pecking order asy me try

8/8/2019 Pecking Order Asy Me Try

1/49Electronic copy available at: http://ssrn.com/abstract=939588

The Pecking Order, Information Asymmetry, and

Financial Market Efficiency

Abu Jalal

This Draft: July 10, 2007

Preliminary Version

Abstract: This paper studies the marginal debt issuance behavior of publicly traded companieswith firm-level data from 42 countries. The focus is on the extent to which measures from theliterature on finance and development can help to explain the observed differences amongcountries in the corporate use of marginal debt financing. Using the pecking order testingframework of Shyam-Sunder and Myers (1999), this study provides empirical evidence thatfinancial market imperfections and institutional development affect the debt issuance decisionsof firms when raising external capital. Country development, Law enforcement, legal origin,shareholder protections, effectiveness of the government, and control of corruption aresignificantly related to marginal debt issuance decisions of firms. Finally, the coefficientestimates of the pecking order regressions are correlated with the long run average growth ratesof the countries and appear to be a powerful objective measure of financial market efficiency.

JEL Classifications: G32, F30

Department of Finance, Carlson School of Management, University of Minnesota, Minneapolis, MN 55455. I

thank John Boyd, Murray Frank and Ross Levine for their guidance and support. I am also grateful to Rajesh

Aggarwal and Jan Werner for many helpful comments and suggestions. I alone am responsible for the contents and

any errors. E-mail: [email protected].


2/49Electronic copy available at: http://ssrn.com/abstract=939588

2

I. Introduction

In this paper I use the pecking order testing framework of Shyam-Sunder and Myers (1999) to

examine the external financing decisions of individual firms in 42 countries over the period

1980-2005. The focus is on the extent to which measures from the literature on finance and

development can help to explain the observed differences among countries in the corporate use

of marginal debt financing within the Shyam-Sunder and Myers (1999) framework. Furthermore,

I propose a new and unique use of the beta estimates obtained from the pecking order regressions

as a measure of financial market efficiency.

The first key result of this paper is that there is a significant relationship between the

efficiency and development of the financial markets, and the dependence of firms on new debt

issuance. As equity markets become larger and more liquid, dependence on marginal debt

financing drops significantly.

The second key result is that the estimated coefficients may provide a useful indicator of

growth prospects in each country. The pecking order coefficients are strongly negatively

correlated with the long run average growth rates. The estimated coefficients provide a robust

summary statistic for the effects of a number of macroeconomic variables and indicators of

development. As such the coefficient estimates can be easily used as an objective measure of

financial market efficiency.

The pecking order theory of capital structure depends, fundamentally, on the notion that

equity market frictions are greater than debt market frictions. These frictions can be due to


3/49

3

information asymmetries, adverse selection, or due simply to institutional costs of bringing a

new issue to market. Many studies of the pecking order theory in recent years have followed the

empirical structure developed by Shyam-Sunder and Myers (1999). They estimate regression

equations of new debt financing on a firms deficit of funds flow.1 The slope coefficient

measures the extent to which marginal debt issues are explained by the external financing needs

of firms. My previous work (Jalal (2006)) shows that high values of the coefficient estimates (or

the pecking order betas) are observed in countries where these problems of information

asymmetry and adverse selection are potentially more severe.2 In this paper, I mainly attempt to

understand these differences in the context of financial market efficiency.

There appears to be a deep connection between the premises of the pecking order theory

and the development and efficiency of capital markets in different countries. In many developing

countries, equity markets are thinly traded or non-existent, and presumably the costs of issuing

equity are comparatively high. As a result, firms in those developing countries will resort to

marginal debt financing more frequently than firms in countries where there are well-functioning

equity markets. Gurley and Shaw (1960) find that as economies evolve, financial markets and

institutions become more sophisticated. Boyd and Smith (1998) show in a theoretical model that

equity markets are not needed in early stages of economic development. In the presence of both

debt and equity markets, the relative importance of debt (as captured by aggregate debt to equity

ratio) appears to fall as economies grow. As a result, it is useful to attempt to use the marginal

debt issuance behavior of firms as a measure of how developed the financial markets of a

1 The funds flow deficit is calculated by subtracting operating cash flows after interest and taxes from a sum ofdividend payments, capital expenditures and net increase in working capital.2 A short review of the existing capital structure literature can be found in Jalal (2006).


4/49

4

country are. My empirical investigations in this paper show that the pecking order betas are

indeed a powerful and objective measure of financial market efficiency.

In the current financial market development literature, cross-country differences in the

size and trading volume of stock markets as a share of GDP, stock market turnover, and number

of listed companies have been widely used as measures of stock market development and

efficiency. However, these variables are based on aggregate country level data and thus, may not

adequately use all available information. On the other hand, there exist different development

indicators and surveys compiled by a number of global research institutions, such as the World

Bank, the IMF, and various think tanks. However, they are inherently subjective and their scopes

tend to be limited.

There have also been a number of efforts to use various asset pricing models to capture

cross country risks and financial market development. The works of Korajczyk and Viallet

(1989), Harvey and Zhou (1993), Ferson and Harvey (1993), and Korajczyk (1994), Bekaert and

Harvey (2000), Demirg-Kunt and Levine (1995), Levine and Zervos (1998), and others are

notable in developing and applying international CAPM and international APT models. The

alpha estimates from these models have been utilized to measure stock market integration and

stock market efficiency. The weaknesses of these measures have been widely discussed. For

example, in a perfectly integrated market, we should observe alpha = 0. However, a failure to

reject alpha = 0 means either the market is not integrated or, more problematically, the

underlying model is mis-specified. In addition to violations of the underlying CAPM


5/49

5

assumptions and the difficulty of finding an appropriate benchmark portfolio,3 these measures

are not particularly useful in developing countries with non-functioning equity markets. In

comparison, the coefficient estimates of the pecking order regressions use all available firm level

data and can be updated or calculated for any country or period. Above all, this would be an

objective measure with solid foundation in a well-explored theoretical model. Such a measure

could be useful to researchers and international agencies, as well as many investors deploying

funds across borders.

To be an effective measure of financial market efficiency, the pecking order betas need to

be able to robustly capture different aspects known to be associated with financial market

development in the existing literature. Thus, I devote a significant portion of this study in

understanding and measuring the effects of these direct (such as, level of development as

measured by GDP, measures of equity market efficiency, etc.) and indirect (such as, legal origin,

shareholder protections, country governance, market vs. bank based systems, inflation, etc.)

measures of development on the marginal financing choices of firms.

i. Market vs. Bank-based Systems:

Previous studies suggest significant differences in the types and magnitudes of external

financing choices of firms in different countries depending on the development of financial

markets and intermediaries. Market-based economies tend to be more developed than bank-based

economies. Shareholder protections and contract enforcements are also more effective in market-

3 One shortcoming with this measure is that as a country becomes more integrated internationally, the relevantbenchmark portfolio shifts away from being a benchmark of domestic assets. The relevant benchmark becomesmore internationalized. Thus, domestic risk mis-pricing as measured by CAPM may rise even as the stock marketbecomes more integrated and efficient. Demirg-Kunt and Levine (1995)


6/49

6

based countries (Demirguc-Kunt and Levine (2001)). I try to find if the marginal debt issuing

decisions of firms differ significantly between market-based and bank-based countries. In the

theoretical models of Boyd and Smith (1996, 1998), as countries develop, their economies

become more market-based. Since firms in developed countries generally prefer equity, we may

expect a lower level of dependence on new debt financing among firms in market-based

economies. My regression estimates support this idea. The pecking order beta estimates are

higher in bank-based countries.

ii. Legal Origin and Legal Environment:

There are a number of studies linking a countrys legal origin and legal environment to

the development of its financial markets. LaPorta, Lopez-de-Silanes, Shleifer, and Vishny

(hereafter LLSV) (1997, 1998) show that countries with different legal origins develop their

financial markets differently. Factors such as shareholder rights, contract enforcement, efficiency

of judicial systems, etc. significantly influence the evolution of financial markets and

intermediaries. In countries with poor protection for shareholders, firms have limited access to

external financing and the capital markets are smaller in terms of size and scope. They also find

that French civil law countries have the least shareholder protection, and thus have the least

developed capital markets compared to British common law origin countries. I examine the

relationships of these indicators of legal origin, legal environment, and enforcements with

external financing choices of firms. I find that, in general, countries where there are more legal

protections for shareholders, the firms tend to depend less on marginal debt issuance. The

pecking order betas successfully capture the effects of these indirect measures of financial

market efficiency.


7/49

7

iii. Inflation:

My data on the issuing activities of publicly traded firms in 42 countries allow me to

examine the relationship between the level of inflation and the firms reliance on marginal debt

issuance. Boyd, Levine, and Smith (2001) find that a high level of inflation is associated with

decreased activity in the financial markets. Inflation exacerbates frictions in both banking sectors

and stock markets. They also find strong evidence of nonlinearity and existence of discrete

thresholds in the effects of inflation on market efficiency. Gillman and Harris (2004), Lee and

Wong (2005) and Rousseau and Wachtel (2002) provide further support for these findings. Using

a similar methodology of threshold regressions, I find empirical evidence consistent with the

results of Boyd, Levine, and Smith (2001). I find that higher long run inflation rates are

associated with greater adherence to marginal debt issuance and that this relationship is

decidedly nonlinear. This is yet another indication that higher levels of marginal debt issuance,

and thus higher values of the pecking order betas, are observed in countries where various

factors, such as inflation, complicate development of functioning financial markets.

The remainder of the paper is organized as follows: Section II provides a description of

the empirical models and specifications used in this study. A detailed description of the data is

available in section III. The results and findings are in section IV and V. Section VI describes

some robustness checks and section VII concludes.


8/49

8

II. Empirical Methodology

2.1 Determinants of Marginal Debt Issuance: Firm-specific Considerations

In this study, I follow and expand upon the empirical methodology employed by Shyam-

Sunder and Myers (1999), and Frank and Goyal (2003). The pecking order theory predicts that

due to asymmetric information, adverse selection, and transaction costs, managers will prefer

internal financing over external financing. If external financing is necessary, managers will

prefer debt over equity, since equity is exposed to significantly higher adverse selection

problems. Therefore, according to the theory, firms will finance their projects with funds in the

following order: retained earnings, debt and new equity issues.

The pecking order theory proposes that firms will use equity relatively rarely as a means

to obtain external financing and will use equity financing only after they have exhausted all debt

sources. As a result, a large amount, if not all, of the variability in marginal debt issuance should

be explained by the funds flow deficit. Thus, Shyam-Sunder and Myers (1999) propose the

following test of the pecking order theory: for a firm i ,

ititPOit eDEFbaD ++= * (1)

where tD is the net debt issued in year t and tDEF is the corresponding funds flow deficit.

If the pecking order theory holds precisely, we should observe 0=a and the pecking order

coefficient 1=POb . That is, in the strictest interpretation of the theory, for every dollar of

external financing needed, the firm will issue one dollar of marginal debt.4

4 The regression equation does not include net equity issue. Therefore, it is not an accounting identity.


9/49

9

The empirical methodology of this study is derived mainly from two sources: Frank and

Goyal (2003) and Petersen (2006). The regression equations are estimated with a panel data from

42 different countries. The variables are scaled by net assets as in Frank and Goyal (2003). This

is especially necessary and convenient in an international sample, since the accounting items are

listed in domestic currency. While scaling by net assets helps make estimating regressions across

countries possible, it may affect the coefficient estimates if the scaling variable is correlated with

the variables in the regression equation.5

To check for any significant bias, I estimate the

regressions with book value of assets or sales (instead of net assets) as the scaling variable.

Consistent with Frank and Goyal (2003), my regression estimates are not seriously affected by

the choice of the scaling variable.

Petersen (2006) examines the common biases associated with capital structure panel

regressions and provides a guideline for addressing these econometric issues. He finds that OLS

and Fama-McBeth (1973) standard errors are biased downward, which may provide misleading

support for the hypotheses. In a panel regression, such as the ones under consideration in this

study, it is necessary to correct for residual dependence created by firm effects. Petersen (2006)

finds that fixed effects models with firm dummies only eliminate bias in OLS standard errors if

the firm effects remain fixed and do not decay over time. A possible solution for this problem

is using clustered standard errors. From a simulation of a panel data, Petersen (2006) shows that

clustering corrects for biases due to firm effects most effectively and this bias goes down as the

number of clusters goes up. Therefore, I use cluster analysis for all the regressions.

5 All variables, except for the dummy variables, are winsorized at the 1 percent level.


10/49

10

In addition to firm effects, it is reasonable to consider that the residuals may be correlated

across time. The Fama-McBeth (1973) method is usually used to address these time effects.

However, since Fama-McBeth (1973) standard errors are biased downward, a realistic and

practicalsolution for this problem is to use time dummies in the regression estimates. Therefore,

all regression equations estimated in this paper address time effects parametrically by including

year dummies and address firm effects by calculating standard errors clustered on individual

firms.

2.2 Determinants of Marginal Debt Issuance: International Factors

An important part of this study is to explore the performance of various macroeconomic

variables in explaining the external financing behavior of firms in those countries. The regression

equations follow this general empirical specification:

itYitPOit eYbDEFbaD +++= ** (4)

where Y is the vector of country-specific variables.

Country-specific macroeconomic variables used in this study can be divided into six

general categories (a) economic development, (b) financial market development and efficiency,

(c) legal origin, (d) shareholder protections, (e) quality of governance including corruption and

rule of law, and (f) inflation. The existing literature in macroeconomics and development suggest

that there are significant differences in the development and functioning of financial markets that

are captured by these variables.


11/49

11

In my regressions, I also include the interaction terms ofDEF and Y wherever possible.

The interactions terms serve an important purpose. Consider the following regression

specification:

ititPYYitPOit eYDEFbYbDEFbaD ++++= **** . (5)

We can rearrange terms,

ititPYPOYit eDEFYbbYbaD ++++= *)*()*( . (6)

It is clear that for a given value of Y, we can easily calculate if the pecking order beta estimate

increases or decreases with Y. In addition to providing a non-linear structure, interaction terms

can explain whether the ability of the funds flow deficit to explain new debt issuance is

magnified (positive slope coefficient) or reduced (negative slope coefficient) by the specific

macroeconomic variable as shown in equations (5) and (6). In other words, the ability of the

funds flow deficit to explain marginal debt issuance is different depending on the values of that

country-specific variable.6

6 Including an interaction term even when there is no interaction effect does not involve any econometric issue. Insuch case, the coefficient estimate of the interaction term will be statistically insignificant. However, not includingan interaction term when there is an interaction effect may create omitted variable bias. The interaction terms areusually identified with the prefix I and followed by the names of the original country-specific variable.


12/49

12

III. Data

The source of international data for this study is Worldscope. The database includes

detailed information on over 40,000 publicly traded companies in about 50 countries. However,

the scope of this study is limited to countries with a significant number of non-financial

companies with available data on debt and equity issuance. The number of international

companies represented in my sample is about 17,000 from 41 different foreign countries. The

data for publicly traded U.S. firms are from Compustat. My sample does not include financial

firms (SIC Code 6000 6999) and highly regulated utility firms (SIC Code 4900 4999). Table

1 lists the countries.

The quality of accounting reports and reconciliation of differences in accounting

standards among various countries are important considerations for any researcher dealing with

international data.7 One advantage of using Worldscope is that there is uniformity in the

presentation of accounting data across companies from different countries. Worldscope has

developed its own templates that take into consideration the variety of accounting conventions

and are designed to facilitate comparisons between companies and industries within and across

national boundaries (Worldscope Data Definition Guide). While Worldscope does not change

the basic valuation methods of the data reported by a company in its accounts, it does change the

presentation of information and reformat the accounts to standardize and to significantly improve

the comparability of different accounting items.8

However, we do need to be very cautious in our

7 A number of earlier studies of international capital structure have depended on Global Vantage, including Rajanand Zingales (1995). The fact that Global Vantage later discarded much of the data used in Rajan and Zingales(1995) as unreliable should be a strong note of caution.8 For example, if a company reports sales without excise taxes and another one reports sales with excise taxes,Worldscope analysts will make necessary changes to make those two data items comparable by following a standard


13/49

13

inferences from the data and undertake various robustness checks to address some of the

empirical issues.

A major source of potential bias to take into account is sample selection bias. Worldscope

was originally developed by a U.S. based global money management firm, as a means to provide

investors with accounting information on publicly traded companies. As a result, early data

represent mainly large companies with high visibility. Over the years, Worldscope has added a

significant number of smaller publicly traded companies. According to Thompson Financial, the

current provider of the dataset, Worldscope covers more than 95% of the worlds market value.

However, this study concentrates on marginal debt and equity issuance. Only about 70% (with a

standard deviation 18%) of the companies in my sample provide issuance data and they tend to

be the larger in size.9 As a result, the results may not fully represent the smaller firms in these

countries. However, I have taken careful steps to empirically control for firm size and to make

sure that it does not severely impact the results.

Table 2 provides a comparison of the average market capitalization, common equity, total

assets, sales, and net income by country. In some countries, such as Brazil and Russia, the

average market capitalization of the firms is relatively high. This is mainly due to the existence

of a few very large companies in these countries. For example, in 2005 the market capitalizations

of the three largest Brazilian companies in the sample Petrobras (Petroleo Brasileiro S.A.),

data coding system. Similarly, minority interests are separated from shareholders' equity and deducted in arriving atnet income even if the original financial statements from some companies did not do so.9 The lowest percentages of companies reporting are from Greece, China, and Japan, with less than 30% of thecompanies reporting marginal debt issuance data. On the other hand, companies from Colombia, Hong Kong,Hungary, India, Netherlands, Pakistan, Singapore, and UK report their marginal debt issuance data in largerfractions more than 90% of the companies reporting.


14/49

14

CVRD (Companhia Vale do Rio Doce), and Ambev (Companhia de Bebidas das Americas) were

approximately $70, $41 and $25 billion respectively. Similarly, in 2004, the four largest firms in

Russia Mechel, Gazprom, Surgutneftegas, and Lukoil had market capitalizations of $99, $55,

$27 and $25 billion respectively. However, my data also include very small companies from

these countries. In my sample, the firm with the lowest market cap in Brazil is about $1 million

and in Russia, about $37 million.


15/49

15

IV. Results

The basic pecking order beta estimates for each country individually are presented in

Table 3. For comparison purposes, the countries are presented according to the magnitude of the

pecking order beta estimates from small to large. The pecking order beta estimates are greater

than zero in all countries and statistically significant everywhere. The lowest beta estimate is for

Australia with 0.1580, followed by UK with 0.1636 and Canada with 0.1977. Australia, UK and

Canada belong to the high per capita income group. On the other hand, the highest pecking order

beta is for Mexico with an estimate of 0.8391, followed by Russia with 0.8317 and Peru with

0.7817. Brazil and Peru belong to the lower middle income group and Russia belongs to the

upper middle income group.

It is obvious from Table 3 that the pecking order beta estimates are, on average, smaller

for developed countries than for developing countries. It appears that the marginal debt issuance

behavior of firms is related to the development of countries. Now, I concentrate on

understanding the relationship between the pecking order beta and a variety of international

factors how the pecking order beta responds to cross-country differences in development and

financial market efficiency.

Conceptually, international factors can be divided into two categories direct and

indirect measures of efficiency. Direct measures provide an assessment of the degree of

development of a country and its financial markets. Such measures employed in the literature

include GDP, private credit provided by the banking system, stock market capitalization, and

total value of stock traded. On the other hand, what I call indirect measures are other


16/49

16

macroeconomic indicators that have been found to be significantly correlated with well-

developed financial markets. These measures include legal origins of countries, shareholder

protection, country governance, and inflation.

4A. Direct Measures: Economic Development and Financial Market Efficiency

The dependence of firms on debt issuance, as captured by the pecking order betas, can be

studied in the context of the evolution of the financial markets. Gurley and Shaw (1960) find that

as economies evolve, financial markets and institutions become more sophisticated. In a

primitive economy, without any banks, all projects are financed through owner equity. As

economies grow and banks are introduced, new capital investments may be financed exclusively

with debt. Boyd and Smith (1998) show in a theoretical model that equity markets are not needed

in early stages of economic development. In the presence of both debt and equity markets, the

relative importance of debt (as captured by aggregate debt to equity ratio) appears to fall as

economies grow.10 The existence of strong equity markets is associated with high levels of

economic development.

Since the costs of information asymmetry and adverse selection are higher in countries

with less transparent and sophisticated financial systems, we can develop some implications

regarding the pecking order beta and the evolution of the financial markets as shown in Figure 1.

10 There do exist some exceptions.


17/49

17

In truly primitive economies where all projects are financed with owner-equity, the value of the

pecking beta obviously equals to zero. In economies with only debt markets, the pecking beta

should jump to the value of one as depicted in Figure 1. Finally, the pecking order beta estimates

should decrease as countries become more developed and firms employ more and more equity.

In this study, it is not possible to verify the relationship between the development of the

countries and the pecking order betas where there are no financial markets ( 0PO ) since no

economies that primitive are included in the sample. However, I can test if the value of the

pecking order beta decreases as countries become more developed with my sample of 42

countries that have both debt and equity markets.

Owner

Equity

+ (Bank) Debt

Market

+ Equity Markets

0

1

Figure 1: The Evolution of Financial Markets and the Pecking Beta

Development

PeckingBeta


18/49

18

Following Demirguc-Kunt and Levine (2001), I constructInitialGDPas a measure of the

level of economic development.11

This is simply the real per capita GDP at the beginning of the

sample in 1980.12 The first equation in Table 4 shows a negative relationship between level of

development and marginal debt issuance decisions of firms. The coefficient estimates for both

InitialGDPand the interaction termI InitialGDPare negative and statistically significant. Thus,

marginal debt issuance is lower among firms in developed countries and the pecking beta

significantly decreases as the level of development increases.

Whether this negative relationship between the pecking order beta and the economic

development of countries also applies to financial market development can be tested with some

aggregate measures of debt market and equity market efficiency widely used in the literature.

Since greater dependence on marginal debt issuance is theoretically associated with lower levels

of development, we should observe higher values of the pecking order beta in countries with

large debt markets. Similarly, we should observe lower pecking order beta estimates in countries

with large equity markets.

A commonly used measure of the level of banking development isPrivo. It is calculated

by taking the amount of private credit provided by deposit money banks and other financial

institution as a share of GDP. I also include an interaction termI Privo. The results are reported

in Column (2) of Table 4. The coefficient estimates for both Privo andI Privo are positive and

11 In development finance literature, initial GDP, as opposed to long term average GDP, is used to obtain a trulyexogenous indicator of the level of development of a country.12 TheInitialGDPdata for Poland and Russia are not available for 1980. So, they were predicted backward frommore recent data using the regression equation GDPt= a + b * Year + e. To capture any interaction between thefinancing needs of firms and the economic development of the country, an interaction term is also included in theregressions.


19/49

19

statistically significant. This supports the idea that in economies where there is a greater

dependence on banks, we observe higher values of the pecking beta.

I consider two measures of equity market development and liquidity. StockMktCap is

total stock market capitalization as a share of GDP. StockTO is a turnover ratio that reflects the

liquidity of the stock market. It is the ratio of the total value traded divided by the market

capitalization. Higher turnover ratio usually implies more liquid stock markets. The estimates are

presented in Columns (3) and (4) of Table 4. The coefficient estimates ofStockMktCap,I

StockMktCap, andI StockTO are all negative and statistically significant. Therefore, countries

where there are well developed and liquid stock markets, firms demonstrate lower levels of

reliance on marginal debt financing.

Finally, I consider a variable that supposedly measures the sophistication of the financial

markets in a country. CAIor the Capital Access Index measures the ability of new and existing

businesses to obtain financing for their projects. In addition to capturing the strength and breadth

of the traditional financial markets and intermediaries, this index accounts for the development

of economic institutions, venture capital funding, private placements, internationalization,

securitization, and macroeconomic environments of different countries.13 The regression

outcomes are presented in columns (6) of Table 4 show that the coefficient estimates for CAIand

I CAIare all negative and statistically significant. Therefore, in countries where it is easier to

access capital and various forms of sophisticated financial instruments, the firms depend less on

marginal debt issuance.

13 The Capital Access Index (CAI) is annually published by Milken Institute. The variable used in this study is anaverage of the index values reported between 2000 and 2006.


20/49


21/49

21

include dummy variables to indicate the legal origins of the countries as controls in the basic

pecking order equation. The results are presented in Table 5.14

As expected, the interaction termI UK Origin isnegative and statistically significant

(column (1)). Therefore, the British countries issue less marginal debt than the non-British

countries. On the other hand, the termsI French Origin andI German Origin are positive and

statistically significant (columns (2) and (3)). As a result, we would observe higher values of the

pecking order beta in French and German origin countries. Column (4) of Table 5 shows that the

pecking order betas for French origin countries fall between British and German origin countries.

These findings are, of course, consistent with the pecking order beta being an inverse measure of

the financial market efficiency.

4B.2. Indirect Measures: Shareholder Protection

In this section, I study the relationship of some indicators of shareholder protection (as

proposed by LLSV (1998)) with marginal debt issuance behavior of firms. I consider four

individual indicators of shareholder rights and two indices used by LLSV (1998): Minor,

Preemptn,Esmreq,Reserve, Srights, and Crights. Interaction terms are also included in the

regression equations.

The dummy variable Minoris an indicator of the rights of the oppressed minority

shareholders (holding less than 10% of the shares) to challenge the decisions of the board of

14 There are disagreements among scholars on how to categorize the countries in terms of their legal origins. In myregressions, I use the definitions of LLSV (1998). Legal origins of China, Hungary, Poland, are Russia are notdefined in LLSV (1998) and thus, I supplement the data from Siems (2006). I also include interaction termscalculated by multiplying the funds flow deficit and the indicators of legal origins.


22/49

22

directors. If the minority shareholders have the right to sue the directors in a court or force the

company to buy their shares, Minortakes a value of 1. The dummy variableProxy identifies

countries where shareholders are allowed to submit their proxy votes to the firm by mail.Esmreq

is the minimum percentage of ownership of shares required to call an Extraordinary Shareholder

Meeting. Lower values ofEsmreq may be considered as more protective of the shareholder

rights.Reserve is the minimum percentage of total share capital mandated by Corporate Law to

avoid a decision to dissolve the firm. Higher values of the variable Reserve are considered to

provide better protection for the creditors. Srights is an index that aggregates the shareholder

rights. The index ranges from 1 to 6 and higher values indicate greater shareholder protection.

On the other hand, Crights is an index of the creditor rights. The value ofCrights ranges between

0 and 4, with higher values indicating greater protection for the creditors of the firm. The

coefficient estimates are presented in Table 6.

The results show that higher pecking order beta estimates are associated withEsmreq and

Reserve. On the other hand, in countries that provide better protection for minority shareholders

as captured by Minorand Proxy, we will find lower values of the pecking beta. These results

indicate that shareholder protection does have a significant relationship with marginal debt

issuance behavior of firms. In the presence of better legal protection for the shareholders, firms

depend less on marginal debt issuance and thus, we can expect to observe smaller beta estimates

in the pecking order regressions.

The relationships between marginal debt issuance and the indices Srights and Crights are

also informative. Countries with higher shareholder rights, as captured by Srights and the


23/49

23

interaction term I Srights, are less dependent on marginal debt issuance. The coefficient

estimates for both Srights and I Srights are negative and statistically significant. On the other

hand, the estimated coefficient for Crights and the interaction termI Crights are both positive

and statistically significant. Thus, as their external financing needs increase, firms depends more

on marginal debt issuance in countries with greater legal protection for the shareholders and less

in countries providing greater protections for the creditors. These findings are interesting and

probably merit further investigation.

4B.3. Indirect Measures: Country Governance, Corruption, and Freedom

It may be argued that without legal protection, contract enforcement, and freedom of

choice, a country may never develop a group of reliable and educated investors (Bardhan (1997),

LLSV (1997), and Shleifer and Vishny (1993)). To study the relationship between these

variables and the marginal debt issuance decisions of firms, I modify the basic pecking order

regression to control for Corruption,Bribery, Government Effectiveness,Rule of Law, Voice and

Accountability, andRegulatory Quality. I also include interaction terms in all the regressions.

CPIis the 2006 Corruption Perception Index Score provided by Transparency

International. Higher values ofCPIindicate lower level of corruption.Bribery Survey is the 2006

Bribe Payers Index that measures the supply side of corruption. It calculates the average

willingness of the companies in the developed countries to provide bribe to foreign entities. The

score is available for 27 mostly developed countries in this sample. Higher values of the Bribery

Survey indicate lower level of willingness to bribe abroad.


24/49

24

The regression estimates are presented in columns (1) and (2) of Table 7. The results

show a statistically significant negative relationship between marginal debt issuance and CPI.

The relationship between marginal debt issuance andBribery Survey is also negative and

statistically significant. The interaction terms are negative and statistically significant in both

cases. Therefore, firms rely more on new debt issuance in countries where companies are willing

to engage in illegal activities, whether it is in the form of corruption or providing bribe to others.

I also consider four variables that depict the quality of governance of the countries in my

sample.

15

The variable Gov Effectiveness measures the level of effectiveness of the bureaucracy

and public servants in the country and the credibility of the government.Rule of Law indicates

whether government agents, including the judiciary, abide by the laws of the country and

whether contracts are properly enforced. The quality of the political process, civil rights, political

rights and the rights of the citizens to choose their representative government are reflected in

Voice and Accountability.Regulatory Quality measures if market-unfriendly policies are enacted

or excessive regulations are imposed on trade and business activities. Higher values of these

variables are better for freedom and business environment. The regression estimates are also

presented in Table 7. The coefficient estimates for all these variables and the corresponding

interaction terms are negative and statistically significant. Therefore, there is a lower level of

dependence on marginal debt issuance among firms in countries where there are greater

effectiveness of the government, enforcement of laws, encouragement of business activities, and

political freedom of the citizens. Again, all these findings are consistent with the pecking order

beta being a summary measure of financial market efficiency.

15 See Appendix A for a description of the sources of these variables.


25/49

25

4B.4. Indirect Measures: Inflation

Boyd, Levine, and Smith (2001) find evidence of negative impacts of inflation on the

performance of the financial markets. Bank lending activity, bank liability issues, stock market

size and liquidity are strongly negatively correlated with inflation. Stock return volatility

increases with inflation. They also show these relations to be nonlinear. There exist discrete

thresholds in the effects of inflation on market performance. Their results have been corroborated

by a number of subsequent studies of inflation and financial markets. In this section, I try to

asses if the pecking order betas robustly capture these effects of inflation. This would be an

important finding since my approach and the dependent variable in this study are totally different

from anything that has been attempted in the current literature.

The regression estimates are presented in Table 8. I test for a linear model with I

Inflation as a control variable in the basic pecking order model. I also calculate two regression

models by including control variables with squared and cubic values of long run inflation. In

column (3) of Table 8, I find a positive coefficient estimate for I Inflation, a negative coefficient

forI Inflation2and a positive coefficient forI Inflation

3. All of the coefficient estimates to

different powers of inflation are statistically significant at the one percent level. This implies that

the pecking order beta increases with inflation at very low (negative values) and very high levels

of inflation as shown in Figure 2. There is a negative relationship between marginal debt

issuance and inflation in the middle.16 This provides evidence of a nonlinear effect of inflation on

marginal debt issuance.

16 I also estimated regression models with higher powers of theInflation variable as controls (not presented here).However, the coefficient estimates lose their relevance if higher powers (greater than the cubic term) ofInflation areincluded in the regression equation.


26/49

26

I also test a threshold regression model similar to Boyd, Levine, and Smith (2001). In

their regressions, they construct a dummy variable (HIPI15) that takes the value of 1 if inflation

rate (PI) is greater than 15%. They also calculate an interaction term (PIHIPI15) by multiplying

inflation rate (PI) and the dummy variable (HIPI15). In their estimates with indicators of stock

market efficiency as dependent variables, they find negative coefficients for PIandHIPI15 and a

positive coefficient forPIHIPI15. Boyd, Levine, and Smith (2001) conclude that as inflation

rises, the performance of equity market diminishes. More importantly, the marginal impact of

additional inflation on stock markets also decreases with increasing levels of inflation. I

construct a set of very similar variables. The indicator variableDummy takes the value of 1 if

long run inflation rate is greater than 6%.17 The variableInteraction is obtained by multiplying

Dummy andInflation. If Boyd, Levine, and Smith (2001) are correct, I should observe marginal

debt issuance to increase in a non-linear fashion as inflation goes up.

0

0.05

0.1

0.15

0 5 10 15 20 25 30 35 40 45 50

Inflation (%)

Pecking

Beta

Polynomial Model Threshold Model

17 The lower threshold may reflect the significant reduction in inflation throughout the world in recent years.


27/49

27

Figure 2: Comparison of the two empirical models involving inflation and pecking beta

(DEF= 0.15,InitialGDP= 10000)

In my estimates (column 4 of Table 8), I find negative and statistically significant

coefficient estimates forI Inflation andI Dummy. The coefficient estimate forI Interaction is

positive and statistically significant. This is very similar to the results obtained by Boyd, Levine,

and Smith (2001). It supports the idea of the existence of discrete thresholds in the effects of

inflation on marginal debt issuance. More importantly, this set of regressions provides more

support for the pecking order beta as a measure of financial market efficiency. At least, it mimics

the highly non-linear results obtained by Boyd, Levine, and Smith (2001) with a different

dependent variable.


28/49

28

V. Pecking Order Beta and Long-run Growth

A large literature in development finance is dedicated to finding factors that are related to

economic growth and thus, providing policy-makers guidelines on how to create and nurture

financial systems that accelerate economic development. To ascertain the ability of the pecking

order beta to predict economic growth, I compare it against six frequently-used well-established

measures of financial market development net interest margin, private bond market

capitalization, stock market capitalization, stock market turnover, stock value traded, and private

credit provided by banks.

Levine and Zervos (1998) show that stock markets and banks play important, but

different, roles in promoting growth. They find that stock market liquidity (measured by total

value traded as a fraction of GDP or stock market turnover), stock market size (measured by

market capitalization as a fraction of GDP), and development of the banking sector (measured by

private credit provided by banks) are strongly related to economic growth. These variables are

widely used as measures of financial market efficiency in the current literature.

Another important measure of cost of intermediation commonly used in development

finance literature is net interest margin. It is the difference between what the bank pays the

depositors and what the bank receives from the borrowers. Demirg-Kunt and Huizinga (2000)

find that as countries become more developed, the net interest margin drops significantly.

Finally, corporate bond market development often indicates the sophistication of the financial

markets in a country. Herring and Chatusripitak (2000) show that bond markets matter for


29/49

29

financial development of a country and it is linked to increased economic efficiency and reduced

probability of financial crisis.

In Panel A of Table 9, I present the correlations between the estimates of the pecking

order beta POb and other conventional variables frequently used as indicators of development

and financial market efficiency. As discussed before, higher POb is associated with lower levels

of equity market efficiency. The correlation between POb andInitialGDPis -0.4979. The

correlation coefficient between the pecking betas and the long run average stock market

capitalization (or stock value traded) is -0.5826 (-0.5088). A commonly used measure of debt

market development isPrivo defined as the long run average value of private credits provided by

deposit money banks and other financial institutions as percentage of GDP. The correlation

between POb andPrivo is -0.6017. These correlations suggest that there is a significant

relationship between the pecking order beta and commonly used measures of financial market

development. In addition, the pecking order beta is strongly correlated with level of development

itself.

In Panel C of Table 9, I estimate regression equations to test the in-sample relationships

between average long run GDP growth and the indicators of growth prospects. All dependent

and independent variables, except forInitialGDP, are calculated for the period between 1980 and

2005. After controlling for the initial size of the economy (InitialGDP), the pecking order betas

are statistically significant (column (1) of Panel C). This statistical significance remains mostly

intact even when I include the measuresPrbond,Netintm, Sttrade, and Stturn. In Panel D of

Table 9, I estimate regression equations to test out-of-sample relationships between long run


30/49

30

average GDP growth and the growth indicators. The dependent variable GDP Growth is the

average real GDP growth during the period 2001 to 2005. All independent variables, except for

InitialGDP96are calculated for the period between 1996 and 2000. After controlling for the

initial size of the economy (InitialGDP96), the pecking order beta and Sttrade appears to be the

only independent variables to maintain a statistically significant relationship with GDP Growth.

The pecking order betas remain statistically significant after controlling other indicators of

growth as shown in columns (8) to (13) in Panel D of Table 9.

Studying economic growth and developing an understanding of the determinants of

growth are important considerations in development finance. These in-sample and out-of-sample

results clearly show that, similar to other successful and widely used measures of financial

market development and efficiency, the pecking order betas are strongly related to economic

growth. The results involving out-of-sample short-term relationships between pecking order beta

and country growth are particularly interesting and requires further investigation.

VI. Robustness Checks

The main robustness check in this study involves cross-country differences in accounting

practices and the biases they introduce in the regression estimates. As mentioned before,

Worldscope modifies the accounting data from different countries to present them in a uniform

reporting format. However, there are limitations to these efforts. These biases are especially

relevant when estimating pooled regressions with data from all countries of the world. One

possible solution is to compute the country-level pecking order betas and use those estimates as a


31/49

31

dependent variable.18 In this way, any bias due to cross-country differences in accounting

standards will be, at least, minimized. I have replicated most of the results presented in this paper

using pecking betas as dependent variable (not presented here). The conclusions derived in this

paper hold robustly regardless of the methodology used.

VII. Conclusion

The pecking order theory of capital structure proposes that due to information asymmetry

and adverse selection costs, firms will rarely use equity when they have financing needs

unfulfilled by internally generated cash. They will almost always depend on marginal debt

issuance. Shyam-Sunder and Myers (1999) provide an empirical framework to estimate the

extent to which new debt issues are explained by the external financing needs of the firms. They

estimate regression equations of new debt financing on a firms deficit of funds flow. The slope

coefficient measures the extent to which marginal debt issues are explained by the external

financing needs of firms. I find that firms in developed countries depend less on marginal debt

issuance when compared to the firms in the developing countries.

There is a statistically significant correlation between the pecking order beta estimates

and the long run financial market development and efficiency. The pecking order betas behave as

if they were a measure of financial market efficiency. Furthermore, the Shyam-Sunder and

Myers (1999) framework robustly captures a number of well-established facts regarding

financial market efficiency. Firms in market-based economies issue less marginal debt than firms

18 This method is used in Gozzi, Levine and Schmukler (2006). They also work with Worldscope data.


32/49

32

in bank-based economies. Firms in British legal origin countries use marginal debt the least and

firms in French origin countries use it more.

I observe a lower level of marginal debt issuance if a country provides greater legal

protection for the shareholders and thus, facilitates a more efficient equity market. The

relationship is exactly opposite for countries providing greater protections for creditors.

Furthermore, firms depend less on marginal debt financing in countries where there are greater

effectiveness of the government, enforcement of laws, encouragement of business activities, and

political freedom. Consistent with the existing literature, I find that the effects of inflation on

marginal debt issuance are nonlinear and that there exist discrete thresholds in the relationships

between inflation and new debt issuance. All these findings are consistent with the notion that

the pecking order beta is a powerful summary statistic for financial market development.

I propose that the pecking order beta can be used as a measure of financial market

efficiency. The coefficient estimates of the pecking order regressions use all available firm level

data, can be updated and calculated for virtually any country or time period. Above all, this is an

objective measure with solid foundation in a well-explored theoretical model. As a result, this is

arguably better than the existing measures of financial market efficiency that depend on

subjective indices and aggregate country-level indicators.


33/49

33

References

Allen, F. and D. Gale, 1999, Comparing financial systems, Cambridge, MA: MIT Press.

Bardhan, P., 1997, Corruption and development: A review of issues, Journal of Economic Literature 35, 1320-1346.

Beck, T., Demirg-Kunt, A. and R. Levine, 2000, A new database on financial development and structure, World

Bank Economic Review 14, 597-605.

Bekaert, G. and C. Harvey, 2003, Emerging markets finance, Journal of Empirical Finance 10, 3-55.

Berglof, E., 1990, Capital structure as a mechanism of control: a comparison of financial systems, The Firm as a

Nexus of Treaties, London, UK: Sage Publications.

Booth, L., Aivazian, V., DemirgcKunt, A. and V. Maksimovic, 2001, Capital structures in developing countries,

Journal of Finance 56, 87-130.

Bose, B., 2002, Inflation, the credit market, and economic growth, Oxford Economic Papers 54, 412-34.

Bose, N. and R. Neumann, 2005, Explaining the trend and the diversity in the evolution of the stock market, CeGE

Discussion Paper 47.

Boyd, J. and B. Smith, 1996, The coevolution of the real and financial sectors in the growth process, World Bank

Economic Review 10, 371-96.

Boyd, J. and B. Smith, 1998, The evolution of debt and equity markets in economic development, Economic Theory

12, 519-60.

Boyd, J., Levine, R. and B. Smith, 2001, The impact of inflation on financial sector performance, Journal of

Monetary Economics 47, 221-48.

Buch, C., 2000, Are banks different? Evidence from international data, Kiel Working Paper 1012.

Demirg-Kunt, A. and H. Huizinga, 2000, Financial structure and bank profitability, World Bank Working Paper

2430.

Demirg-Kunt, A. and R. Levine, 1995, Stock Market Development and Financial Intermediaries: Stylized Facts,

World Bank Policy Research Working Paper 1462.

Demirg-Kunt, A. and R. Levine, 1999, Bank-based and market-based financial systems: cross-country

comparisons, World Bank Working Paper 2143.

Demirg-Kunt, A. and R. Levine, 2001, Financial structure and economic growth: A cross-country comparison of

banks, markets, and development, Cambridge, MA: MIT Press.


34/49

34

Demirg-Kunt, A. and V. Maksimovic, 1998, Law, finance, and firm growth, Journal of Finance 53, 2107-2137.

Demirg-Kunt, A. and V. Maksimovic, 2000, Funding growth in bank-based and market-based financial systems:

evidence from firm-level data, World Bank Working Paper 2432.

Eckbo, B. and R. Masulis, 1992, Adverse selection and the rights offer paradox. Journal of Financial Economics 32,

293-332.

Erb, C., Harvey, C. and T. Viskanta, 1996, Expected returns and volatility in 135 countries. Journal of Portfolio

Management 22, 46-58.

Fama, E. and J. MacBeth, 1973, Risk, return and equilibrium: empirical tests, Journal of Political Economy 81, 607-

636.

Ferson, W. and C. Harvey, 1993, The risk and predictability of international equity returns, The Review of Financial

Studies 6, 527-566.

Frank, M. and V. Goyal, 2003, Testing the pecking order theory of capital structure, Journal of Financial Economics

67, 217-248.

Frank, M. and V. Goyal, 2004, Capital structure decisions: which factors are reliably important? Working paper,

University of British Columbia.

Frank, M. and V. Goyal, 2005, Tradeoff and pecking order theories of debt, Working Paper, University of

Minnesota.

Gillman, M. and M. Harris, 2004, Inflation, financial development and endogenous growth, Working Paper, Monash

University, Australia.

Gurley, J. and E. Shaw, 1960: Money in a theory of Finance. Washington, DC: Brookings Institute.

Harvey C. and G. Zhou, 1993, International asset pricing with alternative distributional specifications, Journal of

Empirical Finance 1, 107-131.

Herring, R. and N. Chatusripitak, 2000, The case of the missing market: the bond market and why it matters for

financial development, ADB Institute Working Paper 11.

Korajczyk, R. and C. Viallet, 1989, An empirical investigation of international asset pricing, Review of Financial

Studies 2, 553-585.

LaPorta, R., Lopez-de-Silanes, F., Shleifer, A. and R. Vishny, 1998, Law and finance, Journal of Political Economy

106, 1113-1155.


35/49

35

LaPorta, R., Lopez-de-Silanes, F., Shleifer, A. and R. Vishny, 1997, Legal determinants of external finance, Journal

of Finance 52, 1131-1150.

Lee, C. and S. Wong, 2005, Inflationary threshold effects in the relationship between financial development and

economic growth: evidence from Taiwan and Japan, Journal of Economic Development 30, 49-69.

Levine R., 1991, Stock markets, growth, and tax policy, Journal of Finance 46, 1445-1465.

Levine, R. and S. Zervos, 1998, Stock markets, banks, and economic growth, American Economic Review 88, 537-

558.

Levine, R., 2002, Bank-based or market-based financial systems: which is better? Journal of Financial

Intermediation 11, 1-30.

Myers, S. and N. Majluf, 1984, Corporate financing and investment decisions when firms have information that

investors do not have, Journal of Financial Economics 13, 187-221.

Myers, S., 1977, Determinants of corporate borrowing, Journal of Financial Economics 5, 147-175.

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

Petersen, M., 2006, Estimating standard errors in finance panel data sets: comparing approaches, Working Paper,

Northwestern University and NBER.

Rajan, R. and L. Zingales, 1995, What do we know about capital structure? Some evidence from international data,

Journal of Finance 50, 1421-1460.

Rousseau P. and P. Wachtel, 2002, Inflation thresholds and the finance-growth nexus, Journal of International

Money and Finance 21, 777793.

Schumpeter, J., 1934, The theory of economic development, Cambridge, MA: Harvard University Press.

Shleifer, A. and R. Vishny, 1986, Large shareholders and corporate control, Journal of Political Economy 94, 461-

488.

Shleifer, A. and R. Vishny, 1993, Corruption, Quarterly Journal of Economics 108, 599-617.

Shyam-Sunder, L. and S. Myers, 1999, Testing static trade-off against pecking order models of capital structure,

Journal of Financial Economics 51, 219-244.


36/49

36

Table 1: Countries in the Sample

Countries are divided in groups by income level as defined by World Bank.

High Income

(OECD)

High Income

(Non OECD)

Upper Middle

Income

Lower Middle

Income

Low Income

AustraliaAustriaBelgiumCanadaDenmarkFinlandFranceGermanyGreece

ItalyJapanKoreaNetherlandsNew ZealandNorwayPortugalSpainSwedenSwitzerlandUK

USA

Hong KongIsraelSingapore

ArgentinaChileHungaryMalaysiaMexicoPolandRussiaSouth AfricaTurkey

BrazilChinaColombiaIndonesiaPeruPhilippinesThailand

IndiaPakistan


37/49

37

Table 2: Mean Company Characteristics

Average Market Capitalization, Common Equity, Total Assets, Sales, and Net Income of firms in the sample of non-financial and non-utility public companies. Data reflects only companies that report debt-issuing activities. Allnumbers are in thousands of US Dollars.

Country Market Cap CommonEquity

Total Assets Sales Net Income Number ofCompanies

Argentina 1,134,443 619,398 1,278,602 655,793 42,461 58

Australia 318,205 164,712 366,677 341,512 14,898 1397

Austria 466,085 320,653 1,061,691 911,136 29,002 99

Belgium 996,957 547,589 1,725,988 1,949,517 52,684 124

Brazil 983,979 1,105,090 2,280,781 1,037,135 65,055 214

Canada 671,980 321,123 836,350 665,253 22,193 1247

Chile 585,296 384,679 697,761 413,354 35,119 87

China 464,304 246,470 532,415 347,695 22,513 462

Colombia 594,048 263,804 444,514 230,216 15,647 33

Denmark 507,783 202,847 496,949 516,576 26,960 183

Finland 771,134 394,710 1,254,382 1,216,864 49,164 236France 1,324,892 666,101 2,655,984 2,284,238 56,617 1054

Germany 1,932,312 926,315 3,638,357 3,719,658 81,308 749

Greece 930,697 368,505 906,113 616,978 35,417 94

Hong Kong 389,261 287,052 573,105 283,149 22,272 914

Hungary 422,176 230,495 451,200 385,341 26,214 32

India 411,155 162,271 420,310 340,760 25,188 490

Indonesia 192,770 87,862 259,493 154,960 7,954 255

Israel 609,183 252,514 865,038 462,340 12,901 135

Italy 1,361,512 674,891 3,252,962 2,034,208 45,699 340

Japan 1,989,018 1,008,583 3,680,689 3,724,613 46,191 1776

Korea 266,447 264,580 1,044,010 996,680 12,741 669

Malaysia 211,819 116,276 273,200 155,061 9,951 815Mexico 1,439,053 734,037 1,739,613 1,151,512 82,839 137

Netherlands 2,920,330 1,029,031 2,785,409 3,090,415 140,997 328

New Zealand 462,280 289,863 757,672 507,797 29,342 103

Norway 434,360 260,594 844,744 698,690 24,773 221

Pakistan 101,440 49,507 122,655 127,753 10,005 107

Peru 264,294 149,197 271,247 144,489 14,682 43

Philippines 193,994 106,374 259,172 128,914 8,044 148

Poland 277,485 149,564 325,372 312,661 12,132 89

Portugal 550,910 221,830 740,325 476,431 22,229 95

Russia 4,321,914 4,604,676 7,267,307 3,390,559 540,609 40

Singapore 285,230 155,907 356,975 205,591 13,914 594

South Africa 551,764 257,569 565,815 661,720 40,561 336

Spain 1,893,679 808,946 2,343,311 1,599,872 83,468 157

Sweden 1,090,359 469,269 1,428,035 1,356,171 54,344 307

Switzerland 3,248,727 1,089,300 2,642,422 2,061,472 145,298 237

Thailand 177,904 76,866 245,027 145,755 6,701 421

Turkey 379,573 142,187 361,479 479,047 20,988 168

UK 716,277 272,616 714,834 708,717 34,879 2809

US 1,037,982 336,517 970,680 938,456 31,342 18662


38/49

38

Table 3: Pecking Order Regressions

Basic Pecking order regression: Dit= a + bPO * DEFit+ it, where D is the net debt issue andDEFis the totalfinancing deficit.DEF= Cash Dividends + Investments + Change in Working Capital Internal Cash Flow. Allvariables are scaled by Net Assets. Financial and utility firms are not included. OLS regressions with robust standarderrors by individual countries, unless otherwise mentioned. All regressions include dummy variables for year and

clustering for individual firms. Countries are sorted according to the coefficient estimates ofbPO, small to large.

Country DEF Constant# Obser(se) (se) (R2)

All Countries 0.2200 0.0233 265836(0.0040)*** (0.0135)* (0.26)

Australia 0.1580 0.0483 8452(0.0100)*** (0.0010)*** (0.18)

UK 0.1636 0.1358 23116(0.0076)*** (0.0012)*** (0.17)

Canada 0.1977 -0.0739 9080

(0.0107)*** (0.0307)** (0.20)US 0.2106 -0.0033 144903

(0.0033)*** (0.0022) (0.25)Hong Kong 0.2585 -0.0269 5488

(0.0190)*** (0.0007)*** (0.25)South Africa 0.2797 0.0114 2506

(0.0264)*** (0.0131) (0.29)Israel 0.2854 -0.0457 690

(0.0383)*** (0.0140)*** (0.31)Netherlands 0.3592 0.0007 2823

(0.0284)*** (0.0059) (0.38)Sweden 0.3719 0.0034 2327

(0.0323)*** (0.0002)*** (0.41)

Austria 0.3740 -0.0106 504(0.0713)*** (0.0030)*** (0.42)

Singapore 0.3982 0.0082 3676(0.0227)*** (0.0003)*** (0.39)

Germany 0.4112 0.0016 4003(0.0222)*** (0.0017) (0.44)

Italy 0.4216 -0.0221 2711(0.0336)*** (0.0013)*** (0.43)

Norway 0.4341 0.0320 1540(0.0403)*** (0.0451) (0.46)

New Zealand 0.4586 0.0463 753(0.0727)*** (0.0208)** (0.53)

Malaysia 0.4697 -0.0643 5546(0.0195)*** (0.0514) (0.47)

Philippines 0.4723 -0.0159 1135(0.0418)*** (0.0098) (0.54)

Denmark 0.4785 -0.0494 1699(0.0417)*** (0.0541) (0.53)

Poland 0.4910 0.0212 477(0.0577)*** (0.0097)** (0.47)

Finland 0.5265 -0.0236 1726(0.0364)*** (0.0541) (0.61)

Country DEF Constant# Obser(se) (se) (R2)

France 0.5317 0.0049 7075(0.0193)*** (0.0131) (0.54)

Korea 0.5427 -0.0620 4534(0.0235)*** (0.0926) (0.61)

Switzerland 0.5685 0.0435 1834(0.0356)*** (0.0036)*** (0.56)

Thailand 0.5810 0.0079 3182(0.0265)*** (0.0114) (0.56)

Indonesia 0.6062 -0.1866 2076

(0.0315)*** (0.1330) (0.60)Greece 0.6125 -0.0042 213

(0.0643)*** (0.0142) (0.72)Belgium 0.6162 -0.0149 722

(0.0425)*** (0.0117) (0.65)Spain 0.6168 0.0070 971

(0.0554)*** (0.0030)** (0.51)Chile 0.6272 -0.0088 867

(0.0329)*** (0.0008)*** (0.65)Japan 0.6409 -0.0526 12097

(0.0158)*** (0.0287)* (0.65)China 0.6519 0.0236 2648

(0.0252)*** (0.0017)*** (0.67)

Pakistan 0.6908 -0.0015 770(0.0700)*** (0.0009) (0.66)Hungary 0.7009 -0.0385 209

(0.0847)*** (0.0350) (0.72)Portugal 0.7039 -0.0242 641

(0.0437)*** (0.0154) (0.70)India 0.7420 0.0030 2881

(0.0213)*** (0.0002)*** (0.77)Argentina 0.7507 0.0023 466

(0.0898)*** (0.0035) (0.70)Brazil 0.7631 0.0002 1582

(0.0332)*** (0.0002) (0.73)Turkey 0.7671 0.0226 633

(0.0440)*** (0.0043)*** (0.80)Colombia 0.7751 0.0092 282

(0.0565)*** (0.0167) (0.74)Peru 0.7817 0.0312 296

(0.0542)*** (0.0250) (0.77)Russia 0.8317 -0.1831 144

(0.0582)*** (0.0105)*** (0.84)Mexico 0.8391 -0.0072 1213

(0.0533)*** (0.0049) (0.64)


39/49

Table 4: Marginal Debt Issuance, Development, and Financial Development Measures

Pecking order regression with controls for development: dependent variable is D the net debt issue scaled by netassets.DEFis the total financing deficit scaled by net assets.InitialGDPis real per capital GDP in thousands of USdollars in 1980.Privo is private credit by deposit money banks and other financial institutions/GDP. StockMktCap isstock market capitalization/GDP. StockTO is stock market turnover ratio calculated as the total value traded divided

by market cap.BankbyStockis relative capitalization of private bank credits compared to stock market cap.Privo,StockMktCap, StockTO, andBankbyStockare averages over the period 1980-2004. Marketis a dummy variableprovided by Ross Levine (2001) to capture if the financial system is market based. CAIis the Capital Access Indexmeasuring the ability of new and existing businesses to access capital. The prefix I indicates the independentvariable listed afterwards is interacted withDEF. OLS regressions with robust standard errors. All regressionsinclude dummy variables for year and clustering for individual firms.

(1) (2) (3) (4) (5) (6)

DEF 0.3721 0.3289 0.4899 0.5115 1.7352 0.5571(0.0109)*** (0.0120)*** (0.0128)*** (0.0157)*** (0.0409)*** (0.0118)***

InitialGDP -0.0004 0.0000 -0.0002 -0.0002 0.0002 -0.0002(0.0000)*** (0.0001) (0.0000)*** (0.0000)*** (0.0001)*** (0.0000)***

I InitialGDP -0.0075 -0.0150 -0.0067 -0.0115 0.0050 -0.0039

(0.0005)*** (0.0008)*** (0.0005)*** (0.0006)*** (0.0007)*** (0.0006)***Privo -0.0066

(0.0013)***I Privo 0.1489

(0.0147)***StockMktCap -0.0046

(0.0009)***I StockMktCap -0.1272

(0.0088)***StockTO 0.0016

(0.0003)***I StockTO -0.0504

(0.0054)***CAI -0.0041

(0.0005)***I CAI -0.2073

(0.0066)***Market -0.0032

(0.0006)***I Market -0.2647

(0.0110)***Constant 0.0267 0.0282 0.0262 0.0223 0.0488 0.0295

(0.0112)** (0.0129)** (0.0102)** (0.0117)* (0.0128)*** (0.0124)**

Observations 268491 265843 268491 268491 268491 265013R-squared 0.25 0.25 0.26 0.25 0.27 0.26

*** indicates significance at the 0.01 level** indicates significance at the 0.05 level* indicates significance at the 0.10 level


40/49

Table 5: Marginal Debt Issuance and Legal Origin

Pecking order regression with controls for legal origin: dependent variable is D the net debt issue scaled by netassets.DEFis the total financing deficit scaled by net assets.InitialGDPis real per capital GDP in thousands of USdollars in 1980. Legal Origins are dummy variables with 1 indicating the country traces its legal origin to thatparticular legal system. Data is from Ross Levine (2001). Country origins for China, Hungary, Poland, are Russia

are supplemented from Siems (2006). The prefix I indicates the independent variable listed afterwards isinteracted withDEF. OLS regressions with robust standard errors. All regressions include dummy variables for yearand clustering for individual firms.

(1) (2) (3) (4)

DEF 0.5624 0.3141 0.3276 0.5513(0.0104)*** (0.0115)*** (0.0112)*** (0.0125)***

InitialGDP -0.0002 -0.0003 -0.0003 -0.0002(0.0000)*** (0.0000)*** (0.0000)*** (0.0000)***

I InitialGDP -0.0025 -0.0048 -0.0055 -0.0025(0.0006)*** (0.0006)*** (0.0005)*** (0.0006)***

UK Origin -0.0056 -0.0028

(0.0006)*** (0.0007)***I UK Origin -0.3020 -0.2919

(0.0094)*** (0.0120)***French Origin -0.0013

(0.0007)*I French Origin 0.2874

(0.0135)***German Origin 0.0074 0.0058

(0.0007)*** (0.0008)***I German Origin 0.2848 0.0267

(0.0121)*** (0.0154)*Constant 0.0280 0.0276 0.0263 0.0272

(0.0155)* (0.0119)** (0.0119)** (0.0144)*

Observations 268491 268491 268491 268491R-squared 0.26 0.26 0.26 0.26



41/49

41

Table 6: Marginal Debt Issuance and Shareholder Protection

Pecking order regression controlling for LLSV (1998) shareholder protection variables: dependent variable is D thenet debt issue scaled by net assets.DEFis the total financing deficit scaled by net assets.InitialGDPis real percapital GDP in thousands of US dollars in 1980. LLSV (1998) data is from Levine (2001). Data is not available forChina, Hungary, Poland, and Russia. The prefix I indicates the independent variable listed afterwards is

interacted withDEF. The prefix I indicates the independent variable listed afterwards is interacted withDEF.OLS regressions with robust standard errors. All regressions include dummy variables for year and clustering forindividual firms.

(1) (2) (3) (4) (5) (6)

DEF 0.5497 0.5099 0.2791 0.2784 0.6382 0.3425(0.0119)*** (0.0095)*** (0.0147)*** (0.0116)*** (0.0159)*** (0.0225)***

InitialGDP -0.0003 0.0001 -0.0003 -0.0002 -0.0002 -0.0002(0.0000)*** (0.0000)* (0.0000)*** (0.0000)*** (0.0000)*** (0.0001)***

I InitialGDP -0.0039 0.0015 -0.0076 -0.0033 -0.0003 -0.0075(0.0006)*** (0.0006)** (0.0005)*** (0.0006)*** (0.0006) (0.0008)***

Minor -0.0010(0.0006)

I Minor -0.2586(0.0105)***Proxy -0.0059

(0.0006)***I Proxy -0.3371

(0.0104)***Esmreq -0.0750

(0.0085)***I Esmreq 0.9931

(0.1127)***Reserve 0.0144

(0.0021)***I Reserve 1.0230

(0.0397)***Srights -0.0015

(0.0002)***I Srights -0.0852

(0.0038)***Crights 0.0004

(0.0004)I Crights 0.0370

(0.0056)***Constant 0.0272 0.0253 0.0342 0.0234 0.0275 0.0223

(0.0131)** (0.0172) (0.0115)*** (0.0141)* (0.0115)** (0.0116)**




42/49

42

Table 7: Marginal Debt Issuance and Country Governance

Pecking order regression controlling country governance variables: dependent variable is D the net debt issuescaled by net assets.DEFis the total financing deficit scaled by net assets.InitialGDPis real per capital GDP inthousands of US dollars in 1980.Log(Inflation) is the natural logarithm of (1 + long run inflation rate). Marketis adummy variable provided by Ross Levine (2001) to capture if the financial system is market based. CPIis the 2006

Corruption Perception Index Score provided by Transparency International.Bribery Survey is the 2006 BribePayers Index that measures the supply side of corruption the willingness of countries to bribe. The score isavailable for 19 mostly developed countries in this sample. Gov Effectiveness,Rule of Law, Voice/Accountability,andRegulatory Quality are different measures of governance performance provided by Governance ResearchIndicator Country Snapshot (GRICS). The variables are averages of their 1996, 1998, 2000, 2002, and 2004 surveys.The prefix I indicates the independent variable listed afterwards is interacted withDEF. OLS regressions withrobust standard errors. All regressions include dummy variables for year and clustering for individual firms.

(1) (2) (3) (4) (5) (6)

DEF 0.8196 1.4295 0.6551 0.6159 0.4782 0.6586(0.0158)*** (0.0475)*** (0.0102)*** (0.0096)*** (0.0099)*** (0.0103)***

InitialGDP -0.0001 0.0001 0.0001 -0.0000 -0.0003 0.0001(0.0001) (0.0001) (0.0001) (0.0001) (0.0001)*** (0.0001)

I InitialGDP -0.0078 -0.0008 -0.0025 -0.0014 -0.0021 -0.0034(0.0005)*** (0.0006) (0.0006)*** (0.0006)** (0.0007)*** (0.0006)***

CPI -0.0000(0.0003)

I CPI -0.0589(0.0024)***

Bribery Survey -0.0037(0.0009)***

I Bribery Survey -0.1663(0.0075)***

Gov Effectiveness -0.0023(0.0007)***

I Gov Effectiveness -0.2129

(0.0078)***Rule of Law -0.0009(0.0008)

I Rule of Law -0.2153(0.0083)***

Voice/Accountability 0.0017(0.0008)**

I Voice/Accountability -0.1660(0.0091)***

Regulatory Quality -0.0040(0.0008)***

I Regulatory Quality -0.2537(0.0092)***

Constant 0.0259 0.0530 0.0269 0.0269 0.0266 0.0269(0.0140)* (0.0165)*** (0.0136)** (0.0144)* (0.0139)* (0.0132)**




43/49


44/49

Table 9: Pecking Order Beta and GDP Growth

Panel A: In-sample CorrelationsPeckingBeta = Beta estimates by country from the basic pecking order regressions (high value implies low equity market effiper capita average growth in real GDP by country over the period 1980 to 2005;Netintm = Net Interest Margin; Stcap = StockSttrade = Stock Market Total Value Traded / GDP; Stturn = Stock Market Total Value Traded / Stock Market Capitalization;

deposit money banks and other financial institutions / GDP;Prbond= Private Bond Market Capitalization / GDP. Values are Significance at 5% level is indicated with bold numbers.

GDP

Growth InitialGDP PeckingBeta Prbond Netintm Stcap

InitialGDP -0.4350 1

PeckingBeta -0.0609 -0.4979 1

Prbond -0.0360 0.5935 -0.4007 1

Netintm -0.3501 -0.3079 0.5613 -0.3504 1

Stcap 0.0657 0.4083 -0.5826 0.2840 -0.3724 1

Sttrade 0.1556 0.4305 -0.5088 0.4544 -0.3943 0.8219

Stturn -0.2019 -0.1179 0.2394 -0.1912 0.2243 -0.0762

Privo 0.2497 0.6002 -0.6017 0.5597 -0.6234 0.7320

Panel B: Out-of-sample CorrelationsPeckingBeta = Beta estimates by country from the basic pecking order regressions for the period 1996 to 2000. GDP Growthother variables, except forInitialGDP96, are calculated as averages for the period 1996 to 2000. Significance at 10% level is

GDPGrowth

InitialGDP96

PeckingBeta Prbond Netintm Stcap

InitialGDP96 -0.6245 1

PeckingBeta 0.1224 -0.5336 1

Prbond -0.4153 0.6307 -0.3872 1

Netintm 0.1351 -0.3871 0.4087 -0.3172 1

Stcap -0.1396 0.4361 -0.4385 0.3083 -0.2543 1Sttrade -0.0884 0.4256 -0.3672 0.4374 -0.2792 0.7969

Stturn 0.0068 -0.1921 0.2427 -0.2198 0.2641 -0.1203

Privo -0.2579 0.5829 -0.3839 0.5318 -0.5513 0.6891


45/49

45

Panel C: In-sample Relations

The dependent variable is the long run average real GDP Growth for the period 1980 to 2005. All variables, except forInitialaveraged over the period 1980 to 2004. Robust standard errors are in parenthesis.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Ln(InitialGDP) -0.74 -0.76 -0.71 -0.66 -0.73 -0.55 -0.68 -0.91 -0.76 -0.77 (0.27)*** (0.29)** (0.22)*** (0.25)** (0.24)*** (0.22)** (0.15)*** (0.30)*** (0.23)*** (0.28)***

PeckingBeta -3.15 -2.92 -0.89 -2.46 (1.37)** (1.38)** (1.25) (1.27)*

Prbond 2.32 1.93(1.17)* (1.06)*

Netintm -30.28 -27.62(10.05)*** (11.28)**

Stcap 0.92 0.50(0.44)** (0.41)

Sttrade 1.96 (0.76)**

Stturn -0.16

(0.11) Privo 2.04

(0.54)*** Constant 10.21 8.22 9.67 7.37 7.83 7.40 6.51 11.20 10.41 9.82

(2.96)*** (2.38)*** (2.30)*** (2.16)*** (2.09)*** (2.17)*** (1.17)*** (2.98)*** (2.54)*** (2.95)***

Observations 42 41 42 42 42 42 41 41 42 42 R-squared 0.29 0.26 0.45 0.26 0.33 0.25 0.34 0.35 0.45 0.31



46/49

46

Panel D: Out-of-sample Relations

The dependent variable is the long run average real GDP Growth for the period 2001 to 2005. The independent variables, excPeckingBeta, are calculated as averages for the period 1996 to 2000. Robust standard errors are in parenthesis.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Ln(InitialGDP96) -1.08 -0.81 -0.93 -0.96 -0.99 -0.89 -0.78 -1.03 -1.10 -1.12 (0.28)*** (0.28)*** (0.26)*** (0.22)*** (0.24)*** (0.24)*** (0.15)*** (0.32)*** (0.30)*** (0.28)***

PeckingBeta -2.45 -2.97 -2.31 -2.22 (1.09)** (1.06)*** (1.10)** (1.17)*

Prbond -0.29 -0.52(0.77) (0.69)

Netintm -6.73 -3.19(7.90) (7.59)

Stcap 0.47 0.27(0.31) (0.33)

Sttrade 0.78 (0.45)*

Stturn -0.08

(0.09) Privo 0.38

(0.44) Constant 13.10 9.60 10.89 10.53 10.73 10.47 8.80 13.11 13.35 13.12

(2.98)*** (2.47)*** (2.68)*** (2.11)*** (2.17)*** (2.35)*** (1.41)*** (3.26)*** (3.26)*** (3.06)***

Observations 42 41 42 42 42 42 41 41 42 42 R-squared 0.45 0.38 0.40 0.41 0.43 0.40 0.35 0.47 0.45 0.46



47/49

Appendix A: Data Description and sources

The accounting data for this study were collected from Worldscope compact discs (CDs).

I merged one CD from each year starting in 1992 and ending in 2006. Each CD contains

historical data from ten (10) previous years. All duplicate observations have been very carefully

examined before removing from the dataset. In case of changes in financial reports, data from the

most recent CD have been taken. Merging these CDs, while challenging, is necessary to obtain

more than 10 years of data. Furthermore, the inactive companies are routinely purged from the

CDs to make room for data from new companies. My sample contains data from 1980 to 2005.

Country-specific indicators of development and other macroeconomic data have been

collected from six major sources:

1. Data from Beck, Demirg-Kunt and Levine (2000), later updated and provided in theWorld Bank website, contain various financial structure metrics including size, activity,

and efficiency of financial markets and intermediaries. The variables are reported in

annual terms during the period 1980 to 2005 for all the countries in my sample.

2. Demirg-Kunt and Levine (2001) provide the LLSV (1998) indicators of legalenvironment and shareholder protections. LLSV (1998) collected these variables from

national bankruptcy and reorganization laws. Data are available for all countries in this

sample, except for China, Hungary, Poland and Russia. This database also contains the

indicator variables for legal origin.

3. The data for macroeconomic controls in my regressions are obtained from 2006 WorldDevelopment Indicators. The variables are available for all countries in my sample and

are expressed in annual terms during the period 1980 to 2004.


48/49

48

4. The indicators of corruption the Corruption Perception Index (CPI) for 2006 and theBribe Payer Index (BPI) for 2006 are obtained from Transparency International (TI). The

CPI is available for all countries and the BPI is available for 27 of the mostly developed

countries in the sample.

5. The data on Capital Access Index (CAI) are collected from the publications of the MilkenInstitute. The variables used in this study are reported annually from 2000 to 2006.

6. Finally, a number of measures of a countrys governance quality such as governmenteffectiveness, rule of law, voice and accountability, regulatory quality, and control of

corruption are provided by Governance Research Indicator Country Snapshot (GRICS).

These variables are based on several hundred individual variables quantifying perceptions

of governance originally collected from 37 separate data sources from 31 different

organizations. The measures are available for all the countries in my sample for the years

1996, 1998, 2000, 2002, and 2004.

Appendix B: Bank-based vs. Market-based Systems

In development finance literature, there are various attempts to differentiate the growth

patterns of countries and their financial systems mainly to understand why we can observe

different levels of development in otherwise very similar economies. One such distinction is

based on whether the country emphasizes its banking system or its equity markets. In bank based

countries, the ratio of private credit provided by banks and other financial intermediaries to GDP

is generally larger than the ratio of equity market capitalization to GDP. Furthermore, the

proponents of the superiority of the market-based systems believe that markets reduce some of

the inherent inefficiencies associated with banks, such as rent extraction by powerful banks and


49/49

collusion of powerful banks with managers of firms in the absence of strong regulatory

environment. Rajan and Zingales (1995) explore some of the implications of the market-based

vs. bank-based systems on the total debt holding decisions of firms. They observe that bank

based countries have smaller financial markets. However, they do not find any significant

differences in the level of leverage between bank-based and market-based economies among G-7

countries. It would be interesting to test if there is a significant difference in the marginal debt

issuance behavior of firms or the pecking order beta estimates in my sample depending on

market-based or bank-based systems. If market-based systems do reduce any of the inefficiencies

in the financial markets, we would observe lower values of the pecking order beta estimates in

market-based countries.

The dummy variable Marketis obtained from Demirguc-Kunt and Levine (2001). They

calculate this variable from an index of financial market structure that compares the long term

size, activity, and efficiency of the banking sector and the equity market. A value of 1 indicates a

market based financial system and 0 indicates a bank based system. Interaction terms for

pecking order asy me try

Documents