Content uploaded by Sverre J. Herstad
Author content
All content in this area was uploaded by Sverre J. Herstad on Oct 20, 2017
Content may be subject to copyright.
Solheim & Herstad, 2017
The differentiated effects of human resource diversity on
corporate innovation
Preprint of article under publication in International journal of innovation and technology
management
This version: 2017
Marte C.W. Solheim
UiS Business School
Centre for Innovation Research
University of Stavanger
N-4036 Stavanger, Norway
Corresponding author. Mail to: marte.solheim@uis.no
Sverre J. Herstad
Nordic Institute for Studies in Innovation, Research and Education
N-0608 Oslo, Norway
Inland Norway University of Applied Sciences
N-2604 Lillehammer, Norway
Abstract
By linking theoretical perspectives on human resource diversity to the distinction between knowledge
exploration and exploitation, this paper contributes to the growing research literature on diversity and
innovation by following up on the original argument by March (1991) that different dimensions of
organizational learning depend on different inputs and processes. Empirically, the paper draws on a
unique dataset constructed by merging Norwegian employer-employee register data for 2004-2008 with
Community Innovation Survey (CIS) data covering the years 2008-2010. Bivariate probit regressions
with controls for innovation strategy find patenting propensities more responsive to the diversity of
experience-based knowledge accumulated in firms, than the propensities to improve production
processes or introduce new products onto the market. The latter depends foremost on firms’ investments
in innovation, and respond positively to human resource diversity only when the financial commitment
of firms to development work is limited.
Solheim & Herstad, 2017
Introduction
This paper deals with the question of whether the human resource bases of firms influence their
innovative capabilities in ways that are reflected in the composition of output from development work
(Østergaard, Timmermans, and Kristinsson 2011). As point of depature for this, it links theoretical
perspectives on human resource diversity to the distinction between knowledge exploration and
exploitation as introduced by March (1991), further elaborated by Levinthal and March (1993) and
reflected in recent work emphasizing that different aspects of organizational performance depend on
different types of knowledge resources and communicative processes (Vaghely and Julien 2010, Dokko,
Wilk, and Rothbard 2009, Herstad, Sandven, and Ebersberger 2015, Jensen et al. 2007).
Ultimately, this knowledge resides with individuals (Nonaka 1994), and reflect their educational
backgrounds and work-life experiences. Still, innovation scholars have traditionally awarded more
attention to network linkages between organizations (e.g. Ebersberger et al. 2012, Spithoven,
Vanhaverbeke, and Roijakkers 2012), than to the skills and interactions of individuals within them
(Rutten and Boekema 2012). Moreover, prior research on the composition of human resource bases has
tended to focus on single performance aspects (e.g. productivity) and how they are associated with
diversity in terms of employee characteristics that are given at birth (e.g. gender and ethnicity), and as
such neither learning-based nor directly task-relevant (Bell et al. 2011). Calls have therefore been made
for this research to move beyond focusing on primary diversity and link different aspects of
organizational performance to the secondary diversity that concerns the composition of organizational
knowledge bases in terms of the skills and insights that individuals have acquired through their
educations and career paths (Bell et al. 2011, Harrison and Sin 2006).
The empirical analysis developed in response to this call uses Norwegian Community Innovation Survey
data (CIS) that has been merged with linked employer-employee (LEED) register data. From the CIS,
three dependent variables have been constructed to capture fundamentally different ways in which
innovativeness is expressed. These measures have been regressed on a set of indicators that describe the
diversity of firms’ human resource bases along the two fundamental dimensions in question.
Research background
Prior research emphasize three ways in which the educational backgrounds and work-life experiences
of individuals might influence the innovation capacities of firms. First, educational background and
career paths allow individuals to acquire skills and insights that reflect the knowledge bases and
organizational routines of past employers. Secondly, both education and experience embed individuals
in enduring interpersonal ties, through which even sensitive information might be transmitted between
past and present places of employment (Agrawal 2006, Oettl and Agrawal 2008, Dahl and Pedersen
2004, Bouty 2000). Thirdly, they shape behavioral attributes, i.e. the ways in which individuals act and
communicate their knowledge (Dokko, Wilk, and Rothbard 2009, Madsen, Mosakowski, and Zaheer
2003). Consequently, the innovation capacities of firms can be understood from the perspective of how
organizational routines transform employees’ individual experiences and networks into collective
resources.
Still, to the extent that innovation and management research has focused on human resources, emphasis
has been put on certain departments or occupational groups, notably R&D departments, researchers and
inventors (Herrera, Munoz-Doyague, and Nieto 2010), project groups (Horwitz and Horwitz 2007), top
management teams (Bantel and Jackson 1989, Pitcher and Smith 2001, Knight et al. 1999, Murray 1989,
Solheim & Herstad, 2017
Wiersema and Bantel 1992) and corporate boards (Miller and del Carmen Triana 2009, Bjørnåli and
Gulbrandsen 2010). Given how modern products and production processes are complex, it is required
that development work draw on a much broader range of knowledge than what reside within individual
experts or contained within certain groups (Grant 1996). This highlights that studies considering how
the human resource bases of entire organizations shape their innovative capabilities are needed.
When approaching this issue conceptually (Harrison and Klein 2007), a distinction can be made between
the ‘cognitive resource diversity’ perspective and the ‘similarity attraction’ perspective (Simons and
Rowland 2011, Horwitz 2005). The cognitive resource diversity perspective holds that that diverse
teams or organizations outperform more homogenous entities (Hong and Page 2004) as they possess a
“broader range of task-relevant knowledge, skills and abilities, giving the group a larger pool of
resources that when combined may generate new insights” (Van Engen and Van Woerkom 2010, 135).
Interaction between individuals with diverse heuristics is one way of creating “kaleidoscope thinking”
(Kanter 1968), entailing that the presence of a variety of perspectives is vital in triggering new
knowledge. However, as emphasized in the literature on absorptive capacity (Cohen and Levinthal 1990)
and reflected in the literature on recruitment (Mawdsley and Somaya 2015), variety, or difference, of
perspectives may also lead to miscommunication, uncertainty, and conflicting views, which could cause
firms to retain rather than adjust well-established practices (Madsen, Mosakowski, and Zaheer 2003).
Acknowledging this, the similarity attraction perspective emphasizes that people prefer to engage in
relationships with other people that are similar to themselves (McPherson, Smith-Lovin, and Cook
2001), and that this similarity eases communication and enable a more efficient execution, not least of
complex tasks. Recent evolutionary contributions incorporate both the similarity attraction perspective,
and the cognitive resource diversity perspective in suggesting that certain levels (cf. Nooteboom et al.
2007 and the concept of 'optimal cognitive distance') or types (cf. Boschma and Iammarino 2009 and
the distinction between related and unrelated variety) of variety is particularly supportive of
organizational performance.
This research has made a strong case that ‘related’ variety, i.e. variety within as opposite to between
larger cognitive domains (e.g. main types of education), is particularly important to the productivity
performances of firms. Yet, productivity is not equal to innovation (Crepon, Duguet, and Mairesse
1998), and other contributions have suggested that communicative challenges and knowledge needs
differ substantially between different yet interlinked phases of innovation processes (Herstad and Brekke
2012, Jensen et al. 2007, Teece 1986) such as idea generation and implementation (cf. Axtell et al. 2000,
269). This raises the question of whether different dimensions, or aspects, or innovation capacity and
performance respond differently to the diversity of human resources.
To approach this question theoretically, one can draw on the distinction between exploration and
exploitation (March 1991, Raisch et al. 2009). Exploitation involves work aimed at refining the
capabilities that are valued in firms’ present markets. It depends on the firm- and sector-specific
knowledge and skills that employees have accumulated (Wang, He, and Mahoney 2009), and take the
form of continuous improvements of product lines, production processes and business models in
response to gradually evolving circumstances. Here, specialized tasks and knowledge might combine
the complexity of work-processes to constrain the ability of firms to incorporate ideas and insights that
differ from those that dominate current work-processes. ‘Exploration’, by contrast, has been described
as ''the pursuit of knowledge of things that might come to be known” (Levinthal and March 1993, 105)
and involve efforts aming to transcend the confines of current technologies, products and organizational
practices. As explorative efforts need not, or should not, at the outset relate to the practicalities and
complexity of daily business operations, they can be assumed more responsive to the presence of skills
Solheim & Herstad, 2017
and mind-sets that are different from each other and from those that are shaped by the ongoing business
processes of firms.
Consistent with this, prior research has found diversity of expertise to be positively associated with non-
routine task environments (Murray 1989, Hambrick, Cho, and Chen 1996). Moreover, it has
demonstrated how inflows of expertise from outside firms` own industry domains strengthen
specifically the explorative efforts that are expressed by firms’ patent output (Herrera, Munoz-Doyague,
and Nieto 2010), without necessarily influencing exploitation where larger rigidities and stronger
absorptive capacity constraints are at play (Herstad, Sandven, and Ebersberger 2015). Explorative
efforts may provide the foundation for radical innovations. Still, for such to materialize, firms may have
to invest substantial effort in adjusting outcomes of explorative efforts to their exploitation capacities,
and vice versa (Jensen et al. 2007, Teece 1986). This underscores the need for, and challenges of,
organizational ‘ambidexterity’ that is capacity to combine exploration and exploitation (March 1991,
Hall, Lotti, and Mairesse 2008, O’Reilly and Tushman 2008), and raises the fundamental question of
whether the human resource bases supporting this (Raisch et al. 2009, He and Wong 2004) are different
from those supportive of either one of its constituent aspects (Bonesso, Gerli, and Scapolan 2014, 392).
The moderating role of innovation efforts
Before empirically addressing these assertions, it should be noted that the relationship between human
resource diversity and innovation might depend on firms’ commitment to development work, as
commonly captured by their expenditures on research and development (R&D). For one, it is widely
acknowledged that the accumulated stock of knowledge available to firms reflect their accumulated
R&D efforts, and increases their capacity to identify, assimilate and transform new ideas and external
technology for productive purposes (Cohen and Levinthal 1989). Similarly, current efforts can be
viewed as reflecting emphasis on mobilization and integration of knowledge, thus potentially increasing
the capacity to integrate diverse human resources.
While this line of reasoning follows the literature on absorptive capacity in suggesting a complementary
relationship between R&D efforts and human resource diversity (Cohen and Levinthal 1990),
substitution effects may also be at play. This is because the composition of firms’ human resources bases
in terms of educational backgrounds and prior work-life experiences might be more visible through its
innovation output when investments in learning are limited, and more important to this output when
firms organize their development work as an integral part of their daily business operations (Jensen et
al. 2007). Moreover, a strong emphasis on internal R&D could reduce the responsiveness of firms to
ideas and insights stemming from other sources than their own R&D departments or project groups (cf.
Katz and Allen 1982 and the 'not-invented-here' syndrome, Laursen and Salter 2006).
Solheim & Herstad, 2017
Empirical analysis
Data
The empirical analysis is based on data that cover innovation activities and outcomes in a representative
sample of Norwegian firms during the three-year period 2008-2010. It was collected by Statistics
Norway in 2010, as an extended version of the harmonized pan-European Community Innovation
Surveys commonly abbreviated ‘CIS’(Eurostat 2010). The questionnaire is based on the definitions of
innovation input (R&D and non-R&D expenditures), external linkages (technology sourcing and
innovation collaboration) and output laid out in the second revised edition of OECD`s Oslo Manual
(OECD 2005). Additional information on individuals working in the firm in 2008, i.e. at the start of the
reference period, has been gathered from Linked Employer-Employee Data (LEED) for the years 2004-
2008.
The complete CIS2010 sample consists of 6595 enterprises in aquaculture, offshore oil & gas extraction,
manufacturing industries, wholesale trade & transportation, hotels & restaurants, energy &
infrastructure, construction and knowledge intensive business services (KIBS). To reduce sector
heterogeneity beyond what can reasonably be accounted for by control variables, the analysis uses only
observations in oil & gas, manufacturing and knowledge intensive business services industries. To allow
computation of diversity measures at the beginning of the CIS reference for which innovations are
observed, it is restricted to the 2942 enterprises sampled in 2010 that could be identified in the
employment registers for 2008. Sample characteristics are summarized in Table 2 below.
Selection and dependent variables
One selection variable and two dependent variables are used in the analysis. The binary selection
variable ENGAGEMENT captures whether the firm is innovation-active, and thus whether the question
of explorative and exploitative content in such work is relevant. Following the structure of the
questionnaire and conventions in the CIS literature, it takes on the value 1 if the firm reported positive
innovation expenditures (R&D or non-R&D), finalized, ongoing or abandoned innovation projects, or
positive innovation outcomes during the period 2008-2010 (Cassiman and Veugelers 2006, Ebersberger
et al. 2012).
For those firms that are engaged, two yet commonly used indicators of innovation output are used to
capture different aspects of development work. The binary dependent variable PATENT takes on the
value 1 if the firm states that it filed a patent application during the reference period. While the use of
patents as innovation indicators is highly debated (e.g. Blind et al. 2006), the assumption here is that the
application itself signal the existence of a new technology or solution that from the perspective of the
firm i) has sufficient potential value to justify the costs involved in patenting, and ii) the novelty content
required for a patent to be awarded. Thus, it signals a certain explorative content in firms’ development
work (cf. Herstad, Sandven, and Ebersberger 2015).
The binary dependent variable INNOVATION takes on the value 1 if a product innovation or a process
innovation is reported, irrespective of patent applications, or novelty content more generally. A product
innovation occurred if the focal firm itself developed, or actively contributed to the development of, a
new or significantly improved product (goods and/or services) during the reference period (OECD 2005)
that was also introduced onto the market. Similarly, a process innovation occurred if the firm itself
developed, or actively contributed to the development of, new production processes or support
Solheim & Herstad, 2017
functions
1
that were also implemented by the firm. Accordingly, it captures changes and improvements
that inherently have an exploitative content.
The diversity construct
According to van Knippenberg and Schippers (2007) diversity research need to move beyond studying
single dimensions of diversity, and rather move on to conceptualizations of diversity as combinations
of different dimensions. In the context herein, this is reflected in a distinction between variety within
(education or experience domains) and variety between (education and experience domains) based on
the use of entropy measures (Jacquemin and Berry 1979).
Each individual working in the firm has been assigned a five-digit code that expresses their educational
background, i.e. the type and level of education obtained. If each firm has n educational types present,
represented by the categories, then the total entropy for each firm is given by Jacquemin & Berry
(1979:360) as:
were Pi is each category’s proportion of the total number of individuals present within the firm. These
categories are structured hierarchically as specialized sub-fields within main aggregate fields. If we have
s main fields, and Ps is the proportion of employees in each main field, then the entropy across main
fields is given by Jacquemin & Berry (1979:361) as:
Entropy within each main field is likewise given by:
The total entropy may be expressed in the following way (see Jacquemin and Berry, 1979:362 for
details):
1
At least one of the following types of innovations, as stated in the CIS questionnaire: i) new or significantly
improved method of production, ii) new or improved method for storing and distributing goods and services, iii)
new or significantly improved support function.
Solheim & Herstad, 2017
or
is a weighted average of the entropy within each main educational field, were the weights are the
proportion of employees in each of the educational classes present within the firm (i.e. the Ps defined
previously). This is hereafter referred to as related educational variety (EDUVAR_REL). It expresses
variety within clearly delineated educational fields, in which a certain overlap of cognitions, languages
and identifies can be expected present. is the entropy across main fields, and is in the following
referred to as unrelated educational variety (EDUVAR_UNREL) because it expresses variety across
fields that cannot be assumed to be characterized by common professional identifies and overlapping
languages. equals the sum of the two, and thus the total educational variety of the focal firm
(EDUVAR_TOT).
(Table 1 about here)
To capture experience diversity, matrixes describing the career paths of employees during the five-year
period ending in 2008, i.e. at the start of the CIS2010 reference period, have been generated, for each
individual firm in the dataset. The firm in the example given in Table 1 with 20 employees in 2008
engaged in the production of engines and turbines (NACE 28.110). Including 2008 and the four years
prior to it gives 20 x 5 = 100 experience-years, of which 74 were associated with employment in the
focal firms’ sector (NACE 28.110). Due to unemployment, five person-years do not count as experience-
years. The remaining 21 experience-years were generated in NACE 09.101 (oil & gas sector drilling
services), NACE 24.421 (primary production of aluminum), NACE 24.422 (aluminum half-fabrics),
NACE 26.110 (electronic components), NACE 26.200 (computers and equipment), NACE 26.300
(communication equipment) and NACE 62.101 (programming services). As the NACE codes are
hierarchically ordered and consists of two-digit main groups with three-digit sub-groups, variety in terms
of accumulated work-life experiences is expressed by entropy measures capturing the total
(EXPVAR_TOT), related (EXPVAR_REL) and unrelated (EXPVAR_UNREL) experience variety of
firms’ workforces as described above.
The distribution of the sample with average diversity and innovation output scores is given in Table 2
below.
(Table 2 about here)
Control variables
From the example in Table 1, it is evident that low employment turnover reduces the experience
diversity in firms’ human resource bases. Conversely, a high turnover might, but not necessarily,
increase this diversity. High turnover rates, and thus low average organizational tenure (Bell et al. 2011),
may work against innovation because it weakens the capacity of firms to accumulate knowledge and
develop knowledge integration routines (Kleinknecht, van Schaik, and Zhou 2014, Zhou, Dekker, and
Kleinknecht 2011, Herstad and Ebersberger 2014, DiMaggio and Powell 1983). Moreover, it reduces
the probability that individuals come to understand the social knowledge, values and expected behaviors
necessary to assume an organizational role (Sturman 2003, DiMaggio and Powell 1983). On the other
Solheim & Herstad, 2017
hand, low turnover might strengthen knowledge accumulation and communication within the firm, at
the cost of new impulses and insights. As turnover and diversity are interlinked and their influences on
innovation may draw in different directions, the control variable REPLACEMENT is included. It
captures the number of employees replaced during the 2008-2010 period as a proportion of employees
present at the start of the period.
In terms of innovation, organizations with higher overall educational levels can be expected to
outperform those with lower educational levels (Bell 2007, Herstad, Sandven, and Solberg 2013) as
“innovation is a relatively more skill-intensive activity than imitation” (Vandenbussche, Aghion, and
Meghir 2006, 97). Although educational levels have been used as a diversity variable in past
contributions (e.g. Amason, Shrader, and Tompson 2006), the distribution of employees across different
educational levels is not likely to increase the breadth of perspectives available for the firm to draw on,
beyond what is associated with educational variety (Bell et al. 2011). Thus, the variable EDULEVEL is
included as a control that captures the mean educational level of the firms’ workforce based on the 8-
level scale used in the public registers.
The entropy of employee experiences and educational backgrounds increase concomitantly as the
number of employees increase. As size might also influence innovation activity and output, the paper
follows conventions and include the logarithm of employment in 2008, i.e. the year for which diversity
is observed, as a control (Grimpe and Kaiser 2010). Different industrial sectors are characterized by
different incentives to engage in innovation activities, different output propensities and differences in
the composition of human resource bases (cf. Table 2). Based on the NACE industry codes provided in
the CIS and reflecting the technology intensity classes of OECD (Hatzichronoglou 1997), manufacturing
firms are divided into 4 sector groups that are distinguished from the 6 main types of services provision
covered by the CIS. Last, petroleum extraction industries are idiosyncratic to the Norwegian economy
and classified as such. This gives 11 industries in total, which are represented by 10 industry dummies
in the regressions (cf. Table 2). Market presence determines potential market size and diversity of market
information exposure, and may therefore influence innovation (Crepon, Duguet, and Mairesse 1998,
Ebersberger and Herstad 2011). Moreover, it provides the reference for when a product introduction is
also a market novelty. MARBREADTH captures the number of world regions specified in the CIS
questionnaire on which the firm indicates a market presence
2
.
Innovation outcomes are strongly determined by the overall emphasis put by the firm on development
work, which in turn may influence the capacity of firms to exploit diverse cognitions. To control for
this, innovation expenditures are captured by the binary variable INNOVINT that takes on the value 1
if reported innovation expenditures (R&D and non-R&D) per employee were above the sample median.
The use of a binary measure is chosen over the option of a continuous measure to allow straightforward
interpretation of interaction effects between INNOVINT and diversity (see Ebersberger and Herstad
2011 for a more elaborate discussion of this point).
Innovation output is also influenced by the extent to which firms strategically use knowledge and
technology from collaboration partners (Grimpe and Kaiser 2010). Because different types of
collaboration partners provide different yet potentially complementary types of knowledge (Ebersberger
and Herstad 2011, Roper, Du, and Love 2008, Nieto and Santamaría 2007), we follow prior studies
(Laursen and Salter 2006, Grimpe and Kaiser 2010) in including a control for the number of different
2
The options included in CIS: Local/regional in Norway, elsewhere in Norway, Other EU, EFTA or EU candidate countries,
and other countries.
Solheim & Herstad, 2017
collaboration partners used by the firm (COBREADTH) during the reference period
3
. Both innovation
strategy controls are constructed from CIS data and information provided assume that ENGAGEMENT
= 1.
Econometric approach
In the first stage of the analysis, probit regression models estimate the selection variable
ENGAGEMENT using only information available for all firms (N = 2942). The estimations are first
conducted with the measures for total diversity included (Model 1), and then with diversity split into
related and unrelated (Model 2). In the second stage, bivariate probit regression models are used to
estimate PATENT and INNOVATION simultaneously (cf. Herstad, Sandven, and Ebersberger 2015),
thus allowing conditional marginal effects to be estimated. This stage of the analysis include only firms
identified in the first stage as innovation active (N=1450).
Technically, this two-step approach is necessary because only engaged firms have provided the
information on innovation expenditures and collaboration needed for innovation strategy controls to be
implemented
4
in estimations of output, and because such output is contingent on the decision to engage.
Moreover, by conducting and reporting the selection stage, the analysis can distinguish clearly between
the human resource bases that generally characterizes innovation-active firms (Step 1) and reflect their
recruitment options and choices; and the human resource characteristics that are supportive of different
output from this work (Step 2) (cf. discussion in Herstad 2017). This provides transparency in terms of
what influences that are exerted and how, and dampens endogeneity concerns in estimations of output
as these are most pressing in the first stage where a corporate choice variable (ENGAGEMENT) which
is subjected to intertemporal persistence (Cefis and Orsenigo 2001) is regressed on human resource
characteristics that also reflect matching of preferences among firms and their employees.
Also for the sake of transparency, estimations in Table 6 and Table 7 are first conducted using only
background information on the firm and total diversity measures (Model 3). Innovation strategy controls
are then included (Model 4), before interactions between INNOVINT and the two measures of total
diversity are included (Model 5). Diversity is then divided into related and unrelated educational and
experience variety (Model 6). In the last set of estimations (Model 7), it is assumed that the capacity of
firms to translate diverse human resources into support for different aspects of their innovation activities
depend on the emphasis put on development work, i.e. on INNOVINT. The size of effects are difficult
to determine from probit coefficients (Hoetker 2007), in particular when interaction terms are involved
(Ebersberger and Herstad 2011). Therefore, average marginal effects of diversity contingent on
innovation intensity are reported in Table 8 and interpreted against the background of predicted
innovation outcome probabilities at different levels of diversity.
A summary of variables used in the two estimation stages is given in Table 3 below.
(Table 3 about here)
3
The options given are: Other units within parent enterprise group, clients, suppliers, competitors, consultancy firms,
universities and other higher education institutions, commercial R&D laboratories, private and public R&D institutes.
4
This translates into a risk that estimates are biased by unobserved determinants of sample selection, i.e. of the decision to
engage. To acknowledge this, supplementary regressions have been estimated using the two-step procedure of Heckman (1979).
In these estimations, the results remain structurally consistent with those reported and discussed below. Due to the absence of
the instrumental variable needed to ensure that the procedure does not build serious multicollinearity into the models, it is not
implemented in the reported regressions (cf. Puhani 2000).
Solheim & Herstad, 2017
Results
Model 1 reported in Table 4 finds ENGAGEMENT positively associated with the size and average
educational level of the firm, and with the breadth of its market presence. It is found to be negatively
associated with the replacement rate. On the one hand, this suggests that high replacement rates reduces
the need to engage in development work, due to learning-by-recruitment effects, or reduces the
incentives of firms to engage due to appropriability problems associated with large outflows of
knowledge (cf. Herstad and Ebersberger 2014). On the other hand, it might simply reflect that active
firms are more attractive employers (cf. discussion of the two-step procedure above).
(Table 4 about here)
Generally, ENGAGEMENT is positively associated with variety in educational backgrounds and work-
life experiences. When the distinction between related and unrelated variety is implemented in Model
2, positive and significant estimates are obtained for related educational variety and unrelated experience
variety. Not only does this underline the importance of making this distinction, but it also suggests that
innovation active firms tend to employ individuals with related educational background and unrelated
experiences; i.e. less diverse educational backgrounds than work-life experiences. Conversely, the
highly exploitative business strategies that are characterized by absence of systematic development work
are associated with specialized human resources in firms.
Table 5 describes the distribution of outcomes from development work, for firms engaged in such. 456
of the 1450 active firms (31 per cent) did not report any outcomes, underscoring the relevance of
distinguishing between innovation activity and innovation results. The most common outcome observed
is the introduction of a product or process innovation during the period (INNOVATION = 1), without
any patent applications filed (PATENT = 0). On the other hand, 6 per cent of the firms filed for a patent
without introducing any innovations. This highlights the need for caution when patent data is used to
describe the innovation capacities of firms (e.g. Herrera, Munoz-Doyague, and Nieto 2010), and
substantiate the relevance of comparing estimates for the two variables in order to capture different
aspects of innovation (e.g. Herstad, Sandven, and Ebersberger 2015).
(Table 5 about here)
Table 6 reports the estimations of PATENT (Equation A), here used to proxy the explorative content of
firms’ development work. The base Model 3 finds this aspect to be positively and significantly
associated with total experience variety within the firm. The inclusion of controls for innovation strategy
in Model 4 (Equation A) does not structurally alter this, beyond underscoring the importance of
commitment to innovation and partner contributions to development work. When interaction effects are
included in Model 5, base and interaction estimates for innovation expenditures are all insignificant.
This suggests that the impact of such expenditures on explorative efforts are dependent on human
resource characteristics that are not captured consistently by the two overall diversity measures.
(Table 6 about here)
This again points to the importance of distinguishing between related (within-domain) and unrelated
(between-domain) variety. When implemented in Models 6 and 7, the results find the probability of
patent application increasing with unrelated experience variety specifically. A positive and weakly
significant base effect of unrelated educational variety is also obtained; yet, the interaction with
Solheim & Herstad, 2017
innovation expenditures is negative and strongly significant. Thus, while financial investments in
development work reduces the receptiveness of explorative efforts to diversity in the educational
backgrounds of employees, firms with high and low commitment levels have in common that this aspect
respond positively to unrelated variety in work-life experiences.
(Table 7 about here)
Estimations of INNOVATION (Equation B), here used to proxy the more exploitative aspect of firms’
development work, are reported in Table 7. The base Model 3 finds innovation associated foremost with
the breadth of market presence, and not influenced by total educational or experience variety. The
importance of investments in innovation and the active contribution of collaboration partners is evident
from Model 4, and a strongly significant positive interaction effect is obtained between INNOVINT and
EDUVAR_TOT in Model 5. Model 6 finds innovation output positively associated with related
experience variety, while Model 7 suggests that INNOVINT and unrelated educational variety are
positively complementary to each other in their effect on innovation.
The strong and positive sign of the interaction is highly notable, because it is contrary to the negative
and significant interaction effect obtained in the mirroring estimation of patenting (Model 7 Equation
A): While efforts and educational variety substitute each other in their effects on the explorative aspects
of development work, they are complementary in exploitation, suggesting that INNOVINT increases
the capacity to exploit cognitive resources available due to employees’ diverse educational backgrounds.
Another striking contrast between Equations A (PATENT) and Equations B (INNOVATION) is the
absence of significant estimates for average educational levels in the latter set, compared to the highly
significant estimate obtained in the former. Consequently, the explorative aspects of development work
are more receptive to formal qualifications among staff than the exploitative aspects, implying that the
latter, in line with the original argument of March (1991), depend more on specialized knowledge and
routines within the firm.
Detailed predicted probability and marginal effects
As stated above, the size and substantial relevance of effects are difficult to determine directly from
probit coefficient estimates (Hoetker 2007), in particular when baseline estimates and interaction effects
must be evaluated jointly (Ebersberger and Herstad 2011). To circumvent this problem, average
marginal effects of educational and experience variety have been computed conditionally on innovation
intensity and reported in Table 8 below. Consider first knowledge exploration capacity expressed by a
patent application. As was evident from the absence of a significant interaction effect in Model 7, it is
positively associated with unrelated experience variety in both sub-samples. For the sample as a whole,
the estimated increase in the probability of patent application is from 14 per cent at zero experience
variety, through 23 per cent at the mean and up to 39 per cent at the cut-point for the 99th percentile.
This equals a factor of 1.64, or 64 per cent.
The negative interaction between unrelated educational variety and innovation expenditures translate
into a significantly positive average marginal effect only amongst firms with a low commitment to
innovation. Thus, it does not weaken the knowledge exploration capacity of firms that are financially
committed to innovation, but strengthens this capacity among firms that are not. In the latter group, the
probability of inventive output increases from an estimated 12 per cent at zero unrelated educational
variety, through 20 per cent at the sub-sample mean and up to 26 per cent at the cut-point for the 99th
percentile. Thus, the probability more than doubles.
Solheim & Herstad, 2017
(Table 8 about here)
Consider then the exploitative aspects of development work captured by INNOVATION. In the sub-
sample of firms that exhibit a weak financial commitment to such work, the probability increases from
an estimated 54 per cent at zero related experience variety, to 70 per cent at the 99th percentile cut point.
Thus, it increases by a factor of 1.3. This effect is notable not least because the selection stage found the
same type of experience variety reducing the initial probability of engagement. Effects on innovation
from related experience variety are absent among firms that exhibit higher levels of commitment to
innovation. This commitment, however, allow firms to translate educational variety into an increase in
the probability of innovation from 55 per cent at zero variety, through 67 per cent at the sub-sample
mean to 73 per cent at the cut-point for the 99th percentile, i.e. by a factor of 1.33.
The last set of average marginal effect estimates displayed in Table 7 demonstrate that the probability
of co-occurring patent application and commercial innovation, here used to approximate the theoretical
concept of ‘ambidexterity’, is positively and significantly associated with unrelated experience variety.
Notably, this effect is independent of firms’ financial commitments to innovation. When this conditional
probability is predicted for the sample as a whole, the estimated increase is from 9 per cent at zero
unrelated experience variety, through 14 per cent at the sample mean to 23 per cent at the cut-point value
for the 99th percentile.
This increase by a factor of 2.56 clearly suggest that firms with ‘ambidextrous’ innovative capabilities
are those that have accumulated a broad range of experience-based knowledge within their
organizations. Underscoring the importance of this final result, firms with such human resource
characteristics were in the initial selection stage also found more inclined to engage in development
work.
Discussion and concluding remarks
Reflecting the cumulative, collective and multi-faceted nature of organizational learning, this paper
investigates how the overall composition of firms’ human resource bases in terms of educational
backgrounds and experiences is reflected in their performances as innovators. The results obtained are
generally supportive of the cognitive resource diversity perspective (Horwitz 2005), and challenge the
now common claim that performance is foremost associated with ‘related’ rather than ‘unrelated’ variety
of skills and knowledge (Timmermans and Boschma 2014). First, the baseline selection Models 1 and
2 found firms that are engaged in innovation activities to be characterized by employees with more
diverse educational backgrounds and work-life experiences, than firms which are not. Thus, firms with
specialized human resource bases, including those characterized by ‘related’ variety of experiences, are
generally less inclined to even attempt to innovate. Second, subsequent outcome regression found firms’
explorative efforts, as signaled by patent applications, to be supported consistently by variety of
experiences among their employees. However, and in line with the similarity attraction perspective,
firms are not necessarily able to benefit commercially from this variety, as the propensity to introduce
innovations was found to depend foremost on the strategies, efforts and thus financial commitment of
firms to development work. Only when this commitment is low, does the ‘related’ experience variety of
their employees matter for innovation.
Still, and most importantly, the results suggest that organizational ‘ambidexterity’, i.e. the ability to
combine exploration and exploitation in organizational learning, is strongly associated with the
Solheim & Herstad, 2017
unrelated variety of human resources along the dimension of the work-life experiences that inherently
have been shaped by the industrial configurations that surrounds firms, in their locations. Importantly,
yet perhaps surprisingly, this holds irrespective of firms’ financial investments in development work,
and is therefore a particularly strong indication of the dependence of modern industrial organizations on
building and maintaining human resource bases characterized by diverse work-life experiences among
well-educated employees. By implication, it underscores how the innovative capabilities of firms are
shaped by expertise available in the labor markets of their locations.
As predicted probability estimations demonstrated that effects, when significant, are also of substantial
size, the results have three important implications. First, they underscore the relevance of moving
beyond the dominant focus of research on the open innovation practices of firms (Spithoven,
Vanhaverbeke, and Roijakkers 2012, Ebersberger et al. 2012), towards a supplementary focus on the
experience-based knowledge, motivations and interactions of individuals not only within organizations,
but also between them, through surrounding labor markets (Eriksson, Lindgren, and Malmberg 2008,
Jøranli and Herstad 2017). In this way, and secondly, the results substantiate the relevance of a
strengthened focus of diversity research on human resource characteristics that are acquired and task-
relevant, i.e. secondary diversity, in addition to those that are given at birth, i.e. the primary diversity
that concern dimensions such as gender and ethnicity. Specifically, the results underscore the need to
account for diversity of experience (cf. Bell et al. 2011). Finally, it has demonstrated that empirical
analyses treating organizational learning and innovation as one-dimensional phenomena are at risk of
severely underestimating the complexity of the interplay between human resources, organizational
processes and management strategies that shape the technological and commercial capabilities of firms.
Solheim & Herstad, 2017
Acknowledgements
Work with this paper was partly funded by the Norwegian Research Council under the DEMOSREG
program (grant numbers 209769 and 209761). The authors gratefully acknowledge the assistance with
data preparation and computation of entropy measures provided by Tore Sandven of NIFU, and
comments on earlier version of the text received from Ron Boschma, Bjørn Terje Asheim, and the
journal. The usual disclaimers apply.
References
Agrawal, A., Cockburn, L., McHale, J. 2006. "Gone but not forgotten: knowledge flows, labor mobility,
and enduring social relationships." Journal of Economic Geography 6:571-591. doi:
10.1093/jeg/lb1016.
Amason, A.C, R.C Shrader, and G.H Tompson. 2006. "Newness and novelty: Relating top management
team composition to new venture performance." Journal of Business Venturing 21:125-148.
Axtell, C.M, D.J Holman, K.L Unsworth, T.D Wall, and P.E Waterson. 2000. "Shopfloor innovation:
Facilitating the suggestion and implementation of ideas." Journal of Occupational and
Organizational Psychology 73:265-285. doi: 10.1348/096317900167029.
Bantel, K.A, and S.E Jackson. 1989. "Top management and innovations in banking: Does the
composition of the top team make a difference?" Strategic Management Journal 10:107-
124.
Bell, S. T. 2007. "Deep-Level Composition Variables as Predictors of Team Performance: A Meta-
Analysis." Journal of Applied Psychology 92 (3):595-615.
Bell, S.T, A.J Villado, M.A Lukasik, L Belau, and A.L Briggs. 2011. "Getting Specific about Demographic
Diversity Variable and Team Performance Relationships: A Meta-Analysis." Journal of
Management 37 (3):709-743. doi: 10.1177/0149206310365001.
Bjørnåli, E, and M Gulbrandsen. 2010. "Exploring board formation and evolution of board
composition in academic spin-offs." The Journal of Technology Transfer 35 (1):92-112. doi:
10.1007/s10961-009-9115-5.
Blind, K, J Edler, R Frietsch, and U Schmoch. 2006. "Motives to patent: Empirical evidence from
Germany." Research Policy 35 (5):655-672. doi:
https://doi.org/10.1016/j.respol.2006.03.002.
Bonesso, S, F Gerli, and A Scapolan. 2014. "The individual side of ambidexterity: Do individuals’
perceptions match actual behaviors in reconciling the exploration and exploitation trade-
off?" European Management Journal 32 (3):392-405. doi:
http://dx.doi.org/10.1016/j.emj.2013.07.003.
Boschma, R, and S Iammarino. 2009. "Related Variety,Trade Linkages, and Regional Growth in Italy."
Economic Geography 85:289-311.
Bouty, I. 2000. "Interpersonal and interaction influences on informal resource exchanges between
R&D researchers across organizational boundaries." The Academy of Management Journal
43 (1):50-65.
Cassiman, B, and R Veugelers. 2006. "In search of complementarity in innovation strategy: Internal
R&D and external knowledge aquisition." Management Science 52 (1):68-82.
Cefis, E., and L. Orsenigo. 2001. "The persistence of innovative activities - A cross-countries and cross-
sectors comparative analysis." Research Policy 30 (7):1139-1158. doi: 10.1016/s0048-
7333(00)00139-6.
Cohen, W.M, and D.A Levinthal. 1989. "INNOVATION AND LEARNING: THE TWO FACES OF R & D."
Economic Journal 99:569-596.
Solheim & Herstad, 2017
Cohen, W.M, and D.A Levinthal. 1990. "Absorptive Capacity: A New Perspective on Learning and
Innovation." Administrative Science Quarterly 35:128-152.
Crepon, B, E Duguet, and J Mairesse. 1998. "RESEARCH, INNOVATION AND PRODUCTIVITY: AN
ECONOMETRIC ANALYSIS AT THE FIRM LEVEL." Economics of Innovation & New Technology 7
(2):115.
Dahl, M.S, and C.O.R Pedersen. 2004. "Knowledge flows through informal contacts in industrial
clusters: myth or reality?" Research Policy 33 (10):1673-1686. doi:
10.1016/j.respol.2004.10.004.
DiMaggio, P.J, and W.W Powell. 1983. "The Iron Cage Revisited: Institutional Isomorphism and
Collective Rationality in Organizational Fields." American Sociological Review 48 (2):147-160.
Dokko, G, S.L Wilk, and N.P Rothbard. 2009. "Unpacking Prior Experience: How Career History Affects
Job Performance." Organization Science 20:51-68.
Ebersberger, B, C Bloch, S.J Herstad, and E.L.S Van De Velde. 2012. "OPEN INNOVATION PRACTICES
AND THEIR EFFECT ON INNOVATION PERFORMANCE." International Journal of Innovation &
Technology Management 9:-1. doi: 10.1142/S021987701250040X.
Ebersberger, B, and S.J Herstad. 2011. "Product Innovation and the Complementarities of External
Interfaces." European Management Review 8:117-135. doi: 10.1111/j.1740-
4762.2011.01014.x.
Eriksson, R., U. Lindgren, and G. Malmberg. 2008. "Agglomeration mobility: effects of localisation,
urbanisation, and scale on job changes." Environment and Planning A 40 (10):2419-2434.
doi: 10.1068/a39312.
Eurostat. 2010. "Community innovation survey - Reference Metadata in Euro SDMX Metadata
Structure (ESMS)." http://epp.eurostat.ec.europa.eu/cache/ITY_SDDS/en/inn_esms.htm (last
accessed 17 Aug 2010, 18:00)
Grant, R. 1996. "Prospering in dynamically-competitive environments: organizational capability as
knowledge integration." Organization Science 7:375-387.
Grimpe, C, and U Kaiser. 2010. "Balancing Internal and External Knowledge Acquisition: The Gains
and Pains from R&D Outsourcing." Journal of Management Studies 47 (8):1483-1509. doi:
10.1111/j.1467-6486.2010.00946.x.
Hall, B. H, F Lotti, and J Mairesse. 2008. "Employment, innovation, and productivity: evidence from
Italian microdata." Industrial & Corporate Change 17 (4):813-839.
Hambrick, D.C, T Cho, and M.J Chen. 1996. "The influence of top management team heterogeneity
on firms` competitive moves." Administrative Science Quarterly 41:659-684.
Harrison, D.A, and K.J Klein. 2007. "What`s the Difference? Diversity Constructs as Separation,
Variety, or Disparity in Organizations." Academy of Management Review 32 (4):1199-1228.
doi: 10.5465/AMR.2007.26586096.
Harrison, D.A, and H-P Sin. 2006. "What is diversity and how should it be measured?" In Handbook of
workplace diversity, edited by A.M Konrad, P Prasad and J.K Pringle. London: Sage.
Hatzichronoglou, T. 1997. Revision of the high technology sector and product classification, STI
Working papers. Paris: OECD.
He, Z-L, and P-K Wong. 2004. "Exploration vs. Exploitation: An Empirical Test of the Ambidexterity
Hypothesis." Organization Science 15 (4):481-494. doi: doi:10.1287/orsc.1040.0078.
Herrera, L, M. F Munoz-Doyague, and M Nieto. 2010. "Mobility of public researchers, scientific
knowledge transfer, and the firm's innovation process." Journal of Business Research 63
(5):510-518. doi: 10.1016/j.jbusres.2009.04.010.
Herstad, S.J. 2017. "Innovation strategy choices in the urban economy." Urban Studies 0
(0):0042098017692941. doi: 10.1177/0042098017692941.
Herstad, S.J, and T Brekke. 2012. "Globalization, Modes of Innovation and Regional Knowledge
Diffusion Infrastructures." European Planning Studies 20:1603-1625. doi:
10.1080/09654313.2012.713334.
Solheim & Herstad, 2017
Herstad, S.J, and B Ebersberger. 2014. "Urban agglomerations, knowledge intensive services and
innovation activity: Establishing the core connections." Entrepreneurship & Regional
Development 26 (03-04):211-233. doi: 10.1080/08985626.2014.888098.
Herstad, S.J, T Sandven, and B Ebersberger. 2015. "Recruitment, knowledge integration and modes of
innovation." Research Policy 44 (1):138-153. doi:
http://dx.doi.org/10.1016/j.respol.2014.06.007.
Herstad, S.J, T Sandven, and E Solberg. 2013. "Location, education and enterprise growth." Applied
Economics Letters 20 (10):1019-1022.
Hoetker, G. 2007. "The use of logit and probit models in strategic management research: Critical
issues." Strategic Management Journal 28 (4):331-343.
Hong, L, and S.E Page. 2004. "Groups of diverse problem solvers can outperform groups of high-
ability problem solvers." Proceedings of the National Academy of Sciences of the United
States of America 101 (46):16385-16389. doi: 10.1073/pnas.0403723101.
Horwitz, S.K. 2005. "The compositional impact of team diversity on performance: Theoretical
considerations." Human Resource Development Review 4 (2):219-245.
Horwitz, S.K, and I.B Horwitz. 2007. "The Effects of Team Diversity on Team Outcomes: A Meta-
Analytic Review of Team Demography." Journal of Management 33:987-1015.
Jacquemin, A.P, and C.H Berry. 1979. "ENTROPY MEASURE OF DIVERSIFICATION AND CORPORATE
GROWTH." Journal of Industrial Economics 27:359-369.
Jensen, M.B, B Johnson, E Lorenz, and B-Å Lundvall. 2007. "Forms of knowledge and modes of
innovation." Research Policy 36 (5):680-693. doi: 10.1016/j.respol.2007.01.006.
Jøranli, I, and S.J. Herstad. 2017. "Urban concentration and labour market linkages in the Norwegian
ICT services sector." European Planning Studies 25 (10):1734-1755. doi:
10.1080/09654313.2017.1337726.
Kanter, R.M. 1968. "Creating the creative environment." Management Review 75 (11-2).
Katz, R, and T. J Allen. 1982. "Investigating the not invented here syndrome - a look at the
performance, tenure and communication patterns of 50 R&D project groups." R & D
Management 12 (1):7-19.
Kleinknecht, A, F.N van Schaik, and H Zhou. 2014. "Is flexible labour good for innovation? Evidence
from firm-level data." Cambridge Journal of Economics. doi: 10.1093/cje/bet077.
Knight, D, C.L Pearce, K.G Smith, J.D Olian, H.P Sims, K.A Smith, and P Flood. 1999. "Top management
team diversity, group process, and strategic consensus." Journal of Strategic Management
20:445-465.
Laursen, K, and A Salter. 2006. "Open for innovation: The role of openness in explaining innovation
performance among UK manufacturing firms." Strategic Management Journal 27 (2):131-
150. doi: 10.1002/smj.507.
Levinthal, D.A, and J.G March. 1993. "The myopia of learning." Strategic Management Journal 14:95-
112.
Madsen, T, E Mosakowski, and S Zaheer. 2003. "Knowledge retention and personnel mobility: The
nondisruptive effects of inflows of experience." Organization Science 14 (2):173-191.
March, J.G. 1991. "Exploration and Exploitation in Organizational Learning." Organization Science 2
(1):71-87.
Mawdsley, J.K, and D Somaya. 2015. "Employee Mobility and Organizational Outcomes: An
Integrative Conceptual Framework and Research Agenda." Journal of Management 42
(1):85-113.
McPherson, M, L Smith-Lovin, and J.M Cook. 2001. "Birds of a feather: Homophily in social
networks." Annual Review of Sociology 27:415-444. doi: 10.1146/annurev.soc.27.1.415.
Miller, T, and M del Carmen Triana. 2009. "Demographic Diversity in the Boardroom: Mediators of
the Board Diversity–Firm Performance Relationship." Journal of Management Studies
46:755-786. doi: 10.1111/j.1467-6486.2009.00839.x.
Murray, A.I. 1989. "Top management group heterogeneity and firm performance." Strategic
Management Journal 10:125-141.
Solheim & Herstad, 2017
Nieto, M.J, and L Santamaría. 2007. "The importance of diverse collaborative networks for the
novelty of product innovation." Technovation 27 (6/7):367-377. doi:
10.1016/j.technovation.2006.10.001.
Nonaka, I. 1994. "A dynamic theory of organizational knowledge creation." Organization Science
5:14-37.
Nooteboom, B, W Van Haverbeke, G Duysters, V Gilsing, and A van den Oord. 2007. "Optimal
cognitive distance and absorptive capacity." Research Policy 36 (7):1016-1034. doi:
10.1016/j.respol.2007.04.003.
O’Reilly, C.A, and M.L Tushman. 2008. "Ambidexterity as a dynamic capability: Resolving the
innovator's dilemma." Research in Organizational Behavior 28:185-206. doi:
http://dx.doi.org/10.1016/j.riob.2008.06.002.
OECD. 2005. Oslo Manual: Guidelines for collecting and interpreting innovation data. Paris: OECD &
Eurostat.
Oettl, A, and A Agrawal. 2008. "International Labor Mobility and Knowledge Flow Externalities."
Journal of international Business Studies 39 (8):1242-1260.
Pitcher, P, and A.D Smith. 2001. "Top Management Team Heterogeneity: Personality, Power, and
Proxies." Organization Science 12:1-18.
Puhani, P.A. 2000. "The Heckman Correction for Sample Selection and Its Critique." Journal of
Economic Surveys 14 (1):53.
Raisch, S, J Birkinshaw, G Probst, and M.L Tushman. 2009. "Organizational Ambidexterity: Balancing
Exploitation and Exploration for Sustained Performance." Organization Science 20 (4):685-
695. doi: doi:10.1287/orsc.1090.0428.
Roper, S, J Du, and J.H Love. 2008. "Modelling the innovation value chain." Research Policy 37
(6/7):961-977. doi: 10.1016/j.respol.2008.04.005.
Rutten, R, and F Boekema. 2012. "From Learning Region to Learning in a Socio-spatial Context."
Regional Studies 46 (8):981-992. doi: http://www.tandfonline.com/loi/cres20.
Simons, S.M, and K.N Rowland. 2011. "Diversity and its Impact on Organizational Performance: The
Influence of Diversity Constructions on Expectations and Outcomes." Journal of technology
management and innovation 6 (3):171-183.
Spithoven, André, Wim Vanhaverbeke, and Nadine Roijakkers. 2012. "Open innovation practices in
SMEs and large enterprises." Small Business Economics:1-26. doi: 10.1007/s11187-012-9453-
9.
Sturman, M. 2003. "Searching for the inverted U-shape relationship between time and performance:
Meta-analysis of the experience/performance, tenure/performance and age/performance
relationships." Journal of Management 29:609 - 640.
Teece, D. 1986. "Profiting from technological innovation: Implications for integration, collaboration,
licensing and public policy." Research Policy 15:285-305.
Timmermans, B, and R Boschma. 2014. "The effect of intra- and inter-regional labour mobility on
plant performance in Denmark: the significance of related labour inflows." Journal of
Economic Geography 14:289-311.
Vaghely, I.P, and P-A Julien. 2010. "Are opportunities recognized or constructed?: An information
perspective on entrepreneurial opportunity identification." Journal of Business Venturing 25
(1):73-86. doi: http://dx.doi.org/10.1016/j.jbusvent.2008.06.004.
Van Engen, M, and M Van Woerkom. 2010. "Learning from Differences. The Relationships between
Team Expertise Diversity, Team Learning, Team Performance, and Team Innovation." In
Workplace Learning. Concepts, Measurement, and Application, edited by M Van Woerkom
and R Poell. NY/Oxon: Routledge.
van Knippenberg, D, and M.C. Schippers. 2007. "Work Group Diversity." Annual Review of
Psychology 58 (1):515-541. doi: 10.1146/annurev.psych.58.110405.085546.
Vandenbussche, J, P Aghion, and C Meghir. 2006. "Growth, distance to frontier and composition of
human capital." Journal of Economic Growth 11:97-127. doi: 10.1007/s10887-006-9002-y.
Solheim & Herstad, 2017
Wang, H.C, J.Y He, and J.T Mahoney. 2009. "Firm-specific knowledge resources and competitive
advantage: The roles of economic and relationship-based employee governance
mechanisms." Strategic Management Journal 30 (12):1265-1285. doi: 10.1002/smj.787.
Wiersema, M.F, and K.A Bantel. 1992. "Top Management Team Demography and Corporate Strategic
Change." The Academy of Management Journal 35 (1):91-121.
Zhou, H, R Dekker, and A Kleinknecht. 2011. "Flexible labor and innovation performance: evidence
from longitudinal firm-level data." Industrial and Corporate Change 20 (3):941-968. doi:
10.1093/icc/dtr013.
Østergaard, C.R, B Timmermans, and K Kristinsson. 2011. "Does a different view create something
new? The effect of employee diversity on innovation." Research Policy 40:500-509. doi:
10.1016/j.respol.2010.11.004.
Solheim & Herstad, 2017
Tables
Table 1: Example of experience diversity matrix. Firm with 20 employees.
Year of observation
Sector of employment in prior years
Employee no
2008
2007
2006
2005
2004
1
28.110
09.101
09.101
09.101
09.101
2
28.110
28.110
28.110
28.110
28.110
3
28.110
28.110
62.020
62.020
62.020
4
28.110
28.110
28.110
28.110
28.110
5
28.110
28.110
28.110
28.110
28.110
6
28.110
28.110
unemployed
unemployed
unemployed
7
28.110
28.110
28.110
28.110
28.110
8
28.110
28.110
28.110
62.020
62.020
9
28.110
28.110
28.110
28.110
28.110
10
28.110
28.110
28.110
28.110
28.110
11
28.110
28.110
28.110
28.110
28.110
12
28.110
28.110
28.110
unemployed
unemployed
13
28.110
28.110
28.110
28.110
28.110
14
28.110
28.110
28.110
28.110
28.110
15
28.110
28.110
24.421
24.421
24.421
16
28.110
24.422
24.422
24.422
24.422
17
28.110
28.110
28.110
26.110
26.200
18
28.110
28.110
28.110
28.110
28.110
19
28.110
28.110
26.300
26.300
26.300
20
28.110
28.110
28.110
28.110
28.110
Unrelated experience diversity (Entropy of distribution between 2-digit groups)
0.830069
+ Related experience diversity (Weighted entropy of distribution within 2-digit groups)
0.100334
= Total experience diversity (Entropy of distribution between 5-digit groups
0.930403
Solheim & Herstad, 2017
Table 2: Description of sample
All firms
ENGAGED firms only
Sample
EDUVAR_REL
EDUVAR_UNREL
EXPVAR_REL
EXPVAR_UNREL
Sample
PATENT
INNOVATION
Oil & gas. Mining
0.052
0.632
1.083
0.167
1.189
0.033
0.521
0.521
HT manufacturing
0.034
0.552
1.090
0.138
0.915
0.055
0.450
0.725
MHT manufacturing
0.105
0.411
1.032
0.109
0.912
0.149
0.389
0.681
MLT manufacturing
0.213
0.389
1.031
0.105
0.818
0.182
0.258
0.686
LT manufacturing
0.181
0.344
1.135
0.132
0.663
0.161
0.116
0.622
Publishing & printing
0.087
0.570
1.392
0.147
0.992
0.087
0.119
0.675
Telecom & ICTs
0.109
0.552
1.184
0.195
1.134
0.135
0.112
0.587
Finance & real-estate
0.071
0.569
1.172
0.170
0.906
0.044
0.016
0.469
Professional services
0.028
0.574
1.072
0.135
1.352
0.032
0.239
0.630
Scientific & technical serv.
0.093
0.591
0.893
0.145
1.050
0.110
0.358
0.547
Other business services
0.028
0.493
1.198
0.121
0.965
0.012
0.111
0.389
All
1
(N=2942)
0.474
1.105
0.138
0.919
1
(N=1450)
0.240
0.626
Note: Above country averages in bold.
Solheim & Herstad, 2017
Table 3: Description of variables used in the two estimation stages
Dependent variable Step 1 (selection variable)
ENGAGEMENT
= 1 if the firm is engaged in innovation activity
Dependent variables Step 2 (applicable only of ENGAGEMENT = 1)
PATENT
= 1 if the filed a patent application during 2008-2010
INNOVATION
= 1 if the firm introduced product or process innovations during
2008-2010
Explanatory variables Step 1 and Step 2 (applicable to all firms)
EDUVAR_TOTAL
The total variety of educational backgrounds among firms’
employees
EXPVAR_TOTAL
The total variety of work-life experiences among firms’
employees
EDUVAR_REL
The related (within main educational groups) variety of
educational backgrounds among firms’ employees
EDUVAR_UNREL
The unrelated (between main educational groups) variety of
educational backgrounds among firms employees
EXPVAR_REL
The related (within main industry groups) variety of work-life
experiences among firms’ employees
EXPVAR_UNREL
The unrelated (between main industry groups) variety of work-
life experiences among firms’ employees
Control variables Step 1 and Step 2 (applicable to all firms)
REPLACEMENT
The number of employees replaced during the 2008-2010 period
as proportion of employees in 2008
EDULEVEL
Average education level of staff in 2008
MARBREATH
The number of different geographical markets on which the firm
is present
SIZE
The log of employment size in 2008
SECTOR
10 dummies included to represent 11 sector groups given in Table
2
Additional control variables Step 2 (applicable only if ENGAGEMENT = 1)
INNOVINT
= 1 if innovation expenditures (R&D and non-R&D) are above
the sample median. Used to construct interaction terms and
compute marginal effects.
COBREATH
The number of different collaboration partners involved
specifically in firms’ innovation activities
Solheim & Herstad, 2017
Table 4: The probability of ENGAGEMENT = 1
Model 1
Model 2
Coeff
SE
Coeff
SE
SIZE
0,112
0,025***
0,109
0,026***
MARBREADTH
0,337
0,025***
0,336
0,025***
EDULEVEL
0,220
0,035***
0,201
0,038***
REPLACEMENT
-0,571
0,191***
-0,562
0,191***
EDUVAR_TOTAL
0,211
0,067***
EXPVAR_TOTAL
0,094
0,052*
EDUVAR_REL
0,360
0,119***
EDUVAR_UNREL
0,124
0,088
EXPVAR_REL
-0,233
0,190
EXPVAR_UNREL
0,165
0,064***
LR Chi2 (df)
621.40(16)***
626.93(18)***
Pseudo R2
0.1524
0.1537
Note: Coefficient estimates and robust standard errors from probit regression models. ***. ** and * indicate significance
at 1 per cent. 5 per cent and 10 per cent levels respectively. 10 jointly significant sector controls are included but not reported.
N= 2942
Table 5: Distribution of inventive and innovative outcomes
PATENT = 0
PATENT = 1
Sum
INNOVATION = 0
456
85
541
INNOVATION = 1
646
263
909
Sum
1102
348
1450
Solheim & Herstad, 2017
Table 6: The probability of PATENT = 1
Model 3
Equation A
Model 4
Equation A
Model 5
Equation A
Model 6
Equation A
Model 7
Equation A
Coeff
SE
Coeff
SE
Coeff
SE
Coeff
SE
Coeff
SE
SIZE
0.167
0.036***
0.144
0.037***
0.144
0.037***
0.161
0.039***
0.164
0.039***
MARBREADTH
0.134
0.039***
0.118
0.039***
0.118
0.039***
0.117
0.039***
0.117
0.039***
EDLEVEL
0.253
0.056***
0.217
0.057***
0.220
0.056***
0.233
0.062***
0.241
0.061***
REPLACEMENT
-0.469
0.320
-0.409
0.320
-0.410
0.325
-0.460
0.321
-0.442
0.328
INNOVINT
0.182
0.079**
0.561
0.342
0.181
0.079**
0.773
0.362**
COBREADTH
0.069
0.019***
0.068
0.019***
0.069
0.019***
0.068
0.019***
EDUVAR_TOT
-0.049
0.116
-0.052
0.116
0.124
0.160
EXPVAR_TOT
0.351
0.082***
0.357
0.082***
0.291
0.111***
INNOVINT*EDUVAR_TOT
-0.328
0.203
INNOVINT*EXPVAR_TOT
0.138
0.159
EDUVAR_REL
-0.146
0.182
-0.153
0.252
EDUVAR_UNREL
0.008
0.149
0.362
0.212*
EXPVAR_REL
-0.103
0.344
-0.433
0.417
EXPVAR_UNREL
0.434
0.102***
0.429
0.138***
INNOVINT*EDUVAR_REL
0.006
0.300
INNOVINT*EDUVAR_UNREL
-0.647
0.273**
INNOVINT*EXPVAR_REL
0.621
0.627
INNOVINT*EXPVAR_UNREL
0.030
0.192
Walds Chi2(df)
238.96(32)***
311.18(36)***
318.20(40)***
319.71(40)***
333.16(48)***
Note: Coefficient estimates from bivariate probit regressions, equation A. Model statistics are for the full models.
***. ** and * indicate significance at 1 per cent. 5 per cent and 10 per cent levels respectively. 10 jointly significant sector controls are included but not reported.
N= 1450
Solheim & Herstad, 2017
Table 7: The probability of INNOVATION = 1
Model 3
Equation B
Model 4
Equation B
Model 5
Equation B
Model 6
Equation B
Model 7
Equation B
Coeff
SE
Coeff
SE
Coeff
SE
Coeff
SE
Coeff
SE
SIZE
0.007
0.033
-0.036
0.035
-0.034
0.035
-0.040
0.036
-0.039
0.036
MARBREADTH
0.081
0.033**
0.058
0.033*
0.059
0.033*
0.062
0.034*
0.062
0.033*
EDLEVEL
0.059
0.048
0.004
0.049
0.003
0.049
0.017
0.054
0.015
0.054
REPLACEMENT
0.232
0.277
0.279
0.283
0.266
0.284
0.284
0.282
0.269
0.284
INNOVINT
0.181
0.070***
-0.380
0.293
0.184
0.070***
-0.473
0.307
COBREADTH
0.130
0.020***
0.130
0.020***
0.131
0.020***
0.132
0.020***
EDUVAR_TOT
-0.010
0.097
-0.017
0.099
-0.191
0.126
EXPVAR_TOT
0.078
0.071
0.080
0.072
0.096
0.095
INNOVINT*EDUVAR_TOT
0.367
0.169**
INNOVINT*EXPVAR_TOT
-0.040
0.138
EDUVAR_REL
-0.111
0.167
-0.218
0.206
EDUVAR_UNREL
0.038
0.124
-0.197
0.165
EXPVAR_REL
0.541
0.269**
0.696
0.343**
EXPVAR_UNREL
-0.011
0.088
-0.032
0.117
INNOVINT*EDUVAR_REL
0.227
0.256
INNOVINT*EDUVAR_UNREL
0.492
0.226**
INNOVINT*EXPVAR_REL
-0.367
0.523
INNOVINT*EXPVAR_UNREL
0.038
0.164
Note: Coefficient estimates from bivariate probit regressions, equations B. Model statistics are given in Table 4.
***. ** and * indicate significance at 1 per cent. 5 per cent and 10 per cent levels respectively. 10 jointly significant sector controls are included but not reported.
N= 1450
Solheim & Herstad, 2017
25
Table 8: Average marginal effects of cognitive diversity on innovation
outcomes
Subsamples
INNOVINT = 0
INNOVINT = 1
Mode 7 Equation A: PATENT = 1
Marg. Eff
SE
Marg. Eff
SE
EDUVAR_REL
-0.035
0.058
-0.041
0.063
EDUVAR_UNREL
0.084
0.049*
-0.080
0.054
EXPVAR_REL
-0.100
0.097
0.053
0.139
EXPVAR_UNREL
0.099
0.032***
0.130
0.039***
Mode 7 Equation B: INNOVATION = 1
Marg. Eff
SE
Marg. Eff
SE
EDUVAR_REL
-0.081
0.076
0.003
0.073
EDUVAR_UNREL
-0.073
0.061
0.100
0.058*
EXPVAR_REL
0.258
0.126**
0.112
0.138
EXPVAR_UNREL
-0.012
0.043
0.002
0.042
Model 7: PATENT = 1 & INNOVATION = 1
Marg. Eff
SE
Marg. Eff
SE
EDUVAR_REL
-0.038
0.042
-0.029
0.051
EDUVAR_UNREL
0.043
0.036
-0.034
0.042
EXPVAR_REL
-0.022
0.069
0.066
0.116
EXPVAR_UNREL
0.064
0.024***
0.095
0.032***
Solheim & Herstad, 2017
26
Table A1: Descriptive statistics & correlations. Engaged enterprises only.
Mean
SD
Min
Max
1
2
3
4
5
6
7
8
9
10
11
12
13
1
PATENT
0,240
0,427
0
1
1
2
INNOVATION
0,627
0,484
0
1
0,150
1
3
SIZE
3,968
1,183
2,303
9,771
0,147
0,004
1
4
MARBREADTH
2,671
1,072
1,000
4
0,180
0,085
-0,001
1
5
EDULEVEL
4,492
1,089
2,429
7,5
0,102
-0,024
-0,181
0,097
1
6
REPLACEMENT
0,184
0,128
0
1
-0,081
0,025
0,190
-0,045
-0,147
1
7
INNOVINT
0,492
0,500
0
1
0,100
0,100
-0,003
0,092
0,055
-0,031
1
8
COBREADTH
1,229
2,011
0
7
0,212
0,185
0,001
0,148
0,062
-0,056
0,107
1
9
EDUVAR_TOT
1,648
0,421
0
2,864
0,022
-0,003
0,115
0,030
0,220
0,052
0,045
0,044
1
10
EXPVAR_TOT
0,527
0,274
0
1,523
0,107
-0,015
0,037
0,094
0,500
-0,091
0,051
0,145
0,679
1
11
EDUVAR_REL
1,122
0,309
0
1,886
-0,066
0,010
0,124
-0,042
-0,145
0,151
0,015
-0,069
0,759
0,036
1
12
EDUVAR_UNREL
1,106
0,517
0
2,791
0,128
0,015
-0,284
0,042
0,268
0,087
0,012
0,019
0,209
0,184
0,121
1
13
EXPVAR_REL
0,143
0,142
0
1,172
0,022
0,030
-0,154
-0,028
0,104
0,044
-0,028
0,013
0,190
0,150
0,125
0,565
1
14
EXPVAR_UNREL
0,963
0,453
0
2,535
0,139
0,008
-0,277
0,057
0,273
0,085
0,022
0,018
0,179
0,164
0,099
0,966
0,332
Note: N = 1450
