Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

NORTH- HOLLAND

Modularity and Productivity


Assessing the Impact of Digital Technology
in the U.S. Telecommunications Industry

S U M I T K. M A J U M D A R

ABSTRACT

This article investigates the impact of digital technology diffusion on the productivity of firms making up
the local operating sector of the U.S. telecommunications industry for the years 1991 to 1993. Diffusion of
digital technology is measured as the extent of line digitalization that has taken place, with digitalization expected
to enable firms to reap significant operating effieiencies. While the diffusion of this particular technology in
the local operating company sector has been relatively low, it is found that, after controlling for the effects of
a number of other covariates also likely to affect firm-level performance, the spread of digital technology
within the U.S. local telephone network plays a significant and positive role in impacting productive efficiency
of the firms making up the network. The investments that have been made in energizing the U.S. information
superhighway, via the digitalization of telephone lines, are justified in productive efficiency terms. The article
contributes to empirical literature in the technical change area which assesses the firm-level productive efficiency
consequences of technology diffusion. © 1997 Elsevier Science Inc.

Introfluetion
T h e a d o p t i o n of a n e w t e c h n o l o g y is a crucial decision for firms, b e c a u s e a n e w
t e c h n o l o g y n o t only changes the way m a n a g e r i a l a n d o p e r a t i o n a l processes are con-
d u c t e d w i t h i n firms, b u t also significantly e n h a n c e s c o m p e t i t i v e n e s s by giving the acquir-
ing firms a source of a d v a n t a g e over others. I n the field of industrial o r g a n i z a t i o n there
has b e e n a c o n s i d e r a b l e b o d y of work researching how particular factors, especially the
type of i n d u s t r y a n d m a r k e t structures involved, help in the g e n e r a t i o n of technical
c h a n g e [1-4].
A large literature [5-10] recognizes that aggregate productivity growth a n d eco-
n o m i c progress is a f u n c t i o n of technical change. But, as N e l s o n a n d W i n t e r ([11], page
235) have r e m a r k e d with respect to firm-level analyses of the c o n s e q u e n c e s of technical
change, " T h e effect of diffusion o n such variables as productivity a n d factor shares have
b e e n less studied, a n d there is virtually n o work that has a t t e m p t e d to tie t o g e t h e r
analysis of diffusion p a t t e r n s a n d productivity changes," a p o i n t that has b e e n re-stressed

SUMIT MAJUMDAR teaches Corporate Strategy at the University of Michigan Business School in
Ann Arbor.
Address correspondence to Sumit K. Majumdar, University of Michigan Business School, Ann Arbor,
MI 48109.

Technological Forecasting and Social Change 56, 61-75 (1997)


© 1997 Elsevier Science Inc. 0040-1625/97/$17.00
655 Avenue of the Americas, New York, NY 10010 PII S0040-1625(97)00023-1
62 S.K. MAJUMDAR

by contemporary researchers in the area of technology [12]. Similarly, Tushman and


Nelson ([13], page 2) state that, "there is no analogous literature to inform our under-
standing of either technological progress or the effects of technological change on
organization and individual outcomes."
The objective of this study is to assess the impact of the diffusion of digital technol-
ogy within firms making up the local exchange sector of the U.S. telecommunications
industry on their productivity, which is a factor correlated with competitiveness. Produc-
tivity is measured as relative efficiency, and evidence on the role that digital technology
deployment plays in influencing firm-level patterns of productive efficiency is generated.
The present study is based on a unique methodological approach that facilitates relevant
firm-level analysis. The approach helps to assess the impact of technical change on
productive efficiency, and in doing so augments the empirical literature on technical
change.
The results generated are also likely to be of contemporary policy significance. The
data will reveal whether the large investments incurred in energizing the information
superhighway by digitalizing the local network has had an impact on the economic
performance of the local exchange companies that make up the land-based telecommuni-
cations backbone of the United States. The article is organized as follows. The second
section contains a discussion of issues relating to technical change and digital technology.
In the third section, empirical issues are discussed, while the fourth section contains an
analysis of the results, and the final section concludes the article.

Technical Change: Theory and Digital Technology


The impact of technical change is classified as having simultaneous impact on the
linkages that firms have with markets and customers [14]. Technical change may involve
variations in the operations or production processes that help eliminate design flaws,
or make the use of existing resources and skills more effective. Conversely, technical
change can have quite the opposite impact by imposing operational requirements that
the existing resources and skills cannot satisfy. With respect to market and customer
linkages, technical change may simultaneously create new applications and uses for the
product or service offerings, allowing firms to enter new markets. Alternatively, technical
change may lead to a firm's serving its existing customers in a superior manner.
Based on the two dimensions analyzed, customer and market linkages, four catego-
ries of technical change are defined: architectural, which involves new technology applied
to new markets; niche creation, which involves refinement in technology applied to new
customer groups and new applications; regular, which involves refinements in technology
applied to existing markets and customers; and revolutionary, which involves a disruptive
change in technology applied to existing markets and customers [14]. Each of these
categories of technical change have different competitive implications. For example,
revolutionary technical change can, in the short-run, change product functionalities and
alter the price-performance tradeoffs for customers; in the long-run, such a change
can quite easily help to re-define industry boundaries. This categorization can help in
understanding the market evolution implications of technical change, but does not throw
much light on the possible intrafirm productivity implications of technical change.
The analysis is extended [15] to address intrafirm operational process concerns.
Technical change can impact on the components within a technological system, either
overturning or reinforcing the technical concepts that are embodied in the components.
Technical change can also have a simultaneous impact on the nature of linkages between
components in a technological system, changing the nature of linkages that exist, or
MODULARITY AND PRODUCTIVITY 63

leaving the linkages unchanged. Radical changes establish a new set of concepts that
are physically embodied in the components. The components need to be linked together
in different ways. Conversely, while incremental changes may extend and improve
existing designs, the nature of concepts embodied in the components, and the linkages
between the components, remain the same. Now two categories of technical change
are: architectural, in which the existing system of linkages between the components
have to be re-configured, but the concepts embodied within the components do not
change; and modular, in which the existing linkages between the components of a
technological system do not change, but the design concepts embodied in the components
undergo a change.
Digital technology involves a modular technical change, because the adoption of
digital technology does not change the overall architecture of a telephone system. A
telephone system still consists of the same components, and these components have to
be linked together in the same way as before. The interface between digital technology
and either analog or electro-mechanical technology has also been designed so that
digital components can be blended into an existing telephone system. What the adoption
of a modular technical change, such as digital technology, does is to change the concep-
tual basis underlying the tasks of transmitting voice and data by altering the nature of
functionalities that are provided by the design concept embodied in the components.
This change can have a positive and significant impact on the patterns of observed firm-
level productive efficiency.
In a telephone system, as people speak the voices actuate transmitters which cause
electric current to flow in proportion to sound pressure, or analogously. Because of the
way people speak, there is little reason to convert analog to digital sounds; economic
reasons which have to do with the cost of equipment, and quality reasons which have
to do with the clarity of transmission of signals, however, make digital equipment the
technology of choice in the telecommunications industry. The system for transmitting
telephone signals digitally is known as pulse code modulation (PCM), and a PCM
channel converts voice or data signals, which are encoded and decoded by central-office
switches, into eight-bit digital words to be transmitted over a line interspersed with
words from 23 other channels. Repeaters then regenerate the 24-channel signal approxi-
mately once every mile to avoid sound losses and retain transmission clarity.
The concept of PCM is over half a century old, having been discovered in 1938.
The first digital transmission systems were built for wartime use, and facilitated communi-
cations between Prime Minister Sir Winston Churchill and President Franklin D. Roose-
velt [16]. Digital systems were first installed commercially in the 1960s. Only recently,
however, have the price drops of integrated circuits and the miniaturization of electronic
components made conversion to digital systems feasible, and since the 1980s a drop in
the price of high-capacity fiber optics has also made the installation of digital transmission
systems economically attractive.
Other than hardware-related lower integrated circuit costs, digital transmission
systems can directly connect with various models of switching systems and data circuits,
thus enhancing the utilization of assets already in the system. In addition, digital transmis-
sion units are more compact and suitable for short- and medium-haul transmission of
voice and data which fall under the purview of local operating companies. Digital
transmission systems also have lower circuit cost for transmission distances ranging up
to 300 to 500 miles. Such transmission distances comprise the maximum length over
which the local and intra-LATA (local access and transport area) toll calls made through
the local operating companies are normally likely to travel.
64 S.K. MAJUMDAR

Digital systems have one drawback. They require greater bandwidth than compara-
ble analog systems. Where an analog system can transport a voice signal in four kilohertz
of bandwidth, a digital system requires 64 kilobits per second. To eliminate bandwidth
capacity constraints requires the use of fiber-optic cabling within a telephone company
network; therefore, the utilization of digital transmission systems within a network
requires that an interaction between the levels of digital and fiber-optic technologies
in use be also evaluated. The use of data and voice compression facilities have obviated
the use of greater bandwidths by digital systems [17]. 1 Other than the improvement in
transmission quality arising from a different methodology of transmission, the major
operational advantage of digital over analog channel banks is the possibility to plug in
modular units, thus enabling a system to be expanded or contracted easily.
The availability of modular plug-in units enhance system flexibility and usage
greatly, because contained within these plug-in units are electrical functions, such as
integrated signaling, wide-band transmission, and data transmission. To provide these
functions would require the use of additional externally placed capital equipment in
analog systems. Contemporary channel banks can be remotely accessed, and units can
be remotely adjusted. Through the remote access facility, transmission levels setting
and maintenance testing can be carried out, enhancing considerably the operating
efficiency of the transmission and the overall telephone system. Additionally, mainte-
nance costs are dramatically reduced because automatic error correction and detection
reduce manpower requirements through the use of digital technology [17].
Other than technical consequences, there are several economic and managerial
implications arising from the deployment of a modular innovation such as digital equip-
ment. From an economic perspective, the adoption of a modular technology is less risky
than adopting a radical technology, even though such a radical technology may lead to
greater cost savings, because of technical compatibility reasons which then lead to
economic benefits. In this respect, it has been noted that the degree of physical compati-
bility of different vintages of capital goods embodying different levels of technical
change has major effects on the economic effects of sunk costs. When piecemeal addition
of new vintages is possible indivisibilities are reduced and firms can fund new vintages,
such as make replacement investments which embody technological innovations on a
marginal basis. When physical compatibility is low and addition of new vintage is difficult
or expensive, the relevance of sunk costs increases as well as the delay in replacement
investment [18].
Therefore, even though an industrial system may consist of complex interconnec-
tions between various components, which have adaptively developed over time, the
flexibility associated with a modular technology does not destroy these interconnectivi-
ties or require the replacement of the complete system infrastructure. Because a tele-
phone or a telecommunications network consists of a number of inter-related compo-
nents, there are systems-scale economies arising as new efficiency-enhancing technology
are deployed within the system [19]. These network effects arise from the reduction in
the cost of training, maintenance, and spare parts. As a greater proportion of digital
technology is used within a given telephone company's network, the benefits of these
economies are accelerated because of a process of increasing returns.

In the United States, PCM digital transmission systems are commonly known as T-! lines. A T-1 consists
of a repeater at each end feeding either twisted-pair copper wire or fiber-optic wire the voice or data signals
supplied after conversion by the channel banks. The core of a digital system is a channel bank which combines
24 data and voice circuits into a T-1 bit stream.
MODULARITY AND PRODUCTIVITY 65

Empirical Analysis
The study uses data for a panel of 45 local operating companies, obtained from
the Federal Communications Commissions Statistics of Communications Common Carri-
ers [20], for the years 1991, 1992, and 1993. These data are contemporary, and the firms
comprise almost the entire population of the sector. For evaluating the impact of
digitization on firm-level productive efficiency, there are several reasons for the use of
data for these years. The standard reporting format of accounting data and operational
plant statistics, in existence for over 30 years, was radically changed after 1987. Until
1987 firms reported the composition of their installed base of switches as to whether
they were electronic or electro-mechanical. Such technology composition data up to
the year 1987 were used to evaluate the impact of the change over from electro-
mechanical to electronic technology on firms' performance [21].
Starting in 1988 data were reported in a new format. In 1991 the format for reporting
plant statistics was again changed, so that different aspects of firm-level plant and
technology composition were being reported. The specific format changes relate to the
reporting of digitalization data, which was reclassified from 1991 onward. Therefore,
digitalization data for the years 1988 to 1990 are not comparable with data for the years
1991 onward. The 45 firms analyzed operate 99% of all local telephone lines in the
United States, and their total revenues generated over $80 billion. The aspect of technical
change that is evaluated in the current study is whether the introduction of digital
electronic technology has a discernible impact on productive efficiency, given that by
1991 electronic technology became ubiquitous.
First, the relative efficiency of the firms are estimated using data envelopment
analysis (DEA). Second, the D E A output is used as the dependent variable in a model
where the key explanatory variable is the extent of line digitalization for each company.
A set of variables which are explanators of variations in efficiency patterns are included
as controls. The log of the productive efficiency measure estimated for each local
operating company using D E A is the dependent variable.
This approach, by first understanding firm-level efficiency analysis, and then ex-
plaining variations in firm-level efficiency patterns using a variety of firm-level variables,
avoids an aggregation problem that characterizes most studies of the telecommunications
industry [22]. In capturing technical change, the study uses a finer measure of intrafirm
technical quality, the level of firm-level line digitalization, which is unlike earlier studies
of the Canadian and the U.S. telecommunications industry. Those studies used the
aggregate and inappropriate or ad hoc measures, such as simple time-trends, annual
R&D expenditures by AT&T, or the percentage of direct-dialled calls to measure
technical change.
To evaluate the impact of line digitalization on firm-level productive efficiency, as
calculated for each firm-level observation using the D E A algorithm of Banker, Charnes,
and Cooper [23] (BCC), panel-data based regression analysis is carried out, with the
efficiency score for each firm being used as the dependent variable in the models, There
are several econometric advantages associated with such a panel-data approach [24].
The dependent variable lies between 0 and 1 and is half-normally distributed. This
distribution is converted to log-normal form by transforming the variable using natural
logs. There are 45 cross-sectional observations and three time periods, giving a total of
135 observations, and a pooled cross section time-series model which corrects for cross-
sectional heteroscedasticity, assuming cross-sectional independence and time-wise auto-
regression is used for estimation purposes.
66 S.K. MAJUMDAR

Charnes, Cooper, and Rhodes [25] developed and Banker, Charnes, and Cooper
[23] extended an efficiency measure first developed by Farrell [26] using a fractional
program where the ratio of the weighted outputs to weighted inputs of each observation
in the data set is maximized. For each observation a single statistic, in which a ratio
measure of how efficient each observation is in converting a set of multiple inputs jointly
and simultaneously into a set of multiple outputs, is calculated. Using only observed
output and input data, and without making any assumptions as to the nature of underlying
technology or functional form, the algorithm calculates an e x - p o s t measure of the
efficiency of each observation, accomplished by constructing an empirically-based fron-
tier, and by evaluating each observation against itself and all other observations included
in the data set.
Each observation rated as efficient is used to define an efficiency frontier, and firms
not so rated are evaluated by comparison with a firm on the frontier, with broadly
similar output or input mixes as the firm being compared. Thus, data from efficient
firms are used to create a frontier based on the principle of envelopment. The efficiency
measure gives an indication of how well each firm performs relative to its potential and
to other firms. The best firms score 1, on a scale of 0 to 1, and for the inefficient firms
the difference between their scores and 1 gives an idea of the efficiency improvement
that is possible.
The generalized D E A model is presented by the following formulation:

Max ek.k (1)

subject to:
ej,k --< 1, Vj; P~rk--> E, Vr; and Vik --> e, Vi;
where j > 1, . . . . n is the index for observations, k being used as the index for the
observation being specifically evaluated and ek.kis the efficiency score for the observation
thus evaluated, r -- 1. . . . . R is the index for the outputs (y~ -> 0 is output r of observation
j), i = 1 . . . . . I is the index for the inputs (x~j -> 0 is input i of observation j), ej,k is the
relative efficiency of observation j when observation k is evaluated, ~,~, Vik are the output
and input weights, respectively, associated with the evaluation of observation k, and
is a non-Archimedean infinitesimal quantity. In (1), the input (xij) and output (y,j) factors
are known quantities observed from the activities of the observations and the factor
weights (P.,k and vik) are the decision variables. Defining,

ej,k = E ~ I tx,k Yrj / E[~t v,k xq (2)


yields the model of Charnes, Cooper, and Rhodes [25] (CCR), which is the basic
D E A model.
Linear programming based approaches [27] used in empirically evaluating economic
phenomena have been criticized for making the assumption of constant returns to scale,
a condition of theoretical import but not of practical use [28]. The original Farrell and
CCR models also suffer from this lacuna, and assume a homogeneous production
function, which is untenable with the postulates of the contemporary capabilities-based
theory of the firm in which heterogeneity plays a key role. Banker, Charnes, and Cooper
[23] show that the efficiency score generated by the CCR model is a composite total
efficiency score which can be broken into two components; one component capturing
scale efficiency, or the ability of each observation to operate as close to its most produc-
tive scale size as possible, and the other component capturing pure managerial efficiency,
MODULARITY AND PRODUCTIVITY 67

which is the ability of a firm to convert of set of multiple inputs jointly and simultaneously
into a set of multiple outputs.
To isolate managerial efficiency, the BCC algorithm assumes that variable returns
to scale exist for firms, and a variable u0 is added in the programming formulation so
that the hyperplanes for each observation do not pass through the origin, unlike in the
CCR model, where hyperplanes pass through the origin because constant returns to
scale are assumed. In the constraint set for the linear programming model, this variable
is kept unconstrained so that it can take on values, depending on the data, which are
negative (denoting increasing returns to scale may exist), 0 (denoting constant returns
to scale may exist), or positive (denoting decreasing returns to scale may exist) for each
fh observation. Therefore, defining the relative efficiency measure as

e j, = E , % ~ , yr, - u,,/E~=~ ~'ik x,j (3)

where u0 is the unconstrained decision variable yields the BCC model. The CCR model
generates a total efficiency score, while the BCC model generates a managerial efficiency
score. Dividing the CCR score by the BCC score generates a measure of scale efficiency
for each observation [29].
The advantage of D E A is in its approach. D E A optimizes for each observation in
place of the overall aggregation of data and thereafter the single optimization performed
in statistical regressions. Instead of trying to fit a regression plane through the center
of the aggregate data, D E A floats a piecewise linear surface to rest on top of all the
observations. The frontier construction is empirically driven by the data, rather than
by restrictive assumptions as to the nature of underlying technology, or as to whether
the functional forms are linear or nonlinear. The only assumptions made are that of
piecewise linearity and convexity of the envelopment surface, and the D E A algorithms
also take each observation's idiosyncrasies into account in the computation of relative
efficiency score, unlike in regression-based estimation techniques where efficiency pa-
rameters are calculated based on an averaging process [30].
Within the D E A framework two estimation approaches are feasible. First, it may
be assumed that firms conserve inputs; then, the algorithm evaluates minimal use of
inputs, with outputs generated kept constant. Second, it may be assumed that given a
finite stock of inputs available, firms seek to maximize outputs that can be generated
with these. For the 45 local operating companies evaluated, the BCC input-conserving
algorithm is used to calculate efficiency scores for each observation. Three outputs:
local calls, inter-LATA toll calls, and intra-LATA toll calls, and three inputs (number
of switches, number of lines, and number of employees) are used in the computations.
The choice of these variables is consistent with industry literature [31, 32].
The present research is concerned with whether the diffusion of digital technology
into the telecommunications network impacts productive efficiency. Prior research, using
data up to 1987, has established that a transition from electro-mechanical to electronic
switching did positively impact firm-level productive efficiency [21]. The current study
differs substantially in providing evidence as to whether the impact of digitalization,
which is an important characteristic describing the technological quality of the contempo-
rary telecommunications network, both in the United States and overseas, has productiv-
ity consequences.
Data on whether electronic switches and transmission equipment were analog or
digital was not available for analysis in the earlier study. Also, between 1987 and 1991
there has been rapid diffusion of electronic technology in the network, principally
68 S.K. MAJUMDAR

because, by mandate, the local operating companies have to provide equal access to
other companies, primarily long-distance firms, who may wish to interconnect, and
electronic technology makes equal access feasible. The digitalization data has been
made available, however, only since 1988; also, digitalization data for the period 1988
to 1990 are not comparable with digitalization data for the period 1991 to 1993 because
collection and reporting definitions changed. Hence, the study is restricted to using data
for a narrow time series of the years 1991 to 1993. On the other hand, these data are
of very recent origin and the results are of contemporary relevance and importance.
The variable D I G I T A L , measured as the percentage of digital lines to total lines,
captures the likely impact of line digitalization on productive efficiency and its impact
is expected to be strongly positive [33]. In spite of declining equipment prices, digitaliza-
tion has not been rapid in the United States [34]. By 1990 less than 10% of the network
was digitalized. One reason is that in the 1980s U.S. local operating companies made
considerable investments in analog electronic technology, of which, consequently, there
is now a very large installed base providing adequate service. Hence, the pace of diffusion
of digital technology has been held up. This pattern of digital technology diffusion also
corresponds with a view that countries which are late-comers to electronic technology
may be able to leapfrog from the electro-mechanical to the digital generation, bypassing
the analog generation completely. On the other hand, early adopters, including the
local U.S. operating companies, have fallen behind in modern technology usage because
of the large analog base in existence, which was installed relatively early in the United
States compared to other nations [35].
An additional evaluation of the productive efficiency impact of digital systems is
to test whether the possession of digital systems when interacted with the quantity of
fiber-optic cabling possessed by local operating companies also influences efficiency
patterns positively. Therefore, a variable D I G I T A L * F I B E R is also created, which is
an interaction term between the level of digital lines and the relative quantity of fiber
optics that the local operating companies possess. The local operating sector of the U.S.
telecommunications industry is heterogenous. It consists of a number of firms possessing
diverse characteristics which are subject to a number of influences. Therefore, variables
are introduced to control for firm-specific effects as well as industry-specific factors that
are likely to impact on firm-level efficiency.
There are a number of holding company groups in the United States. For example,
the Regional Holding Companies (Baby Bells), such as Bell Atlantic, Pacific Telesis,
or U.S. West, the GTE group, the United group owned by Sprint, the CENTEL, and
the Continental groups. Each of these groups operate a number of local operating
companies in their areas of jurisdiction. The number ranges from 1, where only one
local operating entity exists, to 7. Where the number is low, it means that the local
telephone operations are centralized under a few local operating company entities. A
high number denotes decentralization of telephone operations in various jurisdictions
through the establishment of several local operating entities.
There are efficiency implications arising from the diversity of patterns as to how
local operating company groups manage their telephone operations. Where a number
of local operating company entities are under the control of a local operating company
group, the spillover of knowledge and experiences gained from operating in diverse
territories in a decentralized manner can be leveraged between the other companies
operating under the group. This can have a positive impact on the efficiency of each
group company. Conversely, the existence of a fewer number of operating company
entities can enhance technological economies of scale; because, however, of potential
MODULARITY AND PRODUCTIVITY 69

coordination and communication problems associated with large size, the knowledge
and expertise acquired in each sub-area may not flow through to the benefit of other
local operating areas in the group as a whole. SIBLINGS, which is an integer variable
denoting the number of local operating companies operated by each holding company
group, is introduced as a variable to control for the specific structure of each local
operating company, and the impact of this variable on efficiency is left to be empiri-
cally determined.
The percentage of debt in the capital structure can have a significant impact on
productive efficiency. Where firms have high debt they tend to get monitored stringently
by the lenders, and are likely to perform better. With higher equity there can be a
collective action problem for numerous shareholders to do the monitoring. On the other
hand, where public utilities, such as telephone companies, are concerned they are
considered low-risk investments and can attract a larger amount of debt in comparison
with firms operating in other sectors of the economy. Therefore, local operating compa-
nies with relatively greater levels of debt may not feel lender pressures and the associated
incentives to perform in a superior manner. While theory has a prediction with respect
to the impact of debt on performance, the institutional context mediates the impact of
such factors and the issue has to be empirically sorted out. DEBT, the percentage of debt
in the total capital of each operating company, is introduced as a key control variable.
A variable, CUSTOMER EXPENSE, defined as the percentage of total marketing
and customer support expenses to total operating expenses, captures the market-devel-
opment orientation of the local operating companies. Since the 1984 divestiture, local
operating companies have had to become significantly market oriented, and such an
orientation is expected to help companies generate greater volumes of output in compari-
son with their compatriots whose comparative levels of spending on customer support
and development activities are lower. This variable is expected to be positively related
to performance.
While local companies have monopoly control over the local network, they have
to permit long-distance and other certified service providers access to their networks,
so that long-distance and other value-added services can be provided to consumers.
For doing so, they receive an access charge from these service providers, which can
become quite substantial. Given that there is large capacity available in local networks,
for which fixed costs have already been borne, access revenues represent an income
arising from the use of surplus network capacity. These revenues are almost like free
cash-flow, and where firms derive a greater proportion of their revenues from access
charges there are lesser incentives to be efficient as these revenues arise in respect of
assets already in place the marginal cost of utilizing which are zero. ACCESS is intro-
duced as a control variable for this aspect of local telephone company operations, and
is the percentage of access revenues to total operating revenues that are earned by each
local operating company.
Market power and efficiency is inversely related because dominant firms, especially
in relatively controlled markets such as local exchange telecommunications, do not
perceive the threat of instant displacement by newcomers [3]. Such perceptions tend
to breed inefficiencies because of the availability of a large customer base to which
costs can be passed on. SWITCH SHARE, measured as the percentage of switches a
company possesses in its given operating area relative to the total number of switches
all firms possess in that operating area, captures the relative share of installed base
possessed and the expected impact of this variable is negative.
70 s.K. MAJUMDAR

The microcomposition of the market can induce efficiencies in a positive way [36].
The threat of bypass, a direct connection between customer's premises and another
carrier or a self-contained system avoiding the operating company's system fully [37],
may induce efficiencies among the local operating companies, because revenues are
lost through bypass but all costs still have to be incurred to maintain the network in
its present existing size and condition. Such reasoning is consistent with the ideas of
Schumpeter [9] who saw micromarket threats as the key spur to superior performance.
Business customers are important for revenue-raising purposes [38] and are the custom-
ers likely to bypass. B U S I N E S S L I N E S , measured as the percentage of a firm's lines
that are business lines, measures the susceptibility of bypass and is expected to positively
impact efficiency.
A number of local operating companies have a large exposure in providing intra-
L A T A toll services within the context of a market which is considerably competitive.
Local operating companies must, in addition to providing the normal telecommunica-
tions infrastructure required for local services, provide trunking infrastructure for provid-
ing toll services. Where the proportion of toll call operations to total telecommunications
operations, measured by T O L L R E V E N U E (the proportion of toll revenues to total
operating revenues) is higher, the infrastructure requirement is also likely to be greater.
The provision of such infrastructure is assumed to be incremental to the provision of
local operations infrastructure; therefore, local operating companies are clamoring to
be allowed entry into long-distance markets. Whether this assumption is valid or not,
in terms of enhancing productive efficiency, is left to be empirically determined.
Finally, ownership can influence efficiency. In the United States, ownership of
telephone operations is vested among a number of private firms, principal among which
are the Baby Bells. Past research has established variations in behavior and performance
between Baby Bell and non-Baby Bell companies [39]. Dummy variables are assigned
if the firms are either one of the Baby Bell companies or belong to the G T E group,
the C O N T E L group, the CENTEL group, or the United group. The base-case firms
left out are three independent local telephone operators: Cincinnati Bell, Rochester
Telephone Company, and Southern New England Telephone Company.

Discussion
The regression estimates where control variables are included in the different
models estimated are given in Table 1. When the log of the D E A scores are regressed
in D I G I T A L alone as a regressor, D I G I T A L is positive and significant when regressed
alone. Thereafter, the regressor used is the variable, D I G I T A L * F I B E R , which captures
the interaction between digital transmission technology and fiber-optics technology since
the simultaneous use of fiber-optics cabling permits greater bandwidth to be made
available. This explanatory variable also turns out to be positive and significant. The
control variables can be considered as exogenous for the purposes of the model; the
relative proportion of firm-level investments in digital lines can depend on the impact
of the various exogenous control variables. The relative deployment of fiber-optics
technology can also be influenced by the exogenous variables. Two-stage least squares
models are also estimated, with D I G I T A L and D I G I T A L * F I B E R treated as endoge-
nous variables and the control variables treated as exogenous. The results for these
models show that D I G I T A L as well as D I G I T A L * F I B E R turn out to be positive
and significant.
As shown in Table 1, models 1 to 5 test for the sensitivity of the estimate of
D I G I T A L to the inclusion of different sets of control variables. In all models, the
MODULARITY AND PRODUCTIVITY 71

TABLE 1
Regression Estimates: Dependant Variable and Log of Productive EIticiency Score
Coefficient estimates
Variable Model 1 Model 2 Model 3 Model 4 Model 5
Constant -0.125 -0.227 -0.154 -0.139 -0.085
(3.74) (4.45) (4.11) (2.05) (1.30)
DIGITAL 0.014" 0.0(18"* 0.011' 0.007**
(3.96) (1.94) (3.36) (2.18)
D I G I T A L * FIBER /).0(/3"
(4.31)
SIBLINGS -0.001 * 0.006*** 0.006***
(2.50) (1.47) (1.67)
DEB T - 0.003* - 0.004* - 0.004*
(7.62) (8.65) (11.33)
CUSTOMER E X P E N S E 0.009* 0.003* 0.001
(8.74) (2.59) (1.02)
ACCE S S -0.002* -0.002** -0.002*
(2.67) (2.02) (2.43)
SWITCH S H A R E -0.001' -0.000 -0.000
(5.54) (1.05) (0.67)
BUSINESS LINES 0.010" 0.004* 0.002*
(8.12) (3.54) (4.51)
TOLL R E V E N U E -0.001"* -0.005 -0.0(Xt
(2./10) (1.24) (0.77)
BABY BELL 0.061"** 0.081 0.014
(1.62) (1.66) (0.33)
G TE -0.100" -0.016 0.072"**
(2.65) (0.28) (1.35)
CONTEL 0.008 0.135" 0.090"*
(0.19) (2.77) (2.00)
CENTEL 0.089** 0.114"* 0.065
(2.13) (2.05) (1.28)
UNITED -0.079"* 0.010 -0.036
(1.68) (0.21) (0.71)
t-Statistics in parentheses: * p < .01; **p < .05; ***p < .10 (one-tailed).

D I G I T A L variable is found to be positive and significant, while in model 5 D I G I T A L *


F I B E R is also found to be significant, again after controlling for the various factors
that may influence productive efficiency. The need for introducing control variables is
important, since productive efficiency can be affected by a number of other factors,
and any one of these could wipe out the impact that technology diffusion alone might
have. The results, however, indicate that the role of technology diffusion is robust in
influencing firm-level productive efficiency.
The results indicate that although digital technology has not diffused completely
into the telecommunications network infrastructure in the United States, it has had a
significant and positive impact on firm-level efficiency. Because there is a large installed
base of older vintage technology in place in the U.S. network, even more significant
productive efficiency gains will be noticeable as greater diffusion of the digital technology
takes place within the U.S. local operating companies' systems. The pattern of results
that have been noted have important implications for observers of the U.S. telecommuni-
cations industry, as well as for scholars of technical change. Digital technology adoption,
to enhance the quality of the infrastructure of telephone lines, is shown to have significant
72 S.K. MAJUMDAR

firm-level payoffs in productive efficiency terms. This finding justifies the large amount
of contemporary investment spending being made.
Digital technology does not involve radical technical change that can fully destroy
firms' competencies. Rather, it involves modular technical change that makes its impact
on firms less dramatically. Because modular change does not render obsolete the linkages
that exist between components of a system, the technology that is being adopted can
be blended in through the use of existing system-management capabilities. The blending,
however, may not be cost free if the transition involves a direct bypass from the
electromechanical generation to the digital generation by skipping the entire analog
electronic generation.
With respect to blending in, Rosenberg ([40], pages 25-26) remarks that: "it makes
an enormous difference whether a new technology requires the purchase and introduc-
tion of new equipment (especially when such equipment involves large fixed costs) or
whether it can be added on or introduced as modification to existing equipment. The
prospects for technology blending will be very much shaped by the ease with which
new technology can be introduced without having to scrap the old. In the extreme case,
if a new technology requires the complete scrapping of an old one in order to take
advantage of it, no blending is possible." The diffusion of digital technology does not
require system scrapping, and augments the scale economies enjoyed by a telephone
system. Also, once diffusion has commenced, the costs involved with adoption are
reduced considerably, as the learning process with respect to how to blend digital
equipment into the network increases over time.
Among the firm-specific control variables S I B L I N G S is negative and significant in
model 1, but positive in models 4 and 5. D E B T is negative and significant in all the
models, while C U S T O M E R E X P E N S E is positive in all models and significant in models
1 and 4. The presence of a number of operating entities under a local operating company
group, measured by the S I B L I N G S variable, can engender experience sharing resulting
in higher efficiency scores for the other entities in the group but there also seem to be
coordination problems in exploiting these possibilities. With respect to the D E B T
variable, greater equity in the capital structure makes firms perform better. This result
suggests that the increasing presence of institutional investors, who do not suffer from
collective action problems plaguing individual shareholders in the ability to monitor
firms, may induce efficiency because firms are under continuous scrutiny.
Another factor may be relevant. In the 1990s, the telecommunications industry has
become highly watched, and the attention paid can have a positive impact in engendering
efficiency. The present industry is one where the possibilities of high returns exist, and
firms are under continuous pressures to achieve these returns. One way to do so is the
enhancement of productivity. Therefore, firms may find themselves at the receiving end
of pressures to achieve superior performance. For firms undertaking higher levels of
customer-development activities in comparison with their compatriots, a market-ori-
ented predilection does result in superior performance as the sign and significance of
the C U S T O M E R E X P E N S E coefficient reveals.
Among the industry-related control variables, A C C E S S is negative and significant
in all models. The presence of relatively larger streams of free cash flow, via the earning
of access revenues do seem to induce a relatively quiet life, with negative consequences
on the abilities of firms to be efficient. S W I T C H S H A R E is found to be negative, as
expected, in all of the regression models and highly significant in model 2, while the
positive sign and significance of BUSINESS L I N E S remains stable in models 2, 4, and
5. The findings for S W I T C H S H A R E and BUSINESS L I N E S are consistent with the
MODULARITY AND PRODUCTIVITY 73

assumptions of industrial organization theory as well as the ideas of Schumpeter [9].


T O L L R E V E N U E is negative, being significant in model 2; the results imply that
spillovers of skills from the provision of local services to the provision of toll services
do not seem to be taking place but much more work is necessary on this issue of whether
scope economies do exist. Finally, the differences among the ownership dummies reveal
the presence of relative heterogeneity among the individual local operating companies'
economic performance.

Conclusion
Based on analysis of contemporary data, the extent of digital technology diffusion
is shown to significantly influence the productive efficiency of firms in the U.S. telecom-
munications industry. Energizing the information superhighway via the spread of digital
technology within firms enables not only the addition of new services, and functionalities
of existing services to be augmented, but engenders operating efficiencies in the telecom-
munications network.
The enjoyment of operating efficiencies is important for the firms as well as for other
players, since many entrants are seeking interconnection to local operating companies'
networks. As line digitalization continues to spur efficiency gains, the benefits provided
will be enjoyed not only by existing customers but by new customers who will be
interconnected to the network. Hence, the welfare consequences of the spread of digital
technology in the U.S. telecommunications industry as a whole are going to be significant.
Finally, there is a relative absence of firm-level work looking at the consequences
of technology diffusion on productive efficiency. Using a novel efficiency estimation
technique, this article demonstrates how similar studies can be conducted in the future.
The technique used enables the isolation of firm-specific efficiency indices. Thereafter,
the assessment of the impact of technology deployment choices on such efficiency
parameters sheds light on the economic performance of firms which are key actors in
the industrial sector of any economy.

References
1. Dasgupta, P.: The Theory of Technological Competition, in New Developments in the Analysis of Market
Structure. J. E. Stiglitz and G. F. Mathewson, eds., The MIT Press, Cambridge, MA, 1986.
2. Kamien, M., and Schwartz, N. L.: Market Structure and Innovation. Cambridge University Press, New
York, 1982.
3. Scherer, F. M., and Ross, D.: Industrial Market Structure and Economic Performance. 3rd ed., Houghton
Mifflin, Boston, MA, 1990.
4. Thirtle, C., and Ruttan, V. W.: The Role of Demand and Supply in the Generation and Diffusion of
Technical Change. Harwood Academic Publishers, New York, 1987.
5. Domar, E. D.: Total Factor Productivity and the Quality of Capital, Journal of Political Economy 71(6),
586-588 (1963).
6. Link, A. N.: Technological Change and Productivity Growth. Harwood Academic Publishers, New
York, 1987.
7. Salter, W. E. G.: Productivity and Technical Change. 2nd ed., Cambridge University Press, Cambridge, 1966.
8. Schmookler, J.: The Changing Efficiency of the American Economy, Review of Economics and Statistics
34(3), 214-231 (1952).
9. Schumpeter, J. A.: Capitalism, Socialism and Democracy. Harper and Row, New York, 1976.
10. Solow, R. M.: Technical Change and the Aggregate Production Function, Review of Economics and Statistics
39(3), 312-320 (1957).
74 S.K. MAJUMDAR

l 1. Nelson, R. R., and Winter, S. G.: An Evolutionary Theory of Economic Change. Harvard University Press,
Cambridge, MA, 1982.
12. Karshenas, M., and Stoneman, P. L.: Rank, Stock, Order and Epidemic Effects in the Diffusion of New
Process Technologies: An Empirical Model, RAND Journal of Economics 24(4), 503-528 (1994).
13. Tushman, M. L., and Nelson, R. R.: Introduction: Technology, Organizations and Innovation, Administrative
Science Quarterly 35(March), 1-8 (1990).
14. Clark, K. B.: Investment in New Technology and Competitive Advantage, in The Competitive Challenge.
D. J. Teece. ed., Harper and Row, New York, 1987.
15. Henderson, R. M., and Clark, K. B.: Architectural Innovation: The Reconfiguration of Existing Product
Technologies and the Failure of Established Firms, Administrative Science Quarterly 35(March), 9-30 (1990).
16. Flamm, K.: Technological Advance and Costs: Computers versus Communications, in Changing the Rules':
Technological Change, International Competition and Regulation in Communications. R. W. Crandall and
K. Flamm, cds., The Brookings Institution, Washington D.C., 1989.
17. Green, J. H.: The Business" One Irwin Handbook of Telecommunications. 2nd ed., Business One Irwin,
Homewood, IL, 1992.
I8. Frankel, M.: Obsolescence and Change in a Maturing Economy, American Economic Review 45, 296-
319 (1955).
19. David, P. A., and Bunn, J. A.: The Evolution of Gateway Technologies and Network Evolution: Lessons
from Electricity Supply History, Information Economics and Policy 3, 165-202 (1988).
20. Federal Communications Commission: Statistics of Communications Common Carriers. (Annual) FCC,
Washington, D.C.
21. Majumdar, S. K.: Does New Technology Adoption Pay? Electronic Switching Patterns and Firm-Level
Performance in U.S. Telecommunications, Research Policy 24, 803-822 (1995).
22. Waverman, L.: U.S. Inter-exchange Competition, in Changing the Rules': Technological Change, Interna-
tional Competition and Regulation in Communications. R. W. Crandall and K. Flamm, eds., The Brookings
Institution, Washington D.C., 1989.
23. Banker, R. D., Charnes, A., and Cooper, W. W.: Some Models for Estimating Technical and Scale
Efficiencies in Data Envelopment Analysis, Management Science 30(9), 1078--1092 (1984).
24. Hsiao, C.: Analysis of Panel Data. Cambridge University Press, New York, 1986.
25. Charnes, A., Cooper, W. W., and Rhodes, E. L.: Measuring the Efficiency of Decision Making Units,
European Journal of Operations Research 2(6), 429 444 (1978).
26. Farrell, M. J.: The Measurement of Productive Efficiency, Journal of the Royal Statistical Society, Series
A (General), 120(3), 253-281 (1957).
27. Dorfman, R., Samuelson, P. A., and Solow, R. M.: Linear Programming and Economic Analysis. McGraw
Hill, New York, 1958.
28. Hicks, J. R.: The Assumption of Constant Returns to Scale, Cambridge Journal of Economics 13, 9-17
(1989).
29. Majumdar, S. K., and Chang, H. H.: Scale Efficiencies in U.S. Telecommunications: An Empirical Investiga-
tion, Managerial and Decision Economics 17, 301-318 (1996).
30. Seiford, L. M., and Thrall, R. M.: Recent Developments in DEA: The Mathematical Programming Approach
to Frontier Analysis, Journal of Econometrics 46, 7-38 (1990).
31. Majumdar, S. K.: Divestiture and Productive Efficiency in the U.S. Telecommunications Industry, European
Transactions on Telecommunications 6(4), 385-395 (1995).
32. Skoog, R. E.: The Design and Cost Characteristics of Telephone Networks'. AT&T Bell Telephone Labora-
tories, Murray Hill, NJ, 1980.
33, Egan, B.: Information Superhighways: The Economics of Advanced Public Telecommunications Networks.
Artech House, Norwood, MA, 1991.
34. Zanfei, A.: Collaborative Agreements and Innovation in the U.S. Telephony Industry, in The Economics
of Information Networks. C. Antonelli, ed., North Holland, Amsterdam, 1992.
35. Antonelli, C.: The Diffuz'ion of Advanced Telecommunications in Developing Countries. Organization for
Economic Cooperation and Development, Paris, 199l.
36. Majumdar, S. K.: The Determinants of Investment in New Technology: An Examination of Alternative
Hypotheses, Technological Forecasting and Social Change 50(3), 153-165 (1995).
37. Weisman, D. L.: Default Capacity Tariffs: Smoothing the Transitional Regulatory Asymmetries in Telecom-
munications Markets, Yale Journal of Regulation 5, 149-178 (1988).
38. Bolter, W. G., McConnaughey, J. W., and Kelsey, F. J.: Telecommunications Policy for the 1990s and
Beyond. M. E. Sharpe, Inc., Armonk, NY, 1990.
MODULARITY AND PRODUCTIVITY 75

39. Majumdar, S. K.: Market Liberalization and the Psychology of Firm Performance, Journal of Economic
Psychology 15(3), 405-425 (1994).
40. Rosenberg, N.: New Technology and Old Debates, in New Technology and Development: Experiences and
"Technology Blending." A. S. Bhalla and D. James, eds., Lyenne Rienner Publishers, Boulder, CO, 1988.

Received 29 July 1996; accepted 2 January 1997

You might also like