1

Why do private firms hold less cash than public firms?

International evidence on cash holdings and borrowing costs

Abstract

We contend, with aid of a stylized model, that high borrowing costs can overwhelm

precautionary motives and induce low cash holdings in private firms. Supportive of our

hypothesis, we find European private firms hold less cash than public firms and this pattern

reflects differential borrowing costs. Results are robust to endogeneity concerns and reveal

private firms use cash flow to pay-down existing debt instead of building cash reserves; private

firms are also less sensitive to precautionary reasons for holding cash. Further, stronger creditor

rights and debt market development lead to convergence in cash policies of public and private

firms.

2

I. Introduction

This paper investigates how fundamental differences between public and private firms affect

their cash policies. Keynes (1936) predicts that cash holdings may be beneficial to firms with

limited access to external capital markets because they can be used to finance future valuable

projects or activities.1 Along these lines, Harford, Klasa, and Maxwell (2014) emphasize that

cash reserves are important in reducing refinancing risk for firms with short-term debt.

Additionally, Pinkowitz and Williamson (2001) show that public firms that borrow from banks

rather than public debt markets tend to hold more cash. This line of research suggests that private

firms should hold higher precautionary cash reserves than public firms given that private firms

have limited access to external capital markets, have more short-term debt and are subject to

monitoring by banks. However, contrary to these predictions, empirical evidence using U.S. data

documents that private firms tend to hold less cash than public firms (Farre-Mensa, 2015; Gao et

al, 2013). In this paper, we propose a borrowing cost explanation for the difference in cash

policies of public and private firms and test our hypothesis with cross-country data.

We use a stylized model of firms’ cash retention decisions to develop our hypothesis that

higher opportunity costs of holding cash for private firms can lead to a cash holdings differential

between public and private firms. The rationale is that firms face a trade-off between retaining

cash for future investment opportunities and paying down debt. By reducing debt, the firm

lowers its borrowing costs (we will interchangeably refer to these as costs of cash) – the

downside is that the firm may have less investment capital in future states when it cannot raise

sufficient external borrowing. Hence, when the current costs of debt financing are significantly

high, the costs of holding cash may outweigh the precautionary benefits of holding cash, and a

1 This benefit is described as a precautionary motive for holding cash. Opler et al. (1999), among others, provide

evidence consistent with this precautionary motive.

3

firm may use funds to reduce debt rather than hold cash. Reducing debt today may also increase

future funding capacity and lower restrictive covenants, providing an additional rationale for

using cash to pay-off debt.

The existing literature documents that the cost of debt financing is higher for private

firms than for public firms (Pagano, Panetta, and Zingales, 1998; Saunders and Steffen, 2011).

Saunders and Steffen, for example, find that private firms face higher borrowing costs than

public firms because of the higher costs of information production, lower bargaining power and

the higher likelihood of shareholder-debtholder conflicts. Thus, private firms may not only face

considerable precautionary benefits of holding cash but also substantial costs of holding cash

because of high borrowing costs. This could lead to lower cash balances for private firms than

public firms.

To test our cost of cash hypothesis, we analyze cash policies of private and public

European firms. This approach allows us to make contributions to the literature on corporate

liquidity not only by utilizing a unique comprehensive sample of private firms but also by taking

advantage of cross-country variation in the development of legal and financial institutions.2

Thus, we are able to perform an in-depth analysis of the role of financing frictions in explaining

cash holdings. We start by documenting that, similar to U.S. evidence, European unlisted firms

hold significantly less cash than listed firms.3 These results are robust to alternative

specifications, including instrumental variable models that address endogeneity concerns of

public status, and hold in both matched and unmatched samples.

Next, we show that listed and unlisted firms differ in their propensity to save cash out of

cash flow. Following Acharya et al. (2007), we jointly estimate responses of firms’ cash and debt

2 Unlike U.S. firms, both public and private European firms are required to report their financial data. 3 We use the terms “listed” and “public” (and “unlisted” and “private”) interchangeably.

4

policies to cash flow innovations. Listed firms display a higher propensity to save cash from cash

flow (i.e., their cash flow sensitivity of cash is significantly higher than that of unlisted firms). In

contrast, unlisted firms are more likely to use cash flow to pay off their debt (i.e., their cash flow

sensitivity of debt is significantly higher than that of listed firms). These results hold in both a

matched sample and a sub-sample of firms that change status from listed to unlisted. These

findings are consistent with the hypothesis that holding cash is relatively costlier for private

firms.

We then explore the determinants of the differential in cash holdings across listed and

unlisted firms to provide additional insight into the economic forces that drive corporate

liquidity. Our cross-sectional analysis shows that the differential in cash holdings across listed

and unlisted firms is related to the degree of debt market development. Consistent with our

model’s predictions, in countries with less developed debt markets and high cost of debt

financing, unlisted firms retain significantly less cash than their counterparts in countries with

more developed debt markets. Furthermore, our time-series analysis shows that variation in the

cost of loans within a country is negatively related to cash holding of unlisted firms, and this

relation is stronger in countries with less developed debt markets. As debt markets develop and

the cost of debt financing is reduced, unlisted firms increase their cash holdings and the

differential in cash holdings between unlisted and listed firms shrinks significantly. We note that

using these country-level proxies for the cost of debt financing helps mitigate endogeneity

concerns in firm-level regressions.

Our findings support the hypothesis that private firms retain relatively little cash because

of the high opportunity costs implicit in the steep borrowing costs they face. While the traditional

(perfect market) approach suggests that cash should be viewed as negative debt, this is not

5

necessarily the case in the presence of market imperfections.4 As our findings illustrate, firms’

preference for higher cash versus lower debt depends on borrowing costs as indicated, for

instance, by whether the firms are listed or unlisted.

Our results are robust to controlling for agency costs (Gao et al, 2013) and unlikely to be

due to the difference in selective disclosure between public and private firms (Farre-Mensa,

2015). We also show that the cash holdings differential between private and public firms is

unlikely to be due to the differences in firms’ hedging needs (Acharya et al (2007)). Our results

highlight the role of financing frictions in affecting firms’ liquidity management decisions. Most

importantly, they suggest that the relation between financing frictions and cash holdings is non-

monotonic. On the one hand, firms with the greatest access to the capital markets, such as large

U.S. public firms with high credit ratings, hold relatively little cash because the benefits are low

(Opler et al., 1999). On the other hand, firms that lack access to public equity markets and face

high cost of debt financing, such as private firms, hold relatively little cash because the

opportunity costs of holding cash are high.

Another important difference in cash policies across listed and unlisted firms that we

observe is related to the precautionary motive for holding cash. Prior literature shows that public

firms with high cash flow volatility (e.g., Opler et al., 1999; Bates, Kahle and Stulz, 2009) are

more likely to hold cash. The precautionary motive predicts that cash holdings allow these firms

to deal with adverse shocks when access to capital markets is costly. While we also find that

listed firms with high cash flow volatility hold more cash, this does not appear to be the case for

unlisted firms. Similarly, we find that the cash differential between listed and unlisted firms is

greater in industries with better growth opportunities: while listed firms increase cash holdings in

4 Acharya et al. (2007), offers a model in which cash flow is stochastic to show that cash is not equivalent to

negative debt. Our model provides another context in which cash is not equivalent to negative debt.

6

response to an improvement in growth opportunities, unlisted firms do not. While the

precautionary motive suggests that firms with better growth opportunities hold more cash (Bates

et al., 2009), this does not appear to be the case for private firms.5 Costly debt financing may

override the benefits to private firms from accumulating cash for precautionary reasons.

The remainder of the paper is organized as follows. Section II presents our model and

develops the cost of cash hypothesis. Section III describes the data and sample selection

procedure. Section IV examines the differences in cash holdings between listed and unlisted

firms. Section V investigates the cash flow sensitivity of debt and cash for listed and unlisted

firms. Section VI examines the determinants of the differences in cash holdings between listed

and unlisted firms. Section VII investigates the precautionary motives for holding cash by listed

and unlisted firms. Section VIII concludes the paper.

II. Costly External Financing and Cash Holdings

Costly external financing affects both the costs and benefits of holding cash (Almeida et

al., 2004). The extant research emphasizes the benefits of holding cash when the cost of external

capital is high due to asymmetric information and agency problems. Holding cash is beneficial

because firms can invest in valuable future projects and activities without the substantial costs

and uncertainty associated with raising new external capital. The literature refers to this as the

precautionary motive for holding cash. Existing empirical literature provides evidence that

factors related to the precautionary benefits of cash such as growth opportunities and cash flow

volatility positively impact cash holdings (e.g., Opler et al., 1999).

5 Firms with better growth opportunities hold more cash because adverse shocks and financial distress are costlier

for them.

7

The opportunity cost of cash for financially constrained firms has received less attention

in the literature than have the benefits of holding cash. Firms’ executives, however, emphasize

costs of holding cash when the cost of external capital is high (Lins, Servaes, and Tufano, 2010).

Lins et al, in a survey of chief financial officers from 29 countries, find that factors related to the

cost of cash are the second most important determinant of cash holdings. As we would expect,

managers recognize that holding cash entails an incremental cost that is equal to the cost of debt

financing (or interest paid on debt) minus the rate of return that the firm earns on its cash

holdings.6 Thus, a firm with a high cost of debt could find it optimal to use funds to pay down

debt.7 Also, paying down debt can enhance the firm’s future debt capacity, especially if the

retired debt included restrictive covenants.

The downside of not retaining cash, however, is that in some future states of the world

the firm may find it costly or difficult to raise capital, e.g., if it lacks sufficient collateral, and

may find itself largely dependent on its cash holdings to undertake investment opportunities. We

illustrate this trade-off using a simple stylized model. We develop testable implications of the

model and, in particular, explore the implications of debt market and banking sector

development on firms’ cash holdings.

A. A Simple Model of Cash Holdings and Cost of Debt

In this sub-section, we propose a simple stylized model to illustrate the tradeoff firms can

face between cash retention and debt pay-down when borrowing is costly and there is the

6 The rate of return that a firm earns on its cash holdings is the risk free rate minus a carry cost (Bolton et al (2011).

A carry cost may arise from a difference between the corporate and personal income tax rates: Corporations

generally face higher tax rates on interest earned on its cash holdings than individuals (Graham, 2000; Faulkender

and Wang, 2006) Also, a carry cost may arise when a firm retains some portion of its cash holdings in non-interest

bearing accounts (Azar et al, 2016) These carry costs are not related directly to the degree of financial constraints.

We investigate the opportunity cost of holding cash that is directly related to the degree of financial constraints 7 Further, paying off debt and decreasing leverage may increase borrowing capacity in the future and reduce

financial constraint.

8

possibility of being constrained in future periods. The model applies most directly to the case of

firms that are largely dependent on external debt financing.

The model has three relevant dates 𝑡 = 0, 1 & 2. A firm decides on its cash retention and

debt pay down decisions on date 0, prior to learning about its investment opportunities. On date

1, the firm learns of its opportunities and makes its investment decision using its retained cash

and additional borrowing to finance investment. Date 2 is the terminal date on which project cash

flows are realized and debt is repaid. The firm begins with cash holdings of 𝐶0 on date 0. The

cash holdings are the result of cash inflow from past investments and the firm’s prior cash

holdings. On this date, the firm has pre-existing debt with a face value of 𝐵0. The debt is due on

date 2. All market participants are taken to be risk-neutral.

There are two sources of friction in the model that constrain borrowing and affect the

firm’s choice of cash that is retained from date 0 to date 1. The first friction is the cost of

borrowing 𝑔. This is the borrowing cost associated with asset verification and monitoring of

loans and is incremental to the time-value-of-money, which we normalize to zero for simplicity.

The borrowing costs are taken to be proportional to the loan amount i.e., the cost of borrowing is

𝑔𝐵, if 𝐵 is the loan amount. Since our analysis is centered on the cash retention/pay-down

decision on date 0, we simplify exposition by assuming that the cost 𝑔 applies only to debt that is

held between date 0 and 1. Further, it is assumed that 𝐵0 > 𝐶0, so that the firm is unable to fully

payoff its outstanding debt on date 2 in the absence of new and profitable investment. As a

result, if existing debt cannot be junior to new debt issued on date 1, the firm will be unable to

raise additional capital unless lenders are persuaded that the firm has a sufficiently profitable

investment opportunity.

9

The second friction we consider is the possibility that, even when the firm has a positive

NPV investment, it may be unable to raise additional capital because potential lenders/banks are

unable to verify the existence of a profitable opportunity. The likelihood that a firm with good

investments is unable to raise external financing is denoted by 𝛼, and provides a measure of

banking and debt market development (or, more specifically, its absence) in the environment in

which the firm operates.8 We assume that the cost of borrowing 𝑔 is an increasing function of 𝛼

and of a firm specific ‘opaqueness’ measure denoted by 𝜃, which is intended to capture the cost

of firm information acquisition and monitoring by lenders. Hence, the cost of borrowing can be

expressed as a function 𝑔(𝜃, 𝛼). The advantage of separate parameters, 𝛼 and 𝜃, to represent

debt-market development and firm specific transparency is that it allows us to develop

predictions on the interaction between firm specific transparency and debt-market development

on cash policies e.g., the possibly differential effect of debt market development on public versus

private firms.

To develop our predictions, we assume a plausible functional form for borrowing costs:

𝑔(𝜃, 𝛼) = 𝑀𝑎𝑥{𝜃𝛽𝛼𝛾, 𝜎} where 𝛽 > 1, 𝛾 > 1 and where 𝜎 denotes some minimum cost of

borrowing that all firms face. For now, we will assume that 𝜃𝛽𝛼𝛾 > 𝜎 and that 𝑔(𝜃, 𝛼) = 𝜃𝛽𝛼𝛾;

we will explicitly take note of the minimum cost 𝜎 in the context of firms with rather low

monitoring and information costs e.g., firms that are listed and well established. The advantage

of a Cobb-Douglas-type functional form 𝜃𝛽𝛼𝛾 is that it implies that 𝜕𝑔

𝜕𝜃> 0,

𝜕𝑔

𝜕𝛼 > 0 and the

cross-partial 𝜕2𝑔

𝜕𝜃𝜕𝛼> 0. The positive cross-partial captures the notion that, when the debt

market/banking sector is less developed (i.e., α is larger), there is greater divergence in the

8 For simplicity, we assume that 𝛼 is reflective of financial market development and unaffected by firm

characteristics 𝜃. Assuming 𝛼 to be a function of 𝜃 leads to more involved expressions but does not affect our

predictions, as long as the effect of 𝜃 on 𝛼 is sufficiently small.

10

overall borrowing costs between more and less transparent firms. In other words, public firms,

with their greater transparency and greater monitoring by, say, institutional investors are

impacted less by weaker creditor rights and debt market development than private firms. An

assumption of this nature is necessary to obtain specific predictions – that we subject to

empirical tests – on the interaction between firm-specific transparency and debt market

development.

In making its cash retention decision on date 0, the firm takes account of future capital

needs to fund investment on date 1. It is assumed that there are no productive investments on

date 0 and, further, that the firm does not distribute cash in the form of dividends/repurchases

since it faces the risk of being cash constrained on date 1. Hence, on date 0, the firm limits its use

of cash 𝐶0 to paying down its outstanding debt, while retaining the rest. The amount of cash used

to pay down debt is denoted by 𝛥, with 𝐶0 − 𝛥 representing the cash retained. The benefit of

paying down debt is the reduction in the borrowing costs by 𝑔𝛥 between dates 0 and 1.

We next turn to the investment opportunity available to the firm on date 1. We assume

that there is a probability 𝜇 > 0 that the firm has a profitable investment opportunity, while with

probability 1 − 𝜇 it has no positive NPV investment. While the firm knows whether it has such

an opportunity, this is not necessarily apparent to outside providers of capital. As indicated

above, a source of friction is that even if the firm has a profitable investment opportunity, there is

still a probability 𝛼 that it will be unable to raise additional financing; while with probability 1 −

𝛼 outside investors will be able to verify the opportunity and provide sufficient funds to invest

optimally. The profitable opportunity is taken to be such that an investment of 𝐼 produces a cash

11

payoff of 𝑓(𝐼) on date 2, where 𝑓′ > 0, 𝑓′′ < 0 and 𝑓′′′ ≤ 0. With this investment opportunity,9

the optimal investment level is 𝐼∗, where 𝑓′(𝐼∗) = 1.

We analyze the firm’s cash retention decision on date 0 by examining the payoff

implications in the three possible states on date 1: (1) With probability 𝜇(1 − 𝛼), the firm has a

project and can invest optimally; (2) with probability 𝜇𝛼, the firm has a project but is unable to

raise additional financing: this is the state in which cash retention proves valuable; and (3) the

firm has no project and cannot raise additional financing.

In state (1), the firm is able to raise sufficient funds to invest optimally. Hence, its payoff

on date 2, assumed to be positive, can be expressed as:

𝜋1 = 𝑓(𝐼∗) − (𝐵0 − 𝛥)(1 + 𝑔) + 𝐶0 − 𝛥 − 𝐼∗ (1)

= 𝑓(𝐼∗) − (𝐵0)(1 + 𝑔) + 𝐶0 + 𝑔𝛥 − 𝐼∗ > 0 (1’)

In expression (1) above, the debt pay-down on date 0 is 𝛥; 𝐶0 − 𝛥 is the retained cash, while

(𝐵0 − 𝛥)(1 + 𝑔) is the portion of the date 0 debt that will be repaid on date 2. To invest 𝐼∗, the

firm raises external financing of 𝐼∗ − 𝐶0 + 𝛥. As noted earlier, for expositional ease we have

assumed that there is no incremental borrowing cost between dates 1 and 2. As expression (1’)

makes evident, the advantage of paying down 𝛥 is that it results in a saving of 𝑔𝛥 in terms of

payoff on date 2.

In state (2), lenders are unable to verify that the firm has a profitable project. Given our

assumption that 𝐵0 > 𝐶0, lenders will not make new loans without being able to verify that the

firm has access to a profitable project (since they would not expect to be repaid otherwise). The

firm is, therefore, able to invest only its available resources 𝐶0 − 𝛥 in the investment

opportunity. We will assume that the firm’s payoff on date 2 is positive after investing in the

9 Note that the optimal investment level 𝐼∗is unaffected by the borrowing cost 𝑔 on account of our simplifying

assumption that the borrowing costs apply only to existing debt between dates 0 and 1.

12

project, since it would otherwise have no incentive to invest. The firm’s payoff can be expressed

as:

𝜋2 = 𝑓(𝐶0 − 𝛥) − (𝐵 − 𝛥)(1 + 𝑔) (2)

Finally, in state (3), the firm has no investment opportunity and its payoff on date 2 is 0 since its

existing cash holdings are below its date 2 obligations.

We can now determine the firm’s optimal choice of 𝛥 on date 0, assuming that the firm

(or manager) seeks to maximize its expected value as of date 2. The date 0 objective function can

be expressed as:

𝑉 = 𝜇(1 − 𝛼)[𝑓(𝐼∗) − 𝐵(1 + 𝑔) + 𝐶0 + 𝑔 𝛥 − 𝐼∗] +

𝜇 𝛼 [𝑓(𝐶0 − 𝛥) − (𝐵 − 𝛥)(1 + 𝑔)] (3)

In expression (3) above, the two terms represent the probabilities × [expected payoffs] in state 1

and state 2 respectively. The payoffs in the two states are as indicated in equations (1’) and (2).

To determine the optimal value of 𝛥 that maximizes the objective function (3), we obtain

the first-order-condition:

𝛿𝑉

𝛿𝛥= 0 = 𝜇(1 − 𝛼)[𝑔 ] + 𝜇𝛼[−𝑓′(𝐶0 − 𝛥) + (1 + 𝑔)], (4)

which gives,

[1−𝛼

𝛼] 𝑔 + (1 + 𝑔) = 𝑓′(𝐶0 − 𝛥), (5)

or,

(1

𝛼− 1) 𝑔 + (1 + 𝑔) = [

𝑔

𝛼] + 1 = 𝑓′(𝐶0 − 𝛥). (5’)

Substituting for the function 𝑔(𝜃, 𝛼) = 𝜃𝛽𝛼𝛾, we can express (5’) as:

𝜃𝛽𝛼𝛾−1 + 1 − 𝑓′(𝐶0 − 𝛥) = 0. (6)

13

Equation (6) represents the trade-off between cash retention and debt pay-down: A unit

of cash holdings imposes an expected marginal cost of 𝜃𝛽𝛼𝛾−1, and yields an expected positive

value of 𝑓′(𝐶0 − 𝛥) − 1.

To obtain the effect of the firm opaqueness and difficulty of monitoring (𝜃) on the firm’s

cash retention choice we take the derivative of equation (6) with respect to 𝜃 (keeping 𝛼 fixed):

𝛽𝜃𝛽−1𝛼𝛾−1 + 𝑓′′(𝐶0 − 𝛥). (𝜕𝛥

𝜕𝜃) = 0. (7)

Since 𝑓′′(𝐶0 − 𝛥) < 0, it follows that (𝜕𝛥

𝜕𝜃) > 0. This leads to our first prediction:

Prediction 1: An increase in firm opaqueness and cost of monitoring, will be associated with a

greater pay-down of debt 𝛥 and lower cash retention 𝐶0 − 𝛥.

We next consider the effect of debt market and banking development on the cash

retention decision by firms. In particular, we are interested in the differential effect that debt

market and banking sector development can have on firms that are more or less transparent. We

do this by showing that 𝑠𝑖𝑔𝑛 [𝜕2𝛥

𝜕𝜃𝜕𝛼] > 0 under our assumptions. The positive sign implies that

the lack of debt market development will have a greater marginal impact on the reduction in the

cash holdings of more opaque firms. Or, conversely, stronger debt market development will have

a greater marginal effect on the cash holdings of less transparent firms.

We proceed by taking the derivative of equation (7) with respect to 𝛼:

𝛽. (𝛾 − 1). 𝜃𝛽−1𝛼𝛾−2 + 𝑓′′(𝐶0 − 𝛥). (𝜕2𝛥

𝜕𝜃𝜕𝛼) − 𝑓′′′(𝐶0 − 𝛥)

𝜕𝛥

𝜕𝜃

𝜕𝛥

𝜕𝛼 = 0 (8)

Since the project is assumed to be such that: 𝑓′′ < 0 and 𝑓′′′ ≤ 0, it follows that (𝜕2𝛥

𝜕𝜃𝜕𝛼) > 0.

We can state:

14

Prediction 2: An increase in debt market and banking sector development will result in a

relatively greater increase in the cash holdings by more opaque and harder-to-monitor firms,

relative to cash holdings by more transparent and easier-to-monitor firms.

We turn next to the case of firms that have low monitoring costs and their borrowing

costs are such that 𝑔 = 𝜎 (i.e., where borrowing costs are such that 𝜃𝛽𝛼𝛾 < 𝜎). It follows that

equation (5’) in this case becomes:

𝜎𝛼−1 + 1 − 𝑓′(𝐶0 − 𝛥) = 0. (5′′)

Taking derivatives with respect to 𝛼, we obtain:

−𝜎𝛼−2 + 𝑓′′(𝐶0 − 𝛥). (𝜕𝛥

𝜕𝛼) = 0. (9)

Since, 𝑓′′ < 0, it follows from equation (9) that for firms with low monitoring and borrowing

costs, we can have (𝜕𝛥

𝜕𝛼) < 0. In other words, for these firms an improvement in the debt market

and banking-sector development (i.e., lower 𝛼) can lead to less cash retention. This is the

opposite of the effect for firms that are costly to monitor and have relatively large borrowing

costs. The reason is that as the likelihood of the firm being able to finance its project increases

(i.e., 1 − 𝛼 increases), it tends to hold less cash since its need for precautionary cash-holding

decreases; while its borrowing costs are relatively insensitive to debt market development. We

can state:

Prediction 3: For firms that are more transparent and have relatively low monitoring and

borrowing costs, an increase in debt market and banking sector development can lead to a

decrease in the firms’ cash holdings.

15

B. Discussion of the Model and Implications for Private and Public Firms

In our empirical analysis, we investigate whether the cost of borrowing can explain firms’

cash policies and, in particular, whether it can account for the differences in cash holdings

between public and private firms. We refer to this interchangeably as the cost of cash or the cost

of borrowing hypothesis.

Given the information production and monitoring functions of the stock market (Dow and

Gorton, 1997; Subrahmanyam and Titman, 1999; and Edmans, 2009), private firms are likely to

have higher ‘opaqueness’ measure, 𝜃, than public firms and thus, according to our model, higher

borrowing costs. Existing empirical literature, indeed, suggests that the cost of debt financing is

higher for private firms than for public firms (Pagano, Panetta, and Zingales (1998) and Saunders

and Steffen (2011)). Saunders and Steffen (2011), for example, find that private firms face

higher borrowing costs than public firms because of the higher costs of information production,

lower bargaining power and the higher likelihood of shareholder-debtholder conflicts. Then,

from Prediction 1, we would expect private firms to have lower cash holdings and to be more

likely to apply cash flow to pay down existing debt.

There is also literature that indicates that lower levels of creditor protection and debt

market development are associated with higher cost of debt (Qian and Strahan (2007) and Bae

and Goyal (2009)), which is captured in our model by parameter α. Interestingly, our model

predicts that creditor protection and debt market development can have a differential effect on

the cash policies of public and private firms. In particular, debt market development is expected

to have a relatively greater impact on the cash holdings of less transparent firms with high

borrowing costs such as private firms (Prediction 2). For highly transparent firms with relatively

low borrowing costs such as public firms, debt market development may actually lead to a

reduction in their level of cash holdings (Prediction 3). Hence, while private firms are expected

16

to increase cash holdings when the debt market is more developed — the impact on cash

holdings of public firms will be less positive and could even be negative. Overall, we expect a

convergence in the cash polices of public and private firms when creditor rights are stronger and

the debt market is more developed.

We note that our model is consistent with the precautionary arguments for accumulating

cash, as it follows directly from the model that a firm with high growth opportunities will hold

more cash. A more productive investment means that 𝑓′(𝐶0 − 𝛥) becomes larger. In

equilibrium (see equation 6, for instance) this implies that firms will choose a lower 𝛥 i.e., there

will be more cash retention by all firms. However, it also follows from the model that the effect

will be muted when borrowing costs are high (as is the case for private firms) that they have a

strong incentive to reduce borrowing costs.10

III. Data Description

In this section, we describe our data and sample selection procedure.

A. The Sample Selection

Our primary data source is the 2011 version of Amadeus, by Bureau van Dijk. This

database provides balance sheet and income statement items for a set of European firms from

1996 to 2011. An important advantage of Amadeus is that it includes data for a comprehensive

set of public and private firms. This advantage is made possible because European law requires

10 To see this, compare a private firm that retains cash 𝐶𝐿 with an otherwise equivalent public firm that retains 𝐶𝐻,

where 𝐶𝐻 > 𝐶𝐿 . Given strict concavity of the investment payoff function we have 𝑓′(𝐶𝐿) > 𝑓′(𝐶𝐻). Now, assume

that the both firms experience a similar increase in marginal productivity of 𝜏 > 0 (i.e., the marginal outputs are

shifted up by a constant). In equilibrium (see equation 6), this will lead to an increase in cash retention for both firms

such that: 𝑓′(𝐶𝐿 + 𝛿𝐿) = 𝑓′(𝐶𝐿) + 𝜏 and 𝑓′(𝐶𝐻 + 𝛿𝐿) = 𝑓′(𝐶𝐻) + 𝜏. Given strict concavity, it follows that 𝛿𝐻 > 𝛿𝐿. This is since, given that 𝑓′(𝐶𝐿) > 𝑓′(𝐶𝐻), a larger change in 𝐶𝐻 has a similar marginal effect as a smaller

change in 𝐶𝐿 .

17

both public and private firms to report financial statements. The data are collected from each

nation’s official public body in charge of collecting the annual financial statements in its country,

and always come from the officially filed and audited accounts.

The Amadeus dataset is divided into three parts. The first contains the largest firms in the

database, the second contains the next largest, and the third contains the remaining firms. Our

sample comes from the first part of the dataset – the largest firms. The dataset includes a flag for

whether the company is listed on a major stock exchange. All other firms are classified as

unlisted. The dataset reports only contemporaneous information rather than historical

information for this variable. We use historical Amadeus DVDs to track changes in listing status

over time. We use the variable “Legal Form” to exclude unlimited partnerships, sole

proprietorships, cooperatives, foreign companies, foundations, and government enterprises.

As in Giannetti (2003), we exclude Eastern European economies since the quality of the

accounting data provided for these economies is poor. We exclude firm-years with total assets

less than 10 million U.S. dollars, and exclude financial and miscellaneous firms (US SIC-

equivalent codes 60-69 and 89). With a few exceptions, we set a variable to missing if its

observations are within its top or bottom percentile, to avoid the effect of outliers. The

exceptions are standard deviation of cash flows and balance sheet items, such as cash scaled by

total assets. In these cases, we set a variable to missing if its values are non-positive or its

observations are within the top percentile. These filters result in 1,004,674 firm-year

observations.

We complement firm-level data with country indexes of financial and legal development.

We measure debt market development using private credit to GDP from Djankov, McLeish and

Shleifer (2007) and the index of creditor rights from La Porta, Lopez-de-Silanes, Shleifer, and

18

Vishny (1998). To gauge the degree of shareholder rights, we use the anti-self-dealing index

from Djankov et al. (2008), which measures how difficult it is for minority shareholders to

thwart the consumption of private benefits by controlling parties. Djankov et al. (2008), argue

that self-dealing is the central problem of corporate governance in most countries. Following

Dermiguc-Kunt and Levine (1996) and Love (2003), we compute an index of stock market

development that equals the sum of standardized market capitalization to GDP, total value traded

to GDP, and turnover (total value traded to market capitalization). We obtain each of the

elements of this index from the World Bank.

B. The Matching Procedure

Only 2.5% of the firm-year observations in our sample are listed firms. All other firms

are unlisted. To make the samples of listed and unlisted firms more comparable in size, we match

listed firms to unlisted firms based on country, industry code, and total assets. We keep our

matching criteria simple to allow for comparisons between public and private firms across

multiple characteristics.

In order to match each listed firm to an unlisted firm, we first consider all listed firms in

2008, choosing this year because it contains the largest number of firms for any given year in our

sample. We then exclude the largest listed companies (total assets of the company exceeds total

assets of the largest unlisted company in the country by 20 million U.S. dollars or more) as these

companies are likely to have easy access to international financial markets and are less likely to

be subject to the constraints imposed by domestic markets (see Giannetti, 2003). Next, we

require exact matches on country and industry code and the closest possible match on total

assets, measured as of 2008. The matching is done without replacement. Our matched sample

includes only the largest private corporations. We perform most of the tests using the matched

19

sample. To the extent that the largest private companies are more likely to behave like public

companies, this procedure may bias our results towards finding no differences between public

and private firms. For robustness, we also perform some of the tests using the full (unmatched)

sample.

C. Descriptive Statistics

We present descriptive statistics in Table 1 for the matched sample of listed and unlisted

firms. The table contains total assets in millions of USD and various balance sheet items as

percentage of total assets. As indicated, unlisted firms hold significantly less cash and cash

equivalents than listed firms: unlisted firms hold 9% in cash as a proportion of total assets, while

listed firms hold 14%. This difference of 5% is statistically significant, and is consistent with

Prediction 1, that private firms will hold less cash because their borrowing costs are higher. We

confirm this univariate result in various tests below. Interestingly, unlisted firms also hold higher

short-term debt, a variable typically associated with higher levels of cash (see Falato,

Kadyrzhanova and Sim, 2013; Harford, Klasa, and Maxwell, 2014). We also note unlisted firms

are somewhat smaller than their listed counterparts even in our matched sample.

In table 2, we present cash holdings across countries. Unlisted firms hold less cash in

most countries, and this difference is statistically significant in 11 out of 16 countries. This

pattern is also very consistent over time. In Figure 1 we plot average cash holdings over time for

listed and unlisted firms, and find that unlisted firms hold less cash than listed firms for every

year in the sample period.

20

IV. Differences in Cash Holdings between Public and Private Firms

A. Baseline Results

In this section, we test whether public firms hold less cash than private firms using

regression analysis. We consider the following model:

Cash Holdingsit = β*Listedit + δ*Xit + εit , (10)

where Cash Holdings is cash and cash equivalents divided by total assets. Listed is an indicator

variable for the firm being listed on a major stock exchange in the country. We include a set of

firm-level control variables, as well as country, industry, and year dummy variables (X). We

include country dummies to ensure we are measuring within-country differences between listed

and unlisted firms as well as controlling for unobserved time-invariant country effects. We also

include industry and year dummies to control for industry wide factors and time trends that may

affect cash holdings. Standard errors are clustered at the firm level.

The set of firm-level controls includes variables that have been found to determine cash

holdings in earlier studies (e.g., Opler et al., 1999, and Bates et al., 2009). Leverage is measured

as total debt divided by the sum of total debt and shareholder funds. Total debt is the sum of

long-term debt plus short-term loans. Size is measured as the log of total assets, where assets are

in USD. Decreasing marginal benefits to holding cash would predict a negative relation between

size and cash. We use sales growth to proxy for growth opportunities. Later in the paper, we

consider alternative proxies for growth opportunities. Firms with growth opportunities may

prefer to hold more cash for precautionary reasons to avoid foregoing growth opportunities due

to financing difficulties. Cash flow to assets is operating cash flow divided by lagged assets.

Firms with high cash flow may be able to accumulate more cash.

21

We compute the standard deviation of cash flows from the current and the past four

values of annual cash flow to assets ratio. If two or more cash flow to assets ratios are missing,

then the variable is set to missing. We expect firms with high cash flow risk to hold more cash

for precautionary reasons. Investment in tangible assets is the change in tangible fixed assets

divided by lagged assets. To the extent that investment in tangible assets increases debt capacity,

it may reduce the demand for cash. Investment in intangible assets is the change in intangible

fixed assets divided by lagged assets. Investment in intangible assets tends to be associated with

higher levels of information asymmetry, and greater difficulty accessing external capital markets.

As a result, it may increase demand for cash retention. Net working capital is current assets

minus current liabilities minus cash divided by lagged assets. Net working capital consists of

assets that substitute for cash, and thus we expect a negative relation between cash and net

working capital. Finally, firm age is observation year minus year of incorporation. More mature

firms typically have more stable cash flows and lower growth opportunities and require less

cash.

Results of the analysis are presented in Panel A of Table 3. We continue to find that

unlisted firms hold less cash than listed firms when controlling for other determinants of cash

holdings. For the matched sample, the coefficient is 0.033 and is statistically significant at the

1% level, indicating that unlisted firms hold 3% less of their assets in cash than listed firms. This

is economically significant since it indicates that cash holdings of listed firms are more than one

third larger than those of unlisted firms that hold about 9% of their assets in cash. This result is

similar for the unmatched sample. The coefficients on the firm-level control variables are as

expected and consistent with previous studies.

22

We should note that the analysis in Panel A of Table 3 does not control for the dividend

policies of listed and unlisted firms. This is because Amadeus excludes dividend variables.

Previous research, such Opler et al. (1999), however, shows that firms that pay dividends hold

less cash. To investigate whether differences in dividend policies explain the difference in cash

holdings between listed and unlisted firms, we obtain data on dividend payouts for listed firm

using Osiris, another dataset distributed by Bureau van Dijk with extended coverage of financial

data for listed firms. Panel B of Table 3, compares cash holdings of listed firms that pay

dividends to cash holding of unlisted firms. If higher propensity of unlisted firms to pay

dividends explains our results in Panel A, then we should find no difference in cash holdings

between listed firms that pay dividends and unlisted firms. This is not the case however: both

listed firms that pay dividends and listed firms that do not pay dividends hold more cash than

unlisted firms.11

Our results also suggest that low cash holdings of unlisted firms are not due to the

inability of the private firms to generate enough cash from operations because we control for

cash flow. This is different from Denis and Sibilkov (2010), who show that some public firms

with limited access to external markets exhibit low cash holdings because of persistently low

cash flows.

B. Instrumental Variable Models

Our matched sample alleviates concerns that our results are driven by the endogeneity of

the listing status. However, matching can only be done on observable characteristics, and private

and public firms may differ along unobservable characteristics. To address this concern, we

employ a two-stage least squares regression and a treatment effects model. Our instrumental

11 See Michaely and Roberts (2012), for a detailed discussion of dividend policies of private firms.

23

variable is the percentage of firms in a country-region that are listed.12 A rationale is that the

presence and preferences of local investors could affect firms’ decision to go public (investors

often exhibit a ‘home bias’ i.e., have a preference to invest locally). We believe the instrument to

be valid since it is expected to be positively related to the likelihood of a firm being listed

(relevance), while there is no clear reason for the instrument to be directly related to cash policy

(exclusion restriction).

Table 4, Panel A presents results for the two-stage least squares regression using the

matched sample. In the first stage, we estimate a linear probability model in which the dependent

variable is an indicator variable for the firm being listed. As indicated, the proportion of listed

firms in the region within a country is estimated with a positive and significant coefficient at the

1% level. The second stage shows that listed firms tend to hold more cash. The coefficient on the

instrumented listed variables is 0.079 and is statistically significant at the 1% level. The

coefficient suggests that a listed firm holds on average 8% more cash as a percentage of assets

than an unlisted firm. This coefficient is about twice as large as the coefficient on comparable

OLS regression.

To investigate further robustness of our results, in Table 4, Panel B, we estimate the

treatment effects model using the two-step Heckman procedure. In the first step, we estimate a

probit model for the probability of being public, and use the proportion of listed firms in the

region as an instrument. In the second step, we estimate the determinants of cash holdings, while

accounting for possible self-selection bias regarding the choice of private/public status. We

continue to find that unlisted firms hold less cash than listed firms when controlling for other

determinants of cash holdings.

12 We obtain the region of the firm’s location from Amadeus. Regions are territorial units that comprise a country.

24

In un-tabulated results we use, in the spirit of Saunders and Steffen (2011), distance to

the country’s main stock exchange as an alternative instrument for the choice of being public.

Stock exchanges are located in the country’s financial capital, and proximity should facilitate

access to public capital markets in that it facilitates how firms match with investors and

underwriters during the going public process. In addition, it is plausible that the instrument

satisfies the exclusion restriction since we would not expect distance to a country’s financial

capital to have a direct effect on cash holdings. To reduce skewness, we use log(distance). We do

this for both the instrumental variables and the treatment effects models. We obtain results that

are qualitatively similar to the results discussed above, with listed firms retaining greater cash

holdings.

As we have discussed, the positive cash holdings differential between listed and unlisted

firms would not be expected within Keynes’s (1936) framework that predicts that cash holdings

serve a precautionary function for firms with limited access to external capital markets (such as

unlisted firms) because cash holdings can help those firms seize valuable projects or activities in

the future. Opler et al. (1999), among others, provide evidence consistent with the precautionary

motive for holding cash among listed firms. Our results indicate, however, that the precautionary

motive tends to be dominated by the high costs of debt for unlisted firms inducing them to hold

little cash compared to listed firms (Prediction 1). In the sections below, we provide more direct

evidence consistent with the cost of borrowing hypothesis.

25

V. The Cash Flow Sensitivity of Debt and Cash for Public and Private

Firms

In this section we provide further support for our cost of borrowing hypothesis. Our

hypothesis implies that private firms will use cash flow to pay off debt, in lieu of accumulating

cash, because private firms bear a higher cost of debt than public firms. To test this hypothesis

directly, we investigate the cash flow sensitivity of debt and cash for listed and unlisted firms.

We expect the cash flow sensitivity of debt to be higher for unlisted firms than for listed firms,

and the cash flow sensitivity of cash to be lower for unlisted firms.

In our empirical analysis, we follow Acharya, Almeida, and Campello (2007), and

endogenize debt and cash policies. Specifically, we estimate the following system of equations

using three-stage least squares:

ΔDebtit = α1Listedit + α2CashFlowit + α3CashFlowxListedit + α4ΔCashHoldingsit

+ a5ΔDebtit-1 + δ*Xit + εit , (11)

ΔCashHoldingsit = α1Listedit + α2CashFlowit + α3CashFlowxListedit + α4ΔDebtit

+ a5ΔCashHoldingsit-1 + δ'*Xit + ε'it , (12)

where Listed is an indicator variable for the firm being listed. ΔDebt is change in total debt scaled

by beginning of the period assets. ΔCashHoldings is change in cash scaled by initial of period

assets. CashFlow is cash flow scaled by lagged assets. We also include a set of firm-level control

variables such as firm size, growth opportunities as measured by sales growth, and country,

industry and year dummies (X). Standard errors are clustered at firm-level.

26

Results are presented in Table 5, Panel A. They are consistent with our predictions. In the

regressions of changes in debt we find that higher levels of cash flow are associated with debt

reductions for unlisted firms. This effect is muted for listed firms, as evidenced by the positive

coefficient on the interaction variable that is of similar magnitude to the negative coefficient on

cash flow. For unlisted firms, debt drops by 1.28% for a one standard deviation increase in cash

flow. For listed firms, debt increases by a meager 0.01% for a one standard deviation increase in

cash flow.13

In the regressions of changes in cash, we find that high cash flows are associated with

increases in the cash account for both listed and unlisted firms, and this effect is more

pronounced for listed firms, as evidenced by the positive coefficient on the interaction variable.

Cash holdings of unlisted firms increase by 1.35% for a one standard deviation increase in cash

flows, while the cash holdings of listed firms increase by an additional 0.58% for a total of

1.93%.

In Table 5, Panel B we examine a sample of firms that change listing status to investigate

how a firm’s propensity to use cash flow to pay off debt varies for a firm as it changes its own

listing status. The sample contains firms that switch status from unlisted to listed and vice-versa.

We re-estimate equations 11 and 12, but include firm and year fixed effects. Standard errors are

clustered at the firm level. This analysis allows us to examine the effects of listing status while

keeping firm time-invariant characteristics constant, and mitigates any concerns that our results

are explained by unobservable characteristics. Our results are consistent with previous findings.

We find that firms that switch their status from unlisted to listed and vice-versa, are more likely

13 Cash flow scaled by total assets has a mean of 0.081 and a standard deviation of 0.105. The economic effect for

listed firms is computed as the sum of the coefficients on the cash flow relative to total assets, -0.122 and its

interaction with the indicator variable for listed, 0.123 multiplied by 1 standard deviation of cash flow scaled by

assets.

27

to use cash flow to pay off debt when they are unlisted, and are more likely to use cash flow to

increase their cash holdings when they are listed. These results suggest that propensity to use

cash flow to pay off debt changes with the change in listing status.

In sum, we find that public firms are more likely to use cash flow to accumulate cash,

while private firms are more likely to pay down debt. These results are consistent with our cost

of cash hypothesis.

VI. Cross-Country and Time-Series Variation in Cash Held by Public and

Private Firms

In this section, we explore cross-country differences in the development of legal and

financial institutions and their effect on cash holdings of public and private firms to test the

predictions of our model. In addition, we explore time variation in the cost of loan financing

within countries to provide further support for our hypothesis. We also discuss alternative

explanations for the cash holding differential.

A. Creditor Protection, Loan Market Development and the Differences in Cash Holdings

We first investigate the effect of creditor protection and loan market development on the

difference in cash holdings between public and private firms. Our cost of borrowing hypothesis

predicts that as the cost of private debt financing declines and the cost of holding cash is reduced,

private firms will increase cash holdings and the difference in cash holdings between public and

private firms is likely to shrink (Prediction 2). The latter condition is because private firms with

their greater opaqueness and limited monitoring by, say, institutional investors are more

impacted by weaker creditor rights and debt market development than are public firms.

28

Results of the analysis are tabulated in Table 6. We proxy for loan market development

and the cost of loan financing at the country level with the amount of private credit as a

percentage of GDP from Djankov et al. (2007), and the creditor rights index from La Porta et al.

(1998). In panel A, we partition the sample in countries with private credit to GDP above the

median and countries with private credit to GDP below the median. In panel B, we partition the

sample in firms with the creditor rights index above the median and below the median. For the

sake of brevity, we do not tabulate the results for firm-level control variables. Control variables

are the same as in Table 3.

Results in Panel A show that listed firms hold more cash in both subsamples; however,

this differential is significantly larger when private credit to GDP is low and the cost of loan

financing is high. Listed firms hold 1.3% more cash than unlisted firms when private credit to

GDP is high; and 5.2% more cash when private credit to GDP is low. These coefficients are

statistically different at the 1% level. Results in Panel B are similar. Listed firms hold more cash

in both sub-samples, but listed firms hold relatively more cash in countries with low creditor

rights. Listed firms hold 1.3% more cash in countries where creditor rights are above the median,

and 4.5% more cash where creditor rights are below the median. As with private credit to GDP,

these coefficients are statistically different at the 1% level. Overall, these findings suggest that

cost of cash is an important economic factor that explains corporate liquidity. To that end, they

are consistent with Azar et al (2016) who find that the carry cost of cash, calculated as a product

of rate of return on Treasury bills and a share of firm’s cash holding kept in zero interest rate

accounts, may explain dynamics of corporate cash holdings over time.

29

B. Country-level Variables and Cash Holdings of Public and Private Firms

In this sub-section, we study the determinants of cash holdings of private and public

firms separately to investigate whether debt market development has a positive effect on the cash

holdings of private firms (Prediction 2), and an insignificant or even negative effect on the cash

holdings of public firms (Predictions 2 and 3). These tests allow us to further understand what

drives the cross-country differences-in-differences in cash holdings that we document in the

previous section, and provide additional support for the cost of cash hypothesis.

Results are presented in Panels A and B in Table 7. Columns 1 and 2 report results for the

sub-sample that includes unlisted firms only, while columns 3 and 4 report results for the sub-

sample that includes listed firms. In addition to proxies for the cost of debt financing, in

regressions 2 and 4 we include a proxy for managerial agency problems at the country-level and

a proxy for stock market development in our regressions. We proxy for managerial agency

problems with anti-self-dealing index from Djankov et al. (2008), which measures how difficult

it is to prevent the consumption of private benefits by controlling parties. We measure stock

market development following Dermiguc-Kunt and Levine (1996) and Love (2003). This

measure is based on trading activity and size of the country’s stock market (see section III for

details). Each regression also includes the firm-level determinants of cash holdings reported in

Table 3, year and industry dummies, and the log of gross domestic product as an additional

control variable for cross-country differences. Standard errors are clustered at the firm level.

In Panel A we proxy for cost of debt using private credit to GDP, while in Panel B we use

creditor rights. We find that debt market development is associated with a higher level of cash

holdings for unlisted firms and lower levels of cash holdings for listed firms. In Panel A, the

coefficient on Private-Credit-to-GDP for unlisted firms is 0.022 (statistically significant at the

1% level), while for listed firms, the coefficient is -0.014 (also statistically significant at the 1%

30

level). The results are consistent with private firms finding it expensive to accumulate cash in

countries where private credit availability is scarce and consequently the cost of debt is high,

while public firms will accumulate more cash when debt market development is low for

precautionary reasons. Our results are robust when we use creditor rights index to proxy for the

cost of debt. In Panel B, the coefficient on Creditor Rights is 0.028 for unlisted firms

(statistically significant at the 1% level), and -0.012 for listed firms (also statistically significant).

In regressions 2 and 4 we include two additional control variables: a proxy for managerial

agency problems (anti-self-dealing index), and a proxy for stock market development. The use of

the anti-self-dealing index is motivated by previous research that links managerial agency

problems and cash holdings, though the direction of the relationship is ambiguous. Self-

interested managers may hold excess cash rather than paying it out to shareholders because cash

reduces firm risk and increases managerial discretion (Gao et al., 2013). However, findings in

Dittmar and Mahrt-Smith (2007) and Hartford, Mansi and Maxwell (2008), suggest that firms

with high degree of managerial agency problems actually have smaller cash reserves because

managers dissipate cash quickly.

The use of the stock market development proxy is motivated by the literature that

emphasizes the precautionary benefits of holding cash. Specifically, the precautionary motive

suggests that firms hold more cash to better cope with adverse shocks when external capital

markets are difficult to tap. As the access to external capital markets, such as the stock market,

improves, firms should reduce cash holdings due to a decrease in the benefits of holding cash.

Each regression also includes firm-level determinants of cash holdings reported in Table 3, year

and industry dummies, and the log of gross domestic product as an additional control variable for

cross-country differences.

31

We continue to find that debt market development is associated with a higher level of

cash holdings for unlisted firms. In regression 2 of Panel A, the coefficient on Private-Credit-to-

GDP is 0.025 and statistically significant at the 1% level. For listed firms, the coefficient on

Private-Credit-to-GDP is 0.001 and is no longer statistically significant (regression 4). The

difference in the coefficients between listed and unlisted firms is statistically significant at the

1% level (untabulated). Our results are robust when we use creditor rights index to proxy for the

cost of debt (see Table 7, Panel B).

The anti-self-dealing index and stock market development have an effect on the cash

holdings of listed firms only. This is expected, since managerial agency problems should be a

bigger concern for public firms due to their more dispersed ownership, and only public firms are

directly affected by the stock market development. Similar to Dittmar et al. (2003), we find that

the anti-self-dealing index is negatively related to the cash holdings of listed firms. Additionally,

we find that stock market development is negatively related to the cash holdings of listed firms.

This result is consistent with the precautionary motive for holding cash, and is broadly consistent

with Prediction 3 from our model.

C. Time-variation in the Cost of Debt Financing and Cash Holdings

In this sub-section, we further investigate whether cash holdings of private firms are

related to the cost of debt financing. We explore time-variation in the cost of loan financing.

During our sample period, European countries made a significant effort to integrate capital

markets. Findings in Bekaert et al. (2007), and Gupta and Yuan (2009), suggest that market

integration helps reduce financial constraints and is likely to reduce the cost of external financing

for private firms. To proxy for the cost of loan financing, we use country-level interest rates

32

reported by the European Central Bank.14 These interest rates are calculated using all outstanding

loans with a maturity of 1 year or less made by monetary financial institutions in a given country to

non-financial corporations. We recognize that a country-level measure is a noisy proxy for the

cost of debt financing at the firm-level; however, in its favor, a country-level measure is less

susceptible to concerns of endogeneity. As before, we expect that the effect of the cost of debt on

cash holdings is stronger for private firms compared to public firms (Prediction 2). One reason is

that the firms likely to benefit the most from an improvement in credit market conditions are the

firms that are least transparent and hardest to monitor (e.g., private firms) (Bae and Goyal

(2009)). This is consistent with the well-recognized pattern that a slowing economy tends to

widen credit spreads (e.g., between investment grade and speculative bonds), while spreads

narrow in a growing economy.

Results are reported in Table 8. In Panel A, we regress cash holdings on loan interest

rates, the set of firm-level control variables included in Table 3, and year and firm fixed effects.

We continue to find evidence that cash holdings of private firms are negatively related to the cost

of debt financing. In Panel B we include an interaction variable, to capture the effect of the cost

of loans when creditor rights are high. We interact the cost of loans with an indicator variable for

the country’s creditor rights being above median. We find that the interaction is positive and

significant for private firms, suggesting that the cost of loans becomes less important for

determining cash holdings when creditor rights are high. For public firms, the interaction is

insignificant. Results are generally consistent with prediction 2 from our model.

14 The link to the data is: http://sdw.ecb.europa.eu/browse.do?node=9484266

33

D. Alternative Explanations

In this sub-section, we first investigate whether differences in the hedging needs between

public and private firms explain the difference in cash holdings we document. Acharya et al

(2007) show that firms with high hedging needs may benefit by hoarding cash, while firms with

low hedging needs may benefit by building debt capacity. High/low hedging needs arise when

investment opportunities tend to arrive in low/high cash flow states.

We follow Acharya et al (2007) to measure firms’ hedging needs. For each firm-year in

the sample we compute the correlation between investment opportunities and the firm’s cash

flow. To proxy for investment opportunities, we compute the median three-year-ahead sales

growth rate in the firm’s three-digit SIC industry. This approach relies on the premise that firms’

perceptions of investment opportunities are related to estimates of future sales growth in their

industries and that those estimates, on average, coincide with the ex-post observed data. We also

create two dummy variables: Low hedging needs, which takes the value of one when the

correlation between investment opportunities and the firm’s cash flow is above 0.20, and High

hedging needs, which takes the value of one when the correlation between investment

opportunities and the firm’s cash flow is below - 0.20. Appendix 1 presents the results of the

analysis. We run a specification similar to the ones presented in Table 3 additionally controlling

for the hedging needs. We continue to find that listed firms hold more cash than unlisted.

In a contemporaneous paper to ours, Farre-Mensa (2015) argues that in the US private

firms hold less cash than public firms because of differences in disclosure requirements. In the

US, private firms can selectively disclose confidential information to private investors, while

public firms cannot. Farre-Mensa contends that these differences in disclosure requirements

make public firms susceptible to mispricing, which results in public firms hoarding more cash to

34

optimize the timing of equity issues. Private firms are not susceptible to the same miss-valuation

waves, and do not need to hoard as much cash.

We highlight that in Europe (1) both private and public firms have an obligation to report

financial statements, resulting in a narrower gap in disclosure between private and public firms,

and (2) the difference in disclosure requirements for private and public firms does not vary

across the countries in our study. As a result, our cross-country evidence that private firms hold

less cash in countries with less developed private debt markets is unlikely to be explained by

differences in disclosure requirements. Nor are disclosure requirements likely to explain the

finding that the difference in cash holdings between private and public firms is wider for

countries with less developed private debt markets.

Overall, the results in this section highlight the fact that financing frictions play an

important role in explaining firms’ cash policies. The previous literature demonstrates that public

firms that are financially constrained accumulate more cash than financially unconstrained public

firms because the precautionary benefits of holding cash are lower for unconstrained firms (e.g.,

Opler et al., 1999). Our results, however, suggest that in extreme cases, when external financing

is significantly costly, firms accumulate relatively little cash because the opportunity cost of

holding cash is high. Thus, the relation between cash holdings and financing frictions is non-

linear. Private firms with limited access to external markets and with a high cost of external

financing behave similarly to financially unconstrained public firms with low financing costs,

and both accumulate relatively little cash, albeit for different reasons, while constrained public

firms accumulate relatively high cash reserves. The identification of this non-monotonic relation

highlights the benefits of an expanded sample such as ours.

35

VII. Precautionary Motive for Holding Cash and Access to Public Equity

Markets

In this section, we take a closer look at the ability of private firms to accumulate cash for

precautionary reasons. As we have noted, an important benefit of cash holdings is that they allow

firms to better cope with adverse shocks and finance valuable projects or activities in the future.

High costs of cash may, however, impair the ability of private firms to respond to this

precautionary motive for holding cash.

The precautionary motive would suggest that high levels of cash flow volatility should be

associated with higher levels of cash since these types of firms are more likely to suffer from

cash shortfalls. We run specifications similar to those in Table 3 separately for listed and unlisted

firms, and compare the coefficients on cash flow volatility between these two types of firms. We

present results in Table 9. In Panel A, we document a notable difference in the coefficients for

cash flow volatility for listed and unlisted firms. Consistent with the precautionary motive, we

find that the coefficient on cash flow volatility is positive and highly statistically significant for

listed firms. This coefficient, however, is insignificant for unlisted firms suggesting such firms

are less sensitive to the precautionary benefits of holding cash. A plausible explanation is that the

incremental borrowing costs outweigh the benefits of cash holdings for precautionary reasons.

We then take a closer look at the relation between cash holdings and growth

opportunities. Previous research suggests that high growth firms would benefit from

accumulating cash for precautionary reasons (Opler et al., 1999, and Bates et al., 2009). Cash

holdings help these firms avoid missing out on valuable growth opportunities due to difficulties

with external financing. Results using sales growth suggest that growth opportunities are

positively related to cash holdings in both listed and unlisted firms. Sales growth, however, is

36

likely to capture not only growth opportunities but also cash flow from assets in place. Thus, we

consider two additional proxies for growth opportunities: industry market-to-book and global PE

ratios from Bekaert et al., (2007).15 Results are presented in specifications 3-6 in Table 9, Panel

A. We find that these proxies for growth opportunities are positively related to cash holdings

only for listed firms.

Further, we investigate the effect of equity dependence on cash policies of public and

private firms to provide additional insights into the ability of private firms to accumulate cash for

precautionary reasons. Equity-dependent firms are rich in growth opportunities and in need of

external financing (Rajan and Zingales, 1998). Results are presented in Table 9, specifications 7

and 8. We compute a firm’s external equity dependence as in Rajan and Zingales (1998), to

measure the industry’s demand for external equity. While we find that external equity

dependence is positively related to the cash holdings in listed firms, this is not the case for

unlisted firms. The coefficient is statistically insignificant.

In Table 9, Panel B we run regressions for unlisted firms for sub-samples of high and low

creditor rights to investigate whether the insensitivity of private firms to the precautionary

motive varies with debt market development. We find that private firms do not respond to the

precautionary motive regardless of the degree of debt market development. The coefficients on

the cash flow volatility and industry market-to-book are statistically insignificant in both

countries with strong and weak creditor rights.16 Our results suggest that when debt market

15 We obtain the annual global PE ratios for each industry from DataStream and manually match the DataStream

industry codes to 3-digit SIC codes available in Amadeus. 16 In untabulated analysis, we find that our results are robust to alternative proxies for growth opportunities.

37

development is high the incremental borrowing costs continue to outweigh the benefits of cash

holdings for precautionary reasons.

In Table 10, we investigate the combined effect of equity dependence and creditor rights

on the difference in cash holdings between listed and unlisted firms. Based on our findings so

far, we expected the difference in cash holdings between listed and unlisted firms to be at its

highest in countries with low creditor rights (when cash holdings for unlisted firms are relatively

low) and industries with high dependence on external equity (when cash holdings for listed firms

are relatively high). In turn, the difference in cash holdings should be at its lowest in countries

where creditor rights are high (when cash holdings for unlisted firms are somewhat higher) and

industries where external equity dependence is low (when cash holding for listed firms are

relatively low). We partition the sample four-ways: firms from countries with creditor rights

above and below median, and firms from industries with equity dependence above and below

median. We run a specification identical to that in Table 3, model 1, and cluster standard errors

at firm level. As anticipated, we find that the difference in cash holdings is the highest for low

credit rights countries in industries with high equity dependence. This difference is 0.06

compared to 0.03 in the full matched sample. The difference in cash holdings is the lowest in

countries with high creditor rights and industries with low equity dependence. In fact, in this

setting the difference in cash holdings between listed and unlisted firms is indistinguishable from

zero.

Overall, our results suggest that private firms are less sensitive to the precautionary

benefits of holding cash than public firms, consistent with the cost of cash argument.

38

VIII. Summary and Conclusions

In this paper, we emphasize that financing frictions lead not only to considerable

precautionary benefits of holding cash but also to substantial costs of holding cash. We propose a

simple model of cash retention policy and hypothesize that private firms, with higher opportunity

costs of holding cash, will hold significantly less cash than public firms. Using a comprehensive

sample of private and public firms, we provide evidence consistent with this cost of cash

hypothesis. The results are shown to be robust to endogeneity concerns. We further show that

private firms tend to use cash flow to redeem existing debt rather than to accumulate cash.

Consistent with our model’s predictions, we show that in countries with less developed

debt markets and higher costs of borrowing, private firms retain significantly less cash than their

counterparts from countries with more developed loan markets. We also show that time-series

variation in the country-level cost of loans is negatively related to cash holdings of unlisted

firms, and this relation is stronger when the country’s debt market is less developed. Overall,

stronger creditor-rights and debt-market development is associated with a convergence in the

cash holdings by unlisted and listed firms.

By analyzing private firms, we are able to show that the relation between financing

frictions and cash holdings is non-linear. The prior literature demonstrates that financially

constrained firms accumulate more cash than unconstrained firms. Our results, however, suggest

that in extreme cases, when external financing is significantly costly, firms accumulate relatively

little cash due to the high cost of holding cash.

39

References:

Almeida, Heitor, Murillo Campello and Michael Weisbach, 2004, “The Cash Flow Sensitivity of

Cash,” Journal of Finance 59, 1777–1804.

Azar, José, Jean-François Kagy, and Martin C. Schmalz, 2016, “Can Changes in the Cost of

Carry Explain the Dynamics of Corporate “Cash” Holdings?” Review of Financial Studies,

forthcoming

Bae, Kee-Hong and Vidhan K. Goyal, 2009, “Creditor Rights, Enforcement, and Bank Loans,”

Journal of Finance 64, 823-860.

Bates, Thomas W., Kathleen M. Kahle and Rene M. Stulz, 2009, “Why Do U.S. Firms Hold So

Much More Cash than They Used To?,” Journal of Finance 64, 1985-2020.

Bekaert, G., C. Harvey, C. Lundblad, and S. Siegel, 2007, “Global Growth and Market

Integration,” Journal of Finance 62, 1081-1137.

Bolton, Patrick, Hui Chen and Neng Wang, 2011, “A Unified Theory of Tobin’s q, Corporate

Investment, Financing, and Risk Management,” Journal of Finance 66, 1545-1578.

Denis, David and Valeriy Sibilkov, 2010, “Financial Constraints, Investment, and the Value of

Cash Holdings”, Review of Financial Studies, 23(11), 247-269

Dermiguc-Kunt, Asli, and Ross Levine, 1996, “Stock Markets, Corporate Finance, and

Economic Growth: An Overview,” World Bank Economic Review.

Dittmar, Amy, Jan Mahrt-Smith and Henri Servaes, 2003, “International Corporate Governance

and Corporate Cash Holdings,” Journal of Financial and Quantitative Analysis 38, 111-133.

Dittmar, Amy and Jan Mahrt-Smith, 2007, “Corporate Governance and The Value of Cash

Holdings,” Journal of Financial Economics 83, 599-634

Djankov, Simeon, Caralee McLeish and Andrei Shleifer, 2007, “Private Credit in 129

Countries,” Journal of Financial Economics 84, 299-329.

Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes and Andrei Shleifer, 2008, “The

Law and Economics of Self-Dealing,” Journal of Financial Economics 88, 430–465.

Dow, J. and G. Gorton, 1997, “Stock Market Efficiency and Economic Efficiency: Is There a

Connection?” The Journal of Finance, 52, 1087–1129

Edmans, A, 2009, “Blockholder Trading, Market Efficiency and Managerial Myopia,” Journal of

Finance¸ 64(6). 2481-2513

Falato, Antonio, Dalida Kadyrzhanova and Jae Sim, 2013, “Rising Intangible Capital, Shrinking

Debt Capacity, and the US Corporate Savings Glut,” Working paper. University of

Maryland.

Farre-Mensa, 2015, “The Benefits of Selective Discloser: Evidence from Private Firms,”

Working paper. Harvard Business School.

Faulkender, Michael. and Wang, Rong, 2006, “Corporate Financial Policy and the Value of

Cash,” The Journal of Finance, 61, 1957–1990

Gao Huasheng, Jarrad Harford and Kai Li, 2013, “Determinants of Corporate Cash Policy:

Insights from Private Firms,” Journal of Financial Economics 109(3), 623-639.

40

Giannetti, Mariassunta, 2003, “Do Better Institutions Mitigate Agency Problems? Evidence from

Corporate Finance Choices,” Journal of Financial and Quantitative Analysis 38, 185-212.

Graham, J. R., 2000, “How Big Are the Tax Benefits of Debt?” The Journal of Finance, 55,

1901–1941.

Gupta, N., and K. Yuan, 2009, “On the Growth Effect of Stock Market Liberalizations,” Review

of Financial Studies, 22, 4715-4752.

Harford, Jarrad, Sandy Klasa and William F. Maxwell, 2014, “Refinancing Risk and Cash

Holdings,” Journal of Finance 69(3), 975-1012.

Harford, Jarrad, Sattar A. Mansi, and William Maxwell, 2008, “Corporate Governance and Firm

Cash Holdings in the US,” Journal of Financial Economics, 87, 535-555

Hart, Oliver and John Moore, 1990, “Property Rights and the Nature of the Firm,” Journal of

Political Economy, 98(6), 1119-1158

Keynes, John Maynard, 1936, The General Theory of Employment, Interest, and Money, New

York: Harcourt Brace.

La Porta, Rafael, Florencio Lopez‐de‐Silanes, Andrei Shleifer and Robert W. Vishny, 1998,

“Law and Finance,” Journal of Political Economy 106, 1113-1155.

Lins, Karl V., Henri Servaes, and Peter Tufano, 2010, “What Drives Corporate Liquidity? An

International Survey of Cash Holdings and Lines of Credit,” Journal of Financial

Economics, 98, 160-1761

Love, Inessa, 2003, “Financial Development and Financing Constraints: International Evidence

from the Structural Investment Model,” Review of Financial Studies 16, 765-791.

Michaely, R. and M. Roberts. 2012, “Corporate Dividend Policies: Lessons from Private Firms,”

Review of Financial Studies 25, 711-746.

Opler, Tim, Lee Pinkowitz, Rene Stulz and Rohan Williamson, 1999, “The Determinants and

Implications of Corporate Cash Holdings” Journal of Financial Economics 52, 3-46.

Pagano, Marco, Fablo Panetta and Luigi Zingales, 1998, “Why do Companies Go Public? An

Empirical Analysis,” Journal of Finance 53, 27-64.

Pinkowitz, Lee and Rohan Williamson, 2001, “Bank Power and Cash Holdings: Evidence from

Japan,” Review of Financial Studies 14, 1059-1082.

Qian, Jun and Philip E. Strahan, 2007, “How Laws and Institutions Shape Financial Contracts:

The Case of Bank Loans,” Journal of Finance 62, 2803-2834.

Rajan, Raghuram G. and Luigi Zingales, 1998, “Financial Dependence and Growth,” The

American Economic Review 88, 559-586.

Saunders, Anthony and Sascha Steffen, 2011, “The Costs of Being Private: Evidence from the

Loan Market,” Review of Financial Studies, 24(12), 4091-4122

Subrahmanyam, A. and Titman, S., 1999, “The Going-Public Decision and the Development of

Financial Markets,” The Journal of Finance, 54, 1045–1082

41

Viral, Acharya V., Almeida Heitor and Campello Murillo, 2007, “Is Cash Negative Debt? A

Hedging Perspective on Corporate Financial Policies,” Journal of Financial Intermediation

16, 515-554.

42

Table 1. Descriptive Statistics: Balance Sheet Data

The table presents balance sheet data for the firms in the matched sample in 2008. Details of the

matching procedure are provided in the text. The data are from the 2011 version of Amadeus.

The sample includes non-financial firms from Western European countries. Accounting items are

scaled by total assets. ***, **, * denote statistical significance at the 1%, 5% and 10% levels,

respectively.

Listed Unlisted

Mean N Mean N Diff. in

means

Assets Fixed assets (of which)

0.50 2,897 0.44 2,886 0.05***

Intangible 0.19 2,893 0.06 2,782 0.14*** Tangible 0.23 2,897 0.25 2,810 -0.02** Current assets 0.50 2,897 0.56 2,887 -0.05*** Cash and cash equiv. 0.14 2,734 0.09 2,428 0.05*** Total assets ($ mill.) 1,745 2,898 1,267 2,898 478** Liabilities Shareholders’ funds 0.47 2,833 0.38 2,685 0.09*** Non-current liabilities (of which)

0.21 2,873 0.24 2,846 -0.04***

Long-term debt 0.14 2,766 0.16 2,641 -0.02*** Current liabilities (of which)

0.34 2,898 0.43 2,888 -0.09***

Loans 0.08 2,789 0.13 2,756 -0.06***

43

Table 2. Descriptive Statistics: Cash Holdings across Countries

The table presents mean cash holdings of listed and unlisted firms across counties for the

matched sample over the 1996-2011 period. Details of the matching procedure are provided in

the text. We present pooled sample means and sample size for each country. The data are from

the 2011 version of Amadeus. The sample includes non-financial firms from Western European

countries. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively.

Cash/Total Assets

Listed Unlisted

Mean N Mean N Diff. in

means

Austria 0.09 151 0.08 152 0.01

Belgium 0.14 785 0.13 771 0.01

Denmark 0.11 344 0.08 356 0.03***

Finland 0.11 732 0.07 610 0.04***

France 0.18 3,095 0.10 2,943 0.08***

Germany 0.15 3,178 0.11 2,107 0.04***

Greece 0.07 1,807 0.07 1,636 0.00

Ireland 0.21 233 0.09 148 0.13***

Italy 0.10 1,486 0.06 1,266 0.04***

Netherlands 0.12 966 0.10 757 0.02***

Norway 0.14 907 0.11 960 0.03***

Portugal 0.04 212 0.04 169 0.00

Spain 0.09 869 0.10 802 0.00

Sweden 0.17 945 0.09 966 0.08***

Switzerland 0.12 249 0.08 211 0.04***

United Kingdom 0.14 7,302 0.10 7,005 0.04***

44

Table 3. Differences in Cash Holdings across Public and Private Firms: Regression

Analysis

The table presents results of OLS regressions for the matched and unmatched samples (Panel A),

and for listed and unlisted firms where listed firms are partitioned into firms that have paid a

dividend in that year and firms that have not paid a dividend in that year (Panel B). The

dependent variable is cash and cash equivalents divided by total assets. Listed is an indicator

variable for the firm being listed on a major stock exchange. Leverage is measured as total debt

divided by the sum of total debt and shareholder funds. Total debt is the sum of long term debt

plus short term loans. Sales growth is computed as the one-year change in sales divided by

beginning-of-period sales. Standard deviation of cash flows is the standard deviation of current

and the past four cash flows to assets. If two or more cash flows to assets are missing, then the

variable is set to missing. Investment in tangible assets is the one-year change in the value of

tangible fixed assets divided by lagged assets. Investment in intangible assets is the one-year

change in the value of intangible fixed assets divided by lagged assets. Net working capital is

current liabilities minus current assets minus cash divided by lagged assets. Firm age is years

since incorporation. The estimation procedures correct standard errors for clustering at the firm

level. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively.

Panel A.

Matched (1)

Unmatched (2)

Listed 0.033*** 0.032***

Leverage -0.132*** -0.128***

Log(total assets) -0.009*** -0.011***

Sales growth 0.016*** 0.011***

Cash flow/total assets 0.261*** 0.286***

Standard deviation of cash flows 0.147*** 0.055***

Investments in tangible assets -0.161*** -0.181***

Investments in intangible assets 0.053* 0.009

Working capital (net of cash) -0.138*** -0.165***

Firm age -0.000*** -0.000***

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 16,377 379,065

Adjusted R2 0.2175 0.2176

45

Table 3. (continued)

Panel B.

Listed Firms

that pay

dividends (1)

Listed firms that

do not pay

dividends (2)

Diff.

Listed 0.036*** 0.031*** 0.005

Firm-level controls Yes Yes

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 12,991 10,408

Adjusted R2 0.2112 0.1927

46

Table 4. Differences in Cash Holdings across Public and Private Firms: Instrumental

Variables Approaches

The table presents results of the two-stage least squares estimation for the matched sample (Panel

A), and the treatment effects model for the unmatched sample (Panel B). The dependent variable

is cash and cash equivalents divided by total assets. Listed is an indicator variable for the firm

being listed on a major stock exchange. Leverage is measured as total debt divided by the sum of

total debt and shareholder funds. Total debt is the sum of long term debt plus short term loans.

Sales growth is computed as the one-year change in sales divided by beginning-of-period sales.

Standard deviation of cash flows is the standard deviation of current and the past four cash flows

to assets. If two or more cash flows to assets are missing, then the variable is set to missing.

Investment in tangible assets is the one-year change in the value of tangible fixed assets divided

by lagged assets. Investment in intangible assets is the one-year change in the value of intangible

fixed assets divided by lagged assets. Net working capital is current liabilities minus current

assets minus cash divided by lagged assets. Firm age is years since incorporation. Proportion of

listed firms in a region is used as an instrument. The estimation procedures correct standard

errors for clustering at the firm level. ***, **, * denote statistical significance at the 1%, 5% and

10% levels, respectively.

Panel A.

Dependent variable

Listed (1)

Cash/total assets (2)

Listed 0.079***

Leverage -0.129*** -0.126***

Log(total assets) 0.023*** -0.011***

Sales growth 0.019** 0.014***

Cash flow/total assets -0.258*** 0.273***

Standard deviation of cash flows 0.184*** 0.135***

Investments in tangible assets 0.218*** -0.171***

Investments in intangible assets 1.804*** -0.028

Working capital (net of cash) -0.001 -0.138***

Firm age 0.002*** -0.000***

Proportion of listed firms 3.558***

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 16,292 16,292

Adjusted R2 0.1387 0.1963

47

Panel B.

Dependent variable

Listed (1)

Cash/total assets (2)

Listed 0.043***

Leverage -0.399*** -0.128***

Log(total assets) 0.343*** -0.011***

Sales growth 0.038*** 0.011***

Cash flow/total assets -0.295*** 0.286***

Standard deviation of cash flows 0.956*** 0.054***

Investments in tangible assets -0.181***

Investments in intangible assets 0.007

Working capital (net of cash) -0.165***

Firm age 0.004*** -0.000***

Proportion of listed firms 14.899***

Country dummies Yes

Year dummies Yes

Industry dummies Yes

N 378,022 378,022

48

Table 5. Cash Flow, and Changes in Cash Holdings and Debt

The table presents results of the 3SLS regressions. Panel A presents results for the matched

sample. Panel B presents results for the sub-sample of firms that changed listing status. Change

in Debt is change in total debt scaled by beginning of period assets. Total debt is the sum of long

term debt plus short term loans. Change in cash is change in cash scaled by beginning of period

assets. Listed is an indicator variable for the firm being listed on a major stock exchange. Sales

growth is computed as the one-year change in sales divided by beginning-of-period sales. We

include country, industry and year dummies. The estimation procedures correct standard errors

for clustering at the firm level. ***, **, * denote statistical significance at the 1%, 5% and 10%

levels, respectively.

Panel A.

Change in Debt (1)

Change in Cash (2)

Listed -0.012*** 0.001

Cash flow/total assets -0.122*** 0.129***

Listed x cash flow/total assets 0.123*** 0.055***

Sales growth 0.041*** 0.006*

Log(total assets) 0.005*** -0.002***

Change in cash 0.159

Change in leverage 0.161**

Lagged leverage -0.013***

Lagged cash

-0.060***

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 24,771 24,771

49

Table 5. (continued)

Panel B.

Change in Debt (1)

Change in Cash (2)

Listed -0.027*** -0.007**

Cash Flow/total Assets -0.078** 0.151***

Listed x cash flow/total assets 0.159*** 0.065***

Sales growth 0.044*** 0.016***

Log(total assets) 0.080*** 0.013***

Change in cash -0.113* -0.044

Change in leverage

Lagged leverage -0.066* Lagged cash

-0.353***

Year dummies Yes Yes

Firm dummies Yes Yes

N 6,233 6,233

50

Table 6. Differences in Cash Holdings of Listed and Unlisted Firms and Private Debt

Market Development

The table presents results of OLS regressions for the matched sample. We run regressions on two

subsamples of firms: firms in countries with above median debt market development and firms in

countries with below median debt market development. In Panel A, we measure debt market

development with Private Credit/GDP, which we get from Djankov et al. (2007). In Panel B, we

use an index on creditor rights from La Porta, et al. (1998). The dependent variable is cash and

cash equivalents divided by total assets. Listed is an indicator variable for the firm being listed

on a major exchange. Each regression includes the same set of control variables that is included

in Table 3. Each regression includes country, year and industry dummies. The estimation

procedures correct standard errors for clustering at the firm level. We test for the null hypothesis

that the coefficients are equal across the two models using seemingly unrelated estimation. ***,

**, * denote statistical significance at the 1%, 5% and 10% levels, respectively.

Panel A.

Private Credit/GDP

Above median (1)

Below median (2)

Diff.

Listed 0.013** 0.052*** -0.039***

Firm-level controls Yes Yes

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 8,439 7,938

Adjusted R2 0.2338 0.2143

Panel B.

Creditor Rights

Above median (1)

Below median (2)

Diff.

Listed 0.013** 0.045*** -0.032***

Firm-level controls Yes Yes

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 7,245 9,132

Adjusted R2 0.2384 0.2148

51

Table 7. Cash Holdings of Listed and Unlisted, and Legal and Financial Development

The table presents results of OLS regressions for listed and unlisted firms for the matched

sample. In Panel A we control for debt market development with Private Credit/GDP from

Djankov, McLeish, and Shleifer (2007), and in Panel B with Creditor rights index from La Porta,

et al. (1998). The dependent variable is cash and cash equivalents divided by total assets. The

anti-self-dealing index is from Djankov et al. (2008). The stock market development index is

constructed from World Bank data following Dermiguc-Kunt and Levine (1996). GDP is from

the World Bank. Each regression includes the same set of control variables that is included in

Table 3. The estimation procedures correct standard errors for clustering at the firm level. ***,

**, * denote statistical significance at the 1%, 5% and 10% levels, respectively.

Panel A.

Unlisted Listed

(1) (2) (3) (4)

Private credit/GDP 0.022*** 0.025*** -0.014*** 0.001

Anti-self-dealing 0.001 -0.030***

Stock market development -0.006 -0.024**

Log(GDP) -0.005 0.049***

Firm-level controls Yes Yes Yes Yes

Year dummies Yes Yes Yes Yes

Industry dummies Yes Yes Yes Yes

N 7,022 7,022 9,355 9,355

R2 0.1622 0.1624 0.2619 0.2682

Panel B.

Unlisted Listed

(1) (2) (3) (4)

Creditor rights index 0.028*** 0.031*** -0.012** 0.001

Anti-self-dealing -0.003 -0.031***

Stock market development -0.006 -0.026**

Log(GDP) -0.012 0.049***

Firm-level controls Yes Yes Yes Yes

Year dummies Yes Yes Yes Yes

Industry dummies Yes Yes Yes Yes

N 7,022 7,022 9,355 9,355

R2 0.1641 0.1642 0.2682 0.2682

52

Table 8. Cash Holdings of Listed and Unlisted, and Cost of Loans

This table presents results of OLS regressions for listed and unlisted firms for the matched

sample. The dependent variable is cash and cash equivalents divided by total assets. Cost of

loans is country-level interest rates from the European Central Bank calculated using all

outstanding loans with a maturity of 1 year or less made by monetary financial institutions in a

given country to non-financial corporations. Each regression includes the same set of firm-level

control variables that is included in Table 3. The estimation procedures correct standard errors

for clustering at the firm level. ***, **, * denote statistical significance at the 1%, 5% and 10%

levels, respectively.

Panel A.

Unlisted

(1) Listed

(2)

Cost of loans -0.008*** 0.002

Firm-level controls Yes Yes

Year dummies Yes Yes

Firm dummies Yes Yes

N 6,465 9,051

R2 0.7897 0.7620

Panel B.

Unlisted

(1) Listed

(2)

Cost of loans -0.014*** 0.004

Cost x (D=1 if high cr. rights) 0.005** -0.001

Firm-level controls Yes Yes

Year dummies Yes Yes

Firm dummies Yes Yes

N 6,465 9,051

R2 0.7899 0.7620

53

Table 9. Precautionary Benefits and Cash Holdings

The table presents results of OLS regressions for the matched sample. Panel A includes separate regressions for listed and unlisted

firms, while panel B presents regressions for the sub-sample of unlisted firms only. The dependent variable is cash and cash

equivalents divided by total assets. Listed is an indicator variable for the firm being listed on a major stock exchange. Leverage is

measured as total debt divided by the sum of total debt and shareholder funds. Total debt is the sum of long term debt plus short term

loans. Sales growth is computed as the one-year change in sales divided by beginning-of-period sales. Standard deviation of cash

flows is the standard deviation of current and the past four cash flows to assets. If two or more cash flows to assets are missing, then

the variable is set to missing. Investment in tangible assets is the one-year change in the value of tangible fixed assets divided by

lagged assets. Investment in intangible assets is the one-year change in the value of intangible fixed assets divided by lagged assets.

Net working capital is current liabilities minus current assets minus cash divided by lagged assets. Firm age is years since

incorporation. We compute an industry’s external equity dependence as in Rajan and Zingales (1998). Creditor rights index is from La

Porta, et al., (1998). Firm-level controls in Panel B are the same as those in Panel A and in Table 3. We test for the null hypothesis that

the coefficients are equal across the two models using seemingly unrelated estimation. The estimation procedures correct standard

errors for clustering at the firm level. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively.

54

Table 9. (continued)

Panel A.

Listed Unlisted Listed Unlisted Listed Unlisted Listed Unlisted

(1) (2) (3) (4) (5) (6) (7) (8)

Leverage -0.192*** -0.097*** -0.187*** -0.105*** -0.183*** -0.102*** -0.194*** -0.097***

Log(total assets) -0.005*** -0.012*** -0.006*** -0.012*** -0.004*** -0.012*** -0.004*** -0.014***

Sales growth 0.010* 0.016*** 0.004 0.016* 0.011** 0.015** 0.015*** 0.019***

Cash flow/total assets 0.291*** 0.207*** 0.291*** 0.221*** 0.321*** 0.203*** 0.273*** 0.190***

Std. Dev. of cash flows 0.290*** -0.008 0.263*** 0.023 0.250*** 0.005 0.285*** -0.008

Inv. in tangible assets -0.158*** -0.154*** -0.121*** -0.138*** -0.165*** -0.152*** -0.156*** -0.158***

Inv. in intangible assets 0.072** -0.104* 0.056 -0.044 0.104*** -0.117* 0.041 -0.100

Working capital (net of cash) -0.130*** -0.150*** -0.145*** -0.148*** -0.122*** -0.148*** -0.115*** -0.145***

Firm age -0.000*** 0.000 0.000*** 0.000** 0.000*** 0.000 -0.003*** -0.000**

Market-to-Book Industry

0.016*** -0.009 Global PE ratio

0.0004** 0.000

External equity dependence

0.021*** 0.005

Country dummies Yes Yes Yes Yes Yes Yes Yes Yes

Year dummies Yes Yes Yes Yes Yes Yes Yes Yes

Industry dummies Yes Yes Yes Yes Yes Yes No No

N 9,355 7,022 3,807 2,808 6,900 5,788 8,517 6,373

Adjusted R2 0.2835 0.1666 0.2927 0.1733 0.2889 0.1649 0.2851 0.1524

55

Table 9. (continued)

Panel B.

Creditor Rights

Above median (1)

Below median (2)

Diff.

Std. dev. of cash flows -0.061 0.117 -0.178

Market-to-book industry -0.022 0.003 -0.025

Firm-level controls Yes Yes

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 1,555 1,253

Adjusted R2 0.2245 0.1340

56

Table 10. External equity dependence, Creditor rights, and Cash Holdings

The table presents results of OLS regressions for the matched sample. The dependent variable is

cash and cash equivalents divided by total assets. Listed is an indicator variable for the firm

being listed on a major stock exchange. We compute an industry’s external equity dependence as

in Rajan and Zingales (1998). Creditor rights index is from La Porta, et al. (1998). Firm-level

controls are the same as those in Panel A and in Table 3. We test for the null hypothesis that the

coefficients are equal across the two models using seemingly unrelated estimation. The

estimation procedures correct standard errors for clustering at the firm level. ***, **, * denote

statistical significance at the 1%, 5% and 10% levels, respectively.

External equity dependence

Above median (1)

Below median (2)

Diff.

Creditor Rights above median

Listed 0.028*** -0.010 0.038***

Firm-level controls Yes Yes

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 3,146 3,479

Adjusted R2 0.2375 0.2325

Creditor Rights below median

Listed 0.059*** 0.035*** 0.024**

Firm-level controls Yes Yes

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 3,113 5,152

Adjusted R2 0.2271 0.2253

57

Figure 1. Cash Holdings for Listed and Unlisted Firms across Time

The figure contains average cash holdings to total assets ratio for listed and unlisted firms from 1997

through 2010.

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

19971998 1999 2000 2001 2002 20032004 2005 2006 2007 20082009 2010

Cas

h h

old

ings

/ T

ota

l ass

ets

Year

Unlisted

Listed

58

Appendix 1. Cash holdings and Hedging Needs

The table presents results of OLS regressions for the matched sample. The dependent variable is

cash and cash equivalents divided by total assets. Listed is an indicator variable for the firm

being listed on a major stock exchange. Hedging needs is the correlation between investment

opportunities and the firm’s cash flow, where investment opportunities is the median three-year-

ahead sales growth rate in the firm’s three-digit SIC industry. Low hedging needs is a dummy

variable that takes the value of one when the correlation between investment opportunities and

the firm’s cash flow is above 0.20, and zero otherwise. High hedging needs is a dummy variable

that takes the value of one when the correlation between investment opportunities and the firm’s

cash flow is below -0.20, and zero otherwise. Each regression includes the same set of firm-level

control variables that is included in Table 3. The estimation procedures correct standard errors

for clustering at the firm level. ***, **, * denote statistical significance at the 1%, 5% and 10%

levels, respectively.

(1) (2)

Listed 0.032*** 0.032***

Hedging needs 0.003

Low hedging needs 0.001

High hedging needs -0.001

Firm-level controls Yes Yes

Country dummies Yes Yes

Year dummies Yes Yes

Industry dummies Yes Yes

N 14,910 14,910

Adjusted R2 0.2116 0.2175