Entrepreneurial risk-taking beyond bounded rationality : risk factors, cognitive biases and strategies of new technology ventures

(1)

Entrepreneurial risk-taking beyond bounded rationality : risk

factors, cognitive biases and strategies of new technology

ventures

Citation for published version (APA):

Podoynitsyna, K. S. (2008). Entrepreneurial risk-taking beyond bounded rationality : risk factors, cognitive biases and strategies of new technology ventures. Technische Universiteit Eindhoven. https://doi.org/10.6100/IR635533

DOI:

10.6100/IR635533

Document status and date: Published: 01/01/2008

Document Version:

Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.

• The final author version and the galley proof are versions of the publication after peer review.

• The final published version features the final layout of the paper including the volume, issue and page numbers.

Link to publication

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain

• You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:

www.tue.nl/taverne Take down policy

If you believe that this document breaches copyright please contact us at: openaccess@tue.nl

providing details and we will investigate your claim.

(2)

Entrepreneurial risk-taking

beyond bounded rationality:

Risk factors, cognitive biases and strategies of

new technology ventures

(3)

CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EINDHOVEN

Podoynitsyna, Ksenia Sergeyevna

Entrepreneurial risk-taking beyond bounded rationality: risk factors, cognitive biases and strategies of new technology ventures / by Ksenia Sergeyevna Podoynitsyna. - Eindhoven : Technische Universiteit Eindhoven, 2008. – Proefschrift. -

ISBN 978-90-386-1280-5 NUR 801

Keywords: Entrepreneurship / New technology ventures / Success factors / Cognitive biases / Risk and uncertainty management strategies / Performance

(4)

Entrepreneurial risk-taking beyond bounded rationality:

Risk factors, cognitive biases and strategies of

new technology ventures

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven,

op gezag van de Rector Magnificus, prof.dr.ir. C.J. van Duijn, voor een commissie aangewezen door het College voor Promoties

in het openbaar te verdedigen op woensdag 11 juni 2008 om 16.00 uur

door

Ksenia Sergeyevna Podoynitsyna geboren te Moskou, Rusland

(5)

Dit proefschrift is goedgekeurd door de promotoren:

prof.dr.ir. M.C.D.P. Weggeman en

prof.dr. X.M. Song

Copromotor: dr. J.D. van der Bij

(6)

i

Acknowledgments

The author would like to thank the alphabet for the letters it kindly provided.

This year is very special for me since I happen to experience a double birth: that of my daughter and that of this thesis. Both of them can be seen as long-term projects characterized by high risk and uncertainty. Despite some similarities, one of the most important lessons I learned during the PhD is to never treat your work as your own child – otherwise you can never improve on it.

I owe this and many other lessons to my supervisors. There have been a total of four of them in different phases of my PhD and it has been an honor of working with them all. My first first promoter, Joop Halman; I am looking back with great pleasure at the beginning of my PhD. You sparked my interest in science and I am grateful for your enthusiasm and insightful comments. My second first promoter, Mathieu Weggeman; thank you for reminding me of the importance of the practitioners view on scientific research. Your feedback allowed me to take an "outside view" on my work.

My second promoter, Michael Song; thank you for giving me the freedom to choose the research paths I was interested in and for making sure that they were scientifically sound. Each discussion of our papers is a challenge I immensely enjoy – they are always unpredictable and stimulating. My daily supervisor, Hans van der Bij; I am truly thankful for your tremendous support and for sharing your knowledge with me. You helped me dare to make my own decisions, not the decisions that ought to be mine. Thank you for being always open for the crazy ideas I could come with and even accepting them so now and then.

(11)

vi

I am grateful to Aard Groen, Joop Halman, Rob Verbakel, Leo Verhoef en Mathieu Weggeman for helping me find "de proefkonijnen" for my case-studies and pre-tests of the two surveys. I am similarly indebted to the nearly 30 entrepreneurs who agreed to share their inspiring stories and answered my numerous questions.

My gratitude to the PhD commission for this dissertation; Anthony Di Benedetto, Geert Duijsters and Mark Parry, thank you for evaluating this thesis and for your understanding when I had to move its defense date due to the problems in my pregnancy.

I owe a lot of warm memories to the company of our PhD students: Ad, Bonnie, Deborah, Elise, Jeroen, Maurice, Michael, Michiel, Mirjam, Rebekka, Stephan and Vareska. I am thankful to the rest of our OSM colleagues who both helped me improve my Dutch during the lunch hours and helped me out scientifically whenever I needed their advice.

The secretary room has been a social center of our group all the time, and it could not be possible without Bianca, Marion and Marjan. Julius Caesar, famous for his multitasking, would be jealous of your abilities to combine things!

I am endlessly grateful to my parents for raising me who I am, for tinkling my curiosity and for being there whenever I needed their advice. Your warm words and (sometimes not so polite) questions helped me through the most difficult times in my PhD.

Dearest Vladimir, the more I know you, the more I treasure our relationship. Thank you for your patience and your support back home! Even more, thank you for not throwing away my laptop in the busiest times of my PhD!

(12)

Introduction 11

Chapter 1 General introduction

Despite all the controversy about what entrepreneurship is about and who an entrepreneur precisely is (Shane and Venkataraman, 2000), one aspect that consistently comes back in the debates is risk-taking. Taking risk is one of the core functions of an entrepreneur (e.g. Knight, 1921). In our exploratory in-depth interviews with entrepreneurs we came across different risks ranging from "not having focus", "not having a proper image" to "accident in the laboratory" and "an essential staff member leaves the firm". No matter how the various risk factors are framed, it all tends to come back to the survival and prosperity of the firm: i.e. to finances.

Although entrepreneurship is the driving force behind new job creation (Shane, 2003), not all the new entrepreneurial firms become new Microsoft's or Dell's. The majority will bleed to death: depending on the industry, only 37-54% of new firms survive the first year (Timmons and Spinelli, 2004). The risks seem to be huge – so how can entrepreneurs improve their risk-taking?

Despite its importance, the risk theme has been under-researched in the entrepreneurship literature. One of the reasons may be the influential study of Brockhaus (1980) that found no differences in risk-taking of entrepreneurs as opposed to managers in traditional organizations. However, the recent meta-analyses of Stewart and Roth (2001, 2004) found that there is still a significant difference, no matter what measures are being used, although these measures do influence the magnitude of the effect.

(13)

12_Introduction

Probably the most serious problem in studying risk is still of theoretical nature. Risk is a concept that can be applied to almost any field and any theory. It is so broad that it can easily become too limited. Each theory may have its own risks (e.g. Johnson and Van de Ven, 2002). Therefore, bringing all the possible risks under the same umbrella would mean creating a "theory soup". A clear theoretical lens should be used for risk studies.

The empirical part of the research on risk also has a number of pitfalls. The term "risk" has a strong negative connotation, a consequence of which is that data about risks are hard to get from the entrepreneurs. As we have observed in our exploratory case studies, entrepreneurs are reluctant to reveal information on risks, which may also explain why there are so few studies explicitly studying risks. A particular risk can be framed both in positive and negative terms. Therefore, in the studies of risk diagnostics researchers are advised to frame the risk statements positively in order not to evoke defensive behavior from the respondents (Keizer, Halman and Song, 2002).

We tried to avoid these pitfalls and intend to contribute to the entrepreneurship field by answering the following three-fold main research question: (1) What kinds of risks do entrepreneurs take? (2) How do they take the risks? (3) What kinds of strategies do they use to manage the risks? We answer these research questions in three core chapters of this dissertation: chapter two, three and four respectively. We answer the two "what" questions at the firm level and the "how" question at the individual level.

1.1 Meta-analysis of success factors

"Luck is one of the key factors in entrepreneurial success." Fern Mandelbaum (Monitor Venture Partners)

"My idea of risk and reward is for me to get the reward and others to take the risks." An unknown entrepreneur

We started this research by asking ourselves: what kinds of risks do entrepreneurs take? However, due to the lack of studies researching entrepreneurial risks directly, we decided to focus on the positive side and consider risks as the opposite of success factors: i.e. if a given factor is truly a success factor, then not possessing it would mean a risk for an entrepreneurial firm. The higher the effect of a given factor on performance, the more severe it becomes for a new technology venture not to possess this factor.

(14)

Introduction 13 In order to answer this question, we conducted a meta-analysis. A meta-analysis is a method to review and integrate existing research on a given topic (Hunter and Schmidt, 1990). One aspect that clearly differentiates it from narrative reviews is its quantitative character. Unlike primary research, in a meta-analysis the data analyzed consist of the findings from previous empirical studies (Camisón-Zornoza et al., 2004). Just as empirical research requires the use of statistical techniques to analyze its data, meta-analysis applies statistical procedures that are specifically designed to integrate the results of a set of primary empirical studies. This allows meta-analysis to pool all the existing literature on a given topic, not only the most influential and best-known studies (Stewart and Roth, 2001, 2004). At the same time, meta-analysis compensates for quality differences by correcting for different artifacts and sample sizes (Hunter and Schmidt, 1990, 2004).

In this meta-analysis we subdivide the main research question into a set of these lower-level research questions:

• What are the success factors for new technology ventures? • Is the literature consistent on their estimates of these factors?

• In cases when the literature is not consistent, what are the potential methodological moderators for these factors?

For our meta-analysis, we used the methodology developed by Hunter and Schmidt (1990, 2004). We searched for studies on new technology ventures performance in ABI-INFORM and on the internet using the following keywords: “new,” “adolescent,” “young,” and “emergent“ to define the “new” axis; "technology", "high-tech", “technology-intensive,” and “technology-based” to describe the technology domain; and “firm,” “venture,” and “start-up” to define the entity. We examined past research studies where the majority of the sample represented such “new” “technology” ventures. In general, the primary studies set the maximum age for new technology ventures at 15 years, yet most primary studies selected cut-off values of 6 and 8 years. The last selection criterion was publication of a correlation matrix with a performance measure, because correlation matrices serve as the main input for the meta-analysis.

One of our conclusions from the meta-analysis is that there is a lack of entrepreneurship studies bridging the strategic management and the deeper cognitive mechanisms of entrepreneurial decision making. As Busenitz and Barney (1997) argued, if certain individuals are cognitively biased in different ways, they may make strategic decisions

(15)

14_Introduction

in different ways. Past research shows that besides susceptibility to cognitive biases, entrepreneurs differ cognitively from managers in more traditional firms on a number of dimensions, including risk-taking propensity and reliance on intuition (Busenitz and Barney, 1997; Stewart and Roth, 2004). Thus, such cognitive mechanisms may represent sources of competitive advantage or disadvantage of the firms (Barney, 1991). We intend to explore the aforementioned gap on two levels: we look at the individual level how entrepreneurial risk-taking propensity is formed and then we explore at the firm level the performance consequences of various risk and uncertainty management strategies new technology ventures pursue.

1.2 The mechanism of entrepreneurial risk-taking

''There is a fine line between confidence and arrogance. […] You have to have confidence in order to take risks, because too many people are knocking you down and if you do not have confidence, you are not going to keep going. But then at some point, you have success and that confidence changes to arrogance. That's where it really gets dangerous. Arrogance indicates that you are not listening to customers, employees and the market. So it is a fine line and you have to stay on the good side!"

Judy Estrin (Packet Design)

"The younger the company, the more opportunity to take risks you have. Big companies do not take risks. This is the advantage you have when you start a new company. Propensity to take risks is what really differentiates an entrepreneur from a manager in a big company."

Randy Adams, serial entrepreneur (AuctionDrop)

In venture-related decisions, entrepreneurs have to cope on a daily basis with ill-structured, uncertain sets of possibilities, while having the ultimate responsibility for each decision (Knight, 1921; Stewart and Roth, 2001). There are a number of risks associated with this kind of decisions and the question is: How do entrepreneurs take these risks? What is the more precise mechanism of entrepreneurial risk-taking?

Entrepreneurship literature provides two alternative answers about how entrepreneurs take these risks. One explanation is that entrepreneurs objectively tolerate more risks, that they are risk-seeking and that they consciously take the risks (Stewart and Roth, 2001; 2004). A competing, cognitive explanation is that in their intuitive decision-making, entrepreneurs are unconscious (or at least not fully conscious) of the actual risks associated with their decisions; they simply do not see them all due to the cognitive biases (Simon, Houghton and

(16)

Introduction 15 Aquino, 2000). These biases are in fact errors in decision-making arising from the use of heuristics.

How can these two be brought closer to each other? Dual process theory provides an answer. By now the entrepreneurship literature (e.g. Covin, Slevin, and Heeley, 2001; Simon, Houghton, and Savelli, 2003) focused on intuition as the opposite side of being rational. Entrepreneurs are seen as predominantly intuitive decision-makers. However, according to the dual process theory, people can be intuitive and rational at the same time (Epstein, Pacini, Denes-Raj, and Heier, 1996; Pacini and Epstein, 1999). This theory postulates that all judgments and behavior of people are a joint output of both intuitive and rational thinking (Epstein et al., 1996). The rational thinking monitors and eventually corrects outputs of intuitive thinking (including the heuristics and biases). It provides an answer itself if no intuitive judgment is available (Epstein et al., 1996; Stanovich and West, 2000). Recent theoretical developments suggest that at least some of the cognitive biases studied until now can be actually conceptualized as biases of human intuition (Kahneman, 2003). Thus, while risk-seeking perspective can be related to the rational dimension in the dual process theory, the cognitive perspective can be related to the intuitive dimension in the dual process theory.

In this individual-level study, we subdivide the main research question into a set of lower-level research questions: we concentrate on how intuitive thinking, rational thinking, heuristics and biases form entrepreneurial risk-taking:

• To what extent can the cognitive biases studied until now be called "intuitive", i.e. deriving from the intuition?

• To what extent can rational thinking (i.e. without any special training) correct the entrepreneurial cognitive biases and improve entrepreneurial decision-making?

• Do the intuitive and rational thinking also directly influence the entrepreneurial risk-taking?

• To what extent do cognitive biases influence entrepreneurial risk-taking propensity?

We integrate these two perspectives in a model where heuristics and biases mediate the relationship between the intuitive and rational thinking and entrepreneurial risk-taking propensity. We answer our research questions by testing the conceptual model using SEM with Maximum Likelihood (ML) estimator on a sample of 289 entrepreneurs from the US.

(17)

16_Introduction

1.3 Risk and uncertainty management strategies

"There are many kinds of risk that you have when starting a company: there's technical risk, market risk, financing risk and many other kinds of risks you can learn in a business school. The trick is to take the risk out as early as possible. And take as few risks as possible."

Jerry Kaplan, entrepreneur (Winster)

"There are risks and costs to a program of action, but they are far less than the long-range risks and costs of comfortable inaction."

John F. Kennedy, president of USA

This chapter is dedicated to the risk and uncertainty management strategies new technology ventures may pursue. We distinguish two major types of risk and uncertainty management strategies: the traditional risk management (Miller, 1992) and the recently emerged real options reasoning (McGrath, 1999; McGrath, Ferrier and Mendelow, 2004). Both types of strategies concern the mitigation of risk and management of uncertainty. An example of the differences between them is that traditional risk management strategies typically target immediate risk reduction, whereas real options strategy delays the full commitment decision and provides flexibility for future decisions. In Chapter 4, we further elaborate on the differences between traditional risk management strategies and real options strategy using the dimensions identified by Bowman and Hurry (1993), namely risk, uncertainty, size and timing of investments.

Despite the integrative review (Miller, 1992), the traditional risk management strategies were hardly ever compared empirically. Recent developments in the real options theory refined the concept of real options strategy, providing the basis for thorough empirical tests (McGrath et al., 2004). However, real options has also received certain critique raising doubts about the value and distinctiveness of this strategy (Adner and Levinthal, 2004a,b; Miller and Arikan, 2004).

In this study, we aim to tackle these issues by developing a scale to measure the real options strategy directly and comparing its performance consequences with those of the traditional risk management strategies. We also examine how market and opportunity characteristics influence the preference of entrepreneurial ventures for each type of strategic risk management strategy. We consider the effects of established versus emerging technology standards as well as effects of markets with high versus low network externalities.

(18)

Introduction 17 In this firm-level study, we subdivide the main research question into two lower-level research questions:

• What are the performance consequences of the risk and uncertainty management strategies?

• How is the effect of these strategies influenced by market and technology characteristics?

We test our conceptual model by OLS regressions for three different performance measures: return on investment, customer retention rate and sales growth rate. For this test, we use the data from 420 new technology ventures from USA.

(19)

(20)

Meta-analysis of success factors 19

Chapter 2 Meta-analysis of success factors

New technology ventures have the lowest survival rate among all the new ventures (Timmons and Spinelli, 2004). To get a more integrated picture of what factors lead to the success or failure of new technology ventures, we conducted a meta-analysis to examine the success factors in new technology ventures. We culled the academic literature to collect data from existing empirical studies and conducted a meta-analysis. We identified 24 most-widely researched success factors for new technology ventures. Among these 24 factors, 8 are consistently estimated as significant success factors for new technology ventures (i.e., they are homogeneous positive significant meta-factors that are correlated to venture performance). They are supply chain integration, market scope, firm age, size of founding team, financial resources, founders’ marketing experience, founders’ industry experience, and existence of patent protection. Of the original 24 success factors, 5 were not significant: the success of technology ventures are not correlated with founders’ R&D experience, founders’ experience with start-ups, environmental dynamism, environmental heterogeneity, and competition intensity. The remaining 11 success factors are heterogeneous. For those heterogeneous success factors, we conducted a moderator analysis. Of this set, 3 appeared to be success factors and 2 were failure factors for subgroups within the new technology ventures’ population. To facilitate the development of a body of knowledge in technology entrepreneurship, this study also identifies high-quality measurement scales for future research. We conclude the article with future research directions.

(21)

20_{The mechanism of entrepreneurial risk-taking}

2.1 Introduction

Technology entrepreneurship is key to economic development. New technology ventures can have positive effects on employment, and can rejuvenate industries with disruptive technologies (Christensen and Bower, 1996).

Unfortunately, the survival rate of new technology ventures is the lowest among new ventures in general. In our most recent empirical study of 11,259 new technology ventures established between 1991 and 2000 in the United States, we found that after four years only 36 percent, or 4,062 companies with more than five full-time employees, had survived. After five years, the survival rate fell to 21.9 percent, leaving only 2,471 firms with more than five full-time employees still in operation. Given this high rate of failure, it is important to identify what factors lead to the success and failure of new technology ventures.

Current academic literature, however, does not offer much insight. Numerous studies focus on success factors for new technology ventures, but the empirical results are often controversial and fragmented. For example, the data on R&D investments alone yield ambivalent conclusions. While Zahra and Bogner (2000) found no significant relationship between R&D expenses and new technology venture performance, Bloodgood, Sapienza, and Almeida (1996) found a negative relationship and Dowling and McGee (1994) found a positive relationship between R&D investments and new technology venture performance.

Similarly, although new technology ventures often develop knowledge-intensive products and services (OECD, 1997), the research results on product innovativeness have been ambiguous. More than two-thirds of the empirical studies have found a positive relationship between product innovation and firm performance, while the remaining studies have found a negative relationship or none at all (Capon, Farley, and Hoenig, 1990; Li and Atuahene-Gima, 2001). Li and Atuahene-Gima (2001) addressed this problem by introducing contingencies into their regression models and indeed found three moderators.

The inconsistent and often contradictory results can stem from methodological problems, different study design, different measurements, omitted variables in the regression models, and noncomparable samples. More than on any one methodology, entrepreneurship theory hinges on its setting (new firms) as its common denominator. Because of that, numerous theoretical streams run through the scholarship (Shane and Venkataraman, 2000). To help resolve this problem, we looked for a method that would operate independently of model composition. Meta-analysis provides a solution (Hunter and Schmidt, 1990, 2004) and

(22)

Meta-analysis of success factors 21 a lens through which we can evaluate the success factors that contribute to new technology ventures’ performance. We based our meta-analysis on studies that explicitly focus on antecedents of new technology venture performance.

This chapter attempts to make several contributions to technology entrepreneurship literature: (1) our integrated quantitative evaluation of the success factors of new technology ventures provides one step toward developing a theoretical foundation for technology entrepreneurship, (2) it identifies universal success factors, (3) it identifies success factors that are controversial and, by moderator analysis, offers some tentative reasons for those controversies, (4) it reports existing high-quality scales that are important for new technology venture performance, and (5) it proposes and provides a new theoretical framework for studying success factors of technology ventures and a road map for future research in technology entrepreneurship.

This chapter is organized in the following manner. First, we explain our methodology. We then present the results of our research, including the results of the meta-analysis, examples of high-quality scales, and the discussion of future research directions. We conclude the chapter with a description of its limitations and some final remarks.

2.2 Data collection and methodology

Meta-analysis is a statistical research integration technique (Hunter and Schmidt, 1990). One aspect that clearly differentiates it from narrative reviews is its quantitative character. Unlike primary research, in a meta-analysis the data analyzed consist of the findings from previous empirical studies (Camisón-Zornoza et al., 2004). Just as empirical research requires the use of statistical techniques to analyze its data, meta-analysis applies statistical procedures that are specifically designed to integrate the results of a set of primary empirical studies. This allows meta-analysis to pool all the existing literature on a given topic, not only the most influential and best-known studies (Stewart and Roth, 2001, 2004). At the same time, meta-analysis compensates for quality differences by correcting for different artifacts and sample sizes (Hunter and Schmidt, 1990, 2004).

There are two main types of meta-analytic studies in the literature. The first focuses on a relationship between two variables or a change in one variable across different groups of respondents. In general, this type of meta-analysis is strongly guided by one or two theories

(23)

(e.g., Palich, Cardinal, and Miller, 2000; Stewart and Roth, 2001, 2004). The second type of meta-analytic studies examines a large number of meta-factors related to one particular focal construct, such as performance. Such meta-analyses aim to integrate all the existing research on that focal construct and are largely atheoretical because the research they combine rests on heterogeneous theoretical grounds (e.g., Gerwin and Barrowman, 2002; Montoya-Weiss and Calantone, 1994). Because the current literature teems with numerous theoretical streams where only the setting (new firms) is the common denominator (Shane and Venkataraman, 2000), we chose the second type of meta-analysis to study the potential success meta-factors of new technology venture performance. We selected independent ventures and collected studies that explicitly focused on antecedents of new technology ventures’ performance.

In our study, we explore—rather than define ourselves—what “new technology venture” means in the literature. Primary studies use such terms as “new,” “adolescent,” “young,” or “emergent“ to define the “new” axis; and “high technology,” “technology-intensive,” and “technology-based” to describe the technology domain. We examined past research studies where the majority of the sample represented such “new” “technology” ventures. In general, the primary studies set the maximum age for new technology ventures at 15 years, yet most primary studies selected cut-off values of 6 and 8 years. Another important selection criterion was the publication of the correlation matrix in the paper, because the correlation matrices serve as the main input for the meta-analysis. All the collected studies investigated surviving new technology ventures; consequently, we do not consider failures in our meta-analysis.

Meta-analysis allows the comparison of different empirical studies with similar characteristics, and thus lets researchers integrate the results. To conduct a meta-analysis it is important to select studies as input for the analysis and follow a meta-analytical protocol to arrive at those results.

2.2.1 Selection of studies as input for the analysis

First, we combed the literature for research that discussed the success factors of new technology ventures, using the ABI-INFORM system and the Internet. We used keywords— “new,” “adolescent,” “young,” “emerging” and “high-tech,” “technology,” “technology-intensive,” “technology-based”—to limit our sample’s age and domain. Finally, to assess the type of firm, we applied the keywords “firm,” “venture,” and “start-up.” We intentionally did

(24)

Meta-analysis of success factors 23 not limit the studies to those recognized as the best in the field, as usually done in a narrative review: this would have betrayed the spirit of meta-analysis (Hunter and Schmidt, 1990). Instead, we collected as much research as possible, corrected later for any quality differences and controlled for missing studies.

After we gathered papers from ABI-INFORM and the Internet, we added cross-referenced studies from them. In total, we collected 106 studies that met our search criteria. Next, we ensured that the articles on our list (1) represented the correct level of analysis, (2) significantly reflected new technology ventures, and (3) reported a correlation matrix with at least one antecedent of performance and one performance measure. This procedure reduced the number of appropriate research studies to 31 due to the absence of correlation matrices. Appendix B2.1 details our study sample by countries of origin, industries, performance measures, the minimum and maximum ages of the ventures, and their sample sizes. In addition, we provide two other features. First, “sample type” indicates the particular characteristics of the sample. This may be new technology ventures that went through initial public offering (IPO), ventures funded by venture capitalists (VC), ventures from a general database, ventures involved in a governmental support program, ventures that have activity abroad, or combinations of these types. Second, “venture origin” indicates whether the venture was actually independent. Although our meta-analysis focused primarily on independent ventures, it also included mixed samples of independent and corporate ventures, where most were independent, and samples where the type of venture was not specified. Appendix B2.2 lists the journals from which the 31 papers originate.

When coding the studies, we took care to refer to the scales reported in the primary studies, so that dissimilar elements would not be combined inappropriately, and conceptually similar variables would not be coded separately, to compensate for the slightly different labels that authors use to refer to similar constructs (Henard and Szymanski, 2001).

2.2.2 Protocol for meta-analysis

We used Hunter and Schmidt’s protocol (1990) for our meta-analysis. Our most important consideration was to the ability to make comparisons across research studies. To do this, we could draw on Pearson correlations between a meta-factor and the dependent variable or the regression coefficient between the meta-factor and the dependent variable. Because regression coefficients depend on the particular variables included into the model and because

(25)

the models vary across studies, we followed the suggestions of Hunter and Schmidt (1990). Hunter and Schmidt strongly encourage using Pearson correlations as the input, because correlations between two variables are independent of the other variables in the model (Hunter and Schmidt 1990). Other meta-analytic studies have made this choice, including Gerwin and Barrowman (2002) and Montoya-Weiss and Calantone (1994).

Another advantage of Hunter and Schmidt’s method (1990) is their use of random effects models instead of fixed effects models (Hunter and Schmidt, 2004; p.201). The distinction is as follows: fixed effects models assume that exactly the same “true” correlation value between meta-factor and dependent variable underlies all studies in the meta-analysis, while random effects models allow for the possibility that population parameters vary from study to study. Given the differences in how new technology ventures were defined in the selected primary studies, the choice for random effects models was appropriate.

Following the procedure of Hunter and Schmidt (1990), our second step was to correct meta-factors for dichotomization, sample size differences, and measurement errors.

1) To correct dichotomized meta-factors: we made a conservative correction by dividing the observed correlation coefficient of the sample by 0.8, because dichotomization reduces the real correlation coefficient by at least 0.8 (Hunter and Schmidt, 1990, 2004).

i. study primary the of n correlatio observed : not; is it if 1 and ed dichotimiz is variable if 0.8 ation; dichotomiz for correction : : where , : ation dichotomiz for ns correlatio observed of correction Individual i i i oo d d d d oo o r a a a a r r = = =

2) To correct sampling error: we weighted the sample correlation by sample size (Hunter and Schmidt, 1990, 2004).

i. study primary the of size sample : where , : ation dichotomiz for corrected ly individual ns correlatio of average Weighted 1 1 i n i i n i o i o N N r N r i

∑

= = =

(26)

Meta-analysis of success factors 25 3) To remedy measurement errors: we used Cronbach’s alphas. We divided the correlation coefficient by the product of the square root of the reliability of the meta-factor and the square root of the reliability of performance. Since reliabilities were not always reported, we reconstructed them by using the reliability distribution (Hunter and Schmidt, 1990, 2004). factor. -meta given a composing ariables depedent v of ies reliabilit of roots square the of average : factor; -meta given a composing variables indepedent of ies reliabilit of roots square the of average : factor; on attentuati compound : : where , * : n correlatio population Real yy xx yy xx o o R R A R R r A r = = ρ

The third step in the meta-analysis protocol was to determine whether a meta-factor was a success factor. To accomplish this, we assessed three conditions. First, the studies should have, in essence, the same correlation. Other meta-analysis procedures often use a Chi-square test to reveal this homogeneity. However, Hunter and Schmidt (1990, 2004) argue against it and state that this test will have a bias because of uncorrected artifacts. They suggest a variance-based test. The total variance in the correlation coefficient has three sources: variance due to artifacts (dichotomization and measurement errors), variance due to sampling error, and real variance due to heterogeneity of the meta-factor. The meta-factor is assumed to be homogeneous, if the real variance is no more than 25 percent of the total variance. According to Hunter and Schmidt (1990, 2004), in that case unknown and uncorrected artifacts account for these 25 percent, so that the real variance is actually close to zero. We describe the used formulas in Appendix C2.1.

For homogeneous meta-factors, we applied two significance tests. First, we determined whether the whole confidence interval (based on the real standard deviation) was above zero. Second, if it was above zero, we calculated the p-value for the real correlation to estimate the degree of significance. Both of these significance tests are necessary, because the p-value is misleading when part of the confidence interval of the real correlation is below zero. Only when all three conditions held did we consider a given meta-factor to be a success meta-factor for new technology ventures.

(27)

For those heterogeneous meta-factors, we conducted a moderator analysis. We divided the data into subgroups according to various methodological characteristics (see Appendix B2.1). Then, we conducted a separate meta-analysis for each subgroup, hoping to find homogeneous meta-factors in the subgroup in two steps. First, we conducted moderator analysis to deal with different performance measures. Second, we checked whether country, industry, sample type, venture origin, or maximum age of the new technology ventures in the sample were possible moderators. Second, we conducted moderator analysis for different meta-factor measures.

Finally, we reviewed the “file drawer” in an attempt to assess any publication bias. Because there is a general tendency to publish only significant results, insignificant results are often abandoned in researchers’ file drawers (Hunter and Schmidt, 1990; Rosenthal, 1991). This “file drawer” technique provides a number, XS, indicating the number of null-result

studies that when added, would make the total significance of a meta-factor exceed the critical level of 0.05. Thus, the higher the value of XS, the more stable and reliable the results are. If

XS, is 0, it indicates that the meta-factors are already insignificant according to the p-value

criterion.

2.3 Analysis and results

2.3.1 Success factors of technology ventures

Our meta-analysis revealed 24 meta-factors related to the performance of new technology ventures. We present the definitions of these meta-factors in Table 2.1.

(28)

Table 2.1. Definitions of the 24 meta-factors

Meta-factors Definitions Selected references

Market and opportunity

1. Competition intensity _{Strength of inter-firm competition within an industry} _{Chamanski and Waagø, 2001}

2. Environmental dynamism _{High pace of changes in the firm's external environment} _{Zahra and Bogner, 2000}

3. Environmental heterogeneity _{Perceived diversity and complexity of the firm's external environment} _{Zahra and Bogner, 2000}

4. Internationalization _{Extent to which a firm is involved in cross-border activities} _{Bloodgood et al., 1996}

5. Low cost strategy _{Extent to which a firm uses cost advantages as a source of competitive}

advantage

Bloodgood et al., 1996

6. Market growth rate _{Extent to which average firm sales in the industry increase} _{Bloodgood et al., 1996; Lee et al., 2001}

7. Market scope _{Variety in customers and customer segments, their geographic range,}

and the number of products

Li, 2001; Marino and De Noble, 1997

8. Marketing intensity* _{Extent to which a firm is pursuing a strategy based on unique}

marketing efforts

Li, 2001

9. Product innovation* _{Degree to which new ventures develop and introduce new products}

and/ or services

Li, 2001

Entrepreneurial team

10. Industry experience _{Experience of the firm's management team in related industries and}

markets

Marino and De Noble, 1997

11. Marketing experience _{Experience of the firm's management team in marketing} _{McGee et al., 1995; Marino and De Noble,}

1997

12. Prior start-up experience _{Experience of the firm's management team in previous startup}

situations

Marino and De Noble, 1997

13. R&D experience _{Experience of the firm's management team in R&D} _{McGee et al., 1995; Marino and De Noble,}

(29)

Table 2.1 (continued). Definitions of the 24 meta-factors

Meta-factors Definitions Selected references

Resources

14. Financial resources _{Level of financial assets of the firm} _{Robinson and McDougall, 2001}

15. Firm age _{Number of years a firm has been in existence} _{Zahra et al., 2001}

16. Firm size _{Number of the firm's employees} _{Zahra et al., 2001}

17. Firm type _{The type of a firm's ownership (corporate ventures or independent ventures)} _{Zahra et al., 2001}

18. Non-governmental financial

support Financial sponsorship from commercial institutes Lee et al., 2001

19. Patent protection _{Availability of firm's patents protecting product or process technology} _{Marino and De Noble, 1997}

20. R&D alliances _{The firm's use of R&D cooperative arrangements. For new technology}

ventures they also correspond to horizontal alliances.

Zahra and Bogner, 2000; McGee et al., 1995

21. R&D investment _{Intensity of the firm's investment in internal R&D activities} _{Zahra and Bogner, 2000}

22. Size of founding team _{Size of the management team of the firm} _{Chamanski and Waagø, 2001}

23. Supply chain integration _{A firm’s cooperation across different levels of the value-added chain, for}

example suppliers, distribution channel agents, and/or customers

George et al., 2001; George et al., 2002; McDougall et al., 1994

24. University partnerships _{The firm's use of cooperative arrangement with universities} _{Zahra and Bogner, 2000; Chamanski}

and Waagø, 2001 * - these two factors are called marketing differentiation and product differentiation in the stream of research stemming from the work of Porter (1980)

(30)

Meta-analysis of success factors 29 Table 2.2 reports the analytic results on the antecedents, or the success meta-factors of new technology ventures’ performance. To be concise and limit the sensitivity of the results to studies not included in our analysis, Table 2.2 presents only the meta-factors found in three or more research studies. The table presents ρ, an estimate of the real population correlation; total N, the aggregate sample size; and K, the number of correlations that build a given meta-factor. Both N and K are conservative: we counted each study only once. Ninety-five (95) percent confidence interval is the spread of the real correlation variance. XS is the critical number of null-results studies.

To make the analysis of the meta-factors more transparent and interpretable, we generate appropriate categories grounded in the literature’s existing frameworks (Chrisman, Bauerschmidt, and Hofer, 1998; Gartner, 1985; Timmons and Spinelli, 2004). These categories are: (a) Market and Opportunity, (b) the Entrepreneurial Team, and (c) Resources. After three researchers reviewed those categories for completeness and appropriateness, we conducted content analysis, a classification technique that assigns variables to a particular category. Two researchers independently assigned each variable to a category. The two researchers agreed on variables' categorizations in 91.2 percent of the cases across 306 variables. A third researcher resolved any disagreements, making the final categorization. At the same time, variables were combined to form meta-factors.

Reflecting the primary studies, the Market and Opportunity category typically described either the market characteristics, such as environmental dynamism, environmental heterogeneity, and competitive strategies based on Porter’s (1980) typology. The Entrepreneurial Team category included characteristics of the new technology venture team, including experience and capabilities, both as individuals and as a team. The Resources category united a broad scope of factors, comprising resources, capabilities, and characteristics of the new technology ventures as firms. Such resources included financial resources, firm size, patents, and university partnerships.

The meta-factors were unevenly distributed across the three categories. The majority fell into the Resources category; the smallest number, into the Entrepreneurial Team category. The Resources category consisted of heterogeneous meta-factors for 55 percent and the Market and Opportunity category for 56 Percent. Only the Entrepreneurial Team category was completely homogeneous.

(31)

Table 2.2. Results of the meta-analysis

Meta-Factor Total N K ρρ ρρ 95 % Confidence Interval Explained Variance a Mode- rators Xs

MARKET and OPPORTUNITY

1 Competition intensity 634 7 0.01 100% 0

2 Environmental dynamism 637 5 0.05 100% 0

3 Environmental heterogeneity 287 3 0.10 100% 0

4 Internationalization 523 7 0.08(*) (-0.21,0.37) 38% Yes 6 5 Low cost strategy 286 4 0.18(**) _(-0.13,0.49) _70% _Yes ₁₀

6 Market growth rate 505 4 0.23(***) (-0.26,0.72) 16% Yes 12

7 Market scope 1046 10 0.21*** 100% 78

8 Marketing intensity 622 6 0.42(***) (-0.19,1.00) 23% Yes 64

9 Product innovation 702 8 0.04 (-0.48,0.56) 55% Yesb 0

ENTREPRENEURIAL TEAM

10 Industry experience 423 4 0.11* 100% 2

11 Marketing experience 381 3 0.11* 100% 2

12 Prior start-up experience 114 3 0.00 100% 0

13 R&D experience 329 3 0.09 100% 0

RESOURCES

14 Financial resources 638 6 0.12** 100% 14

15 Firm age 1890 15 0.16*** (0.08,0.23) 87% 157

16 Firm size 1360 11 0.26(***) (-0.31,0.83) 10% Yes 197

17 Firm type 715 4 0.09 (-0.15,0.33) 31% Yesb 0

18 Non-governmental

financial support 405 4 0.20

(***)

(-0.15,0.55) 31% Yes 16

19 Patent protection 453 5 0.11* 100% 1

20 R&D alliances 571 5 0.03 (-0.52,0.58) 31% Yesb 0

21 R&D investments 863 9 0.05(*) _(-0.49,0.60) _19% _Yes ₃

22 Size of founding team 332 5 0.13** 100% 6

23 Supply chain integration 604 6 0.23*** (0.12,0.35) 89% 41 24 University partnerships 330 3 -0.04 (-0.25,0.17) 50% Yes 0

a

– explained variance lower than 75% means that the meta-factor has moderator(s)

b

– see Table 2.3 for suggested moderators

*

p < 0.05; ** p < 0.01; *** p < 0.001. 1-tailed test statistic. Direction depends on the sign of ρ p-values indicated by (*), (**) or (***) mean that the meta-factor is heterogeneous

(32)

Meta-analysis of success factors 31 Results in Table 2.2 reveal eight universal success factors (i.e., they are homogeneous positive significant meta-factors that are correlated to venture performance):

• supply chain integration (ρ = 0.23, p<0.001) • market scope (ρ = 0.21, p<0.001)

• firm age (ρ = 0.16, p<0.001)

• size of founding team (ρ = 0.13, p<0.01) • financial resources (ρ = 0.12, p<0.01) • marketing experience (ρ = 0.11, p<0.05) • industry experience (ρ =0.11, p<0.05) • patent protection (ρ =0.11, p<0.05)

One success factor represented Market and Opportunity, five success factors represented Resources, and two success factors were part of the Entrepreneurial Team category.

Results in Table 2.2 also suggested that the following five factors have no significant effects on technology venture performance: 1) R&D experience, 2) prior start-up experience, 3) environmental dynamism, 4) environmental heterogeneity, and 5) competition intensity. Three of these meta-factors represented Market and Opportunity and two represented the Entrepreneurial Team category.

2.3.2 Moderators

As Table 2.2 indicates, 11 of the 24 meta-factors had heterogeneous correlations (i.e., the importance of the factors depend on situations). Therefore, we conducted moderator or subgroup analysis for differences in performance measures, meta-factor measures, venture origin, maximum age of venture in the sample, sample type, country, and industry.

Table 2.3 presents those results from the moderator analysis, including ρ, an estimate of the real population correlation; total N, the aggregate sample size1; K, the number of correlations that build a given meta-factor; the 95 percent confidence interval of the real variance; and XS, the critical number of null-results studies.

1_{Since some studies used multiple measures of performance, sum of performance moderator}

(33)

Table 2.3 also presents the variance explained by dichotomization of meta-factors, measurement, and sampling error. This variance must be more than 75 percent to yield a homogeneous factor. In that case, the real variance is less than 25 percent of the total variance of correlations from the primary studies. The remaining variance is likely due to other unknown and uncorrected artifacts, and therefore it can be neglected (Hunter and Schmidt, 1990, 2004). To keep overview, for the moderator, or subgroup analysis, we only report the meta-factors with at least two subgroups that have no overlapping confidence intervals; each subgroup consists of at least two studies.

Table 2.3. Suggested moderators

Meta-factor Moderator a ρ ρ ρ ρ Total _N K 95 % Confidence Interval Explained Variancea XS RESOURCES Firm type 0.09 715 4 (-0.15,0.33) 31% 0 Performance operationalization Profit based -0.01 572 3 (-0.03,0.01) 98% 0 Sales based 0.27*** 464 2 100% 18 R&D alliances 0.03* 571 5 (-0.52,0.58) 31% 0 Venture origin Independent ventures -0.36*** 262 2 100% 10 Mixed origin 0.37*** 309 3 100% 28

MARKET and OPPORTUNITY

Product innovation 0.04 702 8 (-0.71,0.79) 12% 0

Venture origin

Independent ventures -0.39*** 263 3 (-0.52,-0.27) 80% 23

Mixed origin 0.44*** 300 2 (0.23,0.65) 43% 23

a

- explained variance lower than 75% means that the meta-factor has moderator(s)

*

p < 0.05; ** p < 0.01; *** p < 0.001. 1-tailed test statistic. Direction depends on the sign of ρ

The results reported in Table 2.3 suggest that of the 11 heterogeneous factors, 3 meta-factors (firm type, R&D alliances, and product innovation) had distinct moderator subgroups (i.e., the effect of these factors on venture performance depends on situation). The relationship between firm type and performance depended on the way performance was measured. Firm type was insignificantly related to the profits of new technology ventures, but significantly and positively related to the sales of new technology ventures. No other (methodologically oriented) moderators affected the firm type.

R&D alliances were negatively associated with performance for independent ventures. However, for ventures of a mixed origin, R&D alliances were positively associated with performance.

(34)

Meta-analysis of success factors 33 Product innovation was moderated by venture origin. For independent new technology ventures, product innovation has a significantly negative association with performance. However, for samples with mixed firm type, product innovation has a significantly positive association with performance.

By examining the results in Table 2.2, eight meta-factors proved inconclusive: internationalization, low-cost strategy, market growth rate, marketing intensity, R&D investments, firm size, non-governmental financial support, and university partnerships. Of these eight meta-factors, market growth rate and non-governmental financial support have only one subgroup with two or more studies when differences in meta-factor measurements are considered. We also found only one suitable subgroup for internationalization when looking at sample type, for marketing differentiation when looking at the country, and for university partnerships when either looking at the sample type or the industry. Further research is needed to validate or disprove these potential moderators. Finally, no methodological moderators were found for R&D investments, low cost strategy, and firm size.

2.4 Identification of high-quality measurement scales

Our high-quality scale is either a ratio/interval measure or a Likert-type scale with a Cronbach’s alpha of at least 0.7 (Nunnally 1978) that consists of at least three items. The last condition ensures that Likert-type scales will be reliable and that they will still hold a certain reserve for future studies in case one of the items does not load. Identification of such scales can assist the work of future researchers in the technology entrepreneurship and alert them to poor operationalization practices. Consequently, one of our study goals was to report on scales from meta-factors that were stable and reliable success factors for new technology ventures.

We selected only significant homogeneous (unmoderated) meta-factors from Table 2.2 or homogeneous subgroups from Table 2.3. This selection resulted in 11 strongly supported new technology venture success factors. To ensure that individual scales would perform well in further studies, within each meta-factor, we selected only scales with an observed correlation significant at the 0.05 level. Marketing experience did not have a significant high-quality scale in the previous studies. Therefore, we report high-high-quality scales found for 10 new technology venture success factors in Appendix A2.1. Further research should be

(35)

conducted on other potentially significant success factors (see moderated meta-factors from Table 2.2) before valid conclusions can be drawn.

2.5 Discussion and future research directions

In this study, we conducted a meta-analysis on antecedents of new technology ventures performance and tried to identify success factors for new technology ventures. To the best of our knowledge, this is the first systematic, quantitative effort to integrate the existing research on this topic. Our study sought to contribute to a more homogeneous theory of technology entrepreneurship. We summarize the results of our meta-analysis in Figure 2.1. In the spirit of meta-analysis, we present the results in four main blocks: significant and insignificant homogeneous factors, heterogeneous factors with moderators and heterogeneous factors without moderators. The latter two blocks are shown by the dotted lines. We also show within each block from which category a given meta-factor originates.

(36)

Figure 2.1. Summary of success factors in new technology ventures

(37)

The results are compelling: eight of the 24 meta-factors remain heterogeneous even after we searched for methodological moderators. They are evenly distributed across Market and Opportunity, and Resources categories. Five of the 24 meta-factors were homogeneous, but not significant. Three of them are from Market and Opportunity, and two meta-factors are from Entrepreneurial Team category. Only eight meta-factors are homogeneous and significant, suggesting that they are the only universal success factors for the performance of new technology ventures. The majority of them belong to the Resources group. Two meta-factors are success meta-factors for sub-groups in the population of new technology ventures and one works only for sales and not for profit-based performance. Therefore, more research is necessary on the heterogeneous, moderated meta-factors listed in Table 2.2. While we have identified some moderators in Table 2.3, future research should also explicitly test the effects of these moderators. To help build the body of the knowledge in technology entrepreneurship, we have also identified high-quality scales of the success factors and presented the measurement scales in Appendix A2.1.

2.5.1 Market and opportunity

Nine success factors represented the market and opportunity category in our meta-analysis. One was homogeneous and significant, five were heterogeneous, and the other three were insignificant. Therefore, we can conclude that based on extant research only one general factor, market scope, clearly enhances new technology venture performance. Moreover, we found only one success factor within the moderator subgroups. Product innovation improves new technology venture performance of corporate ventures, but it is detrimental for independent new technology ventures. A radical innovation strategy may be too risky for independent ventures, while corporate ventures can share risks with their parent companies.

Examining the number of heterogeneous meta-factors, one might conclude that the new technology venture population is generally too heterogeneous to examine the success factors. This idea was supported by the fact that for a number of meta-factors, no methodological moderators were found, suggesting that there may be other moderators that have not been reported in published research studies. As a follow-up of this research, we conducted 16 case studies of new technology ventures. We found a striking difference in the strategies used by these entrepreneurs dependent on their background—technical or business. In the first scenario, the entrepreneurs were usually the inventors of the venture technology

(38)

Meta-analysis of success factors 37 and focused on it rather than on its market. In the second scenario, entrepreneurs paid close attention to financials and the product market, while their ventures could do little or even no R&D, and yet these new technology ventures still produced high-technology products by following a sort of “me too“ strategy. Thus, the background of entrepreneurs leading to different involvement in the technological development of the products may be a missing moderator.

Another worthy direction for future studies is further contingency research. Until now, scholars in technology entrepreneurship have focused on product differentiation strategy and its interaction with different environmental characteristics, such as competition intensity and environmental dynamism (Li, 2001; Li and Atuahene-Gima, 2001; Zahra and Bogner, 2000). Other competitive strategies have received considerably less attention in studies of environmental contingencies.

Existing meta-factors describe opportunity in a rather indirect way. Thus, another possible direction of future research could focus more closely on opportunity, the key concept of entrepreneurship (Shane and Venkataraman, 2000). For example, opportunity sources vary in the amount of uncertainty and thus have different degrees of success predictability (Drucker, 1985; Eckhardt and Shane, 2003). Future researchers may want to consider a greater range of opportunity dimensions. The question of how to measure an opportunity remains open. In general, the technology entrepreneurship research does not address the multiple dimensions of the entrepreneurial opportunity concept and generally overlooks the interaction effects of the strategies followed by new technology ventures, opening these two themes to future researchers.

2.5.2 Entrepreneurial team

In our study, four types of experience described the characteristics of the Entrepreneurial Team: marketing, R&D, industry, and prior start-up experience. Only experiences in marketing and industry were significant, suggesting that acquiring more experience in these areas may lead to higher new technology venture performance. However, both prior start-up experience and R&D experience were insignificant at the 0.05 level. The former finding may be further evidence of overestimation of the role of prior start-up experience, ironically one of the most profound venture capitalist evaluation criteria (Baum and Silverman, 2004). It should be noted that the latter finding might have been caused by

Entrepreneurial risk-taking beyond bounded rationality : risk factors, cognitive biases and strategies of new technology ventures

Entrepreneurial risk-taking beyond bounded rationality : risk

factors, cognitive biases and strategies of new technology

ventures

Entrepreneurial risk-taking

beyond bounded rationality:

Risk factors, cognitive biases and strategies of

new technology ventures

Entrepreneurial risk-taking beyond bounded rationality:

Risk factors, cognitive biases and strategies of

new technology ventures

Table of contents

Acknowledgments

Chapter 1

General introduction

1.1

Meta-analysis of success factors

1.2

The mechanism of entrepreneurial risk-taking

1.3

Risk and uncertainty management strategies

Chapter 2

Meta-analysis of success factors

2.1

Introduction

2.2

Data collection and methodology

∑

∑

2.3

Analysis and results

2.4

Identification of high-quality measurement scales

2.5

Discussion and future research directions