AMBIDEXTERITY
AND BOARDROOM
DIVERSITY
The impact of boardroom diversity on
organizational innovation strategy
Demian Klapwijk Student number: s2056402 Master Change Management Supervisor: Björn Mitzinneck Co-assessor: Ileana Maris-de Bresser
Date: 2-7-2019 Word count: 9686
Abstract
The importance of combining explorative innovation and exploitative innovation resulting in organizations sporting so-called “ambidextrous designs” has been established since late last century. To enable such designs a heavy duty sits with top management to effectively integrate these two opposing strategies. Yet, little is known
regarding the top management characteristics which can handle such paradoxical tensions. This paper looks to identify such characteristics by borrowing insights from
related research areas. Based on a sample of S&P500 firms I find evidence that diversity of political ideology, gender diversity, and insiders versus outsiders diversity
within boards of directors have a significant impact on the overall innovation strategy of organizations. Whether the impact on innovation strategy constitutes a push towards ambidexterity depends on individual perception of balanced innovation. Nonetheless, the findings provide concrete guidance regarding inclusion of diversity dimensions in board of directors for organizations which know the direction they wish
1
Abstract
The importance of combining explorative innovation and exploitative innovation resulting in
organizations sporting so-called “ambidextrous designs” has been established since late last century. To enable such designs a heavy duty sits with top management to effectively integrate these two opposing strategies. Yet, little is known regarding the top management characteristics which can handle such paradoxical tensions. This paper looks to identify such characteristics by borrowing insights from related research areas. Based on a sample of S&P500 firms I find evidence that diversity
of political ideology, gender diversity, and insiders versus outsiders diversity within boards of directors have a significant impact on the overall innovation strategy of organizations. Whether the
impact on innovation strategy constitutes a push towards ambidexterity depends on individual perception of balanced innovation. Nonetheless, the findings provide concrete guidance regarding inclusion of diversity dimensions in board of directors for organizations which know the direction they
wish to develop.
Introduction
2
Within this paper I look towards previous research into political ideology and diversity to argue that boards rich in political diversity should be better equipped to tackle paradoxical tensions regarding innovation strategy and thus attain ambidexterity. First, I look to political ideology scholars who have identified the conservatism-liberalism dimension to be an important pillar of individual values (Chin, et al. 2013; Feather, 1979).
Among the preferences indicated by the conservatism-liberalism dimension are the propensity for risk taking and R&D spending (Hutton, Jiang, & Kumar, 2014). Both of those elements are logically determinants of an individual’s opinion on innovation strategy, since determining how much to spend, and how much risk to take are essential parts of innovation strategy.
Secondly, group diversity research is considered. Diversity research suggests that diversity will lead to task conflict, which is an essential factor for improving decision making (Stahl, Maznevski, Voigt, & Jonsen, 2010). Diversity in political ideology can thus be expected to create conflict regarding innovation strategy, which opens the discussion on how to balance and combine paradoxical tensions.
Based on the previous arguments, a hypothesis is developed that diversity in political ideology within a board of directors will positively impact the development of ambidexterity. This hypothesis will be tested in a sample of S&P500 companies. This is enabled through the operationalization of ambidexterity by patent-based measures and political ideology through political donations.
Subsequent analysis showed significant impact of political diversity and other diversity characteristics on innovation strategy. Although findings regarding ambidexterity specifically will depend on individual perception of balanced innovation strategy, political diversity, gender diversity, and diversity of insiders versus outsiders clearly impact the balance between explorative and
exploitative innovation.
3
Literature review
Paradox theory
Research with a focus on paradoxical tensions experienced by organization are part of paradox theory. Paradox theory has provided an alternative approach to opposing demands experienced by organizations, focusing on how solutions which combine tensions can improve overall performance (Smith & Lewis, 2011; He & Wong, 2004). As such it has provided practitioners with an additional perspective on organizational design next to contingency theory and postmodern lenses. With the emergence of global and digital markets organizations face progressively more dynamic environments and the accompanying paradoxical tensions. As a result, paradox theory has become increasingly relevant over the last decades.
Previous research has provided insights into paradoxical tensions within a number of domains, such as innovation (Tushman & O’Reilly, 1996), change (Seo & Creed, 2002), and identity (Fiol, Pratt, & O’Connor, 2009). These previous works have in common their argument that long-term sustainability requires a combination of the paradoxical demands experienced by organizations. From these domains, the research focusing on innovations has identified a paradoxical tension which has gathered much attention.
The paradoxical nature of explorative innovation aimed at developing new products versus exploitative investments to maximize profits of current products has been acknowledged for a long time (March, 1991). More than that, arguments for the importance of combining these two as a requirement for sustained performance in dynamic environment have been brought up since the early 90s (March, 1991). Within the next section the methods in which organization can combine these according to previous research is briefly summarized.
Organizations what wish to combine exploration and exploitation can aim to develop
4
and cultures into business units so that the firm can both explore and exploit” (Smith & Tushman, 2005). The balancing of exploration and exploitation “contributes to firm performance through more structured control of performance risk” (Cao, Gedajlovic, & Zhang, 2009, p. 783) The architectures required to enable ambidexterity combine highly differentiated units with top management team integration (He & Wong, 2004). The structural differentiation enables organization to employ opposing strategies simultaneously which is integrated into the overall business through the top management team (Smith & Tushman, 2005).
Smith & Tushman (2005) define two important tasks for top management teams in order to integrate opposing perspectives; (1) making balanced trade-offs when distributing resources between existing products or innovating new products; and (2) identifying synergies which will be beneficial for both products. This does however bring into question the kind of boards and top management teams which can handle such paradoxical demands. It is important to recognize that structural, psychological and social barriers can impede top management teams from effectively managing paradoxical tensions (Van de Ven, Poley, Garud, & Venkataraman, 1999).
He & Wong (2004) find that to enable ambidextrous designs, top management teams are required which can host internal inconsistencies. The rationale for internal inconstencies is elaborated upon by Smith & Tushman (2005); they explain that when top management teams focus on resolving inconsistencies this leads to distribution of resources rather than finding shared value. Since the inherent inconsistencies between exploitation and exploration cannot be eliminated (Cameron & Quinn, 1988), the conflict should be used to improve strategic decision making (Eisenhardt, Kahwajy, & Bourgeo, 1997).
5
Despite the insights which have been provided into the hypothetical board make-up to enable ambidextrous designs little is known about the team composition or personal characteristics of a board which could handle the previously described paradoxical tensions. Smith & Tushman (2005) confirm this statement as they write in their conclusion “there is limited literature on the characteristics of the senior team that can manage these complex strategies” (p. 534). The identified importance of conflict and variety of perspectives is the main guidance in solving this issue.
Transition to diversity
Task conflict and variety of perspectives are both elements which are associated with boardroom diversity. Stahl, Maznevski, Voigt, & Jonsen (2010) found, through meta analysis, that increased team diversity is asssociated with increased creativity and conflict. As a whole, a diverse group of directors represents a variety of competencies and capibilities which represent the social capital of an organization (van der Walt & Ingley, 2003).
The following three arguments have been made in support of the benefits of diveristy; (1) Andringa & Engstrom, (1998) establish that by employing a diversity of experience, gender and ethnicity organization take advantage of their differences can work succesfully together. (2) Thus, directors with variety of backgrounds can offer unique perspectives to improve strategic decision making (Bryan, 1995). And ultimately, (3) directors with unique perspectives are more likely to voice dissenting opinions (Carver, 2002), a quality which has been described as desired within the paradox theory.
This is not the only area on which paradox theory and diversity research see overlaps; Van der Walt & Ingley (2003) echo the earlier identified sentiments that more sophisticated approaches are required within the increasinly complex economy. Although their foundation is the nature of the multicultural and gender sensitive society, they also identify a more demanding and complex organizational environment.
6
dimension of diversity is impactful on the discussion regarding innovation strategy. The logical next step to consider is the specific characteristics which are strongly related to heterogeneity in opinions regarding innovation strategy. Research focusing on political ideology has interesting findings relevant to this question.
Inclusion of political ideology
Chin, Hambrick, & Treviño, (2013) established that political donations within the US can be used to measure the underlying political idealogy of liberalism versus conversatism. This result held tremendous significance as the conservatism-liberalism dimension is considered to be one of the central pillars of an individuals’ core values (Chin, et al. 2013; Feather, 1979). Additionally, as US data on political donations has to be recorded and is public domain this allowed researchers to measure personal values of corporate elites, something which had previously been difficult (Bonica, 2016).
Subsequent research by Hutton, Jiang, & Kumar (2014) found that political preference
influenced corporate policies, amongst which policies regarding innovation strategy. Specifically, they found that republican managers tend to have lower R&D spending and prefer investments which have lower risk associated with them. These discoveries provide a foundation for the linkage between political ideology and (opinions on) innovation strategy.
Combining elements
In summation; paradox theory has identified innovation as an environment in which the combination of paradoxical tensions is essential for sustained performance. This can be achieved through the implementation of ambidextrous designs. To enable such designs, it is important organizations harness conflict to improve decision making. Research into diversity confirms the importance of conflict and identifies diversity as an important source for conflict. Researchers with a focus on political ideology determined the conservatism-liberalism to predict preference of innovation strategy.
7
H1. Higher diversity of political ideology within an organization’s board will result in higher ambidexterity
Furthermore, based on previous findings that inclusion of multiple perspectives can increase decision making an additional hypothesis is established;
H2. The greater the size of the board the higher the ambidexterity will be
The support for the above hypothesis is certainly much less defined, still it could provide interesting results and is thus included with the primary analysis of hypothesis 1.
Methodology and data
Data
To test the hypothesis the companies and board of the S&P500 are used as a sample. Given the size of the included companies they can generally be anticipated to be involved with innovation activities. This makes the sample appropriate to analyze the dependent variable ambidexterity.
Additionally, a lot of data is available for the S&P500 companies which allows me to employ a variety of control variables.
For the construction of the dependent variable, patent citation data is used. Trajtenberg, et al. (1997) find that, as patents are part of a cumulative process which builds on previous research, cited sources and future citations will reveal information about the nature of research and its impact. Patent based measures have previously also been employed by Rosenkopf & Nerkar, (2001) and Katila & Ahuja, (2002) to measure ambidexterity.
8
2009). Given the scope of the research (all of the S&P500), the timeconstraints and the logistic difficulties of getting in contact with CEOs such an approach was not feasible for this research.
To measure the independent variable, political diversity, donation based data is used which has previously been established to measure political ideology (Chin, Hambrick, & Treviño, 2013; Bonica, 2016).
Variables
Dependent variable.
The dependent variable is ambidexterity, which is measured by the absolute distance between scores for explorative and exploitative innovation as proposed by Cao, Gedajlovic, & Zhang, (2009) and applied by Parida, Lahti, & Wincent (2016). The approach of, measuring ambidexterity on the basis of two separate dimensions for exploitation and exploration has been the general standard (e.g. Tushman & O’Reilly, 1996; He & Wong, 2004).
A. Exploration
To measure the explorative dimension the originality measure developed by Trajtenberg, Henderson, & Jaffe, (1997) is employed. Originality measures the total citations made by a patent within its own patent class compared to citations outside of its own patent class. The measure is calculated as one minus the Herfindahl concentration index and is aggregated per company per year for the uses of this research. The value ranges from 0 to 1 with a higher value indicating higher focus on exploration.
9
The originality data is included in the Patent Data project (PDP) of the National bureau of Economic Research (NBER1). The Patent Data Project was established Under the leadership of
Trajtenberg Henderson & Jaffe to enable future research (Lerner & Seru, 2017), it has been updated under leadership of Bronwyn Hall and Jim Bessen with patent data until the end of 2006.
B. Exploitation
Exploitative innovation is measured as the ratio of self citations versus total citations per company per year as previously employed by Kim, Song, & Nerkar (2012), Rosenkopf & Nerkar (2001), Song, Almeida, & Wu (2003) and Sorensen & Stuart (2000). The value ranges from 0 to 1 with a higher value indicating higher focus on exploitation.
Selfcitations represent the extent to which companies attempt to exploit their previous inventions through renewed investments. They are an indication of the extent to which organizations follow up to previous inventions which are “the conduit that lead to the appropriation of returns” Trajtenberg, et al. (1997, p. 29). As such they have been frequently adopted as a measure for
exploitative learning (Kim, Song, & Nerkar, 2012, Rosenkopf & Nerkar, 2001, Song, Almeida, & Wu, 2003, Sorensen & Stuart, 2000).
The ratio of selfcitations is calculated using patent data from the PDP. As this measure is not readily included within the dataset the score is calculated by looking at the complete citations records. Any citations is counted as a selfcitation if the citing and cited patent both belong to the same
organization as indicated by their GVKEY identifier. The selfcitations are totaled per GVKEY and year and divided by the total citations made per organization per year to calculate the ratio. The patent data is available until the end of 2006.
1https://sites.google.com/site/patentdataproject/ - no clear source reference available for the database as it
10 C. Ambidexterity
As stated previously, ambidexterity is calculated as the absolute difference between the scores for exploration and exploitation. In order to mitigate the possibility for multicollinearity the scores for exploitation and exploration are mean centered around zero.
The next step is to calculate the difference between the two by subtracting the score for exploration from exploitation. The resulting product is an indication of companies’ tendency to focus on a surplus of either explorative or exploitative innovation strategies; negative values indicate
explorative, values close to zero indicate ambidexterity and positive values indicate exploitative. I will refer to this specific variable as “innovation orientation” from this point onwards. Innovation
orientation is transformed into a variable which is positive for all values so that it displays the absolute difference between exploitative and explorative. Finally, to facilitate interpretation the ambidexterity score is subtracted from 1 such that a higher value indicates higher ambidexterity.
The previously mentioned mean centering of exploitative and explorative innovation is tremendously impactful on the final results. This is due to the exact characteristics of the two
variables; When both explorative and exploitative innovation are measured on a scale from one to zero the mean of explorative innovation, is much higher (more than 3 standard deviations higher) in
comparison to the mean of exploitation innovation (see table 3: descriptive statistics). As such, without mean centering the variables, practically any variable which either decreases explorative innovation or increases exploitative innovation will increase ambidexterity (See appendix: figure 10 & 11 for visualization through kernel density estimate graphs).
However, mean centering both values around the same value is not a perfect solution. Mean centering the values will match the mean value of explorative innovation with the mean value for exploitative innovation. However just because both are the mean value does not guarantee that this is the intersection at which exploitative and explorative innovation are truly balanced and thus
11
It is difficult to resolve the issue since the exact combination of measures for exploitative and explorative innovation used in this paper has not previously been used, hence there is no impartial frame of reference with which to establish “true” ambidexterity. Alternatively, there is also no sufficient data to establish long-term performance, which would be another way to establish
ambidexterity through confirmed links with performance (He & Wong, 2004). However, this does not make the measure meaningless, but it does mean that interpretation of ambidexterity is “subjectively” based on mean values of S&P500 companies.
As such, individual perception whether S&P500 companies are on average overly reliant on exploitative innovation, or conversely explorative innovation, will impact the interpretation of these results. To visualize results based on different assumptions additional analyses are made with the values for exploitative innovation shifted incrementally 0.1 positive and negative. Positive values are simulations if you consider organization in general too exploitative and negative values for if you consider them too explorative. In doing so it is possible to see the breakpoints until which the impact of independent variables improves. If such a breakpoint exists where the correlation first becomes stronger and more significant and subsequently weaker that breakpoint is a likely indication for the equilibrium value for that specific independent variable. To clarify consider the below snippet from the regression analysis results of political diversity’s effect on ambidexterity:
The effect of political diversity starts out at small significant effect size and is not significant. However, as the assumption for ambidexterity is shifted as indication by “adjusted by” it becomes much more significant. However, this significance has a breakpoint at an adjustment of -0.2 after which the relation weakens, thus indicating that political diversity “pushes” towards an equilibrium of -0.2. The kernel density graph (see appendix: figure 10 & 11) may be used to develop a feel for how much an increment of 0.1 represent in relation to the entire dataset. For additional context, the
non-Adjusted by none -0.1 -0.2 -0.3 -0.4 -0.5
12
mean adjusted scores of explorative and exploitative innovation are located at approximately -0.37 adjustment point.
Additionally, in order to reduce the ambiguity of the results from the main analysis alternative analyses with dependent variable “innovation orientation” are frequently considered. This variable is less impacted by the previous issue since it simply indicates a net reliance on exploitative or
explorative innovation based on the independent variables without defining a balanced state. For the same reason exploitative and explorative innovations are also tested as dependent variable.
Independent variable.
The independent variable is the political diversity of the board. As an antecedent for
calculating the political diversity the political leaning of individual board directors is measured using the methodology developed by Bonica (2016). The methodology assigns a score between -2 and 2 to each indivual to indicate their political leaning based on the intensity and direction of their political donations. The political diversity is measured as the standard deviation of the political ideology per board per year.
13
Bonica (2016) developed the Database on Ideology, Money in Politics, and Elections (DIME2)
to calculate the political ideology scores for fortune 500 directors by matching publicly available information on executives and board memberships against the database of contribution records using automated record-linkage methods. To expand the data set with S&P500 directors a similar
methodology was applied. Fuzzy matching techniques matched donation data to S&P500 board directors as included in the BoardEx database, subsequently the potential matches were manually checked by two independent coders. Finally, any coder disagreements were cooperative resolved one by one. The BoardEx database holds info from the year 2000 onwards. Combined with the availability of the dependent variable until 2006 the window of analysis will be 2000 until 2006.
Control variables
As the dependent variable is constructed based on two separate measures the control variables can be for either the combined ambidexterity variable or either of the components.
A. Controlling for the differences per industry
The effect of industry type has been widely recognized as impactful as different types of industries require different innovation strategies (Trajtenberg, Henderson, & Jaffe, 1997, Rosenkopf & Nerkar, 2001, Sorensen & Stuart, 2000). The industry effect should thus be controlled for in order to produce meaningful results.
14
In order to enable comparison between the different industries it is necessary to adjust both the variable for exploitation and exploration in such a way that effects on the component variables
themselves have been controlled for. In order to accomplish this, the value of exploitative and explorative innovation will be mean centered per industry category before they are used to establish the ambidexterity variable.
Finally, innovation strategy is not only exploitation and exploration separately but also the combined ambidexterity. Because of that industry type is also expected to impact ambidexterity separately from the effects on its components. As such, industry fixed effects on ambidexterity will be included as a control variable for the analysis of board diversity’s effect on ambidexterity.
B. Controlling for citation availability
Despite the frequent usage of self citation to measure exploitation (e.g. Kim, Song, & Nerkar, 2012, Rosenkopf & Nerkar, 2001), employing selfcitations comes with a few complications. The first issue to take into considersation is that companies without any previous patents will by default always have a selfcitations ratio of zero. Similarly, organizations with very few patents are much much more restricted in selfcitations and will accordingly see much lower self citation ratios.
Previous researchers have used a variety of methods to control for this effect; they have either focused on the companies within a dataset with the most patents, excluded any companies without significant innovation activities (as measured by patents) or used a dataset which only includes companies with at least 10 patents per year on average (Rosenkopf & Nerkar, 2001; Kim, Song, & Nerkar, 2012).
15
First, any organizations with on average lower than two patents per year for the past ten years will be excluded from analysis. This will (1) filter out companies which have too little patents to be able to speak of an innovation strategy and (2) will filter out likely skewed results due to low sample size.
Secondly, observations for exploitative innovation are categorized based on the citation availability. To facilitate grouping I look at the logarithmic function of self-citation availability which is shown in figure 1:
Figure 1: Density of citation availability
16
To check if the established variable has also a significant effect on explorative innovation the process was repeated for the variable of explorative innovation for which the patent availability fixed effects dummies also showed significance (p<0.01).
In order to enable comparison between the different groups in the ambidexterity variable it is necessary to adjust both the variable for exploitation and exploration in such a way that effects on the component variables themselves have been controlled for. In order to accomplish this, the value of exploitative and explorative innovation will be mean centered per group before they are used to establish the ambidexterity variable.
C. Controlling for innovation intensity
Controlling for the overall level of innovation is an extension to citation availability as they are based around the same argument with very similar measures. Innovation intensity is measured by patent count which has been extensively used for this purpose (Geiger & Makri, 2006; Hall, Jaffe, & Trajtenberg, 2001). As citation availability was constructed based on patent count of previous years the two measures are strongly related with a correlation level around 0.85. This brings along two questions, (1) are the variables too strongly correlated to both be included, and if so (2) which variable should logically be included?
17
adjustment showed no clear patterns and much lower level of variance based on number of patents compared to beforehand.
Overall, innovation intensity is not significant as a control for either exploitation or
exploration after the implementation of citation availability. As such the conclusion can be drawn that citation availability and innovation quantity are too highly correlated to both be included in the analysis.
The remaining question to be answered is whether innovation quantity should be included over citation availability. The downside of using innovation quantity is that it will not be able to control for young companies which have quickly grown and have high innovation quantity while not having any citation availability. Such companies could skew the exploitative innovation measure, because of that possibility citation availability provides a superior control variable for the purpose of this study.
18
D. Company age
The age of a company (measured as the years since founding) has also been previously established to be of significant impact on citation based measures (Cao, Gedajlovic, & Zhang, 2009). This is logically consistent as younger companies have less previous work to cite. The effect on exploitative innovation is especially significant for the first 20 years (see appendix: figure 6 & 8). Figure 7 & 9 (in the appendix) shows the relationship after adjustments have been made by the previously discussed controls. From the dispersion it is clear that, through the correlation of the other controls with company age, the most severe variance of the youngest companies has already been filtered out. In order to further control for this impact companies founded after 1998 are excluded as the short timeframe makes measuring exploitation not feasible. Since most companies are significantly older for the timeframe 2000 until 2006 only 14 observations are excluded.
0 .2 .4 .6 m e a n o f e x p lo ra ti v e i n n o v a ti o n 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 .1 .2 .3 m e a n o f a d ju s te d _ e x p lo it _ in n o v 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 .2 .4 .6 m e a n o f e x p lo ra ti v e i n n o v a ti o n 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Figure 2: Exploitative innovation per innovation intensity category
Figure 3: Adjusted exploitative innovation per innovation intensity category
Figure 4: Explorative innovation per innovation
19
To capture the potential residual effect of company age two actions are undertaken; first, the company’s age will be included as a control variable in the final analysis. Secondly, an alternative analysis is included in which all companies younger than 20 years are excluded.
E. Common control variables, available resources and organizational performance
Cao, Gedajlovic, & Zhang, (2009) argue that larger available resource provides organizations with more freedom to not balance their explorative and exploitative innovation as they have the slack resources to take such risks. To control for this effect comparative size within the S&P500 is included. However, since all companies included within this sample can be considered large this variable might show reduced significance.
Additionally, Cao, et al. (2009) point out the significance of organizational performance for ambidexterity as organizations are likely to prioritize differently depending on recent performance. To control for this effect, return on assets is included as a proxy for recent organizational performance.
F. Analyzing other dimensions of diversity
In order to determine if diversity in general impacts ambidexterity or the exact dimension of diversity is significant, two previously established dimensions of diversity will be included as independent variables. The included diversity dimensions are gender and insiders versus outsiders. Both dimensions have previously been established to be of significance for the performance of boardrooms (Gender: Adams & Ferreira, 2004; Insiders: Ingley & van der Walt, 2001). Additionally the data is available within the BoardEx dataset which allows for convenient usage.
20 G. Potential multicollinearity
None of the control variables were excluded while running panel data in Stata indicating that there is no severe problem with multicollinearity. To be certain there were no issues with
multicollinearity the variables were double checked by first running a regression and then checking their VIF values (table 1). All variables displayed VIF values lower than 1.5 which is extremely acceptable.
Variable VIF 1/VIF
Board size 1.41 0.706972 Comp age 1.3 0.767851 Company size 1.29 0.777104 Diversity gender 1.19 0.836837 Innovation inensity 1.15 0.871848 Diversity insiders 1.08 0.927619 Organizational performance 1.03 0.967163 Industry fixed effect 1.03 0.970379 Political diversity 1.03 0.973355 Mean VIF 1.17
Table 1: VIF scores
Testing hypotheses
In order to accurately assess the impact of any variable on the dependent variable it needs to be forwarded by several years. This is required as the measure is constructed based on granted patent data. As the patent application process takes on average 2.4 years (Hall, Jaffe, & Trajtenberg, 2001) impact of any independent variable will only be reflected within the dependent variable after multiple years. Additionally, the independent variable has an opposite effect since a change in board
21
However, a time gap beyond three years poses a problem, as the dataset only spans from 2000 until 2006 and increasing the gap beyond three years results in a significant loss in data and thus the predictive power of analysis. Luckily, the year to year variation in board composition is on average rather low (Bonica, 2016) which means a three-year gap is can reasonably expected to be sufficient as on average the boards had very similar composition in prior years.
For the analysis of time series data, a hierarchical panel regression with random-effect estimator was chosen. As the datasets are longitudinal panel regression is the suitable method of analysis. The hierarchical approach adds one additional variable at a time which allows the additional effect to be more clearly observed. Random-effect estimator was chosen based on the characteristics of the available data; As board compositions are relatively stable (Bonica, 2016) the change in board diversity from year to year is expected to be low. This, in combination with a timeframe of only four years the expectation for the dependent variable to vary based on the independent variable is low. As such fixed-effects panel regression, which looks at the effect of the dependent variable based on variance of the independent variable, is not suitable. Random-effects estimator, which compares between entities, is applied.
Results
Political diversity results
The results from the main analysis (table 2) are based on 573 observations on 175 different companies with an overall R2 of 0.0637. Within the main analysis I find no significant impact of
political diversity on ambidexterity. Alternative analysis without mean centered variables for
22
Using alternative dependent variables, innovation orientation, exploration, and exploitation (See appendix: table 6-8) the results show significant (p < 0.05) positive impact of political diversity on innovation orientation, significant negative impact on exploration and non-significant positive impact on exploitation.
Results from analyses with alternative assumptions (See appendix: table 9 & 10) show a breakpoint around -0.2 adjustment indicating that political diversity significantly pushes towards that balance of exploitative and explorative innovation.
Board size results
Board size did not show significant impact in any of the analyses it was included in, nor did it noticeably impact any of the other included variables. The variable does reach near significance levels (p = 0.104) at +0.1 adjustment levels but overall there is not enough support for board size to draw conclusions on.
Alternative dimensions of diversity
A. Insiders
The inclusion of insiders versus outsiders as a diversity dimension compared to inclusion as a ratio showed very similar results to each other. This is due to the fact that on average only 30 percent of board directors are insiders (see table 3: descriptive statistics). As such, in general higher
23 DV: ambidexterity 1 2 3 4 5 6 7 8 9 9b z P> |z| z P> |z| z P> |z| z P> |z| z P> |z| z P> |z| z P> |z| z P> |z| z P> |z| z P> |z| Political diversity 0.70 0.486 0.37 0.708 0.59 0.555 0.52 0.603 0.28 0.776 0.19 0.846 0.19 0.850 0.16 0.875 0.36 0.716 0.73 0.464 innovation intensity 2.02 0.043* 2.11 0.035* 1.77 0.077 1.01 0.313 0.98 0.328 1.01 0.313 1.09 0.276 1.19 0.236 0.79 0.430 industry fixed effect 4.17 0.000*** 4.22 0.000*** 3.42 0.001*** 3.46 0.001*** 3.50 0.000*** 3.52 0.000*** 3.28 0.001*** 2.91 0.004** boardsize 1.59 0.112 1.38 0.167 0.88 0.378 0.87 0.382 0.48 0.635 0.63 0.527 0.58 0.564 company size 0.78 0.435 0.69 0.489 0.70 0.486 0.70 0.482 0.72 0.473 0.62 0.533 company age 1.85 0.064 1.81 0.070 2.07 0.038* 2.52 0.012* 1.41 0.159 org. Perf. 0.60 0.549 0.57 0.570 0.68 0.495 0.42 0.671 Diversity insiders 1.64 0.100 1.31 0.189 1.60 0.109 Diversity gender -2.38 0.017* -2.09 0.037* constant 49.50 0.000 35.35 0.000 -0.29 0.773 -0.43 0.665 0.69 0.492 0.62 0.537 0.59 0.557 0.52 0.600 0.78 0.433 1.13 0.259 * p < 0.05; ** p < 0.01; *** p < 0.001
24
Table 3: descriptive statistics
Table 4: Correlation matrix
Ambidexterity Political diversity Innovation intensity Industry fixed effect Comp. size Comp. age Org. perf. Diversity insiders Diversity gender Board size Ambidexterity 1 Political diversity -0.04 1 Innovation intensity 0.10 0.09 1 Industry fixed effect 0.25 -0.08 -0.02 1 Board size 0.12 0.05 0.22 -0.01 1 Company size 0.11 0.01 0.32 0.01 0.40 1 Company age 0.14 0.07 0.07 -0.05 0.37 0.15 1 Organizational performance 0.07 0.02 -0.05 -0.08 0.06 0.01 0.11 1 Diversity insiders 0.01 0.00 0.02 0.00 0.12 0.10 -0.14 -0.02 1 Diversity gender -0.06 0.12 0.09 -0.14 0.19 0.09 0.33 0.14 -0.14 1
Variable Obs Mean Std. Dev. Min Max
Ambidexterity 686 0.61 0.16 0.16 1 Innovation orientation 686 0.10 0.18 -0.38 1.03 Exploitative innovation 686 0.26 0.12 0.00 1 Explorative innovation 686 0.64 0.12 0.13 0.99 Political diversity 2,925 0.65 0.24 0.02 1.69 Innovation intensity 1,313 3.84 1.52 1.10 8.2
Industry fixed effect 2,018 0.64 0.04 0.56 0.77
Insider diversity showed near significant positive effects (p > 0.1 & p < 0.2) within the main analysis (Table 2). Additionally, insider diversity showed strongly positive significance (p < 0.001) on innovation orientation, significant (p < 0.05) positive impact on exploitative innovation and significant negative impact on explorative innovation. The significance and effect of insider diversity peaks at an adjustment level -0.3.
B. Gender
When looking at ratio of males versus gender diversity the two show opposite results across the different analyses. This is for a similar (but opposite) reason as the insiders’ dimension; on average of 87 percent of board directors are male, as a result lower ratio of males is essentially the same thing as more balanced gender diversity. Once again, the diversity dimension shows better predictive power and significance leading to the inclusion of the diversity measure.
Gender diversity showed significant negative correlation with ambidexterity. Additionally, gender diversity showed clear significant impact (p < 0.01) on exploitative innovation.
Looking at alternative interpretation adjustment scores, gender diversity never reaches a positive breakpoint. There is however a negative breakpoint around +0.1 adjustment values thus indicating that the absence of gender diversity pushes towards that value as its equilibrium. Given the earlier discussed details of gender diversity it is likely this is the equilibrium point for ratio males.
Company age
Discussion
This paper applied knowledge from group diversity and political ideology research to build support for the potential contribution of diversity of political ideology within boards on the
development and, once established, maintaining of organizations with ambidextrous designs. The main hypothesis was tested by analyzing the degree to which organizations balanced their explorative and exploitative innovation.
Impact of political ideology
Results did not support the direct link between political ideology within the board of directors and the degree to which ambidexterity was achieved. However, the data does support the notion that political diversity pushes towards an alternative equilibrium which relatively has moderately stronger emphasis on exploitative innovation and less on explorative innovation. This state is achieved through increased focus on exploitative innovation. When the distance between exploitative and explorative innovation crosses the moderate threshold, the impact of political diversity on ambidexterity weakens.
Altogether, the theorized impact of political diversity on conflict regarding innovation strategy is likely to be true as political diversity has a clear impact on innovation strategy. However, whether this impact also constitutes improved decision making and increased ambidexterity cannot be determined without defining which exact combination of explorative and exploitative innovation constitutes ambidexterity.
Looking to the influence of board size
Testing hypothesis two, the effect of board size on ambidexterity, sees no significant results. There are however some near significant results which indicate larger board size might push towards an innovation equilibrium with slightly more emphasis on explorative innovation, potentially
Gender and insider diversity
The two established diversity dimensions which were included in analyses, gender diversity and insiders versus outsiders, provided interesting discoveries. First, insiders diversity was found to strongly impact innovation orientation towards a more exploitative balance through increased focus on exploitative innovation and decreased focus on explorative innovation. This results in a drive towards an equilibrium which has a higher emphasis on exploitative innovation, even more so than the
equilibrium of political diversity.
An important question to consider is the likelihood that the equilibrium towards which insider diversity pushes might represent true ambidexterity. In general, the further the equilibrium lies from the mean centered ambidexterity (at zero adjustment) the more divergent the strategies is from the majority. It seems unlikely that ambidexterity would be very far removed from the majority which is centered around the mean. This could be an indication that insider diversity pushes companies too much towards an exploitative innovation state resulting in unbalanced innovation. The same can be said for political diversity, although to a lesser extent as the equilibrium is more centered towards the mean. However, as ambidextrous designs are known to be difficult to achieve with limited information on the topic, it is certainly possible that the average organization is quite a long way from achieving balance.
Although the impact of insider diversity has not been discussed in detail in the literature review of this paper, logically the results make a lot of sense; Those working inside of an organization are likely to have a different view on where their organization should go and which innovation strategy is required to get there compared to those that are outsiders. These divergent opinions can create conflict regarding the topic of innovation strategy, which is linked to altered decision making through group dynamics as was discussed with regard to political diversity.
establish with certainty, it is a diversity effect rather than a “more insiders” effect, organizations which have more insiders on their board then outsiders have to be examined in more detail.
Gender diversity is shown to have an especially strong impact on exploitative innovation. This effect is in fact so strong that only under the assumption that organizations are much too explorative gender diversity reaches near significant levels of support in favor of ambidexterity. A positive breakpoint is not observed indicating that gender diversity is strictly pushing towards more exploitative organizations without regard for a balanced state of exploitative and explorative innovation. Conversely, a negative equilibrium is found at an innovation strategy with a slight
emphasis on explorative innovation, signaling a balanced state towards which boards with low gender diversity are driven. Considering that S&P500 boards are on average very much male dominated, this can in most cases also be interpreted as the balanced state towards which organizations are pushed through the inclusion of males on the board.
These results confirm findings from previous research that gender is impactful on board performance and innovation strategy. The negative effect of gender diversity on ambidexterity is possibly connected to a preference for different conflict resolution methods by women. Holt & De Vore, (2005) confirmed through meta-analysis that women focus more on compromises compared to their male counterparts. Previously the importance of using conflict to increase the quality of decision making was established as a mechanism to enable ambidextrous designs. Hence it is possible that through too much focus on making compromises there is not enough conflict in order to enable
ambidextrous designs. Alternatively, it is possible that women place much higher value on exploitative innovation or, in a related vein of thought, they are more focused on finding ways to implement additional exploitative innovation together with the existing explorative innovation leading to lower ambidexterity.
Implications
board of directors as these have significant impact. Similarly, those interested in explorative
innovation should look to political diversity and insider diversity. The most important implication is that organizations must realize that different diversity characteristics have distinct impacts on
innovation strategy and move towards different equilibriums. When establishing a board of directors, criteria for new members should be determined based on the current position of the organizations innovation strategy and where they wish to move.
From an academic perspective this paper has shown that a variety of diversity characteristic impacts innovation strategy and the resulting balance between exploitative and explorative innovation. Although interpretation of results has been slightly less straightforward due to the inability to pinpoint a single universal intersection between explorative and exploitative innovation which can be
considered ambidextrous, the findings that diversity of insiders, diversity of ideology, and proportion of males on boards of directors all drive innovation strategy towards certain balanced states have contributed to the understanding of how board room dynamics impact innovation strategy.
Limitations and suggestions for further research
This study has several limitations that need to be addressed. The first of these is regarding the validity of ambidexterity. The balanced state of innovation which is identified as ambidexterity is based on mean centered explorative and exploitative innovation as present in the S&P500. Although this is a reasonable assumption to make, there is no guarantee that this intersection of variables reflects true ambidexterity within the real world. Accordingly, interpretation of results involving ambidexterity should be in reference to your opinion on whether S&P500 organizations generally focus too much on exploitation or exploration.
confirmative study with alternative measures or an alternative, preferably more longitudinal, sample could alleviate these concerns.
Thirdly, the timeframe of four years is not sufficient to make statements on long-term effects which significantly limits the extent to which conclusions can be drawn.
A fourth limitation is the industry effect on innovation strategy. This research was done based on aggregated data from a wide variety of industries. Although rigorous effort was undertaken in order to control for industry effects, results on innovation strategy should ideally be considered within a single industry to grasp the full context.
Acknowledgements
Appendix
0 .0 5 .1 m e a n o f e x p lo it a ti v e _ in n o v a ti o n 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 0 .1 .2 .3 m e a n o f a d ju s te d _ e x p lo it _ in n o v 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 0 .2 .4 .6 .8 m e a n o f e x p lo ra ti v e i n n o v a ti o n 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 0 .2 .4 .6 .8 m e a n o f a d ju s te d e x p lo ra ti v e i n n o v a ti o n 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Figure 9: Mean of adjusted explorative innovation for companies younger than 20 years
Figure 8: Mean of explorative innovation for companies younger than 20 years
Figure 7: Mean of adjusted exploitative innovation for companies younger than 20 years
dv:ambidexterity Coef. Std. Err. z P>z [95% Conf. Interval] political_diversity 0.070 0.030 2.34 0.019 0.011 0.129 innovation_intensity 0.019 0.007 2.83 0.005 0.006 0.032 industry_fixed_effect 1.807 0.472 3.83 0.000 0.882 2.732 boardsize -0.001 0.003 -0.44 0.663 -0.007 0.004 company_size -0.002 0.006 -0.35 0.730 -0.015 0.010 comp_age 0.000 0.000 0.71 0.477 0.000 0.001 org_performance 0.006 0.019 0.33 0.740 -0.032 0.044 d_insiders 0.137 0.040 3.45 0.001 0.059 0.215 d_gender 0.109 0.085 1.28 0.201 -0.058 0.276 _cons -0.694 0.305 -2.28 0.023 -1.292 -0.096
Table 5: analysis results without mean centered variables
Innovation orientation Coef. Std. Err. z P>z [95% Conf. Interval] Political diversity 0.069 0.031 2.22 0.026 0.008 0.130 Innovation intensity 0.012 0.007 1.86 0.063 -0.001 0.026 Board size -0.001 0.003 -0.19 0.848 -0.006 0.005 Company size -0.003 0.007 -0.39 0.697 -0.016 0.011 Comp age 0.000 0.000 0.05 0.958 0.000 0.001 Org performance 0.005 0.020 0.26 0.795 -0.034 0.044 Diversity insiders 0.135 0.041 3.3 0.001 0.055 0.215 Diversity gender 0.148 0.089 1.67 0.094 -0.025 0.322 _cons -0.211 0.348 -0.61 0.544 -0.892 0.471
Table 6: Results analysis with dependent variable innovation orientation
dv:explorative Coef. Std. Err. z P>z
[95% Conf. Interval] political_diversity -0.056 0.023 -2.42 0.015 -0.101 -0.011 Innovation intensity -0.008 0.005 -1.57 0.117 -0.017 0.002 boardsize -0.001 0.002 -0.43 0.665 -0.005 0.003 comp_age 0.000 0.000 1.35 0.175 0.000 0.001 org_performance -0.003 0.015 -0.17 0.862 -0.033 0.027 d_insiders -0.081 0.032 -2.53 0.011 -0.143 -0.018 d_gender 0.060 0.067 0.89 0.373 -0.072 0.191 _cons 0.085 0.032 2.67 0.008 0.022 0.147
dv:exploitative Coef. Std. Err. z P>z [95% Conf. Interval] Political diversity 0.021 0.022 0.93 0.354 -0.023 0.064 Innovation intensity 0.000 0.005 0.09 0.930 -0.009 0.010 Board size -0.002 0.002 -0.84 0.401 -0.006 0.002 Comp age 0.000 0.000 0.14 0.887 0.000 0.000 Org performance 0.008 0.015 0.52 0.606 -0.021 0.037 Diversity insiders 0.062 0.031 2.01 0.045 0.001 0.122 Diversity gender 0.188 0.064 2.92 0.004 0.062 0.315 _cons -0.044 0.030 -1.44 0.150 -0.104 0.016
Table 8: Results analysis with dependent variable exploitative innovation
Table 10: results assuming companies are too exploitative
Adjusted by none -0.1 -0.2 -0.3 -0.4 -0.5 z P> | z | z P> | z | z P> | z | z P> | z | z P> | z | political_diversity 0.36 0.716 1.92 0.055 2.74 0.006 2.66 0.008 2.32 0.020 2.29 0.022 innovation_intensity 1.19 0.236 2.17 0.030 2.63 0.008 2.58 0.010 2.92 0.003 3.08 0.002 industry_fixed_effect 3.28 0.001 3.97 0.000 3.72 0.000 3.74 0.000 3.84 0.000 3.91 0.000 boardsize 0.63 0.527 0.05 0.963 0.02 0.982 -0.22 0.825 -0.49 0.624 -0.58 0.562 company_size 0.72 0.473 0.35 0.730 -0.22 0.825 -0.28 0.781 -0.36 0.716 -0.40 0.691 comp_age 2.52 0.012 2.37 0.018 1.49 0.136 0.95 0.342 0.61 0.539 0.43 0.668 org_performance 0.68 0.495 1.11 0.269 0.57 0.570 0.39 0.696 0.34 0.732 0.34 0.735 d_insiders 1.31 0.189 2.58 0.010 3.16 0.002 3.53 0.000 3.32 0.001 3.25 0.001 d_gender -2.38 0.017 -0.34 0.735 0.68 0.498 1.04 0.300 1.33 0.183 1.40 0.161 _cons 0.78 0.433 -0.73 0.467 -1.47 0.140 -2.00 0.046 -2.36 0.018 -2.62 0.009
Companies too explorative
Table 9: results assuming companies are too explorative
Adjusted by +0.1 +0.2 +0.3 +0.4 +0.5 z P> | z | z P> | z | z P> | z | P> | z | z P> | z | z political_diversity -1.13 0.259 -1.36 0.175 -1.79 0.074 -2.17 0.030 -2.23 0.026 innovation_intensity -0.84 0.400 -2.63 0.008 -3.21 0.001 -3.15 0.002 -3.15 0.002 industry_fixed_effect 2.82 0.005 3.61 0.000 3.88 0.000 3.75 0.000 3.73 0.000 boardsize 1.63 0.104 1.07 0.283 0.72 0.469 0.62 0.535 0.61 0.541 company_size 0.41 0.685 0.28 0.783 0.40 0.687 0.40 0.692 0.41 0.683 comp_age 0.75 0.453 0.05 0.963 0.01 0.989 0.04 0.969 0.02 0.987 org_performance -0.64 0.520 -1.31 0.189 -0.40 0.689 -0.33 0.745 -0.29 0.768 d_insiders -1.40 0.163 -2.58 0.010 -3.21 0.001 -3.26 0.001 -3.25 0.001 d_gender -2.51 0.012 -1.64 0.101 -1.35 0.177 -1.50 0.133 -1.46 0.143 _cons 0.94 0.349 -0.76 0.448 -1.44 0.151 -1.29 0.197 -1.15 0.249
Citations
Adams, R. B., & Ferreira, D. (2004). Gender Diversity in the Boardroom. European Corporate Governance Institute; Finance Working Paper.
Bonica, A. (2016). Avenues of influence: on the political expenditures of corporations and their directors and executives. Business Politics, 367-394.
Bryan, J. H. (1995). Allegiance to a Diverse Board,. Directors & Boards, 6-8.
Cameron, K., & Quinn, R. (1988). Organizational paradox and transformation. Cambridge: Ballinger Publishing Company,.
Campbell, K., & Minguez-Vera, A. (2008). Gender Diversity in the Boardroom and Firm Financial Performance. Springer, 435-451.
Cao, Q., Gedajlovic, E., & Zhang, H. (2009). Unpacking organizational ambidexterity: dimensions, contingencies, and synergistic effects. Organization Science, 20, 781-796.
Carver, J. (2002). On Board Leadership. New York: Jossey-Bass, John Wiley.
Chin, M. K., Hambrick, D. C., & Treviño, L. K. (2013). Political Ideologies of CEOs: The Influence of Executives' Values on Corporate Social Responsibility. Administrative Science Quarterly, 198-232.
Eisenhardt, K. M., Kahwajy, J. L., & Bourgeo, L. J. (1997). Conflict and strategic choice: How top management teams disagree. California Management Rev. 39, 42-62.
Feather, N. (1979). Value correlates of conservatism. Journal of Personality and Social Psychology, 37, 1617-1630.
Fiol, C. M., Pratt, M. G., & O’Connor, E. J. (2009). Managing intractable identity conflict. Academy of Management, 32-55.
Geiger, S. W., & Makri, M. (2006). Exploration and exploitation innovation processes: The role of organizational slack in R & D intensive firms. The Journal of HIgh Technology Management Research, 97-108.
Hall, B. H., Jaffe, A. B., & Trajtenberg, M. (2001). The NBER Patent Citation Data File:Lessons, Insights and Methodological tools.
He, Z.-L., & Wong, P.-K. (2004). Exploration vs. Exploitation: An Empirical Test of the Ambidexterity Hypothesis. Organization Science, Vol. 15, No. 4, 481-494.
Heenan, D., & Bennis, W. (1999). Co-Leaders: The Power of Great Partnership. New York: John Wiley & Sons.
Holt, J. L., & De Vore, C. J. (2005). Culture, gender, organizational role, and styles of conflict resolution: A meta-analysis. International Journal of INtercultural Relations: 29, 165-196. Hutton, I., Jiang, D., & Kumar, A. (2014). Corporate Policies of Republican Managers. JOURNAL OF
Ingley, C. B., & van der Walt, N. T. (2001). The Strategic Board: The Changing Role of Directors in Developing and Maintaining Corporate Capability. Corporate Governance: An International, 174-185.
Katila, R., & Ahuja, G. (2002). 2002. Something old, something new: A longitudinal study of search behavior and new product introduction. Academy of Management Journal 45, 1183-1194. Kim, C., Song, J., & Nerkar, A. (2012). Learning and innovation: Exploitation and exploration
trade-offs. Journal of Business Reserach 65, 1189-1194.
Lerner, J., & Seru, A. (2017). The Use and Misuse of Patent Data: Issues for Corporate Finance and Beyond. National Bureau of Economic Research.
Lewis, M. W. (2000). Exploring Paradox: Toward a More Comprehensive Guide. The Academy of Management Review, 760-776.
March, J. (1991). Exploration and exploitation in organizational learning. Organ. Sci. 2, 71-87. Miller, K. D., & Leiblein, M. J. (1996). Corporate Risk-Return Relations: Returns Variability versus
Downside Risk. The Academy of Management Journal, Vol. 39, No. 1, 91-122.
Miller, T., & Triana, M. d. (2009). Demographic Diversity in the Boardroom: Mediators of the Board Diversity–Firm Performance Relationship. Journal of Management Studies, 755-786. Milliken, F. J., & Martins, L. (1996). Searching for Common Threads: Understanding the Multiple
Effects of Diversity in Organizational Groups. Academy of Management Review. Parida, V., Lahti, T., & Wincent, J. (2016). Exploration and exploitation and firm performance
variability: a study of ambidexterity in entrepreneurial firms. International Entrepreneurship and Management Journal, 12 , 1147-1164.
Rosenkopf, L., & Nerkar, A. (2001). Beyond local search: boundary-spanning, exploration, and impact in the optical disk industry. Strategic Management Journal 22, 287-306.
Seo, M., & Creed, W. (2002). Institutional contradictions, praxis, and institutional change: A dialectical perspective. Academy of Management Review, 27:, 222-247.
Smith, W. K., & Lewis, M. W. (2011). Toward A Theory of Paradox: A Dynamic Equilibrium Model of Organizing. THE ACADEMY OF MANAGEMENT REVIEW, 381-403.
Smith, W. K., & Tushman, M. L. (2005). Managing Strategic Contradictions: A Top Management Model for Managing Innovation Streams. Organization Science, Vol, 16 No. 5,, 522-536. Song, J., Almeida, P., & Wu, G. (2003). Learning-by-hiring: when is mobility more likely to facilitate
interfirm knowledge transfer? Management Science 49, 351-365.
Sorensen, J., & Stuart, T. (2000). Aging, obsolescence, and organizational innovation. Administrative Science Quarterly 45, 81-112.
Stahl, G. K., Maznevski, M. L., Voigt, A., & Jonsen, K. (2010). Unraveling the effects of cultural diversity in teams: A meta-analysis of research on multicultural work groups. Journal of International Business Studies, 690-709.
Tushman, M. L., & O’Reilly, C. A. (1996). Ambidextrous organizations: Managing evolutionary and revolutionary change. California Management Review, 38(4):, 8-30.
Van de Ven, A. D., Poley, R., Garud, S., & Venkataraman. (1999). The Innovation Journey. New York: Oxford Press.
