Upload
vudan
View
219
Download
7
Embed Size (px)
Citation preview
Research designResearch design
How does the “breadth” of innovation approach matter for innovation performance? Objectives Sources of knowledge Sources of knowledge Geographic location
Finnish CIS and R&D survey + employment register (skills data)
Cross-sectional with lagged explanatory variablesvariables
Collaborative work with Constance Helfat(Dartmouth College)(Dartmouth College)
R&D surveyR&D survey
Collected by Statistics FinlandCo ected by Stat st cs a d Targeting all Finnish R&D performing manufacturing firms, plus a set of (R&D g pperforming) service firms
Data on R&D investments, employees, units and location; commercialization of innovations
Every 2 years
3
Community Innovation SurveyCommunity Innovation Survey
Collected by Statistics Finland Collected by Statistics Finland Survey instrument and data collection techniques developed by Eurostattechniques developed by Eurostat
All Finnish manufacturing firms with more than 100 employees, plus random sample stratified by100 employees, plus random sample stratified by size and industry of the remainder
72 percent response rate in 1997p p Data on commercialization of innovations, innovation objectives, knowledge sources in innovation, R&D, other information
4
Advantages Advantages Representative data for manufacturing (not just pharmaceuticals!)
Innovation outcomes (not intermediate outputs such as patents!)
Unique data from Finland: innovation objectives Unique data from Finland: innovation objectives (CIS 2); domestic R&D units
Disadvantages Cross‐sectional setup (with lags) difficult to argue exogeneity
Limited set of organizational variables
Aija Leiponen, Cornell UniversityAija Leiponen, Cornell UniversityConstance Helfat, Dartmouth College
6
1 Parallel paths and sampling1. Parallel paths and sampling Nelson 1961; Evenson & Kislev 1976; Baldwin
and Clark 2003 Technological opportunity is characterized as a
distribution of innovation outcomesdistribution of innovation outcomes E.g. balls in an urn
Each innovation project is a draw from the di ib idistribution The more times you draw, more likely to find good outcome
Probability of success is improved by conducting multiple parallel searches, but there are decreasing returns to draws/projectsthere are decreasing returns to draws/projects
2 Cognition and uncertainty2. Cognition and uncertainty
Prahalad & Bettis; Gavetti & Levinthal; Tversky & Prahalad & Bettis; Gavetti & Levinthal; Tversky & Kahneman
Dominant logic constrains innovative search and gmakes local search more likely than distant search
Availability heuristic: rely on easy‐to‐retrieve y y yinformation
Adjustment and anchoring: estimate uncertain events by adjusting an initial value/reference point
People (and firms) tend to search narrowly
3 Innovative search3. Innovative search
March: Exploration and exploitation (OS 1991)p p ( ) Katila and Ahuja (AMJ 2002)
Search depth (reuse) and scope (breadth) and their interaction are pos. related to new product introductions
Decreasing returns to depth, not breadthg p , Laursen and Salter (SMJ 2006)
Broad search facilitates incremental improvement; h d th i i t d ith t th ldsearch depth is associated with new-to-the-world
innovation Decreasing returns to both types of searchg yp
What about breadth in innovation objectives? Objective: a technical goal (Cohen and Malerba Objective: a technical goal (Cohen and Malerba, 2001)
A research program or project A research program or project With a particular objective, e.g., develop a new product reduce labor costsproduct, reduce labor costs
Analogy to technical trials and potential routes to a single innovation (Nelson 1961 Evenson anda single innovation (Nelson, 1961; Evenson and Kislev, 1976)
10
CIS Innovation ObjectivesCIS Innovation Objectives1. Replace outdated 6. Increase flexibility of p
products2. Improve product quality
E d d
yproduction
7. Reduce labor costsR d f i l3. Expand product
assortment4. Enter new markets or
8. Reduce use of materials9. Reduce use of energy10 Mitigate environmental4. Enter new markets or
increase market share5. Fulfill government
l ti t d d
10. Mitigate environmental damage
regulation or standards requirements Scale: not important/not
used – very important (0–3)
11
HypothesesHypotheses
1 Firms that have greater breadth of innovation1. Firms that have greater breadth of innovation objectives experience greater innovation success.
Multiple draws increases probability of successMultiple objectives counteracts diminishing returns within each objective
2. Firms that have greater breadth of knowledge g f gsources for innovation experience greater innovation success.
As the number of innovation objectives or knowledge sources increases the positive impact on innovationsources increases, the positive impact on innovation success diminishes.
12
Innovation Success (R&D survey)Innovation Success (R&D survey)
INNOVATION INNOVATION Binary (0,1) variable indicating whether the firm introduced any new‐to‐the‐firm technological productintroduced any new to the firm technological product or process innovations in 1996‐98 Probit regression
NEW PRODUCT SALES Percentage of firm sales revenue in 1998 from technologically new products introduced in 1996‐98 Tobit regression Tobit regression
13
Objectives and Sources (CIS)Objectives and Sources (CIS)
OBJECTIVES Sum of the binary scores for innovation objectives that
bt i l ti f 2 3 l f 0 3obtain an evaluation of 2 or 3 on a scale of 0‐3
SOURCES SOURCES Sum of the binary scores for knowledge sources that obtain an evaluation of 2 or 3 on a scale of 0 3obtain an evaluation of 2 or 3 on a scale of 0‐3
14
Control VariablesControl Variables
Log of number of firm employees Log of number of firm employees Log of R&D expenditures Business group (0 1 subsidiary of larger firm) Business group (0,1—subsidiary of larger firm) Ratio of export to total sales revenues % employees with PhDs % employees with technical college degrees Industry dummy variables
17
Results OBJECTIVES is significant in tobit for product sales; SOURCES is significant in both probit forsales; SOURCES is significant in both probit for prob(innovation) and tobit for product sales multicollinearity
Diminishing returns not very clear; stronger for knowledge sources than objectives
Optimal breadth is quite high! and many firms don’t seem to reach it – cognitive li i i ?limitations?
Sensitivity analyses with product and process objectives; individual objectivesobjectives; individual objectives
18
ImplicationsImplications Innovation objectives (and objectives more Innovation objectives (and objectives more generally) matter Tend to be overlooked in innovation researchTend to be overlooked in innovation research
Need to decompose innovation activity into underlying componentsunderlying components Provides greater precision regarding the determinants of innovation successof innovation success
Improves the ability to derive meaningful managerial implicationsp
19
Aija Leiponen, Cornell UniversityAija Leiponen, Cornell UniversityConstance Helfat, Dartmouth College
20
What Does Multilocation of R&D Entail?
Multiple geographic locations of R&D activity within a single firm
Geographically separated from headquarters some degree of organizational decentralization
“GEOGRAPHIC DECENTRALIZATION”“GEOGRAPHIC DECENTRALIZATION”
21
Benefits vs. costs of geographic decentralization
1. International Business/FDI and Knowledge-Based View (KBV):Based View (KBV):
Focus on benefitsbenefits
2. Organizational Economics (organization form and incentives):form and incentives):
Focus on costscosts
22
1. International Business & Knowledge-Based View of the FirmBased View of the Firm
Firms need information about foreign markets Firms need information about foreign markets, technologies This knowledge is often tacit
f Requires that firms co-locate geographically
Technology transfer is associated with gysubstantial transaction costs (Teece 1977)
The multinational corporation arises out of superior efficiency in knowledgeknowledge transfertransfersuperior efficiency in knowledgeknowledge transfertransferacross borders (Kogut & Zander 1992, 1993)
Reference pointReference point:: outsourcing or alliances (communicationoutsourcing or alliances (communication Reference pointReference point: : outsourcing or alliances (communication outsourcing or alliances (communication across org. boundariesacross org. boundaries))
23
Implications of FDI + KBVImplications of FDI + KBV
Clear advantages for knowledge Clear advantages for knowledge acquisition of multiple R&D locations Access to more knowledge sourcesAccess to more knowledge sources Access to more knowledge sourcesAccess to more knowledge sources Greater likelihood of successful innovationGreater likelihood of successful innovation Wider range of innovation outputWider range of innovation output Wider range of innovation outputWider range of innovation output
24
2 Organizational Economics 2. Organizational Economics Geographic distance requires at least some
d l ti d d t li tidelegation and decentralization At least of dayAt least of day--toto--day operationsday operations And perhaps strategically as wellAnd perhaps strategically as well And perhaps strategically as wellAnd perhaps strategically as well This increases the costs of monitoring and costs of monitoring and
coordinationcoordination
Knowledge transfer itself is costly too.
Reference point: single R&D locationReference point: single R&D location
25
When is it advantageous to decentralize?
More market- and customer-specific (more applied) R&D○○ Less need to transfer knowledge within the Less need to transfer knowledge within the
firm (and bear the associated costs)firm (and bear the associated costs)○○ Decentralized incentives, decision makingDecentralized incentives, decision making
26
Benefits AND Costs: Hypotheses
1. Firms that have multiple locations of R&D activity i i i b hexperience greater innovation success, but there
are diminishing returns to the number of R&D locationslocations
-- FDI/KBV + costs of organizationFDI/KBV + costs of organization
(C diti l H1) A iti i ti2. (Conditional on H1) Any positive association between multilocation of R&D activity and innovation success reflects access to a larger o at o success e ects access to a a genumber of different sources of knowledge outside of the organization
FDI/KBV ( f /i ti t )FDI/KBV ( f /i ti t )-- FDI/KBV (vs. org. form/incentive arguments)FDI/KBV (vs. org. form/incentive arguments)
27
Benefits vs. Costs: Hypotheses
3. (a) Firms that have multiple R&D locations generate a widerwider range of innovation output than firms that have a i l R&D it (KBV)single R&D unit (KBV)
Vs.Vs.(b) Firms that have multiple R&D locations generate a ( ) p gnarrowernarrower range of innovation output (org. form + incentives)
4. Multilocation of R&D activity is associated with greater innovation success for “imitative” than “novel” innovations (org. form + incentives)-- Implications for research on patents, which reflect novel innovationsImplications for research on patents, which reflect novel innovations
28
Empirical setting
Manufacturing sector in Finland Manufacturing sector in Finland
Uniquely detailed data on innovation Uniquely detailed data on innovation outcomes and R&D locations of individual firmsfirms No information on command and control No information on command and control
structure within firmsstructure within firms
Arguments apply within as well as between countries
29
Can we study geographic decentralization within one small country?
YesYes! Finland is geographically and ! Finland is geographically and economically diverseeconomically diverse Long distances between major cities and hot Long distances between major cities and hot
spots – surface area about the same as CA; 50% larger than the UK
A few specialized “hot spots” – electronics in the North; A few specialized hot spots electronics in the North; pulp, paper & machinery in South East; medical research on the West coast
4 technical universities, 10 universities in distinct ,locations
Helsinki metro area = largest market; two other major industrial concentrations (Turku Tampere)major industrial concentrations (Turku, Tampere)
30
Primary sources of datayAn almost representative sample of R&D An almost representative sample of R&D performing manufacturing performing manufacturing firmsfirms
R&D survey 1998 R&D locations R&D spendingR&D spending Size (employees) Other controls: firm structure (business group; M&A activity;
divestments), exports
Innovation survey 1998-2000 Innovation output measures Importance of knowledge sources
○○ Customers, suppliers, competitors, universities, government Customers, suppliers, competitors, universities, government research institutes, patents and databases, trade and professional research institutes, patents and databases, trade and professional meetingsmeetings
Technological innovations only32
Characteristics of the samplep
469 R&D performing manufacturing firms 469 R&D-performing manufacturing firms 354 employees on average in 1998 (5–22,000) R&D/sales average 4.3% in 1998 (0–93%)g ( ) 67% product innovators (“new to the firm”) 46% process innovators
13% had multiple R&D locations in Finland (in 2000) 2 locations: 7 5%2 locations: 7 5% 2 locations: 7.5%2 locations: 7.5% 3 or more locations: 5.3%3 or more locations: 5.3%
33
Dependent variables Binary indicators of innovation success
Indicate whether the firm introduced at least one inno ationinnovation
Any type of innovation: process, new-to-the-market product, new-to-the-firm product
Product sales revenue from innovation All types of products, new-to-the-market, or new-
t th fi lto-the-firm only
Breadth of innovation impact1 BOTH d t d i ti1. BOTH product and process innovations2. Innovation impact survey measures (product range,
quality, market share, new markets, production flexibility, capacity costs environmental effects regulations andcapacity, costs, environmental effects, regulations and standards)
34
Explanatory variablesp y
Number of R&D locations Number of R&D locations 2 locations; 3 or more locations
Number of important external knowledge sources Original responses on a Likert scale
35
Control variablesControl variables
Firm size (log employees) Firm size (log employees) R&D expenditures (log) Business group subsidiary or parent Business group subsidiary or parent Export revenues (log) S l th f M&A l Sales growth from M&A or sales
reduction from divestment Foreign subsidiaries (log) Foreign subsidiaries (log) Industry
36
Findingsg
1. Multiple R&D locations are associated with greater innovation successgreater innovation success
Diminishing returns to multiple locationsDiminishing returns to multiple locations
2 Effect of multiple R&D locations on innovation2. Effect of multiple R&D locations on innovation success is correlated with that of external sources of knowledge
C i i h di iC i i h di iConsistent with mediationConsistent with mediation
3. Multiple R&D locations are associated with a b d i t f i tibroad impact of innovations
4. Multiple R&D locations are associated with t th fi (i it ti ) b t t t thnew-to-the-firm (imitative) but not new-to-the-
world innovations37
Conclusion: Towards a Nuanced Approach to R&D Location
R&D l il i i i d i h R&D multilocation is associated with: Greater innovation success but diminishing returns
○○ Both information benefits and organizational costs matteBoth information benefits and organizational costs matterrgg Wider access to external sources of knowledge, and Wider applicability of resulting innovations
○○ Consistent with KBVConsistent with KBV○○ Consistent with KBVConsistent with KBV
Benefits from R&D multilocation for imitativeinnovationsinnovations
○○ As predicted by org form/incentivesAs predicted by org form/incentives
38
Contributions of the study
1. Bring together two largely separate literatures regarding geographic decentralization of R&Dgeographic decentralization of R&D Organizational economics and international business/KBVOrganizational economics and international business/KBV
2. Measures of innovation success beyond patents2. Measures of innovation success beyond patents Across many industries; representative sample Across many industries; representative sample Compare Compare novelnovel (new(new--toto--thethe--world) vs. world) vs. imitativeimitative(new(new--toto--thethe--firmfirm) )
innovationinnovationinnovationinnovation
3. Test whether access to more external knowledge sources explains the effects of R&D multilocation on innovation outcomes Prediction of the knowledgePrediction of the knowledge--based viewbased view
39
Methodological issuesMethodological issues
Measurement – survey data Measurement survey data Cross-sectional setup (with lagged
explanatory variables) – endogeneityexplanatory variables) endogeneityissues
A lagged dependent variable – biased A lagged dependent variable – biased coefficients
What really causes innovation?What really causes innovation?
Not search Not search Not location Not R&D Not R&D
h ld f i tit tishould we focus more on institutions, incentives, opportunities, cost of resources for innovation?resources for innovation?
Research opportunitiesResearch opportunities
Combine CIS with other sources of Combine CIS with other sources of representative data
Need conceptual/theoretical novelty to break p / ythrough to major journals
Geography is a hot (overheated?) area g p y How do firms (other than pharma, software) deal with globalization?
Service innovation may respond differently?
Cognition: Apply insights from behavioral econ, psych into empirical research on innovationpsych into empirical research on innovation