SOMETHING OLD, SOMETHING NEW: A LONGITUDINAL STUDY OF SEARCH

® Academy of Management Journai2002, Vol. 45, No. 6, 1183-1194.

SOMETHING OLD, SOMETHING NEW:A LONGITUDINAL STUDY OF SEARCH BEHAVIOR AND

NEW PRODUCT INTRODUCTION

RIITTA KATILAStanford University

GAUTAM AHUJAUniversity of Michigan

We examine how firms search, or solve problems, to create new products. Accordingto organizational learning research, firms position themselves in a unidimensionalsearch space that spans a spectrum from local to distant search. Our findings in theglobal robotics industry suggest that firms' search efforts actually vary across twodistinct dimensions: search depth, or how frequently the firm reuses its existingknowledge, and search scope, or how widely the firm explores new knowledge.

In this study, we examined how firms search, orsolve problems (Nelson & Winter, 1982), to createnew products. The ability to create new products isan important component of firm innovative capa-bilities. New products are a central mechanismwhereby organizations diversify, adapt, and rein-vent themselves in changing market and technicalconditions (Schoonhoven, Eisenhardt, & Lyman,1990). Research has also demonstrated how newproducts improve the market share, market value,and survival of firms (Banbury & Mitchell, 1995;Chaney & Devinney, 1992). Yet, despite the attrac-tiveness, firms find it difficult to create new prod-ucts. Here, we explain a firm's performance in cre-ating products as a function of its search behavior.

Organizational learning researchers have some-times argued that in their search for solutions toproblems, firms position themselves in a unidi-mensional search space that spans the spectrumfrom exploitation to exploration (March, 1991). Wesuggest that firms' search, or problem-solving ef-

We thank Avi Fiegenbaum, Jim Fredrickson, AndyHenderson, George Huber, David Jemison, Paul Mang,and Ken G. Smith for comments and discussions on theearlier versions of this work. We are also grateful to anumber of experts and researchers in the robotics indus-try for their time, insights, and suggestions. We espe-cially thank Risto Miikkulainen for assistance in writingthe programs that were used to combine and analyze thepatent and product data. Financial support from theAbell-Hanger Foundation, the Center for Customer In-sight at the University of Texas at Austin, and the Na-tional Science Foundation (Grant #0115147) is acknowl-edged. An earlier version of this paper was selected as afinalist in the Strategic Management Society Best PaperGompetition 2000.

forts, actually vary on two distinct dimensionsrather than one. Firms can vary in their degree ofuse and reuse of their existing knowledge, just asthey can vary in their exploration of new knowl-edge. We call the first dimension, which describeshow deeply a firm reuses its existing knowledge,search depth. We call the second dimension, whichdescribes how widely a firm explores new knowl-edge, search scope. In the sections that follow, wedevelop and apply this framework to the context ofnew products and argue that a firm's ability tocreate new products is determined by the indepen-dent and interactive effects of search depth andsearch scope.

CONCEPTUAL BACKGROUND

The core technical and user service features of aproduct are customarily called a product's design(Saviotti & Metcalfe, 1984). In this study, a newproduct introduction was defined as any change ina product's design. New products represent the po-tential commercial value of a firm's R&D activities;most innovations do not influence firm perfor-mance until they are introduced to the market. Aconstruct of product introductions also comple-ments other, more intermediate proxies for firminnovation, such as knowledge, R&D investment,and scientific publications. Yet relatively few lon-gitudinal studies have explored the determinants ofnew product introductions.

Search in organizations is one part of the organi-zational learning process through which firms at-tempt to solve problems in an ambiguous world(Huber, 1991). Organizations engage in a wide va-

1183

1184 Academy of Management Journal December

riety of searches: they search for superior organiza-tional designs (Bruderer & Singh, 1996), for optimalmanufacturing methods (Jaikumar & Bohn, 1992),and for the best ways to implement new innova-tions (von Hippel & Tyre, 1995). We focused hereon one specific type of search, search for new prod-ucts. Drawing on the work of Winter (1984),we defined product search as an organization'sproblem-solving activities that involve the creationand recombination of technological ideas. In takinga search perspective and viewing firms as problemsolvers, we build on research that describes prod-uct development as problem solving (e.g., Dough-erty & Hardy, 1996).

Prior work has used two notions of search,local search and distant (exploratory) search. Or-ganizations that search locally address problemsby using knowledge that is closely related to theirpreexisting knowledge bases (e.g. Helfat, 1994;Martin & Mitchell, 1998; Stuart & Podolny, 1996).At the other end of the spectrum, exploratorysearch behaviors involve a conscious effort tomove away from current organizational routinesand knowledge bases (March, 1991; Miner, Bass-off, & Moorman, 2001). Although this traditionalcharacterization of search in terms of scope—thatis, the degree to which it entails the explorationof new knowledge—is useful, it is, however, in-complete. The search efforts of firms can vary notjust in their scope (local versus distant) but alsoin their depth, which is the degree to whichexisting knowledge is reused or exploited. In thesearch for solutions to new problems, certainfirms may use some of their existing elements ofknowledge repeatedl5^ while others may usethem only once. These differences in depth ofsearch can lead to varying degrees of familiaritywith the knowledge and eventually have impli-cations for firms' ability to craft new solutions.Huber presented a similar argument: "No distinc-tion has been made between focused search forsolutions and focused search for informationabout already identified solutions" (1991: 99). Insum, although the different levels of explorationof new knowledge have been studied in somedetail (e.g., Rosenkopf & Nerkar, 2001), relativelylittle is known about the different levels of ex-ploitation of existing knowledge.

Thus, in this work we propose that instead of asingle dimension that represents a trade-off be-tween exploitation and exploration, there actuallyexist two distinct underlying dimensions of search,which we call depth and scope. In the following

section, we develop hypotheses about the effects onnew product introductions of firms' choices alongthese two search dimensions.

HYPOTHESES

Our general proposition is that search depth andscope are distinct dimensions of search and that thechoice to pursue one or both of these dimensionsaffects a firm's ability to introduce new products. Inthe following hypotheses, nev/ product innovationis defined as the number of new products a firmintroduces. Search depth is defined as the degree towhich search revisits a firm's prior knowledge.Search scope is defined as the degree of newknowledge that is explored.

Search Depth

Increase in the depth of search can positivelyaffect product innovation thiough three kinds ofexperience effects. First, using the same knowledgeelements repeatedly reduces the likelihood of er-rors and false starts and facilitates the developmentof routines, making search more reliable (Levinthal& March, 1981). Increased experience is also likelyto make a search more predictable, as the knowl-edge to be searched is familiar and the require-ments the product should meet are better under-stood. Consequently, the product development taskcan be effectively decomposesd into solvable sub-problems, activities can be sequenced in efficientorder, and unnecessary steps can be eliminated(Eisenhardt & Tabrizi, 1995). Third, repeated usageof a given set of concepts can lead to significantlydeeper understanding of those concepts and boost afirm's ability to identify valuable knowledge ele-ments within them, to develop connections amongthem, and to combine them in many different andsignificant ways that are not apparent to less expe-rienced users of those concepts.

Excessive depth can also have negative conse-quences. The literature identifies at least two neg-ative effects of excessive depth: limits to improve-ment along a technological trajectory, and rigidity(Argyris & Schon, 1978; Dosi, 1988). We argue be-low that these negative effects of depth at somepoint exceed the benefits discussed above and,thus, the relationship between depth and innova-tion is, indeed, nonlinear. The first negative effectof depth is a consequence of diminishing returns tobuilding on the same knowledge. It seems that im-provement by focusing on the same knowledgeelements is possible only until the intrinsic perfor-mance limit of that knowledge trajectory is encoun-

2002 Katila and Ahuja 1185

tered (Dosi, 1988). When the limits of the trajectoryare approached, benefits from subsequent productdevelopment efforts increase at a declining rate. Atsome point, further developments based on thesame knowledge elements become increasingly ex-pensive and the solutions excessively complicated,leading to the costs of depth eventually exceedingits benefits. Second, depth can also hurt searchsince beyond a point, reusing existing knowledgecan make an organization rigid: solutions and prob-lem-solving strategies that once made firms greatcan turn into problems to be resolved. For example,to preserve a status quo, organization membersmay try to hide problems related to the approachthat has been traditionally used (Argyris & Schon,1978). Thus, rigidity can eventually lead to a de-crease in product output.

Consequently, we propose that the number ofnew products will first increase with depth insearch but that, beyond a point, additional depth insearch will cause a fall in product output.

Hypothesis 1. Search depth is curvilinearly(taking an inverted U-shape) related to thenumber of new products introduced by a firm.

Search Scope

Search with high scope affects product innova-tion positively through at least two mechanisms.First, search with high scope enriches the knowl-edge pool by adding distinctive new variations.New variations are necessary to provide a sufficientamount of choice to solve problems (March, 1991).Evolutionary theorists call this effect the "selectioneffect of variation." Second, search scope increasesa firm's number of new products through enhanc-ing recombinatory search (Fleming & Sorenson,2001; Nelson & Winter, 1982). There is a limit to thenumber of new ideas that can be created by usingthe same set of knowledge elements. An increase inscope adds new elements to the set, improving thepossibilities for finding a new useful combination.

However, the literature also suggests two nega-tive consequences of extremely high levels ofscope: dynamically increasing knowledge integra-tion costs, and decreasing reliability. First, highscope can hurt innovativeness through the dynam-ically increasing costs of integrating new knowl-edge. As the amount of search scope and, conse-quently, the proportion of new knowledge to beintegrated into a firm's knowledge base increases,so do the technological and organizational chal-lenges in integration. Technologically, common in-terfaces need to be established among knowledge

elements. Organizationally, new knowledge re-quires changes in networks of relations and com-munication relationships both within and outsidean organization (Henderson & Clark, 1990). In priorwork, it has been argued that the wider the scope ofthe knowledge to be integrated, the more complexare the problems of creating and managing integra-tion (Grant, 1996: 377). Thus, eventually, the costsof integration will exceed the benefits of acquiringnew knowledge.

Second, researchers have argued that excessiveincrease in search scope can hurt product outputthrough decreasing reliability (e.g., Martin & Mitch-ell, 1998). A firm's reliability (its ability to respondto new information correctly) is "a negative func-tion [of distance] from an agent's immediate expe-rience or from its local environmental situation"(Heiner, 1986: 84). Thus, innovation projects inwhich the proportion of new knowledge is high areless likely to succeed than projects that searchclosely related knowledge (Cyert & March, 1963).

The mechanisms described above suggest thatsearch scope, as measured by the proportion of newknowledge elements in search, is curvilinearly re-lated to subsequent product innovation. The fol-lowing hypothesis is proposed:

Hypothesis 2. Search scope is curvilinearly(taking an inverted U-shape) related to thenumber of new products introduced by a firm.

Combination of Depth and Scope

The above hypotheses focus on the distinct ef-fects of depth and scbpe on innovation. In thissection, we propose that these variables are mutu-ally beneficial and have interactive effects. We sug-gest two mechanisms that underlie this positiveinteraction: absorptive capacity and uniqueness.

The absorptive capacity literature discusseshow firms can use their accumulated knowledgeto recognize and assimilate new knowledge (e.g.,Cohen & Levinthal, 1994). Relatedly, Winter(1984: 293) argued that the new knowledge firmstypically obtain by searching their external environ-ments is a collection of fragments of possibly usefulknowledge. However, the number and quality ofthese fragments are likely to be less than what isneeded, and therefore assimilation and further devel-opment of novelty require complementary problem-solving efforts by the acquiring firms. Thus, existingknowledge may facilitate both the absorption andfurther development of new knowledge, suggesting apositive relationship between relatively high levels ofdepth and scope, and product irmovation.


A combination of depth and scope search canalso increase the uniqueness of recomhinations. InHypothesis 2, we discuss the relationship betweenscope and new product output. However, increasesin scope can be costly: the probability of findingvaluable new knowledge elements is small and,even if a firm succeeds in doing so, it is possiblethat the same product idea has already been dis-covered. By combining firm-specific accumulatedunderstanding of certain knowledge elements(depth) with new solutions (scope), firms are morelikely to create new, unique combinations that canbe commercialized (Winter, 1984: 293),

Hypothesis 3. The interaction of search depthand scope is positively related to the number ofnew products introduced by a firm.

METHODS

Sample and Data

The research sample was drawn from the pop-ulation of industrial robotics companies in Eu-rope, Japan, and North America, A number ofconsiderations motivated the choice of industrialrobotics as the setting of the study. The highresearch intensity of the robotics industry makesit a good place to analyze the effects of newproduct search (Katila, 2000), And since roboticstechnology is a combination of multiple rapidlychanging technological disciplines such as elec-tronics, new materials, and optics, and since theindustry lacks product standards (Dahlin, 1993),search activities in the industry require complexproblem solving. Finally, although the diffusionof robots and the effects of robotization on work-ers are well documented, few studies have exam-ined how robots are developed.

We searched through robotics trade magazinesand catalogues and talked to industry experts toform a comprehensive list of companies in the in-dustrial robotics industry. This method assuredthat we were not sampling on the dependent vari-able: all relevant companies were included,whether or not they were innovative. Availabilityof yearly data on the control variables reduced thefinal sample to 124 firms, of which 78 were Japa-nese, 27 were American, and 19 were European.The firms varied widely in size: the average firmhad 39,000 employees, and the smallest had fewerthan 100 employees. Industry entry and exit datafor each company were collected from Predicasts,trade journals, and industry reports.

We used two main data sources: new product

introduction announcements and patent data. Inassembling the product data, a method intro-duced by Coombs, Narandren, and Richards(1996) was applied. Following this method, weobtained product introductions and their charac-teristics from both editorially controlled newproduct announcement sections of robotics tech-nical and trade magazines and from roboticsproduct catalogues. The use of several sources fora single introduction assured the reliability of thedata. For a final verification, we contacted oursample companies and asked them to verify theirindividual product records in our data. To collectthe patent data for the search variables, we useddata from the United States Patent and Trade-mark Office, We went througli Who Owns Whom,a set of directories, to create the patent portfoliosfor each firm. Yearly patent data, by applicationdate, were used.

Measures

Dependent variable: Number of new products.To measure the dependent variable we used Mar-tin and Mitchell's (1998) definition of a newproduct as change in design characteristics, Arobot was defined as new if one or more of itsdesign characteristics differed from those of theproducing firm's previous ]3roducts. Thus, anexisting design introduced into a new geograph-ical area, for example, did not qualify as a newproduct.

Independent variables. Obtaining data on in-trafirm problem-solving behaviors over a ten-yearperiod is a major challenge. Data with which toassess intrafirm search activities over time areusually not public, or, even if they are available,are often extremely resource-corisuming to as-semble (Cohen, 1995), In this study, we usedfirms' patenting activities to measure their depthand scope of search. Since a paterit by definitionincludes a description of a technical problem anda solution to that problem (Walker, 1995), patentdata gave us a detailed and consistent chronologyof how firms solve problems—or search. Recog-nizing these attributes, several authors have usedpatent data as an indicator of search activity (seeKatila, 2002; Rosenkopf & Nerkar, 2001; Stuart &Podolny, 1996),

Using patent data as a measure of search also hassome limitations. Previous studies liave shown thatthe propensity for patenting varies considerablyacross industries (e,g,, Cockburn & Criliches, 1987),However, this was not a problem in our study sincewe focused on one industry, industrial robotics.Patents have been shown to be an inaportant appro-


priability mechanism in the robotics industry(Grupp, Schwitalla, Schmoch, & Granberg, 1990)and in the industrial machinery industry in general(Arundel & Kabla, 1998; Cockburn & Griliches,1987),

The variable search depth describes accumula-tion of search experience with the same knowledgeelements. We argued above that the more fre-quently a firm has used knowledge, the moredeeply it knows it. Thus, search depth was mea-sured as the average number of times a firm repeat-edly used the citations in the patents it applied for.We created the depth variable by calculating thenumber of times that, on the average, each citationin year t—1 was repeatedly used during the pastfive years. Prior research has shown that organiza-tional memory in high-technology companies isimperfect: knowledge depreciates sharply, losingsignificant value within approximately five years(Argote, 1999), The following formula was used:

repetition countj

^ "~^

y= (-6

total

The variable search scope, which corresponds tothe theoretical notion of exploration of new knowl-edge, was the proportion of previously unused ci-tations [new citationsn_-i] in a firm's focal year's listof citations. We assessed the share of citations in afocal year's citations that could not be found in theprevious five years' list of patents and citations bythat firm. Values for this variable, which was cal-culated as follows, range from 0 to 1:

new citationSit-i'"̂ total citationSjt-i'

The use of the search depth and scope measurescan be illustrated by considering a firm with ten pat-ents. Each of the ten patents further cites ten otherpatents. On the average, eight out of the ten citationsare new to the firm; that is, it has not used themduring the past five years. The firm's search scope isthus 0,8, Of the remaining two "old" citations in eachpatent, on the average, the firm has used one of themtwice and the other three times. Thus, the searchdepth for this firm is 0,5,

Control variables. Firms often face constraintsin developing innovations in-house (e,g,, Ahuja,2000a), especially if the technology is complex.Thus, we included the number of a sample firm'sfactory automation collaborations as a control, la-beling the variable collaboration frequency.

Prior work suggests that financial performance

may affect innovation in two ways. Search theoristsargue that increase in performance encourages ex-ploration for new innovations (Levinthal & March,1981), Prospect theorists, on the other hand, pre-dict the opposite: when financial performance isgood, managers are less likely to explore (Gyert &March, 1963), We thus included return on assets(ROA) as a firm performance measure. Other inno-vation studies (e,g,, Ahuja, 2000b; Hitt, Hoskisson,Johnson, & Moesel, 1996) have used ROA as a per-formance measure, and its use in this research thusfacilitates comparisons across studies. Moreover,innovation studies have shown that return on as-sets is highly correlated with other performanceindicators, such as return on sales (e,g,, Steensma &Corley, 2000),

We used a firm's yearly R&'D expenditure in mil-lions of dollars as a proxy for a firm's total R&Dinputs to the innovation process. These data canalso describe the amount of the firm's search activ-ities (Cohen, 1995), complementing the patent dataused in predictor variables,

A firm's degree of product diversification canhave both positive and negative effects on newproducts. Since diversified firms possess more op-portunities for the internal use of new knowledge,innovativeness may increase through economies ofscope. On the other hand, as Hoskisson and Hitt(1988) showed, as firms become more diversified,corporate management understands the firm's R&Dactivities less, so innovation decreases. We used anentropy measure (Chatterjee & Blocher, 1992),

National technological characteristics such asR&D infrastructure and historical resource endow-ments, intensity of competition, and national cul-ture affect the innovative behavior of firms (e,g,.Nelson, 1993; Shane, 1992), The propensity to in-troduce new products rather than new processescan also vary across nations (Hayes & Wheelwright,1984; Teece, 1987), In this study, we controlled forsuch effects by including dummy variables for na-tionality; European firm and American firm werethe categories, with "Japanese firm" as the omittedcategory.

Previous research findings on the effects of sizeon product innovation have been mixed. Althoughmost studies have reported a positive effect of sizeon product innovation (e,g,, Chaney & Devinney,1992), some studies have shown a negative effect(Mansfield, 1968), or no effect at all (Clark, Chew, &Fujimoto, 1987), Our measure of firm size wasnumber of corporate employees.

Market conditions and the general economic en-


vironment can vary over time, making it more orless attractive to introduce new products. To con-trol for such period effects, we used year dummies(1985-95). "Year 1996" was the omitted category.

Statistical Method and Analysis

Since the dependent variable of the study, num-ber of new products, included counts of new prod-ucts, we used a panel Poisson regression (McCul-lagh & Nelder, 1989). A Poisson specificationensures that zero values of the dependent variableare incorporated into a model rather than implicitlytruncated, as they are in OLS regression. This esti-mation technique is common in new product intro-duction studies (e.g., Blundell, Griffith, & VanReenen, 1995).

To control for firm heterogeneity, we used thegeneralized estimating equations (GEE) regressionmethod. This method accounts for autocorrela-tion—owing to repeated yearly measurements ofthe same firms—by estimating the correlationstructure ofthe error terms (Liang & Zeger, 1986). Aone-period-lagged dependent variable was also in-cluded as an additional control for firm heteroge-neity (Heckman & Borjas, 1980). Additionally, toaccount for any overdispersion in the data, we re-port all results with robust, or empirical, standarderrors.

Table 1 shows the descriptive statistics and cor-relations for all variables. All independent and con-trol variables are lagged by one year. Grupp andcolleagues (1990) provided qualitative evidence ofone- to two-year lags in introductions of roboticsproducts to market. As the descriptive statistics on

the control variables indicate, the companies inthe sample differ widely in size, R&D efforts, andperformance. The low, nonsignificant correlation(-.003) between search depth and search scope isalso noteworthy; it suggests that these two variablesrepresent two distinct dimensions of search. Thecorrelation matrix suggests that the coilinearityamong the main variables is low. However, firmsize and R&D expenditure are exceptions, and sub-sequently we entered these variables into separatemodels.

A regression approach was used for testing thehypotheses. The regression analysis pertains to theyears 1985-96. As several authors have recom-mended, we centered the independent variables,search depth and search scope, on their meansbefore creating the interaction term (e.g., Gronbach,1987).

In total, 1,898 new robotics introductions wereincluded in the analysis. On the average, the sam-ple companies introduced about one new robotyearly. Some companies had no new introductionsin a given year, while others; introduced over 20new robots. About three-quarters of the patent ci-tations that the firms used were new (averagesearch scope), and each citation was repeatedlyused about 0.22 times within the next five years(average search depth).

RESULT!*

H5rpothesis Tests

Table 2 reports the results of the GEE Poissonregression analysis. Two of our three hypotheses

TABLE 1Descriptive Statistics and Correlations"

Variable

1. Number of newproducts

2. Search depth3. Search scope4. Collaboration frequency5. Firm performance6. R&D expenditure7. Diversification8. European firm9. American firm

10. Firm size

Mean

0.92

0.220.740.160.020.320.640.150.22

39.66

s.d.

2.36

0.240.300.500.050.780.370.360.41

100.77

Min.

0.00

0.000.000.00

-0.730.010.000.000.000.03

Max.

24.00

1.741.006.000.456.781.621.001.00

876.80

1

.04

.08**

.33***-.01

.07*-.02-.06*-.15***-.03

2

-.003.02.04.22***.20***

-.06*-.02

.15***

3

.07*

.03

.07*

.15***

.06*-.06*

.08**

4

.04

.07*

.03-.01

.01

.06*

5

.03-.02

.03

.04

.05

6

.11***

.17***

.05

.86***

7

.20***-.19***

.14***

8

-.22***.29***

9

.17***

' n = 1,185.* p < .05

** p < .01'** p < .001


TABLE 2Results of Regression Analysis for Number of New Products^'

Variable

Intercept

Search depth

Search scope

Search depth x searchscope

Search depth squared

Search scope squared

Collaboration frequency

Firm performance

R&D expenditure

Diversification

European firm

American firm

Number of new products(lagged)

Firm size

Model 1

-0 ,58(0,39)

0,30**(0,11)

-1 ,42(1,07)0,40**

(0,10)-0 ,50(0,36)

-1 ,14*(0,50)

-2 ,05***(0,43)

Model 2

-0 ,57(0,40)0,55^

(0,34)

0,30**(0,11)

-1 ,46(1,08)0,40*

(0,10)-0 ,54(0,38)

-1 ,08*(0,50)

-2,04***(0,43)

Model 3

-0 ,57(0,39)

0,14(0,15)

0,30***(0,11)

-1 ,39(1,08)0,40**

(0,10)-0 ,51(0,36)

-1 ,14*(0,50)

-2 ,05***(0,43)

Model 4

-0 .57(0,40)0,59*

(0,34)0,22+

(0,16)

0,30**(0,11)

-1 ,43(1.11)0,40*

(0,10)-0 ,55(0,38)

-1 ,07*(0,50)

-2 .03***(0,43)

Model 5

-0 ,57(0,41)0,95*

(0,44)1.34**

(0.49)5.69**

(2,22)

0,29**(0,10)

-1 ,44(1,14)0,40*

(0,10)-0 ,56(0,38)

- 1 , 0 5 *(0,51)

-1 ,97***(0,40)

Model 6

-0 ,630,423,07***

(1,04)0,76*

(0,44)3,10"^

(1,97)- 3 , 0 5 *(1,35)

0,25*(0,11)

-1 ,46(1,21)0,30*

(0,10)-0 ,58(0,38)

- 1 , 0 5 *(0,51)

-1 ,97***(0,40)

Model 7

-0 ,620,422,70

(3,24)1,32

(4,25)3,33^

(2,13)-2 ,71(2,92)

-0 ,51(4,05)0,25*

(0,11)-1 ,47(1,21)0,30*

(0,10)-0 ,58(0,38)

-1 ,04*(0,51)

-1 ,97***(0,41)

Model 8

-0 ,630,423,11***

(1,05)0,76*

(0,45)3,08"^

(2,01)-3 ,12*(1,40)

0,26*(0,11)

-1 ,46(1,22)0,30*

(0,10).-0 ,58(0,37)

- 1 , 0 3 *(0,50)

-1 ,95***(0,39)0.01

(0,03)

Model g

-0 ,530,423,45***

(0,96)0,85*

(0,42)3,48*

(1,83)-3 ,32**(1,22)

0,23*(0,12)

-1 ,12(1,26)

-0 ,60(0,39)

-0,97"^(0,57)

-1 ,82***(0,39)

0,0001(0,002)

Model 10

-0 ,510,423,24***

(0,90)0,82*

(0,39)3,43*

(1,70)-2 ,99**(1,09)

0,21(0,13)

-1 ,14(1,20)

-0 ,59(0,39)

-1,01+(0,58)

-1 ,86***(0,40)

-0 ,03(0,06)0,0001

(0,001)

Deviance 2,694,1 2,685,2 2,679,4 2,667,2 2,646,8 2,618,8 2,617,5Difference in log 8,9** 14,7*** 26,9*** 47,3*** 75,3*** 76,6***

likelihood vis-a-vis thebase model

df 19 20 20 21 22 23 24 24 23 24

" The table gives parameter estimates; the standard error is below each parameter estimate in parentheses,^ There were 124 firms and 1,185 firm-year observations. Year dummies were included but are not shown,

+ p < ,10* p < ,05

** p < ,01***p < ,001Two-tailed tests for controls, one-tailed tests for hypothesized variables.

were supported {Hypotheses 1 and 3), but the cur-vilinear relationship predicted in the second hy-pothesis was not fully borne out in that we found alinear rather than a curvilinear relationship be-tween search scope and new products. In the anal-ysis presented in Table 2, number of new productsis the dependent variable. The first column reportsthe baseline model in which collaboration fre-quency, ROA, R&D expenditure, diversification,and the nationality and year dummies were in-cluded as control variables. In models 2-4, weintroduced search depth and search scope to assessthose variables' possible effects on new products.

In model 5, we included the interaction of searchdepth and search scope, and in models 6 and 7 weadded the squared terms of the two search vari-ables. Although we hypothesized a curvilinear ef-fect for search scope, the inclusion of search scopesquared did not significantly improve model fit(model 7). Accordingly, we dropped the squaredterm from the final model. We therefore base ourdiscussion of the results on the full model, model 6.

In Hypothesis 1, we propose that search depthwill have a curvilinear (inverted U-shaped] rela-tionship with new product innovation. In model 6in Table 2, the coefficient for search depth is posi-


tive and that for the squared term of depth is neg-ative and significant, supporting the hypothesis.Hypothesis 2 proposes a curvilinear relationshipbetween search scope and new products. This hy-pothesis was not supported, since the squared termof scope fails to provide a good fit in model 7.However, the linear coefficient for scope in model 6is positive and significant, thus suggesting a posi-tive, linear relationship between scope and productinnovation. We examine the possible explanationsfor the linear effect of scope in the discussion sec-tion. In Hypothesis 3, we predict that search depthand search scope leverage each other, yielding acombined positive effect on product innovation.The estimated positive interaction between depthand scope in model 6 provides support for thishypothesis. Finally, the log-likelihood statisticsprovide evidence that adding depth and the inter-action variables (models 4-6) significantly im-proves the model fit over the model with the scopevariable only (model 3), supporting the idea thatsearch is indeed a two-dimensional construct. Theeffects are also substantively significant. For a hy-pothetical firm at the mean of the depth (0.22) andscope (0.74) variables, a 10 percent increase indepth (an increase of 0.022) leads to a 10 percentincrease in new product introduction. For the samefirm, a 10 percent increase in search scope (anincrease of 0.074) leads to an 11 percent increase innew product introduction.

Overall, the effects of the control variables wereas expected: R&D expenditure and collaborationfrequency were found to increase the number ofrobotics product introductions. The result that Jap-anese robotics firms innovated more than theircompetitors in Europe and in the United Statessupports previous findings reported in the litera-ture (e.g., Mansfield, 1988).

Sensitivity Analyses

The sensitivity of the results was also tested inseveral ways. We tested the sensitivity of the re-gression models by including additional measuresof the industry-level effects. Adding a variable fordemand for industrial robots, measured as yearlyworldwide industrial robot installations, did notchange the results. To examine the effect of firmsize on product introductions, we tested the modelby including firm size instead of the R&D expendi-ture measure. The results from this sensitivity test,reported in model 9 of Table 2, consistently sup-port the main findings of the study. We also ranadditional analyses by (1) including both R&D and

firm size in the same model, (2) replacing R&Dexpenditure with R&D intensity (R&D divided bysales; Helfat, 1994), and (3) adding a size-R&D fac-tor (using factor analysis to reduce the highly cor-related firm size and R&D viariables to a singlefactor). We also modified the models by controllingfor the number of firms that efich sample firm hadacquired. This control was included since acquisi-tions can potentially substitute for internal innova-tion search (Ahuja & Katila, 2001). All of these testsconsistently supported the main re.sults.

We also used several alternative data sources totest the validity of our search depth and scopemeasures. In addition to the U.S. patent citationsused in the original results, we also conducted theanalysis by including both U.S. and foreign citationdata. We also measured the search variables byincluding six past years of patents (instead of thefive years used in the original measures) and byexcluding the company's self-citations from thesearch measures to account for any differences be-tween external and internal citations. We also usedrobotics application area data to substitute for thepatent data used in the search measures (Interna-tional Federation of Robotics, 1999). All of theseresults exhibited a pattern similar to that of theresults reported previously.

Finally, we tested the robustness of the resultsagainst unobserved heterogeneity. Theoretically,the organizational learning and search literaturesare based on the premise that firms differ in theirsearch behaviors and that most firnis do not searchin perfect ways: organizations learn at differentrates and forget (Argote, 1999), and their searchactions are inertial and rationality-bounded (Cyert& March, 1963). We used three separate approachesto control for such unobserved heterogeneity. First,we included proxy variables to capture unobservedinfluence; this is most comn:ionly done by usingprevious values of the dependent variable as anadditional regressor. Second, we modeled the un-observed heterogeneity parametrically, by assum-ing a statistical distribution. Third, we correctedand controlled for the serial correlation that arisesif unobserved heterogeneity is not directly ac-counted for.

In the first approach, we constructed two types ofproxy variables: the lagged dependent variable(Heckman & Borjas, 1980) and the presample vari-able (Blundell et al., 1995). In Table 2 in model 8we report the results obtained using a lagged de-pendent variable method. We also performed ananalysis by including a presample control variable.This presample covariate was constructed from the


dependent variable values in the periods immedi-ately preceding the study period and served as afixed effect for the firms in the panel. Both of theseresults strongly supported the original findings.Second, we applied a commonly used parametricapproach to handling unobserved heterogeneity inPoisson regressions: we presumed that the unob-served error followed a gamma distribution andestimated a negative binomial model. The negativebinomial results again exhibited the same patternas the original results. Third, as discussed earlier,to account for any remaining serial correlation weused a generalized estimating equations (GEE)method in all models.

DISCUSSION

In this study, we examined how firms search fornew products and made a distinction based on twodimensions of search: depth and scope. We havedescribed the mechanisms underlying the differentsearch approaches in detail and distinguished theireffects on performance.

This study has theoretical implications for theorganizational learning and resource-based per-spectives. From the perspective of the organiza-tional search and learning literature, we contributeto knowledge on new product search and its com-ponents. Few authors have examined the perfor-mance effects of search approaches as is done inour study. This contribution is important, since, asArgote stated, "We know relatively more aboutknowledge retention and transfer than we knowabout knowledge creation in organizations" (1999:203).

Prior work on search has frequently focused onthe exploration/exploitation dichotomy (e.g.,March, 1991). A key contribution of our study isthe idea that exploitation is a more comprehen-sive concept than it is usually considered to be.Specifically, we have distinguished levels of ex-ploitation, or search depth. We argue that firmscan differentiate themselves not only as to theextent to which they explore new things, but alsoas to the extent to which they master the old ones.Thus, we extend the unidimensional conceptof exploitation versus exploration into a two-dimensional framework. Relatedly, we draw at-tention to the fact that exploitation is important,not just for fine-tuning and economizing the ef-ficiency of an existing technology (Levinthal &March, 1981: 311), but also for creating newknowledge. Although exploratory search has a

key role in knowledge creation, that of providingcompletely new solutions, exploitation also has arole, that of combining existing solutions to gen-erating new combinations (Schumpeter, 1934).

This study also expands the work on the dy-namic capabilities of firms by examining in detailone such capability: that of problem solving, orsearch. The study supports the notion that afirm's dynamic problem-solving capabilities canbe an important source of resource heterogeneity.Our results also indicate that firms differ in howthey search and that these variations can lead tovariations in performance. Further, finding a sta-tistically significant effect of the interaction termof depth and scope suggests that at least someorganizations are able to engage in both searchapproaches simultaneously. Thus, this studycontributes to an understanding of search pro-cesses within organizations. However, we exam-ined search processes largely through archivalpatent data. Although this limitation was almostunavoidable in our longitudinal setting, it sug-gests the need for future research using comple-mentary approaches to measure search, such assurveys and case studies. Understanding the dif-ferences between organizations that manage theproductive combination of scope and depth inrelation to those that fail may be a fruitful direc-tion for further research in this area.

The unexpected result of this study, the lineareffect of search scope on new product innovation,instead of the expected nonlinear effect, deservesmore attention. A possible explanation is that inthe empirical sample of this study, few companies"oversearched" along this dimension, because in-tensive scope search was costlier than extensivedepth search. Consequently, instead of a curvilin-ear relationship, only the linear, increasing part ofthe curve was detected. This result is also in linewith the proposition that firms search locally (Hel-fat, 1994; Stuart & Podolny, 1996), and with thehuman tendency to try to reduce uncertainty. Wewould expect this tendency to be especially strongin innovation search when firms need to decidebetween searching in known, well-tried directionsand searching in uncertain, new directions. AsThompson observed: "If such tendencies appear inpuzzle-solving as well as in everyday situations, wewould especially expect them to be emphasizedwhen responsibility and high stakes are added"(1967: 4). More explanations for this result shouldbe explored in future work.

The robotics industry was an interesting empiri-


cal setting for the study, given the complexity of theinnovation search activities in the industry and thelack of prior large-scale organizational research inthe industry. Although our findings are likely togeneralize to other high-technology industries,such as Pharmaceuticals, telecommunications, andthe computer industry, where search problems arecomplex and the ability to bring new products tothe market is a key determinant of success, futureresearch to confirm how the framework developedin this study applies to these industries would behelpful. Further research is also needed to examinethe usefulness of our propositions for other types oforganizational search.

A better understanding of new product searchalso has important implications for managers. Mas-tering internal search can provide a source for com-petitive advantage: internal capabilities such as theability to search effectively can be a more stablebasis for strategy formulation than those acquiredexternally (Grant, 1996). Two aspects of this studymake it especially useful in this context. First, priorresearch suggests that many organizations arelikely to develop a natural tendency to specialize inone form of search behavior—they either exploit orthey explore (Levinthal & March, 1993). Thisstudy's arguments and results draw managers' at-tention to the notion that the most fruitful approachlies at the intersection of these two activities andthat some firms are able to achieve this balance andare rewarded for it. This reminder of the impor-tance of balance is especially relevant in the con-text of the two mutually conflicting prescriptionsoften presented to practitioners: "stick to the knit-ting" and work to improve, or cast aside all that isfamiliar and work to develop revolutionary newproducts and modes of thought. In contrast, wesuggest that search is most likely to be productivewhen it uses both familiar and unfamiliar elements.Second, we provide a practical mechanism manag-ers can use to monitor the degree to which a firm isable to maintain this balance, both longitudinallyand cross-sectionally. The patent-based metrics fordepth and scope developed in this study can becomputed from public information for any firm andits leading competitors over time, and it can helpprovide a frame of reference that managers can usefor tracking, focusing, and redirecting search efforts.

REFERENCES

Ahuja, G. 2000a. The duality of collaboration: Induce-ments and opportunities in the formation of inter-

firm linkages. Strategic Management Journal,21(special issue): 317-343.

Ahuja, G. 2000b. Collaboration networks, structuralholes, and innovation: A longitudirial study. Admin-istrative Science Quarterly, 45: 425-455.

Ahuja, G., & Katila, R. 2001. Technological acquisitionsand the innovation performance of acquiring firms:A longitudinal study. Strategic Management Jour-nal, 22: 197-220.

Argote, L. 1999. Organizational learning: Creating, re-taining and transferring knowledge. Boston: Klu-wer Academic.

Argyris, G., & Schon, D. 1978. Organizational learning.Reading, MA: Addison-Weskiy.

Arundel, A., & Kabla, I. 1998. What percentage of inno-vations are patented? Empirical estimates for Euro-pean firms. Research Policy, 27: 127-141.

Banbury, G., & Mitchell, W. 1995. The effect of introduc-ing important incremental innovations on marketshare and business survival. Strategic ManagementJournal, 16(special issue): 161-182.

Blundell, R., Griffith, R., & Van Reenen, J. 1995. Dynamiccount data models of technoLjgical innovation. Eco-nomic Journal, 105: 333-344.

Bruderer, E., & Singh, J. 1996. Organizcitional evolution,learning, and selection: A genetic-algorithm-basedmodel. Academy of Management Journal, 39:1322-1349.

Ghaney, P., & Devinney, T. 1992. Nevif product innova-tions and stock price perform.ance. Journal of Busi-ness Finance &• Accounting, 19: 677-695.

Ghatterjee, S., & Blocher, J. 1992. Measurement of firmdiversification: Is it robust? Academy of Manage-ment Journal, 35: 874-888.

Glark, K., Ghew, B., & Fujimoto, T. 1987. Product devel-opment in the world auto industry. Brookings Pa-pers on Economic Activity, special issue on micro-economics (3): 729-771.

Gockburn, I., & Griliches, Z. 1987. Industry effects andappropriability measures La the stock market'svaluation of R&D and patents. Working paper no.2465, NBER, Gambridge, MA.

Gohen, W. 1995. Empirical studies; of innovative activity.In P. Stoneman (Ed.), Handbook of the economics ofinnovation and technological change: 182-264.Oxford, England: Blackweil.

Gohen, W., & Levinthal, D. 1994. A fortune favors theprepared firm. Management Science, 40: 227-251.

Goombs, R., Narandren, P., & Richards, A. 1996. A liter-ature-based innovation output indicator. ResearchPolicy, 25: 403-413.

Gronbach, L. 1987. Statistical tests for moderator vari-


ables: Flaws in analyses recently proposed. Psycho-logical Bulletin, 102: 414-417,

Cyert, R., & March, J. 1963. A behavioral theory of thefirm. Englewood Cliffs, NJ: Prentice-Hall.

Dahlin, K. 1993. Diffusion and industrial dynamics in therobot industry. International Journal of TechnologyManagement, 8(3-5): 259-281,

Dosi, G, 1988. Sources, procedures, and microeconomiceffects of innovation. Journal of Economic Litera-ture, 26: 1120-1171.

Dougherty, D., & Hardy, C. 1996. Sustained product in-novation in large, mature organizations: Overcominginnovation-to-organization problems. Academy ofManagement Journal, 39: 1120-1153.

Eisenhardt, K. M., & Tabrizi, B. N. 1995. Acceleratingadaptive processes: Product innovation in the globalcomputer industry. Administrative Science Quar-terly, 40: 84-110.

Fleming, L., & Sorenson, O. 2001. Technology as a com-plex adaptive system: Evidence from patent data.Research Policy, 30: 1019-1039.

Grant, R. 1996. Prospering in dynamically-competitiveenvironments: Organizational capability as knowl-edge integration. Organization Science, 7: 375-387.

Grupp, H., Schwitalla, B., Schmoch, U., & Granberg, A.1990. Developing industrial robot technology inSweden, West Germany, Japan, and the U.S.A. InJ. Sigurdson (Ed.), Measuring the dynamics oftechnological change: 106-129. London: PinterPublishers.

Hayes, R., & Wheelwright, S. (Eds.) 1984. Restoring ourcompetitive edge: Competing through manufactur-ing. New York: Wiley.

Heckman, J., & Borjas, G. 1980. Does unemploymentcause future unemployment? Economica, 47: 247-283.

Heiner, R. 1986. Uncertainty, signal-detection experi-ments, and modeling behavior. In R. Langlois (Ed.),Economics as a process: Essays in the new institu-tional economics: 59-115. Gambridge, England:Gambridge University Press.

Helfat, G. 1994. Evolutionary trajectories in petroleumfirm R&D. Management Science, 40: 1720-1747.

Henderson, R., & Glark, K. 1990. Architectural innova-tion: The reconfiguration of existing product tech-nologies and the failinre of established firms. Admin-istrative Science Quarterly, 35: 9-30.

Hitt, M., Hoskisson, R., Johnson, R., & Moesel, D. 1996.The market for corporate control and firm innova-tion. Academy of Management Journal, 39: 1084-1119.

Hoskisson, R., & Hitt, M. 1988. Strategic control systems

and relative R&D investment in large multi-productfirms. Strategic Management Journal, 9: 605-621.

Huber, G. 1991. Organizational learning: The contribut-ing processes and a review of the literatures. Organ-ization Science, 2: 88-115.

International Federation of Robotics. 1999. World robot-ics. Geneva: International Federation of Robotics.

Jaikumar, R., & Bohn, R. 1992. A dynamic approach tooperations management: An alternative to static op-timization. International Journal of ProductionEconomics, 27: 265-282.

Katila, R. 2000. Innovation search determinants of newproduct introductions and their radicality: Thecase of industrial robotics. Unpublished doctoraldissertation. University of Texas at Austin.

Katila, R. 2002. New product search over time: Past ideasin their prime? Academy of Management Journal,45: 995-1010.

Levinthal, D., & March, J. 1981. A model of adaptiveorganizational search. Journal of Economic Rehav-ior and Organization, 2: 307-333.

Levinthal, D., & March, J. 1993. The myopia of learning.Strategic Management Journal, 14: 95-113.

Liang, K.-Y., & Zeger, S. L. 1986. Longitudinal data anal-ysis using generalized linear models. Biometrika,73: 13-22.

Mansfield, E. 1968. The economics of technologicalchange. New York: Norton.

Mansfield, E. 1988. The speed and cost of industrialinnovation in Japan and the United States; Externalvs. internal technology. Management Science, 34:1157-1168.

March, J. 1991. Exploration and exploitation in orga-nizational learning. Organization Science, 2: 71 -87.

Martin, X., & Mitchell, W. 1998. The influence of localsearch and performance heuristics on new designintroduction in a new product market. ResearchPolicy, 26: 753-771.

McGullagh, P., & Nelder, J. 1989. Generalized linearmodels. New York: Ghapman & Hall.

Miner, A., Bassoff, P., & Moorman, G. 2001. Organiza-tional improvisation and learning: A field study. Ad-ministrative Science Quarterly, 46: 304-337.

Nelson, R. 1993. (Ed.). National innovation systems.New York: Oxford University Press.

Nelson, R., & Winter, S. 1982. An evolutionary theory ofeconomic change. Gambridge, MA: Harvard Univer-sity Press.

Rosenkopf, L., & Nerkar, A. 2001. Beyond local search:Boundary-spanning, exploration, and impact in the


optical disc industry. Strategic Management Jour-nal, 22: 287-306.

Saviotti, P., & Metcalfe,). 1984. A theoretical approach tothe construction of technological output indicators.Research Policy, 13: 141-151.

Schoonhoven, G., Eisenhardt, K., & Lyman, K. 1990.Speeding products to market: Waiting time to firstproduct introduction in nev̂ r firms. AdministrativeScience Quarterly, 35: 177-207.

Schumpeter, J. 1934. The theory of economic develop-ment. Gambridge, MA: Harvard University Press.

Shane, S. 1992. Why do some societies invent more thanothers?/ournai of Business Venturing, 7: 29-46.

Steensma, K., & Gorley, K. 2000. On the performance oftechnology-sourcing partnerships: The interactionbetv r̂een partner interdependence and technology at-tributes. Academy of Management Journal, 43:1045-1067.

Stuart, T. E., & Podolny, ). M. 1996. Local search and theevolution of technological capabilities. StrategicManagement Journal, 17(special issue): 21-38.

Teece, D. J. 1987. Gapturing value from technologicalinnovation: Integration, strategic partnering, and li-censing decisions. In B. Guile & H. Brooks (Eds.),Technology and global industry: 65-95. Washing-ton, DG: National Academy Press.

Thompson, J. 1967. Organizations in action. New York:McGraw-Hill.

von Hippel, E., & Tyre, M. 1995. How learning by doingis done: Problem identification in novel processequipment. Research Policy, 24: 1-12.

Walker, R. 1995. Patents as scientific and technicalliterature. Metuchen, NJ: Sccirecrow Press.

Winter, S. 1984. Schumpeterian competition in alterna-tive technological regimes. Journal of Economic Be-havior and Organization, 5: 287-320.

Riitta Katila ([email protected]) is an assistant pro-fessor at Stanford University. She received her Ph.D. instrategic management from the University of Texas atAustin. Her research interests includei technology strat-egy, organizational evolution, and organizational searchand innovation.

Gautam Ahuja is the Michael R. ;and Mary Kay HallmanFellovir and an associate professor of corporate strategyand international business at the University of MichiganBusiness School. He received his Ph.D. from the Univer-sity of Michigan. His research focuse.s on acquisitions,alliances, and technology strategies in large corporations.

_/^

Documents

SOMETHING OLD, SOMETHING NEW: A LONGITUDINAL STUDY OF SEARCH