See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228313919

Accounting Quality and Firm-Level Capital Investment

Article in The Accounting Review · June 2006

DOI: 10.2308/accr.2006.81.5.963

CITATIONS READS
733 2,572

2 authors:

Gary C. Biddle Gilles Hilary

University of Melbourne INSEAD
61 PUBLICATIONS 5,045 CITATIONS 37 PUBLICATIONS 4,250 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Gilles Hilary on 21 May 2014.

The user has requested enhancement of the downloaded file.

Accounting Quality and Firm-level Capital Investment

Gary Biddle and Gilles Hilary

biddle@ust.hk acgh@ust.hk

Department of Accounting
Hong Kong University of Science and Technology
Clear Water Bay, Kowloon
Hong Kong

Forthcoming, The Accounting Review, October 2006

We are grateful for the research assistance of Fenny Cheng and for helpful comments

received from two anonymous reviewers, Mary Barth, Jean de Bettignies, Donal Byard,

Songnian Chen, Dan Dhaliwal (the editor), Clive Lennox, Steve Orpurt, Douglas Skinner,

Guochang Zhang and workshop participants at Baruch College, Cass Business School,

Hong Kong University of Science and Technology, London School of Economics,

Singapore Management University, University of British Columbia, and Waseda

University. Biddle and Hilary thank the University Grants Committee of Hong Kong for

financial support (HKUST6435/05H).

Accounting Quality and Firm-level Capital Investment

Abstract: This study examines how accounting quality relates to firm-level capital

investment efficiency. Our first hypothesis is that higher quality accounting enhances

investment efficiency by reducing information asymmetry between managers and outside

suppliers of capital. Our second hypothesis is that this effect should be stronger in

economies where financing is largely provided through arm’s-length transactions

compared with countries where creditors supply more capital. Our results are consistent

with these hypotheses both across and within countries. They are robust to alternative

econometric specifications, different measures of accounting quality and investment-cash

flow sensitivity, and numerous control variables.

Keywords: Accounting Quality, Capital Investment, Corporate Governance

1
Accounting Quality and Firm-level Capital Investment

1. Introduction

A significant body of prior literature examines relations between accounting

quality and financial market characteristics, yet little prior research exists that analyzes

the effects of accounting quality on investment. Here, we examine how accounting

quality relates to firm-level capital investment efficiency, a key determinant of economic

productivity. Our focus on investment in productive assets complements and extends

findings on how accounting relates to public financial markets’ operating characteristics.

For example, Bhattacharya et al. [2003] find that accounting opacity is associated with a

higher cost of publicly traded equity capital across 34 countries. However, institutional

features may be related to firm-level capital investment differently than to equity market

operations. For instance, private financing from banks in Germany (e.g., Gorton and

Schmid [2000]), keiretsu in Japan (e.g., Hoshi et al. [1991]), and families in East Asia

(e.g., Claessens et al. [2000]) may obviate or alter the effects of institutional features

from their operational roles in public equity markets. Hence, how accounting quality

affects the efficiency of firm-level capital investments remains an open question of

fundamental importance.

In the neoclassical setting, managers (i.e., firms) endowed with capital invest until

the marginal return is zero, allowing for adjustment costs (e.g., Tobin [1969], Hayashi

[1982]). In this baseline setting, we should not observe an association between internally

generated cash flows and investment. But several frictions contradict this efficient result.

One is capital rationing, which is generated by information asymmetry between managers

1
and investors. Since at least Myers [1984], it has been shown that if managers can exploit

private information to issue securities at inflated prices, investors rationally withhold

capital. A resulting reliance on internal funding increases the sensitivity of investments

to cash. A second friction arises from agency problems, when managers pursue perquisite

consumption and “empire building” rather than returning excess cash to investors (e.g.,

Jensen [1986], Blanchard et al. [1994]).1 This behavior also may increase the sensitivity

of investments to cash flows but, in this case, the sensitivity is due to an excess of cash.

Certain institutional features may serve to mitigate these deviations from the

optimal investment policy. For example, Rajan and Zingales [2000] observe that “to

function properly, a financial system requires clear laws and rapid enforcement, an

accounting and disclosure system that promotes transparency and a regulatory

infrastructure that protects consumers and controls risk.” Transparent accounting should

reduce both adverse selection (i.e., the tendency to issue securities at an inflated price)

and moral hazard (i.e., perquisite consumption using assets in place) by improving

contracting and monitoring. 2 Thus, higher quality accounting may serve to enhance

investment efficiency by mitigating these frictions.

We examine this hypothesis by considering the effects of differences in

accounting quality both at the country level and at the firm level within countries. To do

so, we first estimate investment-cash flow sensitivities for firms from 34 countries. We

then consider how the average investment-cash flow sensitivity across countries varies

with accounting quality. We measure accounting quality using three proxies for earnings

quality described by Bhattacharya et al. [2003], as well as a measure of accounting

1
Note that the existence of the agency problem ex post may lead to rationing of capital ex ante.
2
See Healy and Palepu [2001] for a review.

2
timeliness from Bushman et al. [2004]. We also are careful to control for the confounding

effects of other institutional features, such as disclosure quality (CIFAR index), legal

origin, creditor and shareholder rights, judicial efficiency, and economic conditions (e.g.,

the degree of economic development). We find that higher accounting quality is

associated with lower investment-cash flow sensitivity. We also find that other

institutional features, particularly creditor rights and disclosure quality, play a similar but

incremental role. These results persist after allowing for the possibility that operating

cash flows convey additional information about short-term profitability (e.g., Alti

[2003]).

Having established that differences in accounting quality are associated with

differences in the efficiency of capital investments across countries, we then examine

how sources of financing (debt versus equity) affect this relation. A priori, we expect

accounting quality to play a stronger role in economies where financing is largely

provided through arm’s-length transactions, for example, where stock exchanges are the

dominant sources of capital, since here, investment decisions rely more heavily on public

accounting disclosures. On the other hand, in economies where creditors play a more

dominant role, banks may be able to obtain information through private channels

(mitigating adverse selection problems), and they may be in a better position to directly

monitor managers once capital is supplied (mitigating moral hazard). Thus, in credit-

dominated economies, the quality of publicly disclosed accounting information should be

less influential in decisions to supply capital than in public equity-dominated economies.

Therefore, we should observe a smaller effect on capital investment decisions when

3
accounting quality is improved in credit-dominated economies than in public equity-

dominated economies.

We test this hypothesis in two ways. Initially, we regress cash flow sensitivity on

accounting quality in our cross-country sample, partitioning observations in two groups

based on the prevalence of public equity versus debt financing. We find that higher

accounting quality is associated with lower investment-cash flow sensitivity in economies

that depend more on public equity financing compared with those more reliant on debt

financing. Then, we examine the effect of accounting quality on investment efficiency at

the firm level in two selected countries. We do this for several reasons. First, it allows us

to determine whether accounting quality operates similarly across and within countries.

Second, it helps us to control for certain differences in characteristics across countries

that could affect our cross-country results, such as correlations between accounting

quality, legal origins, creditor rights and administrative efficiency. Finally, as explained

below, this firm-level analysis allows us to address certain econometric issues.

In our within-country tests, we examine two polar cases – the US and Japan – for

several reasons. First, they are the two largest economies in the world with abundant

external financing available to firms. Second, they provide large and diversified firm

samples with long time-series data that enable us to estimate firm-specific parameters

with a reasonable level of confidence. Comparable datasets are generally not available for

other countries. However, the US and Japan differ along one important dimension. Public

equity capital plays a much more dominant role as a source of firm financing in the US

than in Japan. Thus, we expect accounting quality to play a more important role in the

largely arm’s-length transactions in the US. On the other hand, Japanese suppliers of

4
capital, such as keiretsu and banks, have non-public sources of information and thus the

quality of public accounting disclosures may be less relevant to their decisions to supply

capital. Our results are consistent with these predictions. Accounting quality matters both

statistically and economically in the US where higher quality accounting is associated

with lower investment-cash flow sensitivity, but we do not observe such a relation in

Japan.

Our findings contribute in at least two ways to the existing literature. First, they

provide empirical evidence that accounting and other institutional features relate to the

economic fundamentals of firm-level capital investment efficiency. In particular,

accounting quality is shown to reduce frictions in the investment process. We find that

this effect exists both across countries and within countries, even in the most liquid

capital market, the US. Second, our findings confirm that this effect is stronger in

economies where public equity capital plays a greater role in capital investment financing

compared with countries dominated by debt financing. This effect is observed in both our

cross- and within-country tests.

We proceed as follows. Section 2 reviews related research. Section 3 develops

our hypotheses. Section 4 describes our tests for relations between accounting quality and

investment-cash flow sensitivity. Section 5 presents our findings. Section 6 concludes

the paper.

2. Related Research

Economists have long studied how financial frictions affect investment decisions

and economic growth. This literature is too extensive to review comprehensively here,

5
though we discuss selected results in Section 4.1 below. 3 More recently, scholars in

accounting, finance and law have focused their attention on the effects of institutional

features (such as legal structure or judicial enforcement) on market frictions and their

consequences for capital investment. For example, Bhattacharya et al. [2003] have shown

that the cost of publicly traded equity across 34 countries is related to three different

measures of earnings transparency. Francis et al. [2005] provide international evidence

that dependence on external financing creates incentives for firms to undertake higher

levels of voluntary accounting disclosure. However, the implication of these findings for

firm-level investment is not straightforward. If equity financing were the only source of

capital, frictions in equity financing would probably lead to frictions in capital investing.

However, firms can access multiple sources of financing. Thus, if one channel is

inefficient (e.g., public equity financing), other sources such as debt, private financing,

state subsidies, and intra-group capital transfers, for example, can substitute. In

equilibrium, there may not be any differences in investment efficiency at the firm level,

but simply cross-sectional differences in financing patterns. Consistent with this view,

Demirguc-Kunt and Maksimovic [2002] find no evidence that a country’s relative

reliance on bank versus stock market financing affects firms’ access to external

financing.

Other studies have considered relations between institutional features and capital

investment at the industry level. For example, Wurgler [2000] measures cross-country

capital allocation efficiency by industry-level elasticity of investment with respect to

value added. He finds this measure to be positively related to the amount of firm-specific

3
See also Hubbard [1998] for a review and representative studies by Bagehot [1873], King and Levine
[1993], Rousseau and Wachtel [1998] and Beck, Levine and Loayza [2000].

6
information available in domestic stock markets when measured by synchronicity,

positively related to minority shareholder rights, and negatively related to state

ownership. Rajan and Zingales [1998] show that industry growth is positively related to a

financial development index.

Our paper advances the existing literature in several ways. First, instead of

considering the effect of institutional features on industry growth (e.g., Rajan and

Zingales [1998], Wurgler [2000]) or on financial market development (e.g., LaPorta et al.

[1997], Demirguc-Kunt and Maksimovic [1998]), we focus on firm-level capital

investment decisions. Second, we concentrate on accounting quality, a feature not

examined in many prior studies (e.g., LaPorta et al. [1997], Demirguc-Kunt and

Maksimovic [1998], Wurgler [2000]). In addition, we consider the differences both

between countries and between firms within a country. Third, our tests do not limit

capital market frictions to their effects on capital rationing exclusively (e.g., Fazzari et al.

[1988]). Rather, we allow accounting quality to reduce either the lack or excess of cash.

Finally, given the possibility that different sources of capital may substitute for each

other, we do not limit our attention to any specific financing channel (e.g., Bhattacharya

et al. [2003]) and instead focus on capital investment behavior.

3. Hypothesis Development

Tobin [1969] theorizes that capital investment is a function of the ratio between

the stock-market valuation of existing real capital assets and their current replacement

cost. Yoshikawa [1980], Hayashi [1982] and Abel [1983] reconcile this theory with the

neoclassical interpretation. In this framework, the marginal Q ratio, q, is the driver of

capital investment policy. For example, Hayashi [1982] summarizes the model by stating,

7
“once q is known […], the firm can decide the optimal rate of investment though the

knowledge of the installation function ψ alone.” He then proceeds to estimate this

baseline model by regressing the investment rate on q. Fazzari et al. [1988] challenge this

view and suggest that firms that are liquidity constrained (i.e., cannot externally finance

their investments) need to rely more on their internally generated funds.

Hoshi et al. [1991] summarize two possible justifications for capital rationing. On

the one hand, moral hazard models suggest that outside financing can dilute

managements’ ownership stakes, thereby exacerbating incentive problems that arise when

managers control the firm but do not own it. This ex post incentive problem reduces the

amount of capital supplied ex ante. On the other hand, Myers and Majluf [1984] propose

an adverse selection model. They suggest that if managers are better informed than

investors about a firm’s prospects, they will try to sell overpriced securities. Rational

investors will, in response, increase the cost of capital, thus decreasing the amount

demanded. Therefore, in both cases, frictions operate to reduce the amount of external

capital supplied to the firm. Firms that can generate cash internally are able to mitigate

these effects, which causes capital investment to be correlated with the availability of

internally generated funds.

There is presently a debate (see, for example, Fazzari et al. [2000] and Kaplan and

Zingales [1997, 2000]) as to whether higher investment-cash flow sensitivity necessarily

implies that firms are more financially constrained. In this paper, we remain agnostic on

this issue since the problem may not be a lack, but rather an excess of cash. Capital

investment can be correlated with internally generated funds simply because managers do

not return to investors excess cash coming from rents (and quasi rents) and other assets in

8
place. Casual empiricism suggests the existence of such overinvestment by managers and

several theoretical explanations have been proposed. For example, Jensen [1986]

suggests that managers have incentives to grow their firms beyond their optimal size.4

Whereas external financing subjects managers to monitoring and disciplining by capital

providers, “financing projects internally avoids this monitoring and the possibility the

funds will be unavailable,” thereby allowing managers to overinvest. Blanchard et al.

[1994] provide empirical support for this view. They consider what managers do when

they receive a cash windfall that does not change the investment opportunity set (i.e.,

Tobin’s Q). In perfect financial markets, managers should return the money to the capital

suppliers. Contrary to this expectation, Blanchard et al. [1994] find that managers tend to

invest in unrelated projects that typically fail.

Notice that all of these imperfections are predicated by the existence of

information asymmetry between managers and outside suppliers of capital. If managers

could commit to revealing all of their private information, outsiders would not ration

capital for fear of buying at an inflated price. Similarly, if higher quality accounting

permitted perfect monitoring, no agency problem would arise (see, for example, Antle

and Eppen [1985] for a formal model of this idea). We would then be back to the baseline

neoclassical model and internally generated cash flows would play no role in investment

decisions. Our main hypothesis is predicated on this idea:

Hypothesis 1: Higher quality accounting reduces the investment-cash flow

sensitivity at the firm level.

4
For simplicity, we use “empire building” as our main way to motivate overinvestment. However, there are
other models leading to a similar pattern such as overconfidence (e.g., Heaton [2002]), the “quiet life”
(Bertrand and Mullainathan [2003]) and reputation (e.g., Baker [2000]).

9
 We further predict, based on the reasoning above, that this effect should be

stronger in economies where financing is largely provided through arm’s-length

transactions, for example, where the stock market is the dominant source of capital. On

the other hand, in economies where creditors play a central role in financing, banks may

be able to obtain information through alternate private channels (and thus mitigate

adverse selection problems). They also may be in a better position to monitor the

managers directly once the capital has been supplied (and thus mitigate the moral

hazard). In this case, accounting quality should be less relevant to their decisions to

supply capital and, as a result, we should observe a smaller effect of accounting quality

on investment efficiency in credit-based economies. his leads to our second hypothesis:

Hypothesis 2: Higher quality accounting reduces the investment-cash flow

sensitivity more in economies dominated by stock markets than in

those dominated by creditors.

Notice that we do not form any prediction as to whether one type of economy is better

than the other in mitigating frictions in the investment process.

4. Cash flow sensitivity: Empirical setting

4.1. Overall approach.

We employ both cross- and within-country tests, each with its own advantages.

In the cross-country tests, average investment-cash flow sensitivities are estimated by

country using firm-level data and then regressed on accounting quality and other

variables of interest. This testing approach is likely to enhance the power of our tests by

increasing both the cross-sectional variation and the magnitudes of financial frictions. It

also enables us to consider the effects of institutional factors, such as creditor rights and

10
legal origin, which cannot be easily studied in a single-country setting. As observed by

Francis et al. [2005], “the United States is generally viewed as having a frictionless

financial market with relatively easy access to external financing, and therefore it is not

clear if U.S. findings necessarily generalize to countries with different financial or legal

systems.”

However, the cross-country approach also has some limitations. First, it is

difficult to obtain sufficient time series data from multiple countries to estimate firm-

specific parameters. Second, different institutional settings and accounting rules make it

difficult to compare firm-based measures across countries. For example, since asset

revaluation is permitted in some countries but not in others, plant, property and

equipment and depreciation measures convey different meanings. To address these issues

and to provide corroborating evidence that the hypothesized effect of accounting quality

on investment efficiency operates within as well as across counties, we further examine

two contrasting countries: the US and Japan. Both countries have large economies with

ample sources of external financing. However, in the US, stock markets play a central

role in providing capital, while in Japan, keiredtsu and banks are the major source of

capital. As documented by Wurgler [2000], the ratio of stock market value to total credit

market value is 0.64 for the US but only 0.15 in Japan. Our focus on the US and Japan

allows us to assess whether the dominant form of financing influences the effect of

accounting quality on investment decisions.

4.2. Samples and data

To enhance comparability with prior research, we utilize data from the same set of

34 countries examined by Bhattacharya et al. [2003]. For our cross-country tests,

11
accounting and financial data are obtained from Compustat Global Vantage for the entire

coverage period of the database (1993 to 2004). Accounting and financial data for the

within-country tests are obtained for the US from COMPUSTAT and for Japan from the

PACAP database (1975 to 2001). he longer time series for the within-country samples

allow us to estimate firm-specific parameters and utilize them in panel specifications.

We focus on industrial firms and, as is customary, exclude utilities and firms in the

financial, real estate, insurance and public administration sectors.5 Our resulting sample

includes only publicly traded companies with access to public sources of capital. As these

firms also can potentially access alternate sources of capital including bank loans,

governmental financing and private equity when public markets are inadequate, this

provides a lower bound for the effects of accounting quality and other institutional

features. Private firms with lesser access to capital and weaker outside monitoring would

benefit relatively more from enhanced institutional features than would firms with more

access to capital and stronger outside monitoring.

4.3 Proxies for accounting quality.

For our cross-country tests, we use four country-specific measures of accounting

quality. Three of these measures are adapted from Bhattacharya et al. [2003] and a fourth

from Bushman et al. [2004]. Earnings aggressiveness is based on the converse of

conservatism as defined by Ball et al. [2000]. Following Ball et al. [2000] and

Bhattacharya et al. [2003], we expect countries with more aggressive accounting

practices to exhibit lower firm-level capital investment efficiency than countries with less

aggressive accounting practices. Loss avoidance is the ratio of the number of firms with

5
We focus on industrial and commercial companies and, following prior studies, exclude US firms with
SIC codes between 4900-4999, 6000-6999 and above 9000, and those trading as American Depository
Receipts. We exclude Japanese firms with INDID codes starting with 05, 06 and 08.

12
small positive earnings minus the number of firms with small negative earnings divided

by the sum of the two. This proxy is derived from Burgstahler and Dichev [1997] and

Degeorge et al. [1999] who employ a similar measure for US firms. Following

Bhattacharya et al. [2003], we expect countries with greater loss avoidance to exhibit

lower capital investment efficiency than countries with less loss avoidance. Earnings

smoothing is based on the cross-sectional correlation between the change in accruals and

the change in cash flows scaled by lagged total assets. Following Leuz et al. [2003] and

Bhattacharya et al. [2003], we expect countries with a greater degree of earnings

management to exhibit lower capital investment efficiency than countries with a smaller

degree of earnings management. Following Bushman et al. [2004], Timeliness is based on

the average ranking of “answers to the following interim reporting questions: frequency

of reporting, count of disclosed items, and consolidation of interim reports.” We expect

countries with timelier reporting to exhibit higher capital investment efficiency than

countries without timely reporting. Because we are agnostic regarding which dimension

of accounting quality plays a more significant role, we aggregate the four accounting

quality measures into a summary index (AQ). In order to combine them meaningfully,

we first form binary specifications. We do so by creating four binary variables for

earnings aggressiveness, loss avoidance, earnings smoothing and timeliness based on

whether they have a better value than the median value in the cross-country sample. We

then create AQ by summing up these four binary variables. Combining the four measures

into an index has the advantage of reducing the effects of measurement errors in the

individual accounting quality measures. This index also is more comparable with the

13
index variables for creditor and shareholder rights developed by LaPorta et al. [1997]

introduced as control variables below.6

In our within-country tests for the US and Japan, we employ firm-specific

accounting quality measures, AQFS, reflecting the uncertainty in accruals following

Dechow and Dichev [2002]. Francis et al. [2005] find that this measure of accounting

quality is the most closely associated with their measure of the cost of capital. In Dechow

and Dichev [2002], accruals quality is measured by the extent to which working capital

accruals explain current-period, prior-period and future-period operating cash flow

realizations. The unexplained portion of the variation in working capital accruals is an

inverse measure of accruals quality (with a larger unexplained portion implying lower

quality accounting). Details of the estimation procedure are provided in the appendix.

We predict that higher quality accruals will be associated with higher capital investment

efficiency especially in countries that rely on capital provided through arms-length

transactions.

4.4. Proxies for investment-cash flow sensitivity.

We use two different but comparable procedures to measure investment-cash flow

sensitivity in the cross-country and within-country tests. In the cross-country tests, we

mirror the prior literature (e.g., Fazzari et al. [1988], Hoshi et al. [1991]) by estimating

investment-cash flow sensitivities after controlling for investment opportunities using

Tobin’s Q (proxied by the market-to-book equity ratio, MTB). Given the heterogeneity of

our cross-country sample, outliers could induce non-linear relations. A standard approach

6
AQ takes values between 0 and 4. Below, we also control for financial disclosure as measured by an
index created by the Center for International Financial Analysis and Research (CIFAR). The CIFAR index
can potentially take values between zero and one hundred. To make AQ and CIFAR more comparable, we
also normalize CIFAR by forming a variable that takes a value between zero and four based on quintiles of
the distribution of CIFAR. Our results (untabulated) are not affected by this transformation.

14
to alleviating this problem is to take the log of the variables, which we apply to market-

to-book ratios. However, operating cash flows and investments often have negative

values for which a log transformation is not defined. For them, we use an arctangent

transformation that in effect “logs negative values.”7 The following model is estimated

(using firm fixed effects) for each country:

Ii,t / Ki,t-1 = βi0 + β1 OCFi,t / Ki,t-1 + β2 MTBi,t + εi,t (1)

where Ii,t / Ki,t-1 is capital investment scaled by the beginning-of-period capital for

firm i in period t, OCFi,t is the operating cash flow, β1 is our measure of

investment-cash flow sensitivity, εi,t is a normally distributed error term with a

mean of zero, and it represents usual idiosyncratic errors.

This specification has two appealing features. First, it does not require a long time

series for each firm to estimate the parameters, which allows us to use pooled cross-

sectional data to estimate an average investment-cash flow sensitivity for each country.

This feature is important because Compustat Global Vantage contains a maximum of

only 11 years of data. Second, this specification has been extensively used in the prior

literature. However, it relies on MTB to proxy for Q. To the extent that MTB does not

fully capture investment opportunities, cash flows may pick up measurement errors that

are correlated across countries with accounting quality.

We address this concern in several ways. In our cross-country tests, we include

future cash flows (Alti [2003]) and we use an instrumental variables approach in some

specifications. In our within-country tests, we use two databases, COMPUSTAT for the

7
The arctangent function is approximately linear over the range [-1;1]. In cases where the absolute value
of investment or cash flows is less than net fixed assets, the data are changed little by the transformation.
However, when the absolute values of these scaled variables increase beyond 1, the transformed value is
bounded by π/2. This transformation provides a convenient way of winsorising the data.

15
US and PACAP for Japan, which provide longer time series. This allows us to use an

alternate test that does not rely on estimating Q. By doing so, we have a natural

robustness check for our cross-country estimations. We proxy for cash flow sensitivity of

investment (CFSI) using the measure recently proposed by Hovakimian and Hovakimian

[2005]. CFSI0 is measured as the difference between the cash-flow-weighted time-series

average investment (CFWAI) of a firm and its un-weighted arithmetic time-series average

investment (AI):

CFSI0,i,t = CFWAI0,i,t - AIi,t =1/n Σs=1t [ (CFi,s / (Σs=1t CFi,s)) * Ii,s] - 1/n Σs=1t Ii,s (2)

where CF is a firm’s cash flow, I is the firm’s investments (where both are scaled

by beginning-of-period net capital), and n is the number of annual observations

for firm i.8

The intuition behind this specification is that there should be no systematic

difference between the weighted and un-weighted average investments for firms whose

investment decisions are not affected by their available cash flows. On the other hand, the

value of CFSI should be higher for firms that tend to invest more in years with relatively

high cash flows and less in years with relatively low cash flows. Unlike the measure of

cash flow sensitivity in Equation (1), we do not have to estimate a regression coefficient

and, therefore, we can avoid the associated econometric estimation issues.

4.5. Estimation procedures

Our cross-country tests regress the 34 investment-cash flow sensitivities from

Equation (1) on our measure of accounting quality, other institutional features (disclosure

quality, creditor rights, shareholder rights, judicial efficiency and legal origin), and

8
Variable definitions are detailed in the Appendix.

16
control variables such as the gross domestic product (GDP), the average tangibility of

firm assets, and firm size, which are all standard in the literature (see Appendix for

variable definitions and estimation details). In the US within-country tests, we also

control for auditing quality and analyst coverage as well as other control variables

suggested by Hovakimian and Hovakimian [2005] and described in the Appendix. In

both the US and Japan within-country tests, accounting quality, investment-cash flow

sensitivity and control variables are estimated using ten-year rolling windows. Standard

errors are corrected for heteroskedasticity and for clustering of observations by firm.

5. Cash flow sensitivity: Empirical results.

5.1. Cash flow sensitivities and between countries differences

Table 1 presents correlations among variables examined in the cross-country tests.

Most correlations are reasonably small, suggesting that multicollinearity is not a

significant concern. In particular, the correlation between AQ and CIFAR is only 0.13,

suggesting that these two variables capture two related but differing constructs.

However, shareholder rights (SR) is highly correlated with both AQ and CIFAR, as is

legal origin (LO) with AQ, CIFAR, SR, and CR. We take these high correlations into

consideration in our later tests.

Table 2 presents tests of our first hypothesis by regressing investment-cash flow

sensitivities at the country level from Equation (1) on accounting quality (AQ) and other

variables as described above. We report several specifications. In column 1, we enter

only AQ. In columns 2 and 3, we control for other institutional features. Column 2

excludes LO, out of the concern that this variable is highly correlated with several other

features. Finally, column 4 introduces controls for economy-wide factors. Results are

17
consistent with our predictions. They indicate significant negative relations between AQ

and the cash flow sensitivity of investment. When we turn our attention to the other

institutional features (CIFAR, JE, CR, SR, and LO), we see that while most are associated

with lower investment-cash flow sensitivity, only CIFAR and CR are consistently

significant at conventional levels. This may be explained by a lack of power due to small

sample size and it is not entirely unexpected given related findings in prior studies (e.g.,

DeFond and Hung [2004], Bushman et al. [2004]). 9 Among the economic control

variables, size has the strongest effect. Our results also appear to be economically

significant, with the specification including all institutional factors (column 5) explaining

44 percent of the variation across countries in investment-cash flow sensitivity. AQ alone

explains 14 percent of the variation (column 1).

As a robustness check, we also employ a “direct” estimation procedure that

examines firm-year observations (untabulated). Specifically, we interact AQ with OCF in

Equation (1) (we also include AQ in this specification). An advantage of this approach is

that countries with larger economies comprise a larger share of the sample, so our sample

is more representative of real economic activity. To prevent bias from measurement

errors in Tobin’s Q from driving our results, we estimate the model using a two-stage

least squares (2SLS) procedure, whereby MTBi,t is instrumented by MTBi,t-2 and Kinti,t-3,

lagged two and three periods, respectively, where Kint is the capital intensity defined as

the ratio of net fixed assets to sales. Specification tests indicate that our instruments are

both relevant and valid. We also apply ordinary least squares (OLS) techniques to ensure

that our results do not hinge on a particular 2SLS specification. Standard errors in our

9
DeFond and Hung [2004] report that international variations in shareholder rights do not affect CEO
turnover. Bushman et al. [2004] find no relation between legal origin and their measure of financial
transparency.

18
pooled 2SLS specifications are adjusted for unspecified heteroskedasticity and serial-

correlation using a two-step efficient generalized method of moments (GMM) estimator

(Hayashi [2000]).10 Results are consistent with our results from the indirect specification.

The z- (or t-) statistics range from -3.81 to -9.98 for the interaction between cash flows

and our accounting quality index, and from -4.71 to -13.08 for the interaction with

CIFAR. As expected, the z-(or t-) statistics are positive and range from 8.37 to 28.13 for

cash flows. Finally, Alti [2003] suggests that there could be cash flow sensitivity even in

an economy without frictions, because contemporaneous cash flows may better reflect

information about short-term profitability than would the market-to-book equity ratio

(MTB). To control for this possible effect, we add operating cash flow (OCF) at time t+1

and t+2 (scaled by fixed assets) or the difference in operating cash flows. Although this

specification reduces the sample size (since we need two future years to estimate future

profitability), results are qualitatively similar. Thus, our cross-country tests support the

first hypothesis that higher-quality accounting reduces investment-cash flow sensitivity at

the firm level.

Our second hypothesis predicts that higher quality accounting should reduce

investment-cash flow sensitivity more in economies dominated by stock markets than in

those dominated by creditors. To test this prediction, we partition our cross-country

sample based on whether the ratio of stock market value to credit market value in the

country is above or below the median in the sample. Consistent with our hypothesis,

accounting quality is statistically significant in the sub-sample of equity-based economies

10
Alternatively, we remove the year dummies and allow for clustering of observations by year to control
for cross-correlation in the error term. Untabulated results are qualitatively similar.

19
(the t-statistic equals -2.06) but less than significant at conventional levels in the credit-

based economies (the t-statistic equals -1.25).11

5.2. Cash flow sensitivities and within-country differences: the US

Results for the US within-country tests reported in Table 3 are consistent with those for

the international cross-sectional tests above. Column 1 reports the results for our panel

specification where CFSI0 is the dependent variable and AQFS is the only independent

variable. In column 2, we introduce the set of control variables described above and in

the Appendix. In column 3, following Hovakimian and Hovakimian [2005], we transform

CFSI0 into a binary variable that takes the value of one if CFSI0 is greater than 0.05 and

zero otherwise, then use a logit specification with firm fixed effects to estimate the

coefficients. All of the specifications consistently indicate that AQFS is associated with a

reduction in the investment-cash flow sensitivity. This result is economically significant –

an increase of AQFS by one standard deviation leads to a decrease of 30 percent in the

average cash flow sensitivity.12 Thus, higher accounting quality is associated with higher

investment efficiency even in the most liquid market in the presence of predominant

public equity financing. Our results (untabulated) also hold under alternate specifications,

for example, when we use boot-strapped standard errors, add year dummies, use

observations from the last year only, control for the analyst coverage, and measure CFSI0

using cash flows lagged by one year. Auditing quality (BigSix) and analyst coverage

(AnalCov) are significantly negative in (untabulated) univariate regressions but these

11
The magnitudes of the coefficients also are different (-0.15 for the equity-based sample and -0.09 for the
credit-based sample). The small sample size prevents us from testing if these coefficient estimates differ
significantly; pooling the two samples is also not feasible because of the high collinearity between the
variable of interest in such a pooled sample. One should not necessarily conclude that accounting quality
plays no role in credit-based economies for the same reason.
12
We obtain this estimate by multiplying the value of the coefficient (1.62) by one standard deviation of
AQFS (0.014) and then dividing by the average CFSI0 (0.076).

20
variables become insignificant in some multivariate specifications. The other control

variables generally have the expected signs and are generally consistent with Hovakimian

and Hovakimian [2005]. For example, we would expect that larger, more profitable and

more stable firms would be more transparent (see, among others, Lang and Lundholm

[1993]). Hence, these firms should exhibit less investment-cash flow sensitivity than

small and unprofitable and unstable firms. Consistent with these expectations, the

coefficients on Size, ROA, Z-score, and CFO are negative, as expected, while the

coefficients on σ(CFO) and M2B are positive.

5.3. Cash flow sensitivities and within-country differences: Japan

Untabulated results for the Japan within-country tests indicate that, contrary to the

US results, there is no statistically significant negative relation between accounting

quality and investment-cash flow sensitivity. Although the Japanese sample size is

smaller than the US sample size (approximately 15,000 observations versus 30,000 for

the US), it is still sufficiently large to rule out statistical power as an explanation. This

lack of statistical significance is consistent with our second hypothesis that accounting

quality should have a smaller effect on investment efficiency in countries where bank

financing and keiretsu are important sources of capital than in countries where equity is a

dominant source of capital, because capital suppliers in credit-based economies have

alternative avenues for reducing information asymmetry.13

13
The mean and the standard deviation of the winsorized investment-cash flow sensitivity estimates are
0.03 and 0.13, respectively, for Japanese firms versus 0.07 and 0.20, respectively, for US firms. This
suggests that homogeneity does not explain the Japanese results.

21
7. Summary and Conclusion

This study examines how accounting quality relates to firm-level capital

investment efficiency. We consider several measures of accounting quality derived from

prior research and two different methods for estimating investment-cash flow

sensitivities. Our first hypothesis captures the intuition that higher quality accounting

should enhance investment efficiency by reducing information asymmetry between

managers and outside suppliers of capital. As argued in prior studies, information

asymmetry can generate liquidity constraints (for example, as investors ration capital to

protect themselves against adverse selection) or excess cash (for example, when it is

difficult for owner-principals to monitor spending by manager-agents). Both situations

generate inefficiencies in the investment process that accounting quality should mitigate.

Our test results are consistent with this prediction both across and within countries under

numerous alternative specifications. We find that the link between internally generated

cash flows and investment is weaker when accounting quality is high. However, our tests

do not address whether higher quality accounting operates primarily to mitigate the effect

of a lack of cash or of its excess. This is a question we leave to future research.

Our second hypothesis is that the effect of higher quality accounting on

investment-cash flow sensitivity should be stronger in economies where financing is

largely provided through arm’s length transactions, for example, where stock markets are

dominant sources of capital. In contrast, in countries where creditors provide more

capital, banks may be able to obtain information through alternative private channels and

may be better positioned to monitor managers directly once capital is supplied. Thus, we

predict a stronger (weaker) relation between accounting quality and capital investment

22
efficiency in countries with predominant equity (bank) financing of firm-level capital

investment. Our test results support this second hypothesis both across and within

countries. Overall, our findings lend support to the argument that accounting quality is an

institutional feature available to policy makers that enhances investment efficiency by

mitigating investment-cash flow sensitivity, and that this effect is stronger under

conditions where capital is provided through arms-length transactions.

23
References

Abel, A. B., 1983. Optimal investment under uncertainty. American Economic Review
73(1): 228-233.

Alti, A. 2003. How sensitive is investment to cash flow when financing is frictionless?
Journal of Finance 63(2): 707-722.

Antle, R., and G. Eppen, 1985. Capital rationing, and organizational slack in capital
budgeting. Management Science 31(2): 163-174.

Bagehot, W. 1873. Lombard Street: A Description of the Money Market. Irwin (1962).

Baker, M., 2000. Career concerns, and staged investment. Working paper, Harvard
University.

Ball, Ray, S.P. Kothari, and A. Robin, 2000. The effect of international institutional
factors on properties of accounting earnings. Journal of Accounting and Economics 29:
1-51.

Beck, T., R. Levine, and N. Loayza. 2000. Finance and the source of growth. Journal of
Monetary Economics 46: 31-77.

Bertrand, M., and S. Mullainathan. 2003. Enjoying the Quiet Life? Corporate governance
and managerial preferences. Journal of Political Economy 111(5): 1043-1075.

Bhattacharya, U., H. Daouk, and M. Welker, 2003. The world price of earnings opacity.
The Accounting Review 78(3): 641-678.

Blanchard, O., F. Lopez-de-Silanez, and A. Shleifer, 1994. What do firms do with cash
windfalls? Journal of Financial Economics 36(3): 337-360.

24
Burgstahler, D., and I. Dichev. 1997. Earnings management to avoid earnings decreases
and losses. Journal of Accounting and Economics (24): 99-126.

Bushman, R., J. Piotroski, and A. Smith. 2004. What determines corporate transparency?
Journal of Accounting Research 42(2): 207-252.

Bushman, R., A, Smith, and F. Zhang. 2005. Investment-cash flow sensitivities are really
capital investment-working capital investment sensitivities. Working paper, University of
North Carolina.

Claessens, S., S. Djankov, and L. H. P. Lang. 2000. The separation of ownership and
control in East Asian corporations. Journal of Financial Economics 58(1): 81-112.

Cobb, C., and P. Douglas. 1928. A theory of production. American Economic Review
8(1): 139-65.

Dechow, P., and I. Dichev. 2002. The quality of accruals and earnings: The role of
accrual estimation errors. The Accounting Review 77 Supplement: 35-59.

DeFond, M., and M. Hung, 2004. Investor protection and corporate governance:
Evidence from worldwide CEO turnover. Journal of Accounting Research 42(2): 269-
312.

Degeorge, F., J. Patel, and R. Zeckhauser, 1999. Earnings management to exceed

thresholds. Journal of Business 72(1): 1-33.

Demirguc-Kunt, A., and V. Maksimovic. 1998. Law, finance, and firm growth. Journal
of Finance 53: 2107-37.

Demirguc-Kunt, A., and V. Maksimovic. 2002. Funding growth in bank-based and

market-based financial systems: evidence from firm-level data. Journal of Financial
Economics 65(3): 337-63.

25
Fazzari, Steven, R. G. Hubbard, and B. Petersen. 1988. Investment and finance
reconsidered. Brookings Papers on Economics Activity: 141-195.

__________. 2000. Investment-cash flow sensitivities are not useful: a comment on

Kaplan and Zingales. The Quarterly Journal of Economics 115(2): 695-705.

Francis, J., I. Khurana, and R. Pereira. 2005. Global evidence on incentives for voluntary
accounting disclosures and the effect on cost of capital. Forthcoming, The Accounting
Review.

Gorton, G., and F. Schmid. 2000. Universal banking and the performance of German
firms. Journal of Financial Economics 20(1): 29-80.

Hayashi, F. 1982. Tobin’s Marginal q and Average q: A Neoclassical Interpretation.

Econometrica 50(1): 213-224.

__________. 2000. Econometrics. Princeton, Princeton University Press.

Healy, P., and K. Palepu, 2001. Information asymmetry, corporate disclosure, and the
capital markets: A review of the empirical disclosure literature. Journal of Accounting
and Economics 31(1-3): 405-440.

Heaton, J. 2002. Managerial optimism and corporate finance. Financial Management

31(2): 33-45.

Hovakimian, A., and G. Hovakimian. 2005. Cash flow sensitivity of investment. Working
paper, City University of New York.

26
Hoshi, T., A. Kashyap, and D. Scharfstein. 1991. Corporate structure, liquidity and
investment: Evidence from Japanese industrial groups. The Quarterly Journal of
Economics 106(1): 33-60.

Hubbard, R. G. 1998. Capital-market imperfections and investment. Journal of Economic

Literature 36(1): 193-225.

Jensen, Michael. 1986. Agency costs of free cash flow, corporate finance, and takeovers.
American Economic Review 76(2): 323-329.

Kaplan, S., and L. Zingales. 1997. Do financing constraints explain why investment is
correlated with cash-flows. The Quarterly Journal of Economics 112(2): 169-215.

Kaplan, S., and L. Zingales. 2000. Investment-cash flow sensitivities are not valid
measures of financing constraints. The Quarterly Journal of Economics 112(2): 707-712.

King, R., and R. Levine. 1993. Finance and growth: Schumpeter might be right. The
Quarterly Journal of Economics 108(3): 717-737.

La Porta, R., F. Lopez-De-Silanes, A. Shleifer, and R. Vishny. 1997. Legal determinants

of external finance. Journal of Finance 52(3): 1131-1150.

Lang, M., and R. Lundholm. 1993. Cross-sectional determinants of analysts’ ratings of

corporate disclosures. Journal of Accounting Research 31(2) (Autumn): 246-271.

Leuz, C., D. Nanda, and P. Wysocki. 2003. Earnings management and institutional
factors: An international comparison. Journal of Financial Economics 69(3): 101-122.

Morck, R., A. Shleifer, and R. Vishny. 1990. The stock market and investment: Is the
market a sideshow? Brookings Papers on Economic Activity 1: 157-215.

27
Modigliani, F., and M. Miller. 1958. The Cost of capital, corporation finance, and the
theory of investment. American Economic Review 48: 261-297.

Myers, S. 1984. The capital structure puzzle. Journal of Finance 39(3): 575-592.

Myers, S., and N. Majluf. 1984. Corporate financing and investment decisions when
firms have information that investors do not have. Journal of Financial Economics 13(2):
187-221.

Rajan, R., and L. Zingales. 1998. Financial dependence and growth. American Economic
Review 88: 559-586.

__________. 2000. The great reversals: The politics of financial development in the 20th
Century. OECD Working Paper.

Rousseau, P., and P. Wachtel. 1998. Financial intermediation and economic performance:
Historical evidence from five industrialized countries. Journal of Money, Credit and
Banking 30(4): 657-78.

Schwartz, R. 1997. Legal regimes, audit quality and investment. The Accounting Review
72(3): 385-406.

Tobin, J. 1969. A general equilibrium approach monetary theory. Journal of Money,

Credit and Banking 1: 15-29.

Yoshikawa, H. 1980. On the “q” theory of investment. American Economic Review

70(4): 739-43.

Whited, T., 1992. Debt, liquidity constraints and corporate investment: Evidence from
panel data. Journal of Finance 47: 1425-1460.

28
World Bank. 2003. World Bank Indicators. Washington, D.C.

Wugler, J. 2000. Financial markets and the allocation of capital. Journal of Financial
Economics 58: 187-214.

29
 Appendix: Details of Variable Estimation

1) Accounting quality.

Cross-country sample

We create AQ by summing binary measures of four variables adapted from

Bhattacharya et al. [2003] and Bushman et al. [2004] as described in Section 4.3 above.

US sample

We first estimate the following model:

TCA1j,t = ϕ0,j + ϕ1,j CFOj,t-1 + ϕ2,j CFOj,t + ϕ3,j CFOj,t+1 + uj,t

where TCAj,t = ∆CAj,t – ∆CLj,t – ∆Cashj,t + ∆STDEBTj,t = total current accruals in year t,

CFOj,t = NIBEj,t – TAj,t = firm j’s cash flow from operations in year t, NIBEj,t = firm j’s net

income before extraordinary items (Compustat #18) in year t, TAj,t = (∆CAj,t – ∆CLj,t –

∆Cashj,t + ∆STDEBTj,t – DEPNj,t) = firm j’s total accruals in year t, ∆CAj,t = firm j’s

change in current assets (Compustat #4) between year t-1 and year t, ∆CLj,t = firm j’s

change in current liabilities (Compustat #5) between year t-1 and year t, ∆Cashj,t = firm

j’s change in cash (Compustat #1) between year t-1 and year t, ∆STDEBTj,t = firm j’s

change in debt in current liabilities (Compustat #34) between year t-1 and year t, DEPNj,t

= firm j’s depreciation and amortization expense (Compustat #14) in year t. We multiply

the variance by minus one, so that a higher value of AQ corresponds to higher accounting

quality.

In addition, we follow Francis et al. [2005] and use the model modified to include

plant, property and equipment (PPE) and change in revenues (scaled by average assets).

McNichols [2002] proposes this extension, arguing that the change in sales revenue and

30
PPE are important in forming expectations about current accruals, over and above the

effects of operating cash flows. She shows that adding these variables to the cross-

sectional Dechow and Dichev [2002] regression significantly increases its explanatory

power, thus reducing measurement error. However, the drawback of using this

specification for our purpose is that it may include a mechanical link with our right-hand

side variables by including capital investment in the regression. As a robustness check,

we estimate a second model:

TCA2j,t = ψ0,j + ψ1,j CFOj,t-1 + ψ2,j CFOj,t + ψ3,j CFOj,t+1 + ψ4,j ∆Revj,t + ψ5,j PPEj,t + ηj,t

where ∆Revj,t = firm j’s change in revenues (Compustat #12) between year t-1 and year t,

PPEj,t = firm j’s gross value of PPE (Compustat #7) in year t. To conserve space, we only

tabulate the results from the first specification but results are very similar when we use

the second one.

Japanese sample

We follow a similar approach in our Japanese sample. Specifically, we use the

following items from PACAP: INC9 and INC8 for NIBE; BAL6 for CA; BAL13 for CL;

BAL1 for Cash; JAF34 for STDEBT; JAF74 for DEPN; INC1 for Rev; and the sum of

JAF16, JAF17, JAF18, JAF19, JAF20 and JAF21 for PPE.

2) Variable I for cash flow sensitivities.

Cross-country sample

To calculate investment-cash flow sensitivities in the cross-country sample, we

estimate the following model:

31
 Ii,t/Ki,t-1 = βi0 + β1 OCFi,t /Ki,t-1 + β2 MTBi,t + εi,t

where Ii,t is the investment in fixed assets for firm i in year t, defined as the change in net

fixed assets (Global Vantage data76) plus depreciation expense (data11). Ki,t-1 is the total

fixed assets for firm i in year t-1. MTBi,t,, our proxy for the Q ratio, is measured as the

ratio of the sum of the market value of equity (data13 multiplied by data3) and the book

value of debt divided (data106, data136 and data137 minus data138 and data139) by the

book value of total asset (data89). One advantage of using the (quasi) market value of

assets instead of their book value is that we do not drop from our sample firms with a

negative equity book value. Consistent with prior literature (e.g., Hoshi et al. [1991]), Q

is calculated at the beginning of the period. Following previous studies, 14 OCFi,t is

calculated as the sum of net income (data32) and depreciation expense (data11).

US Sample

To calculate the cash flow sensitivity in our within-country samples, we estimate

the model proposed by Hovakimian and Hovakimian [1985]:

CFSI0,i,t = CFWAI0,i,t - AIi,t

=1/n Σs=1t [ (CFi,s / (Σs=1t CFi,s)) * Ii,s] - 1/n Σs=1t Ii,s

where n is the number of annual observations for firm i and t indicates the time period. I

denotes investment, defined as capital expenditures (Compustat Item 128) divided by the

beginning-of-period net capital. CF denotes cash flow and is defined as the sum of the

income before extraordinary items (Compustat Item 18) and depreciation and

14
See, among others, Morck et al. [1990], Hoshi et al. [1991], Whited [1992], Kaplan and Zingales [1997,
2000], Hovakimian and Hovakimian [2005]. We purposely do not correct our estimates of cash flows for
any “working capital” effect for two reasons. First, we want to be consistent with prior literature on cash
flow sensitivity of investment. Second, some firms lack the relevant information in our international
database; thus, correcting our estimates of cash flows would introduce a selection bias. Our approach also
has been used in prior international studies (e.g., Ball et al. [2000]).

32
amortization (Item 14), divided by the beginning-of-period net capital (Item 8). As an

untabulated robustness check, we also consider a cash flow measure from Bushman et al.

[2005] that excludes working capital accruals. Our results still hold. To avoid negative

and extreme weighted values, we follow Hovakimian and Hovakimian [2005] and set the

negative cash flows to zero. The variable is estimated using a ten-year rolling window.

This specification does not require the estimation of Q. By doing so, we have a natural

robustness check for the efficiency of our various statistical corrections in our cross-

country sample.

In addition, since investment and cash flow are measured over an annual period,

their exact timings may not coincide. In order to account for the possibility that

investment may be financed with cash flows from the previous fiscal year, we also

estimate CFSI based on OCF, which is lagged relative to investment:

CFSI-1,i,t = CFWAI-1,i,t - AIi,t

= 1/n Σs=1t [ (CFi,s-1 / (Σs=1t CFi,s-1)) * Ii,s] - 1/n Σs=1t Ii,s

Results (untabulated) are not affected by using this alternative definition.

Japanese sample

We follow a similar approach in our Japanese sample. Specifically, we use the

following items from PACAP: change in PPE plus JAF74 scaled by beginning period

PPE for I and NIBE plus JAF74, divided by the beginning-of-period net PPE for CF.

3) Control variables in the cash flow sensitivity regressions.

Cross-country sample

The institutional feature of primary interest is accounting quality, but we also

consider financial disclosure, creditor rights (CR), shareholder rights (SR), judicial

33
efficiency (JE) and legal origin (LO). As in many previous studies (e.g., Battacharya et

al. [2003]), we measure financial disclosure quality using an index created by the Center

for International Financial Analysis and Research (CIFAR). CR and SR are based on the

summary measures of creditor rights (Cr5) and shareholder rights (Sr8) obtained from La

Porta et al. [1997]. Our measure of judicial efficiency, JE, is the sum of the five proxies

for judicial efficiency (Rl1 - Rl5) found in La Porta et al. [1997]. LO takes the value of

one if the country has a legal origin rooted in the Common law tradition (zero otherwise).

We also control for economic characteristics: Log(GDP/Capita) is gross domestic

product per capita as reported by the World Bank [2003]; K-Intensity is the average

capital intensity for a given country defined as the ratio of net fixed assets to sales; Size is

the average firm size in the country measured as the log of the sum of the market value of

equity and the book value of debt.

US sample

As proposed by Hovakimian and Hovakimian [2005], we control for size (log of

Compustat item 6), market-to-book ratio (item 6 plus the product of items 25 and 199

minus item 60 and item 35, scaled by item 6), ROA (a measure of profitability calculated

as the ratio of Compustat item 170 divided by item 6), dividend payout ratio (a dummy

variable that takes the value of one if item 21 or 127 is greater than zero, zero otherwise),

leverage (item 9 scaled by data 9 plus the product of items 25 and 199), Z-score (a

measure of bankruptcy risk defined as 3.3 times item 170 + item 12 + one fourth of item

36 plus one half of the difference between item 4 and item 5, scaled by item 6), the

standard error of CFO, tangibility (a measure of bankruptcy cost defined as the ratio of

item 8 and item 6), and R&D (a dummy variable that takes the value of one if item 46 is

34
greater than zero, zero otherwise). In addition, we include two measures of financial

slack, CFOsale (the ratio of CFO divided by item 12) and Slack (the ratio of item 1 and

item 8). Finally, we control for auditing quality by including BigSix, a dummy variable

taking the value of one if the firm is audited by a Big Six auditor (data # 159 between 10

and 89), zero otherwise. Generally, we expect bigger firms, with higher and more stable

cash flows, more tangible assets and fewer investment opportunities to be less sensitive to

internally generated cash flows. We also control for analyst coverage in an untabulated

robustness check. To measure analyst coverage, we use data obtained from the I/B/E/S

Historical Summary File. We use AnalCov, the log of the number of analysts (plus one

since the log of zero is not defined) reported by IBES as covering the firm, as our proxy

for analyst coverage. For each year, we set the number of analysts following a firm as

the maximum number of analysts who make annual earnings forecasts in any month over

a twelve-month period. We assume that firms not covered by I/B/E/S are not covered by

analysts.

Japanese sample

We follow a similar approach in our Japanese sample. Specifically, we use the

following items from PACAP: log of BAL9 for size; the sum of BAL9 and market value

minus BAL21 scaled by BAL9 for market-to-book ratio; the sum of INC9, INC8 and

INC7 scaled by BAL9 for ROA; MKT1 for calculating dividend payout ratio; the ratio of

BAL14 divided by market value and BAL14 for leverage; 3.3 times the sum of INC9,

INC8 and INC7 plus INC1 plus 0.25 time BAL20 plus 0.5 times the difference between

BAL6 and BAL13, scaled by BAL9 for Z-score; the ratio CFO over INC1 for tangibility;

35
the ratio of CFO divided by INC1 for CFOsale; and the ratio of BAL1 scaled by the sum

of JAF16 through 21 and 23 for Slack.

36
 Table 1: Correlations among summary measures1

AQ CIFAR CR SR JE
0.13
CIFAR (0.48)
-0.17 0.04
CR (0.33) (0.83)
0.31 0.39 0.22
SR (0.08) (0.03) (0.20)
-0.16 0.44 -0.15 -0.08
JE (0.37) (0.01) (0.41) (0.66)
0.27 0.35 0.38 0.66 -0.11
LO (0.12) (0.04) (0.03) (0.00) (0.55)

1
AQ is a measure of accounting transparency calculated as the sum of Earnings aggressiveness, Loss
avoidance, Earnings smoothing (Bhattacharya et al. [2003]) and Timeliness (Bushman et al. [2004]).
CIFAR is based on a disclosure index calculated by the Center for International Financial Analysis and
Research. CR is a summary measure for creditor rights (as reported by La Porta et al. [1997]). SR is a
summary measure for shareholder rights (as reported in La Porta et al. [1997]). JE is a measure of judicial
efficiency based on the five measures found in La Porta et al. [1997]. LO is an indicator variable that takes
the value of one if a country’s legal system has as a Common law origin, zero otherwise (as reported by La
Porta et al. [1997]). The p-values are reported in parentheses.

37
 Table 2: Investment-cash flow sensitivity – Cross-country sample1

Predicted
Variable Dependent Variable: Investment-cash flow sensitivity
Sign
(1) (2) (3) (4)
Intercept (+) 0.527 1.725 1.558 1.994
(5.07) (3.66) (3.12) (3.69)
AQi (-) -0.120 -0.141 -0.129 -0.113
(-2.74) (-2.63) (-2.40) (-2.01)
CIFARi (-) -0.014 -0.013 -0.018
(-2.29) (-1.98) (-2.52)
JEi (-) -0.008 -0.010 -0.016
(-0.63) (-0.81) (-1.19)
CRi (-) -0.089 -0.071 -0.069
(-2.48) (-1.96) (-1.88)
SRi (-) 0.038 0.056 0.075
(0.92) (1.24) (1.52)
LOi (-) -0.126 -0.207
(-1.33) (-1.86)
Log(GDP/capita) (-) -0.000
(-1.37)
K-Intensity (-) 0.002
(0.06)
Size (-) -0.030
(-3.39)
Number of
34 33 33 33
observations
R-square 13.96% 42.05% 43.94% 55.32%

1
AQ is a measure of accounting transparency calculated as the sum of Earnings aggressiveness, Loss
avoidance, Earnings smoothing (Bhattacharya et al. [2003]) and Timeliness (Bushman et al. [2004]).
CIFAR is based on a disclosure index calculated by the Center for International Financial Analysis and
Research. JE is a measure of judicial efficiency based on the five measures found in La Porta et al. [1997].
CR is a summary measure for creditor rights (as reported by La Porta et al. [1997]). SR is a summary
measure for shareholder rights (as reported by La Porta et al. [1997]). LO is an indicator variable that takes
the value of one if a country’s legal system has as a Common law origin, zero otherwise (as reported by La
Porta et al. [1997]). Log(GDP/Capita) is gross domestic product per capita. K-Intensity is the average
capital intensity for a given country defined as the ratio of net fixed assets to sales. Size is the average
firm size in the country. Regressions are estimated using ordinary least squares. The t-statistics with
standard errors adjusted for heteroskedasticity are reported in parentheses.

38
 Table 3: Investment-cash flow sensitivity – US sample1

Variable Dependent Variable: Investment-cash flow sensitivity

(1) (2) (3)

AQFS -2.48 -1.624 -27.57
(-9.17) (-6.24) (-10.26)
BigSix -0.111 -0.17
(-2.41) (0.43)
LogAsset -0.028 -0.559
(-6.85) (-9.48)
Mkt-to-Book 0.048 1.257
(5.92) (11.76)
ROA -0.434 -6.124
(-3.92) (-5.88)
σ(CFO) 0.639 11.916
(5.18) (12.14)
Z-score -0.029 -0.401
(-2.46) (-2.91)
Tangibility 0.131 3.209
(3.18) (5.45)
R&D 0.000 -0.043
(0.03) (-0.22)
K-structure -0.036 1.33
(-1.09) (3.59)
Mean K-structure 0.225 1.129
(3.47) (1.03)
CFOsale -13.717 -48.145
(-3.72) (-1.60)
Slack 0.016 0.272
(1.73) (5.03)
Dividend 0.010 0.395
(0.67) (2.64)
Number of observations 28,353 28,353 12,420
R-square 68.24% 70.10%

1
The dependent variable is a measure of cash flow sensitivity as defined by Hovakimian and Hovakimian
[2005]. AQFS is a measure of accounting quality as defined by Francis et al. [2005]. BigSix is a dummy
variable that takes the value of one if the firm is audited by a Big Six auditor (data # 159 between 10 and
89), zero otherwise. LogAsset is the log of Compustat item 6. Mkt-to-Book is item 6 plus the product of
item 25 and 199 minus item 60 and item 35, scaled by item 6. ROA is a measure of profitability calculated
as the ratio of Compustat item 170 divided by item 6. σ(CFO) is the standard deviation of cash flows from

39
 operations. Z-score is a measure of bankruptcy risk defined as 3.3 times item 170 + item 12 + one fourth of
item 36 plus one half of the difference between item 4 and item 5, scaled by item 6. Tangibility is a
measure of bankruptcy cost defined as the ratio of item 8 and item 6. R&D is a dummy variable that takes
the value of one if item 46 is greater than zero, zero otherwise. K-structure is item 9 scaled by data 9 plus
the product of item 25 and 199. Mean K-structure is the mean of K-structure at the industry level. Slack
the ratio of item 1 and item 8. Dividend is a dummy variable that takes the value of one if item 21 or 127 is
greater than zero, zero otherwise. Columns 1 and 2 are estimated using ordinary least squares. Column 3
is estimated using a logit specification. All specifications use firm fixed effects. The t-statistics are
reported in parentheses.

40

View publication stats