Chapter – 1

Introduction

1.1 Background of the Study

The private corporate sector has emerged as one of the most important sectors
of any modern economy in the world. Its size, strength, influences and dominance
have grown significantly. In this context Galbraith (1985) has said “The modern

r
corporation has acquired tremendous power over the markets, community, beliefs and
the state. It has become a political entity, which is different in form but not in essence

pu
from the state itself. It has become the state within the state”. The governments of
ar t
both developed and developing countries have taken different corporate friendly
h
measures to develop their respective private corporate sector. In response to that

ag
policy measures the private corporate sector in each country has grown in terms of
Kh rig
number of companies and their paid-up capitals and it has significantly contributed to
the economic growth process through increased corporate savings, investments and
employment.
Therefore, the studies all over the world have continuously tried to analyze
y

certain issues which are responsible for enhancing the value of the companies. The
IIT p

various research issues in the area of corporate finance comprises of the financing
decision, investment decision and dividend policy. Among these different issues of
o

corporate finance, the research on financing decision or determination of capital

structure has gained the prime attention all over the world. The capital structure of a
C

company is defined as the ratio of debt to equity of that company. The modern
literature on corporate capital structure has started with the seminal work of
Modigliani and Miller (1958), which states that the capital structure does not affect the
value of the firm. The key Modigliani and Miller’s irrelevance proposition rests on
some simplifying assumptions like the existence of perfect market, tax free world, no
agency costs etc and that the firm’s capital structure can have an effect on the value of
the firm if these assumptions don’t hold or other complications are encountered. The
theories those developed after the Modigliani and Miller’s proposition can be
classified into two broad categories. One group of theories argue that firms would

1
 select the optimal or target debt ratio, which can maximize the value of the companies
and this can be achieved through the cost and benefit analysis of debt financing.
However, alternative theories remain plausible. These theories reject the notion of
timely convergence toward a target debt ratio and argue that firm’s financing decision
has always been driven by the firms’ historical profitability, investment opportunities
and market conditions.
The trade-off theory suggests that firms have a unique optimal capital structure
that balances between the tax advantage of debt financing and the costs of financial
distress and bankruptcy. In the latter phase studies have tried to estimate the speed at
which companies try to reach the optimal or target capital structure. The research

r
based on the estimation of speed of adjustment to target capital structure has been

pu
termed as dynamic trade-off theory. The pecking order theory formulated by Myers
and Majluf (1984) and Myers (1984) does not predict an optimal capital structure of
ar th
the company, which can maximize its value. Under the assumption of asymmetric
information problem in the market, i.e. when investors know less about the investment
ag
Kh rig
opportunity and value of the firm than incumbent managers, firms avoid issuing
external equity which is sensitive to mis-pricing and adverse selection. This theory
predicts a strict hierarchical corporate financing pattern in which new investments are
financed by internal funds first, then by low risk debt and hybrid securities such as
y

convertibles and external equity as the last resort. Therefore, there is no target capital
structure and the observed debt ratio is the cumulative outcome of the pecking order
IIT p

financing behaviour over time.

After a long debate on trade-off vs. pecking order theory of capital structure,
o

Fama and French (2005) conclude that it is probably time to stop running empirical
C

horse races between them as stand-alone stories for capital structure. Perhaps it is best
to regard the two models as stable mates with each having elements of truth that help
explain some aspects of financing decisions, (pp. 580–581). Barclay and Smith (2005)
also maintain that although the pecking order theory is incapable of explaining the full
array of financial policy choices, this does not mean that information costs are
unimportant in corporate decision-making. On the contrary, such costs will influence
corporate financing choices and, along with other costs and benefits, must be part of a
unified theory of corporate financial policy, (p. 16). Therefore, using the idea of
modified pecking order theory given by Myers (1984), a nested model has been

2
 specified which have comprised all the variables used to test the pecking order theory
and dynamic trade-off theory of capital structure simultaneously.
The market timing theory advocated by Baker and Wurgler (2002) argues that
the firms do not adjust their leverage ratios towards the target and it is the equity
market timing effects i.e., firms issue shares when considered to be overvalued and
firms repurchase own shares when considered to be undervalued gets accumulated and
finally drive the firm’s capital structure. Arguing that market-to-book ratio proxies for
stock market misevaluation, they have used the variable external finance weighted
average historical market-to-book ratio (EFWAMB) to capture the firms’ market
timing attempts. In the later phase the studies have tried to analyze the impact of

r
historical market to book ratio as a growth opportunity proxy on target capital

pu
structure in a partial adjustment framework (see, Flannery and Rangan 2006 and Liu,
2009). Welch (2004) proposes the inertia theory which argues that managerial inertia
ar th
permits stock price changes to have a prominent effect on the market valued debt
ratios. Therefore, observable debt ratios of firms vary closely with fluctuations in their
ag
Kh rig
own stock returns, and equity price shocks seem to have a long-lasting effect on
corporate capital structure. Therefore, the firms don’t adjust to target capital structure
and the capital structure is fully induced by change in stock price.
During the recent years most of the literature has found strong support of the
y

trade-off theory for explaining the leverage ratios of both developed and developing
countries, which maintains that market imperfections generate a link between leverage
IIT p

and firm value. This theory suggests that firms do have the target capital structure,
which balances the costs and benefits of leverage and factors like size, tangibility,
o

profitability, non-debt tax shield, liquidity, and industry median of the leverage are the
C

major factors which determine the optimal capital structure of a company. But the
observed capital structures frequently deviate from the target level because adjustment
costs are keeping firms away from reaching their target structure every year.
Therefore, firms take different steps to offset deviations from their optimal debt ratios.
The speed with which firms reverse deviations from their target debt ratios depends on
the costs of adjusting leverage. The speed of adjustment weighs rebalancing costs
against the costliness of deviating from the target. Slower adjustment speeds are
predicted in the presence of higher adjustment costs and faster adjustment speeds are
predicted when deviations are more costly. The broad factors like financial constraint,
external financing cost, financial distress, financial deficit or surplus, distance between
3
 the observed and optimal debt ratio affect the adjustment costs of the companies
which in turn affect the adjustment speed to the target capital structure.
It is also argued that corporate capital structure varies over time. Therefore,
business cycle or macroeconomic condition is also an important factor for explaining
the financing choice of the firm with other firm and industry specific variables. A
business cycle is defined as the periodic but irregular up and down movements in
economic activity which is normally measured by fluctuations in real gross domestic
product (GDP) and other macroeconomic variables like term spread and default
spread. The tax benefit of debt always depends upon the level of cash flows which
always depend on whether economy is an expansion or in a contraction and the

r
expected bankruptcy costs also depend on the probability of default and loss given

pu
default and both of them are highly influenced by the state of the economy. As a result
variations in macroeconomic conditions should induce variations in optimal leverage
ar th
(Hackbarth et al., 2006). Macroeconomic conditions affect the speed of adjustment to
target leverage as well because in the good macroeconomic condition the companies
ag
Kh rig
find it easier to raise capital from the market, but when the market conditions are not
favourable raising funds becomes quiet costly for the companies. Therefore, the speed
of adjustment to target leverage is higher when economic prospects are good.
Therefore, firms should adjust their capital structure faster in booms than in
y

recessions.
The research on the impact of business group affiliation on capital structure
IIT p

dynamics has gained a lot of attention in the recent years. Group affiliation is an
important firm ownership feature of the private corporate sector of many emerging
o

and developed economies. A business group is defined as an organisational structure

C

which constitutes a collection of firms that are legally independent bounded either by
formal or informal ties and are assumed to take coordinated actions (Khanna and
Rivkin, 2000). In general, these corporate groups are characterized by their diversified
operations, extensive commercial and personnel links, cross-equity shareholdings, and
mutual board representation. Business group affiliation has played a significant role
for financing decision of the group affiliated companies because the group help the
companies to raise the capital when the capital markets are imperfect or there is an
external market failure (Leff, 1976, 1978; Aoki, 1990; Berglof and Perotti, 1994 and
Weinstein and Yafeh, 1998), which means that groups play an intermediation role in
the inefficient capital markets (Lomnitz & Perez-Lizaur, 1987; Pan, 1991).
4
 The policy distortion theory of business group put forward by Ghemawat and
Khanna (1998) argues that business groups affiliated firms may be able to create
alternative tax shields for their members by engaging in internal business transactions
aimed at moving profits from one member firm to another and they can reduce risk
and create value through diversification at the group level. Groups may also reduce
the expected cost of default by providing loan guarantees to their members, thus using
the assets of one group member as collateral for another. It can also be argued that
companies that are part of a business group are expected to have better access to
capital markets than comparable stand-alone companies and therefore, they should be
able to adjust their capital structure more frequently. Dewenter and Warther (1998)

r
argued that group affiliation always reduces the asymmetric information problem for

pu
the group affiliated firms and therefore, reduces the relevance of pecking order
hypothesis. It is also found that for the group affiliated companies, there is the
ar th
existence of internal or virtual capital market and this market serves as an alternative
source of finance for the group affiliated firms.
ag
Kh rig
1.2 Rationale behind the Study
In the wake of new economic policy in 1991 a series of economic reforms
have taken place in India and many of them either directly or indirectly lead to the
substantial liberalization of Indian corporate sector. All these reforms included the
y

abolition of the Industries Development and Regulation Act (1951), delicensing of

IIT p

industries, public sector dereservation and easing of Monopolies and Restrictive Trade
Practices Act (MRTP), import delicensing, reduction in import tariffs and
o

decanalization of imports etc. These reforms increased the foreign investment

opportunities in India and the Indian companies were able to import new technologies
C

and capital goods, allowed to expand their productive capacity, and introduce new
products without obtaining industrial licenses. Indian companies were also allowed to
enter into joint ventures with multinational companies freely. Therefore, it is very
much imperative to know how the Indian companies have grown over the period of
liberalization and how they have contributed for economic development.
The other reform measures implemented such as opening up of the capital
market with the abolition of the Capital Issues Control Act 1947 led to free pricing of
capital issues and made equity financing an attractive source of finance for the
corporate sector. Access to foreign funds have also increased with the liberalization of

5
 External Commercial Borrowings (ECB) policy and allowing the instruments like
American Depository Receipts (ADRs), Global Depository Receipts (GDRs) and
Foreign Currency Convertible Bonds (FCCBs) for investments. The Foreign Direct
Investment (FDI) and Foreign Institutional Investment (FII) policies have changed
frequently to make the Indian market foreign investors friendly and this has helped to
access the foreign technology and foreign capitals for investments. Financial sector
reforms also aimed at strengthening Public Sector Banks through recapitalisation and
implementation of prudential norms. Competition has been fostered through the entry
of private sector banks and freeing of deposit and lending rates of banks. Banks have
been given greater freedom in their operations through the abolition of the

r
requirement of the credit authorisation by the central bank, compulsory consortium

pu
lending by banks and the centrally imposed “maximum permissible bank finance”
system. Therefore, the importance of bank as a debt-financing source has increased.
ar th
Further concessional funds have been withdrawn from the development financial
institutions and their lending rates have been freed. Auctions for government
ag
Kh rig
securities have been introduced in a move to make government borrowings at market
related interest rates and pre-emption in the form of Cash Reserve Ratio (CRR) and
Statutory Liquidity Ratio (SLR) have been reduced. Further the use of money market
instruments such as Commercial Paper (CP), Certificate of Deposits (CDs) and Inter
y

Corporate Deposits (ICDs) have been encouraged. These instruments have enlarged
the corporate sectors’ access to funds from the financial system and enabled it to
IIT p

choose between various sources of funds being priced competitively. In this context it
is also very much important to know how the financing patterns of the Indian
o

companies have changed during the period of liberalization due to availability of

C

various alternative sources of finance.

In India most of the studies are based on the testing of the trade-off and
pecking order theories of capital structure, but research on modified pecking order,
market timing theory and inertia theories of capital structure has been very limited.
Therefore, it is imperative to test these theories in the context of Indian companies. It
is important to find out the most suitable capital structure theory for Indian enterprises
because in the changing scenario, identification of the suitable theory may guide the
managers to consider certain important factors which have the greater influence on the
leverage ratio, and therefore, it will be easier for them to optimize their financing
decisions.
6
 Economic cycles have become the characteristic features of market-oriented
economies either in the form of alternating expansions and contractions that
characterise a classical business cycle, or in the form of alternating speedups and
slowdowns that mark cycles in growth. With the progress of the liberalisation process
Indian economy has been transformed into more of a market-driven economy, and
therefore, change in the business cycle has become one of the prominent features
which affect the economic decision process. The studies have shown that business
cycle indicators like growth rate of real GDP and term spread etc have been
fluctuating during the period of liberalization. As we know business cycle or the
macroeconomic condition plays an important role in determining the capital structure,

r
in India, studies on the role of macroeconomic conditions on determination of capital

pu
structure and the speed of adjustment to target capital structure are not available.
Therefore, a study on role on business cycle or macroeconomic condition on capital
ar th
structure is very much required in the context of Indian corporate sector.
In the various studies it is argued that business group affiliation has the
ag
Kh rig
implications for determination of corporate capital structure. For Indian corporate
sector, the importance of group affiliation has been quite relevant due to the certain
specific characteristics such as (i) Indian corporate sector has more than 400 business
groups which not only vary in terms of size but also in terms of diversification, (ii) in
y

India normally the business group membership and affiliation are both same unlike in
other countries like Chile, Spain, Belgium where firms may have group affiliation to a
IIT p

particular business group but may have multiple memberships, and (iii) business
groups in India are representative of business groups in many emerging markets in at
o

least two ways i.e. they are often linked to a particular family and control is typically
C

achieved either through appointing family members and friends to directorship and top
managerial positions, or through direct and indirect ownership. Most of the studies
available on business group focus on the advantages of group affiliation of a firm but
the availability of studies on the impact of group affiliation on financing pattern have
been very few throughout the globe including India. The limited studies available on
business group and capital structure have focused only on the role of business group
on determination of capital structure and testing the trade-off and pecking order
theories of capital structure. Some of the previous studies on capital structure in India
have also found that there is the existence of optimal capital structure for Indian
companies and the companies want to reach that optimal with a reasonable adjustment
7
 speed (see Mahakud and Bhole, 2003; Mahakud and Mishra, 2010, and Mukherjee
and Mahakud, 2010). But none of the studies have tried to determine the role of group
affiliation on determination of adjustment speed to target or optimal capital structure.
Therefore, a fresh study is required to analyze the role of group affiliation on
determination of adjustment speed to target capital structure.

1.3 Scope and Objectives

Looking at the rationale behind the study, there are many questions or issues
that needs to be addressed in the context of the corporate sector in India. For example
whether the new economic policies have led to the economic growth and development
of the corporate sector? Is the opening up of the economy and new policy measures

r
lead to change in financing pattern of the Indian companies? Which is the most

pu
suitable theory of capital structure to explain the financing pattern of the companies in
ar t
India? How the business cycle changes affect the financing pattern of Indian
h
companies? Is there any role of ownership pattern in the form of group affiliation on

ag
the capital structure dynamics of the companies in India?
Kh rig
To address all these questions, the objectives of the study are as follows:
1. To study the growth, significance and financing pattern of the private corporate
sector in India across the different types of companies such as public limited, private
limited and foreign direct investment companies using the aggregate data.
y

2. To analyse the trends in the leverage ratios and the financing pattern of the select
IIT p

891 manufacturing companies.

3. To test the different theories of capital structure such as dynamic trade-off theory,
o

pecking order theory, modified pecking order theory, market timing theory and the
inertia theory and to determine the role of firm specific proxies for financial
C

constraints and external financing cost on determination of adjustment speed to target

capital structure.
4. To analyse the financing pattern of companies across different macroeconomic
conditions.
5. To study the role of the macroeconomic condition on the determination of capital
structure and adjustment speed to target capital structure.
6. To analyse the financing pattern of the standalone and business group companies.
7. To determine the role of business group affiliation on determination of capital
structure and adjustment speed to target capital structure.

8
 1.4 Data
The financing pattern of Indian private corporate sector has been studied in respect
of respective samples of 1501 to 2131 Public Limited Companies (PULCos), 701 to
1113 Private Limited Companies (PRLCos), and 208 to 537 Foreign Companies
(FRCos) during the period of liberalization i.e. 1993-94 to 2007-08. The changing
sample size of the companies would not distort our findings because all our variables
are in ratio form. The major sources of data are the Reserve Bank of India (RBI)
monthly bulletins and Handbook of Statistics; Government of India Economic Survey
and Ministry of Corporate Affairs Annual Reports.
For the econometric analysis, our sample targets all the companies available in the

r
PROWESS data base maintained by Centre for Monitoring the Indian Economy

pu
(CMIE). However, we have made several adjustments because of data constraints and
ar t
other specific fundamental reasons. Following Rajan and Zingales (1995), we have
h
excluded all financial firms because their financing policies are determined by many

ag
exogenous factors. Since leverage ratio is one of the significant concerns for
Kh rig
manufacturing companies, this study has emphasized on that specific companies. We
have selected those companies which have continuous data for the period 1992-93 to
2007-08, the period of liberalization in India in order to have a balanced panel. While
calculating the certain variables we have lost one year data for all the companies.
y

Therefore, the period of analysis of the study has been from 1993-94 to 2007-08. We
IIT p

have found 891 manufacturing companies which have the continuous data for the
above-mentioned period. All these companies belong to nine sub-industry groups
o

within the manufacturing industry. The shares of these sub industries to the total
number of select companies have been as follows: food and beverages (8.98 percent),
C

textiles (6.96 percent), chemicals (29.4 percent), non-metallic mineral products (8.31
percent), metal and metal products (11.34 percent), machinery (20.99 percent),
transport (8.41 percent), miscellaneous manufacturing (4.94 percent), diversified (0.67
percent).
Further the total sample has been divided into two sub samples on the basis of
macroeconomic conditions to study the role of macroeconomic condition on financing
pattern of the companies. We have defined the macroeconomic conditions on the basis
of growth rate of gross domestic product (GDP) and term spread, where term spread
has been measured as the difference between the yields of fifteen years and one year

9
 government bond. The years which have values of growth rate of GDP greater than
the mean value are considered as the good years and the years having the values less
than the mean are termed as the bad years. For term spread also, we have followed the
same criteria. To find out the impact of business group affiliation on capital structure
dynamics we have again divided the whole sample into two types such as group
affiliated companies which belong to certain business groups and standalone
companies, which are not affiliated to any business group in India. Out of total 891
companies 398 companies were identified as group affiliated and 493 as standalone
companies.

1.5 Methodology

r
Over the years the literature on financing behaviour has seen the major shift

pu
from macro modelling towards the micro modelling of financing. For the micro
ar t
modelling, the use of Panel Data models has the major significance, which take care
h
of the various econometric problems associated with the other models like time series

ag
and cross section models. Therefore, various static and dynamic panel data models
Kh rig
have been used for our analysis. For static analysis, pooled data and fixed effect
models have been used and for the dynamic analysis, we have used the Generalized
Method of Moments (GMM) technique as suggested by Arellano and Bond (1991).
1.5.1 Panel Data Models
y

A panel data set is one that follows a given sample of individuals over time,
IIT p

and thus provides multiple observations on each individual in the sample. Panel data
sets for economic research possess several major advantages over conventional cross-
o

sectional or time-series data sets. Hsiao (1986) has pointed out that “by using
information on both the intertemporal dynamics and the individualities of entities
C

being investigated, one is better able to control in a more natural way for the effects of
missing or unobserved variables. Longitudinal data allow a researcher to analyze a
number of important economic questions that cannot be addressed using cross-
sectional or time-series data sets alone”. Chirinko, Fazzari, and Meyer (1999) have
noted that sometimes the studies at aggregate level fail to find an economically
significant relationship between the two variables due to problems of simultaneity or
firm heterogeneity which may be better addressed at micro level data.
Other advantages of panel data set are outlined as follows: Panel data control
the individual heterogeneity. Therefore, the risk of obtaining biased results comes

10
 down. Klevemarken (1989) and Solon (1989) have argued that panel data gives more
information, more variability, less collinearity among the variables, more degrees of
freedom, and more efficiency. It is sometimes argued that cross-section data reflect
short-run behaviour, while time series-data emphasize long-run effects. By combining
the two sorts of information, one can study the dynamics of economic behaviour at an
individual level. When dynamic models are estimated using time series or cross-
sectional data, the usual least squares methods do not lead to consistent estimates for
the parameters of the model as the disturbance terms are serially correlated in these
models which cause the lagged endogenous variable to be correlated with those
disturbances. Freeman (1984) has argued that the panel data set are better able to

r
identify and measure effects that are simply not detectable in pure cross-section or

pu
time-series data set. Panel data models allow us to construct and test more
complicated behavioural models than purely cross-section and time-series data. In
ar th
other words it can be said the technical efficiency of the economic behaviour is better
studied and modelled with panel data (Baltagi and Griffin, 1988; Cornwell, Schmidt
ag
Kh rig
and Sickles, 1990; Kumbhakar, 1991, 1992). Odedokum (1996) argued that panel data
estimation yields more robust effects of independent variables on dependent variables
than time-series estimation.
Including various benefits of the use of panel data set there are some
y

limitations also. These limitations of panel data include the design and data collection
problems (Kasprzyk et al., 1989 and Bailar, 1989). The related limitation is the
IIT p

distortions of measurement errors which may arise because of faulty responses due to
the unclear questions, memory errors, deliberate distortion of responses, inappropriate
o

informants, misreporting of responses, and interviewer effects (Kalton, Kasprzyk and

C

McMillen, 1989). Other limitations are selectivity problems which include self
selectivity (Hausman and Wise, 1979), non response, and attrition (Bjorklund, 1989
and Ridder, 1990; Ridder et al., 1990).
In a panel data set, a given sample of N individuals is observed at different
time periods and thus provides multiple observations on each individual in the sample.
Here we assume a linear model,
Yi,t =α + X/it β + uit ------------------------------------------------------------------------- (1.1)
Where, i stands for ith cross sectional unit and t for the tth time period. α = intercept
term, u= error term and .

11
 If each cross sectional unit has the same number of time series observations
then such a panel is called a balanced panel. If the number of observations differs
among panel members, we call such a panel an unbalanced panel. In this thesis we
have dealt with the balanced panel. Estimation of 1.1 depends on the assumptions we
make about the intercept, slope coefficients and error term. If we assume that the
intercept and slope coefficients are constant across time and individual and the error
term captures differences over time and individuals, then that model is called as
simple pooled data model. The simplest and possibly naïve approach is to disregard
individual and time dimensions of the pooled data and just estimate the usual ordinary
least square regression. If we assume that the slope coefficients are constant but the

r
intercept varies over individuals then that model is defined as fixed effects model

pu
(Gujarati, 2004). On the other hand if we assume that individual heterogeneity is
random rather than systematic in the model then this model is called as random effects
ar th
model. Both the fixed and random effects models are discussed below.
1.5.2 Fixed Effects Model
ag
Kh rig
The fixed effects model is simply a linear regression model in which the intercept
terms vary over the individual units i , i.e.
yit   i  xit   uit , uit IID ( ,  u2 ), ………………………………………….(1.2)

where it is usually assumed that all xit are independent of all  it .

y

Writing (1.2) in vector form, we have

IIT p

 y1   e   0  0  x1   u1 
   0  e  0 x   
Y       1     2       N   2      ,……………………….(1.3)
o

            
           
 y N  0  0  e  xN  u N 
C

where
 yi1   x1i1 x2i1  xki1 
y  x x2i 2  xki 2 
yi   i2 
, X i   1i 2 ,
T 1    T k    
   
 yiT   x1iT x2iT xkiT 

e '  1,1,  ,1, ui'   ui1 ,  , uiT  ,

1T 1T

Eui  0, Eu u   u2 IT , Eui u 'j  0 if i  j,

'
i i

IT denotes the T  T identity matrix.

12
 Given the assumed properties of uit , we know that the ordinary least squares (OLS)

estimator of (1.3) is the best linear unbiased estimator (BLUE). The OLS estimators of i
and  are obtained by minimizing
N N
S   ui' ui   ( yi  e i  X i  ) '( yi  e i  X i  ) ……………………………… (1.4)
i 1 i 1

Taking partial derivatives of S with respect to  i* and setting them equal to zero, we have
^ _ _
 i  yi   ' xi i  1, , N , ……………………………………………………… (1.5)
where

r
_
1 T _
1 T
yi   yit , xi   xit ,

pu
T i 1 T i 1
Substituting (1.5) into (1.4) and taking the partial derivative of S with respect to  , we
ar t
have
h
1
^ N T  _ _ ' N T  _ _ '
   
 FE
ag     xit  xi  xit  xi      xit  xi  yit  yi   ………………… (1.6)
Kh rig
 i 1 i 1      i 1 i 1    
^
Where  FE = the fixed effects estimator
The computational procedure for estimating the slope parameters in this model does not
y

require the dummy variables for the individual (and/or time) effects actually be included
in the matrix of explanatory variables. We need only find the means of tome-series
IIT p

observations separately for each cross-sectional unit, transform the observed variables by
subtracting out the appropriate time-series means, and then apply the least-squares method
o

to the transformed data. Hence, we need only invert a matrix of order K  K .

The foregoing procedure is equivalent to premultiplying the ith equation
C

yi  e i  X i   ui ………………………………………………………………. (1.7)
by T  T idempotent ( covariance) transformation matrix
1
Q  IT  ee ' …………………………………………………………………….. (1.8)
T
Qyi  Qe i  QX i   Qui
………………………………………………….. (1.9)
 QX i   Qui , i  1,...., N .
Applying the OLS procedure to (1.9), we have
1
^ N  N 
 FE    X i'QX i    X i'Qyi  , ………………………………………………... (1.10)
 i 1   i 1 

13
 which is identically equal to (1.6). Because (1.2) is also called the analysis-of-
covariance model, the least-squares dummy variable (LSDV) estimator of  is
sometimes called the covariance estimator. It is also called the within-group estimator,
because only the variance within each group is utilized in formation this estimator.
1.5.3 Random Effects Model
It is commonly assumed in regression analysis that all factors that affect the
dependent variable, but that have not been included as regressors, can be appropriately
summarized by a random error term. In our case, this leads to the assumption that the
i are random factors, independently and identically distributed over individuals.

r
Thus we write the random effects model as
IID (0, 2 );  i IID (0, 2 ), …………………… (1.11)

pu
yit    xit    i   it ,  it

Where  i   it is treated as an error term consisting of two components: an individual

ar t
specific component, which does not vary over time, and a remainder component, which is
h
assumed to be uncorrelated over time. That is, all correlation of the error terms over time
ag
Kh rig
is attributed to the individual effects i . It is assumed that i and  it are mutually

independent and independent of x js (for all j and s ). This implies that the OLS estimator

for  and  from (equation A) is unbiased and consistent. The error components

structure implies that the composite error term  i   it exhibits a particular form of
y

autocorrelation (unless  2  0). Consequently, routinely computed standard errors for

IIT p

the OLS estimator are incorrect and a more efficient (GLS) estimator can be obtained by
exploiting the structure of the error covariance matrix.
o

To derive the GLS estimator, first note that for individual i all error terms can be
stacked as  iT   t , where T  (1,1,....,1) of dimension T and  i  ( i1 ,....,  iT ). The
C

covariance matrix of this vector is;

V { iT   i }     2TT   2 IT , ..………………………...……………. (1.12)

Where IT is the T - dimensional identity matrix. This can be used to derive the
generalized least squares(GLS) estimator for the parameters in (equation A). For each
individual, we can transform the data by premultiplying the vectors yi  ( yi1 ,........, yiT )

etc. by 1 , which is given by

1 2   2 
    IT  2 T T
  ………………………………………………(1.13)
    T  2 

14
 Which can also be written as

 1  1 
 1   2  IT  TT    TT  ,
 T  T 
Where

 2
 .
 2  T 2

Noting that IT  1 T   
T T transforms the data in derivations from individual means and

 1T   takes individual means, the GLS estimator for

T T  can be written as
1
 N T N
   T  x  x  x  x  

r
ˆGLS    it  x  xi  xit  xi   it i it i 
 i 1 t 1 i 1  ……………….. (1.14)

pu
N T N
 
    xit  xi  yit  yi    T   xit  xi  yi  yi   ,
 i 1 t 1
ar t i 1 
Where x  (1 ( NT )) i ,t xit denotes the overall average of xit . It is easy to see that for
h
ag
  0 the fixed effects estimator arises. Because   0 if T  , It follows that the
Kh rig
fixed and random effects estimators are equivalent for large T .

1.5.3 Hypothesis Testing with Static Panel Data

y

Before reporting the results obtained by estimating the above-mentioned static

panel data models we present the various hypothesis testing results to know the
IIT p

significance of the unobservable individual firm specific effects in the data set, and to
find out a suitable panel data method for the estimation of the static model. Likelihood
o

Ratio (LR) test (Gourieroux, Holly and Monfort, 1982) has been carried out to identify
C

the existence of individual firm specific effects in the data set. Lagrange Multiplier
(LM) test (Breusch and Pagan, 1980) has been used to test the acceptability of panel
data models over classical regression models. The LM test statistic follows a chi-
square distribution. Hausman specification test (Hausman,1978) has been carried out
to choose a suitable panel data model (fixed effects model vs. random effects model).
Hausman test statistics also follows a chi-square distribution. For all static panel data
estimations, we present (i) the value of adjusted R2, (ii) the F-test statistics, (iii)
Durbin-Watson statistics.

15
 1.5.5 Dynamic Panel Data Model
Many economic models suggest that the current behaviour depends upon past behaviour
(persistence, habit formation, partial adjustment etc.), so in many cases we would like to
estimate the dynamic models on an individual level. The linear dynamic panel data
model can be specified as
yit  xit'    yi ,t 1   i   it -------------------------------------------------------- (1.15)

Where it is assumed that it is IID (0, 2it).

When the dynamic model as specified in equation (A) is used, the lagged
dependent regressor is correlated with the error term even if the errors themselves are

r
uncorrelated. This renders fixed effect estimators to be biased and inconsistent (for fixed
T). The random effect estimator is also biased in this context. Simpler instrumental

pu
variable estimators may be used but if the errors are correlated then generalized method
ar t
of moments (GMM) technique is more appropriate.
h
For a model without the exogenous variables, Arellano and Bond (1991)

ag
suggest that the list of instruments can be extended by exploiting the additional
Kh rig
moment conditions and letting their number vary with t . For example, when T  4 , we
have
E (i 2  i1 ) yi 0   0
y

as the moment condition for t  2 . For t  3 , we have

E (i 3  i 2 ) yi1  0
IIT p

but it also holds that

o

E (i 3  i 2 ) yi 0   0

For period t  4 , we have three moment conditions and three valid instruments
C

E (i 4  i 3 ) yi 0   0

E (i 4  i 3 ) yi1  0

E (i 4  i 3 ) yi 2   0

All these moment conditions can be exploited in a GMM framework. To introduce the
GMM estimator define for general sample size T ,
 i 2  i1 
 
i    
  
 i ,T i ,T 1 

16
 as the vector of transformed error terms, and
  yi 0  0  0 
 
0  yi0 , yi1  0
Zi   
   0 
 
 0  0  yi 0 ,..., yi ,T  2  

as the matrix of instruments. Each row in the matrix Zi contains the instruments that
are valid for a given period. Consequently, the set of all moment conditions can be
written concisely as
E Zi' i   0.....................................................(1.16)

r
Note that these are 1  2  3    T  1 conditions. To derive the GMM estimator,

pu
write this as

ar t E Zi' (yi  yi ,1  0

h
Because the number of moment conditions will typically exceed the number of
unknown coefficients, we estimate  by minimizing a quadratic expression in terms of
ag
Kh rig
the corresponding sample moments, that is
N ' N
1 '  1 ' 
min 

N

i 1
Z ( yi  yi , 1 )  WN 
i
 N
 Z (y  y
i 1
i i i , 1 )
 -------------------------- (1.17)

Where WN is a symmetric positive definite weighting matrix. Differentiating this with

y

respect to  and solving for  gives

IIT p

1
^  N   N   N   N 
 GMM     yi', 1Z i  WN   Z i' yi , 1      yi' ,1Zi  WN   Z i' yi , 1 
  i 1   i 1    i 1   i 1  ……… (1.18)
o

The properties of this estimator depend upon the choice for WN , although it is
C

consistent as long as WN is positive definite, for example, for WN  I , the identity

matrix.
The optimal weighting matrix is the one that gives the most efficient estimator, i.e.
^
that gives the smallest asymptotic covariance matrix for  GMM . From the general

theory of GMM we know that the optimal weighting matrix is (asymptotically)

proportional to the inverse of the covariance matrix of the sample moments. In this
case, this means that the optimal weighting matrix should satisfy
1 1
p lim WN  V Z i' i   E Z i' i i' Z i 
N  ……………………………… (1.19)

17
 In the standard case where no restriction are imposed upon the covariance
matrix of i , this can be estimated using a first-step consistent estimator of  and

replacing the expectation operator by a sample average. This gives

1
^ opt 1 N ^ ^' 
'
WN   Z  i  i Z i 
i
N i 1  ………………………………….. (1.20)
^
where i is a residual vector from a first-step consistent estimator, for example using
WN  I .
If the models contain exogenous variables, depending upon the assumptions
made about xit , different sets of additional instruments can be constructed. If the xit

r
are strictly exogenous in the sense that they are uncorrelated with any of the  is error

pu
terms, we also have that
ar th E  xis  it   0 for each s,t,

so that xi1 ,..., xiT can be added to the instruments list for the first-differenced equation
ag
Kh rig
in each period. This would make the number of rows in Z i quite large. Instead, almost

the same level of information may be retained when the first-differenced xit s are used

as their own instruments. In this case, we impose the moment conditions

y

E xit  it   0 for each t

and the instrument matrix can be written as

IIT p

  yi 0 , xi' 2  0  0 
o

  
 0 '
 yi 0 , yi1 , xi 3  0 
Zi    -------------- (1.21)
C

   0 
 
 0  0  yi 0 ,..., yi ,T  2 , xiT'  
 
If the xit variables are not strictly exogenous but predetermined, in which case current

and lagged xit s are uncorrelated with current error terms, we only have that

E  xit  is   0 for s  t. In this case, only xi ,t 1 ,...., xi1 are valid instruments for the first-

differenced equation in period t. Thus, the moment conditions that can be imposed are
E  xi ,t  j  it   0 for j  1,..., t  1 (for each t). To examine the validity of the

instruments used in the analysis we have used the Sargan (1958) test. The Sargan test

18
 statistics follows the chi-square distribution. We have also reported the Walt test and
Z2 statistics to test the relevance of the variables used for the analysis and to test the
autocorrelation problem.

1.6 Organisation of the Study

This study is organized as follows. The review of literature on growth,
significance, and financing pattern of corporate sector, theories of capital structure,
determinants of capital structure, role of macroeconomic condition and business group
affiliation on determination of capital structure and factors affecting the adjustment
speed to target capital structure has been provided in the Chapter-2. In Chapter-3, an

r
attempt has been made to study the growth, significance and financing pattern of the

pu
private corporate sector in India across the different types of companies such as public
ar t
limited, private limited and foreign direct investment companies using the aggregate
h
data. The different direct and indirect policy measures taken to develop the corporate

ag
sector have been discussed in details in this chapter. The Chapter-4 discusses the
Kh rig
trends in the leverage ratios and the financing pattern of the selected 891
manufacturing companies of India. We have used various static and dynamic panel
data models to test the trade-off, pecking order, modified pecking order, market timing
and inertia theories of capital structure. We have also carried out the Wald test to
y

identify the best suitable theory of capital structure which can better explain the
IIT p

financing pattern of Indian companies. An attempt has also been made to study the
factors affecting the adjustment speed to target capital structure of the manufacturing
o

companies in India.
C

In Chapter-5 the trends in the financing pattern of select companies have been
analyzed across different macroeconomic conditions. Different econometric models
have been specified and estimated to identify the role of the macroeconomic condition
on the determination of capital structure and adjustment speed to target capital
structure. Chow test also has been carried out to confirm the difference of the
financing pattern of the companies in good and bad conditions of the economy. In
Chapter-6, the trends in leverage and the financing pattern of the standalone and group
affiliated firms have been analyzed. Both static and dynamic panel data models have
been specified and estimated to determine the role of business group affiliation on

19
 determination of capital structure and adjustment speed to target capital structure.
Chow test also has been carried out to differentiate between the financing pattern of
business group affiliated and standalone companies. Chapter 7 provides the major
findings and contributions of the study, policy recommendations, limitations of the
study and the scope for further research.

r
pu
ar th
ag
Kh rig
y
IIT po
C

20