ppl management31

191

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

sbr 66 april 2014 191–222

dominik steinkühler/matthias d. mahlendorf/malte brettel*

how self-JusTifiCaTion indireCTly drives esCalaTion of CommiTmenT – a moTivaTional perspeCTive**

absTraCT

self-justification is the most examined and empirically supported explanation of esca- lation of commitment. based on motivated reasoning theory, we argue that the need for self-justification affects escalation of commitment indirectly via other cognitive pro- cesses. we suggest that the need for self-justification represents a strong motivation for the continuation of a failing project. Thus, it influences the decision maker’s selec- tive perception, sunk cost effect, and overoptimism, which in turn foster escalation of commitment. we investigate escalation in the venture capital industry and thereby – in addition to our theoretical contribution – strengthen the external validity of previous studies in the laboratory.

Jel classification: m10, G24, G30.

Keywords: escalation of commitment; motivated reasoning; overoptimism; selec- tive perception; self-Justification; sunk costs.

1 inTroduCTion

Some of the most difficult decisions that executives face involve decisions on whether to abandon or continue key initiatives that are not succeeding. When decision makers are asked to re-evaluate a previously chosen course of action, they often tend to remain committed to it even when new information indicates that it should be terminated. This phenomenon, known as escalation of commitment, is defined as a “…tendency to be-

* Prof. Dr. Matthias D. Mahlendorf, Frankfurt School of Finance & Management, Sonnemannstraße 9–11, 60314 Frankfurt am Main, Phone: +49 (69) 154008-837, Fax: +49 (69) 154008-4837, E-Mail: [email protected]; Dr. Dominik Steinkühler, RWTH Aachen, Templergraben 64, 52056 Aachen, Germany, E-Mail: steinkuehler@ win.rwth-aachen.de; Prof. Dr. Malte Brettel, Professor, RWTH Aachen, Templergraben 64, 52056 Aachen, Germany, E-Mail: [email protected]

** This paper is based on Dominik Steinkühler’s PhD thesis, completed at RWTH Aachen. We would like to thank the editor, one anonymous reviewer, and the participants of the session at the Academy of Management Confer- ence 2009 in Chicago for their valuable feedback.

d. sTeinKühler/m. mahlendorF/m. breTTel

192 sbr 66 april 2014 191–222

come locked in to a course of action, throwing good money after bad or committing new resources to a losing course of action” (Staw (1981)). Thus, instead of terminating failing projects, decision makers escalate their commitment by continuing to invest resources (money, time, work, etc.) and “…persist beyond an economically rational point” (Staw and Ross (1987, 41)). The phenomenon of escalation of commitment is not limited to any particular area; it has been well documented in a wide variety of contexts, including information technology projects (Keil (1995); Keil, Mann, and Rai (2000); Montealegre and Keil (2000)), large-scale construction projects (Ross and Staw (1993)), banking (Mc- Namara, Moon, and Bromiley (2002); Staw, Barsade, and Koput (1997)), employee per- formance appraisal (Bazerman, Beekun, and Schoorman (1982)), sports events (Camerer and Weber (1999); Staw and Hoang (1995)), and new product development (Boulding, Morgan, and Staelin (1997); Schmidt and Calantone (2002)).

In this paper, we make two contributions to the literature on escalation of commitment. We provide a new perspective on how self-justification leads to escalation; and, following calls in the literature (e.g. Bobocel and Meyer (1994, 363); Staw (2005, 229)), we add to the limited number of studies that investigate escalation outside the laboratory.

Staw (1976) was the first to discuss the escalation of commitment. Since then, escala- tion has attracted sustained interest among researchers who, over time, have proposed a plethora of drivers of escalation tendencies in decision making. Self-justification is the most examined and empirically supported explanation for the escalation of commitment. However, many studies investigate only the direct effect between these two variables. In contrast, based on motivated reasoning theory (Kunda (1990)), we argue that the need for self-justification affects escalation of commitment indirectly via other cognitive pro- cesses. We suggest that the need for self-justification represents a strong motivation for the continuation of a failing project, and thus influences the decision maker’s selective perception, sunk cost effect, and overoptimism, which in turn foster escalation of com- mitment. Establishing mediators between the need for self-justification and escalation shifts our understanding of the role of self-justification. Although many studies see the need for self-justification as a main driver of escalation of commitment, we argue that self-justification alone is not sufficient to explain escalation in professional settings. Con- tinuing investments because of “…decision makers’ unwillingness to admit that their prior allocation of resources to the chosen course of action was in vain” (Brockner (1992, 39)) would not be acceptable for either the organization or for any decision maker with a sense of responsibility. Therefore, we suggest that the effect of the need for self-justifi- cation on escalation manifests through three other cognitive processes that we consider in this study. This perspective moves research beyond the predominant single-variable explanations and adds to the understanding of the escalation phenomenon and the direct and indirect effects of its drivers.

Most research on the escalation of commitment has used laboratory experiments. Al- though there is no doubt that experiments are preferable for testing causal relationships, experiments often deviate substantially from real corporate decisions, for example with

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

193sbr 66 april 2014 191–222

respect to the time span of the decision process, the expertise of the decision maker, his or her ability to learn, the complexity of the decision task, and incentives that may mitigate escalation in organizational decision making (Guler (2007); Mahlendorf and Wallenburg (2013)). Hence, Staw (2005, 229) critiques the state of research by pointing out that “… there is no guarantee that the variables manipulated in the laboratory have captured the reality of escalation”. In contrast, we investigate the escalation of commitment based on survey data on real venture capital investments. Doing so allows us to cross-validate the results of previous laboratory experiments and to provide support concerning the external validity of prior research.

The venture capital industry provides an attractive setting for studying escalation of com- mitment because the potential for escalation in this setting is high (Guler (2007); Ryan (1994, 1995)). Venture capitalists routinely must decide whether to continue investing resources (both monetary and non-monetary) in risky ventures: “The staging of rounds of funding and the reinvestment decisions […] provide an ideal setting to examine the potential for escalation of commitment” (Birmingham, Busenitz, and Arthurs (2003, 220)). Additionally, the investment process is relatively consistent across firms (Gomp- ers and Lerner (2004)). In our study, we refer to the process of a venture capitalist firm’s initial investment in a portfolio company and its subsequent support and follow-on fi- nancing rounds as a “project.” In the venture capital context, escalation of commitment implies that venture capitalists fail to terminate underperforming projects in due time. To study the influence of different escalation drivers, the empirical design of our study focuses on the perspectives of individual decision makers and their involvement in the decision on whether to continue or terminate a project.

The paper is structured as follows. In Section 2, to provide the background for our re- search model we review studies on the escalation of commitment and the motivational side of self-justification. In Section 3, we develop the hypotheses. In Section 4, we outline our method and describe our sample, measures, and methods of empirical analysis. In Section 5, we present the results of our statistical analysis. Section 6 concludes with a discussion of our findings, their managerial implications, and suggestions on possible avenues for further research and limitations.

2 liTeraTure review

2.1 selF-JusTiFicaTion

Self-justification is the most examined and empirically supported explanation of escalation of commitment (Brockner (1992); Sleesman, Conlon, McNamara, and Miles (2012)). The mechanism of self-justification, first discussed by Staw (1976) in the context of es- calation, is rooted in the theory of cognitive dissonance (Festinger (1957)). This theory proposes that people have a strong need to reduce sources of cognitive inconsistency and seek to justify the rationality of past decisions both to themselves and to others. The idea

d. sTeinKühler/m. mahlendorF/m. breTTel

194 sbr 66 april 2014 191–222

was triggered by the U.S. involvement in Vietnam. However, it is applicable to many situations in which people have invested money or effort into a failing course of action, as Staw (2005, 218) so cogently summarizes:

“Like the Vietnam War effort, people may invest in stocks, careers, or even marriages, and when these investments do not pay off, they may not necessarily withdraw from the situation. Instead, people may actually invest further so as to turn the situation around – to prove that their prior decision was indeed and accurate ore appropriate one.”

In other words, the concept of self-justification posits that decision makers may experi- ence psychological discomfort from cognitive dissonance as a direct result of negative feedback on previous investment decisions (Staw and Ross (1987)). To rationalize their past decisions and reaffirm the correctness of their initial investments, decision makers will continue to commit resources to a failing course of action and thus avoid the psycho- logical costs of failure (Staw and Ross (1978)).

To illustrate this stream of research, we refer to a study conducted by Zhang and Baumeis- ter (2006). In their laboratory experiment, undergraduate students took part in a game in which they had a chance to win a jackpot. To continue playing, participants had to periodically invest more money. The experiment tested the ego-threat condition and the control condition. In the ego-threat condition, participants’ need for self-justification was increased by telling them “If you’re the kind of person who usually chokes under pressure or if you don’t think that you have what it takes to win the money, then you might want to play it safe [i.e., quitting the game to avoid escalation]. But it’s up to you.” The results indicated that students in the ego-threat condition lost significantly more money than did those in the control condition. In other words, in a situation that increases the need for self-justification, people have a strong tendency to escalate their commitment.

2.2 selecTiVe percepTion

The concept of selective perception has long since been established as a factor that influ- ences decision-making. As early as 1954, Hastorf and Cantrill (133) stated that “…out of all occurrences going on in the environment, a person selects those that have significance for him from his own egocentric position.”

In the context of escalation of commitment, Biyalogorsky, Boulding, and Staelin (2006) show that one important driver of escalation is biased belief updates. In their experiment, MBA students initially receive case information about a new product, the history of the market, competitors, etc. Based on this information, they have to decide whether the or- ganization should invest $2.5 million to improve and relaunch the product. In the second part of the case, students receive negative new information about the net present value forecast for the investment. The results of Biyalogorsky et al. (2006) indicate that the students do not correctly update their beliefs based on the new information. Instead, their

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

195sbr 66 april 2014 191–222

final evaluation of the investment is strongly influenced by their initial beliefs. Hence, selective perception seems to have occurred.

Keil, Depledge, and Rai (2007) conducted a laboratory experiment with undergraduate students that investigated selective perception concerning software quality and its subse- quent effects on escalation. Those students who were asked to play the role of marketing staff showed a considerably lower perception of software quality problems than did a comparison group that was assigned the position of quality assurance staff. This selective perception led to a higher escalation of commitment, measured as a single item that asked the students to recommend whether the project should continue as planned.

2.3 sunk cosT eFFecT

A considerable amount of evidence shows that individuals take sunk costs into account when faced with decisions on continuance or termination of a project (Arkes and Blumer (1985); Dilts and Pence (2006); Garland and Newport (1991); Garland, Sandefur, and Rogers (1990); Moon (2001)). Sunk costs are defined as resources that have already been irrevocably committed and cannot be recovered. According to classical economic and normative decision theory, sunk costs are irrelevant for project continuation decisions, which should be made only on the basis of incremental gains or losses (Bonini (1977); Howe and McCabe (1983); Robichek and Van Horne (1967)). However, in the context of escalating projects, the studies mentioned above show that once resources such as money, effort, or time have been invested, decision makers show a greater tendency to continue a course of action (Arkes and Blumer (1985); Astebro, Jeffrey, and Adomdza (2007)). This behavior is referred to as the “sunk cost effect.”

For example, Astebro et al. (2007) investigate data from investment decisions that re- ceived feedback from the Inventors’ Assistance Program at the Canadian Innovation Cen- tre. This feedback comprised substantiated advice on whether to continue or terminate efforts. The Astebro et al. (2007) analysis suggests almost a one-to-one relation between resource commitments before receiving the advice to stop and the resource commitment afterwards. In other words, those who had already spent, for example, $10,000, when receiving the feedback to terminate the investment, spent another $10,000 after having received this advice. Thus, sunk costs seem to increase escalation of commitment.

2.4 oVeropTimism

Overoptimism, i.e., the tendency of decision makers to systematically overestimate the probability of good performance and underestimate the probability of bad performance (Heaton (2002)), has been demonstrated in several different areas (Camerer and Lovallo (1999); Camerer and Weber (1999); De Meza and Southey (1996); March and Shapira (1987); Weinstein (1987)) and is accepted as a robust decision making bias. In the con-

d. sTeinKühler/m. mahlendorF/m. breTTel

196 sbr 66 april 2014 191–222

text of the escalation of commitment, overoptimism leads decision makers to systemati- cally overestimate the probability of success and therefore to terminate failing projects too late (Ross and Staw (1993); Staw (1997); Staw and Ross (1987)).

In the escalation literature, we can distinguish two different perspectives on optimism. One perspective is interested in optimism as a personality trait, i.e., being an optimistic person; the other perspective focuses on optimism concerning specific project or busi- ness outcomes, i.e., being optimistic with regard to a specific future event. Studies such as that by Carver, Scheier, and Segerstrom (2010, 879) that take the first perspective are interested in the effect of generalized favorable expectancies for the future on escalation tendencies. In this vein, Astebro et al. (2007) find that above-average optimists spend significantly more time and money on projects on which they have received advice to cancel than do below-average optimists. An example for the second perspective is the study by Mahlendorf and Wallenburg (2013), which provides evidence that the relation between public justification requirements and investments in failed projects is moderated by optimistic outcome expectations. The study suggests that, public justification increases investments in failing projects in the case of low optimistic outcome expectations, but not in the case of highly optimistic outcome expectations.

3 hypoThesis developmenT

From our review of literature, we draw two main conclusions that pertain to escalation commitment.

First, when we consider methods, we see a need to complement the findings from experi- ments in the laboratory with data from the field. Decision making in organizations might be affected differently from individual biases than would be decisions in the laboratory (Guler (2007)). For example, organizations can apply a set of monitoring and incentive mechanisms that promote economically rational decisions. They can also hire experts and establish learning routines that might improve rational behavior beyond those observed in the laboratory. And decision makers who are facing suboptimal performance during actual practice will usually gather more evidence, such as expert advice or additional analysis of performance indicators (Schultze, Pfeiffer, and Schulz-Hardt (2011)). Taken together, these arguments suggest that the experimental studies should be complemented with evidence from the field.

Bobocel and Meyer (1994, 363) summarize the trade-off that exemplifies many studies on escalation conducted with experiments in the laboratory:

“Because our primary aim was to disentangle previously confounded variables, we sought to maximize internal validity in the present research. […] Nevertheless, the precision gained in the present paradigm comes at the expense of external validity. It is thus imperative that field research be conducted to assess the generalizability of our findings.”

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

197sbr 66 april 2014 191–222

Second, we believe that more refined models are necessary to link the disparate expla- nations for escalation of commitment. Since Staw’s (1976) paper, many theories have been proposed on what drives escalation of commitment “…almost to the point that any mechanism used to predict social behavior in general […] has become a viable candidate for explaining behavior in escalation situations” (Staw and Ross (1987, 42)). Most of the research focuses on the influence of single factors on escalation. This approach nei- ther embraces the multidetermined nature of escalation of commitment (Staw (1997)) nor helps to advance the understanding of the relationships among escalation drivers. In fact, “…to further advance our understanding a more integrative approach is needed to construct and test a richer model of the [escalation] phenomenon that includes causal pathways between constructs” (Keil et al. (2007, 393)).

Motivated reasoning theory suggests that motivation affects reasoning “…through reli- ance on a biased set of cognitive processes” (Kunda (1990, 480)). Accordingly, the mo- tives of a decision maker strongly influence his or her cognitive processes (Kunda (1987); Pyszczynski and Greenberg (1987); Sorrentino and Higgins (1986)). Kunda (1990, 480) suggests that “…any wish, desire, or preference that concerns the outcome of a given reasoning task” may bias “…strategies for accessing, constructing, and evaluating beliefs”. For example, defense motives induce selective processing in favor of the desired conclu- sion (Ditto and Lopez (1992)). The motivation to invoke a certain impression will also bias systematic processing towards satisfying the motivational goal (Agrawal and Ma- heswaran (2005)).

Given that cognitive dissonance theory on which self-justification is based is “fundamen- tally motivational in nature” (Elliot and Devine (1994, 382)), it may not be sufficient to consider just a direct behavioral effect of the need for self-justification, as has been done in prior research. In fact, dissonance theory explicitly proposes that individuals can re- duce dissonance by changing their cognitions (Festinger (1957)). Hence, decision makers who feel the need to justify their failed decisions may attempt to cope through a variety of cognitive defense mechanisms (Sivanathan, Molden, Galinsky, and Ku (2008)). We propose that the need for self-justification motivates people to arrive at the conclusion that the respective project should not be abandoned. Self-justification will bias their cognitive processes and thus indirectly influence escalation through affecting aforementioned cog- nitive escalation drivers. Figure 1 illustrates our suggested mediation model.

As noted above, the effect of selective perception on escalation of commitment has primarily been investigated in laboratory experiments. Among the few insights from the field are the qualitative results from Guler (2007) and Ryan (1994). In her study on escalated venture capital projects, Guler (2007, 258) argues that “…investors who evaluated a venture early and worked with it over time might fail to spot critical prob- lems in a timely manner”. However, the correct perception of portfolio companies’ per- formance and challenges is crucial, because these aspects determine whether a venture capitalist will continue to support or abandon a portfolio company (De Clercq and Sapienza (2006)).

d. sTeinKühler/m. mahlendorF/m. breTTel

198 sbr 66 april 2014 191–222

Figure 1: Graphical Representation of the Research Model

In a similar vein, describing her case study results, Ryan (1994, 77–78) reports that “… persistent investors were able to maintain a rigid focus on the sequence of variables that they believed would result in the achievement of bright promise. They did so by selec- tively and, at times, positively perceiving critical events in the company’s history. These biased interpretations in turn reinforced escalaters’ belief in a successful outcome”. To cross-validate the results from the laboratory and the qualitative studies, we reinvestigate the following hypothesis:

H1. With increasing selective perception, project escalation increases.

In the venture capital industry, the sunk cost effect often gives rise to what are called “protective investments”, which Guler (2007, 259) defines as those that are made by investors to recover their initial investments and “…demonstrate an inability to disregard sunk investments”. Investors have a strong desire not to walk away from a company and take a certain loss but instead will invest further funds to keep the company alive (Ryan (1995); Valliere and Peterson (2005)). Building on these previous studies, we attempt to cross-validate the following hypothesis:

H2. With an increasing sunk-cost effect, project escalation increases.

As described in the literature review, some studies on escalation analyze overoptimism in the sense of a general personality trait; others focus on project specific overoptimism. We take the latter perspective, because, as we note below in hypotheses 5c and 5d, project- specific overoptimism can be influenced by motivational aspects such as self-justification.

Two case studies examine escalated venture capital projects. Guler (2007) provides qua- litative evidence that overoptimism occurs as one potential driver of escalation in the

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

199sbr 66 april 2014 191–222

venture capital setting. Guler shows that several interview partners believe that venture capital firms can achieve turn arounds with follow-up investment. In her qualitative field study Ryan (1994) identifies decision makers’ overoptimism as a source of enduring per- sistence and continued financing of failing ventures. Thus, we reinvestigate the following hypothesis:

H3. With increasing overoptimism, project escalation increases.

Similar to the other decision-making biases introduced above, some research suggests that the need for self-justification influences venture capital decision makers (Guler (2007); Ryan (1994, 1995)). According to Ryan (1995, 340), investors often continue to support a failing venture “…to buy time [which] is consistent with self-justification explanations of escalation behavior”, thus justifying the correctness of their initial resource allocations. Apart from this qualitative evidence, most research on self-justification uses the experi- mental setup originally introduced by Staw (1976). However, a study by Schulz-Hardt, Thurow-Kroning, and Frey (2009) questions whether the effects of self-justification on escalation in this type of experiment are driven by preference for the initial decision. Hence, it seems desirable to cross-validate the following hypothesis using a different method:

H4. With increased need for self-justification, project escalation increases.

Previous research usually looks at direct effects on escalation of commitment, but we propose an indirect relation between the need for self-justification and escalation of com- mitment based on motivated reasoning theory (Kunda (1990)) and “…relying on the prevalent view of motivation as having its effects through cognitive processes” (Jain and Maheswaran (2000, 358)).

One way to reduce the cognitive dissonance associated with failing projects is to avoid situations in which negative information might become salient. “[W]hen dissonance is present, [… the decision maker] will actively avoid situations and information which would likely increase the dissonance” (Festinger (1957, 3)). Prior experimental studies show that decision makers with a need to self-justify become increasingly selective in their search for information and will ignore negative feedback (Caldwell and O’Reilly (1982); Conlon and Parks (1987); Schmidt and Calantone (2002); Staw, Sandelands, and Dut- ton (1981)). Furthermore, managers who must cope with self-threats may apply selective cognitive effort when analyzing information (Ditto, Scepansky, Munro, Apanovitch, and Lockhart (1998)), pay selective attention to the reliability of the received information (Doosje, Spears, and Koomen (1995)), and discredit the ability of others who attempt to communicate information on a failing course of action (Staw and Ross (1987)). Hence, we hypothesize that a need for self-justification is positively related to selective percep- tion.

H5a. With increasing need for self-justification, selective perception increases.

d. sTeinKühler/m. mahlendorF/m. breTTel

200 sbr 66 april 2014 191–222

Defensive motives, such as the need for self-justification, increase people’s reliance on heu- ristics to support their desired conclusion (Giner-Sorolla and Chaiken (1997)). Thus, the need for self-justification may strengthen “…powerful judgmental heuristics, such as ‘waste not…’ on the sunk cost problems” (Klaczynski (2001, 305)). Additionally, when choosing between continued investment and project termination, decision makers may be “…chang- ing the attractiveness of the alternatives” (Festinger (1957, 42)). In the context of project escalation, the perceived attractiveness of alternatives often inversely relates to the non- retrievable prior investments (i.e., sunk costs). Thus, decision makers may pay increased at- tention to the sunk costs of the project to lessen the “attractiveness” of a termination (Arkes and Blumer (1985); Conlon and Leatherwood (1989)). Therefore, we hypothesize:

H5b. With increasing need for self-justification, the sunk-cost effect increases.

In addition to reducing the attractiveness of termination, decision makers can also in- crease the attractiveness of continuation, thus mitigating cognitive dissonance. They could, for example, consider any setbacks as only temporary and “…continue their belief in the bright promise of future gains” (Ryan (1994, 105)), i.e., engage in overoptimism. More generally, the need for self-justification increases the desirability of a positive project outcome. This effect in turn increases desirability bias, which is one cause for overopti- mism (Krizan and Windschitl (2007)). Accordingly, we expect that the need for self- justification affects overoptimism. We express this conjecture in the following hypothesis:

H5c. With increasing need for self-justification, overoptimism increases.

So far, few studies have rigorously addressed the influence of self-justification on the cognitive processes of decision makers who escalate commitment, but the considerations we mention above suggest potentially important relationships. In fact, re-examining the roots of self-justification in cognitive dissonance theory and its motivational relation to a decision maker’s cognitive processes may even open up an entirely new perspective on the mechanism through which the need for self-justification exerts an influence on esca- lation of commitment. The key tenet of self-justification reflected in Hypothesis 4 is that decision makers will actively, and therefore consciously, escalate their commitment to rationalize past decisions, as classically conceptualized in the respective literature. How- ever, reinterpretations of cognitive dissonance theory (for reviews see Cooper (2007); Harmon-Jones and Harmon-Jones (2007)) on which self-justification is based may cast doubts on this direct causal effect. For example, the self-consistency model of cognitive dissonance proposes that people hold expectations for competence and morality derived from “…the conventional morals and prevailing values of society” (Thibodeau and Ar- onson (1992, 592)). Dissonance is aroused when a discrepancy is detected between those self-expectancies and a behavior. Because people can generally be expected to hold posi- tive self-expectancies, incompetent, immoral, or irrational behavior will thus create disso- nance (Harmon-Jones and Harmon-Jones (2007)). Similarly, without including concepts of the self, the “New Look” model of cognitive dissonance proposes that dissonance is triggered whenever behavior is inconsistent with general societal or normative standards

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

201sbr 66 april 2014 191–222

of competent or moral behavior (Cooper and Fazio (1984)). The self-standards model, which Maertz, Hassan, and Magnusson (2009, 68) describe as “…the most comprehen- sive [and] integrative view of cognitive dissonance to date”, adopts those views and pro- poses that any behavioral inconsistency can provoke dissonance based on either societal normative standards or uniquely personal standards of competence and morality (Stone and Cooper (2001)).

Considering these advances of cognitive dissonance theory and motivated reasoning, the classic conceptualization of self-justification seems questionable. Simply yielding to the need for self-justification and actively escalating commitment to a losing course of action clearly constitutes an irrational and non-normative act, which would thus arouse further dissonance. Since, according to the theory, the decision maker will avoid situations that increase dissonance, he or she should not be expected to behave in this manner. However, a different mechanism may account for the influence of the need for self-justification on escalation of commitment. This mechanism may have gone unnoticed so far due to limitations of prior studies, which do not examine potential causal relationships between different drivers of escalation. According to classic cognitive dissonance theory, a deci- sion maker may adapt cognitions to reduce dissonance (Festinger (1957)). Therefore, the cognitive “defense mechanisms” described in Hypotheses 5a-c may actually transmit the influence of the need for self-justification on project escalation, i.e., the need for self-justification influences project escalation through the cognitive biases introduced in this study. In this respect, Kunda (1990, 480) states that “…this view of motivation as cognitively mediated has always been integral to the understanding of dissonance reduc- tion”. Research on motivated reasoning, which frequently proposes that motivation trans- mits its influence through cognitive processes by determining which cognitive processes will be used on a given occasion in order to arrive at a desired conclusion, supports this perspective (Jain and Maheswaran (2000); Kunda (1987); Molden and Higgins (2005); Pyszczynski and Greenberg (1987); Sorrentino and Higgins (1986)). These mechanisms allow decision makers to maintain what Kunda (1990, 483) calls the “illusion of objectiv- ity,” which describes their personal belief that the conclusion at which they arrived (in the present case, to continue a project) is objectively the right one, even though the cognitive processes that led to it were actually biased. We summarize the theoretical considerations in the following hypothesis:

H5d. The cognitive processes selective perception, sunk cost effect, and overoptimism (at least partially) mediate the effect of the need for self-justification on project escalation.

4 meThod

4.1 sample and daTa collecTion

Sample. To test the proposed model, we conducted an online survey among European venture capital professionals. In the first step, we chose companies active in venture capi-

d. sTeinKühler/m. mahlendorF/m. breTTel

202 sbr 66 april 2014 191–222

tal financing located in Europe based on the member directories of the European Private Equity and Venture Capital Association (EVCA) as the target population. Our second step was to study the homepage of each venture capital firm as well as press announce- ments on portfolio investments. Doing so enabled us to identify the most appropriate executives within each company who were involved in making the investment decisions and who were thus in a position to provide meaningful information on the decision pro- cesses (Campbell (1955)). This procedure eventually led to a contact database of 2,058 venture capital professionals.

Survey design and administration. To develop our multiple item-based measurement in- strument, we followed the recommendations of Churchill (1979), DeVellis (2003), Dia- mantopoulos (2005), Nunnally (1978), and Rossiter (2002).

We conducted an extensive literature review to identify, wherever possible, measures that would appropriately capture the constructs we wished to include in our study. For several constructs, we had to develop new scales. In the development of an initial item pool, we also relied on prior case studies on escalation behavior and used content analysis to detect phrases frequently used by decision makers in relation to the escalation mechanisms stud- ied in this paper. We conducted several pretests to improve the measurement instruments. These tests included an item-sorting procedure (Anderson and Gerbing (1991); Hinkin (1995)) performed by 17 raters; content validity ratings (Lichtenstein, Netemeyer, and Burton (1990); Zaichkowsky (1985)) by 24 raters; expert interviews with six academics and professionals; a large scale pretest of the survey with 287 German technical manag- ers that we used to validate and purify the measurement models (Homburg and Giering (1996)); and seven additional interviews with venture capital professionals to assess the adequacy of the item formulations in the venture capital context.

During data collection, we followed the procedures prescribed by Dillmann (2000). We gathered our data by sending out an email containing a personalized link to our online survey instrument to 2,058 participants in our contact database. We instructed respon- dents to complete the questionnaire for a single project that failed (for a similar ap- proach see Dilts and Pence (2006)). Our reason for focusing on failed projects was that escalation of commitment is explicitly defined as continuing a failing course of action (Staw (1976)). This approach produced 182 valid responses after the initial email and two subsequent reminders. An additional 300 respondents stated either that company policies prohibited the participation in surveys or that they had no experience with a failed venture capital investment. Because of missing values, we excluded five data sets from the further analysis, leaving 177 valid responses in our sample, which implies a net response rate of 8.6%.

We checked for nonresponse bias by comparing early and late respondents using Mann-Whitney U Tests (Armstrong and Overton (1977)). We found no significant differences.

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

203sbr 66 april 2014 191–222

4.2 measures

We measure the variables in this study as latent variables with multiple indicators. Unless otherwise indicated, all items are measured on a 7-point Likert scale. Based on the crite- ria proposed by Jarvis, MacKenzie, and Podsakoff (2003), all constructs in this study are specified as reflective. Appendix 1 presents details of all measures.

Selective perception. Selective perception describes a systematic tendency of decision mak- ers to filter out or downplay negative incoming information and to focus on feedback that conforms to their positive thinking. Similar to Keil et al. (2007), we develop four items that measure whether decision makers focus their attention on certain areas of informa- tion, neglect information that contradicts their views, or both.

Sunk-cost effect. The sunk-cost effect describes the effect of making the decision to con- tinue a project based on investments made in the past. To measure this construct we use five items that capture the extent to which the presence of sunk costs opposes project termination in the perception of the decision maker. Our items draw specifically on the operationalization of the sunk-cost effect in Keil et al. (2000) and on phrases that other researchers, in their studies on the sunk-cost effect, find are typical for decision makers (Arkes and Blumer (1985); Keil, Truex, and Mixon (1995)). Based on Guler (2007), we include an additional item to capture the notion of protective investments in the venture capital industry.

Overoptimism. The effect of overoptimism leads decision makers to systematically over- estimate the probability of success and thus believe in an ultimately positive outcome. Hence, negative developments are considered, unwarrantedly, as only temporary or man- ageable, and decision makers believe that the project will eventually develop as expected. We develop five items to tap into those notions and capture the decision makers’ attitudes towards negative events as well as their outcome expectancy for the project.

Need for Self-justification. According to Keil et al. (2000), the direct assessment of psycho- logical self-justification is difficult. Therefore, we follow their advice and limit ourselves to the assessment of “…observable behaviors that could serve as reliable indicators of the [...] need to self-justify” (Keil et al. (2000, 641)) and measure the factors that influence the perceived need of the decision maker to justify her/his decisions. For our five item scale, we adapt one item from Keil et al.’s (2000) need for social self-justification scale and leverage the ten-item measure of decision involvement from Biyalogorsky et al. (2006).

Project escalation. Project escalation is characterized by two key features: the continued investment of resources in a failing project that leads to economically irrational delayed project terminations. Our operationalization of project escalation evolves around these two characteristics. Due to the complexity of the construct, we operationalize project escalation as a second-order construct of Perceived Project Escalation, Perceived Resource Investment, and Invested Resources.

d. sTeinKühler/m. mahlendorF/m. breTTel

204 sbr 66 april 2014 191–222

Perceived Project Escalation taps directly into the key escalation characteristic of delayed project termination. Hence, it captures whether the project went on for too long. We measure this feature by using a three-item scale that reflects the ex-post assessment of the decision maker about the delay of the project termination (Mahlendorf (2013)). A sample item is “It was held on to the project too insistently.” The two other constructs, Perceived Resource Investment and Invested Resources, relate to the other key characteris- tic of escalation, the continued investment of resources that should have been avoided. Consistent with our understanding of escalation, delayed project termination should be manifested in a “growth in the cumulative amount of resources invested over time” (Keil et al. (1995, 372)). Therefore, borrowing from project management literature (Ashley, Lurie, and Jaselskis (1987); Dvir, Raz, and Shenhar (2003); Pinto and Slevin (1988)), the three items assess Perceived Resource Investment as the cumulated overinvestment of the relevant resources time, work effort, and money over the course of the project as perceived by the decision maker. A sample item is “Too many financial resources were invested in the project.” Additionally, building on Dilts and Pence (2006) and Mahlendorf and Wal- lenburg (2013), we also measure Invested Resources, a more objective measure of project escalation, by using three items, and compare the actual resource investment (time, work effort, and money) to the initial investment planned at the start of the project. This ap- proach follows the logic that the initial decision to invest was made based on a certain return calculation and that escalating projects will show increasing budget overruns, mak- ing continued investment uneconomical. Budget overruns are an indicator of negative developments. Hence, the items capture investments despite negative information, which is one of the key characteristics of escalation of commitment. The formulation is “How do the resources invested into the project until its termination and/or exit compare to the expenditures expected at the initiation of the project?” with separate answer options for man-days, costs, and length of the project ranging from “1 = much less than expected” to “7 = much more than expected”.

In the absence of an objective measure of escalation of commitment in real investment projects, we believe that this second-order construct is a useful proxy for escalation of commitment in failed projects. Moreover, we believe that this construct addresses esca- lation in greater detail than do prior measures (e.g., Armstrong, Williams, and Barrett (2004); Ku (2008); Ku, Malhotra, and Murnighan (2005)).

Control Variables. Our analysis includes six control variables. We include the relative size of the investment compared to the average investment per portfolio company to control for possible effects of financing size on the level of commitment of the venture capital firm (Dilts and Pence (2006)). In addition, to capture potential effects of hierarchical position, we use a dummy variable that takes the value of one if the respondents posi- tion is at the top level. We also use age of the venture capital firm as a control variable, because control systems develop over time (Aerts (2005); Davila (2005); Davila and Fos- ter (2007)) and might reduce escalation behavior. Further, since larger venture capital companies might apply more formal controls, so we include company size measured as capital under management (cumulated over all funds, in million euros). Additionally, we

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

205sbr 66 april 2014 191–222

included an industry group dummy variable for the portfolio company (1 = manufactur- ing/production; 0 = services and other) to capture general industry effects (Birmingham et al. (2003)). And we control for the role of the venture capital firm (non-lead investor, co-lead investor, or lead investor) to control for syndication effects that may bias the re- sults (Birmingham et al. (2003)).

4.3 meThod oF analysis

To test the hypothesized relationships depicted in Figure 1, we use the partial least squares (PLS) technique (cp. e.g. Berson, Oreg, and Dvir (2008); Cording, Christmann, and King (2008); Cron, Gilly, Graham, and Slocum Jr. (2009)), using SmartPLS 2.0 M3 (Ringle, Wende, and Will (2005)). To evaluate the significance of parameter estimates, we implement a bootstrapping procedure with 1,000 samples, as recommended (Chin (1998)).

5 resulTs

5.1 eValuaTion oF The measuremenT model

Table 1 provides a descriptive analysis of the means, standard deviations, ranges of the constructs, and the correlations between them.

We assess the reliability and validity of the measurement models following the procedures for PLS analysis recommended by Chin (1998). Out of all items, we eliminate only two, due to low factor loading. The subsequent calculation of Cronbach’s alpha (threshold 0.7), composite reliability (threshold 0.7), and average variance extracted (threshold 0.5) shows that the constructs exceed the thresholds for all criteria (see Appendix 1), with only one exception (i.e., the construct invested resources slightly misses the threshold for Cronbach’s alpha with a value of 0.62). Overall, our results indicate satisfactory reliability for the constructs.

We also analyze the discriminant validity on the item and the construct level. We follow Fornell and Larcker (1981), who argue that a construct shows sufficient discriminant va- lidity when it shares more variance with its measures than with other constructs. Table 1 suggests satisfactory discriminant validity for all constructs. We then test item discrimi- nant validity by examining whether all items share more variance with their own con- struct than with any other construct. The matrix of factor cross-loadings in Appendix 2 shows that each item’s loading is higher on its assigned construct than on the other con- structs, which supports adequate discriminant validity.

d. sTeinKühler/m. mahlendorF/m. breTTel

206 sbr 66 april 2014 191–222

Table 1: Descriptive Statistics and Construct-Level Correlations

Correlations and Square Root of Average Variance Extracted (AVE)

Variables 1 2 3 4 5 6 7

1. perceived project escalation 0.91

2. perceived resource investment 0.62 0.81

3. invested resources 0.35 0.59 0.75

4. selective perception 0.23 0.20 0.20 0.83

5. sunk cost effect 0.16 0.25 0.26 0.28 0.82

6. overoptimism 0.19 0.20 0.21 0.32 0.41 0.83

7. need for self-Justification 0.20 0.16 0.07 0.20 0.32 0.25 0.85

Summary Statistics

Number of items 3 3 3 4 4 4 5

Mean 4.75 4.62 4.97 3.40 3.42 4.36 3.02

standard deviation 1.85 1.90 1.40 1.60 1.82 1.48 1.83

Min 1 1 1 1 1 1 1

Max 7 7 7 7 7 7 7

note: Table depicts square root of ave on the diagonal and correlations on off-diagonal.

5.2 common meThod Bias consideraTions

Although the use of key informants is common in research, self-reported data are poten- tially vulnerable to common method biases (Podsakoff and Organ (1986)). Surveying decision makers on failed projects is an emotionally charged topic that could be subject to social desirability bias. We attempt to reduce this problem by implementing some of the procedural remedies (e.g., response anonymity) recommended by Podsakoff, MacKenzie, Lee, and Podsakoff (2003). Additionally, we collect data from two particular groups of informants, key decision makers who were personally involved in the decisions about project continuation and termination and thus evaluated their own decisions, and from non-decision makers who evaluated the decisions of another key-decision maker. This procedure makes it possible to compare the answers of the two groups of respondents and thus detect significant differences between the informants. This classification led to 130 key-decision makers and 47 non-decision makers. We conduct an ANOVA for all items, which shows that the groups of informants differ significantly only on one out of 28 indicators. In addition, we conduct statistical analyses to assess the potential influence of common method bias. First, we perform Harman’s single-factor test (Podsakoff and Organ (1986)). The results of the principal component factor analysis show that no single factor emerges and no single factor accounts for the majority of the covariance among

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

207sbr 66 april 2014 191–222

the measures. Additionally, following Podsakoff et al. (2003) and Williams, Edwards, and Vandenberg (2003), we include a common method factor in the PLS model (see Liang, Saraf, Hu, and Xue (2007) for a detailed description). Appendix 3 shows that the average variance explained by the substantive factors is 0.698, while the average variance ex- plained by the method factor is 0.005. Hence, the ratio of substantive to method variance is about 137:1. Given the invariant answering behavior of key-decision makers and non- decision makers, as well as the small magnitude and nonsignificance of method variance, we contend that common method bias is unlikely to be a serious concern for this study.

5.3 eValuaTion oF The sTrucTural model and TesTing oF hypoTheses

To evaluate the structural model, we estimate the coefficient of determination R², the Stone-Geisser-Criterion Q² (Geisser (1975)), and the path coefficients with their respec- tive t-values, in PLS. Table 2 presents the results for the overall sample.

Table 2: Results of PLS Analysis

Hypothesized Path Standardized Path Coefficient t-value

selective perception → project escalation 0.14 1.86**

sunk cost effect → project escalation 0.14 1.71**

overoptimism → project escalation 0.10 1.46*

need for self-justification → project escalation 0.07 1.18

need for self-justification → selective perception 0.20 2.46***

need for self-justification → sunk cost effect 0.32 4.23***

need for self-justification → overoptimism 0.25 4.02***

Controls

relative investment size → project escalation 0.09 1.42*

role of venture capital firm → project escalation –0.08 1.42*

hierarchical level → project escalation 0.01 0.22

industry → project escalation 0.03 0.64

Venture capital firm size → project escalation –0.12 1.66**

Venture capital firm age → project escalation 0.08 0.47

* p < 0.10, ** p < 0.05, *** p < 0.01 (one-tailed).

Considering the fact that the escalation mechanisms discussed in our paper are only just a few among the many factors influencing project escalation (other factors include, for example, organizational and external effects, see Staw (1997, 2005)), and that the empiri- cal design of our study focuses on individual decision makers, leaving aside the influence

d. sTeinKühler/m. mahlendorF/m. breTTel

208 sbr 66 april 2014 191–222

of the decisions of other people involved in the project, the R² value for project escalation of 0.16 indicates satisfactory explanatory power of our model. Thus, our data substantiate the importance of the escalation mechanisms. In addition, the positive Q² value of 0.06 suggests that our model has predictive relevance (Chin (1998)), lending further support to the important role of the escalation mechanisms.

Our data support seven out of the eight research hypotheses proposed in the paper. All three cognitive escalation drivers, selective perception, the sunk-cost effect, and overop- timism, are significantly related to project escalation, thus supporting Hypotheses 1 to 3.

Hypothesis 4, which postulates a direct influence of the need for self-justification on pro- ject escalation in addition to the mediating effects, is not supported. However, Hypo- theses 5a-5c, which postulate the indirect influences of the need for self-justification, are strongly supported. To further explore this mediated relationship between the need for self-justification and project escalation, we utilize the recommendations by Preacher and Hayes (2008) for assessing indirect effects in multiple mediator models and also fol- low the procedure laid out by Iacobucci, Saldanha, and Deng (2007) for the analysis of mediated relationships in structural equation models using the product-of-coefficients approach.

As can be seen from the z tests (Sobel (1982)) in Table 3, the indirect effects of the need for self-justification on project escalation are significant. Given that the direct effect of the need for self-justification on project escalation is not significant, this result indicates that its influence is fully mediated by the three cognitive processes selective perception, sunk-cost effect, and overoptimism. Hence, Hypothesis 5d is supported.

We also test for mediation by following the causal-steps approach recommended by Baron and Kenny (1986), and find the same results. In the direct-effect model (with- out mediators), the need for self-justification is significantly related to project escalation (path coefficient = 0.19; p < 0.01), fulfilling the first condition according to Baron and Kenny. As Table 3 shows, the other two conditions of a significant relation between the independent and the intervening variables, and between the intervening variables and the dependent variable, are also fulfilled. Full mediation occurs because after introduction of the intervening variables the need for self-justification has no significant effect on project escalation.

To test the robustness of our results, we calculate three additional models. In each of these models, we drop one of the three mediators (selective perception, sunk-cost effect, and overoptimism) from the analysis. In all three models, the significance of the effects of the two remaining mediators on project escalation becomes stronger. Moreover, the direct effect of need for self-justification on project escalation becomes marginally significant (p < 0.1, one-tailed), indicating partial mediation instead of full mediation. These results provide support for the robustness of our model and suggest that all three mediators are relevant.

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

209sbr 66 april 2014 191–222

Table 3: Results of Mediation Analysis

Mediated Relationship Mediator z-value

need for self-Justification → project escalation

selective perception 1.38*

sunk cost effect 1.65**

overoptimism 1.40*

* p < 0.10, ** p < 0.05, *** p < 0.01 (one-tailed).

We find the following results for the control variables. Venture capital firm size has a significant negative relation with project escalation (path coefficient = –0.12, p < 0.05, one-tailed). Hence, larger companies might have more effective mechanisms to reduce escalation; for example, more formal project evaluation procedures. Relative investment size is marginally and significantly related to project escalation (path coefficient = 0.09; p < 0.1, one-tailed). This result is in line with the sunk-cost effect and supports the no- tion that the larger the size of invested resources, the higher the tendency to terminate a project too late. The role of the venture capital firm has a negative marginally significant relation with project escalation (path coefficient = –0.08; p < 0.1, one-tailed). This result suggests that lead investors are somewhat less susceptible to escalation compared to non- lead investors. All other control variables (i.e., hierarchical level, industry and venture capital firm age) do not have a significant impact on project escalation.

6 disCussion

Our study examines the relationship between project escalation and several factors that prior literature identifies as important drivers of escalation. Analyzing the effects of dif- ferent escalation drivers simultaneously, we address the multidetermined nature of esca- lation of commitment. We are able to assess their direct and indirect effects and move beyond the predominant single variable explanations of escalation. Based on motivated reasoning theory (Kunda (1990)), in this study we provide what we believe is a significant conceptual contribution: we take a motivational perspective on the influence of the need for self-justification on escalation of commitment and propose a mediation by cognitive processes.

Our tests support our theoretical predictions and results from laboratory experiments on the direct effects of selective perception, sunk-cost effect, and overoptimism on es- calation. Hence, the empirical data from this survey in the venture capital industry provide evidence on the external validity of the former research results. This finding is important, because it suggests that companies are not yet able to overcome individual biases, for example, by means of monitoring, incentives, and employing experts. Fur- thermore, our data support the hypothesis that the need for self-justification strongly influences the three cognitive escalation drivers included in our research model. Although self-justification has long been considered the main driver of escalation of commitment

d. sTeinKühler/m. mahlendorF/m. breTTel

210 sbr 66 april 2014 191–222

(Brockner (1992)), we find no direct effect on project escalation. Instead, our results suggest that cognitive processes fully mediate the influence of self-justification. Thus, our empirical data support the predictions derived from motivated reasoning theory. Thus, the need for self-justification represents a motivation to continue an ambiguous project. This motivation in turn affects reasoning “…through reliance on a biased set of cognitive processes” (Kunda (1990, 480)). Selective perception, the sunk-cost effect, and overop- timism are powerful mechanisms through which the need for self-justification indirectly drives escalation of commitment. To satisfy their need for self-justification, decision mak- ers utilize cognitive defense mechanisms that distort their assessment of the situation and eventually lead them to believe that continued investment is the appropriate decision. Recent findings support the view that escalation is strongly rooted in individual percep- tion (Biyalogorsky et al. (2006); Keil et al. (2007)), which is also consistent with classic dissonance theory: a decision maker with a need to self-justify will reduce dissonance by adapting his or her cognitions (Festinger (1957)).

6.1 implicaTions For managemenT

Our study has implications for management practices. First, our results support earlier suggestions that managers should implement practices to reduce the need for self-justi- fication. Self-justification might be even more important than previously assumed, be- cause our analysis suggests undesirable effects of self-justification on selective perception, sunk-cost effects, and overoptimism. Hence, reducing the need for self-justification is key for avoiding not only deliberate opportunistic behavior, but also subconscious cognitive coping strategies. For example, to reduce the need for self-justification, managers might consider leadership rotation (Boulding et al. (1997); Schmidt and Calantone (2002)). This procedure could be helpful, because a newly assigned leader has no share of the responsibility for initiating the original investment and thus has a lower pressure for self- justification. (However, we note that sharing decision authority can also lead to greater levels of escalation under certain conditions (Sleesman et al. (2012)).)

The indirect mechanism proposed by our study has a second important implication for de-escalation efforts in managerial practice. Even if the need for self-justification cannot be completely avoided, its effects on escalation behavior can be reduced by decreasing the mediator‘s selective perception, sunk-cost effect, and overoptimism. To mitigate selective perception, managers should receive explicit advice on how to search for disconfirm- ing evidence, i.e., information that suggests it would be wise to cancel the investment (Kadous and Sedor (2004)). To reduce the sunk-cost effect, alternative investment pos- sibilities should be highlighted (Keil et al. (1995)). To avoid overoptimism, (Lovallo and Kahneman (2003)) suggest that managers could apply the advocatus diaboli (Devil’s Ad- vocate) concept.

how selF-JusTiFicaTion indirecTlY drives escalaTion oF commiTmenT

211sbr 66 april 2014 191–222

6.2 implicaTions For FuTure research

Our analysis suggests that cognitive processes mediate the relationship between the need for self-justification and escalation of commitment. Future studies may wish to investi- gate these mediating effects in more detail, perhaps by going back to the laboratory to untangle the interactions. In any case, we can endorse the skepticism of Staw (2005), which we introduced at the beginning of this paper, that prior laboratory research might have failed to capture the complexity of real-world escalation. For a better understanding of the escalation phenomenon, future research might consider scenarios that are more re- alistic and which cater to the apparent interactions between different drivers of escalation. In addition, while our model is a static one, dynamic approaches may generate valuable insights on how the effects of different mechanisms evolve over time as a function of the level of commitment.

Schulz-Hardt et al. (2009) find that a decision maker’s preferences may be an underes- timated driver of escalation. Since preferences represent one facet of motivation (Kunda (1990)), motivated reasoning theory may have to offer additional insights concerning Schultz-Hardt et al.’s findings. Like our study, future investigations might explore wheth- er cognitive processes, such as selective perception, the sunk-costs effect, and overopti- mism also mediate the relation between preferences and escalation.

6.3 limiTaTions

Although we integrate several explanations of escalation of commitment into our frame- work, we cannot claim that our model gives a full understanding of this complex phe- nomenon. We limit our model to four mechanisms that have repeatedly been shown to influence escalation of commitment in experimental contexts. However, there are ad- ditional escalation drivers, and they might be included in subsequent research. We also highlight that our model focuses on psychological drivers of escalation. In real-world set- tings, a lot of aspects influence termination decisions, e.g., organizational determinants, external factors, etc., and researchers suggest that many of these promote escalation of commitment (Staw (1997); Staw and Ross (1987)).

A limitation of this study might be its focus on the venture capital industry, which is a highly specialized industry. However, investment decisions are one of the primary ac- tivities of venture capital firms (Fenn, Liang, and Prowse (1997); Gompers and Lerner (2004)) whose managers have both the experience and incentives to maximize profits. Additionally, venture capitalists are usually distanced from the operations of their in- vestments and thus are more likely to assess them objectively compared to managers in other organizations. Hence, the results of our study may represent the upper bound on the quality of termination decisions and should, therefore, be generalizable beyond the venture capital industry.

d. sTeinKühler/m. mahlendorF/m. breTTel

212 sbr 66 april 2014 191–222

Another limitation of our study might be its focus on the perspective of an individual decision maker. To the extent that the decision to continue or terminate a project is made by a group, there may be other escalation mechanisms besides the individual cognitive processes included in our research model (Bazerman, Giuliano, and Appelman (1984)). However, in group decision-making, it still is important to pay attention to the individual decision makers, as the decision of the group reflects the characteristics and biases of its members (Hambrick and Mason (1984)). Even more important, in our venture capital context, investment decisions are typically made in an investment committee in which each partner can veto any deal (Guler (2007)), so that each decision maker needs to make an individual decision on whether to continue or terminate a project. Still, group effects in escalation situations are a promising avenue for further research, and one which could provide insights into the discussion on whether the individual biases are reinforced (Whyte (1993)) or negated within the group (Smith, Tindale, and Steiner (1998)).

appendix 1

Construct Measurement

Constructs and their respective indicators Cronbach’s Alpha

Compo- site Reli- ability

AVE Load- ing

Selective Perception 0.85 0.90 0.69

if new information on the project appeared, …

1. … i paid more attention to information, which supported my opinion on the project.

0.76

2. … i paid less attention to information that contradicted my view.

0.88

3. … i only turned my attention to certain information (i.e., i was subject to selective perception).

0.86

4. … i focused my attention on my“pet subjects”. 0.82

Sunk Cost Effect 0.84 0.89 0.68

1. i viewed termination of the project as a waste of resourc- es, due to the irrecoverable investments already made.

0.82

2. i argued for the continuation of the project, because if terminated, the previous investments would have been in vain.

0.83

3. The more resources had been invested, the more expen- sive a termination of the project seemed to me.

elimi- nated

4. by continuing the project, i wanted to protect my previ- ous investments.

0.82

5. i regarded a termination of the project as a destruction of prior invested resources.

0.81

Copyright of Schmalenbach Business Review (SBR) is the property of Fachverlag der Verlagsgruppe Handelsblatt and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.