Size, Age & Firm Growth: The

∗
Computer Industry in India

Vinod Mishra#

Vinod.Mishra@BusEco.monash.edu.au

Abstract
This paper analyses the growth trends in the Indian Computer Industry for the period

1991-2002. We focus on the significance of size and age on growth of firms and

whether or not the law of diminishing returns to learning holds for a high tech

industry like computers. An important difference between the computer software and

hardware industry is also taken into account: The software industry is more service

oriented, mainly targeting exports whereas hardware is a goods oriented industry,

targeting the home market. Hence we analyse the hardware and software industry

separately. It is found that the law of diminishing returns to learning does not hold for

the software industry, but does hold for the hardware industry. Current size also

negatively affects firm growth and more diversified hardware firms are more

successful in the Indian context.

JEL Codes: L11, L86, C33

Keywords: Computer Industry, Hardware, Software, India, Panel data, Size, Age,
Firm growth

∗
The author wants to thank Dr. Kausik Chaudhuri and Prof. Russell Smyth for valuable comments and
suggestions at various stages of this work.
#
Room No. H4.42, Department of Economics, Monash University. 900 Dandenong Road, Caulfield
East, VIC. 3145, Australia
1. Introduction

The Computer Industry is a rapidly expanding industry in India. It has shown

remarkable growth in a very short span of around 10 years, in spite of frequent and

abrupt shocks due to world market fluctuations. The big software slowdown, which

shook the world in the years 2001 and 2002, also unsettled the Indian software

industry. Many upcoming firms could not sustain this shock and were forced to wind

up. Global upheaval is not the only reason for the Indian disaster. In the Indian

market an important reason is the inter industry structure. Most of the Indian

computer hardware firms are importers of machine parts and intermediate goods (such

as motherboards, processors, memory and hard-disks) and they assemble and sell the

machine systems under a brand name in the Indian local market. The software firms,

on the other hand, rely mostly on export markets. While the domestic market has been

flourishing due to rapid computerization, it is catered for by ready-made software

manufacturers (usually large multinationals (MNCs) such as Sun, or Microsoft).

There is no substantial demand for specifically designed application software In India.

This paper empirically studies the impact of firm’s size and age on the firm’s growth

prospects for Computer Industry in India. We have also attempted to identify the role

of various factors such as product diversification, R&D expenditure, market

penetration, ownership effects and the level of export orientation in determining a

firm's growth in this industry. We conclude with a comparison between the software

and hardware industries in terms of their growth patterns and the factors affecting

each of them.

2
The first part of this paper models the relationship between a firm's growth and its age

and size. The second part of the paper expands the basic model by including the

number of products manufactured and firm ownership as two more relevant variables

for analysing growth. The third part further broadens the analysis by including some

other variables of interest such as net exports, R&D expenditure, market penetration

and product diversification. Finally interpretations of the various findings of the

model are offered and the main conclusions summarized.

The computer industry has features of both the manufacturing and services sector. The

computer hardware industry is more like a manufacturing industry whereas the

software industry has characteristics of the services sector. The computer hardware

industry requires huge initial investment - sunk costs in setting up a plant and a large

number of employees - whereas the computer software industry does not require large

start-up costs or a large base of employees but it requires maintenance and update and

support services for the software produced. One reason for examining the computer

industry in the current study is the level of intra-industry variations exhibited by this

industry. The other reason is the phenomenal growth rate of this industry in India over

the last two decades. Hence an interesting question is whether the laws governing firm

dynamics in the manufacturing sector are also applicable in a high-growth, high – tech

industry. To this point such industries have largely been ignored in the literature

pertaining to firm growth.

2. The Existing Literature

The relationship between firm growth and firm size has been an issue in the

theoretical as well as the empirical literature on firm growth. According to Gibrat’s

3
law firm growth is independent of firm size. But empirical studies suggest that

Gibrat’s law does not hold when applied to real data. Empirical scrutiny of Gibrat’s

law began as early as 1956 by Hart and Prais (1956) and their results did not support

the law. Singh and Whittington (1975) also tested Gibrat’s’ law for firms in the

manufacturing, construction and distribution sectors in the UK and found that the

growth of the firm decreases with size.

Evans (1987a) explored the relationship between firm growth, firm size and the firm’s

age for a panel of US manufacturing firms for the time period of 1976-1982. He found

that firm growth decreases with firm age, which is consistent with Jovanovic’s (1982)

theory of firm growth. According to Jovanovic’s (1982) theory, firms uncover their

true efficiencies, over time, with a Bayesian learning process.. But Evans (1987a)

other finding was that that firm growth decreases with size, which is not consistent

with Gibrat’s law. Moreover Evans also concluded that the relationship between

firm’s growth and size is a highly non-linear one. In another study, during the same

year, Evans (1987b) – increased the data set to include all the firms in 100

manufacturing industries in the US and found that the above mentioned results are

robust to alternative assumptions concerning the effects of sample censoring and

functional form.

The above findings have been confirmed in other studies. For example Bronwyn and

Hall (1987), came to the conclusion that Gibrat’s law is weakly rejected for publicly

traded firms in the US manufacturing sector. It was rejected for the smaller firms but

was accepted for the larger firms. Nurmi and Satu (2004), using plant level data from

the Finnish Manufacturing sector, found that Gibrat’s law fails to hold. Shanmugam et

4
al. (2002) used data for the Indian manufacturing sector and reached the same

conclusion as Evans (1987a).

To this point in most of the studies related to the firm growth, only the manufacturing

sector has been analysed (see Sutton, 1997 for an overview). There are few studies of

Gibrat’s law for the services sector. A study by Audretsch et al. (2004) considers a

large sample of Dutch firms in the hospitality industry and discovers that the growth

rates are independent of the firm size i.e. Gibrat’s law holds. Hence they conclude that

the dynamics of industrial organisation for services may not simply mirror that for

manufacturing. Piergiovanni (2003) used data for the firms in the Italian small scale

services sector and found that out of five business groups considered, Gibrat’s law is

rejected for three business groups and was accepted for two. Hence, the services

sector has mixed results, unlike the manufacturing sector.

There are also many studies which have explored firm growth dynamics in greater

detail using other factors influencing firm growth. The R&D activities taken by firms

can give them a positive edge in the product market and thus contribute to firm

growth. Studies by Amirkhalkhali and Mukhopadhyay (1993) and Del Monte et

al.(2003) explore the role of R&D in determination of firm’s growth and found that

the firms with strong commitment to R&D have higher rate of growth. Apart from

R&D, other factors considered in various studies are export orientation and foreign

affiliation (Pfaffermayr (2004)) and market share (Koot et al.(1970)).

Only a few studies, have attempted to analyse growth dynamics for a so-called “high-

tech” industry such as the computer or information technology. One study on the

5
computer industry is Das (1995), focussed on examining the relationship between

growth and size1, age and the lagged size for the computer hardware industry in India,

using data for the period 1983 - 1988 in a fixed effect panel data setup. Her results

showed that current size and lagged size have a negative impact on the growth of the

firm and that the age of the firm positively affects the firm’s growth. Das &

Srinivasan (1997) looked at the impact of diversification and public ownership on

duration of survival of a firm in the computer hardware industry in India for the

sample period 1982-1992. This study found that diversification of firms and public

ownership has little effect on the duration of survival of the firm. Das (1998) focuses

on the impact of quality, foreign collaboration and ownership on firm's growth in

India using data for the computer hardware industry for the period 1983 to 1988 and

concludes that firms that use foreign technology produce higher quality but fewer

products than firms using domestic technology. She finds that joint ventures provide

better quality as well as better value products than firms with licensed technology and

that domestic R&D is an inferior substitute for foreign technology purchase. Public

sector firms, moreover, are observed to produce below-average quality but above-

average value products.

Here it is worthwhile to note that each of the above three articles assume the Indian

computer hardware industry is an infant industry. It was indeed so during the period

of study (1983 to roughly 1991-92). However, financial liberalization followed by the

elimination of important restrictions in the early 1990s allowed foreign collaboration

in technology sector and initiated a huge computerization drive in India. This resulted

in the Indian software industry becoming one of the leading software exporters in the

1
Size measured by the annual sales of the firm.

6
world. The infant industry argument no longer holds now. In this paper we analyze

the relevance or impact of different factors influencing growth separately for a fast

growing industry with a high rate of innovation.

3. An overview of the Indian computer industry

The computer industry of India, led by its software sector, started on a high growth

path in mid 1980s. With the New Industrial Policy of 1984-85, the Indian software

industry was liberated from heavy government regulations (namely, the Monopolies

and Restrictive Trade Practices Act) on import of foreign technology and was allowed

to form collaborations with foreign firms. In 1986, as restrictions on software imports

were further lowered, the Indian software industry changed its trade outlook from

inward orientation to outward. Indian firms were now permitted to be distributors of

foreign software packages and, by 1990, the software industry practically moved out

of the domestic market, leaving it to the distributors of equivalent foreign packages.

The Indian software industry turned its force into the exports market and became the

preferred destination for global software outsourcing2. The Enterprise Resource

Planning (ERP) boom and the upshots of the Y2K concerns further contributed to the

increase in its exports3 and, during 1991-96, the software industry of India grew at a

rate that was ten times higher than the growth of her GNP.

The burst in the dot com bubble of 2000 resulted in a drastic fall in software exports

and, as a slowdown set over the industry, many young companies were forced to shut

down or reduce production. This slowdown, however, lasted only a couple of years

2
In 1997-98, 158 of the FORTUNE 500 companies had outsourced their software projects to India.
3
This resulted in a positive externality for the local training industry and firms (e.g. NIIT, APTECH)
that rapidly expanded their business and turnovers. APTECH grew from 27 centers to 850 between
1991 and 1997 and its turnover went up from Rs. 100 million to Rs. 2.02 billion.

7
and the industry gained an upward swing in late 2002 and early 2003 and is booming

at present. The companies that survived the 2001-02 shock emerged more firm in

business and are now giants in the industry. The Indian software industry specializes

in the application software category, whereas its share in communication software,

firmware and consultancy activities has stagnated or even gone down. The hardware

industry is fully confined to the domestic market but the software industry is

becoming increasingly export oriented with professional services and turnkey projects

being the key activities. The domestic software market is more products and packages

based.

India’s IT industry can be characterized thus: “first, a software dominated IT industry;

second, a predominantly on-site development services provider for the export market

and a dominant products and packages developer for the domestic software segment;

and third, more important as an application-oriented software developer than a major

player in communications and consulting segments”.(Venkatesan and Malvea, 2002)

4. The Model Specification and Estimation Methodology

4.1 Data & Variables

The data are from the CMIE4 Prowess data set. It includes data on 350 computer firms

for the period 1991-2002. Of these 350 firms, 43 are hardware firms and 307 software

firms. Here, it is to be noted that there are many firms which lie in both hardware and

software categories as they produce both. But for the purpose of analysis, we have

used the categorization done by CMIE, which does not show overlaps5. During this

4
Centre for Monitoring of Indian Economy.
5
Prowess classifies companies based on the economic activity that constitutes more than 50% of the
company’s income. If a company is into both hardware & software then if the income from hardware

8
period of 10 years many firms exited from the industry and many firms entered.

Hence the data set has many missing observations and it requires an Unbalanced

Panel setup.

Most of the panel data sets, especially on individual firms, have missing years for

some of the cross-sectional units, thus giving rise to the unbalanced panel data model.

The missing observations arise due to mergers, acquisitions and exit of the firms

during the sample period. The methodology of fixed panel estimation with an

unbalanced panel does not differ much compared to the balanced panel. Both the

estimation methods i.e. the LSDV estimation and the within estimation, can be carried

out in the usual way, by using the Ti instead of T for each cross-sectional unit. The

total number of observations is then T1 + T2 + ⋅ ⋅ ⋅ ⋅ +Tn , instead of NT. Hence

appropriate adjustment is required in degrees of freedom calculation and subsequently

standard errors and test statistics calculations. However most of the regression

packages that do the fixed effects estimation make appropriate adjustment for this loss

of degrees of freedom.

4.2 Model specification

The dependent variable is Growth = ln(Sit+1) - ln(Sit). The explanatory variables are

Size (given by sales, denoted by lnS), Age (Current year - year of Incorporation;

denoted by lnA), Net exports (NEX), R&D Expenditure (RD), Number of products

(NOP) and dummies for Ownership. Also time dummies are included to account for

aggregate policy shocks and firm specific effects are taken care of by firm specific

activities constitutes more than 50% of it's income then Prowess classifies it as a hardware company
and vice-versa.

9
dummies. The interaction term6 is generated between the firm size and firm’s age, to

capture the joint effect of size and age on firm’s growth. The squares of size and age

are also included as explanatory variables to capture the possible non-linear relation

ship between a firm’s growth with its age or size or both. The details of units and

exact formulations for different variables used and the expected signs of the

coefficients can be found in TABLE-1 at the end.

4.2.1 Specification of Growth equation

The basic model used in this study is the same as that used by Das (1995) and by

Evans (1987). The General form of the function is:

ln S t +1 = ln F ( At , S t ) + u t (1)

Where ut is the disturbance term.

A second order logarithmic expansion of F(.) in the above equation gives

ln S t +1 = a0 + a1 ln S t + a 2 ln At + a3 (ln S t )(ln At ) + a 4 (ln S t ) 2 + a5 (ln At ) 2 + u t

(2)

The above equation can be written in the form of a firm growth equation by

subtracting ln S t from both sides

ln S t +1 − ln S t = a 0 + (a1 − 1) ln S t + a 2 ln At + a3 (ln S t )(ln At ) + a 4 (ln S t ) 2 + a5 (ln At ) 2 + u t

(3)

6
The rationale for using the interaction terms in the regression is to capture the marginal effects present
between the different explanatory variables. If we consider a simple case of
Y=a + b X1 + c X2 + d X1 X2 + u then dY/dX1 = b + d X2, which means that the marginal effect of
X1 on y depends on X2, where d captures the sign of this effect. If d & b are positive, it means that Y is
increasing in X1 at a rate that is increasing in X2. If d is negative & b is positive, it means that Y is
increasing in X1 at a rate that is decreasing in X2. If b is negative & d is positive, it means that Y is
decreasing in X1 at a rate that is increasing in X2.

10
Lagged size ( ln S t −1 ) is another explanatory variable and ai is a firm specific constant

that accounts for the Unobserved heterogeneity, our first equation to be estimated is

the following:

ln S it +1 − ln S it = ai + a1 ln S it + a 2 ln Ait + a 3 (ln S it )(ln Ait ) + a 4 (ln S it ) 2 + a5 (ln Ait ) 2 +

a 6 ln S it −1 + u it

(4)

TABLE1-1 gives the expected signs of the coefficients, based on the sign obtained by

previous studies and the assumptions of diminishing returns to scale and diminishing

returns to learning. We expect a negative sign for the current size and the age due to

DRS and DRL. We expect net exports, market share, diversification (captured by

number of products) and the investment in research and development to positively

affect growth. It is difficult to predict the sign for the particular ownership dummies

as previous studies that have used the ownership structure as an explanatory variable

in similar analysis have had no clear cut evidence of ownership explaining firm’s

growth.

4.2.2 Extensions of the Basic model

The size and age of a firm by themselves seem quite insufficient to explain firm

growth, in an industry such as the computer industry where there are large network

effects present, unlike traditional industries such as automobiles or machinery. Hence

we introduce in our first extension of the basic model two new variables namely No of

Products and the Ownership Effect. The rationale for the inclusion of these two

variables is explained in the following paragraphs.

11
The computer industry (both hardware and software) is characterized by the presence

of network effects and brand loyalties. Hence, customer's purchasing decisions also

involve issues of compatibility and brand. It is, therefore obvious to think that more

diversified firms will grow faster than less diversified firms, because more diversified

firms will have bigger networks and more established brand names. Moreover,

customers usually buy complete systems (i.e. combination of products like CPU,

Monitor, Keyboard, Mouse etc.) rather than individual machine parts. Hence, a firm

producing many products should have better growth prospects than a firm producing a

single product.

The Ownership of firms has a crucial role to play in the growth of the firm. The

computer industry is relatively new in a developing country like India. Hence, during

the late 1980s and early 1990s, the government promoted a massive computerization

drive in government undertakings such as banks and railways. This computerization

drive is still going on. The study by Dedrick et. al (1993) indicates that most of the

government contracts for hardware and/or software were given to public companies.

Infact, up to now the government (or public companies) is the single largest domestic

buyer of both software and hardware. Besides government intervention, the level of

technology is reflected in the ownership effect. Private foreign firms and joint

ventures are expected to use advanced technology, compared to their Indian

counterparts. The purely Indian firms try to fill this technological gap by investing in

R&D. However the earlier study by Das and Srinivasan (1997) found no significant

effect of firm’s ownership on firm’s duration of survival.

12
4.2.3 First Extension to the Basic Model

The first extension of the basic model (given by equation (4)) includes number of

products (NOP) and dummies for ownership. We have classified all the firms into six

categories based on their ownership viz. Private Indian, Private Foreign, Public,

Indian Group, Foreign Group & Joint Ventures. After including these new variables

our model takes the following functional form:

ln S it +1 − ln S it = ai + a1 ln S it + a 2 ln Ait + a3 (ln S it )(ln Ait ) + a 4 (ln S it ) 2 + a5 (ln Ait ) 2 + a6 ln S it −1 +

5
a7 NOPi + ∑ d j OD i +u it
j =1

(5)

4.2.4 Second Extension to the Basic Model

We have further extended the basic model by including some more variables i.e. Net

Exports, R&D Expenditure and Market Penetration.

The Indian Software industry is mainly export oriented because, in spite of having

skills & English speaking fluency comparable to their counterparts in developed

countries like the` US & Europe (Heeks(1996)), Indian software professional receives

a salary 4 to 5 times less than their foreign equivalents. Government policies also

support “high export orientation” (Heeks(1996)). Moreover out of top 400 companies,

more than 250 have already acquired ISO 9000 certifications7. Out of the 54

companies in the world that have acquired SEI CMM Level 5 Certification8, 27 are

located in India. Hence, the quality and standards in software production are assured.

7
Source: NASSCOM and The Week Magazine.
8
Software Engineering Institution Capability Maturity Model.

13
Thus net exports (NEX) are quite a pertinent variable to include for explaining

growth.

Research and Development behaviour has a vital role to play in an industry dominated

by rapid technological change. Global competition requires keeping up with the pace

of innovation by developed countries. Firms spending more on R&D are, in this

respect, expected to perform better than those who are not spending or spending less.

Similarly, Market Penetration at a certain point of time can be treated as a kind of

proxy for capturing the network effects due to the market share of a firm. Market

Penetration (MP) is calculated by the following formula:

Salesit
MPit =
Total _ Sales _ in _ year _ t

After including these new variables, our final extension to the basic model comes to

the following:

ln S it +1 − ln S it = a i + a1 ln S it + a 2 ln Ait + a 3 (ln S it )(ln Ait ) + a 4 (ln S it ) 2 + a 5 (ln Ait ) 2 + a 6 ln S it −1 +

5
a 7 NOPi + ∑ d j OD i +a 8 NEX it + a 9 RDit + a10 MPit + u it
j =1

(6)

5. Econometric Analysis

We began through using a Fixed Effect Panel model to estimate the above equations.

Here we have assumed that the error terms are independently and identically

distributed i.e. u it → iid . We used OLS as the starting point for estimating all the

models. Subsequently we ran the regressions using dummies for capturing the

aggregate policy shocks. Finally we introduced firm dummies to account for firm

specific effects (unobserved heterogeneity). The Hausman's Specification test rejected

the Random effect model for the fixed effect model. Hence we finally used the Fixed

14
Effect Unbalanced Panel Data Setup for our analysis. The results of our estimations

of all the three models (without time dummies, with time dummies only, with both

time and firm dummies) for the hardware, software and the overall computer industry

are documented in TABLE 2 - TABLE 4.

5.1 Main findings for the Basic Model

Current size has definitely a negative impact on the firm's growth, as it is coming

significant for hardware software and the overall industry (with negative sign). So our

finding confirms that of Das (1995). The lagged size has no significant impact on

hardware but it has a significant negative impact in case of software industry, thereby

indicating that fixed factors of production have important roles to play in the software

sector.

Age is not coming significant in any of the estimations, which shows that age doesn't

explain growth. The Diminishing Returns to Learning doesn't seem to hold here as it

is a booming industry with rapid technology changes. The interaction term between

size and age is significant in all the three cases.

5.2 Main findings of First Extended Model

After introducing the new variables of product diversification and ownership, we find

that age has a strong negative impact on growth for hardware industry whereas it has a

strong positive impact for software industry. This further clarifies the picture that

diminishing returns to learning is applicable to hardware but not to software. Also we

see that the diversified firms are more successful in the hardware industry, where the

issues of compatibility and brand loyalty actually matter. The ownership variables are

coming significant for hardware.

15
5.3 Main findings of Second Extended Model

Inclusion of net exports, R&D and the market penetration as new explanatory

variables doesn't significantly improve the model. But we can make some inferences

on the possible nature of impact of these variables, from the sign of the coefficients.

Net exports negatively affect hardware and have positive impacts for the software

industry, which is quite expected as discussed earlier in section 4.2.2. Market

Penetration has a positive effect for the hardware firms, which confirm the influence

of brand names, network effects etc. in the hardware sector. The R&D variable is not

coming significant in any case, which could be because of the difficulty in separating

R&D expenditure from other manufacturing expenditures in such industries. Hence

the R&D expenditure reported doesn't reflect the actual R&D activity taken up by

firms.

6. Discussion of results
Current size has a negative impact on growth for the hardware, software and overall

industry regressions. Thus, despite being a “high- tech” industry the Law of

Diminishing Returns to Scale or Bounded Efficiency holds well. The smaller firms

will grow faster than the large firms until they attain a size after which it is not

possible to improve efficiency any further. However, we also find that the square of

the size term is significant for both hardware and software. Moreover, it has a

negative sign for hardware but a positive sign for software. This signifies an important

difference between the software and hardware industry i.e. after attaining a certain

threshold the software firm grows with an increase in size. This finding could reflect

the situation when a large software firm will be capable of spending more on the latest

16
technology (either through R&D or through collaborations) or will have more quality

assurance certificates or will be able to retain its employees9.

Age, in our analysis, was found not to have significant impact on firm growth. But

again the signs are opposite for the hardware and software industry. We find that the

interaction term between age and size always has` a significant impact on growth.

Thus, we can infer that age and size together have an impact on growth. Moreover age

is significantly correlated with size (this correlation is 0.355 at 1% level of

significance, in our data).

Ownership has an important role to play in this industry. We see that in hardware the

Private Indian groups & Private Foreign are doing better than others. This is because

of the strong effect of brand names present in hardware markets. Here established

Indian brands (like HCL, Wipro etc.) and established foreign brands (like DELL,

IBM, SUN) are more successful than public and private Indian standalone firms

(involved in selling assembled PC's). This is because of the fact that big groups have

huge networks for providing service and support to customers. Moreover they are

capable of handling large projects that are spread over wide geographical areas (i.e.

projects by Government Agencies or Big Public Enterprises). The ownership effects

are not significant in the software industry, which shows that ownership does not

matter here. As opposed to the hardware industry, the software industry mainly relies

on exports and since most of their clients are corporate firms situated abroad, they do

not need large network of support services to fulfil the needs of the clients. Moreover

the service and support related to a software product can be delivered, by a firm even

9
The job switching rate is very high in software industry. So the small firms are not able to get the

17
if it is located geographically far away. So the software industry has to compete in the

world markets for the projects requiring high degree of specialisation, hence the

ownership does not matter.

A similar kind of difference is observed between hardware and software regarding the

impact of product diversification. Firms producing more numbers of products are

more successful in the hardware industry because of the fact that customers usually

buy computer systems (consisting of monitor, keyboard, CPU etc), rather than

individual components. Hence firms producing all of the components10 have a better

chance of selling the products (rather systems). The case of software is somewhat

different in the Indian context. Here most of the firms rely on selling customized

software designed to cater to the needs of a particular client. Hence there is no

production of general purpose packaged softwares. Therefore diversification does not

have significant impact on the software industry.

Most of the time dummies are significant, reflecting the aggregate shocks affecting

the industry. The time dummies for the years 1999 & 2000 are consistently significant

in all our estimates. These were the two years when the industry was doing better than

in subsequent years. The big software slump started at the beginning of 2001 and

lasted until the end of 2002. This was due to the failure of the huge dot-com

revolution in which mushrooming dot-com firms could not live up to expectations,

and crashed. With this crash of the dot-com companies the newly trained Java

benefits of spending on training of their employees. The job switching rate is less in large firms as
employees feel more secure.
10
Here we are using the term components though most of these devices (i.e. computer peripherals) are
end products in themselves.

18
professionals, all of a sudden, became jobless. This brought about the well known

Software market bust throughout the world.

7. Conclusions
There are striking differences between the growth patterns of the hardware and

software industries. The hardware industry, on one hand, follows a trend similar to

traditional industries like manufacturing, machinery and textiles. On the other hand

the software industry is somewhat similar to the services sector. Knowledge and

innovation have much more greater roles to play in software than hardware. Product

diversification is a lot more important in hardware than in software. The software

industry requires specialization in providing software solutions in one particular field.

This analysis can be done in a much more effective way by including some

employment statistics such as salary, organizational structure within a firm, the

experience and education level of employees because in a knowledge intensive

industry these variables really matter a lot. Similarly there were many outliers and

growth miracles in the data which can influence our analysis. Removing the outliers,

and may be also truncating the growth variable, will perhaps give us a better

perspective on the general growth patterns in this industry.

References

Acs, Z. J. and D. B. Audretsch (1990). "The Determinants of Small-Firm Growth in

U.S. Manufacturing." Applied Economics 22(2): 143-153.

Amirkhalkhali, S. and A. K. Mukhopadhyay (1993). "The Influence of Size and R&D

on the Growth of Firms in the U.S." Eastern Economic Journal 19(2): 223-

233.

19
Audretsch, D. B., L. Klomp, et al. (2004). "Gibrat's Law: Are the Services Different?"

Review of Industrial Organization 24(3): 301-324.

Baltagi, B. H. (2001). Econometric analysis of panel data. New York, John Wiley.

Cabral, L. (1995). "Sunk Costs, Firm Size and Firm Growth." Journal of Industrial

Economics 43(2): 161-172.

Das, S. (1995). "size, age and firm growth in an infant industry : The computer

hardware industry in India." International Journal of Industrial Organization

13: 111-126.

Das, S. and K. Srinivasan (1997). "Duration of firms in an infant industry: the case of

Indian computer hardware." Journal of Development Economics 53: 157-167.

Das, S. (1998). "Foreign Vs. Domestic Technology, Private Vs. Public Ownership and

Product Quality : The Case of Indian microcomputer Industry." Journal of

quantitative Economics 14(2): 97-121.

Dedrick, J. and K. L. Kraemer (1993). India's Quest for Self Reliance in Information

Technology : Cost and Benefits of Government Intervention, Centre for

Research on Information Technology and Organization (CRITO), University

of California, Irvine.

20
Evans, D. S. (1987a). "Tests of Alternative theories of Firm Growth." Journal of

Political Economy 95: 657-674.

Evans, D. S. (1987b). "The Relationship between Firm Growth, Size, and Age :

Estimates for 100 Manufacturing Industries." Journal of Industrial Economics

35(4): 567-581.

Geroski, P. A. and S. Machin (1993). "Innovation, Profitability and Growth over the

Business Cycle." Empirica 20(1): 35-50.

Hall, B. H. (1987). "The Relationship between Firm Size and Firm Growth in the U.S.

Manufacturing Sector." Journal of Industrial Economics 35(4): 583-606.

Heeks, R. (1996). India's Software Industry: State policy, Liberalization and Industrial

development, Sage Publications.

Monte, A. D. and E. Papagni (2003). "R&D and the Growth of Firms: Empirical

Analysis of a Panel of Italian Firms." Research Policy 32(6): 1003-14.

Nurmi, S. (2004). "Plant Size, Age and Growth in Finnish Manufacturing." Finnish

Economic Papers 17(1).

Pfaffermayr, M. (2004). "Export Orientation, Foreign Affiliates, and the Growth of

Austrian Manufacturing Firms." Journal of Economic Behavior and

Organization 54(3): 411-423.

21
Shanmugam, K. R. and S. N. Bhaduri (2002). "Size, Age and Firm Growth in the

Indian Manufacturing Sector." Applied Economics Letters 9(9): 607-613.

Singh, A. and G. Whittington (1975). "The Size and Growth of Firms." Review of

Economic Studies 42(1): 15-26.

Sutton, J. (1997). "Gibrat's Legacy." Journal of Economic Literature 35(1): 40-59.

Venkatesan, R. and S. V. Malvea (2000). Banglore : IT Cluster in India. Information

Technology in Asia : New Development Paradigms. C. S. Yue and J. J. Lim.

Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data, MIT

Press.

Wooldridge, J. M. (2003). Introductory econometrics : a modern approach, South-

Western College Pub.

22
 Tables

Table 1: Description of Variables

Expected Std.
Variable
Description sign Obs. Mean Dev.
GROWTH ln S it +1 − ln S it 1280 0.341 1.004
S Annualized Sales in Rs. Crore 1755 69.624 250.53
Age in years (i.e. current year - year
A of incorporation) 4200 5.443 8.212
lnS ln(Sales) - 1656 2.109 2.334
lnA ln(Age) - 3086 1.802 0.853
lnAlnA (ln A)2 + 3086 3.975 2.935
lnSlnS (ln S)2 + 1656 9.899 11.245
lnSlnA (ln S)x(ln A) - 1612 4.686 5.766
LlnS One period lag of log(Sales) + 1456 2.055 2.318
NOP No of Products + 4200 1.982 2.875
Dummy; takes a value of 1 if
DPI Private Indian Firm, 0 Otherwise 4200 0.657 0.474
Dummy; takes a value of 1 if
DPIG Private Indian Group, 0 Otherwise 4200 0.268 0.443
Dummy; takes a value of 1 if
DPF Private Foreign Firm, 0 Otherwise 4200 0.051 0.221
Net Exports (Rs.Crore) divided by
NEX Sales + 1656 0.068 0.801
Total R&D Expenditure(Rs. Crore)
RD divided by Sales + 1656 0.025 0.784
Market Penetration =
Salesit
MP Total _ sales _ in _ year _ t + 1755 0.006 0.021

23
 Table 2: Estimation of Basic Model♣

Variable Hardware Software Industry

0.1542 -0.1265 -0.6342* -0.0603 -0.0657 -0.4455** -0.0706 -0.0182 -0.420**
Ln(S)
(0.136) (0.182) (0.246) (0.066) (0.065) (0.124) (0.069) (0.071) (0.102)
0.1644 -1.2774* -0.9323 0.2837* 0.1375 0.0894 0.2986** 0.0185 0.3408
Ln(A)
(0.223) (0.515) (1.63) (0.111) (0.304) (0.816) (0.133) (0.237) (0.705)
-0.1061 0.124 -0.2096 -0.0454 -0.0153 0.2231 -0.0509 0.0125 0.0418
(ln A)2
(0.093) (0.107) (0.753) (0.048) (0.081) (0.445) (0.064) (0.070) (0.374)
-0.0119 -0.0105 -0.0586* 0.0426** 0.0409** 0.0573** 0.0396** 0.0378** 0.0589**
(ln S)2
(0.016) (0.017) (0.034) (0.006) (.005) (0.009) (0.007) (0.007) (0.008)
0.0621 0.1701** 0.3683** -0.0670* -0.0596* -0.0792 -0.0528* -0.0456* -0.0832*
(ln S)x(ln A)
(0.050) (0.064) (0.115) (0.027) (0.026) (0.054) (0.034) (0.034) (0.044)
-0.1753** -0.1403* -0.036 -0.0498 -0.0481 -0.1001* -0.0601* -0.0549* -0.0904*
L ln S
(0.065) (0.065) (0.076) (0.035) (0.035) (0.042) (0.037) (0.036) (0.036)
Time
NO YES YES NO YES YES NO YES YES
Dummies
Firm
NO NO YES NO NO YES NO NO YES
Dummies
R-Squared 0.1038 0.2966 0.5397 0.1627 0.2495 0.6032 0.1490 0.2363 0.5851

♣
** And * indicate significance at 1% and 10% level respectively. The values in parentheses represent
the standard errors.

24
 Table 3: Estimation of First Extension of Basic Model

Variable Hardware Software Industry

-0.2496 -0.2104 -0.7576** -0.0741 -0.0698 -0.3859** -0.0878 -0.09 -0.407**
Ln(S)
(0.199) (0.186) (0.233) (0.067) (0.066) (0.120) (0.060) (0.059) (0.101)
-1.0191* -1.2385* -2.0591** 0.4330* 0.1416 1.0472* 0.4095* 0.0709 -0.196
Ln(A)
(0.583) (0.549) (0.735) (0.189) (0.315) (0.596) (0.168) (0.264) (0.397)
0.0259 0.0852 0.2688 -0.0889 -0.0181 -0.3843 -0.0846 -0.0024 0.2881*
(ln A)2
(0.122) (0.116) (0.186) (0.061) (0.084) (0.269) (0.053) (0.070) (0.143)
-0.0096 -0.0112 -0.0393 0.0419** 0.0411** 0.0567** 0.0381** 0.0371** 0.0574**
(ln S)2
(0.015) (0.014) (0.034) (0.005) (0.005) (0.009) (0.004) (0.004) (0.008)
0.2077** 0.2071** 0.3987** -0.0662* -0.0616* -0.1024* -0.4092* -0.0441* -0.0737*
(ln S)x(ln A)
(0.072) (0.067) (0.112) (0.027) (0.026) (0.052) (0.023) (0.023) (0.044)
-0.1144* -0.1282* -0.0807 -0.0578 -0.0518 -0.0887* -0.0644* -0.0573* -0.0632*
L ln S
(0.068) (0.066) (0.075) (0.035) (0.035) (0.041) (0.031) (0.031) (0.036)
0.0079 0.0102 0.3266** -0.0078 -0.0159 0.6651 0.0024 -0.0006 0.0319
NOP
(0.008) (0.007) (0.092) (0.017) (0.017) (0.459) (0.009) (0.008) (0.023)
1.9481* 1.8304* 2.6881** 0.0905 -0.0353 6.7943 0.0714 -0.0771 -1.492*
DPIG
(0.799) (0.755) (0.901) (0.167) (0.188) (5.23) (0.152) (0.175) (0.857)
1.9202 1.8568* 1.2696 -0.1458 -0.1678 5.1329 -0.1142 -0.1764 -2.6196*
DPI
(0.809) (0.765) (1.092) (0.168) (0.186) (4.67) (0.153) (0.173) (1.10)
2.3712* 2.3926** 4.6271** 0.0251 -0.0991 6.9936 0.0674 -0.0612 -0.5619
DPF
(0.863) (0.832) (1.117) (0.220) (0.230) (4.35) (0.202) (0.214) (0.860)
Time
NO YES YES NO YES YES NO YES YES
Dummies
Firm
NO NO YES NO NO YES NO YES YES
Dummies

R-Squared 0.1453 0.3120 0.5121 0.1718 0.2536 0.6010 0.1557 0.2388 0.5764

25
 Table 4: Estimation of Second Extension of Basic Model

Variable Hardware Software Industry

-0.2576 -0.1786 -0.7547* -0.0733 -0.0622 -0.3704** -0.0765 -0.0724 -0.376**
Ln(S)
(0.196) (0.184) (0.331) (0.067) (0.066) (0.125) (0.064) (0.059) (0.110)
-0.6232 -0.9094* -0.5909 0.3514* 0.1563 1.0568* 0.3713* 0.0525 -0.2283
Ln(A)
(0.567) (0.547) (0.985) (0.195) (0.313) (0.597) (0.173) (0.262) (0.402)
-0.0746 0.0106 -0.346 -0.0703 -0.0211 -0.3861 -0.075 0.0044 0.3094*
(ln A)2
(0.121) (0.115) (0.354) (0.062) (0.084) (0.273) (0.054) (0.069) (0.147)
-0.0299* -0.0255 -0.0776* 0.0472** 0.0558** 0.0581** 0.0428** 0.0488** 0.0595**
(ln S)2
(0.017) (0.017) (0.034) (0.006) (0.007) (0.010) (0.005) (0.005) (0.048)
0.2430** 0.2143** 0.4681** -0.0689* -0.0692** -0.1107* -0.0569 -0.0586* -0.0875*
(ln S)x(ln A)
(0.076) (0.073) (0.153) (0.027) (0.026) (0.053) (0.024) (0.023) (0.048)
-0.1092 -0.1230* -0.0348 -0.0610* -0.0631* -0.0895* -0.0652* -0.0621* -0.0629*
L ln S
(0.066) (0.064) (0.074) (0.035) (0.035) (0.041) (0.031) (0.031) (0.036)
0.0012 0.0044 -0.1273 0.0064 0.014 0.6291 0.006 0.0062 0.0315
NOP
(0.007) (0.007) (0.328) (0.019) (0.019) (0.462) (0.009) (0.009) (0.023)
1.4667* 1.3176* 2.9638* 0.1356 0.048 7.022 0.0984 -0.0226 -1.4971*
DPIG
(0.794) (0.757) (1.259) (0.169) (0.191) (5.265) (0.155) (0.177) (0.858)
1.4891* 1.3625* 3.4088 -0.1109 -0.0709 5.3361 -0.1009 -0.1223 -2.6275*
DPI
(0.802) (0.762) (2.325) (0.171) (0.190) (4.700) (0.156) (0.176) (1.10)
2.2629* 2.0621* 0.0295 -0.0743 7.0543 0.0558 -0.0702 0.4777
DPF DR
(0.864) (0.827) (0.220) (0.229) (4.370) (0.202) (0.213) (1.06)
-0.5778 -0.5376** -0.3592 0.1083 0.1355 0.222 0.0018 0.044 0.0471
NEX
(0.175) (0.164) (0.238) (0.116) (0.112) (0.184) (0.093) (0.040) (0.150)
0.7693 0.3465 -2.0182 -0.0016 -0.0147 -0.0249 0.0001 -0.0117 -0.0238
RD
(1.85) (1.82) (2.064) (0.028) (0.027) (0.028) (0.026) (0.025) (0.027)
3.4371* 2.0674 3.5666 -4.0782 -9.7052 -0.0095 -2.5221* 5.5445** -1.7943
MP
(1.43) (1.66) (3.278) (2.80) (3.128) (4.71) (1.44) (1.588) (2.64)
Time
NO YES YES NO YES YES NO YES YES
Dummies
Firm
NO NO YES NO NO YES NO NO YES
Dummies

R-Squared 0.2168 0.3598 0.5579 0.1756 0.2657 0.6026 0.1586 0.2496 0.5773

26