Firm heterogeneity in complex problem solving: A knowledge‐based look at invention

• Date: 01 September 2017
• DOI: http://doi.org/10.1002/smj.2615
• Published date: 01 September 2017
• Author: Turanay Caner,Frits Pil,Susan K. Cohen

Strategic Management Journal

Strat. Mgmt. J.,38: 1791–1811 (2017)

Published online EarlyView 22 December 2016 in WileyOnline Library (wileyonlinelibrary.com) DOI: 10.1002/smj.2615

Received 7 May 2013;Final revision received7 October 2016

FIRM HETEROGENEITY IN COMPLEX PROBLEM

SOLVING: A KNOWLEDGE-BASED LOOK AT

INVENTION

TURANAY CANER,1†SUSAN K. COHEN,2†and FRITS PIL3*†

1The Peter J Tobin School of Business, St. John’s University,Queens, New York,

U.S.A.

2Katz Graduate School of Business, University of Pittsburgh, Pittsburgh,

Pennsylvania, U.S.A.

3Katz Graduate School of Business/Learning Research Development Center,

University of Pittsburgh, Pittsburgh, Pennsylvania, U.S.A.

Research summary: The knowledge-based view suggests that complex problems are best solved

under hierarchical (within-rm) governance. We examined why rms assumed to be in general

alignment with this theory might nonetheless produce solutions of varying usefulness. We

theorize that a rm’s internal knowledge variety (IKV) is associated with its capacity to

support cross-domain knowledge ows during search, and its ability to identify and explore

promising areason the solution landscape. We further theorize that partner knowledge in familiar

(unfamiliar) domains can offset specic weaknesses in searching rugged landscapes, inherent

with low or high (moderate) IKV. We nd support for these ideas in the context of drug

discovery,extending KBV’s focus on governance alignment to explain variation in problem-solving

effectiveness within hierarchy.

Managerial summary: Firms that concentrate their inventive efforts in a few technological

domains, but also dabble in several others, have problem-solving advantages: they can better

support knowledge transfer and recombination across domains. Firms that focus too narrowly

or spread their inventive efforts thinly across many domains lose these advantages, but might

compensate through alliance partnerships. Our study of drug discovery shows that while rms

with very low or high knowledge variety tend to produce weaker solutions than rms in the

moderate range, their inventive performance improves when alliance partners afford them access

to additional knowledge in familiar domains. Weexplain how the combination of rm and partner

knowledge enables rms to better identify, evaluate, and implement alternative solutions to

complex problems. Copyright © 2016 John Wiley & Sons, Ltd.

INTRODUCTION

As a perspective on why rms exist,and as an expla-

nation of their boundaries, the knowledge-based

Keywords: knowledge-based view; alliance partner

knowledge; search; complex problem solving; rm

inventive performance

*Correspondence to: Frits Pil, Katz Graduate School of Busi-

ness/Learning Research Development Center, Universityof Pitts-

burgh, 256 Mervis Hall, Roberto Clemente Drive, Pittsburgh, PA

15260, U.S.A. E-mail: fritspil@pitt.edu

†All authors contributed equally to this paper.

Copyright © 2016 John Wiley & Sons, Ltd.

view of the rm (KBV) maintains that rms have

particular advantages in governing the search for

solutions to complex, ill-structured problems1

(Leiblein and Macher, 2009; Nickerson and

1For simplicity, we use the term “complex” to refer to prob-

lems comprised of many elements with poorly understood inter-

dependencies. Limited understanding of interdependencies means

a problem cannot be fully decomposed to simplify solution search

and rms must develop theories to guide search. Problems remain

nearly or nondecomposable (i.e., ill-structured) while knowledge

of interdependencies is incomplete (Pil and Cohen, 2006; Rivkin,

2001).

1792 T. Caner, S. K. Cohen, and F. Pil

Zenger, 2004). To make the search for solutions

to complex problems efcient, “theories” or per-

spectives on the interdependencies that can affect

solution quality must be developed (Gavetti and

Levinthal, 2000). Hierarchical governance supports

rich knowledge ows across disparate domains,

providing an ideal context for these theories to

emerge (Felin and Zenger, 2014; Grant, 1996).

Consistent with the KBV, empirical evidence

suggests that rms choose to internalize the

search for solutions to complex problems and that

they realize performance benets from doing so

(Hoetker, 2005; Kapoor and Adner, 2012; Leiblein,

Reuer, and Dalsace, 2002; Macher, 2006; Macher

and Boerner, 2012). However, the KBV does not

explain why some rms produce better solutions

than others, through hierarchy.

The KBV ascribes the high-bandwidth com-

munication channels supported by hierarchical

governance to shared organizational identity or

codes (Arrow, 1974; Kogut and Zander, 1992,

1996) and to the use of low-powered incentives

and rm ownership of collaborative outputs

(Nickerson and Zenger, 2004; Williamson, 1991).

The presence of overlapping domain experience

also facilitates knowledge exchange (Lane and

Lubatkin, 1998; Szulanski, 1996) and we suggest

this helps explain rm problem-solving effective-

ness. We theorize that a rm’s internal knowledge

variety (IKV), which reects the dispersion of

inventive activity across technological domains,

affects the prevalence of shared domain knowledge

and thereby the bandwidth of the rm’s commu-

nication channels. In turn, this affects a rm’s

capacity to access external knowledge for problem

solving.

To study complex problem solving, we focus

on drug discovery, which is characterized in the

literature as a search for solutions to complex

problems (Dougherty and Dunne, 2011; Hender-

son, 1994; Pisano, 2006). The knowledge relevant

to drug discovery is rapidly evolving, and rms

draw upon knowledge from external partners to

supplement their internal knowledge. We capture

a rm’s exposure to knowledge from novel and

familiar technological domains through their

alliance partners using external knowledge distance

(EKD)— a metric indicating the degree to which a

rm and its partners patent in the same technology

classes (Jaffe, 1986; Sampson, 2007). We theorize

that access to external knowledge can ll gaps in

knowledge accessible internally, but only if it is

made available through sufciently high-bandwidth

communication channels.

We use the terminology of the NK modeling lit-

erature to describe the rugged solution landscapes

for complex problems (Kauffman, 1993). Peaks on

these landscapes represent the most useful solutions

that can be attained given the search elements (N)

rms can draw upon, and the interdependencies (K)

among those elements (Ethiraj and Levinthal, 2004;

Gavetti and Levinthal, 2000; Levinthal, 1997).

According to its level, IKV supports certain aspects

of searching rugged solution landscapes better than

others. We hypothesize how exposure to knowl-

edge through alliances can offset IKV-related weak-

nesses and augment strengths. When rms invent

with high IKV they lack domain focus and have

difculty identifying attractive peaks (i.e., they are

prone to “peak picking” mistakes, frequently ending

up with inferior solutions). We hypothesizethat low

EKD can improve a rm’s ability to pick attractive

starting points for search by adding to its domain

knowledge. Firms inventing with low IKV are

extremely domain focused and have difculty nd-

ing the highest point on the peaks they identify (i.e.,

their capacity for “peak scaling” is limited by overly

local search). Alternative perspectives on familiar

search elements (through low EKD) can augment

local search, leading to more successful efforts to

scale peaks. When rms invent with moderate IKV,

they can pick and scale peaks competently but may

still miss “peak switching” opportunities because

rm-specic theories (e.g., Felin and Zenger, 2016)

direct search toward certain peaks and obscure oth-

ers. Exposure to novel perspectives (through high

EKD) can trigger new theories about how to search

the landscape and redirect attention to higher peaks.

Using data on the inventive activities of 229

publicly traded pharmaceutical rms over a 20 year

period, we employ a within rm estimation to

capture the inuence of variation in IKV and EKD

on the usefulness of the solutions a rm produces.

To test our hypotheses, we calculate the predicted

value for inventive performance for all the cases

in our sample with low or high EKD. Then, we

assess whether inventive performance outcomes

at low, moderate, and high IKV, are statistically

different between low and high EKD. Consistent

with our theory, we nd that high (low) EKD

moderates IKV in such a way that it steepens

(attens) the inverted-U relationship between IKV

and inventiveperformance. Firms with either low or

high IKV derive greater benet when their alliances

Copyright © 2016 John Wiley & Sons, Ltd. Strat. Mgmt. J.,38: 1791–1811 (2017)

DOI: 10.1002/smj