You are on page 1of 22

The current issue and full text archive of this journal is available on Emerald Insight at:

www.emeraldinsight.com/1355-2554.htm

IJEBR
23,1 A measure of entrepreneurial
bricolage behavior
Per Davidsson
Department of Management,
114 Queensland University of Technology, Brisbane, Australia
Received 30 November 2015 Ted Baker
Revised 8 March 2016 Department of Management and Global Business,
Accepted 18 March 2016
Rutgers University, Newark, New Jersey, USA, and
Julienne Marie Senyard
Department of Marketing, Griffith University, Brisbane, Australia

Abstract
Purpose – The majority of emerging and young firms work under resource constraints. This has made
researchers highlight the importance of resourcefulness. Perhaps the most important theoretical development
in this context is the emerging, behavioral theory of entrepreneurial bricolage. However, although academic
interest is increasing, research on entrepreneurial bricolage has been hampered by the lack of robust
instruments that allow large-scale theory testing. The purpose is to help fill this void. The purpose of this
paper is to describe the development and contents of a novel measure of entrepreneurial bricolage behavior
and assesses its validity. The measure is intended to be applicable in broadly based, quantitative studies.
Design/methodology/approach – The instrument was developed as a unidimensional, reflective measure.
Standard protocols for scale development were followed. The validation uses primary, longitudinal data from
four samples of nascent and young firms as well as published, cross-sectional evidence from another four
samples representing different contexts and variations to the data collection procedure.
Findings – Promising results are reported concerning the reliability as well as the discriminant and
nomological validity of the measure. Based on the pre-testing and validation experiences guidelines are also
provided for attempts at further improvements.
Originality/value – This paper presents a novel measure developed by the authors, which holds promise for
being a useful tool for future research on the prevalence, antecedents, and consequences of entrepreneurial
bricolage. Previously, no established measure of entrepreneurial bricolage behavior existed, and the few partial
measures appearing in the literature have not been comprehensively evaluated. Thus, we offer a comprehensive
and elaborate presentation of a measure only briefly introduced in Davidsson (2016) and Senyard et al., (2014).
Keywords Entrepreneurs, Start-ups, Research methods, Bricolage
Paper type Research paper

1. Introduction
Business founders face resource constraints for many reasons. For example, they may operate
in resource-poor environments, lack personal or family wealth, be unwilling to borrow, or be
personally attracted to opportunities that are not attractive to lenders or investors. In fact, it
has been suggested that “Resource-constrained founder-run firms are the most common form
of business globally” (Powell and Baker, 2014a, p. 1406). An important theme in research
attending to this is entrepreneurial resourcefulness, broadly constituting ways that
entrepreneurs attempt “to deal with problems or opportunities despite ostensibly inadequate
resources” (Powell and Baker, 2014b). Perhaps the most important current theme in this
literature is the emerging theory of “entrepreneurial bricolage.”
Although the number of studies employing the “entrepreneurial bricolage” construct is
International Journal of growing, progress has been held back by the absence of validated operationalizations that
Entrepreneurial Behavior &
Research can be applied in quantitative studies. One such measure – the Baker-Davidsson bricolage
Vol. 23 No. 1, 2017
pp. 114-135
scale – has recently been used in multiple studies (including our own, see Senyard et al.,
© Emerald Publishing Limited
1355-2554
2014). However, the previously published studies focus on substantive relationships and
DOI 10.1108/IJEBR-11-2015-0256 provide only rudimentary description and evaluation of the scale. Therefore, in this paper
we provide a thorough description of the development and contents of the Baker-Davidsson Entrepreneurial
bricolage scale, as well as a comprehensive evaluation of its validity based on data from up bricolage
to eight separate samples. By so doing, we offer the following contributions. First, we behavior
provide an instrument which holds promise of being a useful and broadly applicable tool for
future research on the prevalence, antecedents, and consequences of entrepreneurial
bricolage. Second, our provision of a theory driven, validated measure may help improve
theoretical precision, thereby counteracting a tendency of the entrepreneurship literature to 115
devolve back to the vague and abstract/metaphorical use of the bricolage concept that
appears in many humanities and social science literatures. Third, our description of the
development process as well as our evaluation of the measure provide guidance for possible
further improvements as well as attempts at developing more precise, domain-specific
measures of entrepreneurial bricolage. Fourth, our thorough illustration of the scale
development process may be of help for researchers who set out to operationalize other
theoretical constructs.
In the next section, we briefly describe the emerging theory of entrepreneurial bricolage
and the state of associated research, arguing the need for a broadly applicable and an
easy-to-use quantitative measure. We then describe the process of developing the measure.
This is followed by a description of the samples used to evaluate the measure and results
pertaining to content validity; reliability/internal consistency; discriminant validity, and
nomological validity, respectively. We conclude the paper with a discussion of our results as
well as limitations and possible future improvements.

2. The emerging theory of entrepreneurial bricolage


As a topic of research, bricolage is usually traced to Structural Anthropologist Claude
Levi-Strauss (1966), who introduced the concept primarily as part of his attempt to explain
how a society borrows structural elements of other cultures – especially elements of myths –
and recombines them to suit its own purposes. Such borrowing often ends up putting the
fragments and elements to different uses than they had in the society from which they were
borrowed. Complementing this “ideational” form of bricolage, Levi-Strauss also described a
more material form of bricolage, portraying the bricoleur as a jack-of-all-trades making do
with the tools, skills, and materials at hand. The bricoleur is contrasted with the idealized
notion of the “engineer,” who obtains and deploys exactly the right inputs for every task
(Stinchfield et al., 2013).
The concept of bricolage has been adopted across a variety of fields, mainly in the
humanities and social sciences, but also on occasion in the life sciences (see Baker and Nelson,
2005 for a brief review). In many cases the concept is used as a rather abstract metaphor, such
as when biologists refer to nature as a bricoleur, using parts of an existing organism as the
basis for the evolutionary development of new physical structures and functions, such as a leg
being developed into a wing ( Jacob, 1977). Early application of bricolage to entrepreneurship
was similarly metaphorical and drew on broad notions of bricolage as “making do with what’s
at hand.” For example Baker et al. (2003, p. 269) defined what they labeled “network bricolage”
as “dependence on pre-existing contact networks as the means at hand,” contrasting this with
idea of going out and building a social network as part of the process of creating a new
venture. Garud and Karnøe (2003, p. 278) used the term bricolage even more broadly to
“connote resourcefulness and improvisation on the part of involved actors,” contrasting the
bricolage of Danish wind turbine designers and producers with the more engineering-like
“breakthrough” path pursued by their counterparts in the USA.
Shortly afterwards, Baker and Nelson (2005, p. 333) drew on the cross-disciplinary
literature on bricolage and their own fieldwork to develop an integrative definition of
bricolage as “making do by applying combinations of the resources at hand to new
problems and opportunities.” This definition is restrictive in the sense that it is intended to
IJEBR exclude forms of making do that do not involve both combinations of the resources at
23,1 hand and application to new problems and opportunities. In subsequent work in
organization studies, this has become the dominant definition of bricolage (Rönkkö et al.,
2013; Senyard, 2015).
During the last decade a substantial body of work in entrepreneurship and organization
studies more broadly has applied the concept of bricolage to a wide variety of topics. Garud
116 and Karnøe (2003) explored how multiple actors become engaged in emerging technological
paths, underscoring themes of distributed and embedded agency. Baker and Nelson (2005)
were primarily interested in exploring how bricolage might contribute to or detract from
venture growth under conditions of substantial resource constraints. Other early conceptual
development cleanly distinguished bricolage from the concept of improvisation with which
it had frequently been conflated (Baker, 2007). Similarly, both Baker et al. (2003) and Fisher
(2012) established the distinctiveness of bricolage from effectuation. This work has made it
clear that, operationalizations of those other constructs should not be used to test
hypotheses about entrepreneurial bricolage, and we therefore do not incorporate a
discussion of such measures in this paper.
Whilst much of the ensuing bricolage research has focused on various aspects of
innovation (Brown and Duguid, 1991; Ferneley and Bell, 2006; Senyard et al., 2014) new
research themes are beginning to emerge. These include topics related to the construction
and functioning of teams (Bechky and Okhuysen, 2011), exploration of bricolage processes
(Desa and Basu, 2013) and the effects of environmental constraints and munificence
(Senyard, 2015). A particularly vibrant subtheme has been the application of bricolage to
social entrepreneurship and entrepreneurship in emerging markets (Bacq et al., 2015; Bojica
et al., 2014; Gundry et al., 2011).
As of early 2016, the Garud and Karnøe (2003), Baker et al. (2003) and Baker and Nelson
(2005) papers, all of which are examples of inductive grounded theory development, have
been cited close to 3,000 times according to Google Scholar. This reflects a growing stream
of research on entrepreneurial bricolage. The vast majority of subsequent studies have been
either additional qualitative field studies or else straightforward applications of the concept
to characterize behavior under resource constraints. Our assessment is that the current
trajectory of work on entrepreneurial bricolage threatens to devolve back into the sort of
abstract metaphorical signal of “making do” that it represents in many of the humanities
and other social science literatures. Of the studies that have attempted any sort of
quantitative tests, some are ad hoc and comprise of but a few items (e.g. Salunke et al., 2012),
often without any evidence of internal consistency (e.g. Banerjee and Campbell, 2009) or
include bricolage items as part of a broader measure (e.g. Hmieleski and Corbett, 2006;
Leybourne and Sadler-Smith, 2006). Others appear context-specific, demand post hoc coding,
and are impossible to replicate based on the information provided (e.g. Desa, 2012; Desa and
Basu, 2013). Although Rönkkö et al. (2013) represent a serious and well-documented scale
development effort a considerable problem with their measure appears to be its practicality.
Its length and response format require mail or online data collection, which in their case
resulted in response rates below 10 percent. Further, the multi-dimensional nature of their
measure poses both opportunities and challenges from a theoretical point of view.
The project of developing a behavioral theory of entrepreneurial bricolage threatens to
become stalled because of a lack of a broadly applicable measurement tool reflecting
bricolage as it is theoretically defined. This paper presents a systematic assessment of one
such measurement tool developed for deductive, theory testing, quantitative studies.

3. Developing the measure


Table I summarizes the process of developing the measure. The text below provides
additional detail.
Time Activity Design issues
Entrepreneurial
bricolage
September-December 2006 Internal generation and Conceptualization. Uni- vs multi-dimensional behavior
discussion of items; dimensions; scale in terms of aspects of bricolage and
response formats domains of use; formative vs reflective
December 2006- Pre-test of 20 Likert-scale items Observed:
January 2007 on convenience sample (n ¼ 71) Some extreme response distributions
Some high item non-response 117
Comprehension of “resources”
Dimensionality
Order effect (fatigue?)
Reverse items questionable
January-February 2007 Pre-test of 20 items organized as Observed:
10 matched pairs on (same) Online response rate lower
convenience sample (n ¼ 44) Some extreme response distributions
Dimensionality
Comprehension of “industry standard”
February-March 2007 “Rolling” pre-test of 9-item Implemented:
(design) “never-always” scale on national Phone interviewing for better response rate
March-May 2007 (data random sample of founder- and data quality
collection and analysis) owners of nascent/young firms Six old and three new items; no reverse items
via telephone (n ¼ 78), Increased blending of different bricolage
representing the first set of aspects within same item
respondents to the CAUSEE Behavioral response scale and increased use
Wave 1 main study of behavioral main verbs
Observed:
Dimensionality (minor)
Response distributions (minor)
June 2007-April 2008 CAUSEE Wave I main study None
July 2008-June 2009 CAUSEE Wave 2 main study Observed:
Fit of Item 9 Table I.
October 2009-June 2010 CAUSEE Wave 3 main study Implemented: Overview of the scale
Change from “acquire” to “access” in Item 9 development process

3.1 Conceptualization
We choose a reflective rather than formative measurement logic (see Coltman et al., 2008;
George and Marino, 2011) for two main reasons. First, attempting to capture and compare
the frequency of manifest bricolage behaviors that “form” the overall extent of bricolage
across relevant domains is not feasible in heterogeneous samples of start-ups. Second, the
reflective logic has a better established set of tools for evaluating operationalization
(Edwards, 2010).
We started with the intention of covering several salient aspects of entrepreneurial
bricolage, namely: making do with resources at hand, recombining resources for new
purposes, refusal to enact limitations, and bias for action. However, we conceive
entrepreneurial bricolage behavior in a holistic fashion, assuming that doing more of just
one of these aspects does not constitute bricolage in the complete absence of the others.
Accordingly, we developed a measure intended to adequately capture the breadth of the
theoretical construct – thus achieving content validity – while behaving as a unidimensional
construct in technical terms (cf. Covin and Wales, 2012).

3.2 Item generation and pre-testing


We followed standard protocols (e.g. DeVellis, 2003) in developing the measure. Two
investigators generated the initial pool of items. Having significant prior research on
entrepreneurial bricolage (Baker and Nelson, 2005; Baker et al., 2003) Baker acted as the
IJEBR “theory expert.” Based on significant prior experience of scale development and survey
23,1 research in general (Brown et al., 2001; Davidsson, 2016; Davidsson and Klofsten, 2003)
Davidsson took the role as “method expert.” The process of item generation was iterative
and included many item and wording variations, as well as consultation with
knowledgeable colleagues. It was led by Baker based on theoretical considerations and
with the intent to include the above-mentioned aspects of bricolage without necessarily
118 capturing them as distinct dimensions.
Two partly overlapping sets of items were pre-tested on a convenience sample
(e.g. alumni start-ups, guest speakers’ businesses). First, 20 Likert-scale items were
administered via phone interviews. As noted in Table I, this test revealed numerous issues.
These were dealt with by dropping and modifying items; altering the response scale, and
creating a stronger introduction to the item package. While exploratory factor analysis
(EFA) yielded as many as seven factors with an eigenvalue W1, further analysis showed
that three separate sets with seven, three and seven items, respectively, held some promise
as summated indices based on near-satisfactory Cronbach’s α coefficients (0.75; 0.69, and
0.66). Two of these consisted pre-dominantly of reverse-worded items (suggesting the
absence or opposite of bricolage) and had several items with non-negligible item non-
response. This raised concerns that “non-bricolage” behaviors do not necessarily have
bricolage as their only possible opposite. Similarly, recent research questions the soundness
of the popular textbook recommendation of mixing positive and negative items (Barnette,
2000; Stewart and Frye, 2004; Sauro and Lewis, 2011; Wong et al., 2003). The third set
consisted of seven positively worded items; that is, descriptions of bricolage practices rather
than their presumed absence or opposite.
We concluded that the only defensible way of including “negative” items would be to
develop contrasting item pairs (cf. Brown et al., 2001; Rönkkö et al., 2013). Therefore, our
second pre-test used an adapted set of ten matched item pairs, delivered as an online survey.
The sample was the same as for the first pre-test, although a markedly lower response rate
was achieved, indicating a distinct disadvantage of the online format. With four factors
having eigenvalue W1 dimensionality problems remained. The results also led to the insight
that many respondents do not have a clear notion of an existing “industry standard” with
which to compare. Hence, we excluded that notion from further use.
Based on these pre-testing experiences we developed a new instrument starting from the
set of seven positively worded items from the first pre-test. We implemented changes as
indicated in Table I. These changes served to increase cooperation and response quality as
well as to strengthen the unidimensional, behavioral nature of the scale. The result was the
nine-item, telephone interview instrument displayed in the left hand, “original items” column
in Table II.
This version of the scale was pilot tested on a national random sample of 78 nascent
entrepreneurs and young firm owners (31 and 47 respondents, respectively) comprising the
first set of respondents to Wave 1 of the Comprehensive Australian Study of
Entrepreneurial Emergence (CAUSEE). Detailed descriptions of this study, which used
the sampling methodology of the Panel Study of Entrepreneurial Dynamics II (Reynolds,
2009) are provided in Davidsson and Steffens (2011), Davidsson et al. (2011) and Grünhagen
et al. (2016).
This final pre-test yielded positive and statistically significant correlations among all
bricolage items, ranging from 0.31 to 0.68. As a summated index the nine items reached a
Cronbach’s α of 0.84, and no exclusion of items would further increase that number.
A factor analysis extracted a first factor accounting for 47 percent of the variance, which
we considered satisfactory given the conceptual breadth of the construct. Further, the
intended use was as a summated index, retaining all the information in the responses.
Although there was also a second factor with an eigenvalue marginally above 1 (1.11) and
READ: The following statements are about how your business uses various kinds of resources to deal with
Entrepreneurial
new challenges. By resources we mean things like materials, equipment, people or anything else that can be bricolage
used to get a job done. By challenges we mean both new problems and new opportunities. When I say “we” or behavior
“our” I mean you personally or anybody else acting on behalf of the business. I want you to respond on a scale
from 1 to 5, where 1 means “never” and 5 means “always”:
Q1 OK, does the following represent how you never, rarely, sometimes, often, or always go about doing things
for your start-up? First, … READ STATEMENT
Response scale: 1, never; 2, rarely; 3, sometimes; 4, often; 5, always (8, refused; 9, don’t know) 119
Possible revisions to be considered (after pre-testing)
Original items in future research

1. We are confident of our ability to find workable 1. We usually find workable solutions to new
solutions to new challenges by using our existing challenges by using our existing resources
resources
2. We gladly take on a broader range of challenges 2. We typically take on a broader range of challenges
than others with our resources would be able to than others with our resources would do
3. We use any existing resource that seems useful to
responding to a new problem or opportunity
4. We deal with new challenges by applying a
combination of our existing resources and other
resources inexpensively available to us
5. When dealing with new problems or opportunities 5. When dealing with new problems or opportunities
we take action by assuming that we will find a we immediately take action by assuming that we
workable solution will find a workable solution
6. By combining our existing resources, we take on a 6. By combining our existing resources, we take on a
surprising variety of new challenges very broad variety of new challenges
7. When we face new challenges we put together
workable solutions from our existing resources
8. We combine resources to accomplish new
challenges that the resources were not originally
intended to accomplish Table II.
9. To deal with new challenges we acquire resources 9. To deal with new challenges we access resources at The bricolage measure
at low or no cost and combine them with what we low or no cost and combine them with what we instructions, items and
already have already have response scale

two items with communalities below 0.5 we concluded there was not strong enough
evidence to guide further revision of the items, considering the limited number of cases
underlying the analysis. Further, interviewer feedback cautioned against increasing the
number of items. Hence, CAUSEE Wave 1 was implemented without further revisions to
the bricolage scale.

4. Testing the measure


4.1 Test samples
Our primary test samples are taken from CAUSEE, which captures a national random
sample of start-ups in all industries by directing screening interviews to next-birthday adult
members of households, sampled via random digit dialing (Reynolds, 2009). In addition to
the national random sample of “nascent entrepreneurs” (Sample 1), CAUSEE also features
additional samples:
• Sample 1: a national random sample of 625 nascent entrepreneurs (NE-Random).
• Sample 2: a national random sample of 559 owners of young firms (YF-Random). These
were captured via the same screening procedure; they were too long in operation for the
NE sample but were included if the firm had been in operation for less than 3.5 years.
IJEBR • Sample 3: a judgment over-sample of 106 “higher potential” nascent entrepreneurs
23,1 (NE-HP). To avoid biases associated with particular sampling sources we identified
these through as many different sources as possible that would encounter such firms
at an early stage (e.g. patent attorneys; incubators; research centers, etc.).
• Sample 4: a judgment over-sample of 120 “higher potential” young firms identified
through the same variety of sources as Sample 3 (YF-HP).
120 These samples constitute the data to which we have direct access, and they are the main
samples used to test our instrument. In the analyses we use them separately, combined, and
in breakdowns into different categories of firms and individuals. The bricolage measure was
included in the first three waves of data collection (W1, W2, W3), conducted at 12-month
intervals. This provides additional testing and validation opportunities. Naturally, the
number of cases in the samples decreases for each wave.
We use additional validation evidence from four published studies by other researchers
who have applied the measure with our permission. These are:
• Sample 5: a sample of 160 female Palestinian business founders (Bojica et al., 2014).
• Sample 6: a sample of 253 family firm advisors in the USA (Davis et al., 2013).
• Sample 7: a sample of 113 social entrepreneurs in the USA (Gundry et al., 2011).
• Sample 8: a sample of 123 social entrepreneurs in the USA (Bacq et al., 2015).
These published studies provide data on different contexts and types of participants as well
as variations in data collection mode and response formats. We do not have access to the
raw data for these samples. Hence our reporting is restricted to published data
supplemented by information from email correspondence. The cross-sectional nature of data
from Samples 5 to 8 means they are prone to common-method bias in estimating
relationships among variables (Podsakoff et al., 2003). Therefore, we use information on
these samples solely for assessing reliability and discriminant validity, and not for
nomological validity.
Figure 1 provides an overview of the sequence of the types of validity tests that follows,
including the specific type of test criteria used and what samples are employed for each test.

4.2 Content validity


Content validity refers to the instrument’s ability to capture the essential contents of the
theoretical construct it is supposed to measure (Haynes et al., 1995). Our most important
approach to securing content validity was to engage an expert in bricolage theory in the
development of the measure. As described above, balancing content validity against the risk
of respondent fatigue, we aimed to attain satisfactory results on technical criteria while
capturing the full breadth of the construct. To this end, we blended different aspects of
bricolage in the items so as to obtain a holistic assessment of bricolage rather than distinct
sub-dimensions. Collectively, the items provide good representation of different theoretical
aspects of bricolage. For example, (making do with) resources at hand is mentioned in Items
1, 2, 3, and 7; resource combinations in Items 4, 6, and 8; putting resources to new uses in
Items 3 and 8; workable (rather than “right”) solutions in Items 1, 5, and 7, and taking on a
broad set of challenges (refusal to enact limitation) in Items 2 and 6. Our subsequent critical
examination revealed, however, that some items may have unnecessary content
(“confident,” gladly,” “surprising”) beyond the core of the theoretical notion of
entrepreneurial bricolage. We will consider this in our discussion of possible improvements.
Respondent comprehension of the items, as indicated by low internal non-response, is
sometimes used as support for content validity (e.g. Noest et al., 2014). Table III indicates
that problems of internal non-response are limited in the final version of our scale.
Entrepreneurial
Content validity (Section 4.2)
Do the items (fully and only) reflect the content of the theoretical bricolage
construct “entrepreneurial bricolage”?
behavior
Test: Conceptual evaluation; item non-response; response distributions

121
Reliability/Internal consistency (Section 4.3)
Do the items reliably reflect the same phenomenon?
Test: Pearson
Test: Factor Test: CFA Criteria correlation across
Test: Cronbach’s
structure/loadings (GFI, AGFI, CFI,
 waves (test-retest
Samples: 1-6 + RMSEA) reliability)
Samples: 1-8
subsamples by types Samples: 1-4, 6-7 Samples: 1-4
Table 3
Table 3 Table 3 Table 4

Discriminant validity (Section 4.4)


Is entrepreneurial bricolage distinct from other measures/phenomena?

Distinct from another behavioral construct Distinct from other multiple-item measures (four
(business idea changes)? dimensions of competitiveness)
Test: Factor structure/loadings Test: Factor structure/loadings
Test: CFA Criteria Test: CFA Criteria
Samples: 1-4 Samples: 1-4
Table 5 Table 6

Nomological validity (Section 4.5)


Does the bricolage measure relate to measures of other constructs as theoretically expected? Figure 1.
Overview of the
Decrease in bricolage over time? More experience→More bricolage? More bricolage→More changes?
Test: t-test of means (paired) Test: t-test of means (indep.) Test: Pearson correlation validation testing
Samples: 1-4 Samples: 1-4 Samples: 1-4 procedure and
Table 7 Reported in body text Reported in body text sequence

However, Figure 2 reveals that in an otherwise normal-looking distribution, 8 percent of


the respondents had the maximum score of 5. The cluster of respondents at the high
extreme could mean that the measure fails to capture the full range of variation at high
levels of bricolage use. Checks against other measures suggest, however, that this group
of respondents have a general tendency toward giving answers at the extremes of scales.
We will consider the response distribution issue in our suggestions for future
improvements, noting that Item 5 (mean ¼ 4.31-4.45) and Item 1 (mean ¼ 4.09-4.26) have
the most positively skewed distributions in each wave. It should be noted, however, that
the problem abates over time. For example, the proportion with the maximum score is
down to 4 percent in W3, Sample 2.

4.3 Reliability/internal consistency


As mentioned above, our measure is intended to be a unidimensional, reflective assessment
of entrepreneurial bricolage behavior. This renders relevant criteria such as factor structure
and Cronbach’s α. Initial analyses demonstrated that Item 9 had a negative influence on the
results according to multiple criteria in several sub-samples. We therefore decided to
exclude this item from further analysis. We also identified the use of the verb “acquire”
(rather than “access” or “obtain”) as a possible reason for its inadequate performance.
23,1

122

samples
IJEBR

Table III.

measure across
reliability/internal
consistency of the
eight-item bricolage
Results pertaining to
Section/item Pct variance
Delivery non-response No. of factors Eigenvalue extracted; Loadings Cronbach’s
Sample n format (%) (eigenvalueW 1) Factor 2 Factor 1 range α GFI AGFI CFI RMSEA

Samples 1-4 combined (W1) 1,357 Telephone 3.8 1 – 45 0.57-0.72 0.82 0.94 0.89 0.90 0.11
Sample 1: NE-Random 596 d:o 4.6 1 – 45 0.62-0.71 0.82 0.92 0.86 0.87 0.12
Sample 2: YF-Random 540 d:o 3.4 1 – 46 0.57-0.73 0.83 0.93 0.88 0.90 0.10
Sample 3: NE-HP 104 d:o 1.9 1 – 47 0.51-0.80 0.84 0.93 0.87 0.94 0.08
Sample 4: YF-HP 117 d:o 2.5 2 1.063 35 0.42-0.74 0.74 0.93 0.88 0.88 0.08
Samples 1-4 combined (W2) 863 d:o 5.8 1 – 47 0.64-0.73 0.84 0.93 0.90 0.94 0.15
Sample 1: NE-Random 319 d:o 6.5 1 – 46 0.59-0.74 0.83 0.93 0.88 0.90 0.10
Sample 2: YF-Random 374 d:o 5.3 2 1.058 49 0.63-0.75 0.85 0.87 0.77 0.82 0.15
Sample 3: NE-HP 81 d:o 4.7 2 1.186 44 0.49-0.80 0.80 0.88 0.79 0.85 0.12
Sample 4: YF-HP 89 d:o 6.4 2 1.122 46 0.57-0.73 0.82 0.86 0.75 0.82 0.14
Samples 1-4 combined (W3) 632 d:o 4.1 1 – 50 0.67-0.78 0.85 0.93 0.87 0.91 0.12
Sample 1: NE-Random 203 d:o 6.0 2 1.042 48 0.55-0.78 0.84 0.90 0.82 0.87 0.12
Sample 3: YF-Random 296 d:o 4.2 1 – 52 0.66-0.80 0.87 0.90 0.83 0.89 0.13
Sample 2: NE-HP 61 d:o 1.6 2 1.121 48 0.56-0.83 0.83 0.85 0.74 0.89 0.11
Sample 4: YF-HP 72 d:o 0.0 2 1.196 44 0.52-0.78 0.82 0.85 0.74 0.83 0.13
Sample 5: palestinian 160 Face-to- n/a 1b – n/a 0.70-0.89b,c 0.82c n/a n/a n/a n/a
a
women face
e e f
Sample 6: US family 253 Online n/a 1 – 48 0.61-0.74 0.88 n/a n/a 0.95 0.06f
business advisorsa,d
Sample 7: social 113 Online n/a n/a n/a n/a n/a 0.84g n/a n/a 0.95f,g 0.06f,g
a,d
entrepreneurs
Sample 8: social 123 Online n/a n/a n/a n/a n/a 0.78 n/a n/a n/a n/a
entrepreneursd
Notes: Except for the last four columns, the results are based on exploratory factor analysis (EFA) with principal components (PC) extraction, varimax rotation, and list
wise deletion of missing cases. All these analyses have Kaiser-Meyer-Olkin (KMO) coefficients in the 0.73-0.88 range and p o0.001 on Bartlett’s test, well exceeding
conventional criteria (Hair et al., 1995; Tabachnick and Fidell, 2007). The last four columns are based on the corresponding confirmatory factor analyses (CFA), which
(at least for Samples 1-4) were run with mean-substitution for missing values due to software (AMOS) requirements (therefore the actual n in CFA deviates slightly from
the figure displayed). aThe study used “I” rather than “we” in the item wording; bbased on the reported, multi-construct CFA; cbased on seven items only (Item 7 excluded
due to marginally better CFA results); dthe study used seven-point agree-disagree response scales rather than our five-point never-always original; econfirmed in personal
communication; fbased on a model with multiple (3-6) factors; gbased on the original nine items (no item dropped)
Entrepreneurial
bricolage
behavior

123

Figure 2.
Distribution of scores
on the summated
bricolage index
1.00 2.00 3.00 4.00 5.00 6.00

Table III displays the results of the factor- and reliability analyses pertaining to the
remaining eight items. From the top, the table first displays results for our main samples
across waves of data collection. This is followed by results reported in published research
by other investigators using the same measure.
EFA provides a tougher test than does confirmatory factor analysis (CFA) in the sense
that EFA reports the best factorial representation of the data at hand without any bias
toward a model favored by the analyst, whereas CFA asks the question “Is it possible that
these data are drawn from a population with factor structure X?” Further, conventional EFA
requires factors to be uncorrelated, which imposes strong demands on discriminant validity
(Tables V and VI). This said, based on assumptions about the extent and nature of
measurement errors some argue that CFA is superior for identifying dimensionality
(e.g. Levine, 2005). Further, CFA is associated with what have become established rules-of-
thumb for evaluating model fit (Hair et al., 2010; Hu and Bentler, 1999). Therefore, we also
ran and report fit measures from CFA analyses. Substantively the CFA results are
near-identical to the reported EFA results.
In many respects the results follow what could be expected from our “rolling” pre-test.
The data for internal non-response (which include dropping out completely earlier in the
interview) do not trigger any alarm. Some analyses yield two factors; however, this typically
occurs in small samples and the second factor only marginally exceeds 1 in eigenvalue. We
find the variance extracted (typically 45-50 percent) and loadings (typically within the 0.50
to 0.80 range) satisfactory considering the conceptual breadth of the construct. Importantly,
breakdown analyses (not displayed) suggest the measure performs well across a range of
different types of ventures and respondents, such as product vs service firms; online vs
brick-and-mortar; solo vs team-based firms; male vs female founders, and founders
representing varying levels of education and experience. All these analyses yield one factor
with eigenvalue W1 which extracts 44-50 percent of the variance, and a Cronbach’s α of
0.80-0.85. Broad applicability is further supported by the results for Samples 5-8. The results
are particularly satisfactory in terms of reliability (internal consistency). With a single
IJEBR exception, Cronbach’s α is above 0.80, which is considered good or sufficient (Nunnally,
23,1 1978; Webb et al., 2006).
Although the EFA results are unsatisfactory for Sample 4, especially in W1, there is no
consistent pattern in the two-factor solutions for Sample 4 across waves or across the
two-factor solutions in general. All analyses using more than 400 cases yield a one-factor
solution and a Cronbach’s α of 0.82 or higher. This suggests that the less satisfactory
124 aspects of the results may be stochastic.
As regards the CFA criteria displayed in the last four columns, various sources suggest
the following critical values: GFI W 0.90/0.95; AGFI W 0.80/0.90; CFI W 0.90/0.95, and
RMSEA o 0.05/0.08/0.10 (Chau, 1997; Hair et al., 2010; Hooper et al., 2008; Hu and Bentler,
1999). With the marginal exception of CFI for Sample 2 in Wave 2, the GFI, AGFI, and CFI
measures obtained meet at least the more permissive cutoffs in all analyses with more
than 400 cases. For smaller sub-samples the values are sometimes below the cutoff.
However, these indicators are known to be sensitive to sample size (Hooper et al., 2008);
a close examination of our results for the combined sample and its constituent parts
supports this interpretation.
The RMSEA values appear more clearly problematic as they are above the most
permissive suggested cutoff for the majority of our tests. However, as we shall see
(cf. Tables V and VI) the RMSEA values are satisfactory in multi-construct models. This
reflects the known fact that RMSEA values are inflated in smaller (e.g. single-factor)
models (Fan and Sivo, 2007). Accordingly, lower RMSEA values (under the conventional
0.10 cutoff, and often also below 0.05) are achieved in multi-factor models than in single-
factor models for all constructs in our tests – not just for the bricolage measure. Similarly,
nominally better CFA performance is also achieved in the external Samples 6 and 7, which
are based on multi-factor models. Across the board the bricolage scale therefore seems to
perform satisfactorily albeit not perfectly on technical criteria. A substantive reason for
less than perfect performance on technical criteria could be that indicators of different
aspects of bricolage (making do, resource recombination, etc.) are not optimally
distributed across the items.
The bricolage questions were repeated in three waves. This provides an opportunity to
assess test-retest reliability (Cronbach et al., 1972; Webb et al., 2006). Although some real
changes in the use of bricolage over time are likely to occur (see further below), a sizable
positive correlation across waves should be expected. Table IV reports the results of our
test-retest analysis. Here, like in all subsequent analyses where a bricolage score is used,
the index has been computed with the MEAN command in SPSS. This means that
respondents with partial item non-response are included. An alternative index based on
summing all eight items for respondents with complete data produces near-identical
substantive results.
As can be seen, all the correlations are positive and sizeable (0.30-0.60). Further, in the
larger samples, the W1-W2 correlation is larger than the W1-W3 correlation, which makes
sense in the presence of real, trend-wise changes over time. Further, the W1-W2
correlation is larger for YF samples than for NE samples. This is consistent with the

Sample W1-W2 correlation (n) W2-W3 correlation (n) W1-W3 correlation (n)

Table IV. Samples 1-4 combined 0.42 (1408) 0.48 (624) 0.39 (641)
Correlations across Sample 1: NE-Random 0.30 (325) 0.48 (202) 0.29 (208)
waves for the Sample 2: YF-Random 0.49 (381) 0.46 (293) 0.41 (299)
eight-item bricolage Sample 3: NE-HP 0.35 (82) 0.44 (60) 0.60 (62)
measure Sample 4: YF-HP 0.44 (91) 0.46 (69) 0.39 (72)
assumption that there is more real change in the use of bricolage at the very early Entrepreneurial
stages of firm development. In all, the results strengthen the confidence that the measure bricolage
does not pre-dominantly reflect random noise or response style. The results remain behavior
supportive also in breakdown analyses by type of venture and type of respondent
(not displayed).

4.4 Discriminant validity 125


Discriminant validity requires a measure to be empirically distinct from measures of other
theoretical constructs (Campbell and Fiske, 1959). Arguably, in our case the most relevant
constructs to test against would be other reflective, multiple-item behavioral constructs. Our
data set offers only one good candidate of this nature, namely a four-item measure of the
number of changes to the business idea that have been undertaken in the last 12 months.
The four items cover changes to the core product or service offered, the customers targeted,
the methods of promoting and selling, and the methods of producing or sourcing the core
product/service. These items yield a one-factor solution extracting 52 percent of the variance
and a Cronbach’s α of 0.69 for Samples 1-4 combined in W1. Results in separate samples and
across waves are similar.
Table V displays the results of factor analyses where the eight bricolage items and the
four changes items were entered simultaneously. Due to the instability of the small-sample,
factor-analytic results revealed in Table III we limit the reporting here to analyses of
Samples 1-4 combined and for Samples 1 and 2, separately.
As can be seen, these two constructs are empirically distinct. The analyses consistently
extract two factors, and the bricolage items consistently load on the same factor. The extra
factor for Sample 2 in W2 appeared in Table III and is not due to cross-loadings between the
two constructs. The results also remain supportive in breakdown analyses by type of
venture and type of respondent (not displayed). All the CFA criteria (last four columns) come
out satisfactory in this analysis.
Being drawn from a venture-level study, Samples 1-4 lack personality measures
or other individual-level scales with well-known properties. However, a multiple item
battery aiming at (perceived/self-reported) “competitiveness” can be used for further
exploration of discriminant validity. We first undertook preliminary analyses identifying
a subset of items producing four empirically, as well as conceptually, clear
competitiveness factors: marketing (two items), costs (three items), product
(three items), and networks (three items). The items use a Likert-type “strongly agree –
strongly disagree” response format. They can serve to detect whether the bricolage
measure shares variance with other dimensions that may reflect response style or
respondent fatigue. The package is particularly well suited for the latter since it includes
many items and our analysis retains (among others) the last six-items asked. Subjecting
the bricolage measure to the competition of four other constructs simultaneously also
provides a tougher EFA test in a technical sense than does the above contrast against just
one other construct.
Table VI reports results for Samples 1 and 2 plus Samples 1-4 combined. These results
reaffirm discriminant validity. The bricolage factor comes out as distinct from the
competitiveness factors in each analysis. The single instance of extracting an extra factor is
not due to cross-loadings with the competitiveness factors, and in total there are only two
cross-loadings above 0.30. Again, all CFA cutoff criteria (last four columns) are satisfactory,
and the results remain supportive also in breakdown analyses by type of venture and type
of respondent (not displayed).
Further support for discriminant validity is provided in the published results from
Sample 6. Davis et al. (2013) identified our eight-item bricolage measure as a clearly distinct
factor in an analysis including five other factors (cf. Table III).
23,1

126
IJEBR

Table V.

for bricolage
concurrently with
business idea changes
Factor analysis results
No. of bricolage No. of cross-
No. of factors w. Pct variance items having highest loadings W0.30
Sample n (eigenvalueW 1) extracted, 2 factors loading on same factor (2-factor solution) GFI AGFI CFI RMSEA

W1: Samples 1-4 combined 1,342 2 47 8 0 0.98 0.97 0.97 0.05


Sample 1: NE-Random 588 2 47 8 0 0.96 0.94 0.94 0.05
Sample 2: YF-Random 536 2 49 8 1a 0.96 0.95 0.96 0.04
W2: Samples 1-4 combined 859 2 49 8 0 0.97 0.95 0.96 0.05
Sample 1: NE-Random 316 2 48 8 0 0.94 0.91 0.92 0.06
Sample 2: YF-Random 374 3 43 8 0 0.95 0.92 0.94 0.06
W3: Samples 1-4 combined 631 2 51 8 0 0.96 0.94 0.96 0.06
Sample 1: NE-Random 203 2 50 8 0 0.92 0.88 0.93 0.06
Sample 2: YF-Random 296 2 52 8 0 0.93 0.90 0.93 0.07
Notes: Except for the last four columns, the results are based on exploratory factor analysis (EFA) with principal components (PC) extraction, varimax rotation, and list
wise deletion of missing cases. All these analyses have KMO above 0.80 and p o0.001 on Bartlett’s test, well exceeding conventional criteria (Hair et al., 1995; Tabachnick
and Fidell, 2007). The last four columns are based on the corresponding confirmatory factor analyses (CFA), which (at least for Samples 1-4) were run with
mean-substitution for missing values due to software (AMOS) requirements (therefore the actual n in CFA deviates slightly from the figure displayed). aOne changes item
loads 0.31 on the bricolage factor
No. of bricolage
Pct variance items w. highest loading No. of
No. of factors extracted, on the same factor cross-loadings W 0.30
Sample Min. n w. eigenvalue W 1 5 factors (5-factor solution) (5-factor solution) GFI AGFI CFI RMSEA

W1: Samples 1-4 combined 1,290 5 62 8 0 0.97 0.96 0.99 0.04


Sample 1: NE-Random 564 5 64 8 0 0.95 0.93 0.96 0.04
Sample 2: YF-Random 514 5 62 8 0 0.94 0.92 0.95 0.04
W2: Samples 1-4 combined 799 5 65 8 0 0.97 0.95 0.97 0.04
Sample 1: NE-Random 377 5 65 8 0 0.93 0.91 0.96 0.04
a b
Sample 2: YF-Random 348 6 65 8 1 0.92 0.89 0.93 0.05
W3: Samples 1-4 combined 598 5 66 8 0 0.97 0.95 0.97 0.04
Sample 1: NE-Random 247 5 65 8 0 0.92 0.89 0.97 0.03
Sample 2: YF-Random 282 5 67 8 1c 0.91 0.88 0.93 0.06
Notes: Except for the last four columns, the results are based on exploratory factor analysis (EFA) with principal components (PC) extraction and varimax rotation, due
to the large number of items included, pairwise deletion of missing cases was employed in these analyses. All these analyses have KMO above 0.77 and p o0.001 on
Bartlett’s test, well exceeding conventional criteria (Hair et al., 1995; Tabachnick and Fidell, 2007). The last four columns are based on the corresponding confirmatory
factor analyses (CFA), which were run with mean-substitution for missing values due to software (AMOS) requirements (therefore the actual n in CFA deviates slightly
from the figure displayed). aDue to the previously observed split of bricolage factor; bbricolage Item 6 cross-loads 0.36 on the product factor; cbricolage Item 8 cross-loads
0.37 on the networks factor
127
Entrepreneurial

types of
bricolage
behavior

competitiveness
concurrently with four
Table VI.

results for bricolage


Factor analysis
IJEBR 4.5 Nomological validity
23,1 Nomological validity – aspects of which are sometimes called “predictive,” “convergent” or
“concurrent validity” – is the degree to which the focal construct relates to other constructs
in a way consistent with a priori expectations (Danneels, 2015). This is an important
criterion. Reliability and discriminant validity only demonstrate that a measure captures
something empirically consistent and distinct. Alongside content validity, nomological
128 validity needs to be demonstrated in order to make the case that this something is the focal,
theoretical construct.
Nomological validity is a particularly challenging criterion when working with an
emerging theory like entrepreneurial bricolage, because there is relatively little firmly
established knowledge about how bricolage should relate to other variables. When trying to
establish nomological validity it is important to rule out reverse causality, response style, or
other sources of common-method bias (Podsakoff et al., 2003). Therefore, time separation is
important, as is associating bricolage with variables that were measured differently.
Combining theorizing and prior research with technical considerations we find reason to test
the expectations that:
(1) The use of bricolage should decrease over time as nascent and young firms mature.
There are at least two theoretical reasons to expect this. First, once a venture starts to
develop routine ways of looking at the world, and responding in repetitive ways to the
sorts of opportunities and challenges it commonly faces, demands for bricolage – which
is a response to new challenges and opportunities – may decline. Second, once a firm
begins generating cash inflows, it may be better positioned to acquire the “right”
resources rather than continuing to make do with what is at hand. Samples 1-4 provide
repeated measurements, and here the issue of similar response format is void.
(2) More experienced business founders use more bricolage than novices at early stages
of venture development. Consistent with Levi-Strauss’s (1966) initial insights, the
use of bricolage appears to be positively affected by the skills and experience of
the bricoleur and by the bricoleur’s understanding of – and learning about – the
characteristics of the resources at hand (Harper, 1987; Senyard, 2015). In addition,
some evidence suggests that individuals develop bricolage skills through content
knowledge of the “trove” (Baker and Nelson, 2005) and the transfer of knowledge
between different activities (Lehner, 2000). We would therefore expect more
experienced founders to engage in more bricolage early on than those who have to
learn both about how to perform bricolage and about the resource troves at hand.
Involvement in prior start-ups is a historical fact and is not assessed on scales
similar to our bricolage measure. The historical experience is factually time-
separated even if reported at the same time as bricolage (W1) and separated also in
terms of time of measurement for W2 and W3.
(3) Business founders who use more bricolage are likely to undertake more changes to
the “business idea” or “business model” they are trying to realize. First, a recurring
theme in existing work is that the use of bricolage increases the likelihood that
ventures will experience surprise events and outcomes (Baker, 2007). Some surprises
will lead to perceived-as-positive new options, leading to changes in the orientation
of the business. Second, adherence to bricolage principles may more often lead to
situations where the solution originally conceived approach simply cannot be
implemented. This may also increase the frequency of observed changes.
Table VII displays results comparing the use of bricolage over time. All contrasts suggest
decrease over time. The effect size is moderate (0.04-0.20 scale steps per year) but
consistently statistically significant in the larger samples. Thus, the results clearly indicate
Entrepreneurial
Sample W1 bricolage – W2 bricolage W2 bricolage – W3 bricolage W1 bricolage – W3 bricolage
bricolage
Sample 1-4 combined 4.00-3.86 3.89-3.78 4.01-3.79 behavior
( p o0.001; n ¼ 879) ( p o 0.001; n ¼ 624) ( p o0.001; n ¼ 641)
Sample 1: NE-Random 4.04-3.85 3.85-3.75 4.04-3.77
( p o0.001; n ¼ 325) ( p ¼ 0.02; n ¼ 202) ( p o0.001; n ¼ 208)
Sample 2: YF-Random 3.90-3.78 3.83-3.69 3.95-3.70
( p o0.001; n ¼ 381) ( p o 0.01; n ¼ 293) ( p o0.001; n ¼ 299)
129
Sample 3: NE-HP 4.07-3.96 4.06-3.96 4.11-3.97
( p o0.10; n ¼ 82) ( p o0.01; n ¼ 60) ( p o0.05; n ¼ 62)
Sample 4: YF-HP 4.15-4.11 4.14-4.04 4.12-4.05
( p W10 (ns); n ¼ 91) ( p W0.10 (ns); n ¼ 69) ( p W0.10 (ns); n ¼ 72)
Notes: Based on paired-samples t-tests. One-tailed significance levels are reported, consistent with p o0.05 meaning Table VII.
less than five percent risk of a false positive result for a directional hypothesis. The mean displayed for a particular Changes in the use of
sample in a particular year may vary because slightly different sets of (surviving) cases are included in the analyses bricolage over time

reduction in bricolage over time, although the combination of small effect size and small
sample size renders the result non-significant for Sample 4.
The results also support the notion that experienced founders use bricolage more than
novices. We compare ventures with novice founders to those whose founders have
experience of either one or multiple prior start-ups. In Samples 1-4 combined these three
groups have bricolage scores of 3.89, 3.97, and 4.10, respectively, in W1 (n ¼ 620/278/497).
Independent sample t-tests show that all three contrasts are statistically significant; for the
novices vs multiple start-up experience cases the result achieves p o0.001. The pattern
replicates for Samples 1 and 2 individually and for each consecutive wave (albeit ns, for
Sample 2 in W2). For Samples 3 and 4 the subgroup sizes are too small for a meaningful
analysis. In all, the expectation that experienced entrepreneurs use more bricolage than
novices at early stages of venture development is supported.
We use the Changes index described above to assess whether bricolage is associated
with undertaking more changes to the business idea. If bricolage is causally related to
making changes, the most relevant tests correlate bricolage with concurrent and next-period
changes. Each sample has three correlations of each type (W1-W3). In Samples 1-4
combined, the average Bricolaget0 with Changest0 correlation is 0.14 (n ¼ 640-1,390;
p o0.001 in each case). In terms of direction this result is replicated for 11 of 12 correlations
in Samples 1-4, separately. The result reaches statistical significance ( p o0.05 or lower) in
four cases. The average Bricolaget0 with Changest1 correlation in Samples 1-4 combined is
0.15 (n ¼ 494-882; p o0.001 in each case). This result replicates in direction for 10 out of 12
correlations in the separate samples, and reaches statistical significance ( p o0.05 or lower)
in seven cases. This all indicates a positive association between bricolage and changes to the
business idea. It should also be noted that these correlations are likely to be substantially
underestimated. First, Pearson correlations are systematically underestimated for variables
measured on non-continuous scales (Holgado-Tello et al., 2010). Second, while the
distribution of the bricolage measure is near-normal, the changes index is zero-inflated. It is
impossible for variables with radically different distributions to have very high correlations.
Regarding other relationships, Senyard et al. (2014) presented competing hypotheses of a
positive or inverted U-shape relationship between bricolage and innovativeness and found
support pre-dominantly for the former, using a trimmed version of Samples 1 and 2
combined. It is also worth noting that there is no difference of meaningful magnitude in the
reported use of bricolage between male and female respondents. To illustrate, the W1
average Bricolage score for male and female respondents in Samples 1-4 combined are very
similar (3.99 vs 3.97, ns, n ¼ 847/561). Social desirability is known to vary systematically by
IJEBR sex in many domains (Hebert et al., 1997; Dalton and Ortegren, 2011). The absence of a
23,1 male-female difference in bricolage is therefore an indirect indication that our bricolage
measure is not strongly affected by social desirability.

5. Discussion
Entrepreneurial bricolage refers to a complex and open-ended set of behaviors. It may
130 manifest itself in a vast number of ways across different contexts. It would therefore be
naïve to think that a survey instrument – any survey instrument – could ever capture and
compare the “amount” of such behavior with very high precision (Toomela, 2008). If we had
all of the correct, behavioral evidence at hand, not even a group of experts would fully agree
on the relative amount of bricolage exercised by farmer A, compared to mobile app
developer B, compared to budding restaurateur C. A survey measure is obviously less likely
than experts to achieve a near-perfect correlation with the “true score” (Lord, 1965). Yet, we
have to try to approach that unreachable ideal because otherwise the avenues to ever testing
our theories remain very restricted.
In short, a measure like this is likely to have substantial measurement error. As long as
this measurement error is random, the measure is still useful. What we can hope for is an
instrument that is accurate enough to lead us to correct conclusions in tests of coarse-
grained predictions like the sign (e.g. positive) and form (e.g. inverted U-shape) of
substantial empirical relationships in large enough samples. It should be remembered that
random measurement error will lead to an underestimation of true relationships, whereas
systematic error (e.g. common-method bias) may inflate them. Therefore, there is reason to
caution against chasing – or reporting – subtle effects that are small in magnitude and/or of
a complex structure, especially if the sample is of a modest size. Such findings have proven
particularly hard to replicate even in contexts fraught with far less challenging
measurement problems (Open Science Collaboration, 2015). This said, the example of
entrepreneurial orientation suggests that a less than perfect measure designed to apply to
broadly based samples can assist in meaningful knowledge development in our field
(George and Marino, 2011; Rauch et al., 2009; Wales, 2016).
At this juncture, we find reason to be hopeful that our suggested bricolage measure is
such an instrument. Our analyses provide several indications that our eight-item measure
represents the theoretical construct “entrepreneurial bricolage” with a meaningful level of
precision. The evidence also indicates that the measure works across a broad spectrum of
contexts and data collection formats. If so, it is a useful tool that should be applied in
future research.
This said, there may be room for further improvement. We do not believe that expanding
the instrument would be a good idea as this would likely lead to respondent fatigue and
non-response problems. There is likely room, nonetheless, for improvements to the current
set of items in order to reduce the response distribution problem and make the items more
strictly focused on the core issue of frequency or amount of bricolage behavior, removing
unwanted influence of the abilities, emotions, motivations, and circumstances associated
with such behavior. In the right-hand column of Table II we offer some suggestions of this
kind. However, we would strongly recommend thorough pre-testing of such design changes
before employing them in full scale data collection. We also recommend that any additional
attempts at modification or addition of items consider: focusing strictly on frequency of
bricolage behavior, and avoiding positive outcome implications in the wording. As regards
Item 9 (which we excluded) further testing with a different or randomized item order may
prove worthwhile after substituting “acquire” with “access.”
To achieve greater precision than our type of measure can realistically achieve
probably requires the development of highly context-specific operationalizations based on
deep, qualitative knowledge of that context. An example is Cliff et al.’s (2006) study of
organizational innovation among law firms. They developed a 15-item instrument Entrepreneurial
capturing deviation from established practices in that particular industry, which they bricolage
subsequently combined into a single score. It is conceivable that equally manifest and behavior
precise measures of entrepreneurial bricolage could be developed, provided that narrow
enough contexts are chosen. Such studies could also offer further (convergent) validation
of our measure, if included alongside the customized, context-specific alternative. When it
is proven that our general, “shorthand” measure correlates highly with several customized 131
measures of that kind, we can truly claim it has achieved a fully satisfactory level of
validation. It should be remembered, though, that this high level of validation is extremely
rare or non-existent for other measures commonly applied in business, management, and
organizational research.
We recommend that any further attempts to improve this scale, or develop entirely new
alternatives, also consider what did not work particularly well in our pre-tests. This includes
the use of negative items or forced-choice between “opposite” pairs of items. Further, it is
important to explain and illustrate the meaning of “resources” before delivering the items.
Asking about bricolage use in different domains or business functions (e.g. to distinguish
between “parallel” and “selective” bricolage) or about deviations from an industry norm
may not be practical unless the study and measure are industry specific (cf. Cliff et al., 2006;
Rönkkö et al., 2013).

6. Conclusions
Capturing a complex construct like entrepreneurial bricolage in a survey measure is no
trivial task and not one in which anyone should expect complete success. Yet, developing
sufficiently valid quantitative measures is essential for theory testing. In this paper we
have presented and evaluated a measure of entrepreneurial bricolage behavior, designed
to be unidimensional and applicable across a broad range of contexts. The eight-item,
reflective Baker-Davidsson measure was developed with an emphasis on content validity
and showed good results in terms of reliability as well as discriminant and nomological
validity. This evidence was collected from several samples representing different
empirical contexts as well as some variations to the measure. We consider this a
promising start. We noted some specific ways in which this measure or future alternatives
may be improved while also cautioning against modifications that are not likely to be
successful. Despite limitations and room for improvement, we believe the Baker-
Davidsson scale as presented provides an important tool for research moving forward the
behavioral theory of entrepreneurial bricolage.

References
Bacq, S., Ofstein, L.F., Kickul, J.R. and Gundry, L.K. (2015), “Bricolage in social entrepreneurship: how
creative resource mobilization fosters greater social impact”, The International Journal of
Entrepreneurship and Innovation, Vol. 16 No. 4, pp. 283-289.
Baker, T. (2007), “Resources in play: bricolage in the toy store (y)”, Journal of Business Venturing,
Vol. 22 No. 5, pp. 694-711.
Baker, T. and Nelson, R.E. (2005), “Creating something from nothing: resource construction through
entrepreneurial bricolage”, Administrative Science Quarterly, Vol. 50 No. 3, pp. 329-366.
Baker, T., Miner, A.S. and Eesley, D.T. (2003), “Improvising firms: bricolage, account giving
and improvisational competencies in the founding process”, Research Policy, Vol. 32 No. 2,
pp. 255-276.
Banerjee, P.M. and Campbell, B.A. (2009), “Inventor bricolage and firm technology research and
development”, R&D Management, Vol. 39 No. 5, pp. 473-487.
IJEBR Barnette, J.J. (2000), “Effects of stem and Likert response option reversals on survey internal
23,1 consistency: if you feel the need, there is a better alternative to using those negatively worded
stems”, Educational and Psychological Measurement, Vol. 60 No. 3, pp. 361-370.
Bechky, B.A. and Okhuysen, G.A. (2011), “Expecting the unexpected? How SWAT officers and film
crews handle surprises”, Academy of Management Journal, Vol. 54 No. 2, pp. 239-261.
Bojica, A.M., Istanbouli, A. and Fuentes-Fuentes, M.D.M. (2014), “Bricolage and growth strategies:
132 effects on the performance of palestinian women-led firms”, Journal of Developmental
Entrepreneurship, Vol. 19 No. 4, pp. 1450023-1-1450023-23.
Brown, J.S. and Duguid, P. (1991), “Organizational learning and communities-of-practice: toward a
unified view of working, learning, and innovation”, Organization Science, Vol. 2 No. 1, pp. 40-57.
Brown, T.E., Davidsson, P. and Wiklund, J. (2001), “An operationalization of Stevenson’s
conceptualization of entrepreneurship as opportunity‐based firm behavior”, Strategic
Management Journal, Vol. 22 No. 10, pp. 953-968.
Campbell, D.T. and Fiske, D.W. (1959), “Convergent and discriminant validation by the multitrait-
multimethod matrix”, Psychological Bulletin, Vol. 56 No. 2, pp. 81-105.
Chau, P.Y. (1997), “Reexamining a model for evaluating information center success using a structural
equation modeling approach”, Decision Sciences, Vol. 28 No. 2, pp. 309-334.
Cliff, J.E., Jennings, P.D. and Greenwood, R. (2006), “New to the game and questioning the rules: the
experiences and beliefs of founders who start imitative versus innovative firms”, Journal of
Business Venturing, Vol. 21 No. 5, pp. 633-663.
Coltman, T., Devinney, T.M., Midgley, D.F. and Venaik, S. (2008), “Formative versus reflective
measurement models: two applications of formative measurement”, Journal of Business
Research, Vol. 61 No. 12, pp. 1250-1262.
Covin, J.G. and Wales, W.J. (2012), “The measurement of entrepreneurial orientation”, Entrepreneurship
Theory and Practice, Vol. 36 No. 4, pp. 677-702.
Cronbach, L., Gleser, G., Nanda, H. and Rajaratnam, N. (1972), The Dependability of Behavioral
Measurements. Theory of Generalizability for Scores and Profiles, Wiley, New York, NY.
Dalton, D. and Ortegren, M. (2011), “Gender differences in ethics research: the importance of controlling
for the social desirability response bias”, Journal of Business Ethics, Vol. 103 No. 1, pp. 73-93.
Danneels, E. (2015), “Survey measures of first‐and second‐order competences”, Strategic Management
Journal, Vol. 37 No. 10, pp. 2174-2188.
Davidsson, P. (2016), Researching Entrepreneurship: Conceptualization and Design, 2nd ed., Springer,
New York, NY.
Davidsson, P. and Klofsten, M. (2003), “The business platform: developing an instrument to gauge and
assist the development of young firms”, Journal of Small Business Management, Vol. 41 No. 1,
pp. 1-26.
Davidsson, P. and Steffens, P. (2011), “Comprehensive Australian study of entrepreneurial emergence
(CAUSEE): project presentation and early results”, in Reynolds, P.D. and Curtin, R.T. (Eds),
New Business Creation, Springer, New York, NY, pp. 27-51.
Davidsson, P., Steffens, P. and Gordon, S. (2011), “Comprehensive Australian study of entrepreneurial
emergence (CAUSEE): design, data collection and descriptive results”, in Hindle, K. and Klyver, K.
(Eds), Handbook of Research on New Venture Creation, Elgar, Cheltenham and Northampton, MA,
pp. 216-250.
Davis, W.D., Dibrell, C., Craig, J.B. and Green, J. (2013), “The effects of goal orientation and client
feedback on the adaptive behaviors of family enterprise advisors”, Family Business Review,
Vol. 26 No. 3, pp. 215-234.
Desa, G. (2012), “Resource mobilization in international social entrepreneurship: bricolage as a
mechanism of institutional transformation”, Entrepreneurship Theory and Practice, Vol. 36 No. 4,
pp. 727-751.
Desa, G. and Basu, S. (2013), “Optimization or bricolage? Overcoming resource constraints in global Entrepreneurial
social entrepreneurship”, Strategic Entrepreneurship Journal, Vol. 7 No. 1, pp. 26-49. bricolage
DeVellis, R.F. (2003), Scale Development: Theory and Applications, 2nd ed., Sage Publications, behavior
Newbury Park, CA.
Edwards, J.R. (2010), “The fallacy of formative measurement”, Organizational Research Methods,
Vol. 14 No. 2, pp. 370-388.
Fan, X. and Sivo, S.A. (2007), “Sensitivity of fit indices to model misspecification and model types”, 133
Multivariate Behavioral Research, Vol. 42 No. 3, pp. 509-529.
Ferneley, E. and Bell, F. (2006), “Using bricolage to integrate business and information technology
innovation in SMEs”, Technovation, Vol. 26 No. 2, pp. 232-241.
Fisher, G. (2012), “Effectuation, causation, and bricolage: a behavioral comparison of emerging
theories in entrepreneurship research”, Entrepreneurship Theory and Practice, Vol. 36 No. 5,
pp. 1019-1051.
Garud, R. and Karnøe, P. (2003), “Bricolage versus breakthrough: distributed and embedded agency in
technology entrepreneurship”, Research Policy, Vol. 32 No. 2, pp. 277-300.
George, B.A. and Marino, L. (2011), “The epistemology of entrepreneurial orientation: conceptual
formation, modeling, and operationalization”, Entrepreneurship Theory and Practice, Vol. 35
No. 5, pp. 989-1024.
Grünhagen, J., Davidsson, P., Gordon, S.R., Salunke, S., Senyard, J., Steffens, P. and Stuetzer, M. (2016),
Comprehensive Australian Study of Entrepreneurial Emergence (CAUSEE). Handbook & User
Manual, Version 7, Australian Centre for Entrepreneurship Research (ACE) at QUT Business
School, Brisbane.
Gundry, L.K., Kickul, J.R., Griffiths, M.D. and Bacq, S.C. (2011), “Creating social change out of nothing:
the role of entrepreneurial bricolage in social entrepreneurs’ catalytic innovations”, Advances in
Entrepreneurship, Firm Emergence and Growth, Vol. 13 No. 7, pp. 1-24.
Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E. and Tatham, R.L. (1995), Multivariate Data Analysis,
4th ed., Pearson, Upper Saddle River, NJ.
Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E. and Tatham, R. (2010), Multivariate Data Analysis,
7th ed., Pearson, Upper Saddle River, NJ.
Harper, D. (1987), Working Knowledge: Skill and Community in a Small Shop, University of Chicago
Press, Chicago, IL.
Haynes, S.N., Richard, D. and Kubany, E.S. (1995), “Content validity in psychological assessment:
a functional approach to concepts and methods”, Psychological Assessment, Vol. 7 No. 3,
pp. 238-247.
Hebert, J.R., Ma, Y., Clemow, L., Ockene, I.S., Saperia, G., Stanek, E.J., Merriam, P.A. and Ockene, J.K.
(1997), “Gender differences in social desirability and social approval bias in dietary self-report”,
American Journal of Epidemiology, Vol. 146 No. 12, pp. 1046-1055.
Hmieleski, K.M. and Corbett, A.C. (2006), “Proclivity for improvisation as a predictor of entrepreneurial
intentions”, Journal of Small Business Management, Vol. 44 No. 1, pp. 45-63.
Holgado-Tello, F.P., Chacón-Moscoso, S., Barbero-García, I. and Vila-Abad, E. (2010), “Polychoric
versus pearson correlations in exploratory and confirmatory factor analysis of ordinal
variables”, Quality & Quantity, Vol. 44 No. 1, pp. 153-166.
Hooper, D., Coughlan, J. and Mullen, M.R. (2008), “Structural equation modelling: guidelines
for determining model fit”, The Electronic Journal of Business Research Methods, Vol. 6 No. 1,
pp. 53-60, available at: www.ejbrm.com
Hu, L.T. and Bentler, P.M. (1999), “Cutoff criteria for fit indexes in covariance structure analysis:
conventional criteria versus new alternatives”, Structural Equation Modeling:
A Multidisciplinary Journal, Vol. 6 No. 1, pp. 1-55.
Jacob, F. (1977), “Evolution and tinkering”, Science, Vol. 196 No. 4295, pp. 1161-1166.
IJEBR Lehner, J. (2000), “Bricolage during implementation of strategies: effects on flexibility”, unpublished
23,1 paper presented to the American Academy of Management Conference, Toronto, 4-9 August.
Levine, T.R. (2005), “Confirmatory factor analysis and scale validation in communication research”,
Communication Research Reports, Vol. 22 No. 4, pp. 335-338.
Levi-Strauss, C. (1966), The Savage Mind, University of Chicago Press, Chicago, IL.
Leybourne, S. and Sadler-Smith, E. (2006), “The role of intuition and improvisation in project
134 management”, International Journal of Project Management, Vol. 24 No. 6, pp. 483-492.
Lord, F.M. (1965), “A strong true-score theory, with applications”, Psychometrika, Vol. 30 No. 3,
pp. 239-270.
Noest, S., Ludt, S., Klingenberg, A., Glassen, K., Heiss, F., Ose, D., Rochon, J., Bozorgmehr, K., Wensing, M.
and Szecsenyi, J. (2014), “Involving patients in detecting quality gaps in a fragmented healthcare
system: development of a questionnaire for Patients’ experiences across health care sectors
(PEACS)”, International Journal for Quality in Health Care, Vol. 26 No. 3, pp. 240-249.
Nunnally, J.C. (1978), Psychometric Theory, 2nd ed., McGraw-Hill, New York, NY.
Open Science Collaboration (2015), “Estimating the reproducibility of psychological science”, Science,
Vol. 349 No. 6251, pp. aac4716-1-aac4716-8.
Podsakoff, P.M., MacKenzie, S.B., Lee, J.Y. and Podsakoff, N.P. (2003), “Common method biases in
behavioral research: a critical review of the literature and recommended remedies”, Journal of
Applied Psychology, Vol. 88 No. 5, pp. 879-903.
Powell, E.E. and Baker, T. (2014a), “Creating slack: institutional constraints and entrepreneurial
discretion”, Frontiers of Entrepreneurship Research, Proceedings from Babson Conference,
Ontario.
Powell, E.E. and Baker, T. (2014b), “It’s what you make of it: founder identity and enacting strategic
responses to adversity”, Academy of Management Journal, Vol. 57 No. 5, pp. 1406-1433.
Rauch, A., Wiklund, J., Lumpkin, G.T. and Frese, M. (2009), “Entrepreneurial orientation and business
performance: an assessment of past research and suggestions for the future”, Entrepreneurship
Theory and Practice, Vol. 33 No. 3, pp. 761-787.
Reynolds, P.D. (2009), “Screening item effects in estimating the prevalence of nascent entrepreneurs”,
Small Business Economics, Vol. 33 No. 2, pp. 151-163.
Rönkkö, M., Peltonen, J. and Arenius, P. (2013), “Selective or parallel? Toward measuring the domains
of entrepreneurial bricolage”, in Corbett, A.C. and Katz, J.A. (Eds), Entrepreneurial
Resourcefulness: Competing with Constraints, Advances in Entrepreneurship, Firm Emergence
and Growth, Vol. 15, Emerald, pp. 43-61.
Salunke, S., Weerawardena, J. and McColl-Kennedy, J.R. (2012), “Competing through service innovation:
the role of bricolage and entrepreneurship in project-oriented firms”, Journal of Business
Research, Vol. 66 No. 8, pp. 1085-1097.
Sauro, J. and Lewis, J.R. (2011), “When designing usability questionnaires, does it hurt to be positive?”,
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems in Vancouver,
pp. 2215-2224.
Senyard, J., Baker, T., Steffens, P. and Davidsson, P. (2014), “Bricolage as a path to innovativeness for
resource‐constrained new firms”, Journal of Product Innovation Management, Vol. 31 No. 2,
pp. 211-230.
Senyard, J.M. (2015), Bricolage and Early Stage Firm Performance PhD, Queensland University of
Technology, Brisbane.
Stewart, T.J. and Frye, A.W. (2004), “Investigating the use of negatively phrased survey items in
medical education settings: common wisdom or common mistake?”, Academic Medicine, Vol. 79
No. 10, pp. S18-S20.
Stinchfield, B.T., Nelson, R.E. and Wood, M.S. (2013), “Learning from levi‐strauss’ legacy: art, craft,
engineering, bricolage, and brokerage in entrepreneurship”, Entrepreneurship Theory and
Practice, Vol. 37 No. 4, pp. 889-921.
Tabachnick, B.G. and Fidell, L.S. (2007), Experimental Designs Using ANOVA, Duxbury Press, Entrepreneurial
Belmont, CA. bricolage
Toomela, A. (2008), “Variables in psychology: a critique of quantitative psychology”, Integrative
Psychological and Behavioral Science, Vol. 42 No. 3, pp. 245-265.
behavior
Wales, W.J. (2016), “Entrepreneurial orientation: a review and synthesis of promising research
directions”, International Small Business Journal, Vol. 34 No. 1, pp. 3-15.
Webb, N.M., Shavelson, R.J. and Haertel, E.H. (2006), “Reliability coefficients and generalizability 135
theory”, Handbook of Statistics, Vol. 26 No. 4, pp. 81-124.
Wong, N., Rindfleisch, A. and Burroughs, J.E. (2003), “Do reverse-worded items confound measures in
cross-cultural consumer research? The case of the material values scale”, Journal of Consumer
Research, Vol. 30 No. 1, pp. 72-91.

Corresponding author
Julienne Marie Senyard can be contacted at: j.senyard@griffith.edu.au

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com

You might also like