University of Groningen Faculty of Economics
Master thesis - International Business and Management
TMT heterogeneity and innovation
Student: Ids Pultrum Student ID: 1359169 Month and year: June 2007
Supervisor: Dr. B. Kibriscikli-Ozcandarli Referent: Drs. H. Stek
Abstract:
This paper examines the relationship between TMT heterogeneity and innovation for the manufacturing industry of electrical machinery and apparatus in the E.U. TMT heterogeneity is measured by five independent variables, namely heterogeneity with respect to age, gender, education and nationality, and a variable called overall heterogeneity. Innovation is measured in two ways. It is measured by R&D expenditures divided by sales as well as by the logarithm of the number of patents. Data was received for 84 companies from the 200 largest companies in this industry. Three of the five hypotheses that were formulated are accepted.
Heterogeneity with respect to gender was not analyzed, because of data problems with this variable. The only hypothesis that is rejected in the analysis is hypothesis 3, which expected a positive correlation between heterogeneity with respect to education and innovation.
Keywords: Innovation, TMT heterogeneity, overall heterogeneity, MNC’s, Upper Echelon
perspective, electrical machinery and apparatus, nace(31)
1. INTRODUCTION
Since the 1980’s the innovativeness of U.S. companies has become a topic of national concern (Bantel & Jackson, 1989). Nowadays it is a very important topic at the national and international level within the European Union (E.U.) as well. For example, in the new Dutch coalition agreement there is a big emphasis on innovation. It is stated that Holland aims for a long-term strategy on innovation and entrepreneurship (innovation}platform, 2007). That it is also an important topic at the international level can be seen in the treaty of Lisbon (2000).
There the E.U. has stated that it wants to become the world’s most competitive and innovative region by 2010. The rationale behind this is that innovation leads to higher growth, more jobs and higher wages (Sirelli, 2000).
Another subject that is under debate is the influence of top management teams (TMT’s).
Some argue that organizational leaders are a product of their environments and therefore have little power to control structural and systemic factors that determine organizational choices (e.g. Aldrich, 1979; Perrow, 1970). Others argue that leaders are powerfull decisionmakers who consciously choose among diverse courses of action (e.g. Weiner & Mahoney, 1981).
Central in this paper is the view of Hambrick and Mason (1984). They share the view of Weiner and Mahony (1981) in that they view leaders as powerfull decisionmakers. In their paper Upper Echelons: The organization as a reflection of their top managers Hambrick and Mason argue that organzational outcomes, i.e. strategic choices and performance levels, are partially predicted by managerial characteristics. They argue that group heterogeneity of top managers should result in more innovation and better performance. This is because a more heterogeneous group is best in handling ill-defined and novel problem solving, because the diversity of opinion, knowledge and background allows for a thorough airing of alternatives.
Many researchers have tested the relationship between heterogeneity of top management
teams and firm performance and most of them have found a positive relationship (e.g. Elron,
1997; Mitchel & Hambrick, 1992; Norburn & Birley, 1988), however, so far I have only
found one researcher that has examined the relationship between TMT heterogeneity and the
amount of innovation (Bantel & Jackson, 1989). Bantel and Jackson have tested this
relationship by looking at the correlation between several managerial characteristics and
innovation of U.S. banks, however so far no researcher has developed an overall
heterogeneity score for the company and compared it with the innovativeness of the company.
This paper will try to fill this research gap by examining the influence of overall heterogeneity of TMT’s on innovation. This will be done by looking at the 150 largest MNC’s in the manufacturing industry of electrical machinery and apparatus in the E.U. This focus can be summarized by providing the following research question:
What is the relationship between top management team heterogeneity of European multinational corporations in the manufacturing industry of electrical machinery and apparatus, and their amount of innovation?
This paper will be structured as follows. First there will be a literature review, which starts with explaining the theories on why innovation is important, followed by how it can be measured. Then theories concerning heterogeneity are examined and the literature review is concluded by a conceptual model. After the literature review the concepts and hypotheses will be discussed, followed by the research design. When the research design has been dealt with, the results will be presented and discussed, followed by several conclusions and limitations.
2. LITERATURE REVIEW
INNOVATION
As mentioned in the introduction, innovation has become of national and international concern. Before dealing with why innovation is important and how it can be measured, it is important to know what innovation is and how it is defined. Innovation has been defined differently by several authors. For example, Schumpeter (1934) has defined innovation as an invention that is put onto the market by an entrepreneur, where inventions are defined as the process during which a new technological artefact is constructed. Another definition mentioned by Drucker (1985), is that innovation is the process of equipping in new, improved capabilities or increased utility. In this paper Schumpeter’s definition of innovation will be followed.
At this stage it is important to make a distinction between inventions and innovations. As
stated above, inventions are new products, whereas an innovation is the new value created by
the invention. In order to turn an invention into an innovation, a firm needs to put its different
types of knowledge, resources, skills and capabilities together. For example, market
knowledge, financial resources, etc.
Innovation is a continuous process, i.e. it is often altered by small incremental changes in the product or process. Once an invention has been developed, it is often tested and then adjusted and improved before it is put on the market. Therefore it usually takes some time for an invention to turn into an innovation.
When dealing with innovation, Henderson and Clark (1990) make a distinction between incremental and radical innovation. According to these authors incremental innovation introduces relatively minor changes to the existing product, exploits the potential of the established design and often reinforces the dominance of established firms. Hollander (1965) argues that although incremental innovations do not come from dramatically new science, it often calls for considerable skill and ingenuity, and over time has very significant economic consequences. Radical innovation, on the other hand, is based on a different set of engineering and scientific principles, and often opens up new markets and potential applications (Henderson and Clark, 1990). Established firms often have great difficulties with implementing radical innovations (Cooper and Schendel, 1976; Daft, 1982; Rothwell, 1986;
Tushman and Anderson, 1986), however, it can be the basis for successful entry of new firms or even a redefinition of an industry (Henderson and Clark, 1990).
That innovation is very important for multinational corporations can be seen in the paper of Doz et al. (2001). In their paper they argue that it is the firm’s innovative capability that is the driver of competitive advantage and that this advantage is based on “identifying, accessing, and utilizing pockets of specialist knowledge drawn from around the world”. They argue that this will lead to new and powerful sources of value creation and competitive advantage, which traditional multinationals do not have. When innovation is considered a source of competitive advantage, it represents a strategic change (Cooper & Schendel, 1976).
Therefore it becomes part of the firm’s strategy and top management’s responsibility (Ettlie, Bridges, & O'Keefe, 1984).
Cheng & Bolon (1993) have also addressed innovation as a highly important feature of
international strategy. They argue that a firm’s ability to develop new products and processes
allows them to better address the needs of diverse overseas costumers as well as to
appropriate the benefits of its innovations more advantageously as compared to simply selling
or licensing its technology.
MEASUREMENTS OF INNOVATION
Innovation can be measured in several ways. The most popular indicator of innovation is the expenditures on research and development (R&D) (Klomp & Van Leeuwen, 1999; Lööf et al., 2001). To come to the R&D intensity of a company, the total R&D expenditures are often divided by total sales. The main advantage of this indicator is that it improves the comparability with other studies, because it is used in many other studies, like Kemp et al.
(2003). Kleinknecht (2000), on the other hand, mentions that this indicator also has several weaknesses. R&D expenditures are an input to the innovation process, however it does not state anything about the results, nor about the efficiency. Another limitation is that R&D expenditures are a minority input compared to total innovation expenditures, varying from 25 to 50 percent. R&D data also tends to underestimate innovations in the service sector.
Another indicator of innovation is the number of employees dedicated to R&D (Kemp et al., 2003). The main advantage is that it better suits service sectors than R&D expenditures.
Kemp et al. mention also several weaknesses of this indicator. Like R&D expenditures it provides no information on efficiency and it is a minority of total expenditures, but it also does not include the quality of the employment input and the time devoted to innovation (Kemp et al., 2003).
Some authors, like Kleinknecht, use the number of patents received as an indicator of innovation (Kleinknecht, 2000). There are two major advantages to this indicator. First, the abundance of publicly available information, and second the minor disturbances in these series. However, there are also some problems with this indicator. There are companies that apply for a patent for a strategic purpose, i.e. to misguide a competitor (Kemp et al., 2003).
However, there are also innovations that cannot be patented or just are not patented. One possible explanation is that companies fear inventing around by competitors (Cohen, Nelson,
& Walsh, 2000). Inventing around means that rival firms are able to read laid open patent applications, and develop similar and competing products (McDaniel, 1999). Another reason is that a company is unwilling to provide its “recipe” of the innovation (Cohen, Nelson, &
Walsh, 2000).
In this paper innovation will be measured in two ways, namely by calculating R&D
expenditures divided by total sales and by looking at the number of patents received. R&D
expenditures divided by total sales is chosen because of its comparability advantage and
because it reflects the choice of the top management team. The top management team chooses
how much they are willing to invest in R&D. Patents, on the other hand, are chosen because
they relate directly to an innovation, i.e. it looks at the innovation output, and therefore covers the efficiency loss generated by R&D expenditures. The only problem is that a patent is not automatically caused by the actions of a management team in a certain year, but it often relates to innovations invented several years before. This will be explained in more detail in the methodology section.
HETEROGENEITY
There are theorists that argue that large firms cannot be controlled by a single leader or a management team (Hall, 1977). Hall argues that large companies only need maintanance and that a leader has no power to steer the organization. This view that leaders do not make a difference is supported by several other authors (e.g. Lieberson & O'Connor, 1972; Lubatkin et al., 1989). Hambrick and Mason (1984), however, have developed a different view. They argue that decision makers do have an influence on the organization. This view is not entirely new, however, the contribution of their article to existing literature has its roots in that they have chosen the management team as the unit of analysis and that they argue that the background characteristics of managers have an influence on the choices they make.
Hambrick and Mason (1984) say that organizational outcomes, i.e. strategic choices and performance levels, are partially predicted by managerial characteristics. This is what they call the Upper Echelon perspective. The underlying idea of this perspective is that individual characteristics, like sex, age, tenure, socio-economic background, and personality attributes have an influence on the preferences and attitudes of members, as well as the resulting team dynamics. These team dynamics have an influence on the strategic choices managers make and therefore affect organizational outcomes (Olie & van Iterson, 2004).
Hambrick and Mason (1984) have not empirically tested their view, but they have made
several propositions for future research. The propositions they have made are based on a
paper of Hambrick & Snow (1977), in which they present their sequential view of the
perceptual process. This sequential view starts with the notion that a manager, or even a team,
is not able to scan every aspect of the organization and its environment. This is because the
manager’s field of vision, i.e. those areas to which his or her attention is directed, is limited,
and it therefore limits his or her perceptions. Second, a manager’s view is further limited,
because he or she only selectively perceives some of the phenomena within his or her field of
vision. Further, the selected phenomena for processing are interpreted through a filter based
on the cognitive base and values of the manager. These cognitive base and values depend on
the background characteristics of the managers (Hambrick & Mason, 1984). Hambrick and Mason specifiy seven characteristics that may be influencial. These characteristics are: age, functional track, other career experiences, formal education, socio-economic background, financial position, and group heterogeneity.
These individual characteristics are not solely examined on their own, but also their variety in a team is discussed (Hambrick & Mason, 1984; Wiersema & Bantel, 1993;
Finkelstein & Hambrick, 1996; Olie & van Iterson, 2004). In this paper a team is defined as follows: “A team is a small number of people with complementary skills who are committed to a common purpose, performance, goals, and an approach for which they hold themselves mutually accountable” (Katzenbach, 1997). When managers in the same management team share the same background, joint experience and common perspective, it is assumed to provide a common vocabulary and a basis for mutual understanding, which will lead to greater cohesion in homogeneous groups. Homogeneous groups are therefore very well suited for routine problem solving (Filley, House, & Kerr, 1976; Keck, 1997). Heterogeneous groups on the other hand, are assumed to benefit from the broader spectrum of views which their members cherish. Heterogeneous groups are therefore considered better decisionmakers when faced with complex problems which call for new solutions that are beyond the cognitive capacities of an individual. Heterogeneous teams will therefore be more innovative and creative than homogeneous groups (Bantel & Jackson, 1989; Wiersema and Bantel, 1993).
The disadvantage of a heterogeneous group is that the variety of skills, knowledge, and values can easily hinder communication (Zenger & Lawrence, 1989; Keck, 1997) and this can result in an increase of conflict and power struggle (Olie & van Iterson, 2004).
There have been many papers that have tested the propositions of Hambrick and Mason, especially the influence of several TMT background characteristics on perfomance (e.g.
D'Aveni, 1990; Haleblian & Finkelstein, 1993). Some of those papers have tried to empirically test the effect of group heterogeneity on performance, (e.g. Elron, 1997). Elron focuses in his paper on the heterogeneity of one variable. In his paper he researches the relationship between different cultures of a top management team and its influence on firm performance. There are also papers that are claiming to measure the effect of multivariable heterogeneity on firm performance (e.g. Hambrick, Cho, & Chen, 1996; Carpenter, 2002).
However, they are not actually measuring group heterogeneity, but they are performing a
multiple regression analysis and claim that if several characteristics are significant,
heterogeneous teams have a better performance.
So far I have only found one paper that has researched the relationship between top management team heterogeneity and the amount of innovation (Bantel & Jackson, 1989).
Bantel and Jackson (1989) have found a positive relationship between the heterogeneity of several TMT background characteristics and the amount of innovation of United States banks. They have conducted their research in the same way as Carpenter (2002), i.e. they have conducted a multiple regression analysis. Bantel and Jackson (1989) have used heterogeneity with respect to age, job tenure, education and functional expertise (i.e. the different functional areas in which they have experience) in their analysis. The researchers have divided innovation into two categories, namely technical and administrative innovation. They then interviewed the bankers, in which the bankers were asked to generate a list of innovations the company has created and to give an indication of the percentage of banks who had adopted their innovation, the amount of costumer acceptance and the financial investment required.
Innovations that these bank executives believed were extremely rare, extremely common, had a very poor costumer acceptance or required very large financial investments, were excluded.
Based on these results, a measure of total innovation was created by summing the innovations of each bank.
This paper will not only research the relationship between several heterogeneity measures of TMT’s and innovation, but it will add another variable, namely overall heterogeneity. How this measure will be constructed will be explained in the methodology section. So far I have not found any other research that has measured the relationship between overall heterogeneity and innovation in this way, nor have I found a research that has tested it for an industry in the E.U.
The four variables that have been chosen to measure heterogeneity are: age, gender, field of education and nationality. Based on these four variables an overall heterogeneity score will be created. How this will be done is explained in the methodology section. These variables are chosen because they all reflect basic demographic characteristics of the managers.
This literature review can be summarized by the following conceptual model.
CONCEPTUAL MODEL
The next section will explain the literature that has been written on these four variables, as well as on overall heterogeneity. This will be accompanied by several hypotheses.
3. CONCEPTS & HYPOTHESES
This section deals with the concepts and hypotheses related to the the research question.
These concepts are age, gender, field of education, nationality and overall heterogeneity, and each concept will be discussed in turn.
AGE
Age is one of the basic demographic variables discussed in the paper of Hambrick and Mason (1984). Managers of different ages are expected to differ in their attitudes, values and perspectives for two reasons. The first reason is that people of different ages experience different social, economic and political environments and phenomena and this plays a fundamental role in shaping a manager’s attitudes and values (Bantel and Jackson, 1989).
Another reason is that a manager’s perspective changes, because of the aging process he is undergoing (Elder, 1975).
Ryder (1965) agrees that age is an important demographic variable, because it helps in predicting an individual’s non-workrelated experiences. People of a similar age typically share such experiences and it therefore leads to shared attitudes and beliefs (Rhodes, 1983).
Managerial characteristics:
- Age - Gender - Education - Nationality TMT
Heterogeneity
Diversity in manager’s:
- Cognitive base - Values
Strategic choices and outcomes:
- Innovation
Assuming that diversity of attitudes and values facilitates creativity, top management teams with more age diversity are expected to be more innovative.
This can be summarized by the following hypothesis:
H1: Top management team heterogeneity with respect to age will be positively correlated to innovation.
GENDER
Gender is also a basic demographic variable, however, it is not discussed in the paper of Hambrick and Mason (1986), nor has it been researched by many researches based on the Upper Echelon perspective. A possible explanation for this is mentioned by Plantinga (2006).
Plantinga argues that in the time Hambrick and Mason wrote their paper, women were almost non-existent in TMT’s and therefore very little was written on differences in decision-making between men and women.
Nowadays, much more is written on this topic. There are psychologists that have argued that sex differences in adult social behaviour can be explained by biological influences, such as the greater prenatal androgynization of men (e.g. Money & Ehrhardt, 1972). Other pshycologists have argued that childhood events are different for males and females, such as experiences of sex-segregated play groups in which boys and girls play in different styles and use different methods to influence each other (Maccoby, 1988). Therefore biological sex differences and different experiences may cause men and women to think and perceive differently, even if they would occupy the same management role. Many researchers have also conducted research in the management style of men and women and most of them have reveiled that women manage in a more democratic and participative way, whereas men manage in a more autocratic and directive style (e.g. Buttner, 2001; Eagley & Johnson, 1990;
Loden, 1985).
Assuming that this diversity of biological differences, different experiences and different management styles leads to more creativity, top management teams that are more diverse with respect to gender will be more innovative. This leads to the following hypothesis:
H2. Top management team heterogeneity with respect to gender will be positively correlated
to innovation.
FIELD OF EDUCATION
Another basic demographic variable mentioned by Hambrick and Mason (1984) is education. Hambrick and Mason dinstinguish between two types of education, namely the level of formal eduction and the formal educational background, ie. the field of education.
There are several studies that have researched the relationship between the level of formal education and innovation, and most of them have shown that the level of education is positively related to receptivity to innovation (Becker, 1970; Kimberly & Evanisko, 1981;
Rogers & Shoemaker, 1971). This paper does not research the receptivity to innovation, therefore it is not very relevant to look at the diversity of the level of formal education. In this paper it would be more relevant to look at the diversity in the field of education and compare that with innovation.
To a certain degree, education indicates a person’s knowledge and skill base. A person who is educated in economics is expected to have a different cognitive base than a person who has studied history or law. Furthermore, if people take their decisions about education seriously, then education can serve to some extent as an indicator of a person’s values, cognitive preferences, etc. (Hambrick & Mason, 1984). Holland (1976) has found empirical evidence that educational curriculum choices of people, correspond to their personalities, attitudes, and cognitive styles. This suggests that top management teams that have more dissimilar fields of education are expected to benefit from this diversity of perspectives of the team members.
This can be summarized in the following hypothesis:
H3: Top management team heterogeneity with respect to the field of education will be positively correlated to innovation.
NATIONALITY
Although nationality is a basic demographic variable, it is not discussed by Hambrick and
Mason (1984). However, nationality is an important variable in this paper, especially because
this paper deals with MNC’s. This is because different views might be needed in the different
countries in which they operate. Hofstede (1980) has performed a research in which countries
receive an average score on multiple value dimensions. These dimensions show that people
with different nationalities have different values. The paper of Hambrick, et al. (1998) has
investigated the influence of nationality in which they argue that the effect of diversity
depends on the task of the group. Multinational diversity causes problems when the task is
coordinated, like monitoring and maintenance, however it can be very beneficial when the task requires creativity. Therefore it is expected that a more nationaly diverse TMT will be more innovative.
This leads to the following hypothesis:
H4: Top management team heterogeneity with respect to nationality will be positively correlated to innovation.
OVERALL HETEROGENEITY
As mentioned above, there are several authors that have found a positive relationship between heterogeneity and firm performance (Hambrick, Cho & Chen, 1996; Carpenter, 2002; D'Aveni, 1990). These studies are based on the paper of Hoffman and Maier (1961), which states that heterogeneous groups come up with qualitative better solutions. Filley, House, & Kerr (1976) agree with Hoffman and Maier and claim that heterogeneous groups are better at handling ill-defined, novel problem solving, because their diversity of opinion, knowledge, and background allows them to carefully analyse the alternatives.
Bantel and Jackson (1989) have empirically tested the relationship between heterogeneity and innovation by looking at 199 U.S. banks and they have found a positive relationship. This paper will look at the relationship between overall heterogeneity and innovation in the manufacturing industry of electrical machinery and apparatus. However, this paper will test overall heterogeneity in a different way. Bantel and Jackson (1989) have looked at four variables of heterogeneity and have performed a multiple regression analysis. This paper will add an overall score to measure group heterogeneity. How this overall score will be calculated will be explained in the methodology section. This paper follows the idea of Bantel and Jackson (1989) and assumes that a more heterogeneous TMT will result in more innovation.
This leads to the following hypothesis:
H5: Overall top management team heterogeneity will be positively correlated to innovation.
The result of the group heterogeneity test will be compared with the results of the other
four independent variables.
4. RESEARCH DESIGN
METHODOLOGY
This paper will research the relationship between TMT heterogeneity and innovation by conducting an empirical analysis. The focus of this paper will be on MNC’s in the European Union that are operating in the manufacturing industry of electrical machinery and apparatus.
The E.U. has been chosen as a unit of analysis, because it has stated in the treaty of Lisbon (2000) that it wants to become the most innovative region by 2010, and also because no research has been conducted so far on this relationship in the E.U. Multinational corporations have been chosen for three reasons. Firstly, because they are usually large companies and those companies are most likely to provide data on the background characteristics (age, gender, etc) of the top management team on their websites. Top management teams are defined as those managers that are directly reporting to the CEO in the company hierarchy.
The second reason is that Datastream (a database in the RUG library) does not have the data on R&D of very small companies. The last reason is that heterogeneity with respect to nationality is included in the analysis and therefore MNC’s become very interesting, because they operate in different countries with different cultures.
The manufacturing industry of electrical machinery and apparatus has been chosen, because this industry is expected to be very innovative. This industry covers the rather traditional electrical engineering industry as well as the more recent electronics industry [Nace(31)]. This industry includes companies like Koninklijke Philips Electronics N.V. who has received 2852 patents in 2005.
In order to be able to draw conclusions from the analysis it is necessary to have a sufficient sample size. In this paper the rule of thumb of Green (1991) will be followed. This rule of thumb states that the sample size (N) should be greater than fifty plus eight times the number of independent variables, i.e. N>50+8*M, in wich M equals the number of independent variables. The amount of independent variables, including the control variables, are nine. Therefore the sampling size should be greater than 122. When collecting the data, there will probably be data that cannot be retrieved. In order to have a sample size of at least 122 MNC’s, data will be collected on the 150 largest MNC’s in the E.U., based on Amadeus (electronic database of the RUG).
The time period that will be used to collect data is 2005, except for the number of patents.
The time period for the number of patents is 2006. The first time period has been chosen,
because it is the most recent time period for which those data is available. For most
companies Amadeus only has data on sales revenues and the number of employees for 2005.
As will explained in the next section, sales revenues are needed to calculate one of the dependent variables and it is also used as a control variable. The number of employees are also used as a control variable. As a result only data of TMT’s will be used of 2005.
Companies that have top management teams that have changed in composition after 2005 will not be included in the analysis. On their websites many companies mention when a manager has taken his or her position in the management team. So if all TMT members have been in this position before 2005, the TMT will be included in the analysis and if a manager has joint the team after 2005, the TMT will not be included in the analysis. When the company website does not state when the managers have taken their positions, an email for clarification will be send to those companies. In case of doubt the TMT will be exluded from the analysis.
The time period for the number of patents has been chosen, because it takes some time before an innovation is created and it also takes a while for an innovation to be patented.
Therefore a patent received in 2005 is not automatically caused by the management team in 2005. There is usually a time lag of several years. Data on top management teams of 2005 are collected and therefore the number of patents of at least 2007 or 2008 would have to be collected. However, at the time of writing we are only half way 2007, so data from those years cannot be collected yet. For this reason 2006 has been chosen. Although patents from 2006 may not be entirely caused by the top management team of 2005, there are also management teams that were already present before 2005. For the management teams that were already present in this composition several years before 2005, the number of patents do reflect their choices. However, not all top management teams were already functioning in this composition before 2005, therefore the number of patents still does not make it a very good and robust predictor of innovation, but in this paper it will be used to check if the results with R&D expenditures as the dependent variable, also holds for the number of patents.
Therefore in the analysis the main focus will be on R&D divided by sales, because the amount of money spent on R&D reflects the choice of a management team in a certain year, whereas the number of patents is an outcome of those choices, but does not reflect the choices in the same year. Therefore the main focus will be on R&D expenditures and the number of patents will be used to get more robust results.
DATA
In order to test the relationship between TMT heterogeneity and innovation, data has to be
collected on the dependent, independent and control variables. The dependent variable in this
research is the amount of innovation. Innovation will be measured in two ways, namely by looking at the logarithm of the number of patents the company has received and the company’s R&D investments divided by total sales. The logarithm of the number of patents will be used, to limit the influence of large variations within the variable. The number of patents may vary heavily, however, a company with 1500 patents is not necessarily 15 times as innovative as a company with 100 patents. For second measurement of innovation, i.e. the company’s R&D investments divided by total sales, no logarithm will be used, because the influence of variations in R&D investments is already limited, because it is divided by total sales. For both measurements a separate regression analysis will be performed.
Data on the number of patents can be found in the European Patent Office, which is available on the internet (http://www.epoline.org/portal/public/registerplus). Data on R&D investments can be retrieved from the Datastream computer in the library of the RUG and from their annual reports. Data on total sales can be foud in the database of Amadeus.
The independent variables in this research are heterogeneity with respect to age, gender, education and nationality, and a variable that indicates overall heterogeneity. Heterogeneity will be measured by computing the heterogeneity score of Blau (1977). This score is a widely used measure for heterogeneity, for example in the papers of Hambrick, Cho, and Chen (1996) and Goodstein, Gautam, and Boeker (1994). This measure is calculated as follows:
n H = 1 - ∑ p
i2t=1
H indicates the extent of concentration of the board members. This score can range from 0
to 1, in which 0 indicates complete homogeneity and 1 complete heterogeneity. P
irepresents
the proportion of the board members in the ith group and n indicates the number of possible
categories. In order to calculate the heterogeneity score, all variables have to be put into
categories, for example gender consists of two categories, namely male and female. The
categories of all the variables can be found in the appendix. The overall heterogeneity score
will be measured by calculating the average heterogeneity score of the other four independent
variables. Data on age, gender, education and nationality can mostly be found on the company
website, in their annual reports or in the curriculum vitea’s on the internet. If the data is not
avaible, a survey will be send to those companies. On their websites companies usually only
include managers that are currently in the management team. The composition of the team may have been different in 2005. Therefore companies with management teams that have changed since 2005 are excluded from the analysis. On the company website it usually stated when the managers within the team have entered their position in the team. If all managers were already in the top management team, it would cause no problems and the TMT will be used for the analysis. When a manager has entered during or after 2005, the company will be excluded from the analysis. If the company website does not state when the manager has joined the TMT, an email will be send to the company in which they will be asked if the TMT composition has changed the last two years. When there is any doubt whether the TMT has changed in the past two years, it will be excluded from the analysis. This way the database only contains data on top management teams that were already active in this composition in 2005.
The environments in which firms compete can be very different, with hostile environments generating more demand for innovation in certain industries (Myers & Marquis, 1969). Not only the demand for innovation could differ accross industries, also the types of innovation can also be very different accros certain industries. Their can be industries in which companies rely heavily on product innovations, whereas others put more effort in process innovations. This paper minimizes these differences by studying firms in a single industry, namely the manufacturing industry of electrical machinery and apparatus.
Many papers have found a positive correlation between organization size and innovation (e.g. Baldridge & Burnham, 1975; Cohen & Mowery, 1984; Rothwell & Zegveld, 1985), however, the explanations for this relationship differ. One explanation is that innovations require large investments, which large companies can better afford. Another explaination is that large size neccessitates innovation in order to cope with the increased uncertainties that come along with growth (Kimberly & Evanisko, 1981). However, there are also studies that did not find a positive relationship or even a negative relationship between firm size and innovation. Kemp et al. (2003) mention in their paper that the influence of organization size on innovation differs among countries. In their paper they have found a positive effect in Norway, a negative effect in Finland and in Sweden the effect was not significant. In order to cope with its influence on innovation, organizational size will be added as a control variable.
Organizational size will be measured in two ways. First, by taking the logarithm of each
firm’s revenues. This is the established way to account for firm size differences when
examining organizational outcomes (Montgomery, 1979). Another way to cope with firm size
differences is to calculate the logarithm of the number of employees of the company. For
example Lööf et al. (2001) has measured firm size by the number of employees. The logarithms are used here for the same reason as for the number of patents, i.e. to limit the influence of large variations within the variable.
Another control variable of interest is team size. Although it has not received much theoretical attention, it is likely to be positively correlated with team heterogeneity (Bantel &
Jackson, 1989). According to Bantel and Jackson a positive correlation is especially likely to exist when the management teams are relatively small. This is because in a small team an increase of one additional person substantially increases the maximum amount of heterogeneity that is possible within the team. The teams that will be examined in this paper are relatively small, because only the top management team is under examination. Therefore team size has to be controlled for as well. Team size will be measured by taking the logarithm of the number of individuals in the top management team. Although team size does not differ heavily in size, the logarithm is used for consistency reasons. All the other control variables are also measured in logarithms.
The last control variable that this paper will deal with is company age. There are mixed results for variable found by different authors. Strebel (1987) argues that their exists a negative relationship between company age and innovation. He argues that mature companies have greater difficulty in organizing for innovation, either because competitive pressure for cost reduction has reduced resources that are made available for innovation, or because companies become more bureaucratic when they grow older. This last issue is based on the assumption that when a company grows older it becomes larger and develops a more standard way of doing business and hence becomes more bureaucratic. However, there are also authors that have found a positive relationship between company age and innovation (e.g. Klomp and Van Leeuwen, 1999). A possible explanation could be that more mature companies have more experience in innovation and therefore know the importance of devote resources to innovation. It could also be that older companies have better knowledge of the industry than younger companies and they therefore know that it is necessary to be innovative to survive in this industry. In order to cope with its possible influence on innovation, company age will be added as a control variable.
It could also be that the amount of innovation is influenced by the country in which a
company is operating. It could be that some countries put more effort into innovation
compared to others, because of e.g. different business systems, or different government
policies towards innovation. Although the European Union has stated that it wants to become
the most competitive region by 2010, it could be that some countries view innovation more
important than others and therefore are more supportive to innovation. The country of origin effect will not be added as a control variable to the analysis, because than the model becomes larger and with the number of observations in this dataset less robust conclusions can be drawn. However, in the analysis it will be checked if countries have an influence on innovation. This will be done in two ways. First, the four major countries in this sample will be taken into consideration and a dummy variable for these countries will be added to the final models. With these dummy variables, it can be determined if these four countries have an influence on innovation or not. Second, the country of origin effect will be measured by their type of capitalism. Hall and Soskice (2001) distinguish in their paper between by the extent to which a country is a coordinated market economy (CME) or a liberal market economy (LME). Based on this paper, Casey (2006) has developed an index in which each country is given a weight to what extent they are a CME or a LME country. To see whether the type of capitalism has an influence on innovation, the CME/LME scores will be added to the models in the analysis section. How this index what constructed will be explained in more detail in the analysis section.
SAMPLING DESIGN
To test the variables, a regression analysis will be performed by using SPSS. The hypotheses all assume linear relationships between the variables, therefore the Ordinary Least Square (OLS) method will be used. In order to be able to draw conclusions about the regression analysis, a multicollinearity check will be performed by looking at the Pearson correlation between the independent variables. If there is too much correlation between the independent variables, then these variables will be removed from the model. This way the results are not explained by a relationship between the variables itself. The dependent variable used in this model is the amount of innovation, which is measured by the logarithm of the number of patents received and by the amount of R&D expenditures divided by total sales.
The independent variables are the heterogeneity scores on age, gender, education, nationality
and an overall score on heterogeneity. The controlling variables are organization size, the
number of employees, team size and company age. This can be summarized in the following
table.
Table 1: Overview of the variables
Variable Type Name Measurment Source
Innovation DV Y
i,tLog nr.of patents &
R&D/total sales
Website EPO &
Amadeus Heterogeneity TMT
age
IV H.age
i,tH-score age Company website
& own calculations Heterogeneity TMT
gender
IV H.gender
i,tH-score gender Company website
& own calculations Heterogeneity TMT
education
IV H.education
i,tH-score education Company website
& own calculations Heterogeneity TMT
nationality
IV H.nationality
i,tH-score nationality Company website
& own calculations Overall
heterogeneity
IV H.TMT
i,tH-score group Company website
& own calculations
Organization size CV Org.size
i,tLog company
revenues
Amadeus & own calculations Nr. of employees CV Employees
i,tLog nr. of
employees
Amadeus
Team size CV Team size
i,tLog TMT members Company website
& own calculations
Company age CV Comp.age
i,tLog company age Company website
& amadeus
Legend: i=Company; t= Time; DV=Dependent Variable; IV=Independent Variable; CV=Control Variable.As explained in the literature review, a more heterogeneous top management team is expected to be more innovative. To test this relationship, the following regression model will be used:
Y
i,t= b
0+ b
1*(H.age)
i,t+ b
2*(H.gender)
i,t+ b
3*(H.education)
i,t+ b
4*(H.nationality)
i,t+ b
5*(H.(a/g/e/n))
i,t+ b
6*(org.size)
i,t+ b
7*(employees)
i,t+ b
8*(team size)
i,t+ b
9*(comp.age)
i,t+ ε
The description of the variables can be found in table 1 above. ε represents the error term, which is normally distributed and has a mean of 0 and a variance of σ
2.
In order to support the hypotheses, all variables are expected to have a positive and significant
correlation with innovation.
DATA COLLECTION
During the data collection there were some problems with data on background characteristics of the managers of the top management teams. This was caused by the fact that less companies then expected had all the data of their TMT members on their websites and the response rate on the emails that were send to those companies was very low (8.3%). As explained earlier in this paper, management teams that have changed in composition since 2005 were also deleted from the sample. This also reduced the sample size. In order to increase the sample size the number of companies was extended to the largest 200 companies in this industry. In the end this led to a sample of 84 companies. This is less than the sample size that would be required based on Green’s rule of thumb (Green, 1991), however the sample does represent the 200 companies in the industry quite well. From these 200 companies, Germany, Italy, France and the United Kingdom (U.K.) are the most active countries with respectively 25, 17.5, 16 and 10.5 percent. This indicates that 69 percent of this industry is dominated by these four countries. In the sample of this paper the top four countries in this industry are the same as above, only Italy is somewhat underrepresented. In the sample of 84 companies, Germany is most active, followed by France, UK and Italy with respectively 24.4, 14.6, 12.2 and 11 percent. So 62.2 percent of the sample is explained by the top 4 countries. This is summarized in the following table.
Table 2. Percentages of active countries
Industry: Sample:
Germany 25% 24.4%
Italy 17.5% 11%
France 16% 14.6%
UK 10.5% 12.2%
Total:
69% 62.2%
The other 37.8 percent of the countries in the sample are not highlighted here, because
only a few companies come from these countries. This other 37.8 percent consists of
companies from 11 different countries.
5. RESULTS
DESCRIPTIVE STATISTICS
In table 3 below the mean, the standard deviation, the minimum, the maximum and the median of the variables used in the regression analysis are presented.
Table 3. Means, standard deviation, minimum, maximum and median of the variables assessed in this study.
Variables: Mean: Standard deviation:
Minimum: Maximum: Median: N:
R&D / Sales 0.09 0.17 0 0.79 0.02 84
R&D exp. 25.5 mln 136.6 mln 0 1.2 billion 4.9 mln 84 Log nr. of
patents
0.77 0.83 0 3.45 0.65 84
Nr. Of patents 72.76 328.33 0 2852 4.50 84
H. TMT age 0.50 0.21 0 0.80 0.50 84
H. TMT gender 0.01 0.07 0 0.50 0 84
H. TMT education
0.17 0.27 0 0.81 0 84
H. TMT nationality
0.51 0.18 0 0.79 0.50 84
Overall Heterogeneity
0.40 0.15 0.15 0.72 0.37 84
Log organization size
7.93 0.71 5.83 8.82 8.10 84
Log nr. of employees
3.07 0.52 1.72 4.70 3.04 84
Log team size 0.61 0.21 0.48 .95 .60 84
Log company age
1.48 0.39 0.48 2.19 1.54 84
Organization size 165 mln 140 mln 0.7 mln 661 mln 126 mln 84 Nr. Of
employees
3017.68 7431.08 52 50119 1084 84
Team size 4.49 1.78 3 9 4 84
Company age 42.64 34.49 3 154 35 84
As can be seen in table 3, the amount of R&D expenditures varies from 0 to 1.2 billion
euro’s, with an average of 25.5 million (SD = 136.6 million). This indicates that there exists a
lot variation among R&D data. When R&D is used as a dependent variable the R&D expenditures are divided by the company’s sales. The table also presents the results on R&D divided by sales and shows that on average a firm invests about 9 percent of its sales in R&D, but there exists a lot of variation among the companies. Some companies invest nothing in research and development, whereas others invest up to 79 percent of total sales in R&D.
However, most companies of the companies invest 25 percent or less of total sales in R&D.
Eight of the 84 companies invest more than 25 percent in R&D. While analyzing the data, I will also look at the relationships when these eight companies are reduced from the sample.
When the number of patents is taken into consideration, it can be seen from table 3 that the average number of patents the companies received were 72.76 (SD = 328.33). The number of patents the companies received ranged in size from 0 to 2852 with a median 4.5. This large standard deviation and the large difference between the maximum amount of patents reveived and the median, indicates that there are probably a few outliers. A closer look at the data reveals that there is one huge outlier with 2852 patents in 2006, whereas the second largest has received 813 patents in 2006. The company with 2852 patents will therefore be deleted from the sample.
Table 3 also reveals that the average heterogeneity score with respect to age is 0.50 (SD = 0.21). As mentioned earlier in this paper, the heterogeneity score can range from 0 to 1. The lowest heterogeneity score with respect to age was 0 and the highest score was 0.8. An average heterogeneity score of 0.5, a standard deviation of 0.21 and a median of 0.5 indicates that there is a lot of variation in heterogeneity scores with respect to age and that there are no real outliers, which is good for the analysis.
As can be seen in table 3, the average heterogeneity score with respect to gender has a mean of 0.01 (SD = 0.07). The heterogeneity scores with respect to gender ranged in size from 0 to 0.50 with a median of 0. This indicates that more than half of the companies had no variation with respect to gender among their top management teams. A closer look at the data reveals that only three companies had women in their TMT. This shows that nowadays the top management teams are still dominated by men. Because only three companies in the sample have women in their TMT, it is not very useful to include this variable in the analysis and therefore gender will not be used in the analysis.
When heterogeneity with respect to education is dealt with, it can be seen from table 3 that
average heterogeneity score with respect to education is 0.17 (SD = 0.27). The median of the
heterogeneity score with respect to education is 0, like the median of the heterogeneity score
with respect to gender. This indicates that more than half of the companies have a
heterogeneity score with respect to education of 0, which means that more than half of the companies consist of team members with the same educational background. For most companies the members of the management teams had degrees in economics, finance, business and engineering. The other categories for education were much underrepresented. A closer look at the data set reveals that 26 companies had managers with different fields of education on their management team and those companies that did have differently educated managers on their management team received very high heterogeneity scores with respect to education. This means that when they a company was diversified among its top management team, then it contained many different fields of education. Although about 69 percent of the companies had no managers with different fields of education, heterogeneity with respect to education will be kept in the analysis, because of the high scores of the companies that do have multiple fields of education among their board members, it is still interesting to see what its relationship is with innovation will be.
According to table 3, the average score of heterogeneity with respect to nationality is 0.51 (SD = 0.18). This is almost the same as for heterogeneity with respect to age, only the variation among age was slightly higher. Also when we look at the minimum, the maximum and the median, it can be seen that these are also almost the same as heterogeneity with respect to age. Just as with heterogeneity with respect to age, this indicates that the heterogeneity scores with respect to nationality vary a lot, but do not indicate that there are outliers, which is good for the analysis.
Overall heterogeneity has an average score of 0.40 (SD = 0.15). It ranges in size from 0.15 to 0.72 and it has a median of 0.37. The overall heterogeneity score is the average score of the other independent variables. In this case it is the average of the remaining three independent variables, because heterogeneity with respect to gender will not be used in the analysis. The overall heterogeneity score has a minimum of 0.15, because there was always at least one of the three variables with a heterogeneity score greater than 0. A heterogeneity score with an average of 0.40, a standard deviation of 0.15 and a median of 0.37 indicates that there is a lot of variation in this variable and that there are no outliers, which is good for the analysis.
As can be seen in table 3, the companies ranged in size from €0.7 million to €661 million,
with an average size of €165 million (SD = 140 million). The median was €126 million. This
indicates that half of the companies have sales revenues of €126 million or less. Having a
maximum of €661 million, an average of €165 million and a median of €126 million could
indicate that there are a few outliers. A closer look at the data reveals that are 11 companies
that have sales revenues of more than €300 million and of those 11 companies, there are 5
companies that have sales revenues of more than €400 million. Although the distribution is quite spread, when there is no significant correlation between company size and innovation, these 5 companies with sales revenues of more than €400 million will be removed from the analysis to see if this changes the results.
Another measurement of company size is the number of employees. Table 3 shows that the number of employees ranged in size from 52 to 50119 and has an average of about 3018 employees (SD = 7431.08). These data show that the number of employees varies a lot. The median of the number of employees was 1084 and the maximum was 50119. This indicates that there may be a few outliers. A closer look through the dataset reveals that there are three companies that have more than 10000 employees. First all the data will be used in the analysis and if there are no significant results, the dataset will be reduced by these three companies to see if it changes the results.
The top management teams of the firms ranged in size from three to nine members, with an average size of 4.49 (SD = 1.87). The median of the team sizes is four. As explained before, heterogeneity is expected to increase with the team size. Therefore the heterogeneity score for a team with nine members can be significantly higher than for teams with three team members. This effect could be reduced by including only the companies that vary in team size from four to six, however this would reduce the sample size by more than half of the database.
Even if only teams of three to six members were included, then the sample would be reduced by 28 companies. This would make the sample too small and therefore all the management teams will be used in the analysis and this variable will be kept as a control variable.
Table 3 also deals with the average age of the companies, and shows that average age of the companies is 42.64 years (SD = 34.49). The youngest company was three years old, whereas the oldest company was 154 years old. The median was 35 years old. This suggests that the ages of the companies vary a lot. This dataset therefore includes very young as well as very old companies and a closer look through these data reveals that there are no real outliers.
Therefore all the company ages will be used in the analysis.
Before starting to analyze the hypotheses, there are certain criteria to the OLS method that
the data have to fulfill. First the variables will be checked for intercorrelation by looking at
the Pearson correlations. The OLS method also assumes that the residuals are normally
distributed and that the residuals have equal variance, i.e. are homoskedastic (Wesselink,
1994). This will be checked for each model when the final models have been formed, by
creating a normal probability plot and perform the Levene test for each model. The normal
probability plot shows if the residuals are normally distributed, whereas the Levene test indicates whether the residuals are homoskedastic or heteroskedastic. This will explained in more detail when the models are checked for normality and homoskedasticity.
According to Anderson, Sweeney and Williams (2002), intercollinearity becomes a problem when the correlation coefficient is greater than +0.70 or less than -0.70. The Pearson correlation shows the amount of correlation between two variables and in what direction they are related, i.e. a positive or negative correlation. If two explanatory variables are too heavily correlated it becomes a problem, because then two explanatory variables explain almost the same variance in the dependent variable. The intercorrelations of the explanatory variable in this study are presented in table 4 below.
Table 4. Intercorrelations of the variables assessed in this study
As can be seen in the table no variables have a correlation coefficient of more than +0.70 or less than -0.70, therefore according to Anderson, Sweeney and Williams (2002) there are no problems with intercollinearity and hence no variables have to be deleted from the models.
R&D / sales
Log nr. of patent s
H.age H.edu cation
H.nati onality
H.TM T
Log comp.
size
Log nr.
of employe es
Log team size
Log compan y age
R&D / sales
1.000 .500 .228 .179 .359 .434 -.233 -.025 .277 -.145
Log nr. ofpatents
.500 1.000 .338 .187 .515 .564 -.254 .423 .298 .012
H.age
.228 .338 1.000 0.115 0.241 0.671 -0.147 0.127 0.580 -0.056
H.education.179 .187 0.115 1.000 0.127 0.467 0.095 -0.142 0.292 0.048
H.nationality
.359 .515 0.241 0.127 1.000 0.650 -0.167 0.303 0.335 -0.004
H.TMT
.434 .564 0.671 0.467 0.650 1.000 -0.141 0.201 0.598 -0.024
Logcomp.size
-.233 -.254 -0.147 0.095 -0.167 -0.141 1.000 -0.262 -0.198 0.123
Log nr. of employees
-.025 .423 0.127 -0.142 0.303 0.201 -0.262 1.000 0.021 -0.096
Log team size
.277 .298 0.580 0.292 0.335 0.598 -0.198 0.021 1.000 -0.096
Log
company age
