Page 1 of 83 UNIVERSITY OF AMSTERDAM
EXECUTIVE PROGRAMME IN MANAGEMENT STUDIES STRATEGY TRACK
Entrepreneurial strategy,
industry dynamism &
corporate performance
A study on effectuation and causation in
corporations
Kirsten Wong 0502464 30 June 2016 Version: Final
Page 2 of 83 Statement of Originality
This document is written by Student Kirsten Wong who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Signature
Kirsten Wong
Page 3 of 83
Abstract
Managers are faced with the continuous challenge to maintain and improve corporate performance. At the same time, corporate entrepreneurship is important for existing firms to rejuvenate, revitalize, grow and expand a firm’s capabilities. Different entrepreneurial theories exist, yet there is no prescription on what is the superior approach.
Two prominent theories of entrepreneurship are effectuation and causation. The difference in underlying logic between the two entrepreneurial strategies is a focus on prediction or control, which can coexist to varying degrees. Effectuation is proposed as being more suitable to
dynamic, nonlinear and ecological environments. Causation is proposed as more suitable to static and linear environments.
This thesis investigates the relationship between entrepreneurial strategy and firm performance. Whilst entrepreneurial strategy can be selected by the corporation, the dynamics of an industry are largely driven externally. Acknowledging that different dynamic environments might enhance or decrease the success of any entrepreneurial strategy, the moderating effect of industry
dynamism on the relationship between entrepreneurial strategy and performance is also tested. Content analysis was applied to annual report text to measure entrepreneurial strategy in a sample of large UK companies. Performance was measured as financial performance for EBITDA
margin, Return on Assets, and Return on Capital Employed; as both average and change in absolute percentage points across 2009 to 2014. Industry dynamism, being the rate of unpredicted change, was measured as a composite variable taking into account industry-specific levels of employment, revenue, innovation and level of competition over a 7 year period.
Although entrepreneurial strategy was found to exist to varying degrees in corporations, there was no statistical evidence found between its existence and performance. The moderating effect of industry dynamism was also found to be insignificant. The implication of these findings is that the effects of entrepreneurial strategy remain elusive beyond performance.
Page 4 of 83
Table of Contents
1 Introduction ... 7
1.1 Theories of entrepreneurship ... 7
1.2 The impact of entrepreneurial strategy ... 8
1.3 Industry dynamism ... 9
1.4 Research question ... 10
1.5 Approach and structure of this study ... 11
2 Literature Review ... 12
2.1 Corporate Entrepreneurship ... 12
2.2 Effectuation and causation ... 13
2.3 Prediction and control ... 15
2.4 Hypothesis development... 18
2.4.1 The moderating role of industry dynamism ... 21
2.4.2 Conceptual model with hypotheses ... 25
3 Method ... 25 3.1 Sample ... 26 3.2 Variables ... 28 3.2.1 Independent variables ... 28 3.2.2 Moderator variable ... 30 3.2.3 Dependent variables ... 31 3.2.4 Control variables ... 32
3.2.5 Other performance variables considered but excluded ... 33
3.3 Content analysis and word coding approach ... 34
3.3.1 Word coding method 1: Word coding of full annual reports... 35
3.3.2 Word coding method 2: Word coding of only strategic sections of reports ... 38
3.3.3 Benefits and limitations of method 2 ... 42
3.4 Industry dynamism ... 44 3.5 Measuring performance ... 45 3.6 Control variables ... 46 4 Results ... 47 4.1 Combined dataset ... 47 4.2 Method 1 results ... 47 4.3 Method 2 results ... 55
5 Discussion and Conclusion ... 63
5.1 Conclusion ... 63 5.2 Theoretical implications ... 64 5.3 Practical implications ... 66 5.4 Limitations ... 67 5.5 Future research ... 70 5.6 Final words ... 72
Page 5 of 83
6 Appendix ... 74
6.1 Industry dynamism index ... 74
6.2 Words used for predictive and control strategy ... 76
6.2.1 Predictive strategy word fields. ... 76
6.2.2 Control strategy word fields. ... 78
6.3 Descriptive statistics ... 79
Page 6 of 83
Table of figures
Figure 1: Conceptual model ... 10
Figure 2 Emphasis on prediction and control in strategy ... 17
Figure 3: Conceptual model with hypothesis ... 25
Figure 4 Average EBITDA margin 2009-14 histogram ... 46
Figure 5 Scatterplot of control and predictive word count method 1 ... 49
Figure 6 Scatterplot of control word count method 1 and industry dynamism ... 50
Figure 7 Scatterplot of predictive word count method 1 and industry dynamism ... 51
Figure 8 Scatterplot of predictive and control word count method 2 ... 57
Figure 9 Scatterplot of control word count method 2 and industry dynamism ... 58
Figure 10 Scatterplot of predictive word count method 2 and industry dynamism ... 59
Table of tables Table 1 Dimensions of causation and effectuation ... 14
Table 2 Word coding method 2 - strategic and non-strategic sections ... 42
Table 3 Descriptive statistics and correlations for method 1 ... 48
Table 4 Regression results for method 1 word coding of entire annual report ... 54
Table 5 Descriptive statistics and correlations method 2 ... 55
Table 6 Regression results from method 2 word count of only strategic sections of annual reports ... 62
Table 7 Industry dynamism calculation ... 74
Table 8 Descriptive statistics for predictive strategy word fields ... 76
Table 9 Descriptive statistics for control strategy word fields ... 78
Page 7 of 83
1
Introduction
This chapter introduces the topic of corporate entrepreneurship and the entrepreneurial theories of effectuation and causation. The role of prediction and control in entrepreneurial strategy is then described. The relationship between entrepreneurial strategy and corporate performance is introduced, before leading into the research question that was addressed. An overview of the research approach is also included to provide context to the subsequent chapters.
1.1
Theories of entrepreneurship
Entrepreneurship is still a field of emerging theoretical perspectives, encompassing numerous theories of entrepreneurial approaches. Broadly, it includes opportunity recognition and creation, innovation, action and new business or markets. As an approach, mindset, technique, process and or strategy, entrepreneurship and being “entrepreneurial” is relevant at both the individual and firm level today. (Fisher 2012)
Corporate entrepreneurship is important for existing firms in order to rejuvenate, revitalize, grow and expand a firm’s capabilities (Kuratko and Audretsch, 2009). Seizing opportunities through corporate entrepreneurship can be a vital source of competitive advantage and growth for firms (Ireland, Covin and Kuratko, 2009). Whilst corporate entrepreneurship is described across multiple actions and behaviours of a firm leading towards an outcome, different approaches can be applied.
Sarasvathy (2001) proposed two theories of entrepreneurship; causation and effectuation. Both explain different entrepreneurial approaches and thought processes. The causation approach sees
Page 8 of 83
effectuation approach sees “a set of means as given and focus on selecting between possible
effects that can be created with that set of means” (p. 245, Sarasvathy, 2001).
The difference in underlying logic between the two approaches is the focus on prediction or control. Prediction is at the forefront of the causation process; “To the extent we can predict
future, we can control it”. Whereas focus on control or creation is at the forefront of the effectuation process; “To the extent we can control future, we do not need to predict it”. (p. 251, Sarasvathy 2001).
Although causation and effectuation are proposed as contrasting approaches, the role of prediction and control in strategizing has been further elaborated as independent options. An entrepreneurial strategy can have high or low degrees of prediction and control at any time, suggesting that entrepreneurial strategy can be a combination of attributes from effectuation and causation (Wiltbank, Dew, Read, and Sarasvathy, 2006; Reeves, Love and Tillmanns, 2012).
1.2
The impact of entrepreneurial strategy
Entrepreneurial strategies explain different approaches and logic that may lead to difference decisions and actions being taken. So what is the impact of these actions?
Studies have explored the traits of effectuation and causation, including further conceptual research and reviews (Fisher 2012; Perry et al 2012), as well as validation studies observing effectuation in practice (e.g. Read, Song and Smit, 2009; Dew, Read, Sarasvathy, and Wiltbank 2009; Chandler, DeTienne, McKelvie, and Mumford, 2011). These studies have found that there are different decision-making processes and actions attributed to effectuation and causation.
Page 9 of 83
Research has gone further to test the relationship between entrepreneurial strategy and performance. The outcomes of these studies have been mixed, with both significant and non-significant evidence of a link between entrepreneurial strategy and performance. Studies have included research measuring the effect to performance of start-ups, such as effectuation and performance in angel investing (Wiltbank, Read, Dew, and Sarasvathy, 2009), as well as the effect on performance in specific application areas such as R&D project performance (Brettel, Mauer, Engelen and Küpper, 2012). Effectuation has also been measured at the firm level through gauging “corporate orientation” of managers and employees (Werhahn, Mauer, Flatten, and Brettel., 2015), as well as through testing the relationship between emphasis on prediction and emphasis on control on company performance (Tillmanns and Mauer, 2012).
1.3
Industry dynamism
The rate of unpredictable change in industry and environment is outside the control of the firm. Yet does it play a role in entrepreneurial strategy and firm performance? Sarasvathy (2001) proposed effectuation as suited to dynamic, nonlinear and ecological environments. Studies have found positive and significant correlation between industry dynamism, strategic variety and new venture growth (Larrañeta, Zahra, González, 2014), as well as a positive moderating effect on the relationship between entrepreneurial orientation (EO) and performance (Rauch, Wiklund, Lumpkin, and Frese, 2009), and a positive moderating effect on the relationship between “transformational leadership” (motivating followers by appealing to their ideals) and new venture performance (Ensley, Pearce, and Hmieleski, 2006).
Industry dynamism has also been tested as a moderator between entrepreneur experience and the degree to which effectuation is used – as perceived dynamism increased, entrepreneurs were
Page 10 of 83
more inclined to use effectuation (Harms and Schiele 2012) – and also found to have three-way interaction was found with EO, access to capital and environmental dynamism in the relationship with small business performance (Wiklund and Shepherd 2005).
1.4
Research question
Managers are faced with the continuous challenge to improve corporate performance. At the same time, corporate entrepreneurship is important to rejuvenate, revitalize, grow and expand a firm’s capabilities. This paper investigates the effect of the entrepreneurial strategy - effectuation and causation - on firm performance.
Whilst entrepreneurial strategy can be selected by the corporation, the level of industry dynamism is driven by industry characteristics as a whole. This paper seeks to understand the moderating effect of industry dynamism on the relationship between entrepreneurial strategy and performance.
Effectuation in the corporate context: What is the relationship between entrepreneurial strategy of a corporation and its performance, and the moderating effect of industry dynamism on this relationship?
Page 11 of 83
1.5
Approach and structure of this study
To measure entrepreneurial strategy, content analysis was applied to the text of annual reports of a sample of 212 large UK firms. Word fields associated with prediction or control were used to measure the emphasis on entrepreneurial strategy for each company. Two methods for word coding were used; method 1 included coding of the entire annual reports, method 2 included only the strategic sections of the reports.
The base year that strategy was measured was 2009 and performance was measured across 2009-14. Performance was measured through financial ratios forming six dependent variables to provide a holistic view of the different views of corporate performance. The dependent variables were EBITDA margin average 2009-14, EBITDA margin absolute increase/decrease 2009-14, Return on Assets (ROA) average 2009-14, ROA absolute increase/decrease 2009-14, Return on Capital Employed (ROCE) average 2009-14, ROCE absolute increase/decrease 2009-14. Industry dynamism was measured across 90 different industries based on a calculation capturing the rate of unpredictable change across industry revenue, number of enterprises, employment, and R&D expenditure over time. Hierarchical regressions were used to test the hypotheses, including moderation regressions.
The literature review in chapter 2 describes the entrepreneurial theories in focus and existing studies applying and testing these theories. It concludes with formulation of the hypotheses. Chapter 3 then describes the overall method that was applied, including the sample and definition of variables. The detailed method for calculating each variable and combining the dataset is then described. The results including descriptive statistics and testing of hypotheses are contained in chapter 4.The final chapter 5 provides the conclusion and discussion, including the theoretical
Page 12 of 83
and practical implications from this study. Additionally, the limitations and suggested areas for future research are proposed.
2
Literature Review
The field of entrepreneurship contains different theories and contexts for entrepreneurship. This research addresses the context of entrepreneurship in corporations, and the theories of effectuation and causation. Within effectuation and causation theory, a focal point of difference in each approach is the emphasis on control versus prediction. The relationship between entrepreneurial strategy and firm performance, as well as the moderating effect of industry dynamism is also discussed in the following sections.
2.1
Corporate Entrepreneurship
Applying entrepreneurship as firm strategy; CE refers to firms whose strategic intent is to purposely and continuously aim to create and seize entrepreneurial opportunities for their growth and competitive advantage (Ireland et al, 2009). Applicable across an organization to decision making, processes and actions, corporate entrepreneurship (CE) is “vision-directed,
organization-wide reliance on entrepreneurial behaviour that purposefully and continuously rejuvenates the organization and shapes the scope of its operations through the recognition and exploitation of entrepreneurial opportunity.” (p. 21, Ireland et al, 2009)
CE can include the approaches of corporate venturing as well as strategic entrepreneurship. Corporate venturing can involve internal, external or cooperative venturing to add or create new businesses. Strategic entrepreneurship refers strategic renewal, sustained regeneration, domain regeneration, organizational rejuvenation and business model reconstruction. (Kuratko et al,
Page 13 of 83
2009; Kuratko, 2010; Dess, Ireland, Zahra, Floyd, Janney, and Lane, 2005). This action and creation orientated view of entrepreneurship can result in the creation of new ventures, new products and new markets (Dew et al 2008). For the purpose of this paper, CE is defined as the action, by corporations, of identifying opportunities and taking actions with the purpose to create new products, markets, ventures or business models.
Whilst researchers have become more specific and detailed to describe and define the purposes and concepts for CE, there is no agreement on a superior approach. Two prominent theories of entrepreneurship are effectuation and causation, described in the following sections. These entrepreneurial strategies are applied in the context of CE in established firms as the focus of this study.
2.2
Effectuation and causation
Effectuation and causation are two theories of entrepreneurship that describe different thought processes, attitudes, goals and perception to risk (Sarasvathy, 2001). There are five contrasting attributes to define the diverging approaches of causal versus effectual entrepreneurship, including view or predictability of the future, goal or means driven basis for action, calculated risk or affordable loss, competition versus alliances and partnerships, and avoiding or exploiting unexpected and unplanned events. These differences are summarised in Table 1. Causation is proposed as suited to static, linear and independent environments, whereas effectuation has the assumption of dynamic, nonlinear and ecological environments; in settings of uncertainty, effectuation is proposed as most relevant.
Page 14 of 83
Table 1 Dimensions of causation and effectuation
Issue Causation Effectuation
View of the future
Predictive. Causal logic frames the
future as a continuation of the past; accurate prediction is both necessary and useful.
Creative. Effectual logic frames the
future as shaped (at least partially) by wilful agents). Prediction is therefore neither easy nor useful.
Reason for taking action
Goal-orientated. Goals, even when
constrained by limited means, determine sub-goals. Goals
determine actions, including which individuals to bring on board.
Means orientated. Goals emerge
by imagining courses of action based on given means. Who comes on board determines what can be and needs to be done. And not vice versa.
Predisposition toward risk and resources
Expected return. Focus is on upside
potential; new venture creation is viewed as pursuing the
(risk-adjusted) maximum opportunity. The required resources are raised to pursue the opportunity.
Affordable loss. Focus is on
limiting downside potential and pursuing adequately satisfactory opportunities without investing more resources than stakeholders can afford to lose.
Attitude toward outsiders
Competitive analysis. Competitive
attitude to outsides. Relationships are driven by competitive analyses and the desire to limit dilution of ownership as far as possible.
Partnerships. Focus is on
partnerships to create new markets. Relationships, particularly equity partnerships drive the shape and trajectory of the new venture.
Attitude to unexpected contingencies
Avoiding. Contingencies are seen as
obstacles to be avoided.
Accurate predictions and careful planning are of importance.
Leveraging. Contingencies are seen
as opportunities to be leveraged. Creativity, agility and flexibility are important.
Source: adapted from Sarasvathy (2001), Dew et al (2009) and Perry et al (2012)
Whilst causation and effectuation are often shown as alternative approaches (even in Table 1), they are actually both part of the human reasoning process and can occur simultaneously (Fisher, 2009). The view of prediction or control as opposing forces is often contrasted when comparing causal and effectual approaches. In causation, one seeks to predict and minimize the unknowns or unknowables, whilst in effectuation, one seeks to exploit and take advantage of what cannot be predicted and to aim to shape and control the future. However prediction and control are not strict alternatives, a firm can choose to include both prediction and control to degrees of high or low
Page 15 of 83
importance. The degree of prediction and control in strategy has been the focus of further research within entrepreneurial strategy, outlined in the following section.
2.3
Prediction and control
The difference in underlying logic between effectuation and causation is the focus on prediction or control. Prediction is at the forefront of the causation process; “To the extent we can predict
future, we can control it”. This contrasts to focus on control or creation as the forefront of the effectuation process; “To the extent we can control future, we do not need to predict it”. (p. 251, Sarasvathy 2001).
For any situation, these different logics pose different thought processes and reasoning: Does the entrepreneur or firm seek to predict the future and mitigate uncertainty, or do they instead focus on what is at hand and controllable to reposition themselves and the firm in the event of change and uncertainty? The prediction approach seeks to accurately and confidently forecast, calculate, and estimate future performance, environmental factors and indicators. On the other hand, the control approach seeks to shape, change and influence the future. In the control approach the entrepreneur or entrepreneurial firm asks themselves who they are, what they know and whom they know; what is most important are the resources, capabilities and networks at hand that can be used to shape the future. (Sarasvathy, 2001; Wiltbank et al, 2009)
Wiltbank et al (2006) highlighted the characteristics of prediction and control as independent within strategy, in the example of business strategic planning. The previous planning schools of thought; “try harder to predict better” or “move faster to adapt better” (Wiltbank et al 2006 p. 983), were viewed as an axis of choice for strategy in terms of the importance of prediction and control. Control was defined as meaning is the degree a firm can construct the future assuming an
Page 16 of 83
endogenous environment. High degrees of controllability would be environments where the firm can shape the development, or even creation (of the “unknowable”) of markets over time. That is, the firm can shape the future through the choice of their actions and their effect. Low controllability would be exampled by mature markets where demand is defined and unchangeable by the actions of a single firm.
Through the reconceptualization of prediction and control as independent options; this created four strategic approaches that a firm strategy can take, as shown in Figure 2. Wiltbank et al (2006) proposed the following classification framework of prediction and control strategies:
• The classical “planning” school of strategy has high emphasis on prediction and low emphasis on control, and aims to “try harder to predict and position more accurately”. It is the oldest school in strategic management and predicts that firms, when faced with increasing uncertainty, will try harder to predict better.
• The “adaptive” school of strategy has low emphasis on prediction, and low emphasis on control, and aims to “move faster to adapt to a rapidly changing environment”. When faced with uncertainty, firms neither aim to predict or control, they should experiment in agile and incremental changes with regular cycles of learning and feedback loops.
• The “visionary” field of strategy has high emphasis on prediction and high emphasis on control and aims to “transform current means into co-created goals with others who
commit to building a possible future”. A firm takes the role to a future market and demand, through developing the vision of the future and then implementing the plan to get there.
• The “transformative” school of strategy has low emphasis on prediction and high emphasis on control and aims to “transform current means into co-created goals with
Page 17 of 83 others who commit to building a possible future”. When faced with uncertainty, the firm does not attempt to predict a clear vision for future and create it, it rather engages in experimental actions (often with co-creation e.g. customers) and emphasising what can be controlled over what can be created – attributed characterised to the effectuation entrepreneurial approach.
(Source for above quotes: p. 983, Wiltbank et al, 2006)
Figure 2 Emphasis on prediction and control in strategy
Source: Wiltbank et al (2006)
A similar framework was also proposed by Reeves et al (2012). Similar to Wiltbank et al (2006), they found that a company’s strategy can be selected along the role of prediction; how far ahead
Page 18 of 83
and how accurately can you forecast the future, and malleability; the degree of control over market, competitors and environment. Consistent with Wiltbank et al (2006), a firm strategy was positioned as able to have high or low emphasis on prediction and control (/malleability).
The emphasis on prediction and control by both Wiltbank et al (2006) and Reeves et al (2012) draw a parallel to the causation and effectuation approaches. The classical planning school aligns to the causal approach; the transformative school of strategy aligns to the effectual approach. By conceptualizing the options for focus on prediction and control as independent axis, this provides guidance for effectuation and causation research that prediction and control can exist in parallel.
2.4
Hypothesis development
When Sarasvathy (2001) first proposed the concepts of effectuation and causation, neither approach was positioned as ultimately superior. In the context of new ventures and the venture creation process, effectuation was proposed as more likely to be used as a way to gauge for success or failure early on, however effectuation itself was not prescribed as a winning formula.
The different patterns of decision making logic - causal and effectual – were observed by Dew et al (2009) at the individual level across experts and novice entrepreneurs (MBA students). In the study on entrepreneurial approach, the spectrum of differentiating characteristics of the effectual and causal approach was measured using protocol analysis. This included measurement of the use of non-predictive as opposed to predictive control, means-driven as opposed to goal-driven action, affordable loss as opposed to expected return, and partnerships as opposed to competitive analysis. Dew et al (2009) found support that experts were significantly more likely to take an effectual approach, using non-predictive and means driven actions, quantifying opportunities based on how much was an affordable loss rather than looking to estimate expected return...
Page 19 of 83
Whilst the study demonstrated the different decision-making approaches, it did not go so far as to measure impact on performance or the outcome of the decisions.
The use of causation and effectuation approaches in new venture creation was also tested by Chandler et al (2009). The study developed measures on which to measure the constructs of effectuation and causation, and tested this in the context of start-ups, to understand entrepreneurial decision making in uncertainty. Four measurement areas were tested, with focus on prediction and control being the first area – focus on short-term experiments to identify opportunities versus trying to predict the future. They found to support pre-commitments and alliances as a shared component to both effectuation and causation – supporting the notion that effectuation and causation are not opposites. However it is important to note in the study, the definition of the indicator was mixed with the measure of prediction and control of an uncertain future.
The effect of effectuation and causation on new venture performance was also investigated through meta-analysis by Read, Song and Smit (2009). Four characteristics of effectuation were measured; means, partnership, affordable loss and leveraging contingency. Evidence supporting a significant and positive relationship between an effectual approach to strategy making and new venture performance was found when measuring means, partnership and leveraging contingency as features of effectuation. However affordable loss was found as not significantly related to new venture performance.
Prediction and control under uncertainty was also tested by Wiltbank et al (2009). They studied the use of strategies emphasising prediction or control by angel investors facing uncertainty, and the impact on size of investment, investment failures and “home runs” earning a large investment return. They found that investors using predictive strategy were significantly more likely to make
Page 20 of 83
a larger initial investment; supporting the notion that predictive strategy takes an effect as given, and focuses on the means to create the effect. However the similar hypothesis for control strategy – that investors with a control strategy would make smaller investments – was unsupported. Although they found that control strategies resulted in fewer investment failures, no statistical evidence was found to support a relationship of prediction to failures or home run investment outcomes.
Focusing on prediction and control, Tillmanns and Mauer (2012) also tested the relationship between emphasis on prediction and emphasis on control on company performance. Using both content analysis and a survey approach, they tested the relationship between emphasis on control and performance and emphasis on prediction and performance. No significant results were found to support the hypotheses that prediction will positively influence performance or that control will positively influence performance. On the other hand, Bierwerth, Schwens, Isidor and Kabst (2015) found innovation, strategic renewal and corporate venturing were significantly and positively related to firm performance, through meta-analysis on entrepreneurship and corporate performance. CE was defined with CE including strategic renewal, innovation, and corporate venturing, with the purpose to renew, revitalize and or react to changes in environment. Whilst the study did not look into the entrepreneurial approach of corporations e.g. how innovation initiated or what was the approach to strategic renewal, it did reveal that CE – in broad shapes and forms – has a positive link to performance. There was also the observation that there was a need for more common definitions and categories in which to measure CE as well as performance.
The mixed results from these studies hint that there is further to be explored and tested in understanding the effect of entrepreneurial strategy on corporate performance. Although
Page 21 of 83
hypotheses were not supported in Tillmanns et al (2012), the many studies linking effectuation to new venture performance (Read et al 2009), as well as corporate entrepreneurship to performance (Bierwerth 2014) suggest the relationship between effectuation or causation and performance remains to be further explored. As entrepreneurial strategies, both effectuation and causation are predicted to have a positive impact on performance. Acknowledging that they are independent options, and can both exist to varying degrees in corporations, this leads to the following hypotheses:
Hypothesis 1: Effectuation strategy (emphasis on control) will have a positive effect on corporate
firm performance
Hypothesis 2: Causation strategy (emphasis on prediction) will have a positive effect on
corporate firm performance
2.4.1 The moderating role of industry dynamism
Dynamic environments are characterized by uncertainty, unpredictable and rapid change, where decision making is challenged with the difficulty to predict the outcomes and consequences of chosen actions (Ensley et al 2006). Industry dynamism refers to the magnitude of change across industry and market conditions. Heterogeneity and complexity increase when industry dynamism increases, and firms are faced with an increased number of options and challenges, as market conditions can change and hyper-competition, new entrants and intensified hostility can arise. (Larrañeta et al, 2014).
Environmental dynamism refers to dynamism that encompasses a wider influence of changes on the surrounding environment of the firm, such as technical and general environment. (Harms et al, 2012). In the studies of entrepreneurship and strategy reviewed, the definitions and measures
Page 22 of 83
for both environmental dynamism and industry dynamism were found as comparable, and are therefore discussed within the topic of industry dynamism as described below.
Sarasvathy (2001) proposed effectuation as suited to dynamic, nonlinear and ecological environments. Although industry dynamism is a characteristic proposed to be relevant to effectuation and emphasis on control, there are few studies that have looked directly at the link between dynamism, entrepreneurial strategy and performance. Categorization of industry environment to different dynamics for emphasis on predictive or control strategy was proposed by Reeves et al (2012), with dynamic industries e.g. internet software and services, electronic equipment, instruments and components, proposed as more suited to a non-predictive control strategy (effectuation). Chandler et al (2009) also found uncertainty to be negatively correlated with causation and positively correlated with experimentation, in support of the notion that effectuation is more suited to dynamic environments. However these studies did not go so far as to test the relationship to performance.
Further studies have also tested the moderating role of industry dynamism on the relationship between entrepreneurial strategy and performance. In a study of entry mode selection by gazelles for international new venture creation, perceived environmental dynamism was tested as a moderator between entrepreneur experience and the degree to which effectuation or causation was used (Harms et al, 2012). Dynamism was defined as the degree and magnitude of expected changes in the market as a proxy for the degree of uncertainty a new market was perceived to exhibit by entrepreneurs. The measurement of dynamism included both dynamism of the technical environment and dynamism of the general environment, gauged through a survey to CEOs. The more dynamic an international market was perceived by the entrepreneur, the more likely effectuation was to be used, in support of Sarasvathy’s (2001) original proposal for
Page 23 of 83
effectuation. In the study causation and effectuation were also found not to be diametrically opposed concepts as a positive correlation was found. Furthermore, there was no evidence to support that in more dynamic environments entrepreneurs were less likely to use causation.
Further studies have investigated the moderating effect of industry dynamism on performance. In a meta-analysis study of the effect of entrepreneurial orientation (EO)1 and performance, the moderating effect of industry dynamism was found to be positively significant between rapidly changing industries, indicating that high-tech industries have a greater positive benefit of EO on performance (Rauch et al,, 2009). Industries considered dynamic were categorised in terms of technology and/or customer demand, so called “high-tech” industries (including computer software and hardware, biotechnology, electric and electronic products, pharmaceuticals and new energy), versus “non-high-tech” industries.
In a study of entrepreneur leadership behaviour and new venture performance, industry dynamism was also found to have a positive moderating effect on the relationship between “transformational leadership” (motivating followers by appealing to their ideals) and new venture performance (Ensley et al, 2006). “Transaction leadership” (motivating behaviour of followers through exchange processes e.g. rewards and punishments) was found to be more effective when environmental dynamism was low, whereas transformational leadership style was found to be more effective when environmental dynamism was high.
Industry dynamism has also been found to have a positive moderating effect on the impact of strategic variety on sales growth of new ventures. In a study by Larrañeta et al (2014), strategic variety was defined as implementing multiple different competitive actions simultaneously,
1
With EO defined as the strategy-making processes used to support a firms decision-making, vision and competitive advantage and including the dimensions of innovativeness, risk taking and proactivenes as well as sometimes the additional dimensions competitive aggressiveness and autonomy. (Rauch, Wiklund, Lumpkin, and Frese, 2009)
Page 24 of 83
compared to few similar competitive actions. Industry dynamism was measured as a construct capturing uncertainty, complexity, and velocity of the industry environment. Strategic variety was more positively associated with new venture growth in industries with higher levels of dynamism. In more dynamic industries, strategic variety was found to be more highly and positively associated with independent venture growth than corporate venture growth.
The role of prediction and control can be likened to the role of market expertise and use of resources at hand in the study of strategic variety by Larrañeta et al (2014). On a sample of large corporations, Tillmanns et al (2012) tested the moderating role of industry dynamism on the relationship between emphasis on prediction and control on performance. In more dynamic environments, emphasis on control was found to positively influence firm performance. When using a survey-based approach, emphasis on prediction was found to negatively influence performance in more dynamic environments. However there were inconclusive results found when testing the same hypothesis with a content analysis approach.
These studies highlight that in certain conditions, industry dynamism has been found to have a moderating effect of on the relationship between entrepreneurial strategy and performance. What is observed from the studies is that the moderating effect can be positive or negative depending on the type of strategy employed. Building on the initial assumption by Sarasvathy (2001), that the context of effectuation is a dynamic, nonlinear and ecological environment, and that causation is more useful in static linear and independent environment, the following hypotheses are proposed:
Hypothesis 3: The positive relationship between control strategy and firm performance is
Page 25 of 83 Hypothesis 4: The positive relationship between predictive strategy and firm performance is
negatively moderated by industry dynamism
2.4.2 Conceptual model with hypotheses
Figure 3: Conceptual model with hypothesis
3
Method
The following section outlines the research design of this study. First, the sample is described which included large UK companies as well as broader industry data. Then the definition and measurement of variables is explained. Next, the analysis method is detailed out, which included two different approaches to content analysis. The first word coding method is described, including the approach for testing reliability in the word fields for prediction and control. A second word coding method is then described, where only the strategic sections of annual reports were included. Comparison between the two word coding methods is also discussed. Then, the method for calculating industry dynamism as a composite measure of different industry variables is described. Lastly, the approach to measure the performance and the control variables is outlined, including checking of outliers.
Page 26 of 83
3.1
Sample
This study focused on large UK companies (>250 employees) that were publicly listed from 2009-2014. The research question required that company information be collected at the firm level and is focused on entrepreneurial strategy in the corporate context, meaning established organizations at least 5 years old.
The sample required companies where financial and descriptive data was available, as well as annual reports in PDF form as the measurement of entrepreneurial strategy was conducted through content analysis of the reports. UK companies were selected for this study as the company information and annual reports are originally published in English, reducing the probability of translation errors.
A five-year time period was selected for this study, as a strategy takes at least 2 years to come into effect, and a strategic planning cycle is typically 5 years2. Measuring the company’s performance over 5 years accounted for the time the entrepreneurial strategy can take to come into full effect. Starting with 2009 as the base year, the variables were measured over the period until 2014.
The AMADEUS database was selected as the data source for company financial and descriptive data. “AMADEUS is a pan-European financial database… Up to 10 years of detailed information
is provided including balance sheet items, profit and loss account items and ratios, address and telephone, senior managers, auditors, number of employees, and, when available, a trade description in the local language and English.” (Wharton research Data Services website) From the entire database of 43 countries, the subset of UK large and mid-sized publically traded
2
Page 27 of 83
companies was extracted. Data fields in the extracted data included company name, country, legal status, date of incorporation, national industry code (primary), number of employees, year established, EBITDA margin, Return on Assets, Return on Capital Employed.
From the AMADEUS database, data for 14,213 UK companies was extracted. From this total, 13719 companies had an active status, 999 were publicly listed companies and only 906 had a date of incorporation prior to 2009. A random number generator was applied to this filtered list from the AMADEUS data set, and based on this, the annual reports of 212 companies were manually extracted from their respective company websites in PDF format. These 212 companies formed the sample for this study. There were actually 280 companies were searched, however the reports were no longer publically available for 68 companies, leaving 212 in the sample.
The data collected to measure industry dynamism was obtained from the OECD Structural Business Statistics Database and OECD Business Enterprise R&D Expenditure database at ISIC Revision 4. Due to the change from ISIC revision 3 to revision 4 in both databases, data was only available for the 7 years 2007-2013 as the categories could not be reconciled between ISIC revisions. More recent data was not available. Industry data for Turnover, Employment, and Number of Enterprises was extracted for the UK, according to 2 digit ISIC 4 code. From the R&D database, total R&D spends per industry was extracted. The data for all variables was the annual figures across the years 2007-2013
All industry data was identifiable at the 2-digit ISIC code level. The ISIC Revision 4 codes could be matched to the AMADEUS company data to the 2-digit NACE industry codes (from the AMADEUS company data) which are compatible with the international codes. In the combined OECD industry data set, 75 industry categories were available (at 2-digit code). The list of industries started at “05: Mining of coal and lignite” and went to “82: Office administrative,
Page 28 of 83
office support and other business support activities”.
3.2
Variables
The selection and definition of the independent, dependent and control variables is explained in the following section.
3.2.1 Independent variables
The hypothesis called for measuring the entrepreneurial strategy of firms; effectuation or causation). This was measured through the emphasis on prediction or control, which is the difference in underlying logic between the two entrepreneurial approaches.
Measurement of a firm’s strategic emphasis on prediction or control was measured using word fields based on the work of Tillmanns et al (2012), whom developed word field measurements for prediction and control strategy. The word fields for each independent variable were as follows (including variations of words):
3.2.1.1 Predictive strategy: count of predictive strategy word fields as percentage of words in annual report
Word fields relating to predictive strategy (as proxy for predictive strategy & causation):
• Aim, aimed, aims, anticipate, anticipated, anticipates, anticipation, blueprint, envision, envisioned, envisions, expectation, expectations, extrapolate, extrapolated, extrapolates, forecast, forecasting, forecasts, foresee, foreseeable, future, futures, goal, goals, long-term, objective, objectives, outlook, outlooks, plan, planned, planning, plans, predict, predictability, predictable, prediction, predictions, predicts, prognosis, projected,
Page 29 of 83
projection, projections, prospect, prospects, roadmap, roadmaps, target, targeted, targets, trend, vision, visionary, visions
3.2.1.2 Control strategy: count of control strategy word fields as percentage of words in annual report
Word fields relating to control strategy (as proxy for control strategy & effectuation):
• Collaborate, collaborated, collaborates, commit, commits, committed, control, controlled, controller, controls, cooperate, cooperated, cooperates, create, created, creates, disrupt, disrupted, disruptive, empower, empowered, empowers, explore, explored, explorer, explores, influence, influenced, partner, partnered, partners, persuade, reshape, reshaped, revolutionary, revolutionise, revolutionised, revolutionized, shape, shaped, unexplored, unseen, untested
From the original word list by Tillmanns et al (2012), the words market, marketplace, industry and sector were not included, as these words are typically used in annual reports to describe a company structure and divisions and markets, and not in the context of effectuation or causation.
Annual reports have been used in managerial studies to measure corporate characteristics through textual analysis (Young, 2014). Content analyses have been used in management research to study organizational texts such as CEO letters to shareholders, annual reports and mission statements, and provide a way to objectively measure corporate characteristics including strategy and management intent (Short, Broberg, Cogliser, and Brigham, 2009). The word fields for control and predictive strategy were applied to the 2009 annual reports of the sample of UK firms.
Page 30 of 83 3.2.2 Moderator variable
3.2.2.1 Industry dynamism
Dynamism is defined as the rate of unpredictable change within the industry that a firm operates in (Dess & Beard, 1984 as cited in Hmieleski and Ensley, 2007). This analysis required a measure that could be calculated across over 75 different industries, was relevant for the UK, and captured different aspects of industry change specific to each industry e.g. value of output, level of competition, to employment and also the level of innovation. Approaches to defining and measuring industry dynamism vary across studies and there is no commonly agreed approach. Rather than a survey based measure of industry dynamism, such as that of Larrañeta et al (2014) or Harms et al (2012), the measure for industry dynamism was based on the approach of Ensley et al 2006, p. 254, and Hmieleski et al (2007) (also adapted from the approaches by Sharfman and Dean (1991) and Dess and Beard (1984). Their calculation of industry dynamism took into account industry-specific level of employment, revenue, innovation and level of competition. This allowed for objective calculation and comparison across multiple industries, and could also be calculated over different time periods, as well as reproduced or updated with more recent data for future studies.
Industry dynamism, being the rate of unpredicted change, was measured by calculating the standard error from regression of each variable (industry revenue, number of enterprises, employment, R&D expenditure) over time with a constant. The standard errors were then divided by their respective mean, then z scores taken and instability of each measure added in order to develop a measure of industry dynamism. Detailed calculation of industry dynamism is explained later in the method section.
Page 31 of 83 3.2.3 Dependent variables
The dependent variables in this study measured firm performance. Existing studies measuring the impact of effectuation or causation on performance are largely focused on start-ups and new ventures, where the focus of performance is on growth and survival. Common measures of performance used in these studies are sales (absolute or growth), number of employees, return on investment and also the survival rate of new ventures. (Read et al 2009)
As this study focuses on large publically listed corporations, the measure of performance needed to be relevant to measure performance in large companies. Corporate performance includes improving efficiency and profitability, compared to the focus on top line growth used in studies of start-ups. How to define and measure corporate performance is a whole research topic in itself, and not the core focus of this study, so rather than measure a single aspect of performance, three financial performance areas were included, which are described further below. For each performance area, performance was measured in average, as well as absolute growth (i.e. how many percentage points increase or decrease) over the time period 2009 to 2014, resulting in six dependent variables. Average performance was taken as the simple average of the variable 2009-2014. Growth was measured as the absolute growth in percentage points 2009 to 2009-2014. The performance variables were:
1. Earnings before interest, tax, amortization and depreciation (EBITDA) margin = EBITDA / revenue
EBITDA margin measures company operating profitability as a percentage of total revenue. As the EBITDA excludes the effects of interest, tax, amortization and
Page 32 of 83
depreciation, earnings performance is focused solely on operations and more closely reflects a company’s cash flows (as e.g. amortization is not cash out).
Higher EBITDA margin shows positive performance as it reflects a higher operating income relative to operating expenses (so this implies a higher percentage of every dollar of sales contributes to earnings).
2. Return on Assets (ROA)
= Net income / Total assets
Calculated by dividing a company's annual earnings by its total assets, ROA indicates how profitable a company is relative to its total assets and how efficient management is at using its assets to generate earnings. ROA is displayed as a percentage.
Higher ROA indicates positive performance through a more profitable return on investment of assets.
3. Return on Capital Employed (ROCE)
= Earnings before interest and tax (EBIT) / capital employed
ROCE measures the efficiency through which capital is employed to generate
profitability. Capital employed is the sum of shareholder’s equity and debt liabilities. Higher ROCE indicates positive performance through more efficient use of capital.
For all variables, the percentage figures were already calculated in the AMADEUS data.
3.2.4 Control variables
Control variables were included to account for the impact of size (by number of employees), age of a firm and industry. The selection was based from the literature review of similar studies from
Page 33 of 83
the literature review, for example Ensley et al (2006), Hmieleski et al (2007), Tillmanns et al (2012), Larrañeta et al (2014).
1. Industry classification
2 digit industry code (from AMADEUS data)
2. Number of employees
Base year (2009) number of employees, taken from the AMADEUS dataset.
3. Year established
Year of incorporation for each firm. Data was taken from the AMADEUS dataset.
3.2.5 Other performance variables considered but excluded
Sales turnover was also considered as a potential variable for measuring performance. However for mature corporations, improved performance does not necessarily reflect a growth in sales. In large corporations, a strategy to improve performance can be based on margin improvement. Also, large companies might already have a leading and dominant market share, so increasing top line growth further might be limited.
EBIT margin was also excluded in favour of EBITDA margin, as EBITDA margin showed a more realistic reflection of underlying operational performance. The EBIT margin data showed many companies with significantly fluctuating figures year on year. When checking for outliers in EBIT figures by looking into annual reports, it was found that with the sources of highly fluctuating EBIT for many of these companies were exceptional events and large amortizations, when actually the underlying performance of the company had a fairly steady growth in sales and
Page 34 of 83
operating costs. For example, Intercontinental Hotels Group (IHG) PLC had 7366.7% EBIT growth from 2009 to 2014 due to one-off exceptional items in 2009.
Growth was initially measured as percentage increase in growth for each variable (with the 2009 figure as denominator). However when calculated as percentage this was misleading, as the magnitude of growth in performance was relative to the starting point in 2009, causing amplified view of performance when the starting point was low. For example, a company with 1% EBITDA margin in 2009 that increased to 3% EBITDA margin in 2014 would show 200% EBITDA margin growth 2009-14. If a similar company had initially 10% EBITDA and increased this to 12% (up 2% points), this would be only a 20% EBITDA margin growth. However, both companies were actually showing the same margin point improvement. For a company with 10% EBITDA, increasing this by 200% would be somewhat extraordinary. When measuring growth this way, the second company’s performance would be viewed as inferior growth performance in comparison to the first, which could lead to misinterpretation of performance. To remove the amplification caused by difference starting levels, the absolute growth in percentage points was used to measure growth for each performance variable.
3.3
Content analysis and word coding approach
This research applied content analysis using computer aided text analysis (CATA) to measure the emphasis on prediction or control strategy (independent variables) in companies, analysing the text of their annual reports. CATA for word field counts enhances reliability as it minimizes human error (Short et al 2009). Content analysis was selected for this study as a method to allow objective analysis of existing managerial text for the selected sample of UK companies. The benefit of this approach is that the study can be repeated on another sample using already existing
Page 35 of 83
data sources (annual reports), allowing one to apply the same procedures to replicate research, check result validity and further expand the research. For this study, as explained in the variables definition, the word fields for predictive and control strategies were based on the research of Tillmanns et al (2012).
Two approaches were used to measure prediction and control strategy through word coding. The first involved CATA of the entire annual reports. The second involved coding the reports into strategic and non-strategic sections, where only the strategic sections of the annual reports were coded to count the words for prediction or control respectively.
3.3.1 Word coding method 1: Word coding of full annual reports
This approach applied word coding analysis to the full text (excluding numbers) of the UK annual reports to count the instances of each word from the word fields for predictive and control strategy. The resulting measurement was the count of predictive and control words as a percentage of the total words in each annual report respectively.
Word coding
First, the PDF annual reports had to be converted to text files. For this, an online program called xPDF was used. An R script was used that was able to iteratively convert all files in one go. Then, based on the approach of Tillmanns et al (2012), word fields relevant to predictive or control strategy were counted through coding. From the list of forty nine words representing predictive or control approach (Tillmanns et al 2012), there were ninety eight words total when word variations were taken into account. A text mining function in R was used that collected all the annual report text files into one corpus variable. The text files stripped away all punctuation, numbers and white space and then converted all to lower case.
Page 36 of 83
An R script was used to count the total words of each report and count each instance of the prediction and control words. All total word counts of each word for each annual report were collected into a term document matrix excel file. The term document matrix (showing the count of each word for each report) was then checked for reliability to check the consistency of measurements.
Reliability
Cronbach's alpha is a measure of the internal consistency of items to measure scale reliability. For this study, Cronbach’s alpha provided insight into the construct of the words selected to measure emphasis on prediction or control in the sample of UK firms. Generally, the higher the score and closer to the maximum value of 1, the more reliable a scale is. However it is important to note that the Cronbach’s alpha is dependent not only on the magnitude of correlations among items, but also the number of items in the scale. As there was a large number of items (words) in the scale (total predictive words and total control words), the Cronbach's alpha is helpful to understand reliability of the word fields, but should be interpreted with a level of caution due to the high number of words in each word field list.
For the list of 55 predictive words, the Cronbach’s alpha was 0.841. Checking the correlation between each item and a scale score that excludes that item, 19 words had Corrected Item-Total Correlation score <0.2. When looking at the original data of word counts, these were words rarely found in the annual report, so that the average count of that word per report was zero or close to zero. Most of these words were the variations of words that were significant e.g. predict was significant but the variations (predictability, predictable) were not. The variable for count of predictive words was re-measured without the omitted words and there was only marginal difference. For the purpose of comparability with the second approach to annual reports and word
Page 37 of 83
coding, the regression results using the measurement of all words is included in the results section. The correlation between the full and the reduced word count for predictive words was 0.997, p<0.01, meaning there was marginal difference between using the full or reduced variable.
For the list of 43 control words, the Cronbach’s alpha was 0.685 indicating moderate reliability. Checking the correlation between each item and a scale score that excludes that item, 25 words had Corrected Item-Total Correlation score <0.2. When looking at the original data of word counts, these were words rarely found in the annual report, so that the average count of that word per report was zero or close to zero. Similar to what was seen in the predictive word list; most of these words were the variations of words that were significant. The variable was re-measured without the omitted words and there was only marginal difference once control words were calculated as % of total word count. For the purpose of comparability with the second approach to annual reports and word coding, the regression results using the measurement of all words for control are included in the results section. The correlation between the full and the reduced word count for control words was 0.995, p<0.01, meaning there was marginal difference between using the full or reduced variable.
A further check of the distribution of the word count matrix found several potential vulnerabilities that were checked in more detail. The company John Lewis Partnership used the word “partners” throughout their annual report when referring to the partnership partners, resulting in a very high count for this word. In Laing O’Rourke PLC it was found the word “explore” was part of their business unit name and also a project. Due to the repeat use of the word explore in a different context to that of strategy, the company was removed.
The single word “control” accounted for 40% of total control words. Two companies showed extremely high count of the word “control” and these were inspected. Checking the annual report
Page 38 of 83
of NATS PLC, it was revealed that they used the word control to refer to the air traffic control industry, and Rotork PLC had a division called Rotork Controls – both these companies were also removed from the sample due to the usage of the word control in another context.
The words “plan”, “plans” and “future” together made up around 50% of total predictor words. When checking in samples of the actual annual reports, the words “plan” or “plans” was used in companies as benefit pension plans and employee incentive plans. The high count of plan was distorted due to the double-meaning in the context of employee or pension plans and not in the context desired to represent predictive strategy, so these words were also eliminated from the word field list. The word future, also used frequently, was used in the context of talking about company outlook, so this word was left in the word count as it represented tendency for predictive strategy.
The resulting word count matrix was matched to the extracted company data from AMADEUS, so that each company had corresponding variables indicating the total count of words, count of predictive, count of control, employee, financial data and year established.
3.3.2 Word coding method 2: Word coding of only strategic sections of reports
A second method to measure the variables predictive and control in the sample of firms was devised after reflection on the limitations and challenges of method 1.
Limitations of word coding method 1
A challenge to content analysis is that there is no standard for a consistent report template in UK annual reports (Young et al 2014). UK annual reports contain a mix of strategic, financial and regulatory text, which can vary in depth and breadth. The amount (or absence) of commentary in an annual report can vary between firms to large degrees. Stripped to the bare minimum, an
Page 39 of 83
annual report can be the financial statements, legal and regulatory statements, and not much more than that. On the other hand, annual reports can be highly decorated and insightful documents, containing lengthy explanation and commentary on the strategy of the business and reflection of performance and strategy for the future.
A potential limitation to the method 1 approach was the inclusion of text from the non-strategic report sections as part of the word count. As the control and predict independent variables were measured as a percentage of total report word count, the measure could be amplified or weakened according to the number words in the non-strategic report sections.
Revised approach – word coding method 2
A revised approach was developed that would provide a more focused measure of entrepreneurial strategy could be applied to the sample of UK annual reports. In their study of construct validation using Computer Aided Text Analysis (CATA) including application, Short et al (2009) proposed that the text being analysed should be an appropriate and relevant setting to the purpose and focus on the content analysis. For this study, the research question focuses on entrepreneurial strategy in existing firms, so the most relevant text for analysis is where a narrative of strategy exists. Ideally, the company strategy would be the data source for content analysis, however this is not often publically available and often a slimmed down version of true corporate strategy, which is confidential.
The Corporate Financial Information Environment (CFIE) project was a 2-year project by the University of Lancaster that included the study of the narrative of UK annual reports3. The CFIE Wmatrx import web tool (“Wmatrix”) was developed to support text analysis of specifically UK
3
Page 40 of 83
annual reports, making the tool suitable to support this research. The Wmatrix tool was used in this study to convert the PDF documents to text files, divided into sections matching the contents page of each individual report. To do this, the tool detected the content page with sections, numbering and headings, then added headers as bookmarks to PDF, and extracted the text for each section. Additionally, the tool was used to analyse the extracted text per section based on user-defined word lists for prediction and control as outlined in the variable definitions.
As only one user defined word list was able to be included for each upload, the upload and text extraction process was repeated twice; once with the word list for predictive words and once with control (as). Only 178 of the 212 annual reports were able to be converted to text file and word counts extracted by the tool4. The result was two excel files showing each report divided into sections according to the contents page of each report; the word count for total words, as well as the result of the word count for predictive and control words per section. As the annual reports were the same for each upload, the word count results were merged to form a new data table for the annual report word counts.
Checking outliers
The company IQE PLC showed a word count 202% less than the R coding approach of method 1, indicating an error in the coding. Checking the histogram of the Wmatrix dataset, IQE PLC was an extreme outlier, both in absolute word count and also the histogram of the variables for predictive and control as percentage of total words. As the correct word count could not be verified IQE PLC was removed.
Defining strategic and non-strategic sections
4
