Amsterdam Business School
The impact of risk behaviour on innovation
Name: Ro Dielbandhoesing
Student number: 10114343
Date: June 23th, 2014
MSc Accountancy & Control, specialization Control
Faculty of Economics and Business, University of Amsterdam
First supervisor: Dr. ir. B.A.C. Groen
Abstract
In order to grasp a better understanding of what drives innovation prior studies in the field of Management Accounting and Control focus on the relationship between Management Control Systems and Innovation. From a different perspective but with the same goal this thesis first examines the relationship between risk behaviour and innovation. Introducing a new variable namely outliers in financial performance, the median and moderate effect this construct has on the relationship between risk behaviour and innovation is examined. The applied research method is survey-based research. The results support the postulate that risk behaviour positively influences exploratory innovation. Furthermore, results provide evidence for a positive relation between outliers in financial performance and exploitative innovation. The predicted moderate effect of risk taking behaviour and outliers in financial performance on innovation was not supported. Furthermore, the results reveals strategy to be a better predictor of innovation than risk taking behaviour or outliers in financial performance.
Table of contents
1 Introduction ... 5
2 Literature Review ... 7
2.1 Innovation ... 7
2.2 Risk taking behaviour ... 8
2.3 Outliers in financial performance ... 9
2.4 Hypothesis development ... 10 3 Method ... 14 3.1 Type of Research ... 14 3.2 Sample ... 15 3.3 Survey instruments ... 18 3.4 Measurement of constructs... 19
3.4.1 Outliers in financial performance ...Error! Bookmark not defined. 3.4.2 Innovation ... 20
3.4.3 Risk taking behaviour ...Error! Bookmark not defined. 3.4.4 Control variables ... 23
3.5 Data analyses ... 23
3.5.1 Basic data analysis ... 23
3.5.2 Reliability of measurement ... 25
3.5.3 Convergent and discriminant validity ... 26
3.5.4 Hypothesis testing ... 29
4 Results... 30
4.1 Descriptive statistics main variables ... 30
4.2 Regression analysis for Exploratory innovation ... 32
4.3 Regression analysis for Exploitative innovation ... 33
4.4 Robustness check ... 35
5 Discussion and conclusion ... 37
5.1 Conclusion... 37
5.2 Theoretical implications ... 38
5.3 Limitations... 39
5.4 Practical implications ...Error! Bookmark not defined. 5.5 Suggestion for future research ... 39
Appendix ... 41
References ... 42
1 Introduction
Innovation and accounting & control is a very interesting combination. Especially in the field of control, there isn’t extensive information regarding this subject. The interesting thing about innovation is the way to control this abstract phenomenon and understanding what drives innovation and why.
When it comes to innovation, companies that come to mind are Google and Apple. These companies have introduced innovative services and products with a huge impact on people’s life. The verb “to google” has even been added to several dictionaries. Davila (2000) argues that outperforming competitors in product development has emerged as a relevant source of competitive advantage.
With the introduction of Activity Based Costing, the Balanced Score Card and Customer Lifetime Value, the field of management control itself is has experienced some innovation in the past decades. However, management control research has studied the relevance of Management Control System (MCS) to the broader scope of research and development (Birnberg, 1988; Brownell, 1985; Rockness & Shields, 1984, 1988). These studies mainly characterize MCS as hindering, or at most being irrelevant in Research and Development settings (Davila, 2000). According to Bisbe & Otely (2004), both the innovation and the MCS research provide inconsistent findings regarding the relationship between formal MCS and product innovation. Their research results do not support the presumption that an interactive use of MCS favours innovation. They suggest this may only be the case in low-innovating firms while the effect is in the opposite direction in high innovating firms.
According to Miles and Snow (1978) a company’s focus on innovation is related to the organizational typology. Distinguishing between Prospectors, Defenders and Analyzers, they find different propensity towards innovation. Damodaran (2008) argue an aspect that requires further examination is the role risk taking plays in creating innovation. In order to add to research on what drives innovation, in this study this aspect is included and added with the dimension of financial performance compared to peers. It is suggested that outliers in financial performance are, stimulated by their position relative to the industry’s average performance, more open to new ideas and products, hence are more innovative compared to at average performing companies.
This leads to the following research question: “To what extent is innovation increased by risk taking behaviour?” Consequently, discriminating in financial performance, what is the relationship, if any, between outliers in financial performance and innovation?”.
In the past decade there has been an increased interest in examining the relationships between product innovation and the use of formal management control systems (Shields, 1997). A significant body of literature has explored the relationships between formal MCS and product innovation within subunits, taking R&D departments, product development teams and product development projects as the level of analysis (Abernethy & Brownell, 1997; Brown & Eisenhardt, 1995; Davila, 2000), but limited emphasis has been placed on the relationship between manager’s risk taking behaviour, the business units as an outlier in financial performance and innovation. Adding a new variable to prior research in order to grasp a better understanding of the predictors of innovation is seen as the contribution to science.
This thesis is organized as follows. In the next chapter the relevant literature is addressed, followed by the research method. After describing the research method, the results are presented. The last chapter contains the conclusion, the findings and limitations.
2 Literature Review
This chapter provides a literature review of innovation, risk taking behaviour and outliers in financial performance. Hypothesis development is presented in the last paragraph of this chapter.
2.1 Innovation
The first thing that comes to mind when thinking about innovation is a company with a huge R&D department, investing large amounts in new ideas, products and services with absolutely no guarantee that they will come up with something that can make up for the investments.
Companies like Google and Apple investments heavily in their R&D departments without a guaranteed return. Some success stories of life changing innovations such as the Ipad reveal the high possible returns. Innovations haven’t brought us all good things. Innovations in the financial industry since the 1980’s have resulted in complex financial products of which no one could really say what they were doing. The lack of transparency didn’t prevent the demand for high yield to increase rapidly. Globalization was first affected in financial markets and seen as a leading example for a borderless approach, with unlimited possibilities. The credit crunch showed the public the downside of globalization and the effect of financial incentives with respect to innovations.
Jansen et al. (2006) distinguish 2 types of innovation namely exploratory innovations and exploitative innovation. Exploratory innovation is defined as the development of new ideas, products and services for new markets and customers. According to Danneels (2002) this type of innovation creates new designs, new distribution channels and new markets. Exploitative innovation is defined as amending or slightly adjusting existing products and services focused on optimizing utilization. In today’s technologically driven society, the rapid change in technological developments and increasing aggressive competition, innovation is essential (Volberda et al., 2001). Increased global competition, fragmented markets and convergence of technology force companies to continuously innovate strategically by improving distressed companies and creating new potential through a combination of resources (Guth and Ginsberg, 1990; Volberda et al., 2001). Innovation is a crucial factor for companies to survive in the long term (Faems et al., 2005). In addition Tidd, Bessant and Pivat (2005) find innovation the way for companies to stay ahead of competitors and to create an competitive advantage.
2.2 Risk taking behaviour
Given the ubiquity of risk in almost every human activity, it is surprising how little consensus there is on the definition of risk (Damodaran, 2008). In 1921, Frank Knight defined only quantifiable uncertainty to be risk. In a paper on defining risk, Holton (2004) argues that there are two ingredients that are needed for risk to exist. The first is uncertainty about the potential outcomes from an experiment and the other is that the outcomes have to matter in terms of providing utility.
Corporate governance standards like COSO (Committee of Sponsoring Organizations of the Tradeway Commission) state that a company’s view and tolerance towards risk should be shaped by a process effected by an entity’s board of directors and applied across the enterprise. This view is aligned with Stulz (2008) stating that a company’s view on the amount or degree of risk to be taken is determined by top management.
Despite the rules and procedures set in place Cools (2009) argues that individual behavior has an enormous impact on an organization. Maccrimmon and Wehrung (1990) found that some executives perceive their tendency to take risk in line with the risk appetite derived from their behavior. Power (2009) considers defining risk appetite with a predominant focus on capital rather than human behavior a flaw. The theorem that risk appetite is based on personal characteristics is supported by these studies.
Powell and Ansic (1997) study shows that females are less risk seeking than males irrespective of familiarity and framing, costs or ambiguity. Cited from Halahan et al (2003) it is generally thought that risk tolerance decreases with age (see Wallach & Kogan, 1961; McInish, 1982; Morin & Suarez, 1983; Palsson, 1996). Halahan et al (2003) analysis reveals that income and wealth are significantly associated with financial risk tolerance. Furthermore, marital status is of influence on risk taking behaviour as evidence suggests that single investors are more risk seeking than married investors (Roszkowski et al 1993).
Risk appetite is ultimately reflected in decision making, resulting in either a risk seeking or a risk averse response. According to Sitkin and Pablo (1992) risk behavior may be characterized by the degree of risk associated with the decisions made to the extent that (i) their expected outcomes are more uncertain, (ii) decision goals are more difficult to achieve, or (iii) the potential outcome includes some extreme consequences.
March and Shapira (1987) explored risk taking in relation to decision making theory. The theory assumes that decision makers deal with risks by first calculating and then choosing among the alternative risk-return combinations that are available. Their study shows three major ways in
which the conceptions of risk and risk taking held by managers differs from what could be expected based on decision theory: (i) managers are quite insensitive to estimates of the probabilities of possible outcomes; (ii) their decisions are particularly affected by the way their attention is focused on critical performance targets; (iii) they make a sharp distinction between taking risks and gambling.
Kahneman and Tverskey (1979) critique of expected utility theory as a descriptive model of decision making under risk has led to the development of prospect theory. Under the expected utility theory decision making under risk focuses on utilities of outcome weighted by their probability. Khaneman and Tversky (1979) found that decision making under risk does not obey the axioms of this theory. In situation where winning is possible but not probable given miniscule probabilities, people’s decisions are based on the prospect of wining or losing, hence prospect theory. Decisions are altered in situations where winning or losing is certain (e.g. 100% probable). Moreover, Kahneman and Tversky (1979) emphasize the role of a reference point in decision making. The reference point is a critical element in prospect theory. It predicts that most individuals exhibit a mixture of risk-seeking and risk-averting behavior when the outcome is either below or above the reference point respectively
Weber, Blais and Betz (2002) study involved risk taking in 5 different domains, namely: financial, health/safety, recreational, ethical and social decisions. They found that the degree of risk taking was domain specific and that the respondents were not consistently risk averse or consistently risk seeking
Risk taking behavior or risk appetite specifically that of the business unit managers examined is this study is defined as the managers’ attitude towards accepting risk.
2.3 Outliers in financial performance
A company’s performance is often measured using financial performance measures. Most of the financial performance measures are common measures and broadly acknowledge and accepted in society. Some examples include net income, return on equity (RoE) and earnings before interest, taxes and amortizations (EBITA). These measures are derived from the company’s financial accounting. Net income or profit is the outcome of the company’s financial records based on accounting principles. The purpose of accounting is to facilitate accountability. Stakeholders such as share holders, investors, tax departments and supervisors require a company to be held accountable for its actions. Company’s fulfill this requirement by the publication of their audited
financial statements. The concept of accountability stretches back to the Agency theory. In the
principal-agent model the principal holds the agent accountable for its actions. On the other hand, is the principal faced with the problem of information asymmetry. The principal will set goals and demand performance in return for their interest in the company. The goal setting process is part of the budgeting and target setting process. Financial performance is compared to the target, but also to peers. Realizing the target could still mean that the company is the industry laggard. Differentiating between peers leads to the designation of at average financial performers and outliers in financial performance.
Matusik and Hill (1998) define environmental competitiveness as the extent to which external environments are characterized by intense competition. Competitive environments have been associated with financial performance due to the intensive pressures for higher efficiency and lower prices that lead to tighter margins and less organizational slack (Matisk and Hill, 1998; Zahra, 1996). Jansen et al (2006) research reveals that pursuing exploratory innovation is more effective in dynamic environments, whereas pursuing exploitative innovation is more beneficial to a unit's financial performance in more competitive environments. Although much research has been done on the relation between innovation and financial performance with results showing causality, little research has been found on the impact of financial performance on innovation.
2.4 Hypothesis development
The expected relation between the constructs: risk taking behaviour, outliers in financial performance and innovation can be illustrated as follows:
Figure 2.1 Relationship between risk taking behaviour, outliers in financial performance and innovation
Innovation is by definition a risk taking process. It requires financial investments with high uncertainty about expected return. There are no guarantees about the success of innovation and
Exploratory innovation Exploitative innovation Outliers financial performance Innovation Risk taking behavior a c b 10
there are no safeguards preventing the investment turning out to be a financial disaster. For a company to engage in risk taking activities either (senior) management has to have a high risk appetite and thus risk taking behaviour or external pressure builds up so high management feel forced to expand their risk appetite. Rationalizing this type of risk driven behaviour is not a hard task. In some cases it can be interpreted as a necessity to survive. On the other hand risk taking behaviour by (senior) management in a company with at average financial performance will not be executed in actual risk taking by the company because there is a lot at stake. These companies often have their products and or services in the cash cow stage based on the Boston Consulting Matrix. Innovating and launching a new, potentially unsuccessful product could cannibalize their current product and impact their business model negatively resulting in financial deterioration. Effectively, these companies face decision making under uncertainty posing a threat to their current stage of harvesting economic benefits.
Miller & Friesen (1982) studied the relationship between two strategic archetypes, which they labelled “entrepreneurial” and “conservative”, and the use of control systems. The firms in their sample were split into two strategic groups (the conservative and entrepreneurial) based on ratings of innovation and risk taking. The entrepreneurial firm experienced more hostile environments and compete through product innovation. The entrepreneurial model applies to firms that innovate boldly and regularly while taking considerable risks in their product – and market strategies. The conservative firms have low product differentiation, operate in homogeneous markets and stable environments. In sharp contrast to the entrepreneurial firms, conservative firms apply little innovation and risk taking. Miller & Friesen (1982) find that considerable risk taking is found in companies non hesitant of radical innovation. Hence, it is to be expected that risk taking behaviour stimulates exploratory innovation. Following Kahneman and Tversky (1979) who revealed that even if gains are achievable with likely probabilities individuals chose the prospect that offers the largest gain, it is assumed that exploitative innovation (seen: as the lower gain compared to exploratory innovation) is negatively influenced by risk taking behaviour. Succeeding Miller & Friesen’s (1982) and Kahneman & Tversky (1979) research the following hypothesis is defined:
H1a: There is a positive relation between risk taking behaviour and exploratory innovation H1b: There is a negative relation between risk taking behaviour and exploitative innovation
It is to be expected that outliers in financial performance increase the propensity to innovate, simply because they have more reason to do so. Outliers are referred as under- and outperformers compared to peers. Underperformers must seek ways of turning things around,
making them more open towards new ideas, products, markets and possibilities. The way they are running things now clearly leaves them behind peers. Outperformers who most likely are market leaders must innovate to stay on top and outperform competition. They feel the pressure of maintaining their competitive advantage. A threat to their current position is a competitor who launches a new idea or service which becomes broadly popular and can put them back.
Miles and Snow (1978) research described three successful organizational types - defenders, prospectors, and analyzers. This typology is based on business strategy, and focuses on the rate of change in products or markets. Defenders have a narrow product range and undertake little product or market development. The functions critical for organizational success are finance, production and engineering with less emphasis on marketing, research and development. Prospectors are described as continually searching for market opportunities and as being the creators of change and uncertainty to which their competitors must respond. The marketing, research and development functions dominate finance and production. Efficiency and profit performance are not as important as maintaining industry leadership in product innovation. Analyzers combine the strongest characteristics of defenders and prospectors.
Based on the different approaches to financial performance and attitude towards innovation as defined by Miles and Snow (1978) the following hypotheses are developed.
H2a: There is a positive relation between companies which are outliers in financial performance and exploratory innovation
H2b: There is a negative relation between companies which are outliers in financial performance and exploitative innovation
A positive relation between financially underperforming business units and exploratory innovation is predicted due to the excepted necessity of a radical change to turn things around. Outperforming business units face the threat of their current products becoming obsolete which require new products and services to be developed. Following Miles and Snow (1978) a positive relation is expected between outliers in financial performance and exploratory innovation for organizations that fit the criteria of prospectors. Finance does not dominate prospectors relieving them from that burden, which based on the agency theory could be enforced by the principal.
The negative relation between companies that are outliers in financial performance is predicted based on the assumption that relative small changes to existing products will not prove to be enough to either get out of their current predicament nor to keep competitors behind. Furthermore, defender organizations strongly focus on financial result, alerting them to the need
for innovation in order to improve their financial performance. A higher awareness of their financial performance relative to peers could incentive innovation.
Chenhall & Morris (1995) and Van de Ven (1986) argue that enhancement of the relationship between product innovation and performance should be particularly strong when product innovation is high, since avoiding the risks of product innovation initiatives that could damage the firms objectives need to be fine-tune and redirect actions should be particularly crucial in contexts where innovative ideas, initiatives and transactions proliferate.
Derived from their findings the following postulate is hypothesized with respect to the moderate effect of risk taking behaviour and outliers in financial performance on innovation. H3a: The relation between risk taking behaviour and exploratory innovation is positive in case of financial outliers
H3a: The relation between risk taking behaviour and exploitative innovation is negative in case of financial outliers
Miller and Friesen (1983, p. 223) argue that extensive risk taking and strong emphasis on novelty (i.e., exploratory innovation) can be hazardous when competitive conditions become more demanding and pursuing such high risk and high-cost innovations would considerably harm the viability of the business unit (Zahra and Bogner 1999). Success, when it had a significant relationship with risk propensity, was always positively associated with risk taking. A higher degree of success differentiated the risk takers from the risk averters (Maccrimmon and Wehrung 1990). Since exploratory innovation involves radical changes and some bold movements it is expected that the influence of risk taking behaviour through financial performance has a positive effect on exploratory innovation and conversely has a negative effect on exploitative innovation.
3 Method
In this chapter the research method is addressed, starting off by categorizing the type of research used for this thesis. Second, an elaboration of the sample is given in which the data collection is explained. Subsequently the main characteristics of the respondents are addressed. The following paragraph contains an overview and explanation of the constructs used. This paragraph mentions how theoretical constructs are operationalized and scrutinizes the validity of the constructs through reliability checks. The last paragraph discusses the statistical analyses applied for hypotheses testing.
3.1 Type of Research
This paragraph deals with the categorization of the type of research. The type of research is categorized through several different dimensions. Per dimension a brief explanation is given which clarifies the selected category. The first categorization is based on the notion of the objective of research and consequential its implications for the type of study. Attention is also given to the accessibility of data and the data type. Furthermore, distinction can be made based on the unit of measure and duration of data measurements. Lastly the type of research is related to the taxonomy of categories of accounting research.
In general the objective of research is either theory building or theory testing. Theory building usually constitutes of a qualitative study which examines objects in its real life context. Theory testing usually constitutes of a quantitative study which formulates hypotheses based on existing theories. The formulated hypotheses are empirically tested with retrieved outcomes based on data collections. Gephart (2004) states that qualitative research focuses on the experiences of every day life realities by the meanings of social members, whereas quantitative research imposes meanings on members to explain a singular, presumed-to-be-true reality. Qualitative research intends to provide a meaningful representation of concepts with the use of words, talks and text. Quantitative research intends to do this with codes, counts and quantification of phenomena. This research is characterised as theory testing using a quantitative research method.
Quantitative studies can be based on either publically accessible data or on proprietary data. Public data is accessible to anyone for any reason. Well-known public databases are Compustat and WRDS. Proprietary data is privately owned and access is limited. The data used for this study is gathered through a survey. Responses are collected in a database which is
managed and owned by the University of Amsterdam. Access is limited to students that meet the applicable requirements. Hence, the type of data used in this research is proprietary data.
The data types can also be distinguished into three types namely time series data, cross sectional data and panel data. Time series data is data on the same entity in different points in time. Cross sectional data is data on different entities in the same point in time. Panel data is cross sectional time series data. Respondents to the survey used for this study stem from different companies. Distribution of the survey’s is based on participation of this specific program and is not restricted to dedicated periods over time. The data type used in this study best fits the criteria of cross sectional data.
According to Burrell & Morgan (1979) and Hopper & Powell (1986) research taxonomy of categories of accounting research, there are 3 types of accounting research in relation to Radical Change or Regulation and subjectivism and objectivism. These categories are a) Critical accounting research, b) Interpretive accounting research and c) Mainstream accounting research. This study can be classified as mainstream accounting research which lies in the quadrant of functionalism and objectivism. The main characteristics of mainstream accounting research are a distinction between observation and theory, the use of a hypothetico-deductive model to establish a scientific explanation and the use of quantitative methods of data collection to provide a basis for generalizations.
3.2 Sample
In this paragraph an elaboration of the sample and data collection is given. The MACS project and process by which the data is obtained is explained. Furthermore the background characteristics of the respondents are presented.
The University of Amsterdam (UvA) offers a project called ‘Project Management Accounting and Control Systems (MACS project)’. Students can enrol in this program which facilitates a survey based study to write a thesis. Students must provide respondents that meet certain criteria. Each participating student must provide at least 10 completed surveys. After gathering the required minimum number of completed surveys the students must insert the data in a formatted Excel sheet and return the filled in Excel sheet to the project board. Students must also provide the project board with the hard copy filled in surveys and document evidence regarding the respondents. Evidence focuses mainly on the identity of the respondent and is usually shaped through submission of a business card or email of the respondent to the hard
copy survey. After approval by the project board access to an anonymized database containing the total dataset of all completed surveys is granted.
The survey contains questions with the purpose of gathering information about the design and use of management accounting & control systems in business units. Respondents to the survey are business unit managers who are responsible for a relatively autonomous and independent unit of an organization which employs at least 15 people. The term relatively autonomous refers to the fact that the business unit manager can take decisions independently on a
number of issues, within the legal and financial framework.
Above stated is noted on the cover page of the survey, by which respondents are informed about the requirements as well as the purpose of the survey. The survey is distributed as a University of Amsterdam survey including patented logo and not as a constructed survey by myself. Respondents are given the opportunity to contact dr. S.P. van Triest and or dr. F.H.M. Verbeeten MBA should they encounter difficulties or have any questions regarding the survey, as their contact information is included on the cover page of the survey as well. Respondents can fill in their name and email address if they are interested in a summary of the final results. The survey ends with a section in which respondents can fill in any remarks or comments. Furthermore, the faculty has drawn up an explanatory letter which can be presented to the respondents. In this letter the respondents are assured that the information provided will we handled confidentially and will be anonymized. They are also reassured that is not possible to come to any conclusions regarding their performance as a manager. It should be noted that respondents did not receive any financial reward for completing the survey.
The demographics (experience, age, education) and other characteristics of the respondents are shown in the table below. Examination of the main characteristics of the respondents is necessary for the identification of variables which potentially could distort the results of this study and therefore need be controlled.
Table 3.1: Characteristics of the respondents
The total number of respondents is 149. Managers have an average of approximately 7 years of experience in the business unit and work an average of approximately 5.5 years in their current
N Min Max Median Mean Standard Deviation
Experience in the business unit (in years) 149 0.4 41 5 7.280 6.751
Experience in the position (in years) 149 0.4 53 4 5.347 5.983
Age 149 24 63 45 44.617 8.107
Education 149 1 7 0 4.040 1.241
Number of FTE's in the business unit 149 9 4500 42 194.255 489.298
Annual sales (x 1.000) 124 0 35,000,000 5,700 488,625 3,533,999
position. Mangers’ average age is 45 years and the average education level is Post HBO. Average manager’s span of control amounts to approximately 194 FTE and annual sales of around 489 million euro’s. It should be noted that the data of the variable Education level has been transformed from nominal data into a ordinal data. Educational level is sorted and scaled ascending. The recoding table is appended in Appendix 1.
The distribution of respondents (managers) per industry is shown in table 3.2 and figure 3.1. A clear concentration of respondents in the financial industry is identified, since 45% of the respondents are active in this industry. The financial services industry is an interesting case for innovation researchers because it has been confronted with the blurring of industry boundaries and with new entrants from, among others, the retail and telecom industries (Flier et al. 2003). These changes have triggered incumbent financial services firms to pursue several exploratory and exploitative innovations, such as the introduction of ATMs, Internet banking, and mobile banking (Han et al. 1998, Pennings and Harianto 1992). This is interpreted as indicative for an additional robustness check which controls for the potential effect the variable industry may have on the result of the study. The robustness check is included in paragraph 4.4.
Table 3.2: Distribution of respondents per industry including the description of the industry code.
Code Description Number of
Respondents
Perc of Total
A Agriculture, hunting and forestry 0 0%
B Fishing 0 0%
C Mining and quarrying 0 0%
D Manufacturing 4 3%
E Electricity, gas and water supply 1 1%
F Construction 3 2%
G Wholesale and retail trade; repair of motor vehicles, motor cycles and household goods 13 9%
H Hotels and restaurants 1 1%
I Transport, storage and communications 10 7%
J Financial intermediation 67 45%
K Real estate, renting and business activities 23 15%
L Public administration and defense, compulsory social security 12 8%
M Education 4 3%
N Health and social work 11 7%
Total = n 149 100%
Figure 3.1: Distribution of respondents per sector; no respondents are active in sector A till C. 3.3 Survey instruments
The survey consists of 4 sections and contains a total of 96 items. The first section contains 11 items which inquires some general information from the respondents and their current position in the business unit. These items relate among others to working experience, gender, age, number of FTE employed by the business unit and type of industry. Section 2 to 4 addresses several Management Accounting and Control constructs. The survey was developed based on existing instruments as used in previous studies. An overview of these constructs sorted by item and section is appended in Appendix 2.
For answering 78 of the total of 96 questions a 5 point Likert scale is used. The definition of the corresponding scale is presented in the header above the relevant questions. Of the remaining 18 questions, 11 questions refer to general information and for 7 questions a percentage is asked.
In order to demonstrate reliability of measurement Cronbach alpha is applied. Cronbach alpha estimates and is the lower bound to the portion of variance attributable to common factors among items. That is, it is an index of common-factor concentration and may be applied as a modified technique to determine the common-factor concentration among a battery of subtest (Cronbanch, 1951). Cronbach’s alpha takes into account differences in the item standard deviations and is smaller than standardidezed item alpha to the extend these differences exist. Cronbach Alpha is a more stable measure of reliability because it is the mean of all possible splits and is not subject to randomness in the particular way one chooses to split the test used. (Cortina, 1993). George and Mallery (2003) provide the following rules of thumb for interpreting
0% 10% 20% 30% 40% 50% 60% D E F G H I J K L M N
Distribution of respondents per industry
alpha’s from which can be concluded that an alpha higher than 0.7 qualifies as “good” and an alpha lower than 0.5 is considered to be unacceptable.
Table 3.3 Cronbach alpa’s scale
Gliem and Gliem (2003) conclude that using Cronbach’s alpha to test reliability, analysis of data must be based on summated scales or subscales and not individual items since Cronbach alpha does not provide reliability estimates for single items. While a high value indicates good internal consistency of the items of scale, it does not mean that the scale is unidimensional. Factor analysis will be used as a method to determine the dimensionality of a scale.
3.4 Measurement of constructs
This paragraph goes into the measurement of constructs. Measurement is limited to constructs applied in this study and the related items from the Survey. However, outliers in financial performance are determined by additional analyses.
3.4.1 Risk taking behaviour
The survey designed by the UvA consist of a 6 items scale measuring risk taking behaviour. Measurement of the construct risk taking behaviour is drawn from instruments defined by Sitkin and Weingart (1995). Their research examined the usefulness of placing risk propensity and risk perception in a more central role in models of risk decision making than had been done previously.
Baird & Thomas (1985) and Bettman (1973) define risk perception as an individual's assessment of how risky a situation is in terms of probabilistic estimates of the degree of situational uncertainty, how controllable that uncertainty is, and confidence in those estimates.
Risk propensity is defined as an individual's current tendency to take or avoid risks. It is conceptualized as an individual trait that can change over time and thus is an emergent property of the decision maker (Sitkin and Weignart, 1995). Their approach to risk propensity as “stable but changeable” builds upon and modifies the traditional constant view. By focusing on the important role of past experience, this conceptualization can account for the capacity for people to adapt without denying that as individuals gain more experience, they may be less susceptible
Cronbach's alpha Qualification Cronbach's alpha Qualification
> 0.9 Excellent > 0.6 Questionable
> 0.8 Good > 0.5 Poor
> 0.7 Acceptable < 0.5 Unacceptable
to contextual influence and more likely to exhibit cross-situational consistency. This definition of risk propensity, which follows Sitkin and Pablo (1992), is related to but departs in a critical way from previous conceptualizations of propensity as a stable and constant dispositional attribute (e.g., Fischhoff et al., 1981; Rowe, 1977).
Sitkin and Weingart (1995) adapted a four item scale from MacCrimmon and Wehrung (1985, 1986a, 1986b) and Wehrung et al. (1989) to measure the amount of perceived risk associated with the decision to be made. A modified version of a widely used decision-making case, Carter Racing (Brittain & Sitkin, 1989), was used to construct a 5 items scale measuring risk propensity. Carter Racing places decision makers in a situation of risky choices. The case masked the facts of the National Aeronautics and Space Administration's (NASA) decision to launch the space shuttle Challenger in the guise of a decision about whether a race car team should compete in the last race of the season. The decision involves risk in a number of domains. Sitkin and Weingart (1995) focus on a specific domain namely on business risk which is defined as the impact of the decision whether to race on the financial viability of the race car team as an organization.
The study for this thesis adapted the scale of items designed by the UvA. Item number 64, measuring the manager’s tendency to choose more or less risky alternatives based on the assessment of lower level managers on whom he or she must trust, is excluded from the construct based on factor analysis.
3.4.2 Innovation
Innovation is distinguished into exploratory and exploitative innovations. Units that engage in
exploratory innovation pursue new knowledge and develop new products and services for new and
emerging customers or markets. Units pursuing exploitative innovation build on existing
knowledge and extend existing products and services for existing customers (Benner and Tushman 2003, p. 243).
Exploratory innovations are radical innovations which offer new designs, create new markets, and develop new channels of distribution. They require new knowledge or departure from existing knowledge (Abernathy and Clark, 1985; Benner and Tush man, 2002; Danneels 2002, Levinthal and March, 1993; McGrath, 2001).
Exploitative innovations are incremental innovations and are designed to meet the needs of existing customers or markets which build on existing knowledge and reinforce existing skills,
processes, and structures (Abernathy and Clark 1985, Benner and Tushman 2002, Danneels 2002, Levinthal and March 1993, Lewin et al. 1999).
According to Jansen et al. (2006) appropriate scales for measurement of exploratory and exploitative innovation were not yet available. Therefore, they reviewed relevant literature and generated a pool of items to tap the domain of each construct. They defined a total of 14 items (7 items per construct) of which 12 (6 per construct) are included in the survey designed by the UvA. Factor analysis conducted by Jansen et al (2006) has led to the exclusion of 2 items.
In this study measurement of the construct exploratory innovation is adapted by exclusion of the item measuring the use of new distribution channels (item number 87). Exclusion is based on factor analysis which revealed that inclusion of this item could compromise the unidimensionality of the construct exploratory innovation.
3.4.3 Outliers in financial performance
According to Cochran and Wood (1984) there is no real consensus on the proper measure of financial performance. In fact, there is a wide range of such measures. However, most measures of financial performance can be classified into two broad categories: investor returns and accounting returns. The basic idea underlying investor returns is that returns be measured from the perspective of the shareholder. The thought behind using accounting returns as a measure of financial performance is to focus on how firm earnings respond to different managerial policies. Cochran and Wood (1984) argue that accounting returns may be the best proxy for financial performance.
A similar point of view is found in Huselid’s (1995) study. He states that prior work on the measurement of corporate financial performance is extensive. Perhaps the primary
distinction to be made among the many alternative measures is between measurements of accounting and economic profits (Becker & Olson, 1987; Hirsch, 1991). Economic profits represent the net cash flows that accrue to shareholders; these are represented by capital (stock) market returns. Accounting profits can differ from economic profits as a result of timing issues, adjustments for depreciation, choice of accounting method, and measurement error.
Additionally, economic profits are forward-looking and reflect the market's perception of both potential and current profitability, but accounting data reflect an historical perspective. Although there is widespread agreement in the literature that capital market measures are superior to accounting data, accounting data provide additional relevant information (Hirschey & Wichern, 1984).
In line with Cochran and Woods (1984) and Huselid (1995) the survey designed by the UvA measures the construct financial performance through a 3 items scale consisting of both accounting and economic or investor returns. The study for this thesis uses outliers in financial performance. The first step in the process of determining outliers is a missing value analysis which revealed that 13 cases are identified with missing values. Due to uncertainty regarding a sufficient population size, the missing values are replaced by the mean score.
Based on Tukey (1977) the upper and lower hinges are calculated using a box plot. The upper and lower hinges are computed by adding the difference between the first and the third quartile multiplied by a g-factor of 1.5 and subsequently adding this distance to the first and third quartile. As shown in the box plot in figure 3.2 the upper hinges has a value of 3.8 and the lower hinges has a value of 2.5.
Values outside the upper and lower hinges were coded as outliers using ordinal data (0 = no outlier; 1 = outlier). The variable financial performance (FINREP) was transformed into outliers in financial performance by multiplying the code for identification of an outlier with the FINREP value.
Figure 3.2: box plot used for computing outliers in financial performance based on Tukey (1977)
3.4.4 Control variables
Control variables are used to increase confidence regarding the outcome of the results in relation to the independent variables. The control variable used in this study, including abbreviations and descriptions, are shown in table 3.4.
3.5 Data analyses
This paragraph contains information regarding the data used in this study, primarily focussing on the data quality and reliability checks. Apart from the descriptive statistics of the respondents presented in paragraph 3.2 the first paragraph gives insight by revealing the outcome of the missing value analysis, negative wording check and the response analysis. In the following paragraph Cronbach alpha is applied to demonstrates reliability of measurement. Convergent and discriminant validity of the constructs applied in the study for this thesis are addressed in paragraph 3.5.3. The paragraph is concluded with the regression models which will be used for hypotheses testing.
3.5.1 Basic data analysis
Statistical analyses are performed using SPSS software version 22. A problem which almost certainly is encountered in data research is missing values. Missing value analysis was conducted in order to identify the size of the population of non-missing values and a possible pattern that could explain which values are missing.
The items scaled to constructs were checked for negative wording. Negative wording items measure the opposite of the defined construct. Negative wording items require reverse coding before computing variables. Table 3.4 contains the results of the conducted checks.
Table 3.4: overview of main – and control variables Variable
Construct Abbreviation Items Scale Negatively worded Missing values % of total Independent Risk taking behaviour* RTB 63,65-68 5 point Likert No 5 3.4%
Outliers in financial performance** OFP 94-96 n.a. n.a 13 8.7%
Dependent Exploratory innovation* ERI 82-86 5 point Likert No 8 5.4%
Expliotative innovation ETI 88-93 5 point Likert No 8 5.4%
Control Age AGE 6 n.a. n.a 0 0.0%
Education** EDU 7 customized scaleRecoded into n.a 0 0.0%
Strategy: Prospectors SRI 34-37 5 point Likert No 2 1.3%
Strategy: Defenders STI 38-41 5 point Likert No 5 3.4%
Short term characteristics STC 71,72,74,75 5 point Likert No 1 0.7% * item deleted afte factor analyis
** item recoded
The items scaled to constructs are not negatively worded and hence none were reversed coded. The missing value analysis demonstrates that there is no cause for concern since values are missing randomly. As explained in paragraph 3.4.1 and 3.4.3 two items were deleted after factor analysis and 2 items were re-coded.
Response analysis was conducted to check for non-response bias by comparing early and late respondents. The data used for this study did not include the return date of the filled in survey’s by the respondents. However, it was noted that the data is updated by consecutive numbering of cases. The first fifty cases are coded as early respondents and the last fifty cases are coded as late respondents. An independent-samples t-test was conducted to compare exploratory and exploitative innovation in early and late respondents’ conditions. The results are shown in table 3.5.
With respect to Exploratory innovation there was a not a significant difference in the scores for early (M=2.8 SD= 0.8) and late respondents (M= 2.8, SD= 0.9) conditions; t(95)= 0.302, p = 0.763. With respect to Exploitative innovation there was also not a significant difference in the scores for early (M=3.4 SD= 0.7) and late respondents (M= 3.2, SD= 0.8) conditions; t(95)= 0.525, p = 0.093. These results suggest that early or late respondents do not have an effect on Exploratory nor Exploitative innovation. Hence, no indication of non-response bias was found. However, based on a 10% confidence level there is a statistically significant difference between early and late respondents with respect to Exploitative innovation.
A Mann Whitney U test indicates that was no difference for Exploratory innovation between early and late respondents (U = 1110, p = 0.638). It also showed that for Exploitative
innovation early respondents had a slightly higher mean rank than late respondents (U = 899.5, p
= 0.046). The back test of the independent two tailed t-test the using Mann Whitney’s U test revealed that there is a possible indicative of non-response bias for Exploitative innovation.
Table 3.5: independent two-tailed t-test
Respondents N Mean Standard
deviation t df p Early 47 2.823 0.757 .302 95 .763 Late 50 2.773 0.879 Early 47 3.446 0.711 .525 95 .093 Late 50 3.187 0.788
* p <0.05; equal variances assumed ERI
ETI
3.5.2 Reliability of measurement
The reliability check is conducted to determine if the items used in the survey are reliable. Cronbach alpha measures the inter-item correlation taking into account the number of items in a construct. It has a value between 0 and 1, with higher scores indicating a higher reliability level. The Cronbach alpha for the constructs applied in this study are shown in table 3.6.
Assessment of the internal consistency for the constructs Exploratory and Exploitative innovation results in an alpha of 0.792 and 0.790 respectively. The resulting alpha’s indicate a qualification of “Acceptable” and lie above the 0.70 level recommended by George and Mallery (2003).
Cronbach’s alpha for the construct Risk Taking Behaviour with a qualification of “Questionable” still supports reliability since an alpha of 0.627 exceeds the lower threshold of 0.50. High reliability for the construct Financial Performance is supported by an alpha of 0.814. Following Gliem and Gliem (2003) Cronbach’s alpha was not applied for single items.
Construct Object of measurement Abbre
viation Alpha Conclusion
Main Variables
Exploratory
innovation The business unit’s degree of innovation of new products and services
ERI .792 Acceptable
Expliotative
innovation The business unit’s degree of innovation of existing products and services
ETI .790 Acceptable
Risk taking
behaviour The manager’s risk appetite compared to other managers in the industry RTB .627 Questionable Financial
Performance The business unit’s financial performance compared to the industry average FINP ERF .814 Good Control Variables Strategy:
Prospectors The degree of emphasis in the business unit’s strategy regarding innovation of new products and services
SRI .636 Questionable
Strategy:
Defenders The degree of emphasis in the business unit’s strategy regarding innovation of existing products and services
STI .849 Good
Short term
characteristics The degree of short term focus in the manger’s personal characteristics STC .732 Acceptable Table 3.6: Cronbach alpha for constructs applied
3.5.3 Convergent and discriminant validity
Convergent and discriminant validity is assessed using factor analysis. Convergent validity indicates the likelihood of one underlying factor that might have had an effect on all item scores. Disriminant validity tests whether concepts or measurements that are supposed to be unrelated are, in fact, unrelated. Table 3.6 shows the factor loads, the eigenvalues and the percentage of common variance explained. Table 3.7 reveals the number of components loaded by the items used to scale the constructs.
Factor analysis results revealed a percentage of common variance explained equal to 54.7% and 49.1% for Exploratory and Exploitative innovation respectively, indicating that the items scaled loaded a single factor which supports the unidemensionality of the measured instrument. For the independent variables (Risk Taking Behaviour and Financial Perfomance) as
well as for the control variables results reveal a percentage of common variance ranging from 40.4% to 73.4%, which support high internal consistency.
Table 3.6: factor analysis convergent validity
Factor analysis used to determine discriminant validity demonstrate the loading of 9 components. Results shown in table 3.7 all have eigenvalues greater than 1 and loadings below
Construct Item Loadings Eigenvalues Percentage of total variance (%) ERI Item 82 .640 2.735 54.694 Item 83 .815 Item 84 .785 Item 85 .753 Item 86 .691 ETI Item 88 .726 2.943 49.055 Item 89 .689 Item 90 .749 Item 91 .732 Item 92 .625 Item 93 .673 RTB Item 63 .631 2.018 40.357 Item 65 .674 Item 66 .621 Item 67 .683 Item 68 .560 FINPERF Item 94 .838 2.202 73.3967 Item 95 .867 Item 96 .865 SRI Item 34 .632 1.947 48.677 Item 35 .815 Item 36 .764 Item 37 .547 STI Item 38 .840 2.766 69.161 Item 39 .823 Item 40 .863 Item 41 .799 STC Item 71 .756 2.238 55.943 Item 72 .796 Item 74 .794 Item 75 .635 27
0.50 are omitted. Factor analysis indicated that for each construct at least 3 items load to one factor. Furthermore it shows that there isn’t only one common factor and items do not correlate stronger with other constructs than one they reflect.
Table 3.7: factor analysis discriminant validity
Component 1 2 3 4 5 6 7 8 9 SRI Item 34 .608 Item 35 .765 Item 36 .758 Item 37 .470 STI Item 38 .781 Item 39 .791 Item 40 .858 Item 41 .707 RTB Item 63 .567 Item 65 .686 Item 66 .626 Item 67 .635 Item 68 .551 STC Item 71 .816 Item 72 .764 Item 74 .733 Item 75 .362 .738 ERI Item 82 .375 Item 83 .628 .498 Item 84 .562 Item 85 .719 Item 86 .786 ETI Item 88 .219 .720 Item 89 .289 .774 Item 90 .579 Item 91 .665 Item 92 .693 Item 93 .602 FINPERF Item 94 .817 Item 95 .821 Item 96 .828 28
3.5.4 Hypothesis testing
In order to test hypotheses H1a till H2b linear regression analysis is used. In using risk taking behaviour and outliers in financial performance as intervening variables, a mediated model allows for testing these hypotheses. In capturing the interaction outliers in financial performance has on risk taking behaviour and innovation a moderate model allows for testing H3 through moderated regression analysis.
Regression models used for testing hypotheses with exploratory innovation (ERI) as the dependent variable are expressed in equation form as follows:
Model 1: ERI = α + β1 RTB + β2 SRI + β3 STI + β4 STC + β5 Age + β6 Education + Ԑ Model 2: ERI = α + β1 OFP+ β2 SRI + β3 STI + β4 STC + β5 Age + β6 Education + Ԑ Model 3: ERI = α + β1 RTB + β2 OFP + β3 (RTB*OFP) + β4 SRI + β5 STI + β6 STC + β7 Age + β8 Education + Ԑ
Regression models used for testing hypotheses with exploitative innovation (ETI) as the dependent variable are expressed in equation form as follows:
Model 1: ETI = α + β1 RTB + β2 SRI + β3 STI + β4 STC + β5 Age + β6 Education + Ԑ Model 2: ETI = α + β1 OFP+ β2 SRI + β3 STI + β4 STC + β5 Age + β6 Education + Ԑ Model 3: ETI = α + β1 RTB + β2 OFP + β3 (RTB*OFP) + β4 SRI + β5 STI + β6 STC + β7 Age + β8 Education + Ԑ
4 Results
In this chapter, first the descriptive statistics of the main variables and the correlation between variables are reported. Secondly, the results of the regression analysis are discussed with the distinction between the dependent variable exploratory and exploitative innovation. The last paragraph contains the results of the robustness check.
4.1 Descriptive statistics main variables
The mean, median and standard deviation of the main variables is reported in the table below (see table 4.1). The mean score for exploratory innovation is 2.880 (SD = 0.841), indicating that respondents on average neither agree nor disagree with the statements regarding exploratory innovation. The mean score for exploitative innovation is 3.344 (SD = 0.742). With the exception of outliers in financial performance, the median and mean of the variables are fairly close together. The minimum score for outliers in financial performance is 0, which represents the non-outliers.
Table 4.1.: descriptive statistics of the main variables.
The bivariate correlations between variables is reported in a correlation table (see table 4.2) The outcome of the two-tailed test of significance is shown in brackets. There is a strong correlation between the independent variables exploratory – and exploitative innovation (r = .410; p < .01). The correlation table shows significant correlations between exploratory innovation and risk taking behaviour (p < .01). Exploitative innovation and risk taking behaviour are relatively well correlating (p < 0.10) and contrary to the expected direction (i.e. positive instead of the expected negative direction).
Data do not support proposition H2a stating that exploratory innovation is positively correlated with outliers in financial performance. As indicated in the correlation table the correlation between these constructs is not significant. The weak correlation between outliers in financial performance and exploitative innovation is demonstrated (r = .013; H2b is not supported).
Variable N Min Max Median Mean Standard Deviation
ERI 146 1 5 3 2.880 0.841
ETI 146 1 5 3 3.344 0.742
RTB 145 1 4 3 2.948 0.563
OFP 149 0 5 0 0.801 1.459
ERI = Exploratory Innovation ETI = Exploitative Innovation RTB = Risk Taking behaviour
OFP = Outliers in Financial Performance
Table 4.2 correlation table
ERI ETI RTB OFP SRI STI STC Age Education
ERI .410*** .269*** .064 .165** .620*** .033 -.069 .012 (.000) (.001) (.444) (.048) (.000) (.691) (.411) (.882) ETI .412*** 0.146* .013 .502*** .449*** .125 -.058 .050 (.000) (.084) (.878) (.000) (.000) (.131) (.488) (.551) RTB .301*** 0.163* .057 .087 .261*** .047 -.111 .000 (.000) (.053) (.493) (.300) (.002) (.571) (.183) (.997) OFP .115 .115 .090 -.040 .011 -.095 .025 -.063 (.165) (.166) (.284) (.628) (.891) (.251) (.767) (.446) SRI 0.147* .432*** .097 .028 .297*** .011 -0.139* .043 (.078) (.000) (.246) (.736) (.000) (.892) (.092) (.603) STI .617*** .444*** .264*** .015 .272*** .050 -.027 -.047 (.000) (.000) (.001) (.858) (.001) (.544) (.749) (.574) STC .022 .168** .071 -0.150* .070 .039 -.164** -0.146* (.794) (.043) (.399) (.067) (.398) (.640) (.045) (.075) Age -.033 .020 .003 -.025 .011 -.070 -.163** -.052 (.695) (.812) (.969) (.759) (.899) (.400) (.047) (.528) Education -.047 -.035 -.107 .021 -.132 -.016 -.163** -.031 (.573) (.674) (.200) (.803) (.110) (.848) (.047) (.710)
Pearson correlations are listed below the diagonal, non-parametric Spearman correlations appear above the diagonal ***P < .01, **P < .05, *P < .10; otherwise not significant
4.2 Regression analysis for Exploratory innovation
The results of the regression models for exploratory innovation are reported in table 4.3 The baseline model (Model 0) contains control variables. Model 1 and Model 2 respectively introduces managers’ risk taking behaviour and the business units’ financial performance. Model 3 examines the moderating effect of risk taking behaviour of the manager and the financial performance of the business unit on exploratory innovation.
Regarding the effects of risk taking behaviour on exploratory innovation, Model 1 shows that the coefficient for exploratory innovation is positive and significant (β = .153, p < 0.01). Hypothesis 1a is supported. The model explained a significant variance in performance of 39.9% (R² = .399, F(6,135) = 14.844, p < 0.01) . Model 2 shows that the coefficient for exploratory innovation is positive but not significant (β = .090, ns), thus hypothesis 2a is not supported. The results of the regression indicate that Model 2 is explains 39,1% of the variance (R² = .391, F(6,138) = 14.795, p < 0.01) Risk taking behaviour increases a business unit’s ability to pursue exploratory innovation. However, financial performance outlying the industries average does not incentive business units to maverick moves as required for exploratory innovation.
The results of the moderated regression analysis contained in Model 3 show that the coefficient for exploratory innovation is negative (β = -.855, ns). The total variance explained by the model of 47.6% is significant at a 90% confidence level, however the results do not support the postulate that risk taking behaviour and out- or under financial performance compared to industry’s average have a moderate effect on exploratory innovation. Since the coefficient is not significant H3a is not supported. The adjusted R square value reveals that the introduction of the moderate variable does not result in a increase of the variance explained.
It should be noted that for all models significant coefficients are found for control variable strategy of defender organizations. Results indicate that this variable is a better predictor of exploratory innovation than out- or under financial performance of industry’s average and the moderating effect included in H3a.
Table 4.3: Regression analysis for exploratory innovation
4.3 Regression analysis for Exploitative innovation
Table 4.4 contains the results of the regression analysis for exploitative innovation. Similar to the structure used for the regression analysis for exploratory innovation Model 0 is the baseline model which consists of the control variables.
The first regression model (M1) relates to hypothesis 1b. The coefficient on the degree of manager’s risk taking behaviour is positive but not significant (β = .041, ns). 32% of the total variance is significantly explained by the model (R² = .320, F(6,134) = 10.795, p < 0.01).
Hypothesis 1b is not supported.
The second regression model (M2) relates to hypothesis 2b. The model explains a significant proportion of variance in exploitative innovation scores (R² = .331, F(6,138) = 11.372, p < 0.01). The coefficient on the degree of out- or under financial performance compared to industry average is positively significant (β = .121, p < 0.10). This means that an increase / decrease of one point in the score for outliers in financial performance leads to an increase / decrease of .121 point in exploitative innovation. However, Hypothesis 2b is not supported as it predicted a negative coefficient.
The third model (M3) which examines the moderating effect of risk taking behaviour and financial out- or under performance compared to industry average on exploitative innovation explains 64.4% of the total variance. However, the coefficient for exploitative innovation is positively insignificant (β = .897, ns). Introducing the moderate variable increases the total variance explained from 30.3% to 64.4%. Despite this large increase and the strong relation (β = .897), the relation did not prove to be significant (p = 0.290). It is remarkable that for all models
Exploratory innovation
Variable Description M0 M1 M2 M3
RTB Risk taking behaviour .153** .521
OFP Outliers in financial performance .090 1.104
SRI Strategy exploratory innovation -.035 -.032 -.035 .066
STI Strategy expliotative innovation .628*** .583*** .624*** .402***
STC Short term characteristics -.018 -.027 -.005 .220
AGE Age .006 -.002 .009 -.161
EDU Education -.050 -.034 -.047 .123
RTB*OFP Interaction RTB and OFP -.855
F 17.293*** 14.844**** 14.795*** 3.299*
R .619 .632 .626 .690
R² .383 .399 .391 .476
Adj. R² .361 .372 .365 .332
***P < .01, **P < .05, *P < .10; otherwise not significant
strategy, both for Prospectors and Defenders are found as significantly relevant for predicting exploitative innovation.
Table 4.4: Regression analysis for exploitative innovation
Exploitative innovation
Variable Description M0 M1 M2 M3
RTB Risk taking behaviour .041 -.201
OFP Outliers in financial performance .121* -.365
SRI Strategy exploratory innovation .325*** .324*** .324*** 0.387***
STI Strategy expliotative innovation .349*** .339*** .344*** 0.368***
STC Short term characteristics .13* .132* .147** .074
AGE Age .058 .050 .062 .075
EDU Education .039 .040 .043 .085
RTB*OFP Interaction RTB and OFP .897
F 12.878*** 10.487*** 11.372*** 6.566***
R .563 .565 .575 .803
R² .317 .320 .331 .644
Adj. R² .292 .289 .302 .546
***P < .01, **P < .05, *P < .10; otherwise not significant
4.4 Robustness check
With reference to the observed concentration of industry (sector) in which the managers operate (see paragraph 3.2) a robustness check is required to determine the influence of this concentration on the results of the study for this thesis. In this paragraph the outcome of the robustness check is presented.
In order to indentify the influence of the sector on the results a dummy variable “Non Financial sector (NFS)” was computed. The ordinal data variable “industry” was transformed by valuing the code for the financial intermediation industry (J) as zero. All other industry codes were given the value 1. The results of the regression analysis including the dummy variable NFS are shown in table 4.5.
Table 4.5.: regression analysis including dummy variable non financial sector
The baseline model (Model 0) contains the control variables including NFS. Regarding the effect of risk taking behavior on exploratory and exploitative innovation, Model 1 shows that the coefficient for exploratory innovation is positive and significant. (β = .150, p < 0.05). Hypothesis 1a is supported. The coefficient for exploitative innovation is positive but not significant (β = .050, ns), subsequently not supporting Hypothesis 1b. Although the manager’s risk taking behavior increases the business unit’s ability to pursue exploratory innovation, it does not support the business unit's exploitative innovations as predicted. Model 1 does demonstrate a significant coefficient for exploitative innovation with regard to the non financial sector.
Robustness check
Variable Description
ERI ETI ERI ETI ERI ETI ERI ETI
RTB Risk taking behaviour .150** .050 .489 -.246
OFP Outliers in financial performance .085 .136* 1.005 -.503
SRI Strategy exploratory innovation -.027 .303*** -.025 .303*** -.028 .301*** .058 .375***
STI Strategy expliotative innovation .622*** .364*** .580*** .349*** .619*** .359*** .421*** .395***
STC Short term characteristics -.020 .136* -.029 .137* -.008 .155** .217 .071
AGE Age .011 .046 .002 .039 .013 .050 -.174 .057
EDU Education -.037 .008 -.024 .013 -.036 .009 .11 .067
RTB*OFP Interaction RTB and OFP -.742 1.055
NFS Non Financial sector .062 -.156** .055 -.150** .054 -.169 -.114 -.159
F 14.526*** 11.820*** 12.778*** 9.813*** 12.733*** 10.885*** 2.968** 6.233***
R .622 .583 .634 .584 .628 .598 .699 .817
R² .387 .339 .402 .341 .394 .357 .488 .667
Adj. R² .360 .311 .371 .306 .363 .325 .324 .560
***P < .01, **P < .05, *P < .10; otherwise not significant
M0 M1 M2 M3
With respect to outliers in financial performance Model 2 shows that the coefficient for exploratory innovation is positive but not significant (β = .085, ns), whereas the coefficient for exploitative innovation is positively significant. Despite the level of significance observed, Hypothesis 2b and obviously 2a are not supported. Hypothesis 2b predicted a negative relation between outliers in financial performance and exploitative innovation.
The moderated effect of risk taking behavior and financial out- or under performance compared to industry average as examined with Model 3 does not shows a significant coefficient for neither exploratory nor exploitative innovation. H3a and H3b are not supported.
The results of the robustness check with respect to supporting or not supporting hypotheses do not differ from the regression analysis without the dummy variable non financial sector. The industry the manager operates in is not of significant influence for the degree of a business unit’s exploratory innovation. The negatively significant coefficient of the non financial sector for exploitative innovation are indicative of some influences caused by the industry the manager operates in.
