrupt as well when
the subsidiary is insolvent, and this induces her to pursue the safer investment project in the
operating company in order not to lose her private benefits from control.
In two out of the three equilibrium the controlling shareholder raises external debt from the
operating company with a lower cash-flow share, so as to use limited liability and to shift onto the
minority shareholders a larger share of the cost of external debt. In the equilibrium cum
commitment to the safer project, debt can be raised by the holding company and the correlation
between debt and the cash flow share is less clear-cut.

3.2 Collateral, Transparency, Managerial Incentives and Legal Provisions.
The model does not account for three aspects, all of which would imply debt in subsidiaries
and a negative correlation between the cash-flow share and debt. The first one is credit rationing,
which the entrepreneur may be able circumvent by pledging collateral. It is therefore possible to
observe safer investment projects and collateralized debt in operating companies.5

3 The model does not allow the holding company – which could be thought of as the consolidation of the parent with
other subsidiaries - to rescue its subsidiary. This requires a minor extension of the state space and does not alter the
results.
4 The literature associates limited liability with an increase in risk taking beginning with Stiglitz and Weiss (1980).
5 The seniority structure of debt is however rather peculiar in a group. Subsidiary debt, which has assets as collateral,
is senior with respect to debt in the holding that has equity claims on such assets as collateral. However, when the
holding goes bankrupt and its lenders win control over the group, debt in subsidiaries effectively becomes ju nior to that
in the holding. Appendix B reports about the Ferruzzi restructuring, in which subsidiaries’ lenders lost money to

7

---

It is known that debt is a discipline device for managers when there is separation of
ownership from control, and managerial objectives differ from shareholders’. In pyramids this is not
the most prominent concern because the Controlling Shareholder is often the group CEO making
use of effective internal control mechanisms. However if monitoring by the CS gets less intense as
her stake in the company falls, we may expect debt to substitute for it. This would lead to negative
correlation between debt and the cash-flow share.
The model also assumes that lenders know the structure of the group, which need not be the
case. The bank may prefer lending to a subsidiary if data on the structure of groups are not available
because it is then easier to analyze a subsidiary creditworthiness rather than the group’s 6.
Finally, legal protection of minority shareholders is overlooked. The rule that the directors
of a subsidiary must act in the interests of their own company is also common to all leading
jurisdictions (Hadden, 1996). Company law usually requires firm directors and managers to behave
in the best interest of firm’s shareholders rather than of the group’s controlling shareholder. If
external debt is raised from subsidiaries, then the holding receives resources from them and it is
difficult to argue in court that this is welfare improving for a subsidiary’s minority shareholders.
However, the CS is able to appoint new directors and fire managers in case the latter do not follow
her prescriptions – given that she usually holds absolute voting majority. A priori it is not clear
whether the protection of minority shareholders is successful. However “there is no evidence of any
general acceptance in practice of the principles …. and plenty of evidence of widespread
avoidance” in major jurisdictions (Hadden, 1996).

4. Empirical Analysis.
4.1 Sample description.

The sample we use in our empirical analysis is obtained by matching two data sets. The first
has information for the years 1992 and 1996 on ownership and control provided by listed
companies to the Italian market supervisory authority (ConSoB). It contains data on all “listed
groups” (which include at least one listed company) and their pyramidal structure, and reports all
shareholders with holdings larger than 2 per cent and all listed companies’ shareholdings exceeding
10 per cent. We are therefore able to construct the variable “cash flow share”, defined as the amount
that is actually invested in the company by the CS divided by total equity capital. This variable is
computed by accounting for the controlling agent’s direct and indirect shareholdings along the

holding companies’ lenders.
6 The degree of transparency varies across countries (ECGN, 1997). In Germany it is still difficult to understand group
structure. In Italy data have systematically been collected in the nineties by the supervisory authorities Banca d’Italia
and CoNSoB.

8

---

control chain. That is, if agent A controls company B with a 50 per cent share and B controls
company C with a 50 per cent share, the cash flow share of A in C is 25 per cent. This data set
originally contains information on approximately 4.000 companies, but in our investigation we
exclude listed companies not belonging to a pyramidal group.

The second data set (Centrale dei Bilanci) reports balance sheet information for a large
sample of Italian companies (approximately 30.000). From this we retrieve information on firm’s
total debt and credit, and in particular on bank debt and on internal debt, i.e. debt towards other
companies in the group. No information is available in the data set on the interest rate applied on
bank loans or on loans received from the internal capital market.

We exclude from our sample bank holding and insurance companies, since their liabilities
are not directly comparable to the debt issued by non-financial firms. Groups in Italy do not have a
“main bank” as in Japan or Germany, however they sometimes have operating firms specializing in
leasing and factoring which we keep in our sample7. We also keep state-owned companies, even if
they might have been subject to political pressures and their debt was guaranteed by the state. In the
regression analysis however we distinguish between public and private companies, and between
financial and non-financial companies.

The number of firms belonging to pyramidal groups for which we also have balance sheet
information is 728 for 1996 (700 for 1992). Our data sources classify them as holding (when their
main activity is to hold shares of other companies of the group) or operating companies. In our
sample 86 (117) are holding companies, 573 (528) are operating companies outside the financial
sector and 69 (55) are operating companies in the financial sector (leasing, factoring …). In 1996
the mean number of companies per group is 35, of which 20.5% are listed. While our sample
contains almost all listed companies, it only has information on 33.3 % of companies per group.
This is because Centrale dei Bilanci reports balance sheets for joint-stock companies and not for
partnerships, which are therefore excluded from our sample.

Summary statistics for the main variables are presented in Table 1 for 1996, and in Table 3
for 1992. We consider three measures of debt financing (scaled with total assets): total financial
debt (debt for short), bank debt and net external debt. The latter is the difference between total
financial debt and total financial credit net of internal debt which is in turn the difference between
total financial debt and total financial credit towards firms belonging to the same group. This
measure accounts for loans obtained from outside the group, and sounds novel to the capital
structure literature because of our use of unconsolidated accounts.

7 The reader is referred to Bianchi, Bianco and Enriques (1999), Buzzacchi and Colombo (1996) and Nicodano
(1998) for information concerning performance, ownership and voting premiums in Italian business groups.
Schiantarelli and Sembenelli (2000) study financing constraints in group-affiliated and independent companies.

9

---

In order to assess how the presence of minority shareholders affects external debt, we
consider separately operating companies which are less than fully owned (which we name “limited”
in the regressions) and a subset of these, namely “listed” companies.
Summary statistics do not reveal major differences between public and private firms as far
as debt financing is concerned, and in our comments below we mainly refer to private companies.
Debt is raised by subsidiaries (.16), however holding and financial companies have higher ratios
(.35 and .27). The same is true for bank debt (.12, .19, .17) and for net external debt ( .15, .19, .31).
This allocation of external debt appears at odds with previous descriptive analyses of internal
capital markets in Italian groups, summarized in Appendix B, which reported little external debt for
private holding companies.
The difference between internal debt and net internal debt shows whether the company acts
as an intermediary with respect to the group. Holding companies intermediate large amounts of
funds (in proportion to their assets), since their average internal debt and net internal debt positions
are .10 and -.08 respectively. These figures are .08 and .03 for financial firms and .09 and .07 for
operating companies. Hence holding companies are net lenders to subsidiaries.
The profitability of a company is measured by its return on investment (the ratio between net
profits plus interest payments and the sum of financial debt and equity). Holding companies show a
markedly lower profitability than operating companies.
The cash flow share of the Controlling Shareholder is as a proxy for the position of a
company in the pyramid: it is decreasing as we proceed from holding companies to financial and
non financial operating companies, and reaches a minimum for companies which are not fully
owned. 8
More than 50% of companies in our sample do not borrow internally (median internal debt
is zero). However, a relevant portion of companies use the internal capital market: firms with either
positive internal debt and/or positive internal credit account for 80% of asset value.

4.2 Regression results for 1996.

The regression analysis is displayed in Tables 2 and 3. The dependent variable in the upper
panel is net external financial debt over total assets, as this captures borrowing from outside-the-
group lenders. It is bank debt over total assets in the lower panel, since it might be measured more
precisely. The use of debt plus equity as scaling variable leads to slightly different results that are
not reported.

8 There are some peculiar companies with very large debt ratios and others with very large losses that reduce the value
of equity below zero. We however use robust estimation methods in order to reduce the impact of outliers on
econometric results.

10

---