NBER

WO~G

PAPER

SERIES

FINANCIAL

DEPENDENCE

AND

GROWTH

Raghuram

G. Rajan

Luigi Zingales

Working

Paper 5758

NATIONAL

BUREAU

OF ECONOMIC Avenue MA 02138 1996

RESEARCH

1050 Massachusetts Cambridge, September

We thank Gene Fama, Peter Klenow, David Scharfstein, Robert Vishny, University comments. of Chicago, the Finance

Krishna Kumar, Canice Prendergast, and participants seminar at UCLA, Bank. We gratefully in the Junior and the NBER excellent research acknowledge

Andres Rodriguez-Clare, lunch group at the Institute for usefi.d A preliminary Finance and support from NSF

Faculty Summer financial

Jayanta Sen and Alfi-ed Shang provided by the World and Growth. This paper is part of NBERs Any opinions Research.

assistance. in Corporate

study was supported grant #SBR-9423645. Economic

research programs

Fluctuations

expressed

are those of the authors and not those

of the National

Bureau of Economic

O 1996 by Raghuram to exceed including 0 notice,

G. Rajan and Luigi Zingales. may be quoted without is given to the source.

All rights reserved. explicit permission

Short sections provided

of text, not

two paragraphs,

that full credit,

NBER

Working

Paper 5758 1996

September

FINANCIAL

DEPENDENCE

AND

GROWTH

ABSTRACT

Does correlation

finance

affect

economic

growth?

A number

of studies financial

have identified

a positive

between

the level of development

of a countrys

sector and the rate of growth does not necessarily facilitates development sectors that with

of its per capita income. imply a causal growth

As has been noted elsewhere, This paper examines

the observed whether

correlation

relationship. by scrutinizing

financial

development

economic reduces

one rationale for such a relationship; Specifically, develop

that financial industrial

the costs of external finance to firms. more in need of external finance financial markets.

we ask whether

are relatively

disproportionately

faster in countries

more developed 1980s. causality,

We find this to be true in a large sample of countries to be driven by omitted variables, outliers,

over the or reverse

We show

this result is unlikely

Raghuram Kellogg

G. Rajan School of Management University Road

Luigi Zingales Graduate University Chicago, and NBER School of Business of Chicago IL 60657 .edu

Graduate

Northwestern 2001 Sheridan Evanston, and NBER

1101 East 58th Street

IL 60208 .uchicago.edu

fac_rajan@gsbvax

luigi@gsblgz.uchicago

There is a large literature sizes the links between the rate of growth financial

- dating at le=t

ss far back = Schumpeter of a countrys

(1911) - that empha-

the state of development income.

financial sector and the level and essentially is that the services the

of its per capita

The argument

sector provides

of reallocating hazard,

capital

to the highest

value use without

substantial

risk of loss through catalyst economic (1967), of economic

moral growth.

adverse

selection,

or transactions between

costs are an essential financial development and

The existence

of a correlation

growth has been documented Goldsmith (1969) and McKinnon

in a number of empirical (1973). Economic

studies starting

with Cameron spurts of

historians

have identified

growth in certain countries Cameron

with the development

of specific financial institutions. the economic

For example,

(1961) and Gerschenkron century

(1962) have associated

growth of France in the Mobilier. 1 Nevertheless, arguing that correlation

middle of the nineteenth some economists

with the est abolishment of the Credit the interpretation North, of these studies, development

have questioned

does not imply causality. growth, it is economic

Paraphrasing

financial

does not cause economic

growth.2 recent paper, Levine and King (1993a) investigate the causality problem

In an important following

a post hoc, ergo pTopteT hoc approach. development is a good predictor

They show that the predetermined of growth over the next

component While

of financial

10 to 30 years.

their paper does suggest that the observed doubts about causality. The sceptic

correlation

is not spurious,

it does not lay to rest all

could still offer a number

of arguments. omitted variable (in

First, both financial development such as the propensity certain macroeconomic surprising that growth of households models)

and growth could be driven by a common in the economy to save.

Since endogenous

savings

affects the long run growth rate of the economy, development are correlated.

it may not be is also theory

and initial financial

This argument

hard to refute with simple cross-country

regressions.

In the absence

of a well accepted

1Adopting a diflerent tack, I{indleberger (1993) discusses the importance of finance to the outcome of wars (for instance, Britains ability to mobilize financial resources after the Glorious Revolution of 1688 may explain why, with one-third Frances manpower, it could defeat the French) and asks Can Finance be relewnt to military outcomes, as widely insisted on, but not to the course of economic development? The early studies have typically been cautious about their findings. For instance, Goldsmith (1969, p48) writes In most countries, a rough parallelism can be observed between economic and financial development if periods of several decades are considered. There is no possibility, however, of establishing with confidence the direction of the causal mechanism, i.e., of deciding whether financial factors were responsible for the acceleration of economic development or whether financial development reflected economic growth whose mainsprings must be sought elsewhere.

of growth,

the list of potential

omitted

variables variables

that financial to include

sector development of conjecture

might be a (on this see

proxy for is large, and the explanatory Paxson, 1996). there is a potential

a matter

Second, messured

problem

of anticipation.

Financial

development

- typically

by the level of credit and the size of the stock m~ket may predict economic capitalizes growth

(but see King and Levine (1993a) financial markets anticipate while may

for other proxies) future growth;

simply because

the stock market

the present

value of growth

opportunities, development

financial

institutions

lend more if they think sectors will grow. Thus financial rather than a causal factor. in a more causal sense it is necessary affects economic

simply be a leading indicator To interpret

these correlations

to identify the mechanism its working. Our

through which financial development paper is an attempt The fundamental to do this.

growth and to document

role played by the financial sector is to facilitate

the reallocation opportunities

of funds towards

from agents (individuals) agents (firms) that financial and investing, extent

with an excess of capital given their investment of funds vis h vis their investment First, by reducing

with a shortage development

opportunities.

This implies costs of saving Second, to the

has two effects.

the transactions in general. problems

it lowers the overall cost of capital markets and institutions

in the economy

that finmcial

help firms overcome

of moral hazard

and adverse selection,

it should reduce the costs of external

finance vis ~ vis the cost of internal

funds such as cash flows. Financial development, finance. thus, should reduce both the cost effect. of capital and the differential If the financial sector works

cost of eztemal through

Our paper focuses finance,

on the second

the cost of external

firms that are typically with well developed on external financial

short of funds given investment financial sectors. The test then are relatively

opportunities is whether better off

should do better in economies firms (or industries) in economies dependent

finance for their growth markets.

with well developed about

A finding that they are could test. a

be the smoking

gun in the debate

causality.

There are two virtues to this simple

First, it focuses on the details of a mechanism stronger test of causality. its value depends Second,

by which finance affects growth, (and industry)

thus providing effects. Though

it can correct for fixed country our micro-economic

on how reasonable

assumptions

are, it is less dependent

on a specific

macroeconomic we propose between

model

of growth. test. We identify an industrys generated need for external finance from data on especially for

Specifically, (the difference U.S. firms.

the following

investments

and internal c=h that capital

from operations) States,

Under

the assumption

markets

in the United this method

the large listed firms we analyze, industrys technological demand

are relatively

frictionless,

allows us to identify an the industries that initially

for external financing. financing

We then examine whether fsster in countries

that are more dependent have better developed

on external

grow relatively

financial consistent

markets and institutions. with this. In a country the difference deviation which is one standard rates between deviation above whose is

We find evidence the mean of financial financial 1% more industries dependence (annually,

development, is one standard

in growth above

an industry

the mean and the average country. Moreover,

industry,

and in real terms)

than in the average

we find that

that generate substantial financial

cash flow from operations The intensity

grow relatively

faster in countries dependent on

with underdeveloped

systems.

of investment

in industries

external finance is disproportionately Finally, firms in financially more developed for a countrys by an omitted financial dependent markets.

higher in countries industries

with more developed

financial markets. with

are disproportionately

larger in countries

Our results are robust to the introduction and human capital, making it unlikely

of various proxies that they are driven

level of development variable.

Let us be careful about

what we find, and about

what we have little to say. Our findings the ex post growth of sectors financial development and

suggest that the ex ante development dependent on external finance.

of financial markets facilitates

This implies that the link between

growth identified financial markets

elsewhere

may stem, at le=t

in part, from a channel identified finance for firms.

by the theory: Of course, our

and institutions

reduce the cost of external

analysis suggests only that financial development funds internally. drives growth, Clearly, are. It is ultimately the availability

liberates firms from the drudgery of profitable investment

of generating that

opportunities

and we have little to say about

where these come from. on how acceptable reuon its underlying why some assumptions industries scale, dethe

the plausibility

of our work depends

One of our assumptions finance

is that there is a technological than others. To the extent

penal more on external

that the initial project

gestation

period,

the c=h

harvest period,

and the requirement

for continuing

investment

differ that

substantially

between industries, differences

this is indeed plausible.

More debatable

is our assumption

these technological However,

persist across countries

despite v=t differences sort hold: before

in local conditions. If pharmaceuticals are harvested a higher

all we really need is that statements

of the following period

require a larger initial scale ad than the textile gestation period industry in Korea. for a countrys

have a higher gestation

cashflows

in the U-S., it also requires

a larger initial scale and h=

Furthermore, that country other words, development of frictions savings

financial development

to have any effect on industry

growth

in In

we have to assume that firms finance themselves only if world capital tiect a countrys markets are not perfectly

largely in their own country. integrated can domestic

financial

growth.

There is a wealth of evidence the extremely

documenting between

the existence a countrys

in international

capital markets: (Feldstein

high correlation the strong

and its investments (French

and Horioka,

1980),

home

bias in portfolio returns other

investments (Bekaert

and Poterba 1995).

, 1991 ), and cross countries

differences

in expected

and Harvey,

We have little else to say about

these two assumptions

than noting findings.

that their failure would weaken the power of our tests but not necessarily

bias our

Finally, we address concerns

about

the interpretation

of our findings.

As argued earlier, the But this would

size of financial markets might reflect the extent to which growth is capitalized. not explain why there is a differential dependent on external finance, effect of financial development development

on the growth of industries to matter even when is that

nor why financial

continues

we use other proxies for it such as a countrys financial markets might develop

accounting

standards.

A second possibility

in anticipation

of the financing

needs of industry.

We regard of markets of de

this as unlikely for a number of reasons. is the outcome the colonial Silmes, of a rational plan.

It is not clear that the state of development colonies, it is determined

In many fomer

by the identity Lopez

power

and the legal and political (1996) ).3

systems

they imposed appear

(see La Porta,

Shleifer and Vishny

At best, markets

to develop

in response

to the

3For instance, India which has a common-law system introduced by the British has had a vibrant stock exchange since the l=t century (La Porta et al. argue that common law systems are more conducive to the development of equity markets). The stock exchange survived, albeit in a low energy state, even when many of the activities requiring external finance such as infrastructure development were taken over by the government

current public

(and past ) needs of industry corporate debt markets

(see Chandler

( 1977)s

discussion

of the development

of

in the United needs.

States in response

to the financing

needs of the

railroads)

rather than anticipated another

But this suggests responsible the current

explanation

of our results, dependent

Perhaps, sectors.

some factor This,

may have been explain explain

for the past growth development

of financially markets, sectors.

in turn, would would

of financial dependent

and the persistence

of the factor

the growth of financially would be correlated channels correlated

Thus the current development dependent sectors, financial

of financial markets any of the is strongly economy. is

with the growth

of financially

even absent development in that

we have discussed.

Indeed we find that a countrys of financially growth

with the ex ante proportion

dependent

industries

Nevertheless, higher

we still find that the relative developed

rate of financially

dependent

industries

in more financially

countries

even when we restrict and hence unlikely

our analysis

to sectors for the

that are small at the beginning state of development we instrument

of the period,

to be responsible

of the financial markets .4 Furthermore, measures

the bmic correlation

persists when markets an

with more exogenous accounting dependence

of the state of development

of capital

such as a countrys industrys financial

standards,

and when we include

the interaction dependent

between

and the ex ante proportion factor).

of financially

industries

in the economy While

(which is a proxy for the omitted is primarily

our interest

in the macro-economic

link between

financial

development

and growth, macroeconomic literature

there are at least two re~ons debate on whether

why our work may also be of interest in the current rationed. traded, First, U.S. much of the empirical companies. to the Second, in this

firms are credit on large, publicly between different

on this issue has focused in liquidity constraints across

(or Japanese)

The difference differences following literature

such firms is likely to be small relative financial markets which is our focus. the papers

between

industries

the influential have typically

work by Fazzari,

Hubbard

and Petersen

(1988),

sorted firms on some ex ante proxy

(such as the payment

of dividends

after India got independence. With the gradual withdrawal of the government from industry in the 1980s, the stock muket h= resumed its role as an important source of finance for young firms. This is consistent with Chandlers account of the financing of U.S. industries. Their growth was partly driven by the atilability of finance in a system that had largely been developed to fund infr=tructure such as railroads and CalldS.

or the existence higher sensitivity

of bank relationships) of investments to being

for financial

constraints.

They

then argue

that the are

to cashflows

for firms that the ex ante proxy of a differential A number wedge between

suggests

more susceptible internal

constrained

is evidence

the cost of

and external

funds among

these firms. really

of questions Second, could

have been raised by cashflows proxy for

this methodology. investment

First,

is the proxy thus driving

exogenous?

opportunities,

the correlation? depends constraints,

Final] y, as Kaplan

and Zingales

(1996)

point out, the validity flow sensitivity hold.

of such an approach

on the monotonicity

of the investment-c=h

to the proxy for financial

which they show does not necessarily test of the effects of financial (and, thus, growth) funds. and suggest a new way of of as more ex(we do constraints

By contr=t,

they argue that a more appropriate the differential level of investments

consists in analyzing a differential

across firms facing

wedge between

the cost of internal and external

The approach investigating ogenous

taken by our paper may deal with this criticism For instance, financial development policy

these issues.

may be thought

to a particular

industry

than its dividend in the literature

or its banking

relationships

not imply that none of the proxies flows proxy for investment evidence of financial

are exogenous).

Furthermore,

even if cashthat is by

opportunities,

it is the interaction

with financial development the Kaplan-Zingales

constraints. in industry

Finally, our paper addresses growth across countries wedge between

critique

testing for differences markets

with differentially

developed

financial So

and, thus, with a differential could be thought

cost of external

and internal funds. do fiect

our evidence

of M reaffirming

that financial work.

constraints

investment,

while finessing the debate surrounding Finally, the influence proportion Demirguc-Kunt of financial

the previous (1996)

and Maksimovic development

also use micro Using firm-level

data to develop data, they

a test of the

on growth. exceeds

estimate

of firms whose rate of growth

the growth

that could

have been supported

only by internal resources. is positively related

They then run a cross country regression and find that this proportion and to a measure of law enforcement. While

to the stock market turnover

their paper is similar in spirit to ours, there are two essential differences. of the internal potentially Another growth rate of a firm is dependent on the firms

First, their estimate While it is

characteristics.

more accurate difference

than our measure of external dependence, cross-country

it is also more endogenous. regressions, they focus on

is that in the spirit of traditional

between-country

differences,

while our focus is on within-country, innovation is = in this paper. We describe

between-industry

differences.

The latter is an important The rest of the paper section 1. In section growth.

follows.

our measure

of financial country

dependency characteristics,

in

2, we present our data on financial In section

development,

and industry and explore

3 we set up our main tests, we discuss the results in section 5. Section 6 concludes.

4,

the robustness

of our findings in section

A Measure

of External

Dependency
it must To

If the presence

of a well developed

financial market reduces the costs of external finance,

have a greater effect on the growth rate of firms that are more dependent investigate this, we have to determine measure

on external funds.

a firms demand for external finance over a certain period. of financial is for 1990. market development is for the year 1980, and of interest to be

Our earliest comprehensive our last systematic 1980-1990. We immediately aggregated

data on growth

This then defines the period

face some problems.

First,

a problem

of data availability;

the most

dis-

comprehensive

data we have for countries limited

is at the industry

level (data at the firm

level, if available, financing

is typically

to large listed firms).

Data on the actual use of external it would

is typically

not available. between

But even if it were, it would not be useable because for external funds and its supply.

reflect the equilibrium is precisely problem different typically

the demand

Since the latter A second needs of finance versus

what we are attempting

to test for, this information study

is contaminated. financing

is that we are not aware of any systematic industries, focuses either cross-sectionally

of the external

or over time.

Empirical

work in corporate

on the choice between

debt and equity, and not on the use of external

internal funds. 5 We, therefore, 1980s. sample start by studying the external for this. financing Compustat traded needs of U.S. companies does not contain firms, which over the

We use data from Compustat of U.S. firms, because

a representative large. market

is limited

to publicly

are relatively capital

Nevertheless,

we regard this = an advantage

for two re~ons.

First, in a perfect

Mayer (1990) and Demirguc-I{unt

and Maksimovic (1996) are exceptions.

Mayer uses country level data.

the supply of funds to firms is perfectly the actual amount in such an idealized the United typically

cl-tic

at the proper risk adjusted

rate. In such a mmket In other words, markets traded in

of external funds raised by a firm equals its desired amount, setting, the identification problem does not exist.

But capital

States are among

the most advanced in accessing

in the world,

and large publicly of external

firms

face the le=t

frictions

finance.

Thus the amount

finance used for but

by large firms in the United States is likely to be a relatively external finace.

pure measure of their demand

Note that we are not assuming that the U.S. financial markets are perfect,

only that they are among the least imperfect. A second resson for using a database on listed firms is that disclosure requirements imply

that the data on financing of external

are comprehensive.

In the rest of the paper, we will take the amount ss a proxy for the desired amount markets foreign

finance used by U.S. firms in an industry would

firms in the same industry developed. Next, interested

have liked to raise had their financial

been more

we have to define precisely

what we mean by external

and internal through

finance.

We are

in the amount of desired investment

that cannot be financed Therefore,

internal sources, on external

i.e., the cash flow generated

by the same business.

a firms dependence

finance is defined as the ratio of capital expenditures operations divided by capital expenditures. c~h flow from operations and increases

(Compustat

# 128) minus cash flow from is broadly defined ss the in inventories, includes changes In fact,

Cash flow from operations (Compustat

sum of Compustat decre~es

# 110) plus decre=es

in receivables,

in payables. 6 Note that this definition

in the non-financial in certain businesses

components

of net working capital = part of funds from operations.

these represent major sources sources of funds.7

(or uses) of funds, that help a firm avoid (or

force it to tap) external

Similarly, the dependence of equity issues (Compustat

on external equity finance is defined m the ratio of the net amount # 108 minus # 115) to capital expenditures. Finally, the investment

This item is only defined for cash flow statements with format codes 1, 2, or 3. For format code 7 we construct it as the sum of items # 123, 125, 126, 106, 213, 217. 71t could be argued that inter-firm trade credit should be viewed as a component of external financing. It iG unclear how much of trade credit is used to reduce transactions costs and how much is used for financing. Much trade credit is granted routinely and repaid promptly and typically, net trade credit for a firm (accounts receivable less payables) is small (see Petersen and Rajan (1996)). This may be why trade credit is typically treated m part of operations in capital budgeting exercises. We adhere to this tradition.

intensity # 8).

is the ratio of capital

expenditure

to net property

plant and equipment

(Compustat

To make these measures comparable we have to choose how to aggregate we sum the numerator firms use of external

with the industry

level data we have for other countries, and over time. For instance, For each firm, we sum the expenditure

these ratios across compmies over time before

and denominator

dividing.

finance over the 1980s and then divide by the sum of capital on external of outliers. finance in the 1980s. ratios

over the 1980s to get the firms dependence temporal however, fluctuations and reduces the effects

This smooths across firms, the

To summarize

we use the industry

median.

We do this to prevent

large firms from swamping

information possible

from small firms; for instance,

we know that IBMs free cash flow does not alleviate firms. of a life cycle of financing firms are more wisdom in

cash flow shortages we are concerned on external

of small computer about

Finally, dependent

the existence

financing

early in their life than later which is common (though we were hard-pressed

the corporate

finance literature

to find formal empirical

studies

of this phenomenon). Figure 1 shows such common wisdom to be true. It plots the median financing and invest-

ment needs across U.S. firms = a function (IPO). Not surprisingly, equity).

of the number of years since the initial public offering firms raise a substantial amount of external

in the year of the IPO, More interestingly,

funds (especially approximately external

this continues net equity

albeit on a smaller scale up to issues go to zero and the usage of appears to be fairly standard

the 10th year.

After that period, zero. Although

finance fluctuates industries,

around

this pattern both

across different stage

the level of external industries.

financing,

at an early stage and at a later the pattern industry of

is quite different and financing

across

For example,

in figure 2, we report in the computer

investment

in two different industries.

Companies

show

an intensive use of external external

funds in the first decade of their existence, later on. By contr=t, external

with occasional

spurts of

and equity financing

finance goes to zero very shortly (and equity finance)

after the IPO in the steel industry,

but further substantial

needs of external

might arise later in life of a company. In Table 1, we look at these differences Standard Classification Code (ISIC) more systematically. fraction We tabulate by International

the median

of investments

U.S. firms financed

externally

(column

I) and financed

with equity

(column

II) during the 1980s. To highlight

the

need for investments, property

in column 111we also report the level of capital expenditures We restrict our attention to those manufacturing We report

divided by net industries for

plant and equipment.

which we have value-added

data from the United Nations Statistics.

these figures for

all firms, mature firms (firms that did an IPO at least 10 years ago), and young firms (firms that went public emerge industry = in the I=t 9 years). Not surprisingly, among mature firms, radio and computers finance, while tobacco expenditures. financing is the This

the two industries

with the largest need for external

with the largest excess amount in its large equity

of internal funds relative to capital The clustering of external

is also reflected

repurchases.

in a few

sectors is even more pronounced Given that the life-cycle of external

among young firms. of external financing differs by industry, is debatable. the most approneeds

pattern financing

priate me=ure

needs in other countries

The financing

of mature companies countries. developing However, countries,

represent a better proxy for the financing young firms may make up a larger fraction

needs of existing of the industry,

firms in other especially in

than they do in the United States.

In the absence

of more data on indusin the U.S. to me=ure

try composition, dependence.

we will use the median we will examine

across all firms in the industry

However,

if the result is sensitive to using just young or old firms.

1.1

Is the

Financial

Dependence rests on financial

of U.S. dependence

Firms

a Good

Proxy? a good proxy for the

Much of our analysis demand for external

of U.S. firms being

funds in other countries. state equilibrium

We think this is reasonable

for four re=ons. funds, as Fig-

First, in a steady ure 1 shows. of technological

there will not be much need for external for external investment

Therefore, shocks

much of the demand that raise an industrys To the extent

funds is likely to arise as a result opportunities beyond what inter-

nal funds can support. represents Second, a good proxy.8

these shocks

are worldwide,

U.S. firms

need for funds

even if the new investment

opportunities

generated

by these worldwide

shocks differ

This amounts to saying that if the invention of persona] computers incre=ed the demand for external funds in the U.S. computer industry, it is likely to incre=e the need for funds in the computer iudustry in other countries as well.

10

across countries,

the amount

of cash flow produced

by existing

firms in a certain

industry

is

likely to be similar across countries.

In fact, most of the determinants the level of demand

of ratio of cash flow to its stage in

capital are likely to be similar worldwide: the life cycle, and its cash harvest period, of internally

for a certain product,

For this reason, we make sure that our results hold generated cash, rather than the difference between

even when we use the amount investments Third, and internally

generated

funds. life cycle that U.S. firms are in is likely is biased toward developing proxy for the

one might argue that the stage of a product from that of foreign firms.

to be different countries developing robustness

Given that our sample

one might countries

think that the U.S. industry position in a product

in the 1970s might For this remon,

be a better

life cycle.

we also explore

of our results to me=uring

the financial dependence

of U .S. firms in the 1970s rather

than in the 1980s. L=t but not least, that we only have a noisy measure of the need for funds creates development on growth. a bias

against finding any effect of financial

Data.
Data on financial development. contained in Table 2. Ideally, we would

2.1

We start by describing

our data on financial development

like a measure of a firms ability to raise money. available, and the efficiency

This depends

on the variety of intermediaries monitoring, certification,

with which they perform functions. Since

the evaluation,

communication appropriately proxies system. Turning

and distribution measured,

there is little agreement

on how these are

and even less data available,

we will have to make do with crude we think vital to a modern financial

even though

they may miss many of the aspects

first to equity

markets,

one memure

of stock

market

development offerings.

may be the Unfortunately, it is well

amount of money raised through this meuure documented is not available

initial public offerings and secondary buis

on a systematic

for many countries.

Moreover,

in the U.S. (see earlier) that mature firms are reluctant

to issue stock, even though is a

they have the potential

to do so. It could then be argued that stock market capitalization

11

better measure of the availability with the traditional literature

of finance.

Another virtue of this measure is that it is consistent (1969) for example) as a measure which h= typically used the

(see Goldsmith

ratio of the size of the financial

sector to output

of financial Product

sector development,

So the ratio of stock market capitalization stock market development

to Gross Domestic

in 1980 is the measure of for all countries data on developed

that we use. We obtain stock market capitalization contains

listed in the EmeTging Stock Mar-kets Factbook (which, interestingly, countries also) published by the International Finance Corporation.g

A comparable credit to Gross from

measure for the state of development Domestic Product, a measure Financial

of credit markets is the ratio of domestic and Levine (1993a). This

also used by King Statistics (IFS)

is obtained Monet~

the International

published

by the International 32f and excluding

Fund. Specifically,

this measure is the ratio of IFS lines 32a through the fraction of credit allocated

32e. Finally, we also compute

to the private sector, which is the credit.

ratio of claims on the non-financial There corporate are, of course, bond markets elements for more

private sector

(IFS line 32d) to total domestic For instance,

we do not capture. than a handful

we do not have data on for the United were typically omission States small in is private

of countries. bond

Except markets

which is not included

in our tests, and Japan, corporate 1995).

the early 1980s (see Rajan and Zingales, equity and private placement with me=ures are not. Panetta, of financial

A potentially

more damaging

of debt (not held by institutions). there are certain sized companies

While they may be correlated countries where they clearly held (see Pagano,

market development,

For instance, and Zingales,

in Italy, most medium

are privately small.

1995) while the stock market because,

is relatively

The lack of this data of of

may not be a serious problem

after all, we are interested of private equity outstanding held companies

in how the availability is generally a reflection

external funds affects firms; the amount the amount of internal finance privately

have secured in the past, and therefore

is not a measure of how developed

the financial markets are.

gStock market capitalization is me=ured at the end of the year, while Gross Domestic Product may =Iue flows through the year. This may be a problem in high inflation countries. We therefore measure GDP as the GDP in constant prices multiplied by the producer price index where the base year for both series is five. years before the year of interest.

12

2.2 -Data
The Gross

on

countries. Product, the Producer Price Index, the exchange Financial rate, and the Index Whenever a

Domestic

of Industrial pmticular

Production

are all obtained

from International

Statistics. the wholesale

series is not available, we use close substitutes price index is not available.

for instance,

price index years

if the producer of schooling

Data on a countrys

human capital in GDP

(average

in population

over 25) and the share of investment from the NBER

in 1980 are obtained

from the Barre-Lee

files downloaded

web site (see Barro and Lee (1994)).

2.3

Data

on

industries. and gross fixed capital Industrial formation for each industry in each country are

Data on value added obtained from

the General Division.

Statistics

(VOI 1) database

put together

by the United of

Nations Statistics

We checked the data for inconsistencies,

changes in classification SIC code.

sectors, and changes in units. obtain the amount with SIC codes. of external

The U.N. data is classified by International finance used by the industry

In order to ISIC codes while

in the U. S., we matched

Typically,

the three digit ISIC codes correspond to three digit SIC codes.

to two digit SIC codes,

the four digit ISIC code corresponds on country specific

In order to reduce the dependence firms

factors like natural resources Table 3 presents summary many countries

we confine our analysis to manufacturing statistics for all these variables. The binding We started constraint

(U.S. SIC 2000-3999).

We would like data on ~ ability of comparable

as possible.

is the availfrom

data on a countrys Factbook.

equity markets. We dropped

with 55 countries

the Emerging

Stock Markets

countries

like Kuwait

that did not report

a stock market capitalization Taiwan because volumes.

till the latter half of the 1980s, We could not use Hong Kong and are not present in the International Financial Statistics

data on these countries countries

We also dropped

for which we did not have data from the G ,1.S. database Switzerland). Finally, Thailand is dropped States is in

that is separated because excluded Table 2.

by at Ie=t

five years (notably,

the U.N. notes that data from year to year are not comparable. from the analysis because it is our benchmark.

The United

This leaves us with the 43 countries

13

2.4

.Data description.
measure of financial credit market development market - which we term capitalization financial development Product. of

Our b=ic

is the sum of domestic Table 2 lists this measure patterns we discernible

and stock

to Gross

Domestic

and its components

for countries

in our final sample.

A number

from an analysis of this table.

First, credit markets are more developed development to GDP

in richer countries. is 0.43 (significant

The correlation at the l% at the l% (though

of log per capita income in 1980 with financial the correlation with the ratio of credit with equity The

level) level),

market

is 0.45 (significant to GDP stock is 0.17

while the correlation

market

capitalization between

significant income

on] y at the 25V0 level). is consistent

weak correlation

markets

and per capita

with La Porta conditions,

et al,s view that colonial of the

influences

and legal tradition, of equity markets.

rather than economic As an example,

are a strong determinant

development law influence) equity between =-0.08,

Singapore

and Malaysia

(under British common

have strong equity markets while Phillipines Second,

(under Spanish civil law) hss weaker (if anything, weak substitution) (correlation Third,

markets.

there seems to be no relationship

the development not significant), of domestic

of credit markets and the development a fact previously credit noted by Demirguc-Kunt sector

of stock markets and Levine correlated

(1995).

the fraction capitalization correlation government enforcement credit going

going to the private significant

is strongly

with market of this that the and

to GDP (correlation=O.5,

at the l~o level).

Possible explanations

are that government intervenes

involvement

in industry

crowds out the private sector,

to fill the cap created

by the absence of the market, or that respect Finally, the fraction

of private property to the private

rights drives both variables. is correlated

of domestic O.27,

sector

with log per capita

income

(correlation=

significant

at the 10% level). dependent industries are likely to be better off in countries but

We want to see if financially with well developed also the ability appropriate industry,

financial sectors. operations

The availability

of finance affects not just investment working capi tal. Therefore, in value added between

to finance

and sales through being better oR

the most for that

me~ure

of an industry

is the growth

i.e., the change in the log of real value added in that industry by deflating

1980 and 1990. Price Index. For

Real value added in 1990 is obtained

value added by the Producer

14

high inflation countries, UN data are measured effective deflator

spurious differences at a different

in value added may be obtained So, instead,

simply because we determine

the the

point from the PPI index. in nominal value added production

by dividing

the growth

for the entire manufacturing

sector in the UN database rate in industrial

by the index of industrial obtained

(which measures the real growth

production)

from the IFS statistics. market development and the growth of the manufac-

The raw correlation turing sector across

between

financial

the 43 countries

is 0.27 (significant (1993a),

at the 10% level).

This is consistent component of

with the results in King and Levine market development

who show that the predetermined

affects future growth.

3
3.1

Financial
The basic

dependence
test

and growth

Our hypothesis relatively

is that industries

that are more

dependent

on external

financing markets.

will have In other between financial

higher growth rates in countries

that have more developed

financial

words, the explanatory

variable whose influence we are interested on external finance (as measured

in is the interaction

an indust rys dependence market development. Our study, therefore, ies of growth Levine simovic (e.g.,

in the U. S.) and local

has one important (1991), Mankiw, Kormendi Romer, is simply

advantage

over recent cross-country (1985), King

empirical

stud-

Barro (1992),

and Meguire

and Levine

(1993a), and Makcountry charac-

and Renelt (1996)). between

and Weil (1992),

and Demirguc-Kunt about

That

advantage

that we make predictions between a country

within

differences teristic.

industries

based on an interaction

and industry

Therefore,

we can correct for country

and industry characteristics to criticism

in ways that previous variable

studies were unable to correct for, and will be less subject bias or model specification. The dependent j in country variable

about an omitted

is the average annual real growth 1980-1990.10 The most effective

rate of value added way of correcting

in industry for country

k over the period

Iowe exclude the United Stat=

in all subsequent regressions. As noted earlier, for some countries data availHowever, for no country do we have data separated by less than 5 years,

ability limits the period even further.

15

and industry

characteristics

is to use indicator

variables,

one for each country

and industry.

The disadvantage

of such a fixed effects

regression

is that the determinants

of relative growth country

other than the interaction characteristics is then

of interest become

less transparent.

So we start by including variables.

and industry characteristics

instead of indicator

The model we estimate

GT~thj,k

= PI..* . Countrykcharacteri of country

sties + fl~+ 1... , IndustryjCharacteTistics+ and industry) + BP+lInteractim share in GDP + E (1)

P~+l..P. (Characteristics Among the country income

characteristics

we include

are investments

in 1980, log over age and stock

of per capita

in dollars of financial

in 1980, the average years of schooling development

in population credit

25, and our measure market capitalization The industry nancing

(which is the sum of domestic

to Gross Domestic

Product). initially is industry js dependence on external of both fi-

characteristic

we include we include

in the U.S. Finally,

two variables

that are characteristic

country

and industry;

industry js share in country

k of total value added in manufacturing on external financing

in 1980 and

the interact ion between market development characteristics,

industry js dependence

in the U.S. and financial and industry specific This will

in country

k. We will then drop the country for each country

specific

and replace

them with indicators

and each industry.

be the basic regression

in the rest of the paper.

Results

from the basic regression.

4, the coefficient estimate for the interaction term is positive standard and

As can be seen in Table statistically significant

at the l% level (throughout The interaction

the paper, the reported

errors are Using the

heteroskedasticity

corrected).

term is akin to a second

derivative.

Dropping thcountries does not qualitatively affect the results. We reduce the impact of outliers by constraining growth between -1 and +1, Three obserwtions are affected. The coefficient estimates for the interaction coefficient are higher and still significant when we do not do this, though the explanatory power of the regression is lower. We also rerun the same specification after winsorizing the l% and 5% tails of the growth rate distribution obtaining virtually identicrd results (except that the explanatory power of the regression is still higher). A potential concern is that we measure growth in value added rather than growth in output. Unfortunately, we do not have data for the latter. While we may not capture increases iu productivity fully, we see no obvious way in which this should bias our results.

16

estimates of financial dependence

in the first column, development,

in a country

which

is one standard rates between

deviation an industry

above whose

the mean financial is lin the impact of

the difference deviation

in growth above

is one standard

the mean and the average Note that this measures industries,

industry,

average country. financial between

Is the effect large or small? on financially

the differential

development financial

dependent

rather than the overall relationship

development

and growth.

Still, the effect is of the same order of magnitude share of GDP and per capita income

(see below)

ss the effect of variables such w investments cross-country growth regressions. variables are also of interest. of a one standard

that populate

The other explanatory indicate variable. (1.6%). negative

The following deviation

numbers

in parentheses

the effect on industry

growth

increase in the explanatory growth rate

The higher investments Developed coefficient countries

share of GDP

in 1980, the higher an industrys rates in manufacturing (-2.5Yo). The number (-0.370),

have lower growth

as evidenced

by the

on the log of per capita negative credit

income

of years of schooling while industries Industries in

has a small and insignificantly countries with more developed dependence significant

correlation

with growth

markets

have a higher growth

(0.570).

that

have a considerable borderline

on external

finance have slower growth

(-lYo) though

this only

statistically

at the 10% level, as do industries sector

that account

for a larger variables

share of a countrys have plausible effects.

manufacturing

in 1980 (-1,370). 11 Thus the explanatory

Next, we include the coefficient explanatory in growth

indicator

variables

for all the countries

in the sample, variables

As seen in Table 4, are stable. The

estimates

for the industry-specific

and interaction suggesting

power of the regression rates is because indicators

goes up substantially, inter-country

that much of the variation the regression and interaction indicators and

of omitted

differences.

We re-estimate

now including variables

for the industries. stable.

The coefficients we estimate

for country-specific the regression

are again surprisingly and industries. is positive

Finally,

including

for all countries the coefficient single country

This is the basic regression

we use in the rest of the paper,

(@= O.069) and significant is responsible

at the l% level. It does not appear that one The fundamental partial correlation

or industry

for the results.

11A ~o~~jble ~xplmation for this ~Wficient estimate is measurement error. Those industries that are incorrectly measured to be small in the beginning of the decade will be seen to have grown faster.

17

seems. robust.

4.1

Influence

of composition of the capital

of markets. market between equity and debt make a difference? sector Does it

Does the break-up matter sector? Some theoretical and growth argues,

how much credit

is channeled

to the government

and how much to the private

studies have argued for a correlation of this (King

between

the existence 1993 b).

of stock markets As Levine (1991)

and there is some evidence raised through equity

and Levine,

capital

issues is long term (for the firm), impacting

while shares can be By contrast, shocks to the To and

liquidated

by the investor

at short notice without

the firms projects.

if the firm finances with non-tradeable, institutions the extent may be transmitted

finite maturity

debt held by institutions,

on to the firm and force liquidation between

of long term projects. shocks of investors

that stock markets create a separation

the liquidity

the investment Growth should

needs of firms, the latter can take a longer term view and invest more efficiently. be higher if equity markets account for a larger proportion evaluated of credit markets.

An alternative

view is that managers

who are constantly

by the stock market tend it is better that they

to take the short term view (see Stein (1989) for a theory) obtain finance for long term projects to review all the arguments institutions predictions from financial

and therefore

institutions.

We do not have the space here merits of markets and

in the theoretical

literature

on the relative discussion).

(see Aoki and Patrick are ambiguous. merits

(1994) for an extensive

It suffices to say that the

Since the relative capitalization dependence to GDP.

are an empirical

matter,

we compute between

the ratio of equity

market

We then include Iinmce

the interaction regression

this ratio and the industrys

on external

in the b~ic

(see Table 5). We find that a dollar of it has only 40% of the effect on the that market capitalization The than

market capitalization growth of financially

is not the same as a dollar of credit; dependent firms. be * Unfortunately, e=ily interpreted

this finding

has less effect than credit cannot reason we must be cautious markets in channeling a dollar of market

as if it were the reverse finding. are better

about interpreting

this to mean that institutions firms is because,

funds to financially

dependent

unlike a dollar of credit, a dollar

capi talizat ion does not necessarilyy mean that some firm obtained

18

of finance. between

The lower effect of market

capitalization

may simply

reflect the translation

ratio

market capitalization

and eventual

finance through

new share issues. 12

While this test does not give unambiguous the form of finance matters. are equity dependent We therefore industry

results, it suggests a further way of testing whether correcting for the other effects, industries that

We test whether,

thrive better in a country

with a greater relative size of the stock market. finance used by U.S. firms in the s=e to size of capital markets. If the form

include the ratio of equity to total external with the ratio of market capitalization the coefficient different

interacted

of finance matters, and insignificantly Put together,

of this term should be positive.

It turns out to be negative

from zero. beneficial

the tests suggest that equity markets do not have disproportionately rate of financially dependent firms. Nevertheless, (1) without

effects on the relative growth are positively correlated

equity markets the interaction insignif-

with growth.

When we re-estimate to GDP included,

equation

term but with the market capitalization icant) relationship to the included between

we find a positive

(though

market capitalization measure of financial

and growth,

over and above its contribution In summary; our results

significant

market development.

suggest growth may be positively unrelated through pendent to the availability equity markets ones.

correlated

with the size of equity markets for reasons that are For example, the better risk sharing effected

of external

finance.

may reduce

the cost of capital

to all firms, not just the financial y dewith well developed stock

This would

imply higher growth differences

rates for countries

markets but not necessarily The effect of channeling in Table 5 by including

in growth rates between

its industries.

funds to the public sector rather than the private sector is captured between an industrys need for external funding and the

the interaction

lZWe ~~e ~ountrie~ only in the Emerging Markets Factbook because otherwise we do not know the size of the stock market. There is, however, a potential selectiou problem of look-back bias in these countries. The IFC may only include countries that, ex post (i.e., in the latter half of the 1980s when the database was initiated), have a reasonably sized stock market. This selection bias does not appear severe because the data set seems to contain every country (we know of) that had a stock market at the beginning of the 1980s. Moreover, some of the stock markets present in the datwet are miniscule in the late 1980s, suggesting that this may not have been an important criterion, Nevertheless, ,we can think about the consequences of a potential selection bias. The bim would be most important in the case of countries with small stock markets in 1980, that based on the hypothesized selection criterion, would have grown the fastest. This would induce a negative correlation between the size of the stock market and growth. While such a correlation is not borne out ill the data, and the selection bias does not directly predict any correlation for the interaction term, this suggests further caution about comparing the coefficients of credit and equity market capitalization.

19

ratio .of domestic negative (coefficient

credit

to the public

sector

to GDP.

The

coefficient

is large and strongly of a dollar

estimate=

-O.16, t=-3.3)

). This implies that the overall impact

of extra credit, if directed dependent industries

at the public sector, is to reduce the relative growth rate of financially

by about half u much as a dollar of credit to the private sector incre~es in a number dependent of ways. It may indicate that public sector

it. This evidence credit

can be interpreted at financially

is not directed

segments,

and perhaps

is misallocated

to those

segments

that do not need it.

Alternatively, crowding

it may be that involvement

by the deep pocket (and thus reduces

government growth)

has a disproportionate

out effect on the private sector

in industries

that are financially

dependent.

4.2

Decomposition

of the

sources

of growth. and because existing firms

An industry grow in size. also reports different

can grow because Does

new firms are added to the industry

the interaction

term affect only one of these or both? This statistic may often

The U.N. database be compiled by a

the number

of firms in an industry.

body

in a country

from the one that produces check on our results. compiling compiling the number value-added

the value-added The disadvantage

data.

The advantage

is that it provides clusification classification

an independent

is that the industry the industry

used by the body used by the body

of firms may differ from data, resulting

in an increase in noise. firms less the is obtained by

The growth in the number log of firms in the beginning dividing

of firms is the difference of period.

in the log of ending period

The average size of firms in the industry

the value added in the industry by the number of firms, and the growth in average size again as a difference in logs. We then estimate the basic regression The coefficient with growth in

is obtained

number of firms and growth in average size as dependent in table) of the InteractIon while the coefficient t..gsize= 1.73). dependent

variables.

(not reported

term 1s almost equal in both cases (~~~mb~~ = .022, @aVgsize = .023), regression is estimated more precisely (t~v~b.~ = 2.06,

in the numbers

The coefficient

estimates

suggest

that the incremental

growth

of financially

firms in more developed

markets is almost equally driven by a growth in the number

of firms and by a growth If the development financially dependent

in average size. markets affects both the growth in the number of firms in

of financial industries

and their average size, it must be that the average size of firms

20

in financially

dependent

industries

is larger in economies regression

with developed

markets.

This is indeed

the case. When industry

we estimate

the b~ic

with the log of the average size of firms in an variable (estimates not reported), so when the

in a country

at the end of the 1980s M the dependent term is positive

the coefficient dependent

of the interaction

and highly significant. of the 1980s.

It remains

variable

is the average size at the beginning of financial markets.

So the scale of firms is.

related to the development

5
There

Robustness
are obviously many differences interaction between the same industry an industrys in different financial countries. That and the

we find such a strong countrys financial

effect between is surprising.

dependence

development

We, therefore,

perform

a number

of robustness is sensitive

checks to rule out more mechanical to redefining dependence in definition dependence. financial dependence

explanations.

First, we check if the correlation

M the dependence

of young firms or old firms, or measuring is not robust to these changes of our memure of

in the 1970s rather than the 1980s. If the correlation of dependence, it makes us less confidant

in the universality

5.1

Is the dence?

correlation

robust

to

changes

in the

measure

of

financial

depen-

We calculate

financial

dependence

restricting

the sample only to mature estimated When

firms (listed for more term, 0.05, with

than 10 years) is statistically dependence

in the United significant

States.

The coefficient 6, column (I)).

for the interaction

(see Table

we estimate

the regression

defined = the dependence on the interaction

of young firms in the United States on external significant,

finance, though

the coefficient

term is similar (0.04) and still statistically (II)). in the pattern. from

only at the 10% level (see Table 6, column Next, we check whether States

there is persistence

If the pattern

of financing

in be

the United unreasonable countries

in the 1980s is very different

the pattern

in the 1970s, it would (especially

to expect

it to carry any information

for other countries

developing concerns

that may use older technologies).

Furthermore,

such a check would alleviate

21

that aberrations

in a few industries between

in the 1980s drive the results. demand for external financing in the 1980s and in the 1970s

The raw correlation its demand

an industrys

in the 1970s is 0.63. When we include the demand

for external financing is positive

in the basic regression significant

instead of demand in the 1980s, the coefficient (III)). The coefficient less precisely, is about

and statistically

(see Table 6, column

40% larger than that estimated significant at the 5%

for the 1980s, Though level. 13

it is measured

it is still statistically

5.2

Is the ment ?

correlation

robust

to

changes

in the

measure

of financial

develop-

As a measure

of the development credit and equity

(and thus efficiency) market capitalization

of capital

markets

we have used the development. fairly well h= while

sum of domestic Although our priors.

to measure

financial

crude, this is the most common There are, however,

measure used in prior studies and captures (see Table 2), For instance, equity

some aberrations

Malaysia markets,

a very high level of financial

development,

driven Iargel y by its strong

Belgium has a lower level of development, (1996) attribute and colonial this to differences

because its equity market is minuscule.

La Porta et al.

in legal systems that emerged legal system

as a result of foreign invwions by British common imposed law on

domination.

The Malaysian

was influenced

(which promotes Belgium strength

equity markets), Rather

while the Rench

civil code (which does not) w= the aberrations

by Napoleon.

than being a source of concern,

are a source of apart

for the study because development.14

they enable us to tell the effect of financial

development

from economic

The proxy for financial efficiency of capital markets.

development

we would really like to have is a direct me~ure

of the

While the combined

size of the credit and equity market is likely to is robust and

be highly correlated to changes

with this, it is incumbent An alternative

on us to see that the effect we me~ure way to me~ure the degree

in this measure.

of sophistication

lQThe statistical significance of the coefficients for all the models in this subsection tends to increase substantially when the dependent variable is trimmed at the l~o and 5% levels. theresults are lAwe do check however, that our results are not driven by strange outliers. In particular,
unchanged if we drop Jordan, which scores surprisingly high on our me~sure of financial development, from the regression,

22

efficiency Financial reports

of a capital market is to look at its accounting Analysis of at le=t and Research creates

standards.

The Center for International by rating the annual Thus The

an index for different on the inclusion

countries

three firms in every country

or omission

of 90 items.

each country

obtains

a score out of 90 with a higher number

indicating

more disclosure.

greater the public information, finance.

the lower should be the cost of external finance relative to internal 15 While more developed there are exceptions. or Canada countries

So this should be a proxy for financial development. accounting standards (see Laporta, States, et al.), Australia

have better Malaysia Belgium between

For instance, while

scores

as high as the United

on this measure,

and Germany

are in the same league = Korea, Phillipines, measure of financial development

or Mexico.

The correlation standards When is 0.39

our capitalization

and accounting

(significant

at the 5% level for the 32 countries between financial dependence (see Table

for which we have this data). standards

we include

the interaction measure

and accounting 6, column

instead of our previous is positive and

of financial

development (p=

(IV)),

the coefficient

highly statistically With

significant

0.16, t=4.35). of financial dependence and financial development,

more confidence

in our me~ures

we now turn to other concerns,

5.3

Focussing

on

substantial

industries in other sources of comparative advantage may dictate the

It may be argued presence, absence,

that differences or growth

of industries

in a country.

Even if this is the case, our results advantage is a

cannot be explained is strongly correlated

unless the dependence with their dependence

of industries on this source of comparative on external funding and financial

development

good proxy for the source of comparative

advantage.

As an example,

the. availability

of natural

resources may drive the growth of the mining sector, and if resource colonial intensive countries developed

If mining is dependent markets

on external finance, by common would drive

financial

(either by chance, of natural resources

heritage,

or by design)

then the presence

or absence

the correlation.
15Unfortunately, we do not have this measure for all the countries ill our sample. Also the Center for International Financird Analysis and Research which produces this data was set up only in 1984. So we do not have a me=ure of the accounting standards at the beginning of the period. This may not be a big problem because the stadards do not change dramatically in a short period.

23

In the ideal thought almost every country

experiment,

we should focus on industries

like cement that are present in or natural resources

and do not require special technology,

human capital,

(but do require finance).

We do not have data in such detail, nor are there enough such industries vaiation to enable us to estimate to minimize our coefficients (even if the data

to give us the cross-sectional were available). source of concern. By restricting availability country, the sample Nevertheless,

we can attempt

the problem,

as well as see if it is a

to manufacturing

firms, we have already reduced has a substantial resources presence

the influence

of

of natural

resources. that country

Also, if an industry h= the necessary

in a particular So in

it is logical

and talents for the industry.

by further restricting

the sample to industries of differences

that are above the median size in the country from differences (V)),

1980, we reduce the problem When we estimate

in growth stemming

in endowment. the interaction

the regression

with this smaller sample (Table 6, column significant dependence (~ = 0.06, t=2.65). or financial

variable is still positive

and statistically

We can also test if either financial thing else. Industries

development

proxy

for somedrugs and

that are highly dependent

on external

finance for example,

pharmaceuticals

may also be highly dependent

on human capital inputs. the observed

To the extent that fiinteraction between between human

nancial market development financial capital dependence dependence

and human capital are correlated, development

and financial

may proxy for the interaction

and the availability an interaction

of trained human capital. the industrys

To check this, we include in on external finance and

the basic regression the measure over age 25).

between

dependence

of the countrys

stock of human capital estimates in Table

(average 7, column

years of schooling I show,

in population on the

As the coefficient

the coefficient

human capital interaction interaction term is virtually

term is small and not statistically unchanged.

significant,

while our fundamental is not a proxy

This suggests that financial

dependence

for the industrys Another simply process

dependence

on human capital. on external financing in the United States

possibility

is that a lower dependence maturity

reflects

the greater

of the industry. industries

An influential

view of the development migrate from devel-

is that ~ technologies to developing

mature, economies

using those technologies Dornbusch,

oped economies (1977)).

(see, for example,

Fischer,

and Samuelson markets,

Since developing

countries

are more likely to have underdeveloped

financial

24

the interaction

effect we document countries.

may simply reflect the stronger growth of mature technologies

in underdeveloped We already

have results suggesting

this cunot is me=ured

be the entire explanation.

The interaction States.

effect is present even when dependence Financial dependence is unlikely

only for young firms in the United maturity for these firms.

to proxy for technological

Furthermore, in the regression. dependence

we can test if financial development We include finance in the b~ic

is really a proxy for economic the interaction GDP between

development the industrys in addition to

regression

on external

and the log per capita

for the country,

our usual interaction

term,

As seen in Table 7, column

(II), the coefficient

of the interaction significant. The

term falls to 0.05 (from interaction statistically between

0.07 in the basic regression) dependence

but is still statistically income is positive

financial

and log per capita

and borderline income captures

significant

at the 10~0 level.

This may be because

log per capita

some aspects of financial development rule out the possibility than log per capita support for financial that financial

that are not captured development

by our proxy.

Of course, we cannot development

is a better proxy the above

for economic

income.

Nevertheless,

taken together,

tests do not offer strong

dependence

being a proxy for technological

maturity.

5.4

Other

explanations:

Reverse

Causality. markets as predetermined and exogenous. An al-

Thus far, we have taken the state of financial ternative explanation of the development

of financial markets is that they arise to accommodate

the financing

needs of finance-hungry is as follows.

industries. there are some underlying Then, country countries specific factors that in these

The argument favor certain factors should

Suppose

industries experience

that happen

to be finance hungry. rates in financially market.

abundant

higher growth

dependent

industries

and - as a rates in effect.

result should develop financially dependent

a strong financial sectors

If these factors persist,

then growth interaction

will persist and we will observe factors than any beneficial over periods

the significant

But here it will result from omitted The l=k of persistence

effect of finance. (see Easterly, growth across Kremer, decades

in country (1993))

growth

of decades of sectoral concern.

Pritchet t, and Summers (Klenow, 1995) suggest

and the low correlation not be a major

that this should

Nevertheless,

we can in-

25

vesti~te

the reverse causality a messure

argument

more direct 1 y. dependence finmcing of a countrys for m industry industrial struc-

We construct

of the average financial on external

ture in 1980; we weight contribution

the dependence

by its relative in the coun-

to value added in the manufacturing of the proportion

sector, ad of financially

sum across industries dependent

try. This gives US a meuure When we sort countries are low on this me~ure

firms in that country.

on this measure, while developed

less developed economies

economies

like Sri Lanka and Indonesia Japan and the U.K. are are high on the list

like Netherlands,

high on this list. But there are exceptions. while Portugal, A regression the weighted results Greece, and Belgium

Singapore,

Israel and Malaysia

are below the median. development in 1980 M the dependent in 1980 u variable and

with a countrys

financial

average financial positive

dependence

of its industries

the explanatory

variable the

in a strong

coefficient

(~ = 2.89, t=2.51, at conventional

R square= O.19).

Interestingly,

correlation for young determine

is not statistically firms. A possible

significant explanation sector.

levels when dependence (and larger) correlation dependence.

is measured

is that it is the older A similarly strong

firms that largely when we

the size of the financial capitalization

is observed on

regress market

against weighted

average equity

the one hand, in the

this lends credence

to our method countries

of extrapolating

the financial dependence financial

of industries

Unit ed States to other develop

as might seem natural, dependent

markets

and institutions

to meet the needs of financially

industries.

On the other hand, this finding

suggests reverse causality If an omitted

may indeed be an issue, both the growth of industries and the development of financial

factor explains

markets, its influence should be especial] y felt for large industries in economies weighted average financial dependence markets is high. These industries would

where the ex ante responsible of the if an

be largely

for the state of financial omitted factor.

and their size is presumably (multiplied

driven by an abundance interaction term)

We include

an indicator

by the standard

industry is above median size and is in an economy dependence. having become If a persistent omitted factor

with above median weighted for financially

average financial industries

is responsible

dependent markets,

large, and also consequently

for the size of the financial term).

the coefficient

of this term should be positive is negative, though

(and absorb a lot of the interaction (see Table 7, column

In fact, the coefficient of the standard

not significant

(111)). The coefficient

26

inter~tion Again,

term is unchanged. if a omitted factor drives our results, the introduction of an interaction between estimate a for a

proxy for the omitted our fundamental

factor and financial dependence Although we cannot

should weaken the coefficient

interaction.

measure this factor directly,

we can obtain

proxy for it. Under the alternative that is responsible development financially

explanation

of reverse causality, dependent

it is this unobserved industries

factor

for the large past growth of financially sector.

and, thus, for the of

of the financial dependent

As a result, the beginning-of-period

relative importance

sectors in each country specification

can be used as a proxy for such an omitted after inserting = a regressor and an industrys the product

factor. between (see

We re-estimate, a countrys

then, the b~ic

weighted (V)).

average external The estimated is virtually

dependence coefficient unchanged

financial

dependence

Table 7, column and financial The

of the interaction (even slightly

between increued),

financial

dependence

development above

two findings

that the state of development weight of industries dependent

of an economys on external growth

financial

sector

is explained fundamental industries

by the relative interaction

finance

and that the dependent sector markets are and

does not simply reflect the continued

of financially

that, in the past, were responsible Taken together,

for the development the possibility

of the financial that financial but then facilitate

not in contradiction. institutions of another

they suggest

may develop to meet the needs of one set of industries, smaller, younger, group of industries. the effect of financial

the growth

To investigate

this, we estimate

development

only for industries

that of the

are small to start out with, and are unlikely to be responsible financial markets. the median So we estimate the basic regression countries. (IV))

for the state of development

for industries

that in 1980 were less than term is positive the economys

size in their respective

The coefficient

of the interaction for whom

and significant financial

(see Table 7, column

even for these industries,

development

is largely predetermined. argument we re-estimate are variables the basic specification by

Finally,

to address the reverse causality variables. The proper

using instrumental

instruments

likely to affect the develop-

ment of financial markets, variables. While

but not to be affected by it, For this reason we resort to institutional may foster the development direction. of capital markets, it is unthat

some legal traditions

likely that the relationship

runs in the opposite

La Porta et al. (1996)

suggest

27

the origin of a countrys mmket. Similarly,

legal system

has an effect on the development of the legal system to be true.

of a domestic

capital for

the efficiency mmket,

and the integrity

are a precondition Therefore,

a sophisticated country eficiency

capital

but the reverse is unlikely

we use the

of origin of the legal system u classified by La Porta et al, (1996) of the judicial system produced by Business International As Table

and an index of the (a country-risk (VI), shows the All this

Co~oration 7, column

rating agency) fundamental

as instruments

for financial

development.

correlation

persists even when we estimate

it using instrumental

variables.

suggests that it is unlikely that reverse causality

is the source of our results.

5.5

Other

Explanations: opportunities

Investment

and

Cost

of Capital. For instance, 1) partly the tobacco because in-

Investment industry vestment generates,

in different industries may be very different. States uses negative in the Tobacco external finance

in the United opportunities

(see Table

industry

are small relative to the cashflows on external finance proxies

the industry primarily for

It may be that our measure of dependence intensity of a particular of savings industry.

the investment may proxy country.

Furthermore,

the size of the financial

sector in that

for the extent

in the country,

and hence for the cost of capital intensive

The interaction

may simply represent

the fact that capital While this is certainly

firms grow faster which

in an environment

with a lower cost of capital. could affect growth,

a channel through

financial development If investment

it is not the channel we focus on. and external finance and internal finance were would be irrelevant in countries that

intensity were all that mattered, generated

equally costly, the cash internally are more financially developed.

by industries

All that mattered

would be the size of the required

investment

and the cost of capital. finance which decreases by industries

By contrast,

if there is a wedge between

the cost of internal and external markets, the cash internally development. developStates Inwith

in the state of development would have its own interaction

of financial

generated

effect with financial dependence generates

A first pass at this is to replace the interaction ment by the interaction (normalized dustries that by property generate between

between

and financial

the internal cash an industry

in the United development. in economies

plant and equipment) more internal cash grow

and the countrys disproportionately

financial faster

underdeveloped

financial markets

(~ = .23, t=-2.4,

see Table 8, column

I). A second pass is to

28

also include the interaction to property,

between

an industrys

investment financial

intensity

(the ratio of investment As Table 8 column half the size of the Since the between the

plant and equipment)

and the countrys interaction

development.

II shows, the coefficient coefficient

on the investment

is 0.09, approximately

on internal cash (-O. 18), though only the latter is statistically are not statistically csshflows different, this supports

significant.

two coefficients investments paper. ad

our use of the difference financing

as our measure of dependence

on external

throughout

A final confirmation development

is that when we include the interaction in our basic regression development (~ = O.10). suggests (not reported),

between investment only the interaction

intensity between

and financial financial estimated

dependence

and financial

is significant,

and its coefficient

is close to that

in the basic regression

Finallyl the discussion dependent variable,

of investments

that instead of using the growth rate as the intensity of an industry in other countries u

we could use the investment We obtain investment

the dependent in the industry

variable.

intensity

by dividing

the gross capital formation variables are

by value added in that industry The coefficient

This

(using U.N. data). of the interaction

suggests that

The explanatory

taken from the basic regression. at the 5% level (estimates

in countries heavily with poorly

term is 0.10 and is significant

cost of external finance

not reported). developed markets

the higher

results

in lower investment

intensities

in industries

dependent

on external

finance. of alternative explanations tests of our results and do not find compelling to the argument financial that

In sum, we test a number evidence financially because in their favor.

Instead, industries

the additional grow faster

seem to add credence with better

dependent the cost

in countries

developed

markets

of external

finance

is lower in such countries.

Conclusion
in this paper to investigate growth. whether financial sector development circumvent some of the problems First, it is difficult

We develop a new methodology has an influence on industrial with the existing to interpret the causality cross-country

In doing so, we partially highlighted

methodology

by Mankiw

(1995).

observed debate

correlations

in cross-country

regressions

in a causal sense. Here, we push through which finance

one step further by finding evidence

for a channel

29

theoretically examine problem

influences

growth.

Also, since we have multiple of causality

observations

per country,

we can

situations

where

the direction methodology

is least likely to be reversed. variables

A second and

with the traditional

is that explanatory of these two problems

are multi-collinear

are me~ured significant observed industry

with error. The combination

may cause a variable to appear with error. effects As a result, and

when it is merely a proxy for some other variable correlations indicators) can be misleading. By looking

me~ured

at interaction

(with country

rather than direct effects, we reduce the number alternative explanations. Third,

of variables

that we rely of

on, as well M the range of possible limited degrees of freedom

there is the problem

- there are fewer than two hundred Our approach partially alleviates

countries

on which the myriad by exploiting within-

theories have to be tested. country vwiation

this problem

in the data.

Our methodology,

may have wider applications,

such as testing

the existence

of channels

through

which human capital can affect growth. contribution, they suggest this papers findings may bear on four different development h= a substantial

Apart from its methodological are= of current research. First,

that financial

supportive

influence on the rate of economic finance to financially

growth and this works, at least partly, by reducing dependent of economic firms. We should (Easterly, add that there is no et al, (1993)) is set serially enable

the cost of external cent radiction

when the lack of persistence of financial in a coutrys

growth

against the persistence uncorrelated) changes

development. investment and thereby

Other factors opportunity enhance

may cause (potentially set. Finance may simply

the pursuit of these opportunities, Second, evidence Third, in the context

long run growth. constraints, this paper provides fresh

of the literature

on financial

that financial

market imperfections

have an impact

on investment

and growth.

our results may be relevant to the current debate should have, of capital

about what form of social security system fosters on a countrys

system a country the development growth,

It has been argued that a private social security If this development as an indirect has beneficial effects

markets.

then this might be counted in the context of industry

benefit of a private system. the findings suggest a potential explanation market regulaof

Finally,

of the trade literature, specialization

for the pattern development

across countries.

To the extent that financial accident or government

(or the lack thereof)

is determined market

by historical in a certain

tion, the existence

of a well developed

country

may represent

a source

30

comparative nance.

advantage

for that country

in industries

that are more dependent development etc.

on external

fi-

Similarly,

the costs imposed

by a lack of financial

may differentially Therefore,

tiect

incumbent financial w well u

firms and new entrants, development

large firms versus small fires, in determining

the level of

may also be a factor

the size composition for future research.

of an industry

its concentration.

These issues are important

~eas

31

References

AOKI,

M.

AND H. PATRICK ED. 1994,

The Japanese

Main

Bank System,

Oxford

University

Press, New York. BARRO, R. 1991, Economic CVI: 407-443,

HARVEY 403-44. 1993, International Comparisons of Educational Attainment , NBER 1995,

Growth

in a Cross-Section

of Countries,

Quarter] y Journal

Of

Economics,
BEKAERT, G.

AND C.R. 1995, pp.

Time-varying

World Market Integration,

Journal of

Finance,
BARRO R.

AND J. LEE, Paper. 1961,

Working CAMERON, R.

France and the Economic Princeton, N.J.).

Development

of Europe

(1800-1914)

(Princeton

University CAM ERON, R. New York.

CHANDLER, A.

Press:

1967, Banking

in the Early Stages of Industrialization,

Oxford

University

Press:

1977,

The

Visible Hand: Cambridge

The Managerial MA.

Stock

Revolution

in American

Business

Harvard

University

Press:

DEMIRGUC-KUNT, Stylized Facts,

A. AND R. LEVINE, World A. Bank, Policy

1995,

Markets

and

Financial

Intermediaries:

Research 1996,

Working Financial World and

Paper. Constraints, Working Uses of Funds Paper. and

DEMIRGUC-KUNT1 Firm DIAMOND, Growth: D.,

AND V.

MAKSIMOVIC,

An International Financial

Comparison,

Bank

1984,

Intermediation

Delegated

Monitoring,

Review

of Eco-

nomic
DIAMOND, Placed DORNEUSCH, ments

Studies 393-414.
D., 1991 Monitoring and Reputation: the Choice Between Bank Loan and Directly

Debt, R.,

Joum.al of Political
S. FISCHER, P.

Economy

99, 689-721,
1977 Comparative of Goods, Advantage, Trade and Pay-

SAMUELSON, with

in A Ricardian

Model

a Continuum

Amefican

Economic

Review

823-839.
EASTERLY, Luck? W., M. KREMER, Growth L. PRITCHETT, AND L. SUMMERS, Shocks, 1993 Good Policy or Good

Country

Performance

and Temporary

Journal

of Monetary

Eco-

nomics

32, 459-484. different about banks,

PETERSEN

FAMA, E. 1985, Whats

FAZZARI, rate FELDSTEIN, S., R. G.

Journal of Monetary
1988, Financing

Economics
Constraints

15, 29-39.
and Corpo-

HUBBARD,

AND B.

Investment M.

, Brookings

Papers

on Economic

Activity

1, 141-195. Capital Flows,

AND HO RIOKA, C.

1980, Domestic

Saving and International

The Economic

Journal 90: 314-329.

32

FRENCH, K. AND J. POTERBA ket, American Economic

1991, Investor

Diversification

and International

Equity

Mar-

Review 81, 222-6. Backwardness in Historical Perspective Harvard Univer-

GERSCHENKRON, A.

1962 Economic MA.

sity Press, Cambridge, GOLDSMITH, R.

1969 Financial

Strwcture and Development Market

Yale University

Press, New Haven. Monitor-

HOLMSTROM BENGT, AND J. TIROLE , 1993, ing, JAMES, C. Journal of Political 1987, Some Economy,

Liquidity

and Performance

Vol. 101, No. 4, 678-709. of Bank Loans, Journal of Financial

Evidence

on the Uniqueness

Economics

7, 217-235. Do Investment-Casflow of Chicago, Sensitivities working Provide Useful

KAPLAN, S. AND L. Z] NGALES , 1996, Measures of Financing Constraints?

University

paper. Press, New

KIND LEBERGER, C. York.

KING, R. AND R.

1993 A Financial

Histo~

of Western

Europe Oxford

University

LEVINE

, 1993a,

Finance

and

Growth:

Schumpeter

Might

Be Right,

The

Quarterly

Journal

of Economics,
, 1993b,

Vol. CVIII,

No. 3, 681-737. and Growth, Journal of A40n-

KING, R. AND R. LEVINE etary Economics, KLENOW, P.

Finance,

Entrepreneurship,

513-542. Industry Evidence on Growth Models,

, 1995, Ideas vs. Rival Human Capital: of Chicago working paper.

University

KORMENDI, R. AND P. MEGUIRE , 1985, Macroeconomic of Monetary

LA PORTA, R., F.

Determinants

of Growth,

Journal

Economics,
LOPEZ

Vol. XVI,

141-163.
A. SHLEIFER,

DE SILANES,

AND R. VISHNY (1996)

Law

and Fi-

nance, LEVINE, R.

NBER , 1991,

Working Stock

Paper 5661. Markets, Growth, and Tax Policy, Journal of Finance, XLVI

1445-1465. LEVINE, R. AND D. RENELT , 1993, gressions, MANKIW,

MANKIW,

A Sensitivity Review

Analysis

of Cross-Countries 942-963. on Economic

to the Empirics

Growth

Re-

American

Economic

, Vol. LXXXII, Brookings

G. G.,

, 1995, The Growth D. ROMER,

The AND D.

of Nations,
WEIL , 1992,

Papers

Actiuity,

Vol 1.

A Contribution

of Economic

Growth,

Quarterly

Journal of Economics, Systems, Corporate IL.

Vol. CVII,

407-437. Development, and Investment, in R. The

MAYER, C. Glenn

1990, Financial Hubbard, cd.: of Chicago

Finance,

and Economic Finance

Asymmetric

InjormationJ

Corporate

University

Press, Chicago,

33

MCmNNON,

R.

1973, Mone~ and Capital in Economic D. C.. R. RAJAN 1996, ~ade Credit:

Development

The Brookings

Institution,

Washington,

PETERSEN, M, AND University

Some

theories

and evidence,

mimeo,

of Chicago. of Corporate Borrowing, Journal of Financial Economics, 5:

MYERS, S. 1977, Determinants 146-175.

MYERS, S. C. AND N. S. MAJ LUF that

1984,

Corporate

Financing

and Investment

Decisions

when

Firms

have Information

Investors

Do Not Have

Journal of Financial

Economics,

13,

187-222.
NORTH, D. 1981, StmctUm

and Change in Economic

History Why

New York, Norton. Do Companies Go Public? An

PAGANO, M., Empirical PAX SON, C.

PANETTA, F. AND L. ZINGALES 1995, Analysis, NBER Working

Paper #5367. from Micro Data, Princeton Working Paper.

1996, Saving and Growth:

Evidence

RAJAN, R. AND L. ZINGALES 1995, What idence From International

SC HUMPETER, J. A. 1911,

Do We Know About of Finance,

Capital

Structure:

Some Ev-

Data

Joumai

December. Harvard University Press: Cam-

A Theo~

of Economic

Development

bridge,
STEIN, J.

MA.
Efficient Capital Markets, Inefficient Firms: A Model of Myopic Behavior,

1989,

Quarterly

Journal

of Economics,

655-669.

34

Table 1:

Pattern of external financing and investment across industries in the U.S. during the 1980s
The table reports the median level of external financing, equity financing, and investments for ISIC industries during the 1980s. operations. External finance is the fraction of capital expenditures not financed with cash flow from funds from operations (item Equity finance is the ratio expenditures to Caah flow from operations is broadly defined as the sum of Compustat and increaaes in payables. Investment of equity issues to capital expenditures.

# 110), decreases in inventories, decreases in receivables, of the net amount net property correspondingly is obtained plan and equipment.

is the ratio of capital

Mature firms are defined as firms that have been public for at lemt 10 years, All the information from CRSP and

young firms are defined as firms that went public less than 10 years ago. The IPO year is defined except for the SIC code which is obtained

as the first year in which a company starts to be traded on the NYSE, AMEX, or NASDAQ. from the flow of funds data in Compustat, with the ISIC code. then matched

ISIC code 311 313 314 321 3211 322 323 324 331 332 341 3411 342 3511 3513 352 3522 353 354 355 356 361 362 369

Industrial

sectors

Food products Beverages Tobacco Textile Spinning Apparel Leather Footwear Wood Products Furniture Paper and products Pulp, paper Printing & publishing Basic exclud fert Synthetic resins Other chemicals Drugs Pet. refineries Pet.& coal products Rubber products Plastic products Pottery Glass Non metal products

All Companies External Equity Finance Finance 0.14 0.00 0.08 0.00 -0.08 -0.45 0.40 0.01 -0.09 -0.09 0.03 0.00 -0.14 0.00 -0.08 0.04 0.28 0.04 0.24 0.01 0.18 0.02 0.15 0.02 0.20 0.03 0.25 0.12 0.16 0.03 0.22 0.02 1.49 0.76 0.04 0.00 0.33 0.06 0.23 0.11 1.14 0.26 -0.15 0.11 0.53 0.02 0.06 0.01

Capex 0.26 0.26 0.23 0.25 0.16 0.31 0.21 0.25 0.26 0.25 0.24 0.20 0.39 0.30 0.30 0,31 0.44 0.22 0.23 0.28 0.44 0.20 0.28 0.21

Mature Companies External Equity Capex Finance Finanu Finance 0.25 -0.05 0.00 -0.15 0.00 0.28 -0.38 -0.21 0.24 0.14 0.24 0.00 -0.02 -1.33 -0.57 0.25 0.33 0.10 0.13 0.14 0.08 -0.23 -0.18 0.03 -0.02 0.16 -0.12 -0.01 0.00 0.00 0.01 0.01 0.02 0.01 0.00 0.04 -0.09 0.00 0.00 -0.00 0.01 -0.01 0.27 0.27 0.23 0.23 0.17 0.23 0.21 0.33 0.24 0.20 0.25 0.32 0.22 0.26 0.21

Young Companies External Equity Capex Finance 0.28 mU.da 0.66 0.00 0.63 0.26 0.66 0.27 0.65 0.34 0.68 0.57 0.22 0.60 0.79 0.79 1.35 2.06 0.85 -0.26 0.50 1.14 1.52 -0.03 0.36 0.15 0.20 0.01 0.26 0.11 0.01 0.20 0.36 0.25 0.33 1.15 0.27 0.30 0.32 0.96 0.48 0.00 0.26 0.37 0.26 0.40 0.29 0.29 0.20 0.41 0.29 0.45 0.46 0.47 0.28 0.22 0.32 0.48 0.33 0.26

0.03 0,15

-0:02 0.03

0.28 0.22

35

ISIC code 371 372 381 382 3825 383 3832 384 3841 3843 385 390

Industrial

sectors

Iron and steel Non-femous metal Metal products M&inery Office, computing Elect. machinery Wdio Transp. Eq. Ship Motor veichle Professional goods Other ind.

All Companies External Equity Finance Finance 0.09 0.01 0.01 0.02 0.24 0.02 0.45 0.11 1.06 0.67 0.77 0.36 1.04 0.46 0.31 0.05 0.46 0.02 0.39 0.01 0.96 0.62 0.47 0.16

Capex 0.18 0.22 0.29 0.29 0.60 0.38 0.42 0.31 0.43 0.32 0.45 0.37

Mature Companies Equity External Capex Finance Finance Finance 0.02 0.16 0.09 0.21 0.07 0.00 0.00 0.25 0.04 0.25 0.22 0.04 0.38 0.26 0.05 0.23 0.02 0.29 0.39 0.02 0.30 0.16 0.01 0.28 0.04 -0.04 0.34 0.33 0.11 0.00 0.19 0.03 0.33 0.28 -0.05 0.01

Young Companies External Equity Capex Finance 0.26 0.12 0.19 0.61 0.46 0.24 0.87 0.25 0.34 0.75 0.41 0.33 1.16 0.78 0.64 1.22 0.78 0.46 1.35 0.74 0.48 0.58 0.16 0.31 1.05 0.52 0.56 0.76 0.02 0.32 1.63 0.94 0.52 0.80 0.37 0.49

36

Table 2:

Financial
reported by the IFC divided

development
Product

across countries.
traded equity at the end of 1980 as credit is the sum deposits (IFS line that year. The level of domestic but excluding interbank

The equity market capitalization

over GDP is the total value of publicly

by the Gross Domestic authority

of assets held by the monetary

and depository

institutions

92a-92f but not 32 e) divided by GDP. The ratio of priwte credit represented obtained by claims against by dividing

credit to domestic

credit is the proportion

of domestic

the private sector (IFS line 92d).

The logarithm

of per capita income is

the 1980 GDP in dollars by a countrys population

as reported by IFS. The exchange rates

are also from the IFS.

Country Australia Austria Bangladesh Belgium Brazil Canada Chile Colombia Costa Rica Denmuk E~pt Finland France Germay Greece India Indonesia Israel Italy Jamaica Japm Jordan Kenya Korea Malaysia Mexico Morocco

Equity market cap. over GDP 0,38 0.03 0.00 0.09 0.05 0.46 0.34 0.05 0.04 0.09 0,01 0.06 0.10 0.09 0.08 0.05 0.00 0.35 0.07 0.02 0.30 0.50 0.00 0.08 0.65 0.07 0.02

Domestic credit over GDP 0.44 0,97 0.20 0.56 0.28 0.52 0.40 0.16 0.49 0.47 0.73 0.46 0.60 0,99 0.66 0.45 0.13 0.83 0.90 0.46 1.00 0.67 0.28 0.55 0.53 0.32 0.38

Private credit over domestic credit 0.63 0,79 0,35 0.51 0.83 0.86 0.89 0.89 0.52 0.89 0.29 1.04 0.91 0.79 0.66 0.55 1.50 0.80 0.46 0.32 0.85 0.81 0.73 0.91 0.91 0.52 0.41

bg of per capita income 9.197 9.165 4.793 9.326 7.409 9.258 7.836 7.047 7.676 9.408 6.332 9.228 9.336 9.421 8.246 5.480 6.207 8.181 8.773 7.111 9.201 7.012 6.034 7.250 7.428 7.883 6.693

37

Country Netherlands New ~and Norway Pakistm Peru Philippine Portugal Singapore South Africa Spain Sri Lanka Sweden Turkey UK

us
Venezuela Zimbabwe

Equity mmket cap. over GDP 0.19 0.33 0.06 0.03 0.06 0.10 0.01 1.62 1.20 0.09 0.06 0.11 0.01 0.38 0.53 0.05 0.45

Dom-tic credit over GDP 0.72 0.26 0.57 0.50 0,22 0.36 0.81 0.34 0.31 0.93 0.38 0.68 0.34 0.40 0.49 0.29 0.56

Printe credit over domestic credit 0.84 0.72 0.60 0.50 0.48 0.77 0.64 1.67 0.84 0.81 0.56 0.61 0.41 0.63 0.72 1.03 0.53

Log of per capita income 9.320 8.921 9.505 5.671 6.736 6.591 7.741 8.447 7.972 8.534 5.528 9.573 6.985 9.170 9.338 8.288 6.088

38

Table 3:

Summary
Industry real growth is the annual compounded 1980-1990 for @ is obtained is obtained 1980 due ISIC industry in each country. the Aue

Statistics.
growth rate in real value added estimated of period. for the period in the log

The growth in the number of firms is the difference

of ending period firms less the log of firms in the beginning by dividing added in the industry in logs. The industrys by the total Aue External again as a difference added of the industry statistics.

The average size of firms in the industry added is computed that year. dividing the

by the number Of firms, and the growth in average size share of total due All these data come not financed
of

added in manufacturing

from the United Nations Compustat in payables. from Euro

finance is the median fraction of capital expenditures Cash flow from operations in inventories, decre=es in receinbles,

with cash flow from operations

for each industry.

is broadly defined M the sum

funds from operations

(item # 110), decremes

and incre=es Both

Equity finance is the median ratio of the net amount of equity issues to capital expenditures. by the authors using Gmpustat. Investments and Lee (1994).

these variables have been constructuted years of sdooling

share of GDP in 1980 is

The log of the 1980 per capita income in U.S. dollars is from the IFS. The average Financial development is the sum credit to GDP (from Efficiency that assessm the by the Center annual reports

in the population over 25 years is from B=ro and Lee (1994). ~iables

of the ratio of total market value of publicly traded equity to GDP and the ratio of domestic IFS and IFC), The origin of the legal system has four indicator of the judicial system is an index from O to 10 developed efficency and integrity Jor International in each country. of the legal environment. Financial The accounting by Bwiness

from La Porta et al. (1996).

Cor-pomtion

International

standards

is an index developed of companies

Analysis & Reseanh ranking the amount of disclosure

are taken from La Porta et al. 11996).

The last two miables

Iar,able Industry annual real growth Growth in number of firms in industry Growth in average firms size in industry Industrys sh~e of total due added Industry external finance dependence Industry equity finance dependence Investments share of GDP in 1980 bg per capita income in 1980 in dollars Average years of schooling Financial development English law origin French law origin German law origin Scandinavian law origin Efficiency of judicial system Accounting Standards

Mean 0.035 0.013 0.000 0.016 0.319 0.112 0.236 7.814 5.900 0.718 0.333 0.462 0.103 0.103 7.716 61.937

Median 0.029 0,008

Std. Dev

Min. -0.447 -0.209 -0.582 0.000 -0.451 -0.090 0.033 4.793 1.681 0.132 0.000 0.000 0.000 0.000 2.500 24.000

Max. 1,000 0.295 0.994 0.224 1.492 0.764 0.377 9.573 12.141 1.962 1.000. 1.000 1.000 1.000 10.000 83.000

Nobs. 1217 1004 1004 1217 36 36 43 43 41 43 39 39 39 39 39 32

0.090
0.047 0.007 0.021 0.406 0.206 0.076 1.340 2.829 0.380 0.478 0.505 0.307 0.307 2.036 12.600

0.000

0.009
0.231 0.024 0.250 7.883 5.442 0.654

0.000 0.000 0.000 0.000

7.500 63.000

Table 4:

Financial
The dependent wiable

dependence

and industry

growth.

is the annual compounded growth rate in real value added estimated for the period
in

1980-1990 for each ISIC industry in each country. Investments share of GDP in 1980 is from Barro and Lee (1994). The log of the 1980 per capita income in U.S. dollars is from IFS. The average years of schooling
the population over 25 years is from Barro and Lee (1994). of publicly traded equity added is computed Financial development total market due share of total Aue manufacturing to GDP and the ratio of domestic is the sum of the ratio of by the total due External added in is the during the errors are

credit to GDP. The industrys dependence

dividing the 1980 value added of the industry

that year (both these data come from the United Nations statistics). not financed

fraction of capital expenditures for the constant

with internal funds by U.S. firms in the same industry are not reported. Heterosced=ticity robust standard

1980s. The interaction mriable is the product of external dependence and financial development. The coefficients terms and indicator variables reported in brackets.

Variable

OLS

Fixed Country Effects

Fixed Industry Effects

Fixed effects for both Country and Industry

Investments 1980

share of GDP in

0.203 ( 0.057)

0.199 (0.055) -0.018 (0.004) -0.001 (0.002) 0.014 (0.009)

-0.607 ( 0.188 )

kg per capita income in 1980 in dollars Average years of schooling population over 25 years Financial development in

-0.019 (0.004) -0.001 (0.002) 0.013 (0.009) -0.607 (0.181) -0.034 (0.021) 0.069 (0.024)

Industrys share of total value added in manufacturing in 1980 Dependence of industry from external finance in the U.S. Interaction (external dependence financird development)

-0.907 (0.227)

-0.912 ( 0.238)

-0.035 ( 0.020)
0.067 ( 0.024 )

0.071 (0.023)

0.069 ( 0.022)

R-squared Number of obsermtions

0.083 1171

0.233 1217

0.144 1171

0.290 1217

40

Table 5:

Industry

growth

and the composition markets.

of capital
for the period dividing

The dependent miable is the annual compounded growth rate in real due
1980-1990 for each ISIC industry in each country. The industrys the 1980 value added of the industry from the United Nations statistics). Externrd dependence by the total value added in manufacturing is the fraction

added estimated

share of total value added is computed of capital expenditures

that year (both these data come not financed credit is the ratio of equity estimates

with internrd funds by U.S. firms in the same industry during the 1980s. Equity dependence less credit to the priwte not reported). sector. All regressions contain fixed country and industry errors are reported in brackets.

finance to total external finance by U.S. firms in the same industry in the 1980s. Public credit is domestic effects (coefficient Heteroscedasticity robust standard

Variable

I -0.932 ( 0.237) 0.107 ( 0.034) -0.066 ( 0.032)

II

Industrys share of total value added in manufacturing in 1980 Interaction 1 (external dependence X financial development) Interaction 2 (external dependence X stock market/GDP) Interaction 2 (equity dependence X stock market/GDP) Interaction 3 (external dependence X public credit/GDP)

-0.931 ( 0.237) 0.107 ( 0.034) -0.066 ( 0.032) -0.007 ( 0.014)

-0.158 ( 0.046)

-0.158 ( 0.046)

R-squared Number of observations

0,294 1217

(J.~95 1217

41

Table 6:

Robustness.
The dependent variable is the annual compounded
period 1980-1990 for each ISIC industry in ea~ growth rate in real value added estimated for the that of

country. The industrys share of total value added is

statistics). External dependence is the fraction

computed dividing the 1980 value added of the industry by the total value added in manufacturing
year (bet h these data come from the United Nations capital expenditures In column column computed companies not financed with internal funds by U.S. firms in the same industry only for companies that have been public for companies that have gone public in the l=t IV external dependence

during the 1980s. 10 years, in 111 is for all

I this ratio is computed

for at le=t

II the ratio is computed

9 years, in column is computed is interacted

for U.S. firms during the 1970s. In columns (as in the bmic specification), standard of each country (divided

IV and V external dependence

but in columns

with the

level of accounting to industries standard

by 100) and in column V the regression above the median. estimates not reported).

is restricted robust

that had a share of total value added in manufacturing and industry effects (coefficient in brackets.

All regressions

contain fixed country

Heteroscedasticity

errors are reported

Variable Industrys share of total value added in manufacturing in 1980 Interaction (external dependence financial development) R-squared Number of observations

I -0.906 ( 0.252)

II -().924 ( 0.277)

III -0.871 ( 0.238) 0.100 ( 0.049) 0.279 1179

IV -0.658 ( 0.200) 0.158 ( 0.036) 0.346 1032

v
-0.602 ( 0.199) 0.062 ( 0.023) 0.499 591

0.050
( 0.019) 0.273 1118

0.042
( 0.021 ) 0.277 1085

42

Table 7:

Is financial dependence
The dependent -iable 1980-1990 for each ISIC industry from the United Nations statistics). interaction income). in each country.

a proxy for other variables?

growth rate in real value added estimated added in manufacturing for the period

is the annual compounded

The industrys share of total tiue added is computed dividing

that year (both these data come not financed with specification the is the fraction of capital expenditures

the 1980 value added of the industry by the total due

External dependence internal funds by U.S. firms in the same industry between external dependence between external the interaction dependence

during the 1980s. Column I adds to the b=ic level of economic multiplying development

and a countrys human capital. and a countrys a wriable

Column II adds to the basic specification (log per capita ~=iable average our fundamental interaction

Column 111 adds to the basic specification

and a dummy equal to one if an industry

is above median size in a country with above median weighted

financial dependence. Column IV adds to the basic specification the interaction between external dependence and a countrys weighted average external dependence. Column V estimates the basic specification respective countries. Column VI estimates The weighted average external dependence is the sum over that in 1980 were less than the median using instrumental wriables. size in their all industries in a country of each industrys share of value added in 1980 weighted by the external dependence. for industries the basic specification

The instruments

are the origin of a country legal system and the efficiency level of a countrys judicial system. All regressions contain fixed country and industry effects (not reported). Heteroscedasticity robust standard errors are reported
in brackets.

Variable Industrys share of total value added in manufacturing in 1980 Interaction X financial Interaction 1 (external dependence development) 2 (external dependence

I -0.948 ( 0.249) 0.063 ( 0.022) 0.004 ( 0.003)

II -0.944 ( 0.237 0.049 ( 0.022

III -0.903 ( 0.241 ) 0.072 ( 0.027)

Iv -0.892 ( 0.240 ) 0.075 ( 0.037)

v -5.890 ( 1.608 0.074 ( 0.033

VI -0.552

( 0.140)
0.041 (0.020)

X average years of schooling) Interaction 3 (external dependence in 1980)

0.013 ( 0.008)

X log of per capita income

Interaction 1 X dummy if industry above median in country with above median fin. dependence Interaction X weighted R-squared Number of observations 4 (external dependence dependence) 0.286 1171 0.295 1217

-0.004 ( 0.013)

-0.101 ( 0.324) 0.290 1217 0.290 1171 0.359 626 0.113 1149

ave. external

43

Table 8:

Cash Flow and Investments

The dependent period 1980-1990 dividing broadly Investment computed year (both operations industry. estimates variable is the annual compounded the 1980 value added of the industry defined (see text) to net property growth rate in real value added estimated The industrys by the total value added in manufacturing Internal cash flow is the ratio of c=h plant and equipment fixed country to property for the that for each ISIC industry in each country. share of total value added is flow from for

these data come from the United Nations). intensity

for U.S. firms in the same plant and equipment effects in brackets. (coefficient

is the ratio of capital expenditures All regressions contain robust standard

U.S. firms in the same industry. not reported).

and industry

Heteroscedasticity

errors are reported

Variable Industrys share of total value added in manufacturing in 1980 Interaction X financial Interaction (internal cash flow

I -0.873 ( 0.236) -0.229 ( 0.085 ) intensiveness

11 -0.893 ( 0.238) -0.187 ( 0.090)

development) 2 (investment

0,093
( 0.096) 0.288 1217 0.289 1217

X financial development)
R-squared Number of observations

44

Figure

1:

Life Cycle of External

The graph plots the median U.S. across 3-digit finance reduction is the amount in inventories, equity SIC industries

Financing
of the number not financed

and Investments
and investments in the of years since the IPO. External with cash flow from operations, of funds Investment is the ratio of

level of external financing, as a function (CAPEX)

equity financing,

of investments or decreases issues divided

in trade credit. by the amount

Equity finance is the net amount of investments.

raised through

The IPO year is defined as the first year in CAPEX to net property, plant and equipment. which a company stats to be traded on the NYSE, AMEX, or NASDAQ. All the information is obtained CRSP. from the flow of funds data in Compustat, except for the SIC code which is from

Median
1.5

Across

Industries

0.5
q

-. -0.5
I

-1

o =

8
#

Equity Financing

12 16 20 ears since the IPO External Finance b-

24

28

Investments

45

Figure 2:

Life Cycle of External Financing and Investments Two Different Industries

The two graphs plot the average level of external financing, the U.S. for two 3-digit industries. financed with c~h External finmce flow from operations, reduction equity financing, and investments (CAPEX) is the amount of investments in inventories, or decreases

in
in not of

in trade credit. by the amount The IPO AMEX, except

Equity finance is the net amount of funds raised through equity issues divided investments. or NASDAQ. Investment is the ratio of CAPEX is obtained to net property, year is defined = the first year in which a company All the information for the SIC code which is from CRSP. Computer

plant and equipment.

starts to be traded on the NYSE,

from the flow of funds data in Compustat, is SIC code 357, Steel is SIC code 331.

Computer
2
% ..s. .

Industry

(SIC

Code

= 357)

1.5

1
,,;,~. ./ . . . ..

0.5
~-.,.,,,,,, .,

an 0

,, ,:,.,, ,:,:. .:

....... .

-0.5

-1

0
=

4 Equity

8 Financing d-

12 years

16 since

20

24

20

External

IPO Finance

Investments

Steel
2 1.5

Industry

(SIC

code

= 331)

0.5

m s

a# m, -0.5 w

q s, qa 1

.,

-.*;

,8 ,9 i

V;

-1 o = 4 Equity 8 # ~~ Financing s~c~

In mm m 1 m a~:m n~ m 9

:Y:Tm s s

# # q

~ s

s 0: m ~ ~ m~

the lP~
ternal Finance B

24

20
Investments