University of Southern California Dissertations and Theses

Cash holdings and corporate diversification

(USC Thesis Other)

Cash holdings and corporate diversification

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

CASH HOLDINGS AND CORPORATE DIVERSIFICATION

by

Ran Duchin

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BUSINESS ADMINISTRATION)

May 2008

Copyright 2008 Ran Duchin
ii

Acknowledgements

I am grateful for very helpful comments from Harry DeAngelo, Gareth James, John
Matsusaka, Micah Officer, Oguzhan Ozbas, Berk Sensoy, René Stulz and seminar
participants at Arizona State University, Hebrew University, University of Arizona,
University of Florida, University of Illinois, University of Michigan, University of
North Carolina, University of Oregon, University of Pittsburgh, University of
Rochester, University of Washington, Virginia Tech, Vanderbilt, and Yale.

iii

Table of Contents
Acknowledgements ii
List of Tables iv
List of Figures v
Abstract vi
Introduction 1
Chapter 1: Literature Review 9
Chapter 2: A Model of Liquidity and Diversification 21
A. Setup 21
B. Analysis 22
Chapter 3: Empirical Setup and Main Results 28
A. Sample 28
B. Empirical Models 29
C. Nonparametric Evidence 34
D. Regression Evidence 39
E. Financial Constraints 42
Chapter 4: Robustness 46
A. Subsamples 46
B. Endogeneity 53
C. Agency 58
D. Firm Fixed Effects 61
Chapter 5: Extensions 63
A. Value Implications 63
B. Net Debt 68
C. The Time Trend of Cash Holdings 76
D. Investment-Cash Sensitivity 81
E. Longitudinal Study 84
Conclusion 92
References 95

iv

List of Tables

Table 1: The Distribution of Divisional Investment Opportunities 22
Table 2: Summary Statistics 33
Table 3: Cross Tabulation of Average Cash Holdings 35
Table 4: The Cross-Section of Corporate Cash-Holdings 40
Table 5: Credit Constraints Classification 43
Table 6: Financial Constraints -- The Cross Section of Cash Holdings 45
Table 7: Equal Number of Segments 47
Table 8: Mature Firms 50
Table 9: SFAS 131 (Change in Segments Reporting) 52
Table 10: Instrumental Variables & Transitory Cash 54
Table 11: Endogeneity - Stable Organization Form Firms 57
Table 12: Agency - The GIM Index 59
Table 13: Firm Fixed Effects 62
Table 14: Value Implications 67
Table 15: The Distribution of Average Net Debt 70
Table 16: Cross Tabulation of Average Net Debt 72
Table 17: The Cross-Section of Net Debt 74
Table 18: Average Industry Correlations and Volatilities 80
Table 19: Cash Sensitivity of Investment 84
Table 20: Longitudinal Study 88
v

List of Figures

Figure 1: Average Corporate Cash Holdings from 1990 to 2004 2
Figure 2: Average Firm Cross-Divisional Correlations 77
vi

Abstract
This study documents a strong connection between corporate cash holdings and
organization form, and explores why such a relation exists. Over the period 1990-
2004, U.S. multi-division firms held 39 percent less cash as a fraction of assets than
stand-alone firms, and the difference persists after controlling for firm size, capital
structure, cash flows, growth opportunities and investments. To explain the
difference, I study the determinants of cash holdings in a sample of 10,380 firms
over 1990-2004 and find that (a) correlations between divisional cash flows or
investment opportunities completely explain the relation between cash holdings and
organization form and (b) their effect is big: an increase of one standard deviation in
the correlations between cash flows (investment opportunities) implies an increase
of 13 percent (10 percent) in cash holdings of the average firm. This suggests that
corporate diversification can serve as a way to economize on cash. The paper also
documents a substantial increase in average cross-divisional correlations from 1990
to 2004, which might explain why diversified firms hold more cash than they used
to.
1
Introduction
Cash holdings of U.S. companies are enormous and growing over time. As
of fiscal year 2004, nonfinancial and nonutility firms in the Compustat universe
reported aggregate cash holdings of over 2 trillion dollars, representing 11 percent of
the total market value of equity of these firms.
1
The growth of cash holdings is
equally impressive. According to Bates, Kahle, and Stulz (2006, BKS henceforth),
the average ratio of cash to assets of listed U.S. industrial firms has increased by 129
percent from 1980 to 2004. This massive increase in cash has captured the attention
of both academics and the media. For example, a recent article in the New York
Times states that “Publicly traded American firms hold so much cash that, as a
group, they could pay off all their debt and still have money left over.”
2

While the dramatic increase in cash holdings is receiving growing attention,
another noteworthy pattern is not widely recognized: diversified firms hold 39
percent less cash as a fraction of assets than standalone firms. Figure 1 breaks the
universe of Compustat firms into four clusters based on their reported number of
business segments, and plots the average cash-to-assets ratio from 1990 to 2004. The
figure shows the dramatic increase in cash holdings that others have noted but also
the pronounced difference in cash policies of diversified and specialized firms, a
pattern that has not received much attention. On average, firms with more business

1
Cash holdings are defined as cash and short-term investments (Compustat Data Item 1). Note that
aggregate cash holdings including financial firms and utilities were 6.9 trillion dollars in 2004,
representing 29.4 percent of their total market value of equity.
2
“Behind Those Stock Piles of Corporate Cash”, by Mark Hulbert, New York Times, October 22,
2006.
2
segments hold less cash as a fraction of assets. The cash-to-assets ratio of two-
segment firms is approximately 32 percent less than that of standalone firms. The
cash-to-assets ratio of firms with four segments or more is approximately 50 percent
less than that of standalone firms. While previous evidence suggests there are
economies of scale in cash holdings (Beltz and Frank, 1996; Mulligan, 1997), more
in depth results below will show that a prominent difference remains after taking
into account economies of scale.
Figure 1: Average Corporate Cash Holdings from 1990 to 2004
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The figure reports the average cash/total book assets ratios for firms
with 1, 2, 3, and 4 or more reported business segments.

.05 .1 .15 .2 .25 .3
Average Cash/Assets

1990 1992 1994 1996 1998 2000 2002 2004
Year
1 Segment
2 Segments
3 Segments
4+ Segments

3
The purpose of this study is twofold: (1) to document and quantify the
connection between organization form and cash holdings, and (2) to investigate why
they are connected. A central finding is that differences between cash holdings of
standalone and diversified firms are associated with cross-divisional correlations in
investment opportunities and correlations in operational cash flows, both of which
capture the opportunities of firms to diversify risk by having multiple divisions.
Opler, Pinkowitz, Stulz, and Williamson (1999) provide evidence that
corporate cash holdings are positively correlated with market-to-book ratios, cash
flow levels and industry volatilities (defined as the average cash flow volatility for
firms in the same industry). These findings suggest that firms with better investment
opportunities and higher cash flows, that face greater cash flow risks, tend to hold
more cash as a fraction of assets. The predominant approach to understanding these
relations is the precautionary saving theory, dating back to Keynes (1936). Almeida,
Campello, and Weisbach (2004) develop the theory by highlighting the importance
of financial constraints. They show that cash holdings are only correlated with cash
flows when firms are financially constrained. If firms are not constrained and can
finance first best investment levels, cash-holdings become a matter of indifference.
Recently, Acharya, Almeida, and Campello (2006) study the correlation between
cash flows and investment opportunities. They find that when this correlation is low,
firms will hoard cash to be able to fund investment in states of the world with low
cash flows. I build on this previous work and extend it to consider separately the
correlation between divisional cash flows and the correlation between divisional
investment opportunities. I show that diversified firms are well positioned to be able
4
to smooth cash flows and investment opportunities, because both the outcomes and
the opportunities of their divisions are not perfectly correlated. Therefore,
diversified firms will hold less cash, and increasingly so as cross-divisional
correlations decrease.
In addition to helping understand cross-sectional variation in cash,
investment opportunity and cash flow correlations can help explain why cash
holdings have increased over time (as shown in Figure 1). BKS find that changes in
industry volatility are correlated with the growth of cash holdings over time.
However, the effect of industry volatility depends on the cross-divisional
correlations since lower (higher) correlations can substantially reduce (increase) the
firm’s risk exposure. I show that average cross-divisional investment opportunity
and cash flow correlations have also increased over time, and suggest that this
further explains why investment opportunities and cash flows of diversified firms
have become more volatile, leading in turn to greater cash balances. Furthermore,
when I examine volatilities and correlations at the industry level, I find that average
volatilities have increased from 1990 to 2004, while average correlations have not.
3

This implies that firms could have reduced their exposure to the increase in industry
volatilities (see Irvine and Pontiff, 2005) by diversifying into uncorrelated
businesses (since average cross-industry correlations did not increase).
In order to structure the empirical investigation, I present a model of
corporate liquidity in which firms hold cash to take advantage of new investment

3
Both industry correlations and volatilities are measured using mean growth and cash flow measures
of all standalone firms within each industry.
5
opportunities and avoid the need to tap costly external capital markets. I extend
existing theoretical work by Almeida, Campello, and Weisbach (2004) and others
(e.g., Kim, Mauer and Sherman (1998)) to consider multi-division firms, taking into
account both investment opportunity and cash flow cross-divisional correlations.
The analysis shows that diversified firms can pool their cash resources and hold a
joint amount of cash at the headquarters level, as long as correlations are less than
perfect. It also highlights the key role of financial constraints – the sensitivity of
cash holdings to investment and cash flow cross divisional correlations crucially
depends on the firm’s financial constrains.
With this as motivation, I explore the financial policies of a sample of
Compustat industrial firms over the period 1990-2004. I construct measures of the
cross-divisional average correlations between investment opportunities and between
cash flows. The nonparametric evidence documents a significant variation in
average cash holdings levels across firms with different numbers of business
segments, different cash flow correlations, and different investment opportunity
correlations. For instance, controlling for capital expenditure and firm size, firms
with at least four divisions hold 42 percent less cash than specialized firms. These
findings are supported by subsequent panel data regressions, which control for
various factors that were shown to affect cash holdings in previous studies.
Interestingly, these regressions also show that the number of divisions loses its
explanatory power in the presence of cross divisional correlations, suggesting that
correlations are the driving force behind the relation between cash and
diversification. The regressions document a significant impact of correlations on
6
cash holdings: average cash holdings increase by 10 percent (13 percent) with an
increase of one standard deviation in market to book (cash flow) correlations.
The results are also consistent with the theoretical significance of financial
constraints. Partitioning the sample into financially constrained and unconstrained
firms shows that the results are much more pronounced for financially constrained
firms. Cash holdings are significantly more sensitive to the levels of investment
opportunities and cash flows and to their cross-divisional correlations when firms
are financially constrained.
The findings are then subjected to a series of robustness checks, and are
shown to hold in subsamples as well as when the endogeneity of cash holdings is
taken into account and when alternative hypotheses such as agency-based models
and firm maturity are considered. I also show that the improvement in firms’
segments disclosure following the new accounting standard introduced in 1997
((SFAS 131) strengthens the results.
This study contributes to two prominent areas of financial research. The first
is cash management, which has seen a growing body of research in recent years.
Opler, Pinkowitz, Stulz and Williamson (1999) emphasize the fundamental
distinction between a tradeoff theory of cash (under which firms have optimal cash
levels that balance between the benefits and costs of holding cash) and a financing
hierarchy theory (under which there is no optimal cash level and firms can increase
their cash holdings by taking on more debt). Subsequent literature highlighted
various benefits and costs associated with cash related to the financing of corporate
investments (e.g., Almeida, Campello, and Weisbach, 2004; Acharya, Almeida, and
7
Campello, 2007), taxes (e.g., Foley, Hartzell, Titman, and Twite, 2006), and agency
(e.g., Dittmar and Mahrt-Smith, 2007; Pinkowitz, Stulz, and Williamson, 2006).
This study focuses on the role of cash in reducing the firm’s exposure to investment
and cash-flow risks, and suggests that firms can use their form of organization as a
cash management strategy.
The second area is corporate diversification. The paper extends previous
work on efficient resource allocation resulting from internal capital markets (e.g.,
Stein, 1997; Matsusaka and Nanda, 2002) to include the allocation of cash holdings.
My results suggest that economies in cash could be an important benefit of
diversification to be weighted against the various costs documented in previous
studies (e.g., Scharfstein and Stein, 2000; Rajan, Servaes and Zingales, 2000;
Matsusaka and Nanda, 2002; Ozbas, 2005). Moreover, the ability of diversified
firms to hold less cash might reduce the agency costs of free cash flow (Jensen,
1986). Thus, firms can use their form of organization as a cash management strategy
and save on the costs of holding cash through diversification. Indeed, my results
provide evidence consistent with diversification adding value through facilitating
lower cash holdings. Bringing the two areas of corporate diversification and cash
management together is interesting not only from a theoretical point of view, but
also in practice, because diversified firms in my sample held more than 70 percent
of aggregate corporate cash holdings as of fiscal year 2004
4
.

4
This evidence is broadly consistent with the findings of DeAngelo, DeAngelo and Skinner (2004),
who document a similar concentration in earnings and dividends.
8
Exploiting the imperfect correlations between divisions is also in line with
the coinsurance effect, introduced by Lewellen (1971). In his work, the imperfect
correlations between divisions' cash flows increase the debt capacity of firms by
reducing the probability of default. Empirical research has failed to find conclusive
evidence that diversified firms hold substantially more debt than specialized firms
(e.g., Berger and Ofek, 1995). This study shows that diversified firms do hold more
net debt (defined as debt minus cash) as a fraction of assets. Consistent with BKS
and the New York Times quote from the opening paragraph, net debt to assets ratios
have dropped substantially from 1990 to 2004.
9
Chapter 1: Literature Review

A large and growing body of literature studies the determinants and
implications of corporate cash holdings. Theoretical works on cash holdings are
generally older, and identify various rationales governing the choice and relevance
of cash holdings. Most of the research, however, has focused on the empirical
properties of corporate cash holdings. Such studies test which factors determine cash
holdings in practice, and what the value implications of cash holdings are. Perhaps
the most intriguing empirical question is whether cash holdings are driven by
efficient or inefficient motivations, and whether cash holdings create or destroy
shareholder value.
Conceptual work on cash holdings dates back to Keynes (1936). Keynes
identified two main benefits from holding liquid assets. First, the firm saves
transaction costs to raise funds and does not have to liquidate assets to make
payments. Second, the firm can use the liquid assets to finance its activities and
investments if other sources of funding are not available or are excessively costly.
Keynes describes the first benefit as the transaction cost motive for holding cash,
and the second one as the precautionary motive.
Keynes’ early work implicitly suggests that cash holdings are valuable when
external funding is costly. The seminal work by Modigliani and Miller (1958)
refines and formalizes this idea. Modigliani and Miller show that in a frictionless
world, the choice of cash holdings does not affect firm value, since cash has a zero
net present value. As they point out, cash levels become significant only in the
10
presence of frictions such as incomplete contracting, agency costs of external
funding, and other potential deadweight costs of internal and external financing. The
costs considered in the literature have evolved from brokerage costs, in the classic
paper by Miller and Orr (1966), to inefficient investment resulting from insufficient
liquidity, emphasized in theoretical models such as Jensen and Meckling (1976),
Myers (1977), and Myers and Majluf (1984), as well as in empirical papers that
build on Fazzari et al. (1988).
Empirical work on corporate cash holdings can be broadly divided into two
categories. The first category focuses on various financial policies and
considerations that affect cash holdings. These include investment and payout
policies, capital structure considerations, and the severity of a firm’s financial
constraints. The main shortcoming of much of this literature is that it fails to
recognize and account for the endogeneity or simultaneous determination of the
different financial policies. The second strand of empirical literature on cash
holdings focuses on the efficiency and value-implications of corporate cash
holdings. This work tests whether cash holdings are related to agency and
governance characteristics of the firm, and whether holding cash enhances or
reduces shareholder wealth. Next, I consider each of the two strands of literature
separately.
Starting with the first category, there is some older literature studying the
financial determinants of cash holdings. Chudson (1945), for example, finds that
cash-to-assets ratios tend to vary systematically by industry, and tend to be higher
among profitable companies. Vogel and Maddala (1967) find that cash balances
11
have been declining over time, and that larger firms tend to have lower cash-to-
assets and cash-to sales ratios. This finding suggests that there are economies of
scale in the transaction motive for cash. Baskin (1987) presents empirical evidence
consistent with the notion that liquid assets are employed both to signal commitment
to retaliate against encroachment and to enable firms to preempt new opportunities
A paper by John (1993) argues that firms wish to hold greater amounts of cash when
they are subject to higher financial distress costs. Using a 1980 sample of 223 large
firms, John finds that firms with high market-to-book ratios and low tangible asset
ratios tend to hold more cash. This observation is consistent with the financial
distress theory if one agrees that a high market-to-book ratio is a proxy for financial
distress costs.
More recently, Opler, Pinkowitz, Stulz and Williamson (1999) examine the
properties of corporate cash holdings in the U.S. between 1971 and 1994. They have
in mind two competing views of cash holdings which they try to contrast
empirically. Under the first view (“the tradeoff view”), firms have an optimal level
of cash holdings, where the marginal benefits of cash equal the marginal costs. Costs
include a liquidity premium and tax disadvantages, while benefits include saving
transaction costs to raise funds and being able to finance activities when external
financing is unavailable or too costly. Under the second view (“the financing
hierarchy view”), there is no optimal cash level and firms can increase their cash
holdings by taking on more debt. In accordance with the pecking order model, firms
increase (decrease) their cash holdings when they have more (less) internal
12
resources, but are indifferent between different cash levels.
5
The authors make
empirical predictions along the lines of these two views, while taking into account
information asymmetries and agency costs of debt. They start by testing whether
optimal cash levels exist and whether changes in internal funds drive changes in
cash levels. Both hypotheses cannot be rejected. They find mean-reversion in cash
levels and cannot reject target-adjustment models. They also find that changes in
internal funds go in the same direction as changes in cash-holdings. They then turn
to run panel regressions predicting firm liquidity levels using contemporaneous
independent variables and include dummy variables for each industry defined by the
2-digit SIC code. They find that cash holdings decrease with size, net working
capital, leverage, whether a firm pays dividends and whether it is regulated. Cash
holdings increase with the market-to-book ratio, cash-flow, capital expenditure,
industry cash-flow volatility and research and development expenditure. They also
note that variables that explain leverage tend to explain cash-holdings as well, but
with the opposite sign.
6

Almeida et al. (2004) focus on the linkage between financial constraints and
liquidity management. They suggest that the cash flow sensitivity of cash
(propensity to save cash out of cash flows) can serve as a measure of financial
constraints. They backup their claim with a theoretical model, under which firms

5
This argument disregards any excess cost of cash holding, but if one recognizes the possibility that
such a cost does exist, the firm should not accumulate cash indefinitely. This implies that an optimal
cash level does exist, beyond which the firm should use its cash to pay dividends or repurchase
equity. Jensen's (1986) free cash flow view implies that managers might be reluctant to give up cash,
thus incorrectly providing empirical evidence that supports the financing hierarchy model.
6
However, Graham (1998) finds that the correlation between excess cash and excess debt capacity is
only 11.4%.
13
that expect to be financially constrained in the future accumulate cash today.
Therefore, such firms tradeoff current valuable investments with future investments
and choose their optimal cash policy to balance the profitability of current and future
investments. As a result of their analysis, they conclude that cash holdings of
constrained firms are positively related to cash-flows. On the other hand, their model
implies that unconstrained firms should not exhibit any systematic relation between
cash savings and cash flows, since they are able to finance all their positive NPV
projects and at the same time face no cost of holding cash. Their empirical
prediction suggests that financially constrained firms should exhibit positive cash-
flow sensitivity of cash, while unconstrained firms should not exhibit any systematic
relation. Their sample consists of manufacturing firms between 1971 and 2000.
They find empirical support for their predictions using five measures of financial
constraints.
Focusing on the connection between cash holdings and debt, Acharya,
Almeida, and Campello (2007) examine whether cash can and should be viewed as
negative debt. They model the interplay between cash and debt policies in the
presence of financial constraints. They show that while saving cash allows
financially constrained firms to hedge against future income shortfalls, reducing
debt - "saving borrowing capacity" - is a more effective way of securing future
investment in high cash flow states. This trade-off implies that constrained firms
will allocate excess cash flows into cash holdings if their hedging needs are high
(i.e., if the correlation between operating cash flows and investment opportunities is
low). However, constrained firms will use excess cash flows to reduce current debt
14
if their hedging needs are low. Their empirical investigation reveals evidence
showing that financially constrained firms with high hedging needs have a strong
propensity to save cash out of cash flows, while showing no propensity to reduce
outstanding debt. In contrast, constrained firms with low hedging needs
systematically channel free cash flows towards debt reduction, as opposed to cash
savings. Their analysis suggests that cash should not be viewed as negative debt.
Debt is not the only financial instrument whose relation to cash has been
called into question. One other such instrument is a credit line. Credit lines did not
receive much attention until recently, mainly because data on credit lines is not
available from Compustat. Sufi (2006) uses 10-K SEC filings to examine corporate
usage of bank lines of credit. He finds that the supply of credit lines is particularly
sensitive to historical profitability. In other words, only highly profitable firms
manage to obtain lines of credit. This is the case because banks employ strict
covenants on profitability and firms lose access to their lines of credit when their
profitability drops. This implies that lines of credit are not unconditional obligations
of banks, and therefore are not perfect substitutes for cash. Consistent with this
finding, Sufi also shows that low profitability firms tend to hold larger cash balances
and retain a higher fraction of cash flow as cash holdings.
A different rationale for corporate cash holdings has been suggested by
Foley, Hartzell, Titman, and Twite (2007). While cash holdings have been typically
justified in the existing empirical literature by transaction costs and precautionary
motives, this study claims that U.S. multinational firms hold cash in their foreign
subsidiaries because of the tax costs associated with repatriating foreign income.
15
Consistent with this hypothesis, this study finds that firms facing higher repatriation
tax burdens hold higher levels of cash. The authors also show that such firms tend to
hold this cash abroad and in affiliates that trigger high tax costs when repatriating
earnings. Furthermore, they find that affiliate cash holdings are more sensitive to
repatriation tax burdens when cash is less important domestically. This is
particularly true for firms that are less financially constrained and more technology
intensive.
Focusing on financial constraints, Denis and Sibilkov (2007) provide
evidence that cash holdings are more valuable for financially constrained firms than
for unconstrained firms and investigate why this is so. Their results indicate that
greater cash holdings are associated with higher levels of investment for both
constrained and unconstrained firms, but that the marginal value of investment is
greater for constrained firms. These findings suggest that higher cash holdings allow
constrained firms to undertake value-increasing projects that might otherwise be
bypassed. As such, the evidence is consistent with the hypothesis that greater cash
holdings of constrained firms are a value-increasing response to costly external
financing.
Faulkender and Wang examine the cross-sectional variation in the marginal
value of corporate cash holdings that arises from differences in corporate financial
policy. They begin by providing semi-quantitative predictions for the value of an
extra dollar of cash depending upon the likely use of that dollar, and derive a set of
intuitive hypotheses to test empirically. By examining the variation in excess stock
returns over the fiscal year, they find that the marginal value of cash declines with
16
larger cash holdings, higher leverage, better access to capital markets, and as firms
choose greater cash distribution via dividends rather than repurchases.
One other empirical regularity that has received attention recently is the
secular increase in corporate cash holdings in recent years. Bates, Kahle, and Stulz
(2007) document that the average cash-to-assets ratio for U.S. industrial firms more
than doubled from 1980 to 2006. A measure of the economic importance of this
increase in cash holdings is that at the end of the sample period, the average firm can
pay back all of its debt obligations with its cash holdings; in other words, the
average firm has no leverage if leverage is measured as net debt. This change in cash
ratios and net debt is the result of a secular trend rather than the outcome of the
recent buildup in cash holdings of some large firms, and it is much more pronounced
for firms that do not pay dividends and for firms in industries whose cash flows
became riskier. The average cash ratio increases over the sample period because
firms change: their cash flow becomes riskier, they hold fewer inventories and
accounts receivable, and are increasingly R&D intensive. The precautionary motive
for cash holdings plays an important role in explaining the increase in the average
cash ratio; in contrast, agency considerations are not successful in explaining the
increase.
A second strand of the empirical literature focuses on the connection
between cash holdings and governance or agency characteristics of the firm.
Underlying this literature is the notion that cash holdings might result from non-
value maximizing aspirations of management rather than an optimal account of the
benefits and costs of cash given the financial conditions and needs faced by the firm.
17
These studies examine how firms spend their cash, and how governance affects the
quantity and value of cash, both domestically and internationally.
Harford (1999) shows that cash-rich firms are more likely to attempt
acquisitions than other firms do. Stock return evidence shows that acquisitions by
cash-rich firms are value decreasing. In fact, Harford shows that cash-rich bidders
destroy seven cents in value for every excess dollar of cash reserves held. Cash-rich
firms are more likely to make diversifying acquisitions and their targets are less
likely to attract other bidders. Consistent with the stock return evidence, mergers in
which the bidder is cash-rich are followed by abnormal declines in operating
performance. Overall, this evidence supports the agency costs of free cash flow
explanation for acquisitions by cash-rich firms.
In a different paper, Harford, Mansi, and Maxwell (2005) examine the
relation between the management of cash holdings and a corporate governance
index that includes various anti-takeover provisions (G-index). Using data from the
Investor Research Responsibility Center for the period 1993 through 2002, they find
that firms with high anti-takeover provisions (weak shareholder rights) have smaller
cash reserves. Further tests suggest that firms with weak shareholder rights dissipate
their cash reserves far more quickly than do managers of firms with strong
shareholder rights, primarily through acquisitions. They find some evidence that
firms with weak shareholder rights raise more cash from financing activities. They
also find differences in the source of the cash from the financing activities based on
governance structure. Firms with weak shareholder rights are more likely to issue
debt and less likely to issue equity.
18
Faleye (2004) focuses on the takeover-deterrence effects of corporate
liquidity and suggests the proxy contest as an effective alternative control
mechanism. He finds that proxy fight targets hold 23% more cash than comparable
firms, and that the probability of a contest is significantly increasing in excess cash
holdings. Proxy fight announcement return also is positively related to excess cash.
Faleye also finds that following a contest, executive turnover and special cash
distributions to shareholders increase, while cash holdings significantly decline.
A number of papers examined the value of corporate cash holdings.
Mikkelson and Partch (2003) examine the operating performance and other
characteristics of firms that for a five-year period held more than one-fourth of their
assets in cash and cash equivalents. Following the five-year period operating
performance of high cash firms is comparable to or greater than the performance of
firms matched by size, industry, or prior performance. In addition, proxies for
managerial incentive problems, such as ownership and board characteristics, are not
unusual and do not explain differences in operating performance among high cash
firms. They conclude that persistent large holdings of cash and equivalents have not
hindered corporate performance.
Studying the effect of corporate governance on the value of cash holdings,
Dittmar and Mahrt Smith (2007) compare the value and use of cash holdings in
poorly and well-governed firms. They show that governance has a substantial impact
on value through its impact on cash: $1.00 of cash in a poorly governed firm is
valued at only $0.42 to $0.88. Good governance approximately doubles this value.
Furthermore, they show that firms with poor corporate governance dissipate cash
19
quickly in ways that significantly reduce operating performance. This negative
impact of large cash holdings on future operating performance is cancelled out if the
firm is well governed.
A different method to examine the relation between cash holdings and
corporate governance is to take advantage of the international cross-section of
shareholder protection. For a sample of more than 11,000 firms from 45 countries,
Dittmar, Servaes, and Mahrt-Smith (2003) find that corporations in countries where
shareholders rights are not well protected hold up to twice as much cash as
corporations in countries with good shareholder protection. In addition, when
shareholder protection is poor, factors that generally drive the need for liquidity,
such as investment opportunities and asymmetric information, actually become less
important. These results strengthen after controlling for capital market development.
In fact, consistent with the importance of agency costs, they find that managers
actually hold larger cash balances when capital markets are better developed. Their
evidence indicates that investors in countries with poor shareholder protection
cannot force managers to disgorge excessive cash balances.
A similar approach is taken in Kalcheva and Lins (2007). Their article uses
managerial control rights data for over 5000 firms from 31 countries to examine the
net costs and benefits of cash holdings. They find that when external country-level
shareholder protection is weak, firm values are lower when controlling managers
hold more cash. Further, when external shareholder protection is weak they find that
firm values are higher when controlling managers pay dividends. Only when
external shareholder protection is strong do they find that cash held by controlling
20
managers is unrelated to firm value, consistent with generally prevailing U.S. and
international evidence.
21
Chapter 2: A Model of Liquidity and Diversification
A. Setup
To formulate the main hypotheses investigated in the paper, I consider the
cash policy decision of a multi-division firm that operates in imperfect capital
markets. For simplicity, I study the case of two identical divisions incorporated as a
single firm. The time line of the model is in the spirit of Almeida, Campello, and
Weisbach (2004) and has three dates, 0, 1, and 2. At time 0, the firm is an ongoing
concern, and each division has a cash flow of
0
X from existing assets. At that date,
each division has the option to invest in a long term project that requires an
investment of
0
I today and pays off ) (
0
I f at time 2. Additionally, each division
expects it might have access to another investment opportunity at time 1, with
probability p . If the division invests
1
I at time 1, it generates a payoff of ) (
1
I g at
time 2. At time 1, each division receives a cash flow from existing assets of
1
X .
The production functions ) ( ⋅ f and ) ( ⋅ g are increasing, concave and
continuously differentiable. I assume that the discount factor is 1, everyone is risk
neutral, and the cost of investment goods at dates 0 and 1 is equal to 1. I suppose
that the cash flows ) (
0
I f and ) (
1
I g are not verifiable and thus cannot be contracted
upon. The firm cannot use these cash flows to raise funds from outside investors.
However, it can raise a limited amount of external finance by using its existing
22
assets as collateral.
7
To keep things simple, I assume that for each division the firm
can raise an amount of
0 0
B B ≤ at time 0, and an amount of
1 1
B B ≤ at time 1.
In this setup, the firm is concerned only about whether or not to store cash
from time 0 until time 1. Let
0
S denote the amount of cash per division that the firm
chooses to carry from time 0 until time 1.
B. Analysis
Due to the concavity of the production functions, the firm will invest an
equal amount in each project. Thus, denote the overall time 0 investment as
0
2I (i.e.,
an amount of
0
I is invested in each project). At time 1, the firm will face zero, one,
or two investment opportunities, depending on the realized state of nature. The joint
distribution of investment opportunities follows a bivariate Bernoulli distribution
with a marginal distribution of p , and a correlation coefficient of π . Let
j
O be an
indicator function for the occurrence of an investment opportunity for segment j, i.e.,
1 =
j
O if an opportunity emerges and 0 =
j
O otherwise. The following table gives
the joint probabilities for the two-segment firm:
Table 1: The Distribution of Divisional Investment Opportunities
This table shows the joint and marginal distributions of divisional investment opportunities
1
2
= O 0
2
= O Marginal
1
O
1
1
= O
11
p
01
p
p
0
1
= O
01
p
00
p () p − 1
Marginal
2
O
p
( ) p − 1

7
This idea is in the spirit of hart and Moore (1994), who argue that the liquidation value of “hard”
assets is verifiable by a court. Therefore, creditors can seize those assets if the firm defaults.
23
I assume that the firm is financially constrained: It cannot finance first-best
investments if both investment opportunities become available at time one.
Formally, this implies that
FB FB
I I B B X X
1 0 1 0 1 0
2 2 2 2 2 2 + < + + + . Since unused debt
has zero NPV in this setup, the firm will borrow up to its limit. Therefore, time zero
borrowing is given by
0 0
2B B = . In order to focus on the role of cash when
investments are abundant, I assume that if the firm encounters only one investment
opportunity, it can achieve the optimal investment level:
1 1 1
2 2 X B I
FB
+ ≤ . Thus,
given that the firm chose to maintain
0
2S at time zero, it only needs to borrow
1 1 0 1
2 2 2 B X S I
FB
≤ − − . The following proposition formulates the main result of the
model:
Proposition 1
All else equal, the cash holdings of diversified firms will increase as the correlations
between their divisional investment opportunities increase. Thus, as long as
investment opportunities or cash flows are less than perfectly aligned across
divisions, diversified firms will hold less cash, as a fraction of total assets, than
standalone firms.
24
Proof
We can write the distributions of the above table in terms of the correlation, π , and
the marginal distribution, p , as follows:
8

()( ) p p p p ⋅ + − − = π 1 1
00

() [] p p p p − + = 1
11
π
) )( 1 (
01
p p p p ⋅ − − = π
This implies that investment level at time 0 is given by
0 0 0 0 0 0 0 0
2 2 2 2 B S X I B S X I + − = ⇒ + − = . At time one, we need to consider the
different possible realizations of the state of the world:
No Investment Opportunities, with probability
00
p : 0
1
= I
One Investment Opportunity, with probability
01
2p :
FB
I B X S I
1 1 1 0 1
2 2 = + + =
Two Investment Opportunities, with probability
11
p :
1 1 0 1
B X S I + + =
Given the above analysis, the firm solves the following problem:

( ) ( )
()
1 1 0 11
0 1 1 01 0 00 0 0 0
2
2 2 2 2 2 max
0
B X S g p
S X B g p S p B S X f
S
+ + +
+ + + + + −
(1)
The first order condition is given by:
( ) ( ) ( )
1 1
* *
0 11
* *
0 1 1 01 00 0
* *
0 0
' 2 2 2 ' 4 2 ' 2 B X S g p S X B g p p B S X f + + + + + + = + − (2)

8
To keep the probabilities plausible, one needs to impose the Frechet-Hoeffding bounds:
(a) {} p p p ≤ ≤ −
11
1 2 , 0 max
(b)
()
()
1
1
,
1
max ≤ ≤
⎭
⎬
⎫
⎩
⎨
⎧ −
−
−
− π
p
p
p
p

25
To see how the correlation between investment opportunities affects the
optimal cash holdings, I use the following partial derivatives:
0 ) 1 (
0 ) 1 (
01
11 00
< − − =
∂
∂
> − =
∂
∂
=
∂
∂
p p
p
p p
p p
π
π π

Differentiating the first-order-condition with respect to π yields the following:
() () []
() [] 1 ' ) 1 (
' ' ) 2 2 ( ' ' 4 ' '
1 1
* *
0
1 1
* *
0 11 1 1
* *
0 01 0
* *
0 0
* *
0
− + + − =
+ + + + + + + −
∂
∂
−
B X S g p p
B X S g p B X S g p B S X f
S
π (3)
(The left hand side comes from the fact that: ( ) ( ) 1 ' 2 '
1
* *
0 1
= = +
FB
I g S B g )
Since ( ) 1 '
1 1
* *
0
> + + B X S g , we have that:
0
* *
0
>
∂
∂
π
S
. (4)
We can also compare between the optimal cash-levels of standalone firms
and diversified firms. By setting the correlations equal to 1, the first order condition
in equation (2) implicitly defines the optimal cash holdings of a standalone firm,
*
0
S
:
( ) ( )
1 1
*
0 0
*
0 0
' ) 1 ( ' B X S pg p B S X f + + + − = + − (5)
Equations (2) and (5) are identical when the two investment opportunities are
perfectly correlated (i.e., 1 = π ). However, when the correlation is less than perfect
(i.e., 1 < π ), we have that p p p < + − π π ) 1 (
2
, which implies that
*
0
* *
0
S S < .
Q.E.D
26
Proposition 1 holds once we control for industry volatilities of investment
opportunities faced by the firms. Claiming that diversified firms should optimally
hold less cash than standalone firms implicitly assumes that these firms face the
same industry business risk; otherwise, there is no common basis for comparison.
The same is true for the claim that lower cross-divisional correlations imply lower
optimal cash holdings. Therefore, the empirical analysis that follows will control for
volatilities in order to draw meaningful conclusions about the relation between cash
and correlations. Additionally, in the above analysis diversification has no
deadweight costs. Thus, this analysis cannot be viewed as a theory of corporate
scope, since it implies that infinite diversification is optimal. Previous literature
introduced various costs of diversification, such as agency costs (e.g., Jensen, 1986)
and inefficient investment policies (e.g. Scharfstein and Stein, 2000; Rajan, Servaes
and Zingales, 2000; Matsusaka and Nanda, 2002; Ozbas, 2005), which would
counter balance the cash holdings benefit of diversification. It is therefore important
to emphasize that the model does not attempt to explain the choice of organization
form, as it seems implausible to base a theory of corporate scope on cash holdings
alone.
Setting out to explicitly consider the correlations between divisional
investment opportunities or cash flows extends the previous literature in two
important ways. First, it allows us to take into account the diversification effect that
has bearing on the firm’s risk-exposure. Ultimately, we are interested in capturing
the firm-specific risk exposure in order to explain the cross-section of cash holdings.
This could also be done by considering firm-volatilities directly, but then one cannot
27
disentangle between the business risk and the role of firm-level diversification in
mitigating this risk. Second, the estimation of business risk in previous literature
only took into account one business line per firm (possibly the main business line or
the original one), thus ignoring its other lines of business. This introduces
measurement error into the estimation of firm business risk. In this study, business
risk is measured as a (sales) weighted average of all firm’s business lines.
Finally, note that the above analysis emphasizes the importance of financial
constraints. As Almeida, Campello, and Weisbach (2004) pointed out, cash holdings
should be affected by cash flows only when the firm is financially constrained. If the
firm is not constrained, it can finance its first best investments and it is a matter of
indifference whether the firm hoards cash or distributes it to shareholders. Thus, the
firm’s exposure to cash flow risk is greater when it is more financially constrained.
The same is true for investment risk exposure. If the firm can finance first best
investments through cheap external financing, it loses the precautionary motive for
holding cash. However, when the firm is more financially constrained, it is more
exposed to investment risk. Thus, the empirical investigation that follows should
show that cash levels are more sensitive to cash flow risk and investment risk when
the firm has greater financial constraints.

28
Chapter 3: Empirical Setup and Main Results
A. Sample
The sample includes all firms available from Compustat's North America
Industrial Annual file and Compustat's Segments file. All data are CPI-adjusted into
2004 dollars. Compustat's Industrial Annual file is used to retrieve data on firms'
cash holdings and short-term securities, book assets, sales, operational cash flows,
market to book ratios, debt, capital expenditures, acquisition activity, dividend
payments and net working capital. I use Compustat's Segments file to retrieve data
on a firm's business segments. In particular, I am interested in counting the number
of business segments within each firm and retrieving the industry affiliation of each
segment (represented by 3-digit NAICS codes). Since the Segments file might
contain repeated data years if the reported segments appear on multiple source
documents, I only consider the latest source year of each segment-year observation.
I intersect the data from the Industrial Annual and the Segments files, and exclude
financial firms and utilities. I also eliminate firm-years for which data on cash
holdings are missing, those where cash holdings exceeded the value of total assets,
and those for which market capitalization was less than $10 million (in 2004
dollars). Furthermore, I eliminate all firm-years observations for which I do not
have each segment's primary NAICS code available. I do so because I calculate
industry volatilities (industry volatility henceforth) and cross-divisional correlations
based on data on all standalone firms operating in each segment's industry (see
29
below). All in all, the sample covers the fifteen years period from 1990 to 2004 and
consists of 70,354 firm-year observations.
B. Empirical Models
I group the various factors affecting corporate cash policy into four
schematic models. The first model relates cash balances to the organizational form
of the firm, i.e., its degree of diversification. The model predicts that the firm's cash
holdings will decrease as the number of business segments increases, since more
divisions imply lower cross-divisional correlations on average. However, an inverse
relation between cash holdings and the number of segments does not necessarily
imply that correlations affect cash holdings. An alternative explanation might be
that firms view their non-core segments (i.e., those segments that operate in
secondary industries) as quasi-liquid entities, which can be liquidated in the event of
an increased demand for liquidity. Furthermore, the number of business segments is
a crude proxy for the cross-divisional correlations. For example, it might be the case
that all segments operate in the same industry, in which case the correlation is equal
to one. For these reasons, I employ direct measures of cash-flow and investment
opportunity cross-divisional correlations..
To understand how correlations are measured, consider an economy with N
industries. At time t, define the correlation between the investment opportunities (or
cash flows) in industries i and j ( { } N j i ,..., 1 , ∈ ) as the correlation between the mean
investment opportunities (cash flows) of standalone firms in industries i and j over
the period [t-15,t-1]. Let
t
j i
CORR
,
denote this correlation. Now consider firm k, that
30
reported M divisions at time t, each affiliated with a certain industry (industries are
not necessarily different across the different divisions) and a certain level of sales.
For each pair of divisions, let { } N j i ,..., 1 , ∈ denote the corresponding industries and
let
i
S denote the sales of division i. I define firm k’s sales-weighted average
correlation at time t as follows:
9

∑∑
⋅ =
≠ i
t
j i
i j
i t
k
CORR
SALES
S
CORR
,
2
1
(6)
where SALES is total firm sales. The summation is divided by 2 because each
correlation is counted twice, once for i and once for j.
In practice, I use the mean investment opportunities or cash flow across all
standalone firms operating in each 3-digit NAICS industry. Though I use sales-
weighted average correlations, my results do not change if I use a simple arithmetic
average. Thus, the correlation of a conglomerate firm is the average of the pair-wise
correlations between all its business segments.
10
A similar procedure is used to
calculate industry investment opportunities and cash flow volatilities:
t
i
M
i
i t
k
SIGMA
SALES
S
SIGMA ⋅ =
∑
=1
(7)
where
t
i
SIGMA is the standard deviation of mean investment opportunities (cash
flows) of standalone firms in industry i over the period [t-15,t-1].
The other three models are employed throughout the empirical analysis as
controls. The second model, which I call the "Lifecycle" model of cash

9
This measure follows standard portfolio formulation, but weights correlation i by the sales of
segment i relative to total sales instead of assets.
10
Other correlation specifications, such as the correlation between the two largest divisions produce
similar results.
31
management, implies that mature firms hold lower cash balances. Mature firms are
generating free cash flow and thus tend to finance investment internally. Mature
firms also tend to have more predictable investment needs and greater agency
problems, and both of these factors put downward pressure on optimal cash
balances. I follow DeAngelo, DeAngelo and Stulz (2006) and use the contribution of
retained earnings to total assets (RE/assets) as a proxy for the firm’s lifecycle stage
(Mature firms have higher retained earnings to assets ratios). I also include under
this model operational cash flows over assets, and whether or not the firm pays
dividends.
The third model, which I call the "Capital Structure" model, posits a link
between the firm's capital structure and its cash balances (see Acharya, Almeida, and
Campello, 2006, for a discussion of this framework). Under this view, cash is an
integral part of the firm's capital structure, and serves as a substitute for debt (the
“negative debt” view) and net working capital (excluding cash). Thus, the relevant
variables for the second model are Leverage over assets and net working capital
(excluding cash) over assets (NWC/assets).
According to the fourth model, which I call the "Investment/Growth" model,
cash serves to preempt investment opportunities, and is therefore related to measures
of growth and capital spending (e.g., Opler, Pinkowitz, Stulz, and Williamson, 1999;
Harford, 1999). Growing firms have a higher ex-ante probability of encountering an
investment opportunity, and therefore have a higher demand for liquidity. The most
commonly used measure of investment/growth opportunities is the market to book
ratio. However, John (1993) argues that firms maintain larger cash balances when
32
they are subject to higher financial distress costs, and uses the market to book ratio
as a proxy for these costs. To distinguish between the two hypotheses, I use two
additional proxies for growth opportunities: the growth rate of assets excluding cash
(AGR) and the growth rate of sales (SGR), taken from DeAngelo, DeAngelo, and
Stulz (2006). In addition, this model includes capital expenditure over assets
(Capital expenditure/assets) and acquisitions activity over assets.
Table 2 describes the various variables employed in this study. The table
reveals a wide variation in cash/assets, with a mean of 0.18 and a standard deviation
of 0.22. The median firm has cash equal to 8 percent of its assets. The dependent
variables in my sample also reveal wide variation. For example, market to book has
a mean of 1.96 and a standard deviation of 2.15, while leverage/assets has a mean of
0.27 and a standard deviation of 0.3. Investment opportunities and cash flow
correlations have means of approximately 0.9 (the sample is dominated by
standalone firms) and standard deviations of 0.22-0.29, which indicate that there is a
significant level of dispersion in correlations. Thus, the sample appears to offer a
wide cross-sectional variation in the dependent as well as the independent variables
in my empirical models.

33
Table 2: Summary Statistics
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. Cash/assets is cash and short-term investments over total assets.
Re/assets is retained earnings divided by total book assets. Cash-flow/assets is measured as earnings
less interest and taxes, divided by total assets. NWC/assets is net working capital minus cash over
total assets. Leverage/assets is short-term debt plus long-term debt over assets. Market to book equals
book value of total assets minus book value of equity plus market value of equity divided by total
assets. The asset growth rate (AGR) is the change in total assets, excluding cash, divided by the
previous year’s level, while the sales growth rate (SGR) is defined analogously with respect to
revenue. M/B correlation is defined as the sales-weighted average correlation between mean market
to book ratios over 15 years of all standalone firms in each division's industry, as defined by the 3-
digit NAICS code. AGR correlation, SGR correlation and CF correlation are defined analogously
with respect to AGR, SGR and operating cash-flows, respectively. Industry M/B volatility is the
sales-weighted average standard deviation of mean market to book ratios over 15 years of all
standalone firms in each division's industry, as defined by the 3-digit NAICS code. Industry AGR
volatility, SGR volatility and CF volatility are defined analogously with respect to AGR, SGR and
operating cash-flows, respectively. Size is the natural logarithm of the book value of total assets in
2004 dollars.

Variable Mean Median Std
Cash/assets 0.18 0.08 0.22
RE/assets 0.1 0.13 0.35
Cash-flow/assets 0.04 0.07 0.18
NWC/assets 0.08 0.07 0.24
Leverage/assets 0.27 0.21 0.3
Market to book 1.96 1.41 2.31
AGR 0.15 0.03 0.86
SGR 0.08 0.05 0.28
Capital expenditure/assets 0.07 0.05 0.08
Acquisition/assets 0.02 0 0.07
M/B correlation 0.91 1 0.22
AGR correlation 0.9 1 0.25
SGR correlation 0.91 1 0.23
CF correlation 0.88 1 0.29
Industry M/B volatility 0.37 0.32 0.25
Industry AGR volatility 0.21 0.09 0.56
Industry SGR volatility 0.07 0.06 0.04
Industry CF volatility 0.03 0.02 0.02
Number of segments 1.47 1 1
Size 5.27 5 1.83

34
C. Nonparametric Evidence

The empirical analysis begins in Table 3 by documenting the mean cash
holdings associated with several firm characteristics. For Panel A, firm-year
observations were sorted into 40 bins based on firm size (10 deciles) and number of
segments (1, 2, 3, 4+), and the mean cash-to-assets ratio for each bin is reported. As
noted in the introduction, cash holdings decline as the number of segments increase,
but here we can also see that this is not size effect. The negative relation between
cash and segments appears (and monotonically) for each size class. The magnitudes
of the effects are nontrivial. For the smallest firms (first decile), cash holdings are
18.4 percent of assets when the firm has one segment and 12.9 when the firm has
four or more segments, a 5.5 percent difference. The difference is 3.8 percent for
medium firms (fifth decile) and 2.7 percent for large firms (tenth decile). All of
these differences are statistically different from zero at the 1 percent level.
In Panel B, firm-years are sorted according to different number of business
segments, size, and levels of cross-divisional correlations between investment
opportunities (using each of the three growth variables: market to book, AGR, SGR)
and correlations between industry cash flows. Consistent with the model, average
cash-to-assets ratios monotonically increase with cross-divisional growth
correlations. The results hold for all 3 growth measures, and are of significant
magnitudes. For the smallest 3-segment firms, cash holdings are 11.1 percent when
the firm is in the lowest M/B correlation bin and 17.7 percent when it is in the

35
Table 3: Cross Tabulation of Average Cash Holdings
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. Size is the natural logarithm of the book value of total assets in 2004
dollars. Three proxies are used to measure growth: Market-to-book ratios, AGR (asset growth rate,
excluding cash) and SGR (sales growth rate). The asset growth rate (AGR) is the change in total
assets, excluding cash, divided by the previous year’s level, while the sales growth rate (SGR) is
defined analogously with respect to revenue. Market to book equals book value of total assets minus
book value of equity plus market value of equity divided by total assets Cash-flow/assets is defined
as earnings less interest and taxes, over book assets. M/B correlation is defined as the sales-weighted
average correlation between mean market to book ratios over 15 years of all standalone firms in each
division's industry, as defined by the 3-digit NAICS code. AGR correlation, SGR correlation and CF
correlation are defined analogously with respect to AGR, SGR and operating cash-flows,
respectively. Industry M/B volatility is the sales-weighted average standard deviation of mean market
to book ratios over 15 years of all standalone firms in each division's industry, as defined by the 3-
digit NAICS code. Industry AGR volatility, SGR volatility and CF volatility are defined analogously
with respect to AGR, SGR and operating cash-flows, respectively.

Panel A: Average Cash/assets by the Number of Business Segments and Size

Number of Business Segments
Size
Deciles
1 2 3 4+
1 0.184 0.158 0.139 0.129
2 0.169 0.145 0.135 0.121
3 0.151 0.135 0.119 0.104
4 0.146 0.134 0.111 0.087
5 0.143 0.132 0.107 0.105
6 0.135 0.131 0.105 0.089
7 0.121 0.101 0.096 0.087
8 0.113 0.104 0.094 0.071
9 0.118 0.083 0.073 0.072
10 0.091 0.075 0.067 0.064

36
medium bin. Cross-divisional cash flow correlations also have a sizable effect: the
cash holdings of the smallest 3-segment firms go up from 12.5 percent to 15 percent
when one moves from the lowest cash flow correlation bin to the medium bin.
In addition to cross-divisional correlations, the firm’s risk exposure is also a
function of industry volatility. Thus, one would expect industry volatility to affect
cash holdings as well. In panel C, firm-years are sorted according to different
number of business segments, size, and industry investment opportunity and cash
flow volatilities. Once again, the results document a substantial effect. Across the
different number of segments and size, the average difference between cash holdings
of firms in the lowest volatility bin and the medium bin is 2.1 percent with market to
book, 1.6 percent with AGR, 1.7 percent with SGR, and 1.7 percent with cash flow.
Thus, Table 3 shows sizeable difference along key dimensions in cash. Cash
holdings monotonically increase with correlations and decrease with the number of
business segments. These results hold after taking into account firm characteristics
such as size, and are statistically significant at the 1 percent level.

37
Table 3, Continued
Panel B: Average Cash/assets by the Rank of Industry Growth and Cash Flow correlations,
Number of Segments and Size

M/B Correlation
Rank
AGR Correlation
Rank
SGR Correlation
Rank
CF Correlation
Rank
Number
of
Segments
Size
1 2 3 1 2 3 1 2 3 1 2 3
Small 0.114 0.159 0.188 0.122 0.150 0.188 0.132 0.139 0.188 0.120 0.152 0.188
Medium 0.089 0.127 0.147 0.092 0.123 0.147 0.108 0.124 0.147 0.085 0.131 0.147 2
Large 0.080 0.083 0.097 0.068 0.099 0.099 0.077 0.088 0.097 0.075 0.090 0.097
Small 0.111 0.177 0.201 0.106 0.166 0.215 0.122 0.165 0.201 0.125 0.150 0.213
Medium 0.095 0.100 0.151 0.088 0.106 0.152 0.102 0.122 0.143 0.084 0.100 0.163 3
Large 0.057 0.066 0.093 0.065 0.073 0.088 0.072 0.077 0.080 0.060 0.071 0.095
Small 0.103 0.125 0.186 0.110 0.126 0.178 0.105 0.131 0.178 0.102 0.127 0.214
Medium 0.083 0.085 0.141 0.080 0.084 0.145 0.083 0.088 0.139 0.075 0.086 0.148 4+
Large 0.061 0.063 0.099 0.058 0.070 0.094 0.062 0.069 0.090 0.056 0.070 0.096

38
Table 3, Continued
Panel C: Average Cash/assets by the Rank of Industry Growth and Cash Flow Sigmas,
Number of Segments and Size

Industry M/B
volatility Rank
Industry AGR
volatility Rank
Industry SGR
volatility Rank
Industry CF
volatility Rank
Number
of
Segments
Size
1 2 3 1 2 3 1 2 3 1 2 3
Small 0.087 0.121 0.219 0.121 0.131 0.181 0.109 0.140 0.173 0.112 0.143 0.202
Medium 0.066 0.100 0.217 0.083 0.105 0.183 0.085 0.112 0.174 0.075 0.101 0.188 2
Large 0.058 0.073 0.133 0.064 0.088 0.108 0.063 0.080 0.116 0.074 0.074 0.115
Small 0.090 0.154 0.220 0.138 0.149 0.201 0.138 0.141 0.189 0.113 0.143 0.205
Medium 0.077 0.089 0.186 0.085 0.097 0.162 0.092 0.106 0.147 0.081 0.102 0.148 3
Large 0.052 0.056 0.106 0.062 0.073 0.078 0.049 0.077 0.087 0.065 0.076 0.086
Small 0.116 0.134 0.192 0.161 0.180 0.195 0.134 0.153 0.178 0.124 0.136 0.174
Medium 0.079 0.080 0.158 0.080 0.085 0.140 0.080 0.085 0.140 0.086 0.099 0.121 4+
Large 0.057 0.060 0.100 0.050 0.078 0.087 0.054 0.065 0.097 0.065 0.070 0.092

39
D. Regression Evidence

The results of the previous subsection suggest that diversification has a
sizable effect on cash holdings. However, it might be the case that cross-divisional
correlations and number of business segments serve as proxies for other variables
that affect cash holdings, such as the firm’s size and its lifecycle stage. For example,
the inverse relation between cash and the number of segments can be a side-effect of
economies of scale in cash management. In order to control for such factors and
estimate the marginal effect of diversification, I estimate multivariate regressions
with various variables known to affect corporate cash holdings. Table 4 reports the
estimation of the diversification model in isolation, as well as the results of
estimating "all inclusive" regression models combining the above four models. I
report the results of OLS regressions (with a year fixed effect, corrected for firm
clustering).
11
All the regressions also control for firm size, to account for economies
of scale in cash management.
Model 1 shows that the number of business segments is inversely related to
cash holdings at the 1 percent significance level, even after controlling for size. This
result is consistent with the nonparametric evidence from the previous table. It
strengthens the results of the previous table by establishing robustness to firm
clustering and year fixed effects. This result suggests a possible link between
organization form and cash holdings, but does not reveal the reason behind such a

11
These and subsequent regressions were also estimated with cash/sales as the dependent variable.
Though not reported, the results were qualitatively similar. Also, estimating the regressions using a
Fama-Macbeth based approach (Newey-West corrected) produced similar results.
40
Table 4: The Cross-Section of Corporate Cash-Holdings
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The table reports the results of multivariate regressions with
cash/assets as the dependent variable and a set of independent variables listed in the leftmost column
(see Table 2 for variable definitions). The independent variables are grouped into 4 liquidity models:
the first three are taken from previous literature, while the fourth model ties between diversification
and liquidity management. The table reports OLS Panel regressions (with a year fixed effect,
corrected for firm clustering). Standard errors are given in parenthesis: 3 asterisks denote 1 percent, 2
asterisks denote 5 percent and 1 asterisk denote 10 percent significance levels.

Diversification Model All 4 Models
Variable
Model 1
Model 2
(M/B)
Model 3
(AGR)
Model 4
(M/B)
Model 5
(M/B)
Model 6
(AGR)
-0.006 0.005 0.000
RE/assets
(0.01) (-0.01) (0.01)
0.017 0.012 0.015
Cash-flow/assets
(0.006)*** (0.007)*** (0.007)***
-0.038 -0.027 -0.030
Paid dividend?
(0.003)*** (0.004)*** (0.003)***
-0.293 -0.252 -0.264
NWC/assets
(0.010)*** (0.013)*** (0.009)***
-0.435 -0.305 -0.369
Leverage/assets
(0.013)*** (0.014)*** (0.008)***
0.009 0.009 0.070 Growth
Opportunities (0.004)** (0.004)** (0.004)***
-0.396 -0.430 -0.430 Capital
expenditure/assets (0.017)*** (0.026)*** (0.017)***
-0.152 -0.127 -0.279
Acquisition
(0.011)*** (0.013)*** (0.015)***
0.094 0.093 0.083 0.081 Growth
Correlation (0.007)***(0.007)*** (0.006)*** (0.006)***
0.081 0.084 0.074 0.079
CF correlation
(0.005)***(0.006)*** (0.004)*** (0.005)***
0.207 0.113 0.140 0.094
Growth volatility
(0.006)***(0.003)*** (0.005)*** (0.005)***
0.367 0.371 0.392 0.343 Cash Flow
volatility (0.081)*** (0.066)*** (0.086)*** (0.073)***
-0.011 -0.001 -0.001 -0.012 -0.001 -0.001
Num of segments
(0.001)*** (0.001) (0.001) (0.001)*** (0.001) (0.001)
-0.030 -0.030 -0.030 -0.005 -0.004 -0.007
Size
(0.001)*** (0.001)*** (0.001)*** (0.001)*** (0.001)*** (0.001)***
Adjusted R
2
0.1 0.13 0.13 0.42 0.47 0.46

41
link. One can think of other reasons beside cross-divisional coinsurance of
investment and cash flow. For instance, it might be that non-core business segments
serve as quasi-liquid entities that substitute cash in case demand for liquidity arises.
Models 2 and 3 try to answer whether coinsurance is responsible for the
correlation between number of segments and cash holdings. The results show that
the cross-divisional correlations between growth opportunities and the correlations
between cash flows absorb the effect of the number of business segments. In both
models the regression coefficient on the number of segments is not statistically
significant even at the 10 percent level. On the other hand, both growth and cash
flow correlations have positive coefficients which are statistically significant at the 1
percent level. One interpretation of this result is that diversification affects cash
holdings through its effect on cross-divisional correlations between investment
opportunities and cash flows. Note that these results hold in the presence of firm
size. This suggests that economies of scale in cash holdings are not the driving force
behind the results. The results also show that industry volatilities of investment
opportunity and cash flow are statistically significant at the 1 percent level. Finally,
note that the diversification models produce an R
2
of 0.13.
Having found support for the validity the diversification model in isolation, I
test how it performs when combined with the other three models, and report the
results in the last four columns of Table 4. First, note that the various variables
unrelated to diversification, excluding retained assets (RE/assets), are statistically
significant at the 1 percent level. Model 4 includes the number of reported business
segments, which comes out negative and statistically significant, and produces R
2

42
value of 0.42. Models 5 and 6 include the cross-divisional correlations and industry
volatilities, which once again come out positive and statistically significant, and
absorb the role of the number of business segments.
12
The regression coefficients are
also economically significant. For example, the regression coefficients in model 5
imply that an increase of one standard deviation in the market to book correlation
increases the cash holdings of the average firm by 10 percent, while an increase of
one standard deviation in the cash flow correlation increases the cash holdings of the
average firm by 12 percent. Furthermore, note that models 5 and 6 produce an R
2
of
0.46-0.47. Thus, the addition of correlations and volatilities to the regression model
significantly increases the overall explanatory power, from an R
2
of 0.42 to an R
2
of
0.46-0.47. This evidence supports the claim that the organizational form of the firm
(and in particular its degree of diversification) is a key determinant of its cash
balances.
E. Financial Constraints
The model emphasizes the role of market imperfections in shaping the cash
policy of firms. As Modigliani and Miller (1958) pointed out, in a frictionless world,
the choice of cash holdings does not affect firm value, since cash has a zero net
present value. Cash levels become significant only in the presence of frictions such
as incomplete contracting, agency costs of external funding, etc. To determine the
impact of financial constraints, I divide the sample into constrained and
unconstrained firms, based on short-term and long term credit rating. I adopt a

12
For brevity, Table III does not report the regression results with SGR as the measure of growth
opportunities. Using SGR does not change the results.
43
similar approach to that of Almeida, Campello, and Weisbach (2004), and retrieve
data on firms’ bond ratings and commercial paper ratings from Compustat. I
categorize firms that never had their public debt rated during the sample period as
financially constrained. Observations from those firms are only assigned to the
constrained sub-sample in years when the firms report positive debt.
Table 5 reports the number of constrained vs. unconstrained firm-year
observations according to the number of business segments. It shows that the above
categorizations classify many firms as financially constrained, regardless of their
number of segments. Furthermore, the short term credit rating classification finds
more firms to be financially constrained than the long term classification. While the
short term classification finds the majority of firms with 4+ segments to be
financially constrained, the long term classification finds the majority to be
unconstrained.
Table 5: Credit Constraints Classification
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. Based on commercial paper and bond ratings, the sample is divided
into short-term and long-term constrained and unconstrained firms. Firms that never had their public
debt rated during the sample period are classified as financially constrained. Observations from those
firms are only assigned to the constrained sub-sample in years when the firms report positive debt.
Firms that issued debt receiving ratings at some point during the sample period are considered
unconstrained.

Number of
Segments
ST Credit
Rating-
Unconstrained
ST Credit
Rating-
constrained
LT Credit
Rating-
Unconstrained
LT Credit
Rating-
constrained
1 13,265 40,069 18,525 34,809
2 1,594 5,959 2,538 5,015
3 1,082 3,584 1,979 2,687
4+ 1,167 2,297 1,800 1,664
44

Table 6 reports the results of the regression estimation in each subgroup of
firm-year observations categorized based on financial constraints. An examination
of the results shows that the coefficients on cash flow are more pronounced for
financially constrained firms. In fact, they are only statistically significant for
constrained firms. This result replicates the findings in Almeida, Campello, and
Weisbach (2004), who show that cash holdings are only sensitive to the cash flow
stream when the firm is financially constrained. A similar pattern emerges for
investment opportunities, measured by market to book. The rationale is similar:
firms will only adjust their cash-balances to the expected level of investment
opportunities when external financing is costly or infeasible.
The level of diversification, as measured by cross-divisional correlations,
also affects the stream of investment opportunities and cash flows. For the same
reason, we would expect it to have a stronger effect on cash holdings when the firm
is financially constrained. This is reflected in the results through the regression
coefficient on market to book correlation and cash flow correlation, which go up
from 0.017 to 0.082 and from 0.029 to 0.063, respectively, when one moves from
financially unconstrained to constrained firms based on short-term debt. Moreover,
the regression coefficients are statistically significant at the 1 percent for constrained
firms and at the 10 percent or 5 percent levels for unconstrained firms. As the table
shows, the same qualitative differences in the magnitude of regression coefficients
prevail for industry volatilities.

45
Table 6: Financial Constraints -- The Cross Section of Cash Holdings
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The table reports the results of multivariate regressions with
cash/assets as the dependent variable and a set of independent variables listed in the leftmost column
(see Table 2 for variable definitions). The regressions are estimated separately for financially
constrained and financially unconstrained firms, based on short term and long term credit ratings (see
Table 5 for classification definitions).The table reports OLS Panel regressions (with a year fixed
effect, corrected for firm clustering). Standard errors are given in parenthesis: 3 asterisks denote 1
percent, 2 asterisks denote 5 percent and 1 asterisks denote 10 percent significance levels.

Variable
ST Credit -
Unconstrained
ST Credit -
Constrained
LT Credit -
Unconstrained
LT Credit -
Constrained
0.070 0.031 0.061 0.029
RE/assets
(0.017)*** (0.008)*** (0.016)*** (0.008)***
0.012 0.151 0.037 0.186
Cash-flow/assets
(0.05) (0.024)*** (0.04) (0.026)***
-0.060 -0.021 -0.044 -0.017
Paid dividend?
(0.009)*** (0.003)*** (0.007)*** (0.004)***
-0.452 -0.183 -0.353 -0.183
NWC/assets
(0.025)*** (0.012)*** (0.023)*** (0.010)***
-0.342 -0.224 -0.346 -0.227
Leverage/assets
(0.024)*** (0.009)*** (0.020)*** (0.010)***
0.002 0.035 0.003 0.037
Growth Opportunities
(0.00) (0.003)*** (0.00) (0.003)***
-0.777 -0.292 -0.616 -0.310 Capital
expenditure/assets (0.048)*** (0.025)*** (0.045)*** (0.024)***
-0.251 -0.119 -0.197 -0.115
Acquisition /assets
(0.035)*** (0.011)*** (0.024)*** (0.013)***
0.017 0.082 0.013 0.075
M/B correlation
(0.014)* (0.005)*** (0.006)** (0.009)***
0.029 0.063 0.025 0.073
CF correlation
(0.049)* (0.050)*** (0.063)* (0.054)***
0.056 0.139 0.051 0.142
Industry M/B volatility
(0.004)*** (0.005)*** (0.003)*** (0.005)***
0.112 0.398 0.092 0.384
Industry CF volatility
(0.053)*** (0.084)*** (0.031)*** (0.083)***
-0.004 -0.002 -0.002 -0.002
Number of segments
(0.003) (0.001) (0.002) (0.002)
-0.019 -0.001 -0.016 0.002
Size
(0.003)*** (0.001) (0.002)*** (0.001)
Adjusted R
2
0.48 0.41 0.45 0.41

46
Chapter 4: Robustness
A. Subsamples
I now take a closer look at the behavior of cash holdings within subgroups of
firms that have the same number of business segments. So far, the results suggest
that the correlations between investment opportunities and cash flows of business
segments are an important driving force behind the variation of cash holdings across
firms with different numbers of reported business segments. Next, I want to examine
whether the variation of cross-divisional correlations among firms with a similar
number of business segments affects the cash holdings of these firms. To this end, I
divide the firms in the sample into four bins according to the number of reported
business segments and estimate the “all-inclusive” cash regressions from the
previous table. Table 7 reports the results of these regressions, concentrating on
market to book as the measure of investment opportunities (the results do not change
if I use AGR or SGR instead). As the table shows, cross-divisional correlations
between investment opportunities and between cash flows remain significant within
the clusters; hence they have explanatory power for the cross-sectional variation
across firms with equal numbers of reported business segments. Furthermore,
unreported results show that they add to the overall explanatory power of the
regression models, increasing the R
2
by approximately 5 percent relative to models
absent the diversification variables.
47
Table 7: Equal Number of Segments
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The table reports the results of multivariate regressions with
cash/assets as the dependent variable and a set of independent variables listed in the leftmost column
(see Table 2 for variable definitions). The regressions are estimated separately for each number of
business segments. The table reports OLS Panel regressions (with a year fixed effects, corrected for
firm clustering). Standard errors are given in parenthesis: 3 asterisks denote 1 percent, 2 asterisks
denote 5 percent and 1 asterisk denote 10 percent significance levels.

Variable 2 Segments 3 Segments 4+ Segments
0.021 0.038 0.040
RE/assets
(0.012)* (0.015)** (0.017)**
0.120 0.116 0.059
Cash-flow/assets
(0.044)*** (0.060)* (0.08)
-0.036 -0.020 -0.030
Paid dividend?
(0.007)*** (0.007)*** (0.011)***
-0.233 -0.217 -0.214
NWC/assets
(0.019)*** (0.024)*** (0.029)***
-0.313 -0.246 -0.235
Leverage/assets
(0.018)*** (0.025)*** (0.025)***
0.022 0.015 0.004
Market to book
(0.004)*** (0.006)** (0.001)**
-0.344 -0.318 -0.298 Capital
expenditure/assets (0.046)*** (0.051)*** (0.070)***
-0.133 -0.107 0.043
Acquisition/assets
(0.022)*** (0.022)*** (-0.06)
0.083 0.081 0.080
M/B correlation
(0.009)*** (0.008)*** (0.005)***
0.072 0.072 0.076
CF correlation
(0.007)*** (0.008)*** (0.007)***
0.133 0.088 0.075 Industry M/B
volatility (0.005)*** (0.005)*** (0.010)***
0.381 0.352 0.326 Industry CF
volatility (0.084)*** (0.074)*** (0.076)***
-0.003 -0.009 -0.009
Size
(0.001) (0.002)*** (0.003)***
Adjusted R
2
0.42 0.36 0.35

48
A different interpretation of the results can be sketched based on a sorting of
firms into young-growth firms and mature-established firms. Growth firms might
have greater cash holdings than mature firms in order to fund their profitable and
less certain investment opportunities. On the other hand, mature firms are generating
free cash flow, have more predictable investment opportunities, and are more prone
to agency problems. These factors might be driving their optimal cash holdings
downward. Thus, it is possible that “few segments” firms tend to have higher cash
balances due to a high incidence of growth firms in the “few segments” sorts.
Moreover, BKS document an increase in cash holdings for firms that do not
pay dividends, but find no time trend for firms that pay dividends. As dividend
payers tend to be mature, established firms (see e.g., DeAngelo, DeAngelo, and
Skinner, 2004), this suggests that the increase in cash holdings is mainly attributed
to young, growth firms. Combined with the findings of Fama and French (2001,
2004), who documented a massive shift in the population of publicly traded firms
towards young, growth firms, this might explain both the time trend and the cross-
section of cash holdings.
To address this issue, I construct a “constant composition” sample of firms
that are present in all sample years. This is a device for factoring out the influence of
imperfect controls for growth versus mature firm status and isolating the effect of
diversification on cash holdings. The first column of Table 8 presents the results of
estimating a cross-sectional regression on a constant composition sample between
1990 and 2004. The results are in line with previous findings, and show that cross-
divisional correlations continue to play a role in explaining the cross-section of
49
corporate cash holdings across firms that exist throughout all sample years.
However, the regression coefficients on correlations are smaller: the coefficient on
M/B correlation decreased from 0.083 to 0.063 and the coefficient on CF correlation
decreased from 0.074 to 0.045 (compared to Table 4). Adjusted R
2
also decreased
from 0.47 to 0.38, suggesting that the effects are somewhat lower in this subsample.
There are alternative clusters that can be examined to deal with the above
concerns. The second and third columns of Table 8 construct a sample of the Top
100 (200) dividends payers on Compustat (along the lines in DeAngelo, DeAngelo,
and Skinner, 2004). Once again, these are largely mature old line firms. The results
are qualitatively similar to those obtained with the “constant composition” sample,
and show that correlations are still significant, but their impact loses some of its
magnitude.

50
Table 8: Mature Firms
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The table reports the results of multivariate regressions with
cash/assets as the dependent variable and a set of independent variables listed in the leftmost column
(see Table 2 for variable definitions). The regressions are estimated separately for different sub-
samples. Top dividend payers are firms that pay out the largest amounts of cash dividends every year.
The constant composition sample only includes firms that are present in the sample all sample years.
The table reports OLS Panel regressions (with a year fixed effect, corrected for firm clustering).
Standard errors are given in parenthesis: 3 asterisks denote 1 percent, 2 asterisks denote 5 percent and
1 asterisks denote 10 percent significance levels.

Variable Constant Composition
Top 100 Dividend
Payers
Top 200 Dividend
Payers
0.060 -0.006 0.030
RE/assets
(0.027)** (0.02) (0.02)
0.109 0.030 0.039
Cash-flow/assets
(0.021)*** (0.11) (0.10)
-0.295 -0.271 -0.266
NWC/assets
(0.054)*** (0.080)*** (0.046)***
-0.324 -0.268 -0.272
Leverage/assets
(0.033)*** (0.051)*** (0.034)***
0.021 0.020 0.024
Market to book
(0.003)*** (0.005)*** (0.005)***
-0.528 -0.373 -0.426 Capital
expenditure/assets (0.094)*** (0.118)*** (0.085)***
-0.164 -0.061 -0.118
Acquisition/assets
(0.037)*** (0.06) (0.042)***
0.082 0.051 0.052
M/B correlation
(0.019)** (0.01)*** (0.01)***
0.063 0.060 0.062
CF correlation
(0.006)*** (0.005)*** (0.005)***
0.045 0.056 0.051 Industry M/B
volatility (0.019)** (0.024)** (0.016)**
0.522 0.701 0.496
Industry CF volatility
(0.277)*** (0.252)*** (0.186)***
-0.001 -0.002 0.000
Number of Segments
(0.00) (0.00) (0.00)
-0.014 -0.014 -0.010
Size
(0.004)*** (0.006)** (0.004)***
Adjusted R
2
0.38 0.37 0.35
51

As noted in the introduction, the Financial Accounting Standards Board
(FASB) issued a new segment reporting standard (SFAS 131) in 1997. The new
standard has been shown to have a positive effect on the amount and quality of
publicly available information about segments (e.g., Berger and Hann, 2003).
Therefore, it is important to examine whether the explanatory power of the segment-
based correlations and volatilities is affected by the change in reporting standards. If
the reporting quality of firms has improved due to the change, we would expect to
find a stronger relation between cash holdings and correlations/volatilities after the
change. Table 9 estimates the cash holdings multivariate regressions before and after
the 1997 change. The results document a significant increase in the magnitude of the
regression coefficients on cross-divisional correlations and volatilities, as well as an
overall improvement in the explanatory power of the model. In particular, the
coefficient on M/B correlation is almost four times larger after the change (0.081
after the change vs. 0.022 before the change). The correlation on CF correlation
went up from 0.063 to 0.078, while the coefficient on CF volatility went up from
0.129 to 0.695. Moreover, adjusted R
2
went up from 0.32 before the change to 0.45
after the change. This adds confidence to the main findings because it shows that a
more accurate disclosure of segment information leads to a stronger relation
between cash holdings and the risk-diversification measures.
52
Table 9: SFAS 131 (Change in Segments Reporting)
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The table reports the results of multivariate regressions with
cash/assets as the dependent variable and a set of independent variables listed in the leftmost column
(see Table 2 for variable definitions). The regressions are estimated separately for the period prior to
SFAS 131, which was issued in 1997 and changed the requirements for reporting business segments.
The table reports OLS Panel regressions (with a year fixed effect, corrected for firm clustering).
Standard errors are given in parenthesis: 3 asterisks denote 1 percent, 2 asterisks denote 5 percent and
1 asterisks denote 10 percent significance levels.
Variable Before Change (1990-1997) After Change (1998-2004)
0.045 0.026
RE/assets
(0.008)*** (0.010)***
0.017 0.028
Cash-flow/assets
(0.007)*** (0.003)***
-0.027 -0.032
Paid dividend?
(0.004)*** (0.005)***
-0.216 -0.235
NWC/assets
(0.012)*** (0.016)***
-0.257 -0.305
Leverage/assets
(0.011)*** (0.015)***
0.027 0.030
Growth Opportunities
(0.003)*** (0.003)***
-0.356 -0.393
Capital expenditure/assets
(0.028)*** (0.034)***
-0.121 -0.158
Acquisition /assets
(0.015)*** (0.017)***
0.022 0.081
M/B correlation
(0.007)** (0.010)***
0.063 0.078
CF correlation
(0.006)*** (0.008)***
0.064 0.062
Industry M/B volatility
(0.014)*** (0.010)***
0.129 0.695
Industry CF volatility
(0.10) (0.157)***
-0.004 0.000
Number of segments
(0.002)** (0.00)
-0.008 -0.004
Size
(0.001)*** (0.002)**
Adjusted R
2
0.32 0.45
53

B. Endogeneity
Another issue is the potential endogeneity of financial and investment
decisions. To address the endogeneity issue, I adopt an instrumental variables
approach using lagged variables as instruments, following Almeida, Campello, and
Weisbach (2004) and Fazzari and Petersen (1993). The set of instruments includes
lagged capital expenditure, lagged acquisitions, lagged net working capital, lagged
debt, lagged dividend payouts and lagged retained earnings. The first column of
Table 10 reports the results of the 2 Stage OLS regression with the above
instrumental variables. The coefficients on the cross-divisional correlations are
statistically significant and of magnitudes similar to those in previous regression
estimations.
Another concern arising from a review of Table 4 is that some of the cash
holdings are transitory because the firm might have raised funds that it is waiting to
spend. To acknowledge the possibility of transitory cash holdings, I follow Opler,
Pinkowitz, Stulz, and Williamson (1999) and add next year’s change in cash
holdings as an explanatory variable. As shown in the second column of Table 10,
introducing this variable does not affect the coefficients of cross-divisional
correlations.

54
Table 10: Instrumental Variables & Transitory Cash
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The table reports the results of multivariate regressions with
cash/assets as the dependent variable and a set of independent variables listed in the leftmost column
(see Table I for variable definitions). The Instrumental Variable (IV) regression is estimated using a
2SLS procedure, with lagged variables as instruments. The Transitory regressions add next year's
change in cash holdings as an independent variable, to account for transitory changes in cash-
holdings. The OLS Panel regression include a year fixed effect and a correction for firm clustering.
Standard errors are given in parenthesis: 3 asterisks denote 1 percent, 2 asterisks denote 5 percent and
1 asterisk denote 10 percent significance levels.

IV Transitory
Variable
2SLS OLS
0.055 0.018
RE/assets
(0.003)*** (0.003)***
0.073 0.073
Cash-flow/assets
(0.013)*** (0.012)**
-0.038 -0.006
Paid dividend?
(0.002)*** (0.001)***
-0.277 -0.084
NWC/assets
(0.006)*** (0.004)***
-0.293 -0.081
Leverage/assets
(0.006)*** (0.004)***
0.005 0.001
Market to book
(0.000)*** (0.001)*
-0.613 -0.093
Capital expenditure/assets
(0.021)*** (0.008)***
-0.156 -0.143
Acquisition /assets
(0.013)*** (0.012)***
0.084 0.089
M/B correlation
(0.003)*** (0.002)***
0.073 0.074
CF correlation
(0.002)*** (0.002)***
0.129 0.131
Industry M/B volatility
(0.005)*** (0.005)***
0.375 0.377
Industry CF volatility
(0.081)*** (0.084)***
-0.003 -0.001
Size
(0.001)*** (0.0001)***
0.783
Change in Next Year's Cash
(0.008)***
Adjusted R
2
0.46 0.77

55
The diversification discount literature has been recently criticized by a
number of studies (e.g. Campa and Kedia (2002), Chevalier (2004), Villalonga
(2004)) for not dealing properly with the endogenous nature of the diversification
decision. These studies argue that the diversification decision might be affected by
the same factors that drive firm value and cause the diversification discount. In such
a case, it is erroneous to make the inference that diversification leads to a discount.
Specifically, Chevalier (2004) shows that merging firms tend to trade in a discount
even before they merge. Campa and Kedia (2002) and Villalonga (2004) apply an
econometric approach that simultaneously estimates the decision to diversify and the
value of diversification, and find that the diversification discount disappears.
A similar critique might apply to this study. In particular, the endogenous
nature of the diversification decision might cause a similar inference problem as
above, namely that diversification and cash policy are affected by unobservable
factors that simultaneously determine both. If cash affects the diversification
decision or is affected by determinants of the decision to diversify, then inferring
that diversification drives cash holdings down might be incorrect.
The longitudinal analysis presented in this work follows firms that that
undergo an acquisition, and tracks their cash balances pre- and post-merger. Such an
analysis is in line with the approach taken by Chevalier (2004). The findings
indicate that while acquiring firms tend to hold positive excess cash balances before
the acquisition, they end up holding negative excess cash 3 years after the
acquisition. These findings remain intact after I take into account the method of
payment. The key results of the analysis are that the negative cash balances vary
56
across acquisitions with different degrees of diversification. The more diversifying
the acquisition, the less cash the firm holds after the merger. Thus, the connection
between cash holdings and the diversification decision does not conflict with the
coinsurance/precautionary cash holdings hypothesis of my work.
In this section, however, I utilize a different approach to factor out the
endogenous nature of the diversification decision. In particular, I consider a sub-
sample of firms that did not change their degree of diversification from 1998 to
2006. In such firms, the self-selection bias raised by the endogenous nature of the
diversification decision is mitigated by the fact that these firms did not make any
active diversification decision during the sample period.
Table 11 presents the results of estimating cross sectional regressions
predicting cash holdings in a sub-sample of firms that reported the same segments
on Compustat from 1998 to 2006.
13
For brevity, I only report the regression
coefficients on the variables of interest, namely the cross-divisional correlations and
the number of segments. The regression models are similar to those in Table 4. The
first two columns do not include control variables, while columns (3) and (4) include
the full set of controls employed in Table 4. The results indicate that cross-divisional
correlations are positively related to cash holdings, while the number of business
segments is only statistically significant when the correlations are excluded from the
analysis. These results are in complete accordance (both magnitudes and statistical
significance) with the main results presented in Table 4. Once again, they are

13
The sub-sample starts in 1998 to avoid potential reporting problems due to the change in the
accounting segment reporting standard that went into effect in 1997.
57
consistent with the hypothesis that diversified firms (with lower cross-divisional
correlations) hold less cash because coinsurance allows them to optimally reduce the
levels of precautionary cash holdings. As the Table shows, these findings continue
to hold in a sub-sample of firms that did not make any diversification decision
during the sample period, thus suggesting that the results are not likely to be driven
by a selection bias that arises due to the endogenous nature of the diversification
decision.
Table 11: Endogeneity - Stable Organization Form Firms
This Table includes a subsample of nonfinancial and nonutility firms that reported the same number
of business segments operating in the same industries for every year from 1998 to 2004. The table
reports the results of multivariate regressions with cash/assets as the dependent variable and a set of
independent variables (see Table 2 for variable definitions). For brevity, the table only reports the
coefficients on the variables of interest. The OLS Panel regression include year fixed effects and a
correction for firm clustering. Standard errors are given in parenthesis: 3 asterisks denote 1 percent, 2
asterisks denote 5 percent and 1 asterisk denote 10 percent significance levels.

Variable (1) (2) (3) (4)
Investment
Opportunity
Correlation

0.085***
(0.009)

0.076***
(0.008)
Cash Flow
Correlation

0.077***
(0.008)

0.066***
(0.007)
Number of
segments
-0.009***
(0.002)
-0.001
(0.001)
-0.010***
(0.002)
-0.001
(0.001)
R
2
0.08 0.12 0.39 0.45
Control Variables No No Yes Yes

The main findings thus far have shown that firms with higher degrees of
diversification (lower cross-divisional correlations) tend to hold less cash as a
58
fraction of their assets. I argue that these findings can be explained by a reduced
need of diversified companies to hold precautionary cash due to the coinsurance
benefit that reduces the ex-ante probability of adverse cash flow shocks or multiple
investment opportunities that require them to hold cash. It is optimal for such firms
to hold less cash because of the costs that accompany cash holdings, namely agency
and tax costs.
C. Agency
A competing hypothesis is that diversified (multi-segment) firms are mostly
mature firms where agency problems are more severe. In particular, managers of
multi-division firms might be more entrenched (e.g. Shleifer and Vishny (1989)). It
might also be the case that the diversification decision itself was driven by agency-
related or empire-building considerations (e.g. Grossman and Hart (1982), Morck,
Shleifer and Vishny (1990)), in which case diversification is a proxy for agency
problems. Given that there are agency costs to holding cash, the finding that
diversified firms hold less cash might be an equilibrium result where potential
agency costs, which are more severe in diversified firms, push their cash holdings
down. In other words, diversification may simple capture agency problems that
make cash holdings more costly for the firm because managers can spend cash on
non-value-maximizing activities. Such costs will drive cash holdings down, in
equilibrium.
To test whether diversified firms optimally hold less cash because of lower
precautionary needs rather than due to increased agency costs, I include a direct
59
measure of managerial entrenchment -- the GIM index (Gompers, Ishii and Metrick
(2003)). This index ranks firms based on 24 corporate bylaws and provisions that
protect minority shareholders and reduce managerial power. The index is available
from the IRRC database, and is available for approximately 1,500 firms included in
the S&P small-cap, med-cap and S&P 500 indexes. Since the data is only available
for 1990, 1993, 1995, 1998, 2000, 2002, 2004, 2006, I follow Gompers, Ishii and
Metrick (2003) and fill in the data for the missing years using the available data.
This leaves me with a much smaller sample of 2,153 firms, but allows me to directly
test the competing agency hypothesis.
Table 12: Agency - The GIM Index
This Table includes a subsample of nonfinancial and nonutility firms with data on the GIM
managerial-entrenchment index taken from the IRRC database. The table reports the results of
multivariate regressions with cash/assets as the dependent variable and a set of independent variables
(see Table 2 for variable definitions). For brevity, the table only reports the coefficients on the
variables of interest. The OLS Panel regressions include year fixed effects and a correction for firm
clustering. Standard errors are given in parenthesis: 3 asterisks denote 1 percent, 2 asterisks denote 5
percent and 1 asterisk denote 10 percent significance levels.

Variable (1) (2) (3) (4)
Investment
Opportunity
Correlation

0.071***
(0.009)

0.063***
(0.005)
Cash Flow
Correlation

0.063***
(0.006)

0.058***
(0.004)
GIM Index
Quintile
-0.023***
(0.004)
-0.018***
(0.002)
-0.017***
(0.002)
-0.015***
(0.002)
R
2
0.05 0.16 0.40 0.46
Control Variables No No Yes Yes

60
Table 12 reports the results of estimating cross-sectional regressions
predicting cash holdings similar to those reported in Table 4. To conserve space, I
only report the relevant regression coefficients on the cross-divisional correlations
and the GIM index. Columns (1) and (2) do not include any control variables, while
columns (3) and (4) include the full set of controls employed in Table 4.
The findings presented in Table 12 suggest that potential agency problems
indeed push firms towards lower cash balances. Keeping in mind that higher values
of the GIM index correspond to weaker minority shareholders protection, the
negative coefficient on the GIM index in all 4 specifications (columns (1) through
(4)) in Table 12 suggest that cash holdings are negatively correlated with agency
problems. This is consistent with the hypothesis that severe potential agency
problems reduce corporate cash holdings in equilibrium. Nevertheless, the main
takeaway from the table is that potential agency problems are not the driving force
behind the negative correlation between cash and diversification. As columns (2)
and (4) in Table 12 show, cross-divisional correlations continue to be positively
related to cash holdings. Thus, more diversified firms, with lower cross-divisional
correlations between cash flows and between investment opportunities tend to hold
less cash even after controlling for the severity of potential agency problems. These
findings combined with my findings that the ability to hold less cash has positive
value implications (see Table 14), suggest that diversification efficiently reduces
cash holdings from a shareholder value maximization point of view, rather than as a
byproduct of increased agency costs in diversified firms.
61
D. Firm Fixed Effects
As an additional robustness check, I examine whether changes in cross-
divisional correlations within a firm affect cash holdings. While this question can be
addressed using a longitudinal study, the drawback of the longitudinal study
employed in my work is that it is limited to a relatively small sub-sample.
Econometrically, I can examine how changes in the degree of diversification affect
the firm’s cash holdings by including firm fixed effects in my cross-sectional
regressions predicting cash holdings.
Table 13 reports the results of estimating cross-sectional regressions
predicting cash holdings similar to those reported in Table 4 with the addition of
firm fixed effects. To conserve space, I only report the relevant regression
coefficients on the cross-divisional correlations and the number of reported business
segments. Columns (1) and (2) do not include any control variables, while columns
(3) and (4) include the full set of controls employed in Table 4.
The results in Table 13 indicate that changes in the degree of diversification,
as measured by changes in the levels of cross-divisional correlations between
investment opportunities and between cash flows, are positively related to cash
holdings. When the correlation between divisional investment opportunities
increases, cash holdings decrease. The same is true for correlations between
divisional cash flows. The results are all statistically significant at the 1 percent
significance level, and the magnitudes are approximately cut in half relative to the
cross sectional results presented in Table 4 which does not include firm fixed
62
effects. Thus, the main takeaway from Table 13 is that changes in the degree of
diversification do affect cash holdings, as the coinsurance hypothesis of
precautionary cash holdings would suggest.
Table 13: Firm Fixed Effects
This table reports the results of OLS panel regressions similar to those in Table 4 with the addition of
firm fixed effects. For brevity, only the coefficients of the variables of interest are reported. Each
column reports coefficient estimates from a single regression. The dependent variable is the firm’s
cash holdings, defined as the ratio between cash and book assets. Significance levels are indicated: *
= 10%, ** = 5%, *** = 1%.

Variable (1) (2) (3) (4)
Investment
Opportunity
Correlation

0.038***
(0.006)

0.033***
(0.004)
Cash Flow
Correlation

0.029***
(0.005)

0.021***
(0.005)
Number of
segments
-0.004**
(0.002)
-0.001
(0.001)
-0.003***
(0.001)
-0.001
(0.001)
R
2
0.49 0.53 0.51 0.56
Control Variables No No Yes Yes

63
Chapter 5: Extensions

A. Value Implications

The results presented so far suggest that coinsurance allows diversified firms
to optimally hold less cash and save the costs that accompany cash holdings.
Therefore, holding everything else constant, the ability to hold less cash due to
cross-divisional coinsurance should positively affect firm value. This statement does
not translate into an overall statement about the value of diversification. Previous
theoretical works have highlighted various potential benefits (e.g. Stein (1997),
Matsusaka and Nanda (2002)) and costs (e.g. Scharfstein and Stein (2000), Rajan,
Servaes and Zingales (2000)) stemming mainly from investment
efficiency/inefficiency brought about by diversification. And while this work has no
predictions for the overall value of diversification, it suggests that the ability to hold
less cash due to diversification adds value to the firm.
The empirical findings regarding the overall value of diversification are
mixed. Earlier work by Lang and Stulz (1994), Berger and Ofek (1995) and others
suggested that diversified firms are discounted relative to their standalone
counterparts, thus implying that diversification destroys value overall. These
findings have been called into question by more the recent works of Campa and
Kedia (2002), Chevalier (2004), Villalonga (2004) and others, who argue that
previous empirical findings are contaminated by a selection bias. They claim that the
diversification decision is endogenous, and therefore the inference that
64
diversification destroys value is not necessarily correct. Consistent with this
observation, their findings suggest that the diversification discount disappears when
properly controlling for the endogeneity of the diversification decision.
The lesson from these studies is that one needs to explicitly recognize the
endogeneity of the diversification decision when estimating the value implications
of diversification. In what follows, I adopt a propensity score based approach to deal
with the selection bias (as in Villalonga (2004)), and proceed in two steps: (i)
estimate the excess values of multi-division firms using propensity score matching,
and (ii) test whether the excess value by the implied reduction in cash due to
coinsurance. I measure the reduction in cash holdings facilitated by diversification
based on the cross sectional regression model estimated in Table 4. The reduction in
cash brought about by diversification in firm i is given by firm i’s cross-divisional
correlation multiplied by the estimated regression coefficient on the correlation.
Thus, based on Table 4, the implied reduction in cash equals 0.083*Q_Corr
i
+
0.074*CF_Corr
i
, where Q_CORR is the cross-divisional correlation between
investment opportunities as measured by Tobin’s Q, and CF_Corr is the cross-
divisional correlation between cash-flows (see Table 4, model 4).
Let us start by briefly describing how excess values are measured. As noted
above, the estimation procedure follows Villalonga (2004). Excess values are
measured in three different ways. First, following Berger and Ofek (1995), I
compute excess values based on asset or sales multipliers as the ratio of the firm’s
actual value to its imputed value. A firm’s imputed value is the sum of the imputed
values of its segments, where a segment’s imputed value is equal to the segment’s
65
assets (sales) multiplied by the average ratio of market value to assets (sales) of
standalone matched firms. Second, following Lang and Stulz (1994), excess value is
computed as the ratio between Tobin’s Q and the imputed Q. The imputed Q is the
asset-weighted average of the hypothetical Qs of the firm’s segments, where a
segment’s hypothetical Q is the average of the standalone matched firms.
Next, I turn to describe the method applied to match between multi-division
and standalone firms based on propensity scores. In essence, propensity score
matching is a two-stage procedure. The first stage estimates firms’ propensity to
diversify based on firm and industry characteristics. The second stage matches
between standalone and diversified firms based on the fitted values of the first stage
regression, also known as propensity scores.
A firm’s propensity to diversify is modeled as a function of the
characteristics of the firm and its industry and is a cross-sectional probability of a
firm to be diversified as in Campa and Kedia (2002). The firm characteristics I use
include firm size, profitability, investment, cash holdings; and dummies that indicate
whether the firm is listed on a major exchange (NYSE, AMEX, or Nasdaq) and
whether it paid dividends. In addition, I include the percentage of outstanding
common stock owned by institutions and insiders, and the firm’s age measured by
the number of years listed in CRSP. The industry characteristics that I use include
the average Tobin’s Q, the percentage of diversified firms in the industry, and the
percentage of sales in the industry accounted for by diversified firms.
I use the probit model to compute propensity scores – the predicted values
from the model. The propensity score matching procedure I use follows Abadie and
66
Imben (2002). Each diversified firm is matched with four standalone firms based on
their propensity scores, where the matching is done with replacement to reduce the
asymptotic bias. The final step involves computing the three excess value measures
outlined above, using the four matched standalone firms for each diversified firm.
Using the excess values from the above procedure, Table 14 reports the
results of estimating cross-sectional regressions of diversified firms with excess
value as the dependent variable. The table has three columns, corresponding to the
three different measures of excess values. The independent variable of interest is the
coinsurance cash reduction, estimated as the implied reduction in cash due to the
firm’s cross-divisional correlations based on the results of the regression model
reported in Table 4, Model 4. The regression model also includes firm size, EBIT,
capital expenditure and leverage as additional control variables.
Let us start with the first column, which reports the results when excess
values are measured using asset multipliers. As the results indicate, the ability to
hold less cash due to lower correlations has a positive effect on the firm’s excess
value. In particular, the regression coefficient implies that a one standard deviation
decrease in the cross-divisional correlations between investment opportunities and
between cash flows, which allows the average firm to hold approximately 23 percent
less cash, improves average excess value by approximately 25 basis points. This
effect is statistically significant at the 5 percent significance level, and is consistent
with the hypothesis that coinsurance allows firms to optimally economize on the
costs associated with holding cash. The results are virtually identical in the next two
columns, where excess values are measured based on sales multipliers or Tobin’s Q.
67
Furthermore, the regression coefficients are statistically significant at the 1 percent
level in both specifications.
Table 14: Value Implications
Each column reports coefficient estimates from a single regression, with standard errors (robust and
clustered by industry) in parentheses. The dependent variable is the firm’s excess value. Each column
corresponds to a different measure of excess value, as indicated at the top of each column. Asset
(Sales) Multiplier is defined as the ratio of the firm’s actual value to its imputed value, where a firm’s
imputed value is the sum of the imputed values of its segments, and a segment’s imputed value is
equal to the segment’s Assets (Sales) multiplied by the average ratio of market value to Assets
(Sales) of standalone matched firms. The Tobin’s Q-based excess value is computed as the ratio
between Tobin’s Q and the imputed Q. The imputed Q is the asset-weighted average of the
hypothetical Qs of the firm’s segments, where a segment’s hypothetical Q is the average of the
standalone matched firms. Coinsurance cash reduction is the estimated coefficient on cross-divisional
correlations estimated in Model 4 of Table 4, multiples by the firm’s correlation value. Significance
levels are indicated: * = 10%, ** = 5%, *** = 1%.

Asset Multiplier Sales Multiplier Tobin’s Q

0.004** 0.003*** 0.004*** Coinsurance cash
reduction (0.002) (0.001) (0.001)
0.169*** 0.160*** 0.197***
Size
(0.042) (0.039) (0.044)
0.065** 0.042** 0.049**
EBIT
(0.026) (0.019) (0.022)
0.029* 0.071** 0.053*
Capital Expenditure
(0.016) (0.032) (0.031)
-0.061** -0.058** -0.059**
Leverage
(0.025) (0.023) (0.025)
2
R
0.143 0.157 0.184

These results imply that coinsurance has a positive effect on firm value
through facilitating a more efficient cash policy that allows the firm to save costs
associated with holding cash. The above analysis does not have any implications for
68
the overall efficiency of diversification, but rather serves as a conservative
estimation of the coinsurance-cash implied benefit taking into account the
endogeneity of the diversification decision. This evidence is consistent with the
efficient, value-maximizing cash-diversification model presented here, and is
inconsistent with the notion that cash holdings of diversified firms are driven by
higher agency costs. An agency-based model of cash holdings does not predict a
positive value effect for diversification.
B. Net Debt
As noted in the introduction, coinsurance between divisions of diversified
firms was introduced by Lewellen (1971), who suggested that cash flow coinsurance
can serve as a purely financial rationale for corporate diversification by reducing the
probability of financial distress. While this implies that diversified firms should hold
more debt than specialized firms, subsequent research did not find a substantial
coinsurance effect on corporate debt (see e.g. Table 2 and section 5.3 in Berger and
Ofek, 1995).
However, equally interesting is the correlation between corporate
diversification and net debt, defined as debt minus cash. Net debt is a more relevant
measure of leverage in the context of coinsurance, since the portion of debt covered
by cash holdings is less exposed to default risk. Table 15 breaks the firm-year
observations in the sample into four clusters based on their reported number of
business segments, and records the average net-debt-to-assets ratio from 1990 to
2004. The table reveals interesting patterns both in the cross section and the time
69
series dimensions. In the cross section, firms with more business segments hold
higher levels of net debt. Moving from firms with a single business segment to firms
with two business segments, net debt increases by more than a 100 percent on
average. This finding lends support to the predicted effect of coinsurance on net debt
balances, and is significantly higher than the documented effect in previous studies
(Berger and Ofek document an effect of only 1.4 percent).
The paper by BKS documents a dramatic drop in corporate net debt from
1980 to 2004 due to growing cash holdings. In fact, the average US firm had no
leverage, as measured by net debt, in 2004. In accordance with their findings, Table
15 shows that net debt has decreased substantially from 1990 to 2004 in all four
business segments clusters. However, Table 15 shows that even in 2004, when
average net debt was negative according to BKS, diversified firms with 3 business
segments or more still had positive net debt balances. This implies that highly
diversified firms still have positive leverage and cannot cover their overall debt
using the cash at their disposal.
Table 15 shows that in every year from 1990 to 2004, diversified firms had
larger net debt balances than standalone firms. It verifies that this pattern persists in
every year during our sample, and in addition highlights the time-series trend of
decreasing net debt balances. The table also shows that even in 2004, diversified
firms did not have negative net debt balances. However, the analysis in Table 15
does not control for other factors known to be correlated with debt balances. In
addition, it does not utilize the direct measure of cross divisional correlation
70
between cash flows, which is the driving force behind the connection between
diversification and debt according to the coinsurance hypothesis.
Table 15: The Distribution of Average Net Debt
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. Net Debt is short-term debt plus long-term debt minus cash divided
by total book assets.

Number of Business Segments
Year
1 2 3 4+
1990 0.190 0.240 0.247 0.268
1991 0.151 0.209 0.236 0.238
1992 0.134 0.213 0.224 0.243
1993 0.085 0.186 0.211 0.226
1994 0.095 0.192 0.196 0.209
1995 0.093 0.177 0.206 0.211
1996 0.053 0.176 0.196 0.208
1997 0.058 0.164 0.180 0.201
1998 0.063 0.153 0.176 0.252
1999 0.035 0.094 0.130 0.207
2000 0.052 0.061 0.154 0.158
2001 0.034 0.052 0.118 0.160
2002 0.018 0.049 0.073 0.145
2003 -0.020 -0.003 0.046 0.123
2004 -0.053 -0.002 0.005 0.080

The next table attempts to overcome both shortcomings by cross-tabulating
average net-debt balances across firms with different numbers of business segments
or levels of cross-divisional cash flow correlations, controlling for firm size. Panel A
compares between the average net debt/assets ratios of firms that reported 1, 2, 3, or
71
4+ business segments, controlling for the size quartile (as measured by book assets).
The main finding in Table 16 is that across all four size quartiles, firms with more
business segments tend to hold more net debt on average. The magnitude of the
effects is impressive. For example, firms in the second size quartile have a net debt
to assets ratio of 0.095 when they report two business segments, and a ratio of 0.170
when they report three segments. Furthermore, all the differences are statistically
significant at the 1 percent level. The table also demonstrates that across the panel,
only the smallest standalone firms actually had an average negative net debt balance
from 1990 to 2004.
Panel B of Table 16 specifically considers the correlation between divisional
cash flows, and reports how average net debt balances vary with the correlation rank
(low, medium, high). The table controls for the number of reported business
segments and the size of the firm, as measured by book assets. The results show that
the correlations are the ones driving net debt balances. Holding the number of
business segments and size fixed, net debt balances decrease when firms have higher
cross-divisional cash flow correlations, i.e. when they are less diversified. Once
again, all the differences are statistically significant at the 1 percent level and have a
sizeable effect. For instance, net debt to assets ratio of medium-size firms with 3
business segments decreases from 0.213 when correlation is low to 0.150 when
correlation is high.

72
Table 16: Cross Tabulation of Average Net Debt
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. Size is the natural logarithm of the book value of total assets in 2004
dollars. Three proxies are used to measure growth: Market-to-book ratios, AGR (asset growth rate,
excluding cash) and SGR (sales growth rate). The asset growth rate (AGR) is the change in total
assets, excluding cash, divided by the previous year’s level, while the sales growth rate (SGR) is
defined analogously with respect to revenue. Market to book equals book value of total assets minus
book value of equity plus market value of equity divided by total assets Cash-flow/assets is defined
as earnings less interest and taxes, over book assets. M/B correlation is defined as the sales-weighted
average correlation between mean market to book ratios over 15 years of all standalone firms in each
division's industry, as defined by the 3-digit NAICS code. AGR correlation, SGR correlation and CF
correlation are defined analogously with respect to AGR, SGR and operating cash-flows,
respectively. Industry M/B volatility is the sales-weighted average standard deviation of mean market
to book ratios over 15 years of all standalone firms in each division's industry, as defined by the 3-
digit NAICS code. Industry AGR volatility, SGR volatility and CF volatility are defined analogously
with respect to AGR, SGR and operating cash-flows, respectively.

Panel A: Average Net Debt/Assets by Size and Number of Segments
Number of Segments
Size Quartile
1 2 3 4+
1 -0.008 0.011 0.153 0.225
2 0.072 0.095 0.170 0.219
3 0.046 0.073 0.184 0.238
4 0.130 0.143 0.203 0.219

Panel B: Average Net Debt/Assets by Number of Segments, Size and CF Correlation
Rank
Number of
Segments
Size
Low CF
Correlation
Medium CF
Correlation
High CF
Correlation
Small
0.132 0.127 0.035
Medium
0.210 0.196 0.134
2
Large
0.217 0.197 0.176
Small
0.109 0.080 0.011
Medium
0.213 0.198 0.150
3
Large
0.281 0.231 0.218
Small
0.195 0.190 0.033
Medium
0.255 0.234 0.171
4+
Large
0.223 0.203 0.143
73

Similar to the cross-sectional analysis of cash holdings, the next step would
be to estimate cross-sectional regressions predicting net debt. This approach allows
us to introduce various control variables previously documented to explain the cross
section of both cash and debt levels, and establish whether the results are statistically
significant. Table 17 reports these results, and utilizes measures of cash flows,
dividends, investment opportunities, firm size, capital expenditure, acquisitions, firm
size and the number of business segments as control variables. The first column does
not include the cross-divisional correlations and shows that indeed the number of
business segments is positively related to the levels of net debt. Columns 2-4 include
the cross-divisional cash flow correlation measure and correspond to the three
measures of growth opportunities (Tobin’s Q, AGR, SGR). Across all three
measures, the findings show that cash flow correlations are negatively related to net
debt balances: firms with higher correlations, which do not enjoy coinsurance as
much and therefore are more exposed to bankruptcy costs, tend to hold less net debt.
The results in Tables 15, 16, and 17 suggest that higher degrees of corporate
diversification, measured as lower correlations between divisional cash flows,
correspond to larger net debt balances. These findings are once again consistent with
coinsurance: lower cross-divisional correlations reduce the ex-ante probability of
distress, and therefore increase the optimal net debt levels of firms trading off the
tax benefits of debt with the deadweight costs of financial distress.
74
Table 17: The Cross-Section of Net Debt
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The table reports the results of multivariate regressions with net debt
as the dependent variable and a set of independent variables listed in the leftmost column (see Table 2
for variable definitions). The table reports OLS Panel regressions (with a year fixed effect, corrected
for firm clustering). Standard errors are given in parenthesis. Significance levels are indicated: * =
10%, ** = 5%, *** = 1%.

Number of
Segments Q AGR SGR

(1) (2) (3) (4)
-0.105*** -0.105*** -0.084*** -0.099***
Cash-flow/assets
(0.018) (0.018) (0.017) (0.019)
-0.073*** -0.075*** -0.098*** -0.094***
Paid dividend?
(0.007) (0.007) (0.007) (0.007)
-0.032*** -0.032*** -0.085*** -0.047** Growth
Opportunities
(0.004) (0.004) (0.012) (0.021)
0.383*** 0.384*** 0.316*** 0.29*** Capital
Expenditure/assets (0.039) (0.039) (0.040) (0.041)
0.575*** 0.575*** 0.724*** 0.591*** Acquisition
/assets
(0.023) (0.023) (0.035) (0.027)
-0.028*** -0.026*** -0.030*** Cash Flow
correlation

(0.005) (0.001) (0.004)
0.03*** 0.031*** 0.034*** 0.033***
Size
(0.002) (0.002) (0.002) (0.002)
0.013*** 0.011*** 0.005* 0.007* Number of
Segments
(0.003) (0.003) (0.003) (0.005)
2
R
0.273 0.283 0.281 0.279

The idea that coinsurance should lead firms to optimally hold more debt is
not new, and has been introduced in early works such as Lewellen (1971) and
75
Weston and Mansinghka (1971). However, subsequent empirical work (e.g. Berger
and Ofek (1995)) did not find sustainable evidence suggesting that indeed
diversified firms hold more debt. The above findings offer a possible bridge between
the theoretical implications of coinsurance and the empirical reality. It might be the
case that while diversified firms do not hold more debt, they hold more net debt,
possibly because it is the portion of debt not covered by cash that is most susceptible
to bankruptcy costs.
Therefore, it might be the case that firms focus on their net debt balances
rather than their debt balances, and adjust the former rather than the latter to their
degree of diversification. Such findings provide a possible explanation between
capital structure policy and diversification, which previous empirical studied did not
find. Nevertheless, it is important to emphasize that these findings are not overall
surprising given the previous evidence presented in my work showing that
diversified firms tend to hold less cash. Given that diversified firms tend to hold less
cash, and that previous empirical works do not document any significant cross-
sectional variation in debt levels based on the degree of diversification, the finding
that diversified firms hold more net debt is almost mechanical. Put differently, the
cross-sectional differences in net debt levels is mostly driven by the variation in cash
holdings, given the documented little variation in debt levels. Thus, my findings that
diversified firms hold more net debt should not be taken as a new set of results, but
rather as a different way to interpret the existing results. Inverting the lenses and
focusing on debt ratios, one can interpret the cross sectional variation in cash levels
within the framework of a static tradeoff theory where firms weight the tax benefits
76
of debt against the expected bankruptcy costs it entails. Higher degrees of
diversification (i.e. lower cross-divisional correlations) reduce the expected
bankruptcy costs and increase the optimal net debt position of the firm.
C. The Time Trend of Cash Holdings

The evidence presented thus far supports the main implications of the
model, namely that diversification and low degrees of cross-divisional correlations
correspond to lower ratios of cash to total assets. However, this does not explain
why diversified firms tend to hold more cash than they used to. As Figure 1 shows,
the average Cash/assets ratio has increased dramatically from 1990 to 2004 in
diversified firms (with two business segments or more). BKS show that increasing
cash flow volatility is one of the main driving forces behind the increase in cash
balances. Indeed, this is consistent with a number of recent studies (e.g., Irvine and
pontiff, 2005) that document a market-wide increase in idiosyncratic risk and cash
flow volatility. Thus, the precautionary demand of firms for cash reserves is greater
due to increasing levels of business risks that cannot be avoided or hedged.
This study argues that corporate diversification can serve as means to reduce
the firm’s risk exposure through coinsurance across divisions. By diversifying into
unrelated industries, with low investment opportunity and cash flow correlations
between them, firms can reduce their exposure to investment and cash flow risks
substantially. This, in turn, will allow them to hold lower cash balances and save on
liquidity costs. Thus, it is of interest to examine whether coinsurance levels have
changed by tracking average firm correlations over the same period of time in which
77
industry volatilities have increased. Figure 2 plots average firm correlations from
1990 to 2004. As Panel A shows, the average correlations between investment
opportunities (measured by market to book ratios) have increased dramatically
between 1990 and 2004. In fact, average correlation has increased from 0.54 to 0.76
in two-segment firms, from 0.47 to 0.75 in three-segment firms, and from 0.5 to 0.72
in firms with four segments or more. Panel B documents a similar trend in cash flow
correlations: average correlation has increased from 0.56 to 0.81 in two-segment
firms, from 0.51 to 0.78 in three-segment firms, and from 0.53 to 0.75 in firms with
four segments or more.
Figure 2: Average Firm Cross-Divisional Correlations
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The figure reports the average investment opportunities and cash
flow correlations for firms with 2, 3, and 4 or more reported business segments.

Panel A: Average Investment Opportunities (Measured by M/B Ratios) Correlations
.5 .6 .7 .8
A verage C orrelations

1990 1992 1994 1996 1998 2000 2002 2004
Year
2 Segments
3 Segments
4+ Segments

78
Figure 2, Continued

Panel B: Average Cash Flow Correlations

.3 .4 .5 .6 .7 .8
Average Correlations

1990 1992 1994 1996 1998 2000 2002 2004
Year
2 Segments
3 Segments
4+ Segments

The findings in the previous paragraph reveal an interesting pattern. The
increase in cash flow volatility documented in previous studies was accompanied by
an increase in investment opportunities and cash flow correlations of multi-division
firms. This might explain why we observe greater cash-balances in diversified firms.
On average, diversified firms have greater investment opportunities and cash flow
correlations than in the past, and therefore do not enjoy the benefits of
investment/cash flow coinsurance as much as before. These results are inline with
the 1990s trend of an increasing percentage of same-industry mergers documented
by Andrade, Mitchell, and Stafford (2001). Thus, these firms are more exposed to
79
investment and cash flow risks, and have a stronger precautionary motive for
holding cash.
However, there is an important distinction between the increase in industry
volatilities and the increase in firm correlations. As their names suggest, the former
are a market-wide phenomenon, over which the firm has no control, while the latter
is a choice variable of the firm. Firms cannot control the level of industry risk, but
they can control the level of correlation by choosing their organizational structure.
This last statement hinges on the availability of uncorrelated industries; if the
increase in industry risk were accompanied by a similar increase in cross-industry
correlations, then multi-division firms would not be able to use diversification and
coinsurance effectively to reduce their risk exposure.
Table 18 compares between average cross-industry correlations and average
industry volatilities of investment opportunities and cash flows.
14
The first two
columns record average cross-industry correlations and reveal that the average
market to book correlation has decreased from 0.191 to 0.121 between 1990 and
2004, while the average cash flow correlation has decreased from 0.161 to 0.115
over the same period. This implies that multi-division firms could still diversify into
uncorrelated industries in 2004 and coinsure their investment opportunities and cash
flows streams. However, as the last two columns of Table 18 indicate, this is not the
case with volatilities. Average industry market to book volatility has increased from
0.354 to 0.615 between 1990 and 2004, while average industry cash flow volatility

14
Both cross-industry correlations and industry standard deviations are calculated using the average
standalone firm within each industry. For a given year, I record the time series of the average
standalone firm’s cash flows and market to book over the past 15 years within each industry. These
are used to calculate average standard deviations and correlations across all industries.
80
has from 0.038 to 0.049. These findings suggest that firms could still diversify into
low-correlated industries, but instead chose to operate in highly correlated industries
and forgo the benefits of coinsurance, and in particular the ability to hold lower cash
balances.
Table 18: Average Industry Correlations and Volatilities
Average industry market to book correlation is defined as the average pair-wise correlation between
industries' mean market to book ratios over 15 years of all standalone firms. Average industry cash-
flow correlation is defined analogously with respect to cash-flows. Average Industry market to book
volatility is defined as the average standard deviation of mean market to book ratios over 15 years of
all standalone firms in each industry. Industry cash-flow volatility is defined analogously with respect
to cash-flow.

year
Avg. Industry
market to book
correlation
Avg. Industry
cash-flow
Correlation
Avg. Industry
market to book
volatility
Avg. Industry
cash-flow
volatility
1990 0.191 0.161 0.354 0.038
1991 0.214 0.155 0.356 0.041
1992 0.187 0.099 0.361 0.042
1993 0.188 0.053 0.365 0.041
1994 0.253 0.044 0.370 0.040
1995 0.248 0.040 0.371 0.040
1996 0.273 0.040 0.452 0.039
1997 0.296 0.039 0.460 0.039
1998 0.288 0.048 0.467 0.040
1999 0.275 0.077 0.499 0.043
2000 0.208 0.089 0.604 0.045
2001 0.190 0.103 0.613 0.048
2002 0.165 0.116 0.643 0.051
2003 0.188 0.113 0.635 0.051
2004 0.121 0.115 0.615 0.049

81
D. Investment-Cash Sensitivity

The main findings thus far suggest that multi-division firms with lower
cross-divisional correlations between cash-flows and lower cross-divisional
correlations between investment opportunities optimally hold a reduced buffer of
cash at the firm level. These results are interpreted as consistent with a coinsurance
benefit which allows firms to economize on the costs associated with holding cash
without giving up the benefits they entail, namely the ability to undertake valuable
investment opportunities even when (i) opportunities are abundant; (ii) cash-flows
cannot fully fund all investments. Thus, coinsurance serves a substitute mechanism
for the precautionary and/or speculative motives of holding cash, and allows the
firms to avoid potential underinvestment costs by efficiently transferring internally
generated cash flows across divisions to fund investments rather than relying on
cash balances to fund those investments. Put differently, coinsurance reduces
underinvestment risk and can therefore be viewed as an alternative to precautionary
cash holdings.
The above interpretation implies higher degrees of diversification and
coinsurance allow firms to rely less on cash in financing their investments.
Therefore, one would expect the correlation between cash and investment to be
lower in more diversified firms. Such firms fund their investments more often using
internally generated cash flows, and are therefore less likely to use their cash
balances to fund investments. Put together, this implies that the dependence (or
sensitivity) of investment on cash is lower in more diversified firms. If firms build
82
cash balances to mitigate possible underinvestment problems, we would expect them
to do less so when they are more diversified. And if firms spend cash on
investments, then we would expect them to do less so when they are more
diversified. This line of reasoning translates into two testable hypotheses:
A1: Lagged cash balances should be positively related to future investments.
However, the correlation should be weaker in more diversified firms.
A2: Contemporaneous cash balances should be negatively related to
investments. However, this correlation should be less negative in more
diversified firms.
Table 19 tests the above hypothesis in a regression a panel regression
framework. The dependent variable in the regression is the firm’s investment level
at time t. The independent variable in the regression is the firm’s cash level at time t-
1 and its cash level at time t. The regressions are estimated separately for firms with
high degrees of diversification and for firms with low degrees of diversification. The
degree of diversification is measured by cross-divisional correlation between
investment opportunities, where multi-division firms are sorted on cross-divisional
correlations into three quantiles. The bottom quantile corresponds to highly
diversified firms, while the top quantile corresponds to firms that are the least
diversified. The regression specification controls for other factors correlated with
investment, namely sales growth, cash flow and Tobin’s Q (see e.g. Shin and Stulz
(1998)).
Let us start with the first column, which corresponds to firms that are highly
diversified, with low cross-divisional correlations. As the table shows, time t cash
83
holdings are negatively related to time t investment, while time t-1 cash holdings are
positively related to time t investment. Both coefficients are significant at the 5
percent significance level, and suggest that even highly diversified firms hoard cash
to finance investment and end up spending cash to fund investment. These findings
are consistent with precautionary motives for holding cash. However, the question
remains how highly diversified firms compare to firms that are less diversified. The
next column answers this question by showing that investment is more sensitive to
cash holdings in less diversified firms with higher cross-divisional correlations. The
second column reports a regression coefficient of -0.095 on contemporaneous cash
(compare to a coefficient of -0.064 for more diversified firms) and a coefficient of
0.065 on lagged cash (compare to a coefficient of 0.012 for more diversified firms).
Both coefficients are significant at the 1 percent significance level (compare to a 5
percent significance level in more diversified firms). The third column reports that
the differences between investment sensitivity to contemporaneous as well as lagged
cash in more vs. less diversified firms are significant at the 1 percent significance
level. Thus, the table is consistent with both hypotheses A1 and A2. Investment is
negatively related to contemporaneous cash holdings but more so in less diversified
firms. These firms end up spending more cash on investment, while highly
diversified firms rely more often on internally generated cash flows to finance
investment. Investment is positively related to lag cash balances but more so in less
diversified firms. These firms have to build cash reserves taking into account future
investment needs because they are less likely to be able to use internally generated
cash flows to avoid potential underinvestment problems.
84
Table 19: Cash Sensitivity of Investment
For each year over 1990-2004, the sample consists of nonfinancial and nonutility firms in the
intersection between Compustat's Industrial annual file and Compustat's segments file, with non-
missing data on cash holdings and the industry codes of each business segment, and with total market
capitalization of $10M or more. The dependent variable is the firm’s investment level, measured by
capital expenditure divided by book assets. The firs two columns estimate multivariate OLS panel
regressions for firms with low and high cross-divisional correlations between investment
opportunities separately. The third column reports the differences between the regression coefficients
in the two subsamples. Significance levels are indicated: * = 10%, ** = 5%, *** = 1%.

Low Q
Correlation
High Q
Correlation
Diff (Low-High)

0.012*** 0.009** 0.003
Sales Growth
(0.004) (0.004) (0.003)
0.016*** 0.011*** 0.005**
Cash Flow
(0.001) (0.001) (0.002)
0.003*** 0.003*** 0.000
Tobin’s Q
(0.005) (0.003) (0.001)
-0.064** -0.095*** 0.031***
Cash
(0.031) (0.003) (0.005)
0.012** 0.065*** -0.053***
Cash
t-1
(0.051) (0.003) (0.007)
2
R
0.235 0.242

E. Longitudinal Study

A negative average cross-sectional connection between diversification and
cash holdings is not evidence per se that diversified firms hold less cash due to a
greater coinsurance benefit. For diversification to be interpreted as driving firms to
hold less cash, diversified firms must have reduced their cash holdings by engaging
in diversification, or at least be holding less cash by staying diversified. Therefore,
85
one needs to look at changes in diversification status, which requires the usage of a
longitudinal approach. A similar approach is taken by papers such as Graham et al.
(2002) and Hyland and Diltz (2002) that study the effect of diversification on firm
value.
Similar to these studies, I focus on those firms that change their
diversification status from single-segment to diversified (multi-segment). There are
two reasons for this selection: first, the diversification discount literature usually
defines diversification as a multi-segment dummy; and second, Lang and Stulz
(1994) and subsequent studies find that the discount is significant between one- and
two-segment firms, but not between two-segment and firms with larger numbers of
segments. These results are consistent with my findings that most of the reduction in
cross-divisional correlation occurs between standalone and two-segment firms. The
marginal reduction in correlations moving to firms with three or more segments is
relatively smaller.
Graham et al. (2002) and Hyland and Diltz (2002) show that the longitudinal
analysis may be contaminated by segment reporting changes in Compustat. In
particular, as noted earlier, there has been a change in the accounting reporting
standard of segments in 1997 (SFAS 131). In an attempt to avoid misclassifications
of reporting changes as diversification changes, I take two preventive measures.
First, I focus on the post 1997 period, which has been shown by previous studies to
be characterized by better segments reporting. Second, I obtain data on mergers and
acquisition from SDC, and only include acquirers that were reported as single-
86
segment firms on Compustat prior to the acquisition, and that went through a
completed acquisition according to SDC.
The goal of the longitudinal approach is to identify and estimate changes in
cash holdings brought about by acquisitions that change the degree of diversification
in the firm. Given that acquisition activity might generate short-term (transitory)
fluctuations in cash holdings due to the costs associated with an acquisition, I allow
the firms to re-adjust their cash balances in the three years following the merger, and
consider the cash balances the firm reports on its balance sheet after three years.
Therefore, my sample includes mergers that took place from 1998 to 2003, allowing
me to observe cash balances in 2006, which is the most recent year for which data is
available. This requires me to eliminate from the sample firms that go through
another acquisitions in the three years that follow the acquisition. Furthermore, I
make a distinction between acquisitions that were paid with cash and acquisitions
that were not paid with cash because the former might hold less cash following a
merger simply because they used their cash to make the acquisition.
An additional source of concern is the economy-wide upward trend in cash
holdings documented previously. Since this study compares between cash balances
before an acquisition and three years after the acquisition, one needs to eliminate the
time trend. To this end, I report “excess” cash balances, defined as the difference
between the firm’s cash holdings and the cash holdings of the average non-acquiring
firm in the same industry and the same size decile.
Finally, I distinguish between acquisitions based on their level of
diversification, as measured by the cross-industry correlation between average
87
industry investment stream and average industry cash flow stream. I sort all
acquisitions into three quantiles based on the degree of diversification as measured
by cross industry cash flow and investment correlations. High correlation
corresponds to low degrees of diversification and vice-versa.
Table 20 provides the results of the longitudinal analysis. Let us start with
Panel A, which reports the results for the whole sample, that is both cash and non-
cash acquisitions. The first column reports average levels of excess cash for low-,
medium, high-diversification acquisitions. As the results show, acquiring firms hold
approximately 2.5 percent excess cash (as a fraction of their assets) relative to
similar non-acquiring firms, but there are no systematic observed differences
between cash holdings of firms that make more diversifying acquisitions and less
diversifying acquisitions. These results are consistent with Harford (1999), who
finds that cash-rich firms are more likely to make acquisitions.
The second column reports excess cash levels three years after the
acquisition has been made. For all three types of acquisitions, post-acquisition
excess cash levels are negative. However, the post-acquisition excess cash levels are
significantly different across firms that went through different degrees of
diversifying acquisitions. In particular, the more diversifying the acquisition, the less
cash the firm holds three years after the acquisition. The magnitude of the
differences is substantial: highly-diversified acquirers hold on average -5.1 percent
excess cash, while non-diversified acquirers hold on average -0.4 percent excess
cash, and they are all statistically significant at the 1 percent significance level. This
88
results is consistent with the hypothesis that coinsurance reduces optimal cash
balances.
Table 20: Longitudinal Study
This table identifies standalone firms that make an acquisition between 1998 and 2003. It sorts
acquisitions based on cross divisional correlations between investment opportunities and then tracks
the cash holdings of the acquiring firm over the next three years. Firms that make an additional
acquisition during this three-year period are excluded from the sample. The table reports the cash
holdings of the acquiring firm three years after the acquisition took place, benchmarked against a
control group of the average non-acquiring firm in the same industry and size decile.

.Panel A: All Firms
Acquisition
Type
Cash Before
Acquisition
Cash After
Acquisition
After - Before
(t-statistic)
Low Correlation 0.024 -0.051
0.075
(6.31)
Medium
Correlation
0.027 -0.019
0.046
(5.59)
High Correlation 0.026 -0.004
0.030
(2.12)
High - Low
(t-statistic)
0.002
(0.62)
0.047
(4.88)
0.045
(5.12)
Panel B: Cash Acquisitions
Acquisition
Type
Cash Before
Acquisition
Cash After
Acquisition
After - Before
(t-statistic)
Low Correlation 0.038 -0.067
0.105
(7.14)
Medium
Correlation
0.039 -0.031
0.060
(7.78)
High Correlation 0.045 -0.012
0.057
(4.59)
High - Low
(t-statistic)
0.007
(1.49)
0.045
(5.33)
0.048
(4.46)
89
Table 20, Continued
Panel C: Non-Cash Acquisitions
Acquisition
Type
Cash Before
Acquisition
Cash After
Acquisition
After - Before
(t-statistic)
Low Correlation 0.022 -0.038
0.060
(5.32)
Medium
Correlation
0.023 -0.013
0.036
(4.07)
High Correlation 0.022 -0.001
0.023
(1.99)
High - Low
(t-statistic)
0.000
(0.00)
0.037
(3.65)
0.037
(3.86)

Panel B reports average excess cash levels for cash-acquisitions. As in Panel
A, the results indicate that highly diversifying acquisitions lead firms to hold less
cash relative to less diversifying acquisitions. In fact, firms that went through a
highly diversifying acquisition hold 4.5 percent less excess cash (as a fraction of
assets) compared to firms that went through the least-diversifying acquisitions. Cash
holdings fall by 10.5 percent in highly diversifying acquisitions and by 5.7 percent
in the least diversifying acquisitions. While qualitatively similar, panels A and B
reveal that cash-acquirers tend to hold approximately 1.5 percent more excess cash
relative to the sample average, and they burn approximately 1.5 percent more cash
in the acquisition. These results indicate that firms build cash reserves to finance
acquisitions (or pushed to make acquisitions due to large unused cash reserves) and
actually spend them on acquisitions.
90
Panel C reports the results of a similar analysis focusing on non-cash
acquisitions. Once again, the results indicate that highly diversifying acquisitions
allow firms to hold less cash. Highly diversified acquirers hold 3.7 percent less
excess cash relative to the least diversifying acquirers and this difference is once
again statistically significant at the 1 percent level. However, non-cash acquirers
generally build smaller cash reserves prior to the acquisition, and end up having
more cash after the acquisition compared to cash-acquirers.
Overall, all three panels are consistent with the hypothesis that diversifying
allows firms to hold a reduced buffer of cash. The longitudinal study shows that
acquirers generally hold positive excess cash balances before making an acquisition,
and revert to a negative excess cash balance 3 years after making the acquisition.
However, the reduction in cash is larger and the actual post-merger excess cash
balance is smaller when the acquisition is more diversifying. These results are in
accordance the cross-sectional results presented earlier, and are consistent with the
interpretation that coinsurance allows firms to optimally reduce their precautionary
cash holdings and save on the costs associated with holding cash.
Put together, these results highlight the role played by cash holdings in
funding investment opportunities. Firms use cash to finance investments and take
potential future investments into account when they determine how much cash they
should hold. These considerations are more important in firms that are less
diversified, because potential underinvestment problems pose a bigger threat due to
lower coinsurance benefits. Thus, coinsurance and cash holdings are two alternative
mechanisms that mitigate potential underinvestment problems, and the existence of
91
one reduces the need for the other. This implies that cash will be less used to fund
investments and therefore less strongly correlated with investments, in firms with
lower cross-divisional correlations where coinsurance is stronger.
92
Conclusion

The interaction between corporate liquidity and corporate diversification is
interesting theoretically as well as practically. From a theoretical point of view,
diversified firms enjoy the benefit of coinsurance, which reduces their exposure to
cash flow and investment risks, and allows them to hold reduced amounts of cash in
comparison to their standalone counterparts. As long as investment opportunities
and cash flows are not perfectly correlated across the various divisions, diversified
firms can save on liquidity costs and hold less cash as a fraction of assets.
Furthermore, lower correlations further reduce the optimal cash holdings. From a
practical point of view, diversified US firms hold a large fraction of total corporate
cash. In 2004, for example, diversified firms held approximately 70 percent of
aggregate corporate cash.
This study presents a model of firms that hold cash to preempt emerging
investment opportunities and to deal with possible future cash flow shortages. Ex-
ante, the firm assesses the distribution of divisional investment opportunities and
cash flows. It then chooses the optimal level of cash that balances between the
benefits of being able to invest more in the future and the costs associated with
forgoing valuable investment opportunities today. This framework highlights the
linkage between diversification and cash holdings, and points out that it is especially
relevant when firms are financially constrained. It suggests that corporate cash
holdings are not only affected by cash-flow and investment industry-level
93
volatilities, but also by cross-divisional correlations between cash flows and
between investment opportunities.
I test these predictions empirically using nonparametric as well as regression
tests. The nonparametric tests record a significant economic impact of
diversification on average cash holdings. Average cash-to-assets ratios decrease as
the number of reported business segments increases, and increase with the cross-
divisional correlations between investment opportunities or cash flows. The
parametric-regression analysis is conducted within the framework of four schematic
liquidity models: the diversification, lifecycle, capital-structure, and investment
models. I find that diversification, through its impact on the exposure to investment
and cash flow risks, is a key determinant of cash holdings. This holds even after
controlling for the various factors associated with the three other models of
corporate liquidity. Moreover, the inclusion of the diversification model is able to
explain 5 percents of the cross-sectional variation of cash holding unexplained by
the previous models.
The sample is then divided into financially constrained and unconstrained
firms and separate regressions are estimated in each subsample. The results suggest
that the effect of diversification is more pronounced for financially constrained
firms, consistent with the predictions of the model. I also consider the interaction
between diversification and net debt, defined as debt minus cash, and show that
average net debt is substantially higher in diversified firms, consistent with a
coinsurance effect that reduces the probability of default.
94
I also show that the dramatic increase in average cash holdings from 1990 to
2004 was accompanied by an increase in average cross-division correlations
between investment opportunities and between cash flows. The increase in average
correlations of diversified firm contributes to the increase in their risk exposure, and
can therefore explain why diversified firms hold more cash than they used to.
My results indicate that the ability to hold less cash due to an increased level
of coinsurance, as captured by lower cross-divisional correlations, has a positive
impact on firm value. Accounting for the selection bias that contaminates empirical
studies of the diversification discount through a propensity-score approach, I show
that excess values increase when firms hold less cash due to lower correlations.
Finally, I investigate the connection between investment and cash holdings
and show that diversification reduces the firm’s dependence on cash to finance its
investments. In more diversified firms, the sensitivity of investment to cash is lower,
consistent with the coinsurance hypothesis which suggests that diversified firms will
more often transfer internally generated funds to finance investments in their various
divisions rather than utilize their cash balances for this purpose.
95
References

Andrade, G., Mitchell, M., Stafford, E., 2001, New Evidence and Perspectives on
Mergers, Journal of Economic Perspectives, 103-120.

Acharya, V., Almeida, H., Campello, M., 2006, Is Cash Negative Debt? A Hedging
Perspective on Corporate Financial Policies, London Business School working
paper, London, England.

Almeida, H., Campello, M., Weisbach, M., 2004, The Cash Flow Sensitivity of
Cash, Journal of Finance, 1777-1804.

Baskin, J., 1987, Corporate Liquidity in Games of Monopoly Power, Review of
Economics and Statistics, 69, 312-319.

Bates, T.W., Kahle, K.M., Stulz, R.M., 2006, Why Do U.S. Firms Hold So Much
More Cash Than They Used To? NBER Working Paper No. 12534.

Beltz, J., Frank, M., 1996, Risk and Corporate Holdings of Highly Liquid Assets,
Unpublished manuscript, University of British Columbia, Vancouver.

Berger, P.G., Hann, R., 2003, The Impact of SFAS No. 131 on Information and
Monitoring, Journal of Accounting Research, 163-223.

Berger, P.G., Ofek, E., 1995, Diversification’s Effect on Firm Value, Journal of
Financial Economics, 39-65.

Campa, J., Kedia, S., 2002, Explaining the Diversification Discount, Journal of
Finance, 1731-1762.

Chudson, W., 1945, The Pattern of Corporate Financial Structure, NBER, New
York.

DeAngelo, H., DeAngelo, L., Skinner, D.J., 2004, Are Dividends Disappearing?
Dividend Concentration and the Consolidation of Earnings, Journal of Financial
Economics, 425-456.

DeAngelo, H., DeAngelo, L., Stulz, R., 2006, Dividend Policy and the Earned
/Contributed Capital Mix: A Test of the Lifecycle Theory, Journal of Financial
Economics, 81, 227-254.

96
Dittmar, A., Mahrt-Smith, J., 2007, Corporate Governance and the Value of Cash,
Journal of Financial Economics, 83, 599-634.

Dittmar, A., Mahrt-Smith, J., Servaes, H., 2003, International Corporate Governance
and Corporate Cash Holdings, Journal of Financial and Quantitative Analysis,
111-133.

Kalcheva, I., Lins, K., 2007, International Evidence on Cash Holdings and Expected
Marginal Agency Problems, Review of Financial Studies, 20, 1087-1112.

Faleye, O., 2004, Cash and Corporate Control, Journal of Finance 59, 2041-60.

Fama, E., French, K., 2001, Disappearing Dividends: Changing Firm Characteristics
or Lower Propensity to Pay? Journal of Financial Economics, 3-43.

Fama, E., French, K., 2004, New Lists: Fundamentals and Survival Rates, Journal of
Financial Economics, 229-269.

Faulkender, M. Wang, R., 2006, Corporate Financial Policy and the Value of Cash,
Journal of Finance, 1957-1990

Fazzari, S., Petersen, B., 1993, Working Capital and Fixed Investment: New
Evidence on Financing Constraints, RAND Journal of Economics, 328-342.

Fazzari, S.M., Hubbard, R.G., Petersen, B., 1988, Financing Constraints and
Corporate Investment. Brookings Papers on Economic Activity, 19, 141-195.

Foley, F., Hartzell, J., Titman, S., Twite, G., 2006, Why do Firms Hold So Much
Cash? A Tax-based Explanation, NBER Working Paper No. W12649.

Graham, J., Campbell, H., 2001, The Theory and Practice of Corporate Finance:
Evidence From the Field, Journal of Financial Economics, 187-243.

Gompers, P., Ishii, J., Metrick, A., 2003, Corporate Governance and Equity Prices,
Quarterly Journal of Economics, 118, 107-155.

Harford, J., 1999, Corporate Cash Reserves and Acquisitions, Journal of Finance,
1969-1997.

Harford, J., Mansi, S., Maxwell, W., 2006, Corporate Governance and Firm Cash
Holdings, AFA 2006 Boston Meetings Paper.

Hart, O., Moore, J., 1994, A Theory of Debt Based on the Inalienability of Human
Capital, Quarterly Journal of Economics, 841-879.

97
Irvine, P.J., Pontiff, J.E., 2005, Idiosyncratic Return Volatility, Cash Flows, and
Product Market Competition, unpublished working paper, Boston College,
Newton, MA.

Jensen, M.C., 1986, Agency Costs of Free Cash Flow, Corporate Finance, and
Takeovers, American Economic Review, 323-329.

Jensen, M.C., Meckling, W.H., 1976, Theory of the Firm: Managerial Behavior,
Agency Costs and Ownership Structure, Journal of Financial Economics 3,
305-360.

John, T.A., 1993, Accounting Measures of Corporate Liquidity, Leverage, and Costs
of Financial Distress, Financial Management, 91-100.

Keynes, J.M., 1936, The General Theory of Employment, In: Interest and Money.
Harcourt Brace, London.

Kim, Chang-Soo, Mauer, D.C., Sherman, A.E., 1998, The Determinants of
Corporate Liquidity: Theory and Evidence. Journal of Financial and Quantitative
Analysis, 305-334.

Lewellen, W.G., 1971, A Pure Financial Rationale for Conglomerate Merger,
Journal of Finance, 527-537.

Matsusaka, J.G., Nanda, V., 2002, Internal Capital Markets and Corporate
Refocusing, Journal of Financial Intermediation, 176-211

Mikkelson, W., Partch, M., 2003, Do Persistent Large Cash Reserves Hinder
Performance? Journal of Financial and Quantitative Analysis, 275-294.

Miller, M.H., Orr. D., 1966, A Model of the Demand for Money by Firms, Quarterly
Journal of Economics, 413-435.

Modigliani, F., Miller, M.H., 1958, The Cost of Capital, Corporation Finance and
the Theory of Investment, American Economic Review, 261-297.

Mulligan, C.B., 1997, Scale Economies, the Value of Time, and the Demand for
Money: Longitudinal Evidence from Firms, Journal of Political Economy,
1061-1079.

Myers, S.C., 1977, Determinants of Corporate Borrowing, Journal of Financial
Economics, 5, 147-175.

98
Myers, S.C., Majluf, N., 1984, Corporate Financing and Investment Decisions When
Firms have Information that Investors Do Not Have, Journal of Financial
Economics, 13, 187-221.

Newey, W., West, K., 1987, A Simple, Positive, Semi-Definite, Heteroskedasticity
and Autocorrelation Consistent Covariance Matrix, Econometrica, 703-708.

Opler, T., Pinkowitz, L., Stulz, R., Williamson, R., 1999, The Determinants and
Implications of Corporate Cash Holdings, Journal of Financial Economics, 3-46.

Ozbas, O., 2005, Integration, Organizational Processes, and Allocation of
Resources, Journal of Financial Economics, 201-242.

Pinkowitz, L., Stulz, R., Williamson, R., 2006, Do Firms in Countries with
Poor Protection of Investor Rights Hold More Cash?, Journal of Finance,
Forthcoming.

Rajan, R., Servaes, H., Zingales, L., 2000, The Cost of Diversity: The
Diversification Discount and Inefficient Investment, Journal of Finance, 35-80.

Scharfstein, D. S., Stein, J.C., 2000, The Dark Side of Internal Capital Markets:
Divisional Rent-Seeking and Inefficient Investment, Journal of Finance,
2537–2564.

Shin, H., Stulz, R., 1998, Are Internal Capital Markets Efficient? Quarterly Journal
of Economics, 531-552.

Sibilkov, V., Denis, D., 2007, Financial Constraints, Investment, and the Value of
Cash Holdings, Working Paper.

Stein, J.C., 1997, Internal Capital Markets and the Competition for Corporate
Resources, Journal of Finance, 111-133.

Stulz, R., 1990, Managerial Discretion and Optimal Financing Policies, Journal of
Financial Economics, 26, 3-27.

Vogel, R.C., Maddala, G.S., 1967, Cross-Section Estimates of Liquid Asset Demand
by Manufacturing Corporations, Journal of Finance, 22, 557-575.

Sufi, A., 2007, Bank Lines of Credit in Corporate Finance: An Empirical Analysis,
Review of Financial Studies, Forthcoming.

Villalonga, B., 2004, Does Diversification Cause the “Diversification Discount”?
Financial Management, 33, 5-27.

Abstract (if available)

Abstract This study documents a strong connection between corporate cash holdings and organization form, and explores why such a relation exists. Over the period 1990-2004, U.S. multi-division firms held 39 percent less cash as a fraction of assets than stand-alone firms, and the difference persists after controlling for firm size, capital structure, cash flows, growth opportunities and investments. To explain the difference, I study the determinants of cash holdings in a sample of 10,380 firms over 1990-2004 and find that (a) correlations between divisional cash flows or investment opportunities completely explain the relation between cash holdings and organization form and (b) their effect is big: an increase of one standard deviation in the correlations between cash flows (investment opportunities) implies an increase of 13 percent (10 percent) in cash holdings of the average firm. This suggests that corporate diversification can serve as a way to economize on cash. The paper also documents a substantial increase in average cross-divisional correlations from 1990 to 2004, which might explain why diversified firms hold more cash than they used to.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Share repurchases: how important is market timing?

PDF

Costs and benefits of "friendly" boards during mergers and acquisitions

PDF

Accrual quality and expected returns: the importance of controlling for cash flow shocks

PDF

Incentives and relative-wealth concerns: theory and evidence

PDF

Workplace organization and asset pricing

PDF

Essays on delegated portfolio management under market imperfections

PDF

Three essays on young entrepreneurial firms

PDF

Evolution of returns to scale and investor flows during the life cycle of active asset management

PDF

Essays in asset demand: dividends, institutions, and price

Asset Metadata

Creator Duchin, Ran (author)

Core Title Cash holdings and corporate diversification

School Marshall School of Business

Degree Doctor of Philosophy

Degree Program Business Administration

Publication Date 05/06/2008

Defense Date 03/25/2008

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag cash,cross-divisional correlation,diversification,liquidity,OAI-PMH Harvest

Place Name USA (countries)

Language English

Advisor DeAngelo, Harry (committee member), James, Gareth (committee member), Officer, Micah (committee member), Ozbas, Oguzhan (committee member), Sensoy, Berk (committee member)

Creator Email duchin@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1230

Unique identifier UC1106354

Identifier etd-Duchin-20080506 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-67839 (legacy record id),usctheses-m1230 (legacy record id)

Legacy Identifier etd-Duchin-20080506.pdf

Dmrecord 67839

Document Type Dissertation

Rights Duchin, Ran

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu