Board Composition and Innovation Performance Does Organizational Form Matter?

Master Thesis

Board Composition and Innovation Performance Does Organizational Form Matter?

– Empirical Evidence from Indian Pharmaceutical Industry–

MSc. International Business & Management Faculty of Economics and Business University of Groningen, the Netherlands

Frantisek Demi (s2966425) E-mail: f.demi@student.rug.nl

Supervisor: Dr. S.R. Gubbi

Co-assessor: Dr. B.J.W. Pennink

January 23

, 2017

2 Board Composition and Innovation Performance: Does Organizational Form Matter?

Empirical Evidence from Indian Pharmaceutical Industry

Frantisek Demi, Faculty of Economics and Business, University of Groningen, January 2017

ABSTRACT

In this study, I examine the impact of gender diversity and director independence on innovation performance. Moreover, also how is this relationship influenced by organizational form. By using firm-level data of Indian pharmaceutical companies, both business groups and standalones, it was found out, that the organizational form indeed partially moderates the relationship between board composition characteristics and firm’s innovation performance. However, the relationship was found to be statistically insignificant and too inconsistent therefore not generalized, as the provided results differed based on the used estimation method.

Keywords: Innovation, gender diversity, director independence, business groups

3 ACKNOWLEDGEMENTS

First of all, I would like to thank my supervisor Dr. S. R. Gubbi for his guidance and support during the entire research process. His comments and feedback were invaluable. Additionally, I would also like to sincerely thank all the friends that helped me and provided me support during the whole studies.

Lastly, I would also like to thank my co-assessor Dr. B. J. W. Pennink for reviewing and assessing the

entire thesis.

4 TABLE OF CONTENTS

INTRODUCTION ... 6

LITERATURE REVIEW ... 9

Innovation ... 9

Corporate Governance in India ... 10

Board composition ... 11

Conceptual model ... 16

RESEARCH METHODOLOGY ... 17

Research context: Indian pharmaceutical industry ... 17

Data source and sample operationalization ... 17

Measures ... 18

Descriptive statistics ... 20

Pearson correlations and multicollinearity test ... 21

Method of analysis ... 22

RESULTS ... 25

Fixed-effects model ... 25

Negative binomial model ... 27

Robustness checks ... 29

DISCUSSION ... 31

CONCLUSION... 33

Theoretical implications ... 33

Limitations and suggestions for further research ... 33

REFERENCES ... 35

APPENDICES: Appendix 1: Peareson correlation matrix ... 42

Appendix 2: Variance inflation factor ... 43

5

Appendix 3: Normality of distribution test of the dependent variable ... 44

Appendix 4: Hausman test ... 45

Appendix 5: Likelihood-ratio tests of Alpha (Models 5-8) ... 45

Appendix 6: Robustness checks – Fixed effect models ... 46

Appendix 7: Robustness checks – Negative binomial models ... 47

6 INTRODUCTION

The composition of board of directors (BOD) as well as organizational innovation performance have received increasing attention from researchers in previous years as they are both considered to constitute important determinants of organizational future performance. Board composition nowadays is one of the cornerstone issues of corporate governance. The BOD is one of the internal governance mechanisms to ensure alignment of interests of shareholders and managers. However, research has shown that the role of corporate boards reaches beyond solely monitoring function to ensure alignment of interests among stakeholders, as it is often argued by agency theorists (Dalton, Daily, Certo & Roengpitya, 2003). Next to that, in line with the resource-dependence perspective (Boyd, 1990; Hillman & Dalziel, 2003), it was found that boards further provide crucial resources to organizations through knowledge, embedded in the human (expertise, experience) and relational capital (networks) of board members (Pfeffer, 1972; Burt, 1992).

Following this perspective, many studies examined the relationship between board composition diversity and firm performance, focusing among others issues related to diversity, such as age or gender of directors or director independence, concluding mostly with mixed results (Sicialiano, 1996; Carter et al., 2003, Hillman, 2002; Erhardt et al., 2003; Adams & Ferreira, 2008 etc), leaving this area of research open for further discussion. Deeper examination also provided evidence of downsides occurring in highly diverse corporate boards. Namely, distrust, lack of cohesion, higher coordination costs or higher conflict occurrence (Huse, 2007; Milliken & Martins, 1996), resulting in the boards’ inability to operatively react to changing market environment due to slower decision-making. On the other hand, board composition with various organizational outcomes, such as better CSR performance (Bear et al., 2010), enhanced problem-solving capability (Murray, 1989, Carter et al., 2003) and higher creativity (Bantel & Jackson, 1989) consequently having positive effect among other also on innovation performance (Miller & Triana, 2009)

The ability to innovate is increasingly important also in the context of emerging market (EM)

countries, as these countries are becoming competitive also on the field of innovations

(Govindarajan & Ramamurti, 2011), eventually resulting in increasing share of innovations

introduced in EMs (Economist, 2011). Together with this shift in global economy, also researchers

started to focus more on EM context. Research related to board composition in EMs recently

provided evidence about prevailing low gender diversity on senior management positions (Sanan,

7 2016; Catalyst, 2012) or about families dominating organizational boards (Chung & Luo, 2008) with a relatively low share of independent directors, leading to the first research questions of this thesis:

(1) How does gender diversity and director independence influence firms’ innovation performance in India?

Innovation performance nowadays is perceived as one of the crucial factors needed to survive on highly competitive markets (Cefis & Marsili, 2006). Firms are forced to relentlessly innovate their products, services and management practices with increasingly shorter innovation cycles (Kock et. al, 2015). However, unlike more developed counterparts, EMs suffer from institutional voids, such as underdeveloped institutional environment, non-functional financial markets, weak legal framework or insufficient patent protection etc. (Khanna and Palepu, 1997). Creating additional obstacles of already risky innovation projects. Organizational response to this business environment resulted in the emergence of business groups (BG) as an organizational form capable of compensating prevailing market imperfections. Another distinctive feature of these groups is the high share of family ownership combined with high family presence on senior management positions. Moreover, BGs are characterized by formal and informal ties among legally independent firms (Khanna & Rivkin 2001). However, they are mainly coordinated from a central entity (Leff, 1978). Therefore, second research question is related to the influence of organizational form on the main relationship

(2) How does the organizational form (BGs x Standalones) influence the main relationship between board composition and innovation performance?

To provide answers to these questions, this thesis examines a sample consisting of 119 (86

standalones and 33 BGs) operating in the Indian pharmaceutical industry. Country and market

selection of this research is critical, as India represents an example of an EM country with strong

presence of BGs, and pharmaceutical industry represents a highly competitive market, with strong

emphasis on innovation performance. This thesis builds upon data collected over a six-year period

spanning between years 2006-2011. The time horizon of this study ends before Company Act 2013

came into force, as this law forced companies to have at least one women represented in their

8 BODs, thus providing the opportunity to explore how “non-enforced” board composition affects innovation performance.

The remainder of the thesis will be structured as follows. First, the theoretical framework be

laid out, where existing literature on innovation, board composition and its influence on innovation,

will be introduced. Followed by developing of hypotheses, where the focus will be held on gender

diversity and director independence as one of the mostly discussed topics of contemporary

corporate governance research. Following by the methodology section, where used dataset will be

introduced together with the chosen method of investigation. Later, the results and the

corresponding discussion will be given. The thesis will conclude with a final section containing

discussion and contribution to the existing body of literature as well as presentation of limitations

faced in conducted research.

9 LITERATURE REVIEW

Innovation

Despite being in the center of attention from researchers and practitioners there is still no consensus about the definition of innovation. Most of the definitions tend to emphasize various aspects of this term. Schumpeter was the first who tried to define this innovation, stresses mainly the novelty aspect, as the innovation was reflected in novelty of outputs (Schumpeter, 1934). Others defined it either as creation and later implementation of new solutions (Damanpour, 1996), successful commercial exploitation of newly generated ideas (UNECE, 2009) or as a new product generation process resulting from the application of new knowledge (Cho & Pucik, 2005). It is worth noting, that innovation is not related solely to outcome but also to the process itself (Crossan & Apaydin, 2010).

Innovation has been for a long time perceived as one of the main determinants of firms’ future prosperity and their survival on highly competitive market (Cefis & Marsili, 2006). Some researchers conclude that firms’ ability to relentlessly innovate through development and introduction of new products is fundamental for company’s’ success (Krugman, 1979). The importance of innovation was emphasized also by the United Nations, as it was stated that innovation is one of the crucial sources of competitive advantage of modern knowledge-based economies, that additionally can contribute to the improvement of living standards and help to mitigate prevailing social and environmental issues (UNECE, 2009). Therefore, the innovation is at utmost interest especially in context of EMs in their attempts to catch-up with developed countries.

EM countries were for a long time perceived solely as a source of natural resources and cheap labor force. However, as EM develop economically, ability of their firms to innovate has become crucial determinant of their competitiveness. Simultaneously with this trend, countries such as China and India no longer import innovations from developed countries but are increasingly capable of exporting innovations even to developed countries, so called “reverse innovation”

(Govindarajan & Ramamurti, 2011). Innovations introduced in EM context often do not involve

technological breakthroughs, that are drivers of innovations in developed markets. Their character

is rather modest, often defined as increamental, based on innovative ways of combining already

existing knowledge and technology (Govindarajan & Ramamurti, 2011).

10 Related to innovations there are additional challenges, due to institutional voids, that firms from EM countries are forced to deal with, such as non-functioning financial markets, as external financing represents important resource to facilitate expensive innovation projects. However, in developing countries are financial markets often underdeveloped and suffer from instability.

Research emphasizes the role of BGs as a response to market imperfections (Khanna & Palepu, 1997). BGs, as a distinctive organizational form, are defined as legally independent firms, “bound together by formal and informal ties” (Khanna & Rivkin 2001), usually with high share of family ownership and coordinated by a central entity (Leff, 1978). Related to previously mentioned issues, BGs are capable of overcoming these burdens by using internal source of financing to compensate for lack of external financial resources, allowing BGs to finance also bigger innovation projects (Ayyagari et al., 2012).

Corporate Governance in India

Corporate boards serve as entities connecting shareholders and managers with the main responsibility towards company shareholders (Cadbury, 1999). The main role of boards in contemporary organizations is usually divided into three main categories: control, strategy and service (Pearce & Zahra, 1992). Nowadays, there are two prevailing forms of board structure.

A one–tier system, with a single board, that holds both executing and non-executive directors together in one unified board. However still allows to differentiate between both groups of directors and their main responsibilities. Another distinctive feature is CEO, as a chairman is in parallelly the CEO of company (Anand, 2007). This governance structure is mostly spread over Anglo-saxon countries. Second board structure is a two-tier board structure, very common in countries of continental Europe, consisting of an executive and supervisory board that are separated from each other (Jungmann, 2006) while prohibiting the parallel membership in both boards (Solomon, 2007).

Similar distinction is also related to CEO, where this position is separated from a board chairman.

India, as a country of focal interest, currently mandates the one-tier board system. The first corporate law entered force in form of Company Act 1956. Since its implementation it has been multiple times amended but the core of this act remained the backbone of Indian corporate law up to its revision by Company Act 2013. Despite its relatively long corporate history, the term

“corporate governance” remained relatively unknown until 1993, where after multiple corporate

scandals and series of bankruptcies, became an increasingly discussed issue during the economic

11 liberalization. Despite seemingly operating with same governance structure as other Anglo-saxon countries, there are significant difference between them. As mentioned in previous section India, as well as many other EM countries, contains many companies with high family ownership, compared to the majority of widely held companies in USA or UK (Mehrotra, 2015). Significant differences were found especially with respect to below examined variables of gender diversity and director independence, where Indian boards were found to be more homogeneous (Mehrotra, 2015). Women representation remains very low and due to founder family members occupying board position, also the share of independent directors is lower compared to developed countries.

This drawback of the Indian governance system has been recognized and therefore legal requirements to increase the share of independent director, to reduce the exploitation of minority shareholders and to enhance gender diversity by enforcing the presence of at least one woman on BOD were introduced in the Company Act 2013 (Goswami, 2002).

Board composition

The composition of the BODs has been receiving increasing attention and remains one of the most important issues currently discussed in corporate governance (Kang et al., 2007). According to Kang et al. (2007), diversity is defined as the variety in composition of BODs and this variety can be categorized into two main categories. First, the observables ones: gender, age or ethnicity and second the less visible ones, such as the educational, professional or functional background of board members.

In line with the resource-dependence perspective (Boyd, 1990) studies provided evidence that board members provide firms with valuable resources embedded in their human capital and relational capital (Korn/Ferry, 1999). Those resources (knowledge, expertise etc.) represent sources of competitive advantage helping organizations to achieve superior performance. For example, research conducted by Bantel & Jackson (1989), found that educational and functional diversity of executive teams positively affect innovation and creativity due to more diverse human capital. Moreover, it was argued that heterogeneous teams should be able to produce a broader range of solutions because of the diverse knowledge they contain (Miliken & Vollrath, 1991).

Within organizational boards this often occurs in form of divergent perspectives the directors

provide as well as higher quality of decisions (Hoffman, 1959; Amason, 1996). In addition to a

more diverse human capital, heterogeneous boards also come with more diverse social networks,

12 further providing firms with crucial information that might be critical to the firms’ success (Granovetter, 1973). Board composition has also been related to innovation as one of the crucial factors of organizations for gaining and maintaining competitive advantages (Hitt et al., 1996), improving firms’ performance (Morbey, 1988), and increasing market share (Franko, 1989). On the other hand, observed downsides of higher board diversity are a possibly conflicting climate, an increase regarding misunderstandings among the board members and an increase in coordinating costs (Huse, 2007). These characteristics result in decreasing the ability of boards to effectively react to new challenges and eventually lead to lower organizational performance.

Following chapters are dedicated to influence of gender diversity and director independence on several organizational outcomes. To build up the framework used as a cornerstone for expected on innovation performance.

Gender diversity

Gender diversity is nowadays arguably one of the most discussed characteristic of board composition in corporate governance (Kang, 2007; Huse, 2007). Studies examining the presence of women within Fortunes 500 companies between the years 2009-2013, found out that women, on average, occupied only one out of seven executive positions (Catalyst, 2014).

The low representation of women led to the introduction of several quota systems in legislations of many countries, aiming to increase the share of women on the highest levels of organizational hierarchies and to make women equally represented, compared to men. The first gender quota came into force in Norway in 2005, aiming to increase female representation up to 40% in corporate boards, quickly followed by similar initiatives spanning all over Europe. A similar enforcement mechanism was signed also in India in form of the Company Act 2013, forcing companies to have at least one female representative present in their BODs.

First studies examining relationship between gender diversity in corporate boards on

organizational performance concluded with various results. Depending on the country and industry

specific context the results varied from positive (Hillman, 2002; Hillman 2014, Bantel & Jackson,

1989; Ross & Deszo, 2012) all over to negative relationship between higher levels of gender

diversity and various organizational financial performance (Laible, 2013; Du Rietz & Henrekson,

2000). Extended research started to link higher gender diversity on corporate boards with various

organizational outcomes, such as better CSR performance (Bear et al., 2010), enhanced creativity

(Bantel & Jackson) and problem-solving capabilities (Murray, 1989, Carter et al, 2003). On the

13 other hand, research also found downsides related to diverse corporate boards. Namely, distrust, lack of cohesion, higher coordination costs or higher conflict occurrence (Huse, 2007), resulting in the boards’ inability to operatively and appropriately react to changing and demanding market environments due to slower decision-making.

According to Krishnan & Part (2005), gender represents a highly complex variable as it also influences the individuals’ cognitive base, as significant differences in the way how women and men process information prevail. Related to firms’ innovation performance, research conducted by Miller & Triana (2009) showed that broader range of perspectives helps companies to identify innovative ideas and opportunities and also new working styles into the boards (Hillman, 2002).

Men generally tend to be more impatient and less risk averse compared to women (Adams &

Ferreira, 2008; Sapienza, Zingales, & Maestripieri, 2009; Frederick, 2005). This behavioural trait of men is amplified in situations when financial incentives are involved (Schubert, Brown, Gysler,

& Brachinger, 1999) resulting in an over-confident evaluation of decisions (Huang & Kisgen, 2013). Moreover, it was argued, that women do not act that impulsively as men and tend to spend more time considering their decisions (Hillman, 2014). On the other hand, findings related to risk- averse behavior of women (Adams & Ferreira, 2008; Sapienza, Zingales, & Maestripieri, 2009) suggests, that women tend to mitigate risks by avoiding projects with unpredictable outcomes. This provides contradictory evidence to studies of Bantel & Jackson (1989), as it implies that higher gender diversity might result in lower attempts of organizations to innovate due to higher propensity of risk-avoidance related to uncertain results of most of the innovative project developments.

Deeper investigation found out that, main the factor influencing this relationship is not solely the presence of women itself but more importantly the share of women appointed as directors.

Research conducted by Torchia (2011) and Kramer (2008) provided evidence, that very low level of gender diversity limits women to tokens, with very little influence on board dynamics.

Significant effect appears once the consistent gender minority is reached, usually consisting of at least three women on corporate boards (Torchia, 2011).

However, in the Indian context there is prevailing inequality between genders especially in

higher management positions. While women represent 36% of the labor force, their share of the

senior management and director position is only around 5%. (Catalyst, 2012). Therefore, the first

hypothesis argues in line with critical mass theory, denying existence of strong relationship

14 between gender diversity and innovation performance. Being based on the assumption of very low female representation on boards, that in most of the boards is not sufficient to constitute critical mass.

Hypothesis 1: Higher gender diversity will have no significant effect on innovation performance.

According to Hamilton & Kao (1990), especially in the EM context with high representation of family-owned enterprises, there is great decision-making power concentrated within board’ as the “inner circle”. Additional research conducted by Chung (2003) focusing on Taiwanese firms discovered, that most of the members of inner circle are either founders’ family members or individuals with prior social ties (former classmates, friends etc.). Women appointed to BG boards usually have ties to founders’ family, often they are either wives, daughters of company founder etc. Despite their low share, compared to women directors in standalone firms, they might have bigger power of exerting influence over board decision-making because of their position within

“inner-circle” position. Therefore, there is no need to achieve consistent minority to exert influence.

Following, the hypothesis focusing on moderating effect of BGs is therefore stated accordingly:

Hypothesis 2: Among business groups higher gender diversity will lead to enhanced innovation performance.

Director independence

Next to gender diversity, an additional variable examined in this thesis is board independence and

its influence on innovation performance. The idea that the BODs of companies should be

dominated by outside directors independent from management is not new in management literature

(Chandler, 1975). The notion of directors’ independence has reached significant importance,

especially after major corporate scandals such as Enron or WorldCom in the early 2000s. Those

scandals emphasized flaws in corporate governance mechanisms that are related to a lack of board

oversight due to close relationships between executive and non-executive directors. Despite

increasing pressures towards increased appointment of independent directors there is still no mutual

consensus related to the definition of the term itself. Many companies, even nowadays, relatively

briefly identify their directors just as “independent” and “non-independent” without disclosing

15 additional information to external stakeholders related to their appointment or responsibilities.

Moreover, as found out by Mahadeo (2012), term “independent director” has especially in countries with recently introduced codes of governance very ambiguous meaning, often leading to labelling directors as independent while being closely related to the company.

Murray (1989) in the research examining inner board dynamics, argues that homogeneous boards act more as a “clans” producing higher conformity among their members. This, subsequently, results in the inability of boards to fulfill their controlling duties and the tendency to act as a “rubber-stamping authority” to decisions made by executives. Moreover, in research of Kang et al (2007) was pointed out that directors are usually selected from “old-boys network”, positioning appointed directors into indebted position to those responsible for their selection.

Majority of implemented governance codes aims to increase the share of independent directors within boards to prevent big governance scandals from repeating. However, as it was argued by Murray (1989), especially from a short-term view on organizational performance, such enforced appointment of directors can harm organizations due to a higher likelihood of misunderstandings, thereby creating conflicts within boards (Murray, 1989). Additional evidence brought by Laing and Weir (1999), concluded that attempts to increase the share of independent directors purely to meet criteria required by quotas creates inertia problems, which further slowdown decision-making processes and lower the boards’ overall efficiency. However, when their share is increasing in unenforced manner, higher share of independent directors is often associated with successful mitigation of occurring agency problems (Mahoney, 1992; Eisenhardt, 1989). A firm innovation strategy can be, due to high levels of risks involved in sunk-cost investments, the subject of agency problems. Especially, because directors might hesitate to make investments into innovation activities that will not pay off in short term (Baysinger et al., 1991: Eisenhardt, 1989). However, in pharmaceutical industry can this reluctant approach lead to lowering of firm’s competitiveness, due to inability to keep pace with more innovative companies, as R&D investments are usually necessary for new product development (Mosakowski, 1993).

The findings of academic studies investigating the impact of independent directors on

innovation performance, usually measured in R&D investments. Zahra (1996) found negative

relationship while Hoskinson et al. (2002) failed to find evidence of any relationship between those

two variables. Lastly more recent study conducted by Kor (2006) provides evidence of positive

relationship between board indepence on R&D investments, that is amplified in cases of separated

16 roles of CEO and Chairman (Kor, 2006). Despite the inconsistent findings and lack of empirical evidence third hypothesis is stated in favor of positive relationship of director independence on innovation. Although the dependent variable in this study is more focused on final outcome compared to previously mentioned studies.

Hypothesis 3: Higher ratio of independent directors will lead to better innovation performance.

Moreover, a study conducted by Chen & Hsu (2009) examined this relationship more into detail by using the sample of Taiwanese companies operating on high-tech industry, focusing solely on family-owned enterprises. This research concluded, with finding negative relationship between family ownership and R&D investments. Suggesting that this kind of ownership tends to discourage risky R&D investments. As BGs in India share a similar trait of high share of family owned enterprises and high family presence on BODs, it will be argued similarly to results provided by Chen (2009). However, as authors acknowledge, these findings can be also interpreted as a more efficient way of using R&D investments, suggesting that impact on newly introduced products could be also opposite. However, also in the case of family-owned enterprises are independent directors brought to firms mainly for their expertise and guidance (Solomon, 2007; Osma, 2008).

Therefore, also in this case will be argued in favor of positive relationship between number of independent directors and innovation performance.

Hypothesis 4: Among business groups higher ratio of independent directors will lead to enhanced innovation performance.

Conceptual model

Board Composition:

Gender diversity Director independence

Innovation Performance Newly introduced products

Organizational form

BG x Standalones

17 RESEARCH METHODOLOGY

Research context: Indian pharmaceutical industry

Context of Indian pharmaceutical industry provides optimal setting to test hypotheses developed based on the assumption of highly competitive market with presence of BGs. Ever since nineteenth century India has strong presence of family-owned enterprises (Pattnaik et al. 2013) and even after massive liberalization in 1991 they continued to play a major role in the local economy.

A pharmaceutical industry in India represents an example of a highly competitive market with strong reliance on innovation. In the previous years, the pharmaceutical market has experienced dramatic growth. The market value increased from 6Bn. $ in 2005 to 16 Bn.$ in 2014 and by the end of 2019 it is estimated that market value should reach 33 Bn. $ (Marketline, 2015) making the Indian pharmaceutical industry one of the fastest-growing industries among emerging countries and key player on a global scale. Today India is one of the largest producers of generic drugs, globally ranked as 14

largest pharmaceutical market measured in value and 3

in volume (McKinsey & Company) responsible for over 20 % of global production.

Data source and sample operationalization

Data required for the analysis were collected via two databases, Orbis and Prowess. The Orbis database was used to retrieve financial data, as it compiles data of most of the companies operating worldwide. The Prowess database contains relevant information about the Indian economy and firms operating in Indian. As additional source of data, the announcements on Bombay Stock Exchange were used to collect information about new newly introduced products. These announcements were used as a proxy for quantifying the innovation performance. Company annual reports were used as a supporting source of information, to verify provided information and possibly add any missing data, if available.

To further proceed with the analysis was first it was necessary to adjust the initial sample. The

first step to create a consistent dataset was to find and delete companies that did not appear over

the entire time horizon. This approach was selected as more parsimonious to the dataset than adding

average values to missing companies. Moreover, initial attempt to use R&D intensity as a control

variable failed because of large amount of missing data related to firms’ R&D expenditures.

18 Lastly, the check for outlying observations was performed. Three companies were identified with dramatically higher number of innovated products over examined period compared to other companies within the sample. Therefore, these outliers were excluded from the sample to not negatively influence the outcome of further regressions. Deeper examination discovered that one company in form of BGs was acquired during examined period by foreign multinational company and therefore excluded from sample. Additionally, there were found several missing observations of variables sales. Here on the other hand, due to very small amount of missing values was selected approach of adding missing values instead of deleting companies with missing observations. The value added were calculated as a mean of firms’ sales during the time horizon of 2006 – 2011.

Year 2006 was chosen due to the implementation of the Trade Related Intelectual Property Rights (TRIPS) that in India came into force a year before. This agreement disallowed reverse engineering and selling knockoff drugs (Agrawal and Saibaba, 2001), therefore it is considered to have strong influence on innovation performance.

Year 2011 was chosen due to the upcoming implementation of the Company Act 2013 that came into force on September 2013. This act covers also the area of corporate governance forcing Indian companies among others, to have at least one female representative in their boards (Ernst &

Young, 2014). Selecting year 2011 as the end of the examined time horizon provides this thesis with reliable data about the voluntary diversity in BODs without potential influence of forthcoming law enforcement.

The final sample prepared for further analysis consists of 119 companies over 6 years, resulting in total of 714 observations, out of which were 86 standalones and 33 BGs.

Measures

Dependent Variable

Innovation performance: Clearly defining this variable is complicated due to its intangible nature.

Some studies tried to capture innovation performance by using the number of granted patents or filed patent applications as a proxy (Rodriguez-Pose & Di Cataldo, 2014; J. Li et al., 2013; Cheung

& Lin, 2004;). There are also some downsides for using this proxy to capture innovation

performance. By focusing solely on number of patents it will not be possible to capture incremental

innovations that do not undergo patenting process (Rodriguez-Pose & Di Cataldo, 2014). Due to

focusing on EM countries in this thesis it would not be possible to capture large number of

19 improved products, as most innovations introduced in EMs do not have breakthrough character.

Therefore, by using this proxy there is a high possibility of undervaluing the innovation performance.

Innovation performance in this research is measured by new product announcements made in examined time horizon and will be focused on products introduced on Indian market. Similarly, research conducted by Liu & Buck (2007), where was this way of measuring innovation performance preferred over patents more suitable, as it captures also improved existing products, that otherwise won’t be captured. Data for this variable were gathered from new product announcements made on Bombay Stock Exchange.

Independent variables – Board diversity variables

Gender diversity: The level of gender diversity within BODs is quantified through the proportion of men and women appointed as directors. For objective measurement was used the Blau diversity index (Harrison, 2006), calculated by following formula:

𝐵𝑙𝑎𝑢 𝑖𝑛𝑑𝑒𝑥 = 1 − ∑

𝑝

Where p is the share of members is each category (in this case share of women) and n stands for total number of board member. The values of Blau index can range from 0, for completely homogeneous samples, to maximum of 0,5 occurring when the board consist of equal number of men and women. Due to low number of appointed female director within examined sample the Blau index was preferred over simply working with share of women.

Independent directors: Similarly, also in this case was considered total number of appointed independent directors afterwards was calculated the share of independent directors. However, in this case the interest was focused on ration of independent directors. Therefore, the Blau index was not utilized as an additional measure.

Control Variables

Created models were also controlled for several firm level variables. First control variables are

Assets and Sales. Both variables are measure in millions and are used as a proxy for measuring

firm size. Based on the research conducted by Cohen (1995) size of the firm is often positively

related to innovation, as some innovation require investments only bigger firms can afford. Yet,

there is also the possibility of reverse relationship, as it might be argued, that innovation facilitate

growth (Bantel & Jackson., 1989).

20 An additional variable is the Board size, measured as a total number of appointed directors. It is expected that the bigger the board is, the higher is the likelihood of diversity among board members can occur. Therefore, this variable often used to control for any board diversity measure (Goodstein, et al., 1994).

Moreover, as argued by previous studies also the firms’ innovation performance might be influenced by firms’ financial situation. To control for this Debt/Equity ratio measured by the ratio of total debts to total sales will be applied.

Lastly, the variable firms age will be used to control for the maturity of the companies, measured in number of years since the date of establishment. Older firms might be less flexible in their adaption to competitors and market environment and therefore also less likely to innovate (Capon, Farley & Hoenig, 1990).

Descriptive statistics

Table below provides brief summarized statistics of used variables. Overall female representation measured as a share of women was low in the examined sample with the maximum value being 33% of all directors in case of only one company with vast majority of companies having no women at all appointed as directors. Overall share of women in BODs was just 3.78%, which is very low compared to other studies (Hunt, 2015.; Catalyst, 2014), and confirmed expectations stated in literature review, suggesting that majority of firms having no women present on their BODs.

Following with the ratio of independent directors, that shows higher volatility, as some boards consisted solely of independent directors while there were also some examples of boards with no independent director. Overall share of independent directors was on average slightly over 62%.

This number is relatively consistent with for example study conducted by Kang et al. (2007) where in the sample of Australian companies was also majority of directors stated as independent.

Domestic innovation as a main dependent variable of this study, measure by number of newly

introduced products on domestic market also suffers from excess occurrence of zeros within the

sample. Maximum value of innovated products is 15 while the average is only 0,36.

21

Variable Obs Mean Std. Dev. Min Max

Innovation Performance 714 .359944 1.557903 0 15

Board gender diversity - Proportion 714 .0378105 .0669493 0 .3333333 Board gender diversity - Blau 714 .0638099 .1091497 0 .4444444

Independent director ratio 714 .6201215 .1531804 0 1

Independent director ratio - Blau 714 .4242789 .0966375 0 .5

Debt/Equity 714 .9444062 1.419789 0 17.5784

Board size 714 8.477591 2.876646 2 18

Firm age 714 26.36975 16.73037 8 110

Assets (mil) 714 131.5539 317.0282 .1093 3787.555

Sales (mil) 714 79.56209 159.5841 .005 1422.639

Table 1: DESCRIPTIVE STATISTICS OF USED VARIABLES

Pearson correlations and multicollinearity test

Before starting the with analyses, Pearson correlations and variance inflation factor (VIF) tests were conducted to check for presence of multicollinearity among the examined variables. Pearson correlations provided initial evidence for little to no presence of multicollinearity, as it shows relatively low correlation coefficients between the variables.

Nevertheless, due to working with panel data, it was not possible variable values in their

original form to conduct variance inflation factor analysis for multicollinearity. To get reliable

measure, it was necessary to use centered values of these variables, that were obtained by

subtracting the average value from each variable. Finally, after adjusting the data VIF test was

conducted showing very no evidence for presence of multicollinearity among the variables, as no

coefficient exceeds the value of 10. Therefore, the regressions could be performed without the need

to drop any variable (VIF provided in Appendix 2).

22 Method of analysis

Fixed-effects model

The analysis was conducted by the statistical software STATA, that was preferred over SPSS, as it allows to work with panel data samples. Firstly, it was necessary to set the dataset as panel data.

Therefore, the company unit of analysis was set (1-119) together with the year as a time variable (1-6). This step confirmed the balance of the examined sample and allowed to proceed with analysis.

In the next step, it was necessary to select the most appropriate regression model, either fixed- effects or random-effects model. Therefore, the Hausman test was conducted as its results will help to identify the right model for regression analysis (Baltagi, 2008). The results of the Hausman test rejected the random effect models. Therefore, the fixed-effects model was selected as preferred panel data model. The result of first the Hausman test of the first model is attached in the Appendix 4, following models achieved similar and therefore were omitted).

The effect of gender diversity on organizational innovation performance was estimated by conducting the following fixed effects models:

(1) Innovation performance

= β

+ β

board gender diversity -blau

+ β

Z

+ 𝑣

+ 𝑒

(2) Innovation performance

= β

+ β

board gender diversity -blau

+ β

board gender diversity × BG dummy

+ β

Z

+ 𝑣

+ 𝑒

(3) Innovation performance

= β

+ β

Independent director ratio

+ β

Z

+ 𝑣

+ 𝑒

(4) Innovation performance

= β

+ β

Independent director ratio

+ β

Independent director ratio × BG dummy

+ β

Z

+ 𝑣

+ 𝑒

Where Z

,

represent the set of control variables, v

,

stands for time invariant effects and 𝑒

represents the residuals.

To investigate how the organizational form influences the relationship between board

composition and innovation performance, a similar approach to study conducted by (Chen, 2009)

was applied. The influence of gender diversity and director independence ratio was examined

firstly independently later also in interaction with the BG dummy variable. Moreover, the

23 dependent variable was defined as a lead variable. The innovation performance from years 2007- 2011 was regressed against control and independent variables from years 2006-2010 to ensure that the direction of causality was in the desired way. Additional evidence from management literature is suggesting that board composition characteristics influence organizational outcomes, rather than the other way around (Miller & Triana, 2009), thus the likelihood of reverse causality is minimal.

However, due to a relatively low number of observations it was not possible to automatically accept the assumption of normal distribution of examined dependent variable. The frequency plot raised the suspicion about non-normal distribution of dependent variable because of a large spike of zero values. The further conducted Shapiro-Wilk test only confirmed this suspicion of non- normal distribution of innovation performance (frequency histogram and Shapiro-Wilk test are attached in the Appendix 3). To compensate for non-normal distribution of the used dependent variable, a count data model was utilized (Hausman et al,. 1995), as a regression method capable of processing dataset with such distribution.

Count data model

Due to the nature of the dependent variable having non-normal distribution, as the majority of observations reaches zero it was therefore necessary to select an appropriate count data model. A zero-inflated model is two-step model assuming that data is a mixture of two separate processes generating zero values, one generating solely zeroes and the other also other values (Zuur et al., 2009). To better understand this process with relation to examined sample of companies, it basically assumes that within examined dataset, there are companies that intentionally did not invest at all in their R&D activities and this lack of effort is therefore one of the reasons behind excess of zero values, next to the unsuccessful product development.

However, the assumption of such passive firm behavior is not realistic, especially in case of highly

competitive markets, where it clashes with the companies’ main goal of maximizing profit

(Demsetz & Lehn, 1985) and sustaining their competitive advantage (Barney, 1991) over its

competitors by relentlessly innovating their products (Kock et al., 2015). This was the main reason

for logically rejecting the zero-inflated model Another method is zero-truncated model. This

method however basically ignores zero values and operates only with non-zero observations. Using

this modelling method would dramatically reduce the size of examined sample, therefore

representing the main reason for abandoning this method. Leaving only Poison and negative

binomial regression models as final candidates.

24 To decide between those two, it was necessary to test the mean and variance of dependent variables, as equality of those measures is the basic assumption for Poisson regression model. The test was done by firstly conducting negative binomial regression model as an example of less restricted method compared to Poisson modelling method and later conduct likelihood-ratio test of alpha. In this case, the test was statistically significant and therefore, the negative binomial model was selected as the most appropriate count data modelling method (results are provided in Appendix 5).

However, attempt to deal with non-normally distributed data leads to another limitation related inability of count data models to operate with panel data samples. This limitation was partially mitigated by controlling for company and year of observation. For presented regression outcome were these control variables dropped from the table.

(5) Innovation performance

= β

+ β

board gender diversity - blau

+ β

Z

+ year dummy

+ company dummy

+ 𝑒

(6) Innovation performance

= β

+ β

board gender diversity - blau

+ β

board gender diversity - blau × BG dummy

+ β

Z

+ year dummy

+ company dummy

+ 𝑒

(7) Innovation performance

= β

+ β

Independent director ratio

+ + β

Z

+ year dummy

+ company dummy

+ 𝑒

(8) Innovation performance

= β

+ β

Independent director ratio

+ β

Independent director ratio × BG dummy

+ β

Z

+ year dummy

+ company dummy

+ 𝑒

Similarly to the previous models, Z

,

and 𝑒

Also in

this case was the dependent variable shifted by one year, so that all independent and control

variables are one year lagged. Minor difference needed to be done, to compensate for the inability

of count data models to work with panel data. Therefore, it was necessary to add time (year dummy)

and company (company dummy) control variables to mitigate such a drawback.

25 RESULTS

Fixed-effects model

As mentioned in the previous sections, the final regression was conducted by using fixed-effects model for panel data. By setting innovation performance as a lead variable the final samples was reduced to 595 observations. After conducting initial regressions all models were tested for the presence of heteroscedasticity by using the modified Wald test for group wise heteroscedasticity.

In all of them null hypothesis was rejected, providing evidence for the presence of heteroscedasticity. Therefore, all models presented were and estimated by using robust standard errors to deal with the presence of heteroscedasticity (Wooldridge, 2002).

As stated earlier, this study focuses on the main areas. First, how does the board composition with respect to gender diversity and director independence influence company’s innovation performance (models 1&3). Second, how does the organizational form of BG affect this relationship (models 2&4). Therefore, was the regression analysis was conducted initially by using solely a board composition measure and later by adding the interaction term between measure and BG dummy, to investigate the moderating effect of BGs.

Model (1) provided results of relationship between gender diversity, measured by Blau index, and innovation performance. Regression analysis suggest statistically significant (at 5% level) negative relationship between diversity and number of newly introduced products and therefore rejecting first hypothesis. Debt/Equity ratio and Board size appear to have negative but statistically insignificant impact, Assets on the other hand show positive coefficient, however also in this case it is not statistically significant. However, last two controls variables, Firm age and Sales contribute positively and even significantly to innovation performance.

Model (3) provides results related to influence of director independence measured by ratio of

independent directors. Also in this case fixed-effect model showed statistically significant

relationship between both variables, however this relationship was found to be negative rejecting

third hypothesis, that argued in favor of positive relationship of independent directors on innovation

performance. In this model, Debt/Equity ratio shows negative impact, variables Board size, Assets

have positive effect, nevertheless the effect of all these variables seems to be statistically

insignificant. Similarly to model (1), just the variables Firm age and Sales were statistically

significant and again contributing positively to the dependent variable.

26

(1) (2) (3) (4)

VARIABLES Innovation

performance

Innovation performance

Innovation performance

Innovation performance Board gender diversity - Blau -0.488** -0.606*

(0.192) (0.370) Board gender diversity – Blau

× BG dummy

1.087*

(0.656)

Independent director ratio -0.215** -0.180

(0.0962) (0.116) Independent director ratio ×

BG dummy

-0.302 (0.495)

Debt/Equity ratio -0.00915 -0.00961 -0.00965 -0.0101

(0.0137) (0.00935) (0.0138) (0.00876)

Board size -0.000283 -0.000353 0.00107 0.000616

(0.00651) (0.00623) (0.00651) (0.00697)

Firm age 0.0316*** 0.0317*** 0.0316*** 0.0316***

(0.00543) (0.00128) (0.00544) (0.00131)

Assets 1.00e-04 0.000100 9.87e-05 9.92e-05

(6.62e-05) (6.84e-05) (6.63e-05) (6.79e-05)

Sales 0.000733*** 0.000733 0.000708*** 0.000686

(0.000243) (0.000445) (0.000243) (0.000443)

Constant -0.439*** -0.442*** -0.349** -0.311**

(0.142) (0.0467) (0.153) (0.123)

Observations 595 595 595 595

Within R-squared 0.150 0.156 0.148 0.149

Number of companies 119 119 119 119

Company FE Yes Yes Yes Yes

Year FE Yes Yes Yes Yes

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1 Table 2: RESULTS OF FIXED-EFFECT MODELS

Related to the interaction effect models model (2) investigates the moderating effect of BG-

affiliation on the relationship between gender diversity and innovation performance. Coefficient of

the interaction variable indicates positive impact of gender diversity on dependent variable,

27 supporting second hypothesis. However, this impact is only marginally significant (at 10% level).

The impact of the rest of control variables is the same, as for model one with minor changes in their significance. In this case however, only variable firm age is statistically significant.

Model (4) as the last fixed-effect model adds again interaction variable to model (2). The coefficient of interaction term is negative and statistic insignificant rejecting last hypothesis. The coefficients of the rest of control variables indicate very similar directions to model (2), again with minor difference in significance level. Similarly to model (3), also in this case the only significant variable is Firm age, the rest is statistically insignificant.

R-squared coefficient is ranging from 0,148 to 0,156, providing information about moderate explanatory of these models. Adding interaction term increased explanatory power of both models, from 0,150 to 0,156 respectively from 0,148 to 0,149. On average these models manage explain 15% of total variability.

Negative binomial model

Following table shows results provided by using negative binomial regression model that was used

to compensate non-normal distribution of depended variable due to excess of zeroes in

observations. Analysis was again conducted by four separate models using each independent

variable firstly alone (models 1 & 3) and later again in interaction with organizational form (models

2 & 4) to examine also moderated effect of organizational form. Additionally, it was necessary to

compensate for inability of negative binomial model to deal with panel data. Therefore, additional

control variables were created to control for year and company unobserved effects but they were

excluded from final table.

28

(5) (6) (7) (8)

VARIABLES Innovation

performance

Innovation performance

Innovation performance

Innovation performance Board gender diversity - Blau 5.804*** 4.124***

(1.626) (1.592) Board gender diversity – Blau

× BG dummy

7.814**

(3.451)

Independent director ratio -0.937 -1.334

(1.359) (1.364) Independent director ratio ×

BG dummy

0.717 (0.650)

Debt/Equity ratio -0.309 -0.207 -0.371 -0.323

(0.233) (0.214) (0.250) (0.242)

Board size 0.186*** 0.147** 0.207*** 0.180**

(0.0709) (0.0686) (0.0767) (0.0799)

Firm age 0.00213 -0.000334 -0.000842 -0.00267

(0.00952) (0.00919) (0.0101) (0.00995)

Assets 0.00189 0.00138 0.00189 0.00164

(0.00277) (0.00257) (0.00324) (0.00308)

Sales -0.000629 0.000773 -0.000717 -0.000837

(0.00477) (0.00457) (0.00551) (0.00526)

Constant -3.442*** -3.171*** -2.265* -1.900

(0.797) (0.763) (1.186) (1.207)

/lnalpha 2.400*** 2.319*** 2.544*** 2.522***

(0.171) (0.175) (0.167) (0.169)

alpha 11.025 10.168 12.729 12.448

(1.889) (1.780) (2.123) (2.100)

Observations 595 595 595 595

Company dummy Included Included Included Included

Year dummy Included Included Included Included

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1 Table 3: RESULTS OF NEGAGITE BINOMIAL MODELS

Using negative binomial models as an alternative estimation method provided slightly different

results to initially used fixed-effect models. Results obtained by model (5) suggest positive and

29 statistically significant impact of gender diversity on innovation performance, not supporting to first hypothesis. However, provided result for main relationship is opposite to those model (1). The impact of Debt/Equity ratio and Sales is negative and statistically insignificant. Variable Board size is showing positive and significant impact on dependent variable. The rest of control variables suggest positive impact, however the impact is statistically insignificant. Examination of model (7) focusing on relationship between director independence and innovation performance showed same results as model (3), also in this case were negative and statistically insignificant and therefore rejecting second hypothesis.

Interaction model (6) provides evidence of positive and statistically significant effect of gender diversity influence on innovation performance among BGs, supporting second hypothesis. Related to control variables their direction and significance is very similar to model (5). With minor difference at Firm age, where the relationship is negative in this model. Final interaction model (8) shows positive however statistically insignificant impact of director independence on innovation performance. In this case is the impact of the rest of control variables similar to model (7) only with small change in statistical significance of variable Firm age, being the only statistically significant variable of the model.

However negative binomial regression does not have an equivalent measure to the R-squared usually found in other OLS regressions. Software provides only results of Pseudo R-squared, nevertheless their interpretation is not consistent with R-squared measure. Therefore, they were excluded from the table.

Robustness checks

In this section, main regressions will be additionally modified and thus will help to find any

misspecification and enhance the validity of drawn conclusions (Lu & White, 2014). Several

robustness checks were conducted to verify the validity of gained results. All robustness checks

were simultaneously conducted by using both regression methods. First set of robustness checks

(Appendix 6) is related to fixed-effect model, the second (Appendix 7) is linked to negative

binomial model. Results of conducted robustness checks confirm the direction of previously

provided results however changes in levels of statistical significance of gender diversity variable

suggest weak link to innovation performance.

30 First set robustness check is related to slightly different set of used control variables. In main regression were variables sales and assets, used as a proxy for measuring firm size, used in their numerical value in millions. However, some researchers prefer to use such variables in logarithmic form, as it allows to cope with their large dispersion and skewedness (McLeod & Malhotra).

However, by introducing logarithmic values into the model, the significance level of several variables changed signaling relatively weak relationship between the independent and the dependent variables.

To further explore the possibility of nonlinear relationship between examined independent variables and innovation performance the squared values of female proportion and independent directors were implemented into the model next to the originally used variables. However, the variables related to gender diversity were replaced by women ratio. Results provide weak evidence suggesting the possibility of U-shape relationship between the variables in case of fixed-effect models with BG dummy interaction (Appendix 6; models 6 & 8) and the possibility of inverse U- shape relationship otherwise (Appendix 6 & 7; models 6 & 8). However, this non-linear relationship was found to be statistically insignificant.

Furthermore, as it was often argued, that corporate boards dominated by independent directors perform significantly better than those where there is only minority of independent directors (Kang, 2007). To examine this issue sample was divided into two subsamples with the threshold of 50%

ratio of independent directors, showing however very similar results to initial regressions. The only observed difference was related to negative binomial regressions, where the coefficients became significant. After previous set of models provided suspicion for possible non-linear relationship between independent and dependent variable, additional regression with changing threshold for independent directors were conducted. For limit set to 65% was found out that the impact of independent directors turned out to be positive, however statistically insignificant.

However, the division into two sub-samples was not possible to for models operating with

women proportion, as their low presence within the boards did not allow to divide sample into

smaller sub-samples. As the highest number of women within one corporate board was only two

and the consistent minority was not achieved (Kramer, 2008, Torchia, 2011) it did not allow to

control for the possible significant impact of critical mass on innovation performance within

examined sample

31 DISCUSSION

The aim of this thesis was to investigate the effect of gender diversity and board independence on the innovation performance and the potential moderating impact of organizational form on this relationship. I try to further extend existing body of literature about the influence of board composition on organizational outcomes, by focusing on innovation, as a predictor of organizational performance. This study follows recent trends in the research by examining this relationship in a context of relatively under researched EMs. The paper focuses on India as an example of EM with strong presence of family-owned enterprises. Moreover, this study was narrowed down to the pharmaceutical industry, as it represents a very competitive market, where companies’ competitiveness is heavily dependent on their innovative performance.

Hypothesis 1 draws from previous market research suggesting a very low presence of women among highest-ranking positions. Moreover, Kramer (2008) and Torchia (2011) suggest that number of women first need to achieve a certain minimum amount. Beyond this limit, they can exert an influence within the BOD, thus positively contributing to the innovation performance.

Results related to this hypothesis however differed with respect to the utilized estimation model.

In both fixed-effect model for panel data and negative binomial model, gender diversity significantly affects innovation performance However, the former shows a negative significant impact on innovation, while the latter a positive one. The opposing results of these estimations can be contributed to different characteristics of used regression models.

Hypothesis 2 investigates to what extent organizational form, examined by introducing the BG dummy variable into the previous model, influences the relationship between gender diversity and innovation performance. Results for this hypothesis showed positive coefficient, thus providing support for this hypothesis, regardless of the estimation method used. The assumptions for this hypothesis were derived by combining studies of Sanan (2016), Hamilton & Kao (1990) and Chung (2003), arguing that due to ties with founder’s family are women part of BG’s inner circle allowing them to exert influence despite their low representation.

The results related to the impact of director independence on innovation performance were

more consistent when using both estimation methods. Both hypotheses suggesting positive effect

of director independence on innovation performance, due to broader range of perspectives and

invaluable knowledge embedded in human capital of these directors. However, the results were

found to be statistically insignificant. With an exception fixed-effect model without BG interaction