You are on page 1of 18

Firm diversication and the value of corporate cash holdings

Zhenxu Tong
X Centre for Finance and Investment, University of Exeter, Rennes Drive, Exeter EX4 4ST, United Kingdom
a r t i c l e i n f o a b s t r a c t
Article history:
Received 6 February 2008
Received in revised form 3 March 2009
Accepted 5 May 2009
Available online 8 May 2009
This paper studies the effect of rm diversication on the value of corporate cash holdings. We
develop two hypotheses based on efcient internal capital market and agency problems. We
nd that the value of cash is lower in diversied rms than in single-segment rms, and that
rm diversication is associated with a lower value of cash in both nancially unconstrained
and constrained rms. We nd that rm diversication has a negative (zero) impact on the
value of cash among rms with a lower (higher) level of corporate governance. These ndings
are consistent with the interpretation that rm diversication reduces the value of corporate
cash holdings through agency problems.
2009 Elsevier B.V. All rights reserved.
JEL classication:
G32
G34
Keywords:
Firm diversication
Corporate cash holdings
1. Introduction
The impact of rm diversication on rmvalue has received considerable attention from economists. Lang and Stulz (1994) and
Berger and Ofek (1995) nd a signicant diversication discount and interpret the results as evidence of value destruction by
diversied rms. Agency problems have beenproposedas anexplanation for the lower valuations associatedwithrmdiversication.
Denis et al. (1997) nd that agency problems are responsible for rms maintaining value-reducing diversication strategies. Shin and
Stulz (1998) and Rajan et al. (2000) show that the internal capital market in diversied rms engages in cross-subsidization by
allocating too much (too little) to divisions with low (high) investment opportunities. These results are in line with the argument
proposed by Jensen (1986) that managers derive private benets from rm diversication. However, these ndings have been
challengedby a number of other papers. For example, Campa and Kedia (2002) and Grahamet al. (2002) argue that the diversication
discount is tainted by endogeneity problems because rms with poor performance choose to diversify.
1
We contribute to the debate in this literature by studying an under-researched channel through which rm diversication can
affect rm value: corporate cash holdings. The study of corporate cash holdings can complement the existing research in the
following ways.
First, cash holdings provide a promising area to examine the implications of agency problems. Managers can get access to cash
holdings with less scrutiny and use them in a discretionary way. Myers and Rajan (1998) argue that more liquid assets can be
turned into private benets at lower costs. The existing literature has largely focused on investigating investments such as capital
expenditures (e.g., Shin and Stulz, 1998; Rajan et al., 2000) to examine the potential agency problems associated with rm
Journal of Corporate Finance 17 (2011) 741758
I would like to thank David Denis (the Editor), an anonymous referee, and seminar participants at the University of Warwick, the University of Exeter, and the
2008 Financial Management Association European Conference for their comments and suggestions.
Tel.: +44 1392 263155; fax: +44 1392 262475.
E-mail address: z.tong@exeter.ac.uk.
1
The literature has identied other costs or benets associated with rm diversication. For example, Amihud and Lev (1981) argue that there is a managerial
preference for rm diversication to achieve risk reduction. See Stein (2003) and Maksimovic and Phillips (2007) for a more comprehensive review of the
literature.
0929-1199/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jcorpn.2009.05.001
Contents lists available at ScienceDirect
Journal of Corporate Finance
j our nal homepage: www. el sevi er. com/ l ocat e/ j cor pf i n
diversication. Since managers can more easily derive private benets from cash holdings, our study can complement the agency
literature on rm diversication by investigating the value of cash holdings.
Second, cash holdings occupy a signicant role in the balance sheet. Bates et al. (2006) nd that the average cash to asset ratio
for industrial rms increases from10.48% to 24.03% between 1980 and 2004. Moreover, Opler et al. (1999) showthat in their data,
mean corporate cash holdings are greater than mean capital expenditures or mean acquisitions.
2
The magnitude of cash holdings
implies that they serve as a potentially important channel through which rm diversication can affect rm value. The literature
has documented a diversication discount at a magnitude of around 15%, using Tobin's Q as the measure of rmvalue (e.g., Berger
and Ofek, 1995). The calculation of Tobin's Q involves a rm's total assets. It implies that we can study the value impact of
diversication not only through rms' capital expenditures, as often examined in the literature but also through their other types
of assets. Given that the value of cash holdings represents a signicant part of the total rm value, we believe that it is a valid
approach to study the value impact of diversication through corporate cash holdings.
Third, cash holdings are more comparable across rms.
3
One dollar in cash in a diversied rmis physically the same as one dollar
in cash in a single-segment rm. This facilitates the comparison of the value of cash between diversied rms and single-segment
rms. One of the focuses inthe debate over the diversicationdiscount relates tothe comparability of investments betweendiversied
rms and single-segment rms. Since divisional investment opportunities are not directly observable, a method has beenproposed in
the literature to obtain the measures of divisional investment opportunities based on the investment opportunities of single-segment
rms (e.g., Rajan et al., 2000). However, Graham et al. (2002) argue that the discount associated with rm diversication by
acquisitions is not attributable to diversication itself but should be attributed to the fact that diversifying rms acquire assets that are
already valued at a discount relative totheir industry benchmarks. Since cashholdings are more comparable across rms, we candraw
conclusions more directly from the results.
We develop two hypotheses in this paper. First, rm diversication increases the value of cash holdings for nancially
constrained rms through efcient internal capital market because more resources are allocated to the divisions with better
investment opportunities (e.g., Stein, 1997). Second, rm diversication reduces the value of cash holdings through agency
problems because rm diversication can be associated with empire building (e.g., Jensen, 1986) and cross-subsidization (e.g.,
Shin and Stulz, 1998; Rajan et al., 2000).
We use a sample of 28,563 rm-year observations from1998 to 2005. We use the methodology in Faulkender and Wang (2006)
to study the value of corporate cash holdings. We nd that the marginal value of one dollar in diversied (single-segment) rms is
$0.92 ($1.08), implying that the same dollar is valued 16 cents less in diversied rms than single-segment rms. This is consistent
with the interpretation that rm diversication on average reduces the value of cash holdings. We divide the sample into
nancially unconstrained and constrained rms and nd that the value of cash is lower for diversied rms in both sub-samples.
We obtain a measure of corporate governance and divide the sample into the rms with a higher and a lower level of corporate
governance. We nd that rm diversication has a negative (zero) impact on the value of cash among the rms with a lower
(higher) level of corporate governance. We obtain consistent results when we use an alternative measure of unexpected change in
cash holdings, three econometric methods to control for the potential endogeneity problem, an alternative measure of corporate
governance, and pre-1997 data.
These ndings are most consistent with predictions of the agency hypothesis. We conclude that rmdiversication reduces the
value of corporate cash holdings through agency problems.
The remainder of the paper is organized as follows. Section 2 develops two hypotheses about the relation between rm
diversication and the value of cash holdings. Section 3 discusses the data and describes the methodology. Section 4 examines how
rm diversication affects the value of cash holdings. Section 5 concludes the paper.
2. Hypotheses
We develop two hypotheses about the relation between rm diversication and the value of cash holdings in this section.
2.1. Efcient internal capital market
For a given amount of capital, the headquarters of a diversied rm have the control rights to allocate more resources to the
divisions with better investment opportunities. Stein (1997) argues that this form of winner-picking increases the efciency of
internal capital allocation. We expect that among nancially constrained rms, the value of corporate cash holdings is higher for
diversied rms than for single-segment rms. The winner-picking argument is relevant to nancially constrained rms. If a rm
is nancially unconstrained, the rm can receive funding from external capital markets to nance good investment projects. We
thus propose the following hypothesis.
Hypothesis 1. Firm diversication increases the value of cash holdings in nancially constrained rms through efcient internal
capital markets.
2
See Opler et al. (1999), p17. They show that the mean of the ratio Cash/Net Assets (dened as Total Assets minus Cash) is 0.170. The mean of the ratio Capital
Expenditures/Net Assets is 0.090, while the mean of the ratio Acquisitions/Net Assets is 0.011.
3
We follow the literature and use cash and marketable securities (Compustat item #1) to construct the measures of cash holdings. Since rms may hold cash
in the form of marketable securities, we argue that cash holdings are more comparable, rather than identical, across rms.
742 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
2.2. Agency problems
Firmdiversication can be associated with agency problems. At the rmlevel, empire-building preferences will cause managers
to spend available funds excessively on unprotable investment projects (e.g., Jensen, 1986). Morck et al. (1990) nd more negative
market reactions when acquirers are engaging in unrelated diversication. At the segment-level, cross-subsidization can exist in
diversied rms. Cross-subsidization refers to the diversion of corporate resources from divisions with good investment
opportunities to divisions with poor investment opportunities. Shin and Stulz (1998) and Rajan et al. (2000) nd evidence of
inefcient cross-subsidization in diversied rms. Correspondingly, Ahn and Denis (2004) nd that corporate spinoffs result in a
signicant increase in investment efciency and create value by breaking up the conglomerates. Because cash holdings are more
liquid assets, managers can use cash in a discretionary way and thus derive private benets more easily. We expect that agency
problems reduce the value of cash holdings in a diversied rm because shareholders anticipate the inefcient use of cash. We
propose the following hypothesis.
Hypothesis 2. Firm diversication reduces the value of cash holdings through agency problems.
2.3. Combining the hypotheses
We combine the above two hypotheses into the following table.
The impact of rm diversication on the value of cash
Efcient internal capital market Agency problems
Financially unconstrained rms 0
Financially constrained rms +
This table summarizes the predictions of the two hypotheses. A plus (minus) sign indicates a positive (negative) impact of rm
diversication on the value of cash holdings. A zero sign indicates no relation between these two. The predictions of the two
hypotheses are different.
We supplement the analysis by comparing the marginal value of $1 with one. From the perspective of agency problems, we
expect that an additional dollar is valued signicantly less than one in diversied rms because of the anticipated misallocation of
cash by managers.
4
3. Data and methodology
In this section, we describe the data and the methodology.
3.1. Data
We obtain segment-level data fromthe Compustat/Segment database. We use the Compustat/Industrial Annual database as the
source for the rm-level data. We obtain stock return data from CRSP. The sample period is from 1998 to 2005. The Statement of
Financial Accounting Standards (SFAS) No. 131 was issued as a new standard for the reporting of segment information in 1997.
Villalonga (2004a,b) and Berger and Hann (2003) show that segment data before and after 1997 are not directly comparable to
each other. Since we use the 2006 version of Compustat, we set the sample period from1998 to 2005, using the data in the period
after the implementation of SFAS 131 to ensure the comparability of the data.
5
We match the Compustat/Segment with the Compustat/Industrial Annual and CRSP databases and exclude rms with
incomplete data. Following the literature on rm diversication (e.g., Berger and Ofek, 1995), we exclude nancial service rms
and rms with nancial service segments (SIC codes between 6000 and 6999). We also require that the sum of the segment sales
be within 1% of the total sales of the rm. We dene diversied rms as the rms that have at least two segments with different SIC
codes.
6
The observations of multi-segment rms with segments in the same SIC codes are treated as single-segment rms. After
these screening procedures, we obtain a nal sample of 6867 rms with 28,563 rm-year observations. The sample contains
10,828 (17,735) rm-year observations for diversied (single-segment) rms.
4
An alternative argument predicts a lower value of cash in diversied rms because diversied rms have marginally fewer protable investment
opportunities than single-segment rms. In this case, we still expect that the marginal value of $1 is not less than one in diversied rms because there is no
waste of cash by managers under this argument.
5
We use pre-1997 data for robustness checks in the later analysis.
6
We obtain similar results by dening diversied rms as the rms that have at least two segments with different two-digit SIC codes.
743 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
3.2. Methodology
Faulkender and Wang (2006) developed a methodology to measure the marginal value of cash holdings. We follow their
method and develop the primary specication as follows.
R
i;t
RB
i;t
= a + b
1

CashHoldings
i;t
MV
i;t 1
+ b
2
FirmDiversification
i;t

CashHoldings
i;t
MV
i;t 1
+ b
3
FirmDiversification
i;t
+ b
4

Earnings
i;t
MV
i;t 1
+ b
5

Net Assets
i;t
MV
i;t 1
+ b
6

R&D
i;t
MV
i;t 1
+ b
7

Interest Expenses
i;t
MV
i;t 1
+ b
8

Dividends
i;t
MV
i;t 1
+ b
9
CashHoldings
i;t 1
+ b
10
Leverage
i;t
+ b
11
CashHoldings
i;t 1

CashHoldings
i;t
MV
i;t 1
+ b
12
Leverage
i;t

CashHoldings
i;t
MV
i;t 1
+ b
13

New Financing
i;t
MV
i;t 1
+ e
i;t
;
1
where X
i,t
indicates the change in the variable X of rm i from year t 1 to t.
(e.g., Cash holdings
i,t
=cash holdings
i,t
cash holdings
i, t 1
)
R
i,t
: stock return over scal year t 1 to t.
RB
i,t
: stock i's benchmark return over scal year t 1 to t. The benchmark portfolio is one of the 25 Fama and French portfolios
formed on size and book-to-market.
MV
i,t 1
: market value of equity at year t 1 computed as price times shares outstanding.
Firm Diversication
i,t
: a dummy variable that equals 1 for diversied rms and 0 for single-segment rms.
Cash Holdings
i,t
: cash and marketable securities at year t.
Earnings
i,t
: earnings before extraordinary items over scal year t 1 to t.
Net Assets
i,t
: net assets (total assetscash holdings) at year t.
R&D
i,t
: R&D expenses over scal year t 1 to t.
Interest Expenses
i,t
: interest expenses over scal year t 1 to t.
Dividends
i,t
: common dividends over scal year t 1 to t.
Leverage
i,t
: leverage (debt / total assets) at year t.
New Financing
i,t
: net new equity issues (equity issued minus repurchases)+net new debt issues (debt issued minus debt
retired) over scal year t 1 to t.
This methodology essentially represents a long-run event study. The event is the unexpected change in cash holdings, while the
event windowis dened as the entire scal year. In the above equation, the dependent variable is excess stock return. It is calculated
as the stock return of rmi during a scal year t (R
i,t
), minus its benchmark return over the same period (RB
i,t
). This left-hand side
variable is interpreted as the cumulative abnormal return during a scal year, which incorporates the market reaction to the
unexpected change in corporate cash holdings. We use the 25 Fama and French portfolios formed on size and book-to-market as
benchmark portfolios. For each rm-year observation, a rmis grouped into one of the 25 Fama and French portfolios based on the
intersection between size and book-to-market. The return of the corresponding Fama and French portfolio is regarded as the
benchmark return for the rm during that year.
We rst dene the unexpected change in cash holdings as the realized change in cash holdings. Then we use an alternative
measure based on the net change in cash holdings. In the primary specication, we measure the change in cash holdings as the
ratio of the change in cash and marketable securities (Compustat item#1) over the scal year to the 1-year lagged market value of
equity. Since both the dependent variable and the independent variable are scaled by the 1-year lagged market value of equity, the
coefcient represents the dollar change in shareholder value resulting froma one-dollar change in the amount of cash held by the
rm. We can interpret this as the marginal value of one dollar to the shareholders of the rm.
We construct an interaction term, Firm DiversicationCash Holdings. Firm Diversication is a dummy variable that equals 1
for diversied rms and 0 for single-segment rms. The coefcient of this interaction term represents the difference in the
marginal value of one dollar between diversied rms and single-segment rms, thus indicating the impact of rm diversication
on the value of cash holdings. We also include the Firm Diversication dummy in the regression to ensure that the estimated
coefcient of the interaction term is due to the interaction, and not due to rm diversication itself.
We include other control variables in the regression, as suggested by Faulkender and Wang (2006). They argue that the value of
cash holdings depends on both leverage and the lag of cash holdings, and they include the interaction terms between these two
variables and the change in cash holdings in the regression. We control for sources of value other than cash (Earnings, Net Assets,
R&D, Interest Expenses, Dividends, and New Financing). We provide more details on the denition of the variables in the
Appendix A. We winsorize the data at 1% and 99% to reduce the impact of outliers.
744 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
4. Results
We report the results in this section. We rst present the relation between rm diversication and the value of cash holdings
in the entire sample. Then we demonstrate the results by dividing the sample into nancially unconstrained rms and
constrained rms. We next investigate the impact of corporate governance on the value of cash between diversied rms and
single-segment rms. Finally, we report the results using an alternative measure of unexpected changes in cash holdings, three
econometric methods to control for the potential endogeneity problem, an alternative measure of corporate governance, and
pre-1997 data. In the regressions, we report the p-value calculated based on the robustness standard errors clustered by rm
(e.g., Petersen, 2009).
4.1. The value of cash holdings
Our primary objective is to measure the impact of rmdiversication on the value of cash holdings. The results are presented in
Table 1. We rst examine the marginal value of cash holdings in the entire sample and report the results in the rst and second
columns. In this regression, an extra dollar in cash will affect a rm's excess return through the item Cash Holdings and the
interaction terms (Cash Holding
t 1
Cash Holdings and LeverageCash Holdings). We use the coefcients of these items to
obtain the marginal value. The calculation is as follows: the mean rmhas a lag of cash holdings equivalent to 16.56% of the market
capitalization of equity, and the mean leverage ratio is 20.14%. Therefore, the marginal value of one dollar to shareholders in the
mean rm is $1.02 (=1.245+(0.72616.56%)+(0.50920.14%)). We conduct the F-test on the null hypothesis that the
marginal value of $1 is one and report the p-value in the brackets. We nd that an additional dollar is valued insignicantly
different from one (p-value=0.43) in the entire sample.
In the third and fourth columns of Table 1, we report the impact of rmdiversication on the value of corporate cash holdings. This
regressionfollows the primary specication, as indicatedinEq. (1). To calculate the marginal value of cashholdings for single-segment
rms, we need to use the coefcients of the same three items as before (Cash Holdings, Cash Holding
t 1
Cash Holdings, and
LeverageCash Holdings). We nd that the marginal value of one dollar to shareholders in single-segment rms is $1.08. The F-test
shows that it is signicantly different fromone (p-value=0.02). To calculate the marginal value of cash holdings for diversied rms,
we need to use the coefcient of the interactiontermFirmDiversicationCashHoldings. Therefore, the marginal value of one dollar
to shareholders in diversied rms is $0.92 (=1.295+(0.160)+(0.74116.56%)+(0.47820.14%)). The F-test shows that it is
signicantly different from one (p-value=0.04).
Table 1
The value of cash holdings.
Coef. p-value Coef. p-value
Intercept 0.003 0.61 0.003 0.64
Cash Holdings 1.245 0.01 1.295 0.01
Firm DiversicationCash Holdings 0.160 0.01
Firm Diversication 0.001 0.98
Earnings 0.649 0.01 0.648 0.01
Net Assets 0.284 0.01 0.284 0.01
R&D 0.742 0.01 0.744 0.01
Interest Expenses 1.305 0.01 1.309 0.01
Dividends 2.814 0.01 2.821 0.01
Cash Holdings
t 1
0.194 0.01 0.195 0.01
Leverage 0.224 0.01 0.223 0.01
Cash Holdings
t 1
Cash Holdings 0.726 0.01 0.741 0.01
LeverageCash Holdings 0.509 0.01 0.478 0.01
New Financing 0.011 0.69 0.007 0.78
Observations 28,563 28,563
Adjusted R
2
0.13 0.14
The following table shows the marginal value of $1, calculated based on the estimates in the regressions (see text for more details). We conduct the F-test on the
null hypothesis that the marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Entire sample $1.02 (0.43) Single-segment rms $1.08 (0.02)
Diversied rms $0.92 (0.04)
This table reports the value of cash holdings. The dependent variable is Excess Return, dened as R
i,t
RB
i,t
. All variables except Excess Return, FirmDiversication, and
Leverage are standardized by the lagged market value of equity. Cash Holdings is cash plus marketable securities. Cash Holdings is the one-year change in cash holdings
(CashHoldings
t
CashHoldings
t 1
). FirmDiversicationis a dummyvariablethat equals 1for diversiedrms and0for single-segment rms. Earnings is theone-year
change in earnings before extraordinary items. Net Assets is the one-year change in total assets minus cash holdings. R&D is the one-year change in research and
development expenses. Interest Expenses is the one-year change in interest expenses. Dividends is the one-year change in common dividends. Leverage is the ratio
of debt to total assets. New Financing is net new equity issues plus net new debt issues. The p-value is calculated based on robust standard errors.
745 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
These results indicate that an extra dollar in diversied rms on average is valued 16 cents (p-value=0.01) less than an extra
dollar in single-segment rms. It implies that rm diversication reduces the value of cash holdings. Moreover, since the marginal
value of $1 is signicantly belowone in diversied rms, this is consistent with the agency hypothesis and reects the anticipated
misallocation of cash by the managers in diversied rms.
4.2. Financial constraints
We proceed to conduct the tests by dividing the sample into nancially unconstrained and constrained rms. We design two
criteria for nancial constraints based on the measures used in the previous literature (e.g., Almeida et al., 2004). We use the
lagged variables to measure the criteria. Since lagged variables are pre-determined, they are less likely to be subject to the
endogeneity problem.
4.2.1. The criteria of nancial constraints
4.2.1.1. Payout. Fazzari et al. (1988) state that nancially constrained rms have lower payout ratios. Firms with higher payout
ratios are more likely to have sufcient internal resources to nance their investments. We use the ratio of the sum of dividends
and stock repurchases to total assets as the measure for the payout ratio. For every year of the sample period, we sort all rms
according to their payout ratios at year t 1 and assign a rm to the nancially unconstrained (constrained) group if its lagged
payout ratio is above (below) the mean of the annual payout distribution.
4.2.1.2. Credit rating. Firms with a credit rating have better access to public debt markets. They are usually better known and have
less difculty in raising external funds to nance their investments. For every rm-year observation, we obtain data for long-term
issuer credit ratings at year t 1 from Compustat, and we assign a rm to the nancially unconstrained (constrained) group if its
lagged long-term issuer credit rating is available (unavailable).
4.2.2. Results
Table 2 presents the results by dividing the sample into nancially unconstrained and constrained rms according to the two
criteria, and it reports the marginal value of cash below the regressions. We show the results using payout as the nancial
constraints criterion in Panel A. We nd that the coefcients of the interaction term, Firm DiversicationCash Holdings, are
negative and signicant, supporting the interpretation that rm diversication reduces the value of cash holdings in both
nancially unconstrained rms and constrained rms.
Among nancially unconstrained rms, we nd that an additional dollar is valued at $1.01, which is insignicantly different
from one (p-value=0.88), in single-segment rms. We also nd that an additional dollar is valued at $0.83, which is signicantly
different from one (p-value=0.06), in diversied rms. The results imply that shareholders place a lower value on the cash
holdings in diversied rms because of the anticipated inefcient use of cash. This is consistent with the predictions of the agency
hypothesis.
Among nancially constrained rms, we nd that an additional dollar is valued at $1.12, which is signicantly different
fromone (p-value=0.01), in single-segment rms. Although the interaction term, FirmDiversicationCash Holdings, is 0.151
(p-value=0.01), we nd that an additional dollar is valued at $0.96, which is insignicantly different fromone (p-value=0.27), in
diversied rms. We report the results using credit rating as the criterion for nancial constraints in Panel B and nd a consistent
pattern.
The interpretation of the results for nancially constrained rms is confounded by the fact that corporate governance and
nancial constraints can concurrently affect the value of cash holdings (e.g., Dittmar and Mahrt-Smith, 2007).
7
Faulkender and
Wang (2006) nd that the difference in the value of an additional dollar between nancially unconstrained rms and constrained
rms can reach up to $0.63.
8
This economic magnitude is consistent with the ndings in Denis and Sibilkov (2009) who document
that the marginal value of cash can be 51 cents higher in constrained rms than in unconstrained rms. These ndings in the
literature imply that it is possible to discover a situation where the negative impact of agency problems and the positive impact of
nancial constraints on the value of cash offset each other, so that we observe a marginal value of cash insignicantly different from
one for nancially constrained rms. Therefore, we conduct further analysis aiming to identify whether agency problems also
prevail in the value of cash holdings for diversied rms among nancially constrained rms.
4.3. Corporate governance
We include a measure of corporate governance in this section and more explicitly examine the impact of agency problems on
the value of cash holdings in diversied rms. From the agency perspective, we expect a negative (zero) impact of diversication
7
See Dittmar and Mahrt-Smith (2007), p612614.
8
See Faulkender and Wang (2006), p1981.
746 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
Table 2
Regressions for nancially unconstrained and constrained rms.
Panel A. The criteria for nancial constraints: payout
Unconstrained Constrained
Coef. p-value Coef. p-value
Intercept 0.045 0.18 0.030 0.01
Cash Holdings 1.202 0.01 1.351 0.01
Firm DiversicationCash Holdings 0.185 0.06 0.151 0.01
Firm Diversication 0.004 0.70 0.001 0.89
Earnings 0.775 0.01 0.632 0.01
Net Assets 0.296 0.01 0.308 0.01
R&D 1.144 0.01 0.692 0.01
Interest Expenses 2.838 0.01 1.677 0.01
Dividends 3.459 0.01 4.026 0.01
Cash Holdings
t 1
0.140 0.08 0.251 0.01
Leverage 0.037 0.48 0.139 0.01
Cash Holdings
t 1
Cash Holdings 0.202 0.05 0.819 0.01
LeverageCash Holdings 1.130 0.33 0.698 0.01
New Financing 0.278 0.01 0.034 0.25
Observations 7011 21,552
Adjusted R
2
0.11 0.14
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Unconstrained Constrained
Single-segment rms $1.01 (0.88) $1.12 (0.01)
Diversied rms $0.83 (0.06) $0.96 (0.27)
Panel B. The criteria for nancial constraints: credit rating
Unconstrained Constrained
Coef. p-value Coef. p-value
Intercept 0.023 0.37 0.090 0.01
Cash Holdings 1.190 0.01 1.462 0.01
Firm DiversicationCash Holdings 0.198 0.03 0.143 0.02
Firm Diversication 0.041 0.03 0.013 0.18
Earnings 0.905 0.01 0.654 0.01
Net Assets 0.219 0.04 0.370 0.01
R&D 0.152 0.87 0.819 0.01
Interest Expenses 1.665 0.01 1.968 0.01
Dividends 4.314 0.01 3.555 0.01
Cash Holdings
t 1
0.295 0.01 0.331 0.01
Leverage 0.084 0.22 0.145 0.01
Cash Holdings
t 1
Cash Holdings 0.774 0.01 1.672 0.01
LeverageCash Holdings 0.295 0.11 0.446 0.01
New Financing 0.147 0.11 0.009 0.77
Observations 6052 22,511
Adjusted R
2
0.13 0.13
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Unconstrained Constrained
Single-segment rms $1.02 (0.85) $1.11 (0.01)
Diversied rms $0.82 (0.05) $0.97 (0.54)
This table reports results using the criteria of nancial constraints. The regressions are presented across the groups of nancially unconstrained and constrained
rms (see text for denitions). The criteria for nancial constraints are payout (Panel A) and credit rating (Panel B). The dependent variable is Excess Return, dened
as R
i,t
RB
i,t
. All variables except Excess Return, Firm Diversication, and Leverage are standardized by the lagged market value of equity. Cash Holdings is cash plus
marketable securities. Cash Holdings is the one-year change in cash holdings (Cash Holdings
t
Cash Holdings
t 1
). Firm Diversication is a dummy variable that
equals 1 for diversied rms and 0 for single-segment rms. Earnings is the one-year change in earnings before extraordinary items. Net Assets is the one-year
change of total assets minus cash holdings. R&D is the one-year change of research and development expenses. Interest Expenses is the one-year change in
interest expenses. Dividends is the one-year change in common dividends. Leverage is the ratio of debt to total assets. New Financing is net new equity issues
plus net new debt issues. The p-value is calculated based on robust standard errors.
747 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
on the value of cash among the rms with a lower (higher) level of corporate governance.
9
We use a corporate governance index
(Gindex) proposed by Gompers et al. (2003). We use the data on corporate charters of takeover defenses from the Investor
Responsibility Research Center (IRRC) database. A higher Gompers et al. index indicates more restrictions on shareholder rights,
thus corresponding to a lower level of corporate governance.
10
The IRRC database only provides the data for a subset of rms
(mostly larger rms) in the sample. We conduct the analysis with this sub-sample that contains 8242 rm-year observations.
We divide the sample into two groups. One group has a higher level of corporate governance (Gindexb7) and includes
observations in the bottom quartile. The other group has a lower level of corporate governance (Gindex7). We separately
estimate the regressions for each group and report the results in Table 3. We nd that the coefcient of the interaction term, Firm
DiversicationCash Holdings, is 0.064 (p-value=0.90) in the group with a higher level of corporate governance. This is
consistent with the interpretation that since good corporate governance alleviates agency problems in diversied rms, we do not
nd a signicant difference in the value of cash between diversied rms and single-segment rms. Correspondingly, we nd that
the marginal value of one dollar is $1.25 ($1.19) in diversied (single-segment) rms. This implies that cash holdings are valuable
to the shareholders in both diversied rms and single-segment rms with a higher level of corporate governance.
However, we nd that the coefcient of the interaction term, Firm DiversicationCash Holdings, is still negative and
signicant in the group with a lower level of corporate governance. Correspondingly, we nd that the marginal value of one dollar
is $0.49 ($0.88) in diversied (single-segment) rms. Both are signicantly different from one. This supports the interpretation
that shareholders place a lower value on cash holdings in rms with lower levels of corporate governance, and that cash is even
less valuable in diversied rms because of the additional agency problems associated with rm diversication. We conduct a
9
A higher level of corporate governance wipes out the difference in the value of cash between diversied rms and single-segment rms caused by agency
problems. We thank an anonymous referee for comments on this point.
10
The Gompers et al. index has a range of possible values from 0 to 24. See Gompers et al. (2003) for more details.
Table 3
The Impact of corporate governance.
Higher governance Lower governance
Gindexb7 (bottom quartile) Gindex7
Coef. p-value Coef. p-value
Intercept 0.050 0.16 0.033 0.04
Cash Holdings 1.515 0.01 1.158 0.01
Firm DiversicationCash Holdings 0.064 0.90 0.389 0.05
Firm Diversication 0.040 0.27 0.016 0.20
Earnings 0.798 0.01 0.715 0.01
Net Assets 0.351 0.01 0.375 0.01
R&D 0.593 0.64 0.709 0.19
Interest Expenses 1.379 0.27 3.280 0.01
Dividends 5.273 0.10 4.442 0.01
Cash Holdings
t 1
0.161 0.26 0.160 0.01
Leverage 0.103 0.34 0.131 0.01
Cash Holdings
t 1
Cash Holdings 1.758 0.01 0.680 0.39
LeverageCash Holdings 0.650 0.22 1.042 0.04
New Financing 0.330 0.18 0.331 0.01
Observations 1608 6634
Adjusted R
2
0.09 0.14
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Higher governance Lower governance
Gindexb7 (bottom quartile) Gindex7
Single-segment rms $1.19 (0.06) $0.88 (0.07)
Diversied rms $1.25 (0.01) $0.49 (0.01)
The following table shows the difference in the effect of diversication on the value of cash between high and lowgovernance rms and reports the p-value of the
t-test in the brackets.
Difference 0.453 (0.01)
This table reports the impact of corporate governance. The dependent variable is Excess Return, dened as R
i,t
RB
i,t
. All variables except Excess Return, Firm
Diversication, and Leverage are standardized by the lagged market value of equity. Gindex is the corporate governance index constructed by Gompers et al.
(2003). Cash Holdings is cash plus marketable securities. Cash Holdings is the one-year change in cash holdings (Cash Holdings
t
Cash Holdings
t 1
).
Earnings is the one-year change in earnings before extraordinary items. Net Assets is the one-year change in total assets minus cash holdings. R&D is the one-
year change in research and development expenses. Interest Expenses is the one-year change in interest expenses. Dividends is the one-year change in
common dividends. Leverage is the ratio of debt to total assets. New Financing is net new equity issues plus net new debt issues. The p-value is calculated based
on robust standard errors.
748 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
t-test of the difference in the effect of diversication on the value of cash between high and lowgovernance rms. We discover
that the difference is signicant (p-value=0.01). Therefore, the results in Table 3 are consistent with the agency hypothesis.
4.4. Financial constraints and corporate governance
We examine the effect of diversication on the value of cash for high and low governance rms separately for unconstrained
and constrained rms in this section. This can conrm whether the value of cash is affected by agency problems in both
unconstrained and constrained rms. We divide the sub-sample covered by the IRRC database into four groups and estimate the
value of cash for each group.
We report the results in Table 4. Among nancially unconstrained rms with a higher level of corporate governance, we nd
that an additional dollar is valued at $1.04 (p-value=0.59) in single-segment rms. We also nd that an additional dollar is valued
at $1.03 (p-value=0.72) in diversied rms. The coefcient of the interaction term, Firm Diversication Cash Holdings, is
0.006 (p-value=0.98). Among nancially unconstrained rms with a lower level of corporate governance, we nd that an
additional dollar is worth $0.83 (p-value=0.05) in single-segment rms, while it is worth $0.46 (p-value=0.01) in diversied
rms. The coefcient of the interaction term, FirmDiversication Cash Holdings, is 0.371 (p-value=0.03). The t-test shows a
signicant difference (p-value=0.04) between high and lowgovernance rms in the effect of diversication on the value of cash.
Therefore, the results are consistent with the interpretation that rm diversication has a negative (zero) impact on the value of
cash among rms with a lower (higher) level of corporate governance for nancially unconstrained rms.
Table 4
Financial constraints and corporate governance.
Unconstrained Constrained
Higher governance Lower governance Higher governance Lower governance
Gindexb7
(bottom quartile)
Gindex7 Gindexb7
(bottom quartile)
Gindex7
Coef. p-value Coef. p-value Coef. p-value Coef. p-value
Intercept 0.019 0.58 0.026 0.18 0.037 0.45 0.046 0.05
Cash Holdings 1.427 0.01 1.118 0.01 1.771 0.05 1.256 0.01
Firm DiversicationCash Holdings 0.006 0.98 0.371 0.03 0.034 0.96 0.309 0.06
Firm Diversication 0.067 0.05 0.035 0.02 0.027 0.59 0.018 0.38
Earnings 0.637 0.01 1.063 0.01 0.867 0.01 0.744 0.01
Net Assets 0.445 0.01 0.324 0.01 0.320 0.03 0.336 0.01
R&D 0.754 0.73 0.337 0.49 0.610 0.70 1.216 0.10
Interest Expenses 3.391 0.02 4.577 0.01 1.365 0.33 3.186 0.01
Dividends 2.317 0.47 3.128 0.03 7.774 0.17 7.028 0.01
Cash Holdings
t 1
0.131 0.34 0.023 0.79 0.270 0.17 0.187 0.01
Leverage 0.118 0.39 0.104 0.11 0.102 0.47 0.103 0.11
Cash Holdings
t 1
Cash Holdings 0.408 0.85 2.294 0.03 2.168 0.03 0.298 0.75
LeverageCash Holdings 1.727 0.02 0.117 0.93 0.233 0.89 1.252 0.01
New Financing 0.362 0.12 0.274 0.01 0.299 0.29 0.237 0.03
Observations 518 2860 1090 3774
Adjusted R
2
0.19 0.14 0.09 0.15
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Unconstrained Constrained
Higher governance Lower governance Higher governance Lower governance
Gindexb7 (bottom quartile) Gindex7 Gindexb7 (bottom quartile) Gindex7
Single-segment rms 1.04 (0.59) 0.83 (0.05) 1.48 (0.03) 0.98 (0.79)
Diversied rms 1.03 (0.72) 0.46 (0.01) 1.45 (0.03) 0.67 (0.05)
The following table shows the difference in the effect of diversication on the value of cash between high and lowgovernance rms and reports the p-value of the
t-test in the brackets.
Unconstrained Constrained
Difference 0.365 (0.04) 0.275 (0.07)
This table reports the results using the criteria of both nancial constraints and corporate governance. The regressions are presented across the groups of nancially
unconstrained/constrained rms and high/low governance rms. Payout is the criterion for nancial constraints. Gindex is the corporate governance index constructed
by Gompers et al. (2003). The dependent variable is Excess Return, dened as R
i,t
RB
i,t
. All variables except Excess Return, Firm Diversication, and Leverage are
standardized by the lagged market value of equity. Cash Holdings is cash plus marketable securities. Cash Holdings is the one-year change in cash holdings
(Cash Holdings
t
Cash Holdings
t 1
). Firm Diversication is a dummy variable that equals 1 for diversied rms and 0 for single-segment rms. Earnings is the
one-year change in earnings before extraordinary items. Net Assets is the one-year change in total assets minus cash holdings. R&D is the one-year change
in research and development expenses. Interest Expenses is the one-year change in interest expenses. Dividends is the one-year change in common dividends.
Leverage is the ratio of debt to total assets. New Financing is net new equity issues plus net new debt issues. The p-value is calculated based on robust standard errors.
749 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
Among nancially constrained rms with a higher level of corporate governance, we nd that an additional dollar is valued at
$1.48 (p-value=0.03) in single-segment rms. We also nd that an additional dollar is valued at $1.45 (p-value=0.03) in
diversied rms. The coefcient of the interaction term, Firm Diversication Cash Holdings, is 0.034 (p-value=0.96).
Among nancially constrained rms with a lower level of corporate governance, we nd that an additional dollar is worth $0.98
(p-value=0.79) in single-segment rms, while it is worth $0.67 (p-value=0.05) in diversied rms. The coefcient of the
interaction term, Firm Diversication Cash Holdings, is 0.309 (p-value=0.06). The t-test shows a signicant difference (p-
value=0.07) between high and low governance rms in the effect of diversication on the value of cash. These results reconcile
the agency hypothesis and the ndings in Table 2 that the marginal value of $1 is insignicantly different from one for diversied
rms among nancially constrained rms. When we divide constrained rms into two groups depending on their levels of
corporate governance, we nd that agency problems also prevail in the value of cash holdings for diversied rms among
nancially constrained rms with a lower level of corporate governance. Therefore, these ndings are consistent with the agency
hypothesis.
4.5. Alternative measure of the unexpected change in cash holdings
Our primary specication essentially represents a long-term event study. The excess return reects the market reaction to
unexpected changes in cash holdings. We followFaulkender and Wang (2006) and repeat our analysis using the net change in cash
holdings dened as the realized change in cash holdings minus the average change in cash holdings in the corresponding
benchmark portfolio over the same period. There is an additional advantage to using the net change in cash holdings. Bates et al.
(2006) document a time trend in the level of cash holdings. We expect the time trend to be present in both a rm's realized change
in cash holdings and the average change in cash holdings in the corresponding benchmark portfolio. We can therefore reduce the
impact of the time trend by using the net change in cash holdings.
Table 5A shows that an extra dollar is worth $1.08 ($0.93) for single-segment (diversied) rms. We repeat the analysis on the
marginal value of cash for nancially unconstrained rms and constrained rms and report the results in this table, using payout as
the criterion for nancial constraints. We obtain consistent results using credit rating as the criterion for nancial constraints. The
results are similar to those presented above. Firm diversication reduces the value of cash holdings in both nancially
Table 5A
The value of cash holdings alternative measure of the unexpected change in cash holdings.
Entire Sample Unconstrained Constrained
Coef. p-value Coef. p-value Coef. p-value
Intercept 0.016 0.01 0.016 0.55 0.065 0.01
Net Cash Holdings 1.394 0.01 1.113 0.01 1.482 0.01
Firm DiversicationNet Cash Holdings 0.155 0.01 0.169 0.02 0.180 0.01
Firm Diversication 0.001 0.97 0.006 0.37 0.010 0.19
Earnings 0.652 0.01 0.777 0.01 0.681 0.01
Net Assets 0.291 0.01 0.260 0.01 0.349 0.01
R&D 0.698 0.01 1.149 0.03 0.863 0.01
Interest Expenses 1.427 0.01 2.314 0.01 1.522 0.01
Dividends 2.926 0.01 2.679 0.01 3.864 0.01
Cash Holdings
t 1
0.250 0.01 0.128 0.14 0.286 0.01
Leverage 0.147 0.01 0.079 0.12 0.135 0.01
Cash Holdings
t 1
Net Cash Holdings 1.211 0.01 0.778 0.02 1.268 0.01
LeverageNet Cash Holdings 0.706 0.01 0.162 0.57 0.718 0.01
New Financing 0.025 0.34 0.175 0.04 0.059 0.07
Observations 28,563 7011 21,552
Adjusted R
2
0.13 0.12 0.13
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Entire sample Unconstrained Constrained
Single-segment rms $1.08 (0.03) $1.02 (0.92) $1.15 (0.01)
Diversied rms $0.93 (0.02) $0.87 (0.08) $0.97 (0.53)
This table reports the value of cash holdings using an alternative measure of the unexpected change in cash holdings. Payout is the criterion for nancial
constraints. The dependent variable is Excess Return, dened as R
i,t
RB
i,t
. All variables except Excess Return, Firm Diversication, and Leverage are standardized
by the lagged market value of equity. Cash Holdings is cash plus marketable securities. Net Cash Holdings is the realized change in cash holdings minus the
average change in cash holdings in the corresponding benchmark portfolio over the same period (see text for more details). Firm Diversication is a dummy
variable that equals 1 for diversied rms and 0 for single-segment rms. Earnings is the one-year change in earnings before extraordinary items (Earnings
t

Earnings
t 1
). Net Assets is the one-year change in total assets minus cash holdings. R&D is the one-year change in research and development expenses.
Interest Expenses is the one-year change in interest expenses. Dividends is the one-year change in common dividends. Leverage is the ratio of debt to total
assets. New Financing is net new equity issues plus net new debt issues. The p-value is calculated based on robust standard errors.
750 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
unconstrained rms and constrained rms. Table 5B presents the results using the net change in cash holdings with the separation
of the sample by the level of corporate governance. Firm diversication has a negative (zero) impact on the value of cash among
the rms with a lower (higher) level of corporate governance. Table 5C presents the results for the sub-samples separated by both
nancial constraints and corporate governance. We get consistent results that agency problems prevail in the value of cash
holdings for diversied rms among both nancially unconstrained rms and constrained rms.
4.6. About the endogeneity
Campa and Kedia (2002) argue that rmdiversication can be an endogenous choice made by a rm, and they use econometric
techniques to control for the potential endogeneity problem. Other papers in the literature have expressed similar concerns (e.g.,
Graham et al., 2002). We follow Campa and Kedia (2002) and used three econometric methods to examine the impact of rm
diversication on the value of cash holdings while controlling for the endogeneity problem.
4.6.1. Methodology
The three econometric methods are the Heckman two-stage estimation, instrumental variables estimation and xed effect
estimation. Each of them addresses the endogeneity problem from a different perspective.
We apply the Heckman (1979) two-stage estimation to control for a rm's self-selection into diversication. In the rst stage,
we estimate a probit regression to model the decision of a rmto diversify. The variables are motivated by Campa and Kedia (2002)
and include various rm-specic, industry and macroeconomic characteristics that can inuence a rm's decision to diversify. We
use some two-year lagged variables (size, EBIT, and capital expenditures) in the probit regression, which reduces the sample size
Table 5B
The value of cash holdings: the impact of corporate governance alternative measure of the unexpected change in cash holdings.
Higher governance Lower governance
Gindexb7 (Bottom quartile) Gindex7
Coef. p-value Coef. p-value
Intercept 0.070 0.04 0.042 0.01
Net Cash Holdings 1.482 0.01 1.122 0.01
Firm DiversicationNet Cash Holdings 0.005 0.99 0.290 0.01
Firm Diversication 0.039 0.29 0.011 0.29
Earnings 0.800 0.01 0.530 0.01
Net Assets 0.350 0.01 0.269 0.01
R&D 0.628 0.62 0.476 0.21
Interest Expenses 1.460 0.25 2.370 0.01
Dividends 5.197 0.10 3.106 0.01
Cash Holdings
t 1
0.146 0.32 0.121 0.01
Leverage 0.107 0.34 0.101 0.01
Cash Holdings
t 1
Net Cash Holdings 1.590 0.02 0.842 0.01
LeverageNet Cash Holdings 0.604 0.25 0.860 0.01
New Financing 0.323 0.19 0.243 0.01
Observations 1608 6634
Adjusted R
2
0.10 0.12
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Higher governance Lower governance
Gindexb7 (bottom quartile) Gindex7
Single-segment rms $1.20 (0.01) $0.86 (0.05)
Diversied rms $1.19 (0.02) $0.57 (0.01)
The following table shows the difference in the effect of diversication on the value of cash between high and lowgovernance rms and reports the p-value of the
t-test in the brackets.
Difference 0.285 (0.01)
This table reports the impact of corporate governance with an alternative measure of the unexpected change in cash holdings. The dependent variable is Excess
Return, dened as R
i,t
RB
i,t
. All variables except Excess Return, Firm Diversication, and Leverage are standardized by the lagged market value of equity. Gindex
is the corporate governance index constructed by Gompers et al. (2003). Cash Holdings is cash plus marketable securities. Net Cash Holdings is the realized
change in cash holdings minus the average change in cash holdings in the corresponding benchmark portfolio over the same period (see text for more details).
Firm Diversication is a dummy variable that equals 1 for diversied rms and 0 for single-segment rms. Earnings is the one-year change in earnings before
extraordinary items (Earnings
t
Earnings
t 1
). Net Assets is the one-year change in total assets minus cash holdings. R&D is the one-year change in research
and development expenses. Interest Expenses is the one-year change in interest expenses. Dividends is the one-year change in common dividends. Leverage is
the ratio of debt to total assets. New Financing is net new equity issues plus net new debt issues. The p-value is calculated based on robust standard errors.
751 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
slightly to 27,767 rm-year observations due to incomplete data. Table 6A shows the probit regression. We obtain the Lambda from
the probit estimates. The calculation of Lambda follows the standard Heckman methodology and is reported in Appendix A. In the
second stage, we estimate the regressions with the Lambda as an additional control variable. This provides the treatment for a
rm's self-selection into diversication.
We use instrumental variables estimation to examine the underlying causal relations. We follow Campa and Kedia (2002) and
use the estimated probability of diversication from the probit model as a generated instrument. In the rst stage, we use all
exogenous variables and the probability of diversication as explanatory variables in the decision to diversify. In the second stage,
we use the tted value from the rst stage as an instrument for diversication to evaluate the effect of diversication on the value
of cash holdings. Campa and Kedia (2002) provide more details on the rationale of this methodology.
We introduce a two-way xed effect estimation to account for the omitted variable problem. Fixed rm effects and year effects
are used to control for unobservable rm characteristics and time effects. We exclude rms with only one observation during the
sample period in order to estimate the xed effect regression, thus leaving 27,151 rm-year observations in the sample.
4.6.2. Results
We report the results using the three econometric methods on the entire sample inTable 6B. The rst and second columns show
the second stage of the Heckman two-stage estimation. We nd that the marginal value of one dollar is $0.93 ($1.08) for diversied
Table 5C
Financial constraints and corporate governance alternative measure of the unexpected change in cash holdings.
Unconstrained Constrained
Higher governance Lower governance Higher governance Lower governance
Gindexb7
(bottom quartile)
Gindex7 Gindexb7
(bottom quartile)
Gindex7
Coef. p-value Coef. p-value Coef. p-value Coef. p-value
Intercept 0.046 0.15 0.049 0.01 0.054 0.25 0.064 0.01
Net Cash Holdings 1.333 0.01 1.244 0.01 1.720 0.06 1.245 0.01
Firm DiversicationNet Cash Holdings 0.031 0.96 0.398 0.05 0.034 0.92 0.341 0.03
Firm Diversication 0.067 0.07 0.041 0.02 0.015 0.75 0.021 0.34
Earnings 0.689 0.01 1.088 0.01 1.022 0.01 0.751 0.01
Net Assets 0.447 0.01 0.330 0.01 0.343 0.03 0.333 0.01
R&D 0.628 0.79 0.337 0.52 0.439 0.79 1.236 0.10
Interest Expenses 3.473 0.02 4.520 0.01 1.358 0.33 3.171 0.01
Dividends 2.244 0.49 2.983 0.05 9.572 0.15 7.025 0.01
Cash Holdings
t 1
0.140 0.35 0.032 0.70 0.225 0.27 0.207 0.01
Leverage 0.124 0.40 0.100 0.13 0.120 0.41 0.119 0.06
Cash Holdings
t 1
Net Cash Holdings 0.019 0.99 2.496 0.02 2.311 0.03 0.264 0.78
LeverageNet Cash Holdings 1.662 0.03 0.656 0.62 0.038 0.98 1.256 0.08
New Financing 0.375 0.13 0.280 0.01 0.340 0.25 0.235 0.03
Observations 518 2860 1090 3774
Adjusted R
2
0.18 0.14 0.10 0.15
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Unconstrained Constrained
Higher governance Lower governance Higher governance Lower governance
Gindexb7 (bottom quartile) Gindex7 Gindexb7 (bottom quartile) Gindex7
Single-segment rms 1.00 (0.99) 0.83 (0.07) 1.45 (0.02) 0.97 (0.73)
Diversied rms 1.03 (0.65) 0.43 (0.01) 1.42 (0.02) 0.63 (0.03)
The following table shows the difference in the effect of diversication on the value of cash between high and lowgovernance rms and reports the p-value of the
t-test in the brackets.
Unconstrained Constrained
Difference 0.432 (0.01) 0.307 (0.02)
This table reports the results using the criteria of both nancial constraints and corporate governance. The regressions are presented across the groups of
nancially unconstrained/constrained rms and high/low governance rms. Payout is the criterion for nancial constraints. Gindex is the corporate governance
index constructed by Gompers et al. (2003). The dependent variable is Excess Return, dened as R
i,t
RB
i,t
. All variables except Excess Return, Firm
Diversication, and Leverage are standardized by the lagged market value of equity. Cash Holdings is cash plus marketable securities. Net Cash Holdings is the
realized change in cash holdings minus the average change in cash holdings in the corresponding benchmark portfolio over the same period (see text for more
details). Firm Diversication is a dummy variable that equals 1 for diversied rms and 0 for single-segment rms. Earnings is the one-year change in earnings
before extraordinary items. Net Assets is the one-year change in total assets minus cash holdings. R&D is the one-year change in research and development
expenses. Interest Expenses is the one-year change in interest expenses. Dividends is the one-year change in common dividends. Leverage is the ratio of debt
to total assets. New Financing is net new equity issues plus net new debt issues. The p-value is calculated based on robust standard errors.
752 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
Table 6A
Probit regression.
Coef. p-value
Intercept 3.592 0.01
Size 0.071 0.01
Size t 1 0.020 0.45
Size t 2 0.085 0.01
EBIT 0.710 0.01
EBIT t 1 0.711 0.01
EBIT t 2 0.160 0.02
Capital Expenditures 0.562 0.01
Capital Expenditures t 1 0.220 0.07
Capital Expenditures t 2 0.017 0.87
S&P 0.115 0.01
Major stock exchange 0.010 0.69
Fraction of Diversied Firms in the Industry 1.889 0.01
Fraction of industry sales by diversied rms 0.555 0.01
GDP t 1 0.001 0.01
GDP growth 0.286 0.65
Observations 27,767
Pseudo R
2
0.14
This table reports probit estimates. The dependent variable takes the value 1 for diversied rms and 0 for single-segment rms. Size is the logarithmof sales. EBIT
is the ratio of earnings before interests and taxes to assets. Capital Expenditures is the ratio of capital expenditures to assets. S&P is a dummy that equals 1 when the
rm is part of the S&P index and 0 otherwise. Major Stock Exchange is a dummy that equals 1 if the rm is listed on Nasdaq, NYSE, or AMEX, and 0 otherwise.
Fraction of Diversied Firms in the Industry is the fraction of all the rms in the industry that are diversied rms. Fraction of Industry Sales by Diversied Firms
is the fraction of industry sales accounted for by diversied rms. GDP is Gross Domestic Product. GDP Growth is the growth rate of Gross Domestic Product. The
p-value is calculated based on robust standard errors.
Table 6B
Three econometric methods.
Heckman two-stage
estimation: second stage
Instrumental variables
estimation: second stage
Fixed effect estimation
(rm effects and year
effects not reported)
Coef. p-value Coef. p-value Coef. p-value
Intercept 0.012 0.18 0.052 0.01
Cash Holdings 1.316 0.01 1.435 0.01 1.278 0.01
Firm DiversicationCash Holdings 0.145 0.01 0.334 0.02 0.118 0.03
Firm Diversication 0.010 0.53 0.084 0.01 0.008 0.61
Earnings 0.672 0.01 0.640 0.01 0.566 0.01
Net Assets 0.293 0.01 0.234 0.01 0.210 0.01
R&D 0.689 0.01 0.385 0.01 0.486 0.01
Interest Expenses 1.416 0.01 1.307 0.01 0.894 0.01
Dividends 3.008 0.01 2.545 0.01 2.069 0.01
Cash Holdings
t 1
0.202 0.01 0.270 0.01 0.804 0.01
Leverage 0.168 0.01 0.160 0.01 0.422 0.01
Cash Holdings
t 1
Cash Holdings 0.797 0.01 1.545 0.01 0.842 0.01
LeverageCash Holdings 0.687 0.01 0.558 0.01 0.320 0.01
New Financing 0.020 0.48 0.062 0.02 0.085 0.01
Lambda 0.004 0.71
Observations 27,767 27,767 27,151
Adjusted R
2
0.13 0.12 0.14
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Heckman two-stage estimation Instrumental variables estimation Fixed effect estimation
Single-segment rms $1.08 (0.01) $1.09 (0.01) $1.07 (0.03)
Diversied rms $0.93 (0.02) $0.76 (0.01) $0.95 (0.08)
This table reports the results using three econometric methods to control for the potential endogeneity problem (see text for more details). The dependent
variable is Excess Return, dened as R
i,t
RB
i,t
. All variables except Excess Return, Firm Diversication, and Leverage are standardized by the lagged market value
of equity. Cash Holdings is cash plus marketable securities. Cash Holdings is the one-year change in cash holdings (Cash Holdings
t
Cash Holdings
t 1
). Firm
Diversication is a dummy variable that equals 1 for diversied rms and 0 for single-segment rms. The tted value is used in the instrumental variables
estimation. Earnings is the one-year change in earnings before extraordinary items. Net Assets is the one-year change in total assets minus cash holdings.
R&D is the one-year change in research and development expenses. Interest Expenses is the one-year change in interest expenses. Dividends is the one-year
change in common dividends. Leverage is the ratio of debt to total assets. New Financing is net new equity issues plus net new debt issues. Lambda is obtained
from the probit estimates in Table 6A. The p-value is calculated based on robust standard errors.
753 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
Table 6C
Heckman two-stage estimation nancial constraints and the impact of corporate governance.
Panel A. Financial constraints
Unconstrained Constrained
Coef. p-value Coef. p-value
Intercept 0.032 0.05 0.012 0.25
Cash Holdings 1.176 0.01 1.379 0.01
Firm DiversicationCash Holdings 0.225 0.06 0.133 0.03
Firm Diversication 0.019 0.47 0.020 0.32
Earnings 0.764 0.01 0.660 0.01
Net Assets 0.282 0.01 0.309 0.01
R&D 1.269 0.01 0.636 0.01
Interest Expenses 2.135 0.01 1.714 0.01
Dividends 2.854 0.01 3.804 0.01
Cash Holdings
t 1
0.146 0.01 0.245 0.01
Leverage 0.093 0.01 0.153 0.01
Cash Holdings
t 1
Cash Holdings 0.709 0.01 0.860 0.01
LeverageCash Holdings 0.112 0.79 0.718 0.01
New Financing 0.189 0.01 0.035 0.25
Lambda 0.019 0.27 0.012 0.36
Observations 6925 20,842
Adjusted R
2
0.12 0.14
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Unconstrained Constrained
Single-segment rms $1.04 (0.40) $1.13 (0.01)
Diversied rms $0.82 (0.07) $0.99 (0.72)
Panel B. Corporate governance
Higher governance Lower governance
Gindexb7 (bottom quartile) Gindex7
Coef. p-value Coef. p-value
Intercept 0.075 0.23 0.043 0.11
Cash Holdings 1.535 0.01 1.156 0.01
Firm DiversicationCash Holdings 0.076 0.89 0.386 0.05
Firm Diversication 0.093 0.44 0.035 0.39
Earnings 0.792 0.01 0.715 0.01
Net Assets 0.353 0.01 0.377 0.01
R&D 0.627 0.62 0.714 0.18
Interest Expenses 1.388 0.27 3.286 0.01
Dividends 5.309 0.10 4.460 0.01
Cash Holdings
t 1
0.147 0.32 0.165 0.01
Leverage 0.094 0.41 0.128 0.01
Cash Holdings
t 1
Cash Holdings 1.807 0.01 0.675 0.39
LeverageCash Holdings 0.678 0.20 1.041 0.05
New Financing 0.341 0.17 0.334 0.01
Lambda 0.036 0.65 0.013 0.64
Observations 1608 6634
Adjusted R
2
0.10 0.14
The following table shows the marginal value of $1, calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and we report the p-value in the brackets.
The marginal value of $1
Higher governance Lower governance
Gindexb7 (bottom quartile) Gindex7
Single-segment rms $1.20 (0.01) $0.87 (0.06)
Diversied rms $1.27 (0.01) $0.48 (0.01)
The following table shows the difference in the effect of diversication on the value of cash between high and lowgovernance rms and reports the p-value of the
t-test in the brackets.
Difference 0.462 (0.02)
754 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
(single-segment) rms. The third and the fourth columns report the second stage of the instrumental variables estimation. The
marginal value of one dollar is $0.76 ($1.09) for diversied (single-segment) rms. We nd that the coefcient of the interaction
term, Firm DiversicationCash Holdings, is 0.334. The instrumental variables estimation reveals that diversication has a
more negative impact on the value of cash. The fth and the sixth columns show the results using xed effect estimation. We nd
that an additional dollar is worth $0.95 ($1.07) in diversied (single-segment) rms. These results are consistent with the
interpretation that rm diversication reduces the value of cash, controlling for the potential endogeneity problem.
Table 6D
Heckman two-stage estimation nancial constraints and corporate governance.
Unconstrained Constrained
Higher governance Lower governance Higher governance Lower governance
Gindexb7
(bottom quartile)
Gindex7 Gindexb7
(bottom quartile)
Gindex7
Coef. p-value Coef. p-value Coef. p-value Coef. p-value
Intercept 0.059 0.32 0.041 0.19 0.111 0.15 0.116 0.10
Cash Holdings 1.410 0.01 1.102 0.01 1.763 0.05 1.262 0.01
Firm DiversicationCash Holdings 0.017 0.98 0.362 0.04 0.015 0.90 0.328 0.04
Firm Diversication 0.155 0.15 0.062 0.18 0.202 0.16 0.152 0.15
Earnings 0.617 0.01 1.061 0.01 0.853 0.01 0.741 0.01
Net Assets 0.446 0.01 0.326 0.01 0.321 0.03 0.342 0.01
R&D 0.801 0.71 0.353 0.47 0.317 0.84 1.156 0.71
Interest Expenses 3.553 0.01 4.593 0.01 1.293 0.35 3.252 0.41
Dividends 2.452 0.44 3.158 0.03 7.817 0.17 7.134 0.18
Cash Holdings
t 1
0.152 0.27 0.032 0.72 0.253 0.20 0.217 0.16
Leverage 0.090 0.53 0.098 0.13 0.064 0.66 0.083 0.72
Cash Holdings
t 1
Cash Holdings 0.470 0.83 2.263 0.03 2.139 0.03 0.325 0.27
LeverageCash Holdings 1.702 0.01 0.102 0.94 0.298 0.86 1.245 0.01
New Financing 0.380 0.11 0.278 0.01 0.340 0.24 0.258 0.14
Lambda 0.059 0.37 0.020 0.54 0.119 0.20 0.098 0.15
Observations 518 2860 1090 3774
Adjusted R
2
0.19 0.14 0.09 0.15
The following table shows the marginal value of $1 calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and report the p-value in the brackets.
The marginal value of $1
Unconstrained Constrained
Higher governance Lower governance Higher governance Lower governance
Gindexb7 (bottom quartile) Gindex7 Gindexb7 (bottom quartile) Gindex7
Single-segment rms 1.02 (0.95) 0.82 (0.06) 1.46 (0.03) 0.98 (0.83)
Diversied rms 1.04 (0.61) 0.46 (0.01) 1.45 (0.03) 0.65 (0.04)
The following table shows the difference in the effect of diversication on the value of cash between high and lowgovernance rms, and reports the p-value of the
t-test in the brackets.
Unconstrained Constrained
Difference 0.379 (0.03) 0.313 (0.04)
This table reports the results using the criteria of both nancial constraints and corporate governance. The regressions are presented across the groups of
nancially unconstrained/constrained rms and high/low governance rms. Payout is the criterion for nancial constraints. Gindex is the corporate governance
index constructed by Gompers et al. (2003). The dependent variable is Excess Return, dened as R
i,t
RB
i,t
. All variables except Excess Return, Firm
Diversication, and Leverage are standardized by the lagged market value of equity. Cash Holdings is cash plus marketable securities. Cash Holdings is the one-
year change in cash holdings (Cash Holdings
t
Cash Holdings
t 1
). Firm Diversication is a dummy variable that equals 1 for diversied rms and 0 for single-
segment rms. Earnings is the one-year change in earnings before extraordinary items. Net Assets is the one-year change in total assets minus cash holdings.
R&D is the one-year change in research and development expenses. Interest Expenses is the one-year change in interest expenses. Dividends is the one-year
change in common dividends. Leverage is the ratio of debt to total assets. New Financing is net new equity issues plus net new debt issues. Lambda is obtained
from the probit estimates in Table 6A. The p-value is calculated based on robust standard errors.
Notes to Table 6C:
This table reports the second stage of the Heckman two-stage estimation. The dependent variable is Excess Return, dened as R
i,t
RB
i,t
. Panel A presents the
results using payout as the criteria of nancial constraints. All variables except Excess Return, Firm Diversication, and Leverage are standardized by the lagged
market value of equity. Cash Holdings is cash plus marketable securities. Cash Holdings is the one-year change in cash holdings. Firm Diversication is a dummy
variable that equals 1 for diversied rms and 0 for single-segment rms. Earnings is the one-year change in earnings before extraordinary items. Net Assets is
the one-year change in total assets minus cash holdings. R&D is the one-year change in research and development expenses. Interest Expenses is the one-year
change in interest expenses. Dividends is the one-year change in common dividends. Leverage is the ratio of debt to total assets. New Financing is net new
equity issues plus net new debt issues. Lambda is obtained from the probit estimates in Table 6A. Panel B reports the impact of corporate governance. Gindex is
the corporate governance index constructed by Gompers et al. (2003). The p-value is calculated based on robust standard errors.
755 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
We report the results using payout as the criteria of nancial constraints and the impact of corporate governance in Table 6C.
This table only shows the results of the Heckman two-stage estimation for brevity. We get consistent results using the other two
econometric methods. We nd that rm diversication reduces the value of cash holdings in both nancially unconstrained and
constrained rms, and that rm diversication has a negative (zero) impact on the value of cash among rms with a lower
(higher) level of corporate governance. We report the results for the sub-samples separated by both nancial constraints and
corporate governance in Table 6D. We nd consistent results. These results give further support to the agency hypothesis
controlling for the potential endogeneity problem.
4.7. Alternative measure of corporate governance
We conduct a robustness check using an alternative measure of corporate governance. Bebchuk et al. (2009) argue that 6 out of
24 provisions in the IRRC database are the most important ones leading to managerial entrenchment. They design an
Entrenchment index based on these six provisions. A higher Entrenchment index indicates more restrictions on shareholder rights,
thus corresponding to a lower level of corporate governance.
11
We use a similar methodology and divide the sample into two
groups. One group has a higher level of corporate governance (Eindexb2) and includes observations in the bottom quartile. The
other group has a lower level of corporate governance (Eindex2).
11
The Entrenchment index has a range of possible values from 0 to 6. See Bebchuk et al. (2009) for more details.
Table 7
The impact of corporate governance alternative measure of governance.
Higher governance Lower governance
Eindexb2 (bottom quartile) Eindex2
Coef. p-value Coef. p-value
Intercept 0.032 0.27 0.046 0.01
Cash Holdings 1.312 0.01 1.206 0.01
Firm DiversicationCash Holdings 0.041 0.97 0.322 0.01
Firm Diversication 0.032 0.27 0.021 0.04
Earnings 0.816 0.01 0.738 0.01
Net Assets 0.344 0.01 0.377 0.01
R&D 0.878 0.43 0.376 0.22
Interest Expenses 2.329 0.03 3.159 0.01
Dividends 4.497 0.16 4.783 0.01
Cash Holdings
t 1
0.100 0.34 0.174 0.01
Leverage 0.087 0.30 0.141 0.01
Cash Holdings
t 1
Cash Holdings 1.101 0.03 0.332 0.34
LeverageCash Holdings 0.064 0.89 1.609 0.01
New Financing 0.237 0.12 0.375 0.01
Observations 2374 5868
Adjusted R
2
0.10 0.14
The following table shows the marginal value of $1 calculated based on the estimates in the regressions. We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and report the p-value in the brackets.
The marginal value of $1
Higher governance Lower governance
Eindexb2 (bottom quartile) Eindex2
Single-segment rms $1.18 (0.03) $0.85 (0.04)
Diversied rms $1.22 (0.03) $0.53 (0.01)
The following table shows the difference in the effect of diversication on the value of cash between high and lowgovernance rms and reports the p-value of the
t-test in the brackets.
Difference 0.363 (0.01)
This table reports the impact of corporate governance using an alternative measure of governance. The dependent variable is Excess Return dened as R
i,t
RB
i,t
.
All variables except Excess Return, Firm Diversication, and Leverage are standardized by the lagged market value of equity. Eindex is the entrenchment index
constructed by Bebchuk et al. (2009). Cash Holdings is cash plus marketable securities. Cash Holdings is the one-year change in cash holdings (Cash Holdings
t

Cash Holdings
t 1
). Earnings is the one-year change in earnings before extraordinary items. Net Assets is the one-year change in total assets minus cash
holdings. R&D is the one-year change in research and development expenses. Interest Expenses is the one-year change in interest expenses. Dividends is the
one-year change in common dividends. Leverage is the ratio of debt to total assets. New Financing is net new equity issues plus net new debt issues. The p-value
is calculated based on robust standard errors.
756 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
We report the results in Table 7. We nd that the coefcient of the interaction term, Firm DiversicationCash Holdings, is
0.041 (p-value=0.97) in the group with a higher level of corporate governance. We also nd that the coefcient of the
interaction term, Firm DiversicationCash Holdings, is 0.322 (p-value=0.01) in the group with a lower level of corporate
governance. The results imply that rm diversication has a negative (zero) impact on the value of cash among rms with a
lower (higher) level of corporate governance. We therefore nd consistent results using this alternative measure of corporate
governance.
4.8. Alternative sample period
We conduct the robustness check using the data from 1990 to 1996, to ensure that the effect is not conned to recent
years. We report the regressions for the entire sample in Table 8. We nd that an additional dollar is worth $1.07 in single-
segment rms, while it is worth $0.92 in diversied rms. Both are signicantly different from one. We nd consistent
results by separating the sample by the criteria of nancial constraints, as well as by using the measures of corporate
governance.
5. Conclusion
We study the relation between rm diversication and the value of corporate cash holdings. We develop two hypotheses
based on different theories of rm diversication. We use the methodology in Faulkender and Wang (2006) and nd that
the marginal value of one dollar in diversied (single-segment) rms is $0.92 ($1.08). The ndings imply that the same
dollar is valued 16 cents less in diversied rms than in single-segment rms. We nd that rm diversication reduces the
value of cash in both nancially unconstrained and constrained rms. We nd that rm diversication has a negative (zero)
impact on the value of cash among rms with a lower (higher) level of corporate governance. We obtain similar results
when we use an alternative measure of the unexpected change in cash holdings, three econometric methods to control for
the potential endogeneity problem, an alternative measure of corporate governance, and the data in a different sample
period.
These ndings are consistent with the predictions of the agency hypothesis. We conclude that rm diversication reduces the
value of corporate cash holdings through agency problems.
Table 8
The value of cash holding robustness check: sample period 19901996.
Coef. p-value Coef. p-value
Intercept 0.002 0.78 0.001 0.99
Cash Holdings 1.130 0.01 1.177 0.01
Firm DiversicationCash Holdings 0.145 0.04
Firm Diversication 0.003 0.68
Earnings 0.621 0.01 0.617 0.01
Net Assets 0.337 0.01 0.339 0.01
R&D 1.016 0.01 1.019 0.01
Interest Expenses 1.220 0.01 1.199 0.01
Dividends 2.246 0.01 2.228 0.01
Cash Holdings
t 1
0.184 0.01 0.157 0.01
Leverage 0.253 0.01 0.255 0.01
Cash Holdings
t 1
Cash Holdings 0.373 0.01 0.262 0.01
LeverageCash Holdings 0.305 0.01 0.307 0.01
New Financing 0.058 0.02 0.049 0.02
Observations 26,861 26,861
Adjusted R
2
0.14 0.14
The following table shows the marginal value of $1 calculated based on the estimates in the regressions We conduct the F-test on the null hypothesis that the
marginal value of $1 is one, and report the p-value in the brackets.
The marginal value of $1
Entire sample $1.01(0.81) Single-segment rms $1.07 (0.07)
Diversied rms $0.92 (0.05)
This table reports the value of cash holdings using the data in 19901996. The dependent variable is Excess Return dened as R
i,t
RB
i,t
. All variables except
Excess Return, Firm Diversication, and Leverage are standardized by the lagged market value of equity. Cash Holdings is cash plus marketable securities. Cash
Holdings is the one-year change in cash holdings (Cash Holdings
t
Cash Holdings
t 1
). Firm Diversication is a dummy variable that equals 1 for diversied rms
and 0 for single-segment rms. Earnings is the one-year change in earnings before extraordinary items. Net Assets is the one-year change in total assets minus
cash holdings. R&D is the one-year change in research and development expenses. Interest Expenses is the one-year change in interest expenses. Dividends
is the one-year change in common dividends. Leverage is the ratio of debt to total assets. New Financing is net new equity issues plus net new debt issues. The
p-value is calculated based on robust standard errors.
757 Z. Tong / Journal of Corporate Finance 17 (2011) 741758
Appendix A. Denition of variables
R
i,t
R
i,t
is rm i's stock return over scal year t 1 to t.
RB
i,t
RB
i,t
is stock i's benchmark return over scal year t 1 to t. The benchmark portfolio is one of the 25 Fama
and French portfolios formed based on rm size and book-to-market.
Excess Return Excess Return is dened as R
i,t
RB
i,t
.
MV
i,t 1
MV
i,t 1
is market value of equity at time t 1 computed as price (Compustat item #199) times shares outstanding (#25).
Cash Holdings Cash Holdings is dened as cash and marketable securities (#1).
Earnings Earnings is dened as earnings before extraordinary items (#18+#15+#50+#51).
Net Assets Net Assets is dened as total assets minus cash holdings (#6#1).
R&D R&D is dened as research and development expenses (#46).
Interest Expenses Interest Expenses is dened as interest expenses (#15).
Dividends Dividends is dened as common dividends (#21).
New Financing New Financing is dened as net new equity issues (#108#115) plus net new debt issues (#111#114).
Leverage Leverage is dened as the sum of long-term debt (#9) and debt in current liability (#34), divided by total assets (#6).
Firm Diversication Firm Diversication is a dummy variable that equals 1 for diversied rms and 0 for single-segment rms.
Lambda We assume that a rm's decision to diversify is determined by the equations
D
it

=Z
it
+u
it
D
it
=1, if D
it

N0
D
it
=0, if D
it

b0,
where D
it

is an unobservable latent variable. D


it
is a dummy variable (1 for diversied rms and 0 for
single-segment rms). Z
it
is a set of variables that affect a rm's decision to diversify, and u
it
is an error term.
We rst estimate the above equation using a probit model to obtain the estimates of denoted by
e
.
Lambda is calculated as follows:
Lambda
it
=
/
e
Z
it


e
Z
it

D
it
+
/
e
Z
it

1
e
Z
it

1 D
it
;
where the density function of the standard normal distribution. is the cumulative distribution function of the
standard normal distribution, and D
it
is a dummy variable (1 for diversied rms and 0 for single-segment rms).
References
Ahn, S., Denis, D.J., 2004. Internal capital markets and investment policy: evidence from corporate spinoffs. Journal of Financial Economics 71, 489516.
Almeida, H., Campello, M., Weisbach, M., 2004. The cash ow sensitivity of cash. Journal of Finance 59, 17771804.
Amihud, Y., Lev, B., 1981. Risk reduction as a managerial motive for conglomerate mergers. Bell Journal of Economics, 12, 605617.
Bates, T., Kahle, K., Stulz, R., 2006. Why do U.S. rms hold so much more cash than they used to? Unpublished working paper, University of Arizona.
Bebchuk, L., Cohen, A., Ferrell, A., 2009. What matters in corporate governance? Review of Financial Studies 22, 783827.
Berger, P., Ofek, E., 1995. Diversication's effect on rm value. Journal of Financial Economics 37, 3965.
Berger, P., Hann, R., 2003. The impact of SFAS No. 131 on information and monitoring. Journal of Accounting Research 41, 163223.
Campa, J., Kedia, S., 2002. Explaining the diversication discount. Journal of Finance 57, 17311762.
Denis, D.J., Sibilkov, V., 2010. Financial constraints, investment, and the value of cash holdings. Review of Financial Studies. 23, 247269.
Denis, D.J., Diane, D.K., Sarin, A., 1997. Agency problems, equity ownership, and corporate diversication. Journal of Finance 52, 135160.
Dittmar, A., Mahrt-Smith, J., 2007. Corporate governance and the value of cash holdings. Journal of Financial Economics 83, 599634.
Faulkender, M., Wang, R., 2006. Corporate nancial policy and the value of cash. Journal of Finance 61, 19571990.
Fazzari, S., Hubbard, R., Petersen, B., 1988. Financing constraints and corporate investment. Brookings Papers on Economic Activity 141195.
Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices. Quarterly Journal of Economics 118, 107155.
Graham, J., Lemmon, M., Wolf, J., 2002. Does corporate diversication destroy value. Journal of Finance 57, 695720.
Heckman, J.J., 1979. Sample selection as a specication error. Econometrica 47, 153161.
Jensen, M., 1986. Agency costs of free cash ow, corporate nance, and takeovers. American Economic Review 76, 323329.
Lang, L., Stulz, R., 1994. Tobin's q, corporate diversication, and rm performance. Journal of Political Economy 102, 12481280.
Maksimovic, V., Phillips, G., 2007. Conglomerate rms and internal capital markets. In: Eckbo, B. (Ed.), Handbook of Corporate Finance: Empirical Corporate
Finance. North-Holland, Amsterdam, pp. 423477.
Morck, R., Shleifer, A., Vishny, R., 1990. Do managerial objectives drive bad acquisitions? Journal of Finance 45, 3148.
Myers, S.C., Rajan, R., 1998. The paradox of liquidity. Quarterly Journal of Economics 113, 733771.
Opler, T., Pinkowitz, L., Stulz, R., Williamson, R., 1999. The determinants and implications of corporate cash holdings. Journal of Financial Economics 52, 346.
Petersen, M., 2009. Estimating standard errors in nance panel data sets: comparing approaches. Review of Financial Studies 22, 435480.
Rajan, R., Servaes, H., Zingales, L., 2000. The cost of diversity: the diversication discount and inefcient investment. Journal of Finance 55, 3580.
Shin, H., Stulz, R., 1998. Are internal capital markets efcient? Quarterly Journal of Economics 113, 531552.
Stein, J., 1997. Internal capital markets and the competition for corporate resources. Journal of Finance 52, 111133.
Stein, J., 2003. Agency, information and corporate investment. In: Constantinides, G.M., Harris, M., Stulz, R. (Eds.), Handbook of the Economics of Finance. Elsevier,
Amsterdam, Netherlands, pp. 111165.
Villalonga, B., 2004a. Diversication discount or premium? New evidence from the business information tracking series. Journal of Finance 59, 479506.
Villalonga, B., 2004b. Does diversication cause the diversication discount? Financial Management 33, 527.
758 Z. Tong / Journal of Corporate Finance 17 (2011) 741758