We’re only human - An exploratory study of biases and strategic problem formulation performance

ABSTRACT

This study looks at how decisions are made during early-stage innovation, specifically how managers, teams, or team leaders formulate problems on which to focus. Problem formulation processes put organisations on paths that incur sunk costs and influence subsequent problem solving and innovation outcomes. Varying human biases may impair these processes. Therefore, factors that have the potential to mitigate biases play a key role in determining problem formulation performance. This exploratory study looks at one of these factors by examining the role of awareness of bias. It offers theoretical insight into and examines empirically the relationships between bias awareness and bias intensity, as well as between bias intensity and problem formulation performance. Using the problem as the unit of analysis, and examining original survey datasets gathered from the US, China, and Finland, we find that two bias types are particularly prone to influence strategic problem formulation: Solution jumping as a cognitive bias and dominance as a motivational bias show negative relationships with problem formulation performance. We also find that unlike other biases, dominance bias appears unaffected by the degree of bias awareness. The insights from our study shed light on how formulating novel problems is subject to the influence of cognitive, motivational, and informational biases.

KEYWORDS:

• Bias
• awareness
• strategic problem formulation
• cognition
• problem solving
• microfoundations
• innovation

Introduction

Innovation processes follow a path constrained by patterns of existing solutions and the underlying choices of strategic innovation problems (Thrane et al., 2010). An individual problem can be framed in different ways, leading to distinct innovation outcomes. A mistake in formulating a problem may result in increased problem-solving costs, missed opportunities, or incorrect resource allocation (Ramaprasad & Mitroff, 1984; Von Hippel & Von Krogh, 2016). Conversely, formulating a good problem – a problem that grounds valuable solutions and is difficult for others to solve efficiently – may yield notable benefits (Baer et al., 2013; Becker et al., 2005; Cowan, 1986; Mohaghegh & Furlan, 2020; J. A. Nickerson et al., 2007; Terwiesch, 2008; Von Hippel & Tyre, 1995). Despite the importance of formulating good problems, conversations often focus on problem-solving rather than problem formulation – the act of identifying fundamental, uniquely valuable challenges for which alternative innovative solutions can be generated (Morais-Storz et al., 2019; J. Nickerson & Argyres, 2018). Underlying factors affecting problem formulation outcomes thus remain poorly understood.

Existing research recognises that within vast solution spaces for innovative problem solving, individuals in different roles adopt different ways to approach problems and opportunities (Hookway et al., 2019). The same argument can be applied to problem formulation. To gain benefits and avoid excessive costs, addressing organisational or team issues during problem identification is essential (Deichmann et al., 2021; Grass et al., 2020). Managers, teams, and team leaders routinely make decisions about relevant problems, but these decisions may not be unbiased (J. A. Nickerson et al., 2017). Cognitive, motivational, and informational biases – impediments stemming from attributes of choices, choosers, or the environment in which choices take place – exert a distorting effect on problem formulation (Baer et al., 2013; Barnes, 1984; Van Knippenberg et al., 2015). They limit individuals’ ability to understand what is truly happening in a given situation, and to recognise and seize upon feasible candidates for good problems to subsequently solve (Deichmann et al., 2021; Mumford et al., 2006; J. A. Nickerson et al., 2007). As part of the strategic planning process, strategic problem formulation is inherently influenced by the limits to the abilities of managers to use and process information (Barnes, 1984). In practice, managers often pay little attention to the quality of processes that guide the initial direction for problem resolution (Korte, 2003).

Recognising the downsides from various individual biases, the existing literature includes insights on how to address them (Baer et al., 2013; Bhardwaj et al., 2018; Mohaghegh & Furlan, 2020; J. A. Nickerson et al., 2017). Seminal works on decision-making and individuals’ information processing highlight bounded rationality and limited managerial attention as (Barnes, 1984; Cyert & March, 1992; Tversky & Kahneman, 1974) and, hence, conscious analytic assessment is found to be relevant (Mumford et al., 2006). The increasing richness of information has been identified to burden the attention capacity of individuals, teams, and organisations (Ocasio, 2011; Van Knippenberg et al., 2015). This applies not only to information used in the problem formulation as an input but also to the knowledge of problem formulation practices. Managers may not be aware of impediments – biases–influencing problem formulation. Awareness of bias emerges as a relevant factor that can reduce bias intensity – the presence and degree of biases (Hurmelinna-Laukkanen & Heiman, 2012). However, as little is known about the microfoundations of problem formulation (Baer et al., 2013), there also is limited understanding of how much influence such awareness has on various types of biases.

In this study, we posit that bias awareness, as reflected in individual- and group-level decisions, relates to the intensity of cognitive, motivational, and informational biases, which, for their part, have effects on problem formulation outcomes (and beyond) (see Banerjee et al., 2019; Mohaghegh & Größler, 2020). Within this framing, we focus on solution jumping (cognitive bias), dominance (motivational bias), and information distortion (informational bias). We also examine perceptual bias to gain additional information on the similarities and differences of individual cognitive biases. We examine how managers’ awareness of the existence of biases during problem formulation activities influences the intensity of varying biases, and what role different biases play in problem formulation performance.

We build on previous discussions of biases in managerial decision-making (e.g., Barnes, 1984; Cyert & March, 1992; Tversky & Kahneman, 1974) and strategic problem formulation (e.g., J. A. Nickerson & Zenger, 2004; J. Nickerson et al., 2012). More specifically, we build on Baer et al. (2013) who developed a framework for strategic problem formulation that combines governance- and process-oriented models and recognises cognition and environmental issues as relevant to problem formulation. Following calls for comparative studies across continents (Karam & Ralston, 2016), we analyse data obtained from firms in three countries – the US, China, and Finland – to explore the topic empirically.

Our findings show that cognitive, motivational, and informational biases may not be equally relevant to problem formulation performance. Information distortion (an informational bias) does not show a similar negative relationship with problem formulation performance as solution jumping (a cognitive bias) and dominance (a motivational bias). We also find that though bias awareness generally mitigates bias intensity, some biases persist. The findings from our exploratory analysis suggest that dominance is a particularly challenging, difficult-to-remedy bias. These provisional findings contribute to the previous theorising on managerial decision-making and problem formulation.

Foundations of strategic problem formulation – literature overview

Research on strategic management and decision-making in organisations (Barnes, 1984; Cyert & March, 1992; Van Knippenberg et al., 2015) recognises that the creation of new knowledge allows the conversion of various inputs into valuable outputs (Nelson & Winter, 1982). This activity entails solving relevant problems in novel ways (see Macher, 2006), including looking at precursors to problem solving and considering practices deployed to select a problem for which solutions are then sought (Baer et al., 2013). A manager’s job is not simply to solve well-defined problems but, on an ongoing basis, to identify potentially valuable new problems to solve (Pounds, 1969).

While putting problems under a lens is not a new approach (Smith & Shaw, 2019; Volkema, 1986), a focus on problem solving rather than problem formulation persists (Mohaghegh & Furlan, 2020), suggesting that the latter remains under-examined despite its influence on problem solving and its outcomes. As a topic of research, problem formulation has appeared in different streams of literature. For example, Simon (1973), Kiesler and Sproull (1982), Ramaprasad and Mitroff (1984), Cowan (1986), M. Lyles (1987, 2014), J. A. Nickerson and Zenger (2004), and J. Nickerson and Argyres (2018) have addressed aspects of problem formulation. M. A. Lyles and Mitroff (1980) were among the first to empirically establish the importance of problem formulation in the strategic management literature. Problem formulation is considered a relevant part of innovation (e.g., Erez et al., 2019) – usually an antecedent for successful problem solving (Baer et al., 2013; cf. von Hippel & von Krogh, 2016).

In the evolution of scholarly work, taking a problem as a unit of analysis has rendered the study of problem formulation more microanalytical (Deichmann et al., 2021; J. A. Nickerson & Zenger, 2004). For example, Cowan (1986) took the discussion from an organisational to an individual level. Capturing the notion of limits to problem formulation, several studies address factors that influence problem formulation outcomes (e.g., Baer et al., 2013; Heiman et al., 2009; Hurmelinna-Laukkanen & Heiman, 2012; Mumford et al., 2006; J. A. Nickerson et al., 2017). Theorising clearly recognises the limits to decision makers’ capabilities (Barnes, 1984; Van Knippenberg et al., 2015). Problem formulation activities involve cognitive, motivational, and informational impediments (J. A. Nickerson et al., 2007). Such biases may lead to a search for problems in wrong places or along suboptimal paths (Baer et al., 2013). The presence of biases is natural, but their prevalence in decision-making processes is a notable concern (Korte, 2003). Effects of prior experiences on perceived outcome probabilities, overconfidence, and preference for certainty over uncertainty are some aspects among the attributes of decision makers, decisions, and environment that may influence the understanding of which problems are worth solving (Barnes, 1984; Baer et al., 2013).

Managerial attention (Ocasio, 2011) is needed to find and employ the right ways to realise the potential residing in problems. Without awareness of biases potentially distorting problem formulation, it is difficult to mitigate their effects (Baer et al., 2013; Hurmelinna-Laukkanen & Heiman, 2012). Knowledge and understanding of biases generally mitigate their harmful effects and lead to better decisions (Korte, 2003). However, Korte (2003) also notes that action to reduce biases’ effects matters, raising the question of whether awareness is sufficient to improve problem formulation performance.

Existing research suggest a need to examine problem formulation processes more closely. Prior studies do not fully explicate the roles of bias awareness and intensity in problem formulation (Hurmelinna-Laukkanen & Heiman, 2012; Mohaghegh & Furlan, 2020). We look at the relevance of bias awareness on intensity of different biases. We explore solution jumping (a cognitive bias), dominance (a motivational bias), and information distortion (an informational bias), as well as perceptual bias (a different cognitive bias) as biases that have been judged relevant (e.g., Baer et al., 2013; Van Knippenberg et al., 2015). The following section discusses our proposed logic in detail.

Bias awareness, bias intensity, and problem formulation performance

Building on conceptual studies (Baer et al., 2013; Hurmelinna-Laukkanen & Heiman, 2012; J. A. Nickerson et al., 2017), we explore how managerial awareness of biases is associated with bias intensity (Stage 1 of our proposed logic), and how bias intensity is associated with problem finding performance (Stage 2).

Baer et al. (2013) suggest that performance is a function of managerial organisational decisions made in response to cognition-altering factors, such as attributes of the problem environment, problem characteristics, and team/individual attributes. Following the literature on discrete governance choices (Williamson, 1996), we treat high problem formulation performance to reflect well-aligned managerial organising (governance) choices. We define strategic problem formulation performance as the extent to which managers successfully formulate problems that (1) are relevant to the firm/project/team, (2) have appropriate scale and scope given the available resources and capabilities, (3) are potentially profitable, and (4) are important for the (future of the) organisation (Björkdahl & Holmén, 2016; J. A. Nickerson et al., 2007).

We define bias intensity as the degree to which a bias is present in an organisation or team. Each bias has a distinct character and corresponding effect, depending on the personnel (teams and individuals) and situations (Baer et al., 2013; J. A. Nickerson & Zenger, 2004).

Whether individuals are generally aware of their cognitive limits and existence of biases affects their intensity and effects (Barnes, 1984; Banerjee et al., 2019). We define bias awareness as the extent to which individuals or teams consciously realise that problematic biases exist. Following Korte (2003), we suggest that a general awareness of bias(es) may have broad effects. We thus explore the effects of general bias awareness on each bias’s intensity. This reflects situational awareness, i.e., being on the lookout for biases (e.g., Adams et al., 1995).

Impact of bias awareness on bias intensity

We suggest that successfully addressing biases necessitates an awareness of the existence of issues. While scholars in strategic management have often neglected an explicit consideration of bias awareness, the few exceptions share similar insights. Curley et al. (1986) addressed awareness of biases in decision-making processes from a psychological perspective. They found that awareness of potentially biased choices improved decision-making quality. Korte (2003) stated, in the context of human resource development, that awareness comprises the initial step in dealing with the effects of biases. Subsequent understanding of biases and corresponding actions to minimise effects lead to better decisions. Bias awareness has also been found to matter for problem solving related to investment decisions (Cowan, 1986). Building on these insights, we expect higher bias awareness to generally ease mitigation of bias intensity. However, we also suspect that variation exists between different types of biases.

While previous literature has identified numerous individual biases (Baer et al., 2013; Heiman et al., 2009; J. A. Nickerson et al., 2007), we examine a specific set of biases, looking at different bias types: cognitive, motivational, and informational. We focus on (1) dominance – a motivational element reflecting the degree to which excessive influence exists, (2) solution jumping – a cognitive factor indicating the degree to which management or teams limit search processes in problem finding, and (3) information distortion – an informational factor encompassing the degree of wrong, missing, or incomplete information a team or individuals experience. We also investigate (4) perceptual bias – a cognitive factor capturing the personal and professional experience of an individual. As cognitive factors are a seminal bias type (see Barnes, 1984), we look at both managerial aspects and individual-level issues in examining the differences and similarities between individual biases.

Dominance

Collaborative teams typically generate multiple ideas about how to structure a problem (Grass et al., 2020; Mohaghegh & Furlan, 2020). Dominance obstructs such activity, reflecting incentives for individuals with a commitment to a particular view to engage in sub-goal pursuit (Borchardt et al., 2022; Brusoni, 2005; Vidal & Möller, 2007) rather than working towards a team’s goal (Barnes, 1984; M. Lyles, 1987). Individuals may be motivated to seek approval of their approach rather than value maximation (see Williamson, 1996). Baer et al. (2013) suggest that dominance arises from heterogeneity of objectives, driven by differences between high-stakes and low-stakes players and the costs and benefits of advocacy. As Montibeller and Von Winterfeldt (2015) suggest, motivational biases can entail distortions of judgements and decisions – emerging as uneven use of power and strong push towards certain directions – driven by self-interest, organisational context, or social pressures. For example, senior managers may be motivated not to change a status quo serving their interests, resulting in bias for risk-sensitive decisions (see Barnes, 1984). An individual may introduce an idea, but its further development requires others’ assistance (Franke & Shah, 2003). Dominance-driven processes serve specific individuals’ interests (Hodgkinson & Sadler-Smith, 2018), leaving those with lower motivation in the background. Dominance stemming from different motivations can give rise to uneven behaviours in promoting ideas and may impair participation of other team members, confounding voluntary processes and diminishing trust.

Awareness of the presence of a dominant individual involved in problem formulation may help diminish the dominance-based behaviours. Underscoring the importance of bias awareness, Mohaghegh and Größler (2020) note that high employee participation can be promoted via managerial awareness and associated leadership action. For example, a team may be initially created or later reconstituted to exclude managers or members known to engage in dominant behaviour. A challenge to remediating dominance exists, however: despite awareness, power differences are often difficult to remove completely in collective activities (Clegg et al., 2006). Nevertheless, we posit that increasing awareness of biases should reduce the intensity of dominance bias.

Solution jumping

Existing studies establish the tendency to simplify the search for and integration of information in decision-making (Baer et al., 2013; Barnes, 1984). This idea maps well to our construct of solution jumping, wherein limited exploration takes place, and there is a rapid movement towards certain outcomes. Solution jumping is a bias related to the idea of ‘plunging-in’ (Bhardwaj et al., 2018), which leads to the premature termination of search for good problems to solve (Baer et al., 2013). Cyert and March (1992) discuss satisficing: searching for problems until a ‘good enough’ problem is discovered. Limiting search processes is a natural response to turbulent business environments requiring quick decisions (Nutt, 2002), and resource-intensity of extensive information gathering and storage (Anderson et al., 2001). For example, team members may only consider technologies and challenges emerging in a specific knowledge area where they already have some experience, and thus decide to focus there without considering more unfamiliar areas. Although an economising behaviour in organisations (Winter, 2000), satisficing behaviour stemming from solution jumping as a cognitive bias constrains environmental scanning and monitoring (Cowan, 1986; Korte, 2003).

We interpret solution jumping as a primarily harmful bias. Awareness of solution jumping bias may facilitate more thorough search processes. We suspect the solution jumping bias to be of lower magnitude as bias awareness increases.

Information distortion

Attributes of information affect managerial organisational choices (Barnes, 1984). Information distortion bias includes reliance on outdated, missing, irrelevant, confused, incorrect, and/or disorganised information (Mohaghegh & Furlan, 2020; J. A. Nickerson et al., 2007). When knowledge and information become impaired, conscious, unconscious, or associational mechanisms may emerge and cause errors (Mumford et al., 2006). Information distortion is derived from needs to reduce the work of evaluating new information, distinguishing between alternatives, and achieving consistency between old and new information (Russo et al., 2008).

Information distortion bias can be mitigated by expertise and knowledgeability (Carley & Lin, 1997). We expect that awareness of information being (potentially) distorted affects the bias intensity. For example, the use of experts to remediate information distortion may facilitate identifying distorted information and deploying appropriate corrective action. Hence, we expect that awareness of bias varies negatively with the intensity of information distortion bias.

Perceptual bias

Another cognitive factor, perceptual bias, offers a broader insight into the role of individual-level biases. Accompanying solution jumping among cognitive and sociological factors, perceptual bias stems from self-protection, desire for certainty, and ‘similar-to-me’ behavioural tendencies (Barnes, 1984; Becker et al., 2005; Payne, 2006; Ramaprasad & Mitroff, 1984). Symptoms of perceptual bias include a lack of understanding of the narrowness of existing knowledge and ignorance of (potentially) poor understanding of project feasibility and subsequent payoffs; notably, different ideas are not seen as valuable (Ahuja et al., 2013; Deichmann et al., 2021). We define perceptual bias as a function of individuals’ distinct cognitive frames constructed from professional and personal social experiences (DiMaggio & Powell, 1983). As a result of perceptual bias, different types of knowledge breed psychological isolation, even when people are in frequent contact (Litchfield & Gentry, 2010).

Payne (2006) notes that individuals are often unaware of their projections of values and attitudes. The diversity of a team plays a role in cognitive processes associated with information surfacing and combination (Martins & Sohn, 2022). In diverse teams, managers’ abilities to manage socio-political processes and their own frames are relevant in strategic problem formulation (M. Lyles, 1987). Our final expectation for the first stage of our proposed logic suggests that managerial awareness (partly) mitigates the intensity of perceptual bias.

Impact of bias intensity on strategic problem formulation performance

Existing (largely theoretical) research establishes that biases influence problem formulation processes and associated outcomes (e.g., Baer et al., 2013; Heiman et al., 2009; J. A. Nickerson & Zenger, 2004). In line with J. A. Nickerson and Zenger (2004), we suggest that bias intensity negatively affects problem formulation performance.

Todorova and Durisin (2007, p. 782) argue for the distorting effects of dominance in organisations’ activities, noting that ‘powerful actors within and outside the organisation may influence knowledge absorption processes to achieve their goals’. Similarly, Clegg et al. (2006) connected dominance with negative outcomes. We expect dominance bias intensity to be negatively associated with problem formulation performance.

Excessive or otherwise unwarranted solution jumping also negatively impacts problem formulation performance. When search processes are exhaustive or nearly so, satisficing is low; more and increasingly diverse alternatives emerge from a problem search (Ramaprasad & Mitroff, 1984). In discussing managerial problem sensing, Kiesler and Sproull (1982) suggest that exhaustive examination and evaluation allow managers to see problems they would otherwise miss. For example, time constraints and tensions between the need for long-term innovation and quarterly financial performance may influence managers to act prematurely to choose a problem, leading to suboptimal outcomes (Amabile, 1988). Hence, we posit a negative association between intensity of solution jumping bias and problem formulation performance.

Information distortion bias has been shown to have a variety of negative consequences (Russo et al., 2008). A problem based on wrong or distorted information may lead to project failure, a forced return to searching, or the abandonment of a project. Established preferences may lead to the distortion of new information in favour of the preferred alternative (Russo et al., 1996). For problem formulation, information distortion bias can be thus expected to limit performance.

Perceptual bias similarly presents challenges to problem formulation performance. An efficient search for problems is easily corrupted because of difficulties with accepting others’ views (Nickerson et al., 2017). While large differences in people’s training and backgrounds are a source of genuinely new combinations of ideas and knowledge, a degree of overlap and shared meaning is needed (Ramaprasad & Mitroff, 1984; Soda & Bizzi, 2012). We expect that perceptual bias decreases the quality of problem formulation performance.

Empirical evidence from three continents

Methods and data

Since theory for problem formulation is still developing and empirical evidence is scant, we chose to take an exploratory approach in our study (see Faems, 2020). The data for the study were collected from the US, China, and Finland. The US and China data were gathered concurrently, while the data for Finland were gathered about 6 months earlier, using a self-administered survey. The time lag is not an issue for this study. The US/China data gathering instruments were dedicated only to this study, while in Finland, the survey instrument was embedded in a larger data gathering effort involving multiple research agendas. As a result, some differences exist between the surveys. To be methodologically conservative, we did not combine the Finnish data with the US and China data but analysed them in two parts. Separate analyses were additionally warranted owing to differences in respondents’ positions: while the US and China respondents were project leaders and team members, Finnish respondents were R&D directors and other senior managers. Although sample bias is a possibility for our US/China convenience samples (practically, random sampling was not feasible), the samples are diverse and broadly representative of their respective populations (see Table 2 below). The Finnish sample consists of a simple random sample of firms.

Table 1. Industry control dummy variables.

Industry	Variable Name	Dataset
Manufacturing	Manufacturing	US/China and Finland
Service	Service	US/China and Finland
Information Technology	InfoTech	US/China and Finland
Education	Education	US/China and Finland
Electronics	Electronics	US/China
Medical	Medical	US/China
Real Estate	RealEstate	US/China and Finland
Trade	Trade	US/China and Finland
Consumer Goods Manufacturing	ConsumerGoodsMfg	Finland
Other Manufacturing and Repair	OtherMfgRepairInstall	Finland
Transport/Storage	TransportStorage	Finland

Table 2. Summary statistics.

	US and China					Finland
Variable	n	Mean	St.d.	Min	Max	n	Mean	St.d	Min	Max
ProblemFormulationPerf	376	0.421	0.844	−3.50	3.00	191	0.003	0.841	−2.05	1.78
DominanceBias	377	1.553	0.864	0.00	4.00	189	−0.005	0.746	−1.94	2.01
SolutnJumpBias	361	1.274	0.934	0.00	4.00	189	−0.007	0.797	−1.76	2.09
InfoDistortionBias	372	−2.530	0.960	−4.00	0.00	198	0.004	0.778	−1.50	2.30
PerceptualBias						203	0.000	0.637	−1.87	1.66
Awareness	367	2.514	0.931	0.00	4.00	202	−0.002	0.571	−1.75	1.24
SearchBreadth	350	2.351	1.068	0.00	4.00
KnowledgeTacitness	293	2.624	0.781	0.00	4.00
InformalIP	302	2.655	0.754	0.00	4.00
HybridIP						154	−0.007	0.869	−1.45	2.38
FormalIP	304	2.761	0.762	0.00	4.00
Position	360	1.547	1.193	0.00	3.00
FirmAge	340	26.335	32.229	3.00	201.00	304	27.530	26.010	2.00	112.00
NumberOfEmployees	330	12440	48402	1.00	400000	305	848.10	2580.3	9.00	28246.0
CountryChinaUS	457	0.381	0.486	0.00	1.00
CompetitiveEnvironment						320	−0.003	0.876	−2.24	1.04
Manufacturing	360	0.139	0.346	0.00	1.00	335	0.310	0.463	0.00	1.00
Service	360	0.331	0.471	0.00	1.00	335	0.060	0.237	0.00	1.00
InfoTech	360	0.119	0.325	0.00	1.00	335	0.084	0.277	0.00	1.00
Education	360	0.050	0.218	0.00	1.00	335	0.048	0.214	0.00	1.00
Electronics	360	0.042	0.200	0.00	1.00
Medical	360	0.100	0.300	0.00	1.00
RealEstate	360	0.019	0.138	0.00	1.00	335	0.042	0.200	0.00	1.00
Trade	360	0.058	0.235	0.00	1.00	335	0.122	0.328	0.00	1.00
ConsumerGoodsMfg						335	0.218	0.413	0.00	1.00
OtherMfgRepairInstall						335	0.036	0.186	0.00	1.00
TransportStorage						335	0.042	0.200	0.00	1.00

Base condition for industry dummy variables (all dummies = 0) is ‘Other Industry’.

The survey instrument made use of primarily Likert-type scale items. The survey was developed first in English and then translated into Chinese and Finnish using back translation to enhance accuracy. Respondents were asked to respond based on a completed project with which they were familiar. The decision to focus on a specific project was made based on a pilot study: When developing survey, we initially relied on existing literature to generate initial groups of items for the constructs. To establish face validity, we used a panel of three experts (business strategy professors) in the US to review and critique questions, and a pilot effort in Finland, comprised of interviews with nine key managers from six firms. These steps allowed us to gain meaningful insights into our survey’s validity and initially assess the reliability of our constructs. Experts consistently informed us that focusing respondents on specific projects rendered our questions comprehensible. The refined questionnaire was reviewed and approved by our experts before launching the survey.

For the survey administered in China, we contacted 340 people and in the US, 225 people. We achieved a 90% response rate (305 responses) in China and an 83% response rate (187 responses) in the US. The Finnish population of firms consisted of 570 organisations. The criteria for inclusion were having more than 100 employees and the authority to make R&D decisions. 335 responses reflect a 58.8% response rate.

To deal with potential common method bias issues, we followed multiple suggestions from Podsakoff et al. (2003). These included proximal separation, reverse items, psychological separation, and removing ambiguous terms. Potential same-source issues for independent/dependent variables were outweighed by the opportunity to obtain data from knowledgeable sources on an otherwise empirically elusive topic. Owing to culture-specific attributes of our three sampled countries (the US, China, and Finland), local adaptations of the survey added built-in diversity of variable measurement, helping mitigate the possibility of common method bias (Podsakoff et al., 2003).

Construction of variables

Problem formulation performance (ProblemFormulationPerf) is measured in terms of a team’s consistent ability to identify valuable projects to work on. Baer et al.’s (2013) critical performance dimensions – relevance and comprehensiveness – refer to orienting search in the right directions, and the thoroughness of search processes. We measure problem formulation performance by assessments of how well a firm or team identifies relevant problems on which to work (see Appendix); the ability to identify relevant problems during stages that precede the search for solutions signifies successful problem formulation.

Bias awareness (BiasAwareness) represents the degree to which people within a firm are aware of the existence of biases in individuals, the team, and the firm. As a novel construct, with a Cronbach’s Alpha of 0.633, we elected to use the measure but note that a more refined measure would be useful in future studies.

Bias intensity constructs measure degree of respondents’ perceived bias. Dominance (DominanceBias) arises when a team member exerts a high degree of influence over project choices. A dominant team member may or may not possess formal authority over other team members. Solution jumping bias (SolutnJumpBias) exists when search processes are limited or truncated. Information distortion (InfoDistortionBias) measures the degree of inaccuracy of information available to a project team and missing information. Perceptual bias (PerceptualBias) measures the extent to which teams and their members are subject to process-distorting issues related to individual professional or personal experiences.

We also included several control variables in our analyses. Most studies of problem formulation acknowledge the importance of the attributes of the problem-finding environment (e.g., Baer et al., 2013; Leiblein & Macher, 2009; J. A. Nickerson & Zenger, 2004). Factors related to intellectual property (IP) may play a role in problem formulation (see Felin & Zenger, 2014; Hurmelinna-Laukkanen & Heiman, 2012). In the US/China dataset, FormalIP covers reliance on legal tools, such as patents, copyrights, trademarks, employees’ loyalty obligations, employment contracts, and/or nondisclosure/confidentiality agreements (Hurmelinna-Laukkanen & Olander, 2014). Poor use of formal IP protection may lead to the setting of arbitrary limits on sharing the sources of ideas (Jaruzelski & Holman, 2011) owing to fear of legal issues related to rights infringement. Conversely, managers may limit problem formulation to those problems with solutions protectable via formal IP mechanisms, following established IP practices in their own field (Yang & Hurmelinna-Laukkanen, 2022). InformalIP, a construct based on seven survey items from the US/China data, comprises access restrictions, limited document distribution, training on IP issues, and calculated trust (Williamson, 1996). For example, if no training on secrecy practices is offered, those involved in problem search activities may choose to treat everything as secret to avoid accusations of confidentiality breaches, limiting knowledge sharing (Olander & Hurmelinna-Laukkanen, 2010). In the Finnish survey instrument, the HybridIP construct is comprised of three survey items combining both formal and informal dimensions.

Attributes of knowledge, including knowledge tacitness in the local search environment, may affect problem formulation performance (Leiblein & Macher, 2009). KnowledgeTacitness, a construct comprising five survey items, measures the extent to which knowledge is undocumented, irretrievable, or whose transfer imposes excessive costs (Zander & Kogut, 1995).

We further expect that search breadth (SearchBreadth) – the scope of search employed in problem finding processes – affects bias intensity and problem formulation performance. Many organisations’ cultures dictate that search activities occur in accord with historical precedents (Ritala et al., 2016). Both narrow and broad searches may yield positive or negative results (Ahuja et al., 2013); hence, we do not predict a sign for this factor’s coefficient.

We additionally added some demographics elements as controls. Position (US/China data only) specifies the job type and level of seniority of each respondent. FirmAge indicates the age of the firm in years. A minimum firm age of three years was required to avoid newness-related measurement issues (Peña, 2002). NumberOfEmployees measures the total number of employees and serves as a control for firm size. To control for country-specific effects in the combined (US/China) dataset, a dummy variable, CountryChinaUS, was included. CompetitiveEnvironment describes respondents’ assessments of how many other firms offer similar or identical products or services (Finnish data only). Table 1 presents a list of industry dummy measures used for the US/China and Finland data samples.

To evaluate data-construct reliability, we present Cronbach’s Alpha, and factor loadings from exploratory factor analysis in the Appendix. Following Cohen (1988) and Sakakibara et al. (1993) for original constructs, we elected to accept a construct as useable when that construct had an alpha of 0.6 or greater. For established constructs, we required alpha values of 0.8 or greater. With one exception, all measures conformed to our standards for inclusion as sufficiently reliable constructs; the original measure PerceptualBias has an alpha of just below 0.6 (0.593). Insights and feedback from the pilot study and a panel of experts, as well as existing theory led us to keep the chosen indicators. For novel constructs, the results from the exploratory factor analysis largely echo the Cronbach’s Alpha findings, with some factors having lower than desirable loadings of their components (see Appendix). For more conventional constructs, the loadings were better. We deem our constructs to be sufficiently reliable for an exploratory study, but we recognise the need for future research to further refine constructs.

Descriptive analysis

Table 2 shows summary statistics for the dependent, independent, and control variables. The values of the summary statistics suggest no apparent issues with the sample datasets, suggesting their suitability for inferential analyses. From the mean values of most of our main measures, bias intensity appears lower on average in Finland than in the US or China. Some of this may be owing to cultural differences. For example, power distance is known to be lower in Finland than in the US and China (Hofstede, 1991), which may partly explain differences in the dominance measure. Also, based on Table 2’s descriptive analysis, we note that the breakdown of industry types in Finland differs from that in the US and China, for example, with a higher comparative presence of manufacturing firms in Finland. Firms in the samples are larger in the US and China than they are in Finland. Overall, our two datasets reflect substantial diversity of firms.

Correlation matrices for both datasets (US/China and Finland) are presented in the Appendix. The highest absolute magnitude of Pearson’s r is .646, less than the convention of 0.8 for identifying a multicollinearity issue. This suggests that further inferential analysis using ordinary least-squares regression is appropriate.

We ran the Shapiro–Wilk test on all non-binary variables for normality. Three measures, IPbias, CompetitiveEnvironment, and InfoDistortion required closer scrutiny, but a visual inspection of these measures’ histograms suggests that they sufficiently closely resemble a normal distribution. All the other constructs were normally distributed. We also performed Breusch–Pagan tests for heteroskedasticity of data, and there was no indication that any data were heteroskedastic (Chi-square statistics were all >0.05).

Regression analysis

For the statistical models, we focused on evaluation of perceptions, and our findings should be conservatively interpreted as indicating associations rather than causality. To gain initial insight into an under-examined issue, we used two distinct regression analyses to examine the relationships between (a) awareness of bias and bias intensity, and (b) bias intensity and problem formulation performance. Since we did not combine the two datasets, this was also done to avoid communication diseconomies incurred by using more complex models (e.g., path analysis, or structured linear systems). Figures 1 and 2 below summarise the findings.

Figure 1. Empirical findings – Bias-awareness and bias intensity.

Figure 2. Empirical findings – Bias intensity and problem formulation performance.

Stage 1 – bias awareness and bias intensity

The statistical results of our Stage 1 examination are presented in Table 3. Model 1 features analyses of both the US/China and Finland datasets, testing whether BiasAwareness is related to DominanceBias. Both F statistics are significant (p < .005 and p < .05, respectively). In Model 1, DominanceBias intensity is not significantly predicted by Awareness in the US/China data. For Finland, Model 1 shows that DominanceBias magnitude varies negatively with BiasAwareness (p < .05). Model 2 shows that for solution jumping, for both the US/China and Finland data, BiasAwareness is negatively associated with SolutnJumpBias. Model 3 demonstrates a significant association between InfoDistortionBias intensity and BiasAwareness for both datasets. Model 4 (Finland only) shows strng support for the effect of BiasAwareness on the magnitude of PerceptualBias (p < .005).

Table 3. Stage 1 regression: bias awareness and bias intensity (continues onto next page).

	Model 1		Model 2		Model 3		Model 4
Stage 1	DominanceBias		SolutnJumpBias		InfoDistortionBias		PerceptualBias
Dependent Variable	US/China	Finland	US/China	Finland	US/China	Finland	Finland
n	243	164	246	165	241	171	174
F (dfModel. dfResid.)	3.40 (14.23)***	1.77 (13.150)*	7.37 (14.23)***	3.87 (13.151)***	2.8 (14.22)***	7.23 (13.15)***	4.59 (13.160)***
R-squared	0.18	0.1329	0.26	0.25	0.15	0.37	0.27
Adj. R-squared	0.13	0.0577	0.22	0.19	0.10	0.32	0.21
Awareness	−0.073 (0.058)	−0.233 (0.107)*	−0.180 (0.064)***	−0.460 (0.109)***	−0.204 (0.066)***	−0.796 (0.093)***	−0.547 (0.082)***
SearchBreadth	−0.246 (0.047)		−0.041 (0.052)		−0.192 (0.054)***
Postion	−0.011 (0.042)		−0.057 (0.047)		−0.035 (0.048)
FirmAge	−0.002 (0.002)	0.001 (0.002)	0.000 (0.002)	0.002 (0.002)	0.000 (0.002)	−0.001 (0.002)	0.000 (0.002)
NumberOfEmployees	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)
CountryChinaUS	0.310 (0.117)***		0.831 (0.129)***		−0.273 (0.132)*
Competitive Env		0.116 (0.071)*		−0.082 (0.072)		−0.002 (0.060)	0.032 (0.053)
Manufacturing	0.252 (0.178)	−0.239 (0.311)	−0.125 (0.196)	−0.333 (0.315)	−0.142 (0.200)	0.197 (0.277)	0.138 (0.246)
Service	0.053 (0.148)	−0.654 (0.374)*	0.024 (0.163)	−0.178 (0.378)	−0.089 (0.168)	0.047 (0.327)	0.007 (0.290)
InfoTech	0.121 (0.180)	−0.290 (0.375)	0.099 (0.198)	−0.869 (0.379)*	−0.208 (0.209)	−0.001 (0.333)	0.043 (0.296)
Education	0.25 (0.258)6	−0.487 (0.385)	−0.015 (0.284)	−0.693 (0.389)*	−0.029 (0.289)	−0.066 (0.342)	0.016 (0.304)
Electronics	0.011 (0.250)		0.205 (0.285)		−0.395 (0.280)
Medical	−0.064 (0.190)		0.179 (0.209)		−0.372 (0.214)*
RealEstate	−0.227 (0.397)	−0.139 (0.389)	0.378 (0.438)	−0.019 (0.393)	−0.364 (0.443)	0.217 (0.338)	0.037 (0.300)
Trade	0.220 (0.228)	−0.443 (0.341)	0.094 (0.251)	−0.658 (0.343)*	0.163 (0.256)	0.053 (0.300)	0.001 (0.265)
ConsumerGoodsMfg		−0.523 (0.319)*		−0.641 (0.322)		0.110 (0.283)	−0.111 (0.252)
OtherMfgRepairInstall		−0.067 (0.439)		0.021 (0.445)		0.377 (0.391)	0.329 (0.348)
TransportStorage		0.382 (0.403)		−0.437 (0.408)		0.326 (0.351)	0.380 (0.312)
Constant	2.156 (0.223)***	0.281 (0.300)	−2.230 (0.246)***	0.436 (0.303)	2.463 (0.252)***	−0.073 (0.267)	−0.049 (0.237)

coefficient and/or standard errors shown as 0 have near-zero values on the order of 1E–06 or less. *p ≤ .05; **p ≤ .01; ***p ≤ .005

Overall, for Stage 1, of seven regressions exploring four distinct bias intensities, six models support our assertion that bias awareness is negatively related to bias intensity. The results are particularly compelling for five of the models in which these models’ coefficients are significant at p < .005. In Models 2 (SolutnJumpBias) and 3 (InfoDistortionBias) both the US/China and Finland models show significant coefficients. Of the control variables, search breadth is significant in one model (3) for the US/China data. Being a US respondent predicts increased dominance and solution jumping biases (p < .005), while being a Chinese respondent predicts that teams experience increased information distortion bias (p < .05), likely reflecting differences in work cultures.

Stage 2 – bias and problem formulation performance

Stage 2 results are presented in Table 4 (Model 5). Both the US/China and Finland regressions in Model 5 have significant F statistics (p < .005). For both DominanceBias and SolutnJumpBias, coefficients are significant, reflecting negative relationships between these bias types and problem formulation performance. On the other hand, InfoDistortionBias and PerceptualBias are not associated with performance. We discuss these findings in more detail below.

Table 4. Stage 2 Regression analysis: Bias intensity and problem formulation performance (continues onto next page).

	Model 5
Stage 2	ProblemFormulationPerf
Dependent variable	US/China	Finland
n	191	138
F (dfModel. dfResid.)	6.24 (19. 171)***	4.24(16.121)***
R-squared	0.40	0.36
Adj. R-squared	0.34	0.27
DominanceBias	−0.467 (0.081)***	−0.1976 (0.107)*
SolutnJumpBias	−0.237 (0.070)***	−0.4595 (0.106)***
InfoDistortionBias	0.072 (0.068)	−0.0155 (0.101)
PerceptualBias		−0.0606 (0.138)
InformalIP	−0.184 (0.131)
HybridIP		−0.0710 (0.075)
FormalIP	−0.138 (0.102)
KnowledgeTacitness	−0.036 (0.095)
SearchBreadth	−0.061 (0.056)
Position	−0.015 (0.049)
FirmAge	−0.002 (0.002)	−0.0037 (0.003)
NumberOfEmployees	0.000 (0.000)	−0.0001 (0.000)*
CountryChinaUS	0.051 (0.156)
Manufacturing	−0.584 (0.187)***	−0.0179 (0.385)
Service	−0.040 (0.166)	−0.1795 (0.441)
InfoTech	−0.215 (0.206)	−0.0075 (0.456)
Education	−0.374 (0.285)	−0.5513 (0.463)
Electronics	−0.246 (0.255)
Medical	−0.019 (0.199)
RealEstate	−0.057 (0.765)	−0.4279 (0.447)
Trade	−0.074 (0.268)	0.1087 (0.426)
ConsumerGoodsMfg		−0.3020 (0.396)
OtherMfgRepairInstall		−0.3212 (0.517)
TransportStorage		0.1503 (0.508)
Constant	−1.432 (0.402)***	0.3539 (0.379)

*p ≤ .05; **p ≤ .01; ***p ≤ .005.

Discussion

Existing literature recognises that appropriate problem formulation activities are an important antecedent to problem solving, and a significant determinant of innovation trajectories in firms (Baer et al., 2013; Thrane et al., 2010). Biases that disrupt problem formulation have potentially serious negative downstream effects on solving chosen problems, and/or firm performance (Deichmann et al., 2021; J. A. Nickerson & Zenger, 2004). Using the lens of the strategic problem formulation (Baer et al., 2013), our work refines and extends previous studies in the fields of the behavioural theory of the firm (e.g., Cyert & March, 1992), strategic management (e.g., M. A. Lyles & Mitroff, 1980), and innovation management (e.g., Terwiesch, 2008). We provide initial insight into organising for strategic problem formulation by focusing on antecedents to problem solving (Baer et al., 2013; Mumford et al., 2006; J. A. Nickerson et al., 2017).

Our provisional findings suggest broadly that biases play a role in problem formulation performance. They also suggest that roles of different biases are not uniform. In line with previous studies suggesting that biases may distort problem formulation (Baer et al., 2013, Barnes, 1984; Brusoni, 2005; J. A. Nickerson & Zenger, 2004), we find that bias intensity is negatively related to problem formulation performance especially for solution jumping and dominance. Conversely, information distortion bias and perceptual bias seem comparatively less important. This points to differences between bias types (motivational, informational, cognitive) and individual biases within these categories (e.g., solution jumping and perceptual bias as cognitive biases).

Our exploration of bias awareness offers insight into these issues. Existing literature acknowledges that good problems are difficult to formulate because managers are frequently ignorant of the importance of problem formulation activities in organisations (Amit & Zott, 2001). While limited managerial attention (low bias awareness) is a potential source of biases in many studies (e.g., Barnes, 1984; Van Knippenberg et al., 2015), we extend previous conversations by suggesting that managerial attention is relevant for remediating biases. Specifically, our preliminary findings indicate that apart from dominance, most biases are quite responsive to increased awareness – consistent with studies suggesting that awareness generally facilitates appropriate managerial action (Banerjee et al., 2019; Hurmelinna-Laukkanen & Heiman, 2012; Mohaghegh & Größler, 2020). Solution jumping, information distortion, and perceptual bias seem particularly sensitive to bias awareness. Understanding that shortcuts have challenges or being alert to possibly wrong information allows the mitigation of negative effects by, for example, putting extra effort into a search (Anderson et al., 2001; Bhardwaj et al., 2018; Tabesh, 2021). Dominance, on the other hand, may impact problem formulation despite awareness of its associated potential problems (Clegg et al., 2006). Motivational issues (related to power differences) influencing behaviours and decision-making (Borchardt et al., 2022; Vidal & Möller, 2007) are difficult to address (Todorova & Durisin, 2007).

Acknowledging the differences described above is essential for practitioners. Our findings suggest that managers should seek to understand the front end of value creation activities. Specifically, as increased awareness generally mitigates bias intensity, training and educating managers to detect bias may yield beneficial results. Especially for more persistent biases, an active search for appropriate remedies becomes a key task for managers. Sometimes this may mean stepping down and letting others take over, or removing members from a team. Understanding the microfoundations of problem formulation in innovation activities increases the potential for achieving high-performance problem formulation.

Conclusions

Strategic problem formulation as a framework for understanding innovation provides useful tools to address scholarly and practical challenges in innovation management. The existing discussion revolves around evaluating pre-selected or pre-existing problems and their attributes rather than formulating problems and remediating the effects of biases in problem formulation processes. This study turns attention to these issues.

We emphasise our work is exploratory in nature and should not be considered definitive, particularly considering some issues around the reliability of our constructs. We nonetheless add to existing understanding of how bias awareness mitigates bias intensity and which biases are prone to challenge problem formulation performance. Our exploratory study indicates that bias awareness can be instrumental in reducing cognitive and informational biases and that motivational biases (dominance, in particular) and behaviours driven by them may be more challenging to address.

The limitations of our study reveal and ground relevant research opportunities. Regarding empirical evidence, larger datasets and more sophisticated constructs will advance research on problem formulation. Although our findings are drawn from original survey datasets from three countries across three continents, further development of concurrent data collection across multiple locations is encouraged. Moreover, we extracted several useful insights from this study, but more reliable constructs than those presented herein would comprise a significant step forward. We constructed all measures with attention to theoretical justifications, the differences in our data sets, the nature of the constructs, and overall model fit. Based on these considerations, we chose to keep some indicators despite limitations. In exploratory research that focuses on generating new, preliminary insights and understanding of relationships among variables, some lower factor loadings or alphas may be expected and acceptable. Nevertheless, we encourage scholars to pursue the development of more refined constructs. For example, dominance-bias might be constructively decomposed to capture different types of dominance (e.g., reflecting distinct effects of opportunistic/altruistic approaches). Also, problem formulation performance might be examined in a fashion that reflects its constituent components.

More sophisticated analytical techniques may also shed further light on the questions we have raised. For example, while we do not expect bias awareness to have a direct connection to problem formulation performance and, hence, we did not pursue examining moderation or mediation effects, later studies are encouraged to explore these potentially more complex relationships. More assumption-free non-parametric analytic techniques might be appropriate as research advances.

Theoretically, more biases (e.g., behavioural ones) remain to be identified and tested empirically, especially under different problem formulation contexts. For example, one interesting contemporary phenomenon, artificial intelligence, will quite likely introduce new types of biases, or remedy some existing ones. Future work should also explore what factors enhance bias awareness, thereby potentially decreasing the impact of bias intensity. There may exist multiple awareness-enhancing practices appropriate for teams and managers, such as training, simulations, and the use of outside contractors. Remedial action may have important implications for decision-making in organisations (Borchardt et al., 2022), improving the quality and outcomes of decisions. The task is certainly challenging, as awareness of some biases, for example dominance, is often suppressed to avoid conflict. Nonetheless, though exploratory, our work supports the idea that a better understanding of the mechanisms of biases may allow managers to rise to the occasion and enhance problem formulation performance.

Acknowledgement

We wish to thank Xuejiao Dutton and Stephen Thompson for their assistance during the data collection. We also would like to express our thanks to professors Aino Kianto and Jorge F.S. Gomes for their help in drafting the questionnaire and Jackson Nickerson for his insightful comments on an early draft of this work.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Appendix

Survey items (US/China).

Table

Constructs	Survey Items; Factor Loadings (where applicable)	Cronbach’s Alpha
Dependent Variables
Problem FormulationPerf	Our firm consistently identifies relevant projects to work on. 0.589	0.653
	Our firm is very good identifying appropriate projects to work on. 0.616
	Our firm is unable to focus on the most important projects. (r) 0.434
	Our firm does not do a good job identifying potentially profitable projects. 0.402
Independent Variables
BiasAwareness (DV at stage 1)	We are well aware in our firm that personal biases negatively influence project selection. 0.725	0.633
	Our team always recognises that information used in a project selection process may be biased. 0.790
	We are well aware in our firm that organisational issues exist that negatively influence project selection (e.g., established routines, ways of organising things). 0.83
DominanceBias (DV at stage 1)	We often end up working on the ‘wrong’ project because one or more members of the team pushes hard in this direction. 0.545	0.654
	The senior manager in our team exerts a high degree of influence over project choice.0.45
	Junior members of our team are hesitant to express opinions on important topics. 0.934
SolutnJumpBias (DV at stage 1)	When defining a project, our teams always search until the most appropriate project is identified -	0.706^†
	When choosing a project, our team always researches the topic extensively. -
InfoDistortion (DV at stage 1)	We have on some occasions relied on information that has later proven to be incorrect. (r) -	-
Control Variables*
SearchBreadth	We mostly work on projects that require having a broad understanding of issues, rather than a deep understanding. -	-
Knowledge Tacitness	It is very hard to teach knowledge related to our product/service/processes. 0.589	0.788
	It is hard to understand features of key product/service/processes by observing/examining them. 0.655
	It is not possible to document knowledge related to the product/service/processes in writing. 0.73
	Core knowledge related to the product/service/process is embedded in our internal routines. 0.727
	Core knowledge related to the product/service/process is embedded in key individuals’ heads. 0.684
FormalIP	Effectiveness of patents for protecting intellectual property 0.734	0.916
	Effectiveness of copyrights for protecting intellectual property 0.807
	Effectiveness of trademarks for protecting intellectual property 0.741
	Effectiveness of long-term collaboration contracts for protecting intellectual property 0.831
	Effectiveness of non-disclosure/confidentiality agreements for protecting intellectual property 0.853
	Effectiveness of inter-firm agreements on not recruiting personnel from alliance partners for protecting intellectual property 0.823
	Effectiveness of employee non-competition agreements for protecting intellectual property 0.779
	Effectiveness of the legal loyalty obligation of employees for protecting intellectual property 0.778
	Effectiveness of legal right of the employer to assign tasks for protecting intellectual property 0.672
InformalIP	Educating personnel on IP rights and secrecy issues. 0.717	0.842
	Motivating personnel to be increasingly committed to the firm. 0.783
	Small amount of personnel turnover. 0.777
	Sharing information with only a few people. 0.759
	Using password(s). 0.694
	Restricting access to meetings. 0.727
	Restricting access to firms’ premises. 0.691

^†We used Pearson’s r to judge construct reliability if the construct is composed of only two items.

*Other controls that employ one measure in both datasets are listed in the Finnish Cronbach’s Alpha table, below.

(r) = reversed wording

The respondents were given the following explanation above the questions: ‘In this questionnaire, a “project” is understood widely, including various stages. In the following questions, we inquire about 1) problem formulation, 2) problem selection, 3) problem solving, and 4) outcomes’.

Survey items, continued (Finland).

Table

Constructs	Survey Items; Factor Loadings	Cronbach’s Alpha
Dependent Variables
Problem	Our firm routinely tries to implement too many distinct major projects 0.704	0.783
Formulation	Our firm is unable to focus on the most important projects. 0.972
Perf	Our firm does not do a good job identifying potentially profitable projects. 0.565
Independent Variables
BiasAwareness (DV at stage 1)	We are well aware in our firm that personal preferences negatively influence project selection. 0.813	0.634
	We are well aware in our firm that organisational issues (e.g., established routines, ways of organising things) influence project selection. 0.84
	Our company employs many programs and projects to increase the employees’ knowledge and skills. 0.61
	There are great shortcomings in the expertise of our employees. (r) −0.86
	Our company knows its strengths and where it is aiming at. 0.28
	In our company problems are not hidden but handled openly. 0.41
	People in our company have a lot of knowledge of the work they do. 0.254
DominanceBias (DV at stage 1)	We often end up working on the ‘wrong’ project because one or more members of the team push hard in this direction. 0.665	0.720
	It is more important who presents a new idea for a project than what the content of the idea is. 0.882
	The senior manager in our team exerts a high degree of influence over project choice. 0.512
	All opinions are of equal value regardless of the person’s status in the organisation. −0.88
SolutnJumpBias (DV at stage 1)	When defining a project, our teams always search until the most appropriate project is identified. 0.593	0.676
	In our firm we do not make extensive enough examinations on alternative projects.; −0.83
	When choosing a project our team always researches the topic exhaustively. −0.499
PerceptualBias (DV at stage 1)	Everyone in our project team understands and uses the same technical language or jargon. (r) 0.608	0.593
	The differences in people’s backgrounds and training create difficulties for identifying the appropriate projects. −0.359
	The different units & functions (e.g., R&D, marketing and production) understand each other well. (r) 0.486
	When planning and making decisions, we look at things from many perspectives. (r) 0.448
	The company management is good at seeing the weak signals in our business. (r) 0.529
InfoDistortionBias (DV at stage 1)	We have a clear understanding of what information and knowledge is the most relevant for achieving the company’s goals. (r) 0.646	0.674
	Our company employs effective and functional information systems. (r) 0.639
	Our company has a lot of useful information in documents and databases. (r) 0.627
Control variables
HybridIP	Barriers exist to choosing some projects because of lack of access to protected intellectual property. 0.608	0.726
	It is only appropriate to choose projects that are likely to result in clear intellectual property rights for the firm. 0.902
	It is easy for us to get access to and use intellectual property belonging to others in our projects. 0.551
Competitive Environment	Our sector generally includes many companies offering similar products/services. 0.777	0.847
	We have many competitors in our home market. 0.896
	We do not have many competitors. (r) −0.753
Other controls – Single-item measures in both datasets
Firmage	2009 - Year of establishment (‘firm age’)
NumberEmpl2007	Number of employees 2007 (‘firm size’)
Industry dummies	Industrial supplies manufacturing Consumer goods manufacturing Other manufacturing & repair and installation Wholesale & retail Transportation & storage Information and communication Construction & real estate Professional, scientific and technical activities Administrative support services
	(Base condition: other industries)

(r) = reversed wording

The respondents were given the following explanation above the questions: ‘In this questionnaire, a “project” is understood widely, including various stages. In the following questions, we inquire about 1) problem formulation, 2) problem selection, 3) problem solving, and 4) outcomes’.

Correlation matrices (US/China Dataset; *p ≤ 0.05)

Table

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
Awareness(1)
DominanceBias(2)	−0.1532*
SolutnJumpBias(3)	−0.2298*	0.3995*
InfoDistortionBias(4)	−0.3097*	−0.0165	0.0389
InformalIP(5)	0.2831*	−0.1042	−0.1616*	−0.3820*
KnowledgeTacitness(6)	0.2537*	−0.2690*	−0.2299*	−0.2563*	0.5630*
FormalIP(7)	0.2240*	−0.0765	−0.1362*	−0.1791*	0.6464*	0.3998*
CountryChinaUS(8)	−0.1320*	0.1648*	−0.0323	0.3352*	−0.0953	−0.1865*	0.1754*
FirmAge(9)	−0.0575	0.0115	−0.0556	0.1396*	−0.0618	−0.1885*	0.0961	0.4014*
Manufacturing(10)	−0.1005	0.0471	0.0111	−0.0689	0.0502	0.0721	0.0359	−0.0991	−0.0691
Service(11)	0.0702	−0.0070	0.0117	0.0240	0.0256	−0.0987	0.0149	0.1276*	0.0352	−0.2822*
InfoTech(12)	0.1271*	−0.0304	−0.0889	−0.0693	0.0468	0.0343	0.0503	−0.0430	−0.0986	−0.1479*	−0.2588*
Education(13)	−0.0954	0.0558	0.0998	0.1018	−0.2101*	−0.1242*	−0.1796*	0.0678	0.1386*	−0.0921	−0.1612*	−0.0845
Electronics(14)	0.0535	−0.0051	−0.0533	0.0058	−0.0300	0.1192	−0.0777	−0.1268*	−0.0075	−0.0837	−0.1465*	−0.0768	−0.0478
Medical(15)	−0.0516	−0.0082	−0.0731	0.0726	−0.0425	−0.0171	0.0505	0.1564*	0.1353*	−0.1339*	−0.2342*	−0.1228*	−0.0765
RealEstate(16)	−0.0487	−0.0180	0.0341	0.0444	−0.0242	−0.0768	0.0033	−0.1052*	−0.0751	−0.0566	−0.0990	−0.0519	−0.0323
Trade(17)	0.0097	0.0610	0.0700	−0.0107	−0.0961	−0.0407	−0.0042	0.0117	0.0185	−0.1000	−0.1749*	−0.0917	−0.0571
NumberOfEmployees(18)	0.0674	0.0054	−0.1156	0.0328	0.1431*	−0.0829	0.1343*	0.2599*	0.4524*	−0.0506	0.0772	0.1004	−0.0439
ProblemFormPerf(19)	0.0000	0.4367*	0.1708*	−0.2702*	0.2087*	−0.0073	0.2047*	0.0361	0.0236	−0.1181*	0.0162	0.0203	−0.0399
JobPosition(20)	0.1979*	−0.1030	−0.1247*	−0.1352*	0.0331	0.0621	−0.1227*	−0.1198*	−0.0596	−0.0866	−0.0230	0.0757	−0.0461
SearchBreadth(21)	0.1663*	−0.3643*	−0.3019*	−0.1417*	0.2474*	0.2218*	0.1230*	−0.1008	−0.0101	0.0517	−0.0317	0.0503	−0.1070
	(14)	(15)	(16)	(17)	(18)	(19)	(20)
Electronics(14)
Medical(15)	−0.0695
RealEstate(16)	−0.0294	−0.0469
Trade(17)	−0.0519	−0.0830	−0.0350
NumberOfEmployees(18)	−0.0559	−0.0446	−0.0386	−0.0520
ProblemFormPerf(19)	−0.0070	−0.0107	−0.0102	0.0244	0.0763
JobPosition(20)	0.0416	−0.0069	0.0237	0.0619	−0.0929	0.0086
SearchBreadth(21)	−0.0213	−0.0375	−0.0044	−0.0339	0.0516	−0.1293	0.0832

Correlation matrices (Finland; *p ≤ 0.05)

Table

	(1)	(2)	(3)	(4)	(5)	(6)	(11)	(12)	(13)	(14)	(15)	(16)	(17)
HybridIP(1)
SolutnJumpBias (2)	−0.004
DominanceBias (3)	0.023	0.3512*
PerceptualBias (4)	−0.124	0.5343*	0.483*
InfoDistortionBias (5)	−0.080	0.3335*	0.209*	0.481*
ProblemFormPerf (6)	−0.039	−0.421	−0.3805*	−0.371	−0.128
Awareness (11)	0.015	−0.405*	−0.2077*	−0.509*	−0.585*	0.114
FirmAge (12)	−0.134	−0.083	0.017	−0.072	−0.079	−0.023	0.084
NumberOfEmployees (13)	0.025	−0.150*	0.026	0.024	−0.062	−0.005	−0.036	0.001
CompetitiveEnvironment (14)	−0.149	−0.166*	0.020	−0.082	−0.093	−0.024	0.1784*	0.119*	0.012
Manufacturing (15)	0.072	0.102	0.064	0.084	0.039	−0.032	0.002	−0.134*	0.049	−0.213*
ConsumerGoodsMfg (16)	0.011	−0.058	−0.084	−0.058	0.056	0.038	−0.079	0.023	0.048	0.058	−0.3542*
OtherMfgRepairInstall (17)	−0.043	0.127	0.000	0.076	0.000	−0.043	0.045	0.069	−0.034	0.028	−0.129*	−0.102
Trade (18)	−0.131	−0.160*	0.001	−0.075	−0.094	0.093	0.091	0.2474*	−0.041	0.192*	−0.251*	−0.197*	−0.072
TransportStorage (19)	−0.068	0.038	0.157*	0.148*	0.113	−0.004	−0.153*	0.036	−0.003	−0.081	−0.140*	−0.110*	−0.040
InfoTech (20	0.008	−0.155*	−0.040	−0.064	−0.090	0.072	0.089	−0.026	−0.031	−0.015	−0.203*	−0.159*	−0.058
RealEstate (21)	0.070	0.108	0.059	0.027	0.060	−0.119	−0.061	−0.051	−0.020	0.085	−0.140*	−0.110*	−0.040
Service (22)	0.003	0.089	−0.124	−0.030	−0.025	0.012	−0.015	−0.027	−0.023	−0.057	−0.169*	−0.133*	−0.049
Education (23)	0.062	−0.119	−0.082	−0.087	−0.125	−0.011	0.161	−0.053	0.003	0.103	−0.150	−0.118	−0.043
	(18)	(19)	(20)	(21)	(22)
TransportStorage (19)	−0.078
InfoTech (20)	−0.113*	−0.063
RealEstate (21)	−0.078	−0.044	−0.063
Service (22)	−0.094	−0.053	−0.076	−0.053
Education (23)	−0.084	−0.047	−0.068	−0.047	−0.056