|
ZEF Discussion Papers on Devlopment Policy 7
Pdf ko'rish
|
| bet | 22/43 | | Sana | 20.05.2024 | | Hajmi | 363,42 Kb. | | #245457 |
Bog'liq zef dp07ZEF Discussion Papers on Devlopment Policy 7
21
The complexity and novelty of these new technologies may necessitate a learning period
of some length before the positive effects of the technologies become manifest and measurable.
Avgerou (1998) suggests that, in terms of the Neo-Schumpeterian theory of long-run business
cycles, IT investments represent a new paradigm in production and technology which renders
previous knowledge of production obsolete. Accordingly, organizations will become efficient
again only after a long period of learning. Associated with this is the view that significant
benefits from IT will be achieved only if the technology is used to support fundamental
organizational changes. Other possible explanations are that difficulties in identifying the value
of information have led to the misallocation of resources and that IT has a redistributive impact,
i.e. firms that invest in IT might benefit, but overall industry productivity and output remain
unaffected.
In an attempt to reconcile this paradox, recent work has attempted to examine the IT-
productivity link using more accurate and comprehensive data from a later time period. For
instance, Brynjolfsson and Hitt (1996) use data from 367 large firms for the 1987-1991 period to
examine the following questions: (i) Are the output contributions of computer capital and
information systems (IS) staff labor-positive? (ii) Do these positive contributions persist after
accounting for depreciation and labor expenses?
29
To answer these questions, the authors jointly
estimate a system of Cobb-Douglas production functions for each of the years in their sample.
The production function is specified as:
t
it
it
it
it
j
t
it
L
S
K
C
Q
ε
β
β
β
β
β
β
+
+
+
+
+
+
=
ln
ln
ln
ln
ln
4
3
2
1
,
(7)
where
i
indexes the individual firm,
t
is the time subscript ranging from 1987-1991,
j
represents
the industrial sector,
Q
is output,
C
represents computer capital,
K
is non-computer capital,
S
is
information services (IS) labor and
L
represents all other labor. The
β
’s are coefficients to be
estimated, and
ε
is the error term.
30
Several variants of this basic equation are estimated. These alternative specifications
relax the cross-equation equality restriction on the coefficients, allow for measurement error and
the endogeneity of computer capital and are estimated on the basis of different sub-samples.
Regardless of the specification, estimates of the output elasticity of computer capital stock are
statistically significant (at least at the 5 % level), and the magnitude of the coefficient ranges
from 0.0113 to 0.0435. The effects of IS labor are less robust (i.e. not always statistically
significant) and range between 0.007 and 0.0191. Based on estimates from their baseline
specification (elasticities of 0.0169 and 0.0178 for computer capital and IS labor, respectively),
an additional dollar spent on computer capital is associated with a yearly increase in output of 81
cents, while an additional dollar spent on IS staff is associated with an increase in output of $
2.62. These figures may be contrasted with the marginal product of non-computer capital, which
29
These large firms generated 1.8 trillion dollars in output in 1991 and thus represent a substantial proportion of the
US economy.
30
The five-equation system is estimated as an SUR (seemingly unrelated regression) model. A cross-equation
equality restriction is imposed on the coefficients. Errors term are allowed to differ and to be correlated across
years. The model is estimated using iterated FGLS (feasible generalized least squares).
|
| |
