Transcript Henderson and Cockburn (1994)

Measuring Competence? Exploring
Firm Effects in Pharmaceutical
Research
Henderson and Cockburn (1994),
Strategic Management Journal, 15: 63-84
Prepared by:
Enrique, Lihong, John, Jongkuk
1
Motivation
The resource-based view of the firm suggests that inimitable
firm-level heterogeneity, or the possession of unique
competencies or capabilities may be an important source of
enduring strategic advantage.
Despite the theoretical interest in these ideas, empirical work
in the area is still at a preliminary stage.
Relatively little empirical work has attempted to combine the
richness of measures of competence derived from field work
with large-scale statistical studies of competition.
2
Objectives
To explore the nature of firm-level effects
Firm-specific, enduring sources of
heterogeneity
To explore the role of competence in
pharmaceutical research
Architectural competence
Component competence
3
Prior Research
Three conditions for an organizational competence to be
a source of competitive advantage:
Heterogeneously distributed within an industry
Impossible to buy or sell in the available factor markets at
less than its full economic value
Difficult or costly to replicate
Unique capabilities in R&D are particularly plausible
sources of competitively important competence.
There are significant and persistent differences across
firms in their ability to conduct R&D.
4
Definitions
Two classes of capabilities that might act as sources of
idiosyncratic firm advantage
Component competence (research program level)
The local abilities and knowledge that are fundamental to dayto-day problem solving
Includes ‘resources’ (Amit and Schoemaker 1993), and
‘knowledge and skills’ or ‘technical systems’ (Leonard-Barton
1992, Teece et al. 1992)
Architectural competence (firm level)
The ability to use these component competencies
Includes ‘integrative capabilities’ (Lawrence and Lorsch 1967),
‘organizational architecture’ (Nelson 1991), ‘dynamic capability’
(Teece et al. 1992), ‘invisible asset’ (Itami 1987), and ‘values
and norms’ (Leonard-Barton 1992)
5
Hypotheses
Two dimensions of component competence
Unique disciplinary expertise
• Disciplinary groups embedded within particular firms
develop deeply embedded knowledge or unique modes of
working together that make the group particularly effective
and that cannot be easily codified.
Knowledge about particular disease areas
H1: Drug discovery productivity is an increasing function of
firm-specific expertise in particular disciplinary areas.
H2: Drug discovery productivity is an increasing function of
component competence in particular disease areas.
6
Hypotheses
Two dimensions of architectural competence
The ability to access new knowledge from outside the
boundaries of the organization
The ability to integrate knowledge flexibly across disciplinary
and therapeutic class boundaries within the organization
H3: Firms with the capability to encourage and maintain an
extensive flow of information across the boundaries of the
firm will have significantly more productive drug discovery
efforts, all other things equal.
H4: Firms that encourage and maintain an extensive flow
of information across the boundaries between scientific
disciplines and therapeutic classes within the firm will have
significantly more productive drug discovery efforts, all
7
other things equal.
Specification of Model
The productivity of drug discovery
Counts of important patents that were granted in two of the three
major jurisdictions: Japan, Europe, and the USA
In science-intensive industries such as pharmaceuticals, patents are
closely related with economic profitability and market value
y=f (x, β)
y: patent count, x: a vector of inputs to the drug discovery process,
β: a vector of parameters
Assume that patent counts are generated by a Poisson process:
E(yit) = λit = exp(xit β)
log(λit) = β log(rit) + δ zit + cit
r: the R&D variables
z: a set of control variables which include measures of competitive
activity and measures of scope and scale
c: a set of variables designed to capture heterogeneous firm-level
8
competencies.
Data
Data sources
Public sources and the internal records of 10 major
pharmaceutical firms
Quantitative data
3,210 observations (1975-88) indexed by firm, research
program, and year
Qualitative data
In-depth field interview (110 individuals) to construct
the measures of organizational structure and process
9
Measures
Component competence
Firm specific expertise in particular discipline: could not be measured.
Component competence in particular disease area
• KPATS: the stock of patents previously obtained in each program
Architectural capability
PROPUB: the degree to which standing in the larger scientific
community was a dominant criterion for promotion of personnel
CROSS: the degree of communication within the program or across
programs within the firm to solve problems
GLOBAL: the degree to which global research was managed as a
seamless whole under a single director
DICTATOR: the degree to which resource allocation within research was
entirely controlled by a single individual
Control variables
Size, Shape, and scope of the research portfolio
The effects of internal and external spillovers
Therapeutic class dummies: to control for differences in opportunity
10
across classes
Results: Firm level
DV: patent output at the FIRM level
Indep. Variables
Model1
Model2
Model3
Model4
Firm Dummy
Not included
Included
Not included
Included
Architectural Capability
Not included
Not included
Included
Included
Control variables
Included
Included
Included
Included
R2
0.490
0.859
0.655
0.863
Model 2: Introducing firm-level dummy variables in the regression
substantially increases the R2 (0.4900.859)
Organizational effects (architectural capability)
Systematic differences across firms in their propensity to patent,
accounting practices, and labor market conditions
Model 3: Architectural Capability explains a substantial amount of the
variance in patenting at the firms level (R2: 0.4900.655)
Support H3 & H4
Models 2, 3, and 4 show that Firm Dummy and Architectural Capability
are not orthogonal, and so the measures of Architectural Capability are
11
the firm effects captured by the Firm Dummy.
Results: Program level
DV: patent output at the research program level
Indep. Variables
Model1
Model2
Model3
Model4
Model5
Firm Dummy
Not included
Included
Not included
Not included
Included
Component Competence
(Program level)
Not included
Not included
Included
Not included
Included
Architectural Capability (Firm
level)
Not included
Not included
Not included
Included
Included
Therapeutic class Dummies
Control variables
Included
Included
Included
Included
Included
R2
0.383
0.503
0.655
0.446
0.693
Model1: Therapeutic class effects and spillovers are strongly significant in
the regressions (R2=0.383).
Model2: Firm dummies significantly increase R2 but they are much less
important than at the firm level (R2: 0.3830.503).
Model3 supports H2: component competence has a very significant impact
on research productivity (R2: 0.3830.655).
Model4: Architectural capability only marginally improve the overall fit of
the equation in the presence of other controls for firm heterogeneity.
12
Model 5: Preferred model
Discussion and Conclusions
Differences in local capabilities may play an
important role in shaping enduring differences
between firms.
There are long-lived sources of heterogeneity in
research productivity across programs.
Econometric Identification Problems
PROPUB and DICATOR research performance
• PROPUB and DICATOR may be measures of results as much as
they are measures of causes.
13
Discussion and Conclusions
The small changes in the way in which research is managed inside
the firm appear to have major implications for its productivity.
The research efforts of firms which score the highest on the use of
publication records as an important criterion in promotion are 38% more
productive than those at bottom end of the scale.
Firm that allocate resources through a process of consensus appear to
be as much as 55% more productive than those which use a ‘dictator’.
It is puzzling that there are such large and persistent differences across
firms in these dimensions.
Explanations
The measured capabilities are fundamentally inimitable.
The failure to adopt efficient techniques for managing research reflects
agency problems.
The measures reflect the quality of the scientists recruited by the sample
companies, rather than any fundamental difference in the quality of the
information flow within the organization.
14
Discussion and Conclusions
Firm-specific knowledge is an important
source of strategic advantage.
The capability to integrate knowledge
both across the boundaries of the firm
and across disciplines and product areas
within the firm is an important source of
strategic advantage.
15