panel studies of new venture creation: a review and suggestions for future research
DESCRIPTION
Per Davidsson & Scott R. Gordon. Panel Studies of New Venture Creation: a Review and Suggestions for Future Research. Paul D. Reynolds. Recipient of the 2004 “Global Award for Entrepreneurship Research”. Panel Studies of Nascent Entrepreneurs/ Venture Creation: Historical roots. - PowerPoint PPT PresentationTRANSCRIPT
www.qut.edu.au
Queensland University of Technology
Panel Studies of New Venture Creation: a Review and Suggestions
for Future Research
Per Davidsson&
Scott R. Gordon
www.qut.edu.au
Queensland University of Technology
Paul D. Reynolds
Recipient of the 2004 “Global Award for Entrepreneurship Research”
www.qut.edu.au
Queensland University of Technology
Panel Studies of Nascent Entrepreneurs/ Venture Creation: Historical roots
Pre-PSEDPilots
PSED
SwPSED;Other country
sudies
e.g. Gartner (1988); Katz & Gartner, 1988;
Bhave, 1994
CAUSEEPSED 2
e.g. Sarasvathy, 2001; Shane, 2000; Baker & Nelson, 2005;
Dahlqvist, 2007
e.g. China; Switzerland
…
GEM
www.qut.edu.au
Queensland University of Technology
PSED-type Research: Basic Procedure
• Approach a very large, random sample of households or individuals
• Screen this sample for ‘nascent entrepreneurs’ = individuals currently involved in an active business start-up that is not yet an up-and-running firm
• Conduct a longer interview with those qualified as well as a randomly selected comparison group
• An important part of the contents concerns the initiation/completion and timing of a range of ‘gestation activities’
• Follow up over time (every 6-12 months over 12-72 months)
www.qut.edu.au
Queensland University of Technology
Basic Rationales• Overcome under coverage of smallest/newest
entities• Achieve international comparability• Overcome selection bias• Overcome hindsight bias/memory decay• Allow the study of process issues• Get temporal order of measurement right for causal
analysis
www.qut.edu.au
Queensland University of Technology
Studies of Nascent Entrepreneurs(hip)
= the study of individuals who are currently involved in a business start-up. They may or may not have started other firms before (‘nascent’ is not equal to ‘novice’)
= the study of on-going start-up efforts during their creation. They may or may not become up and running firms.
• Level of analysis issue• Representativeness of sample issue
www.qut.edu.au
Queensland University of Technology
Reviewed Works
• All 78 PSED-type manuscript that have been published or accepted into peer reviewed, scholarly journals
• Based on data sets from US, Canada, Netherlands, Norway and Sweden
• Plus a peep into other types of publications (not ignoring, but not striving for completeness outside of the journals)
www.qut.edu.au
Queensland University of Technology
Articles published by data set and journal outlet
Publishing period
1992 - 1997 1998 - 2001 2002 - 2005 2006 - 2010
Arti
cles
pub
lishe
d
0
10
20
30
40
50
Ann
ual a
rticl
e ra
te b
y ou
tlet
0
1
2
3
4
5
Total articlesNon-US articlesSSCI journalsSSCI entrepreneurship journals
www.qut.edu.au
Queensland University of Technology
Article Sa Gb Nc Wd Theorye Comparison FindingsCarter et al. (2003) US A 558 1 Motivation NEs vs. GPf NEs no different than GP in seeking independence, but less motivated by roles and
recognition. Male NEs likely cite innovation and financial success as career reasons.
Cassar & Craig (2009) US A 198 4 Cognitive biases NE subgroups NEs exhibit hindsight bias. Unsuccessful NEs correct their previous overestimation of
their chances of success. More highly educated exhibit less overconfidence.
Cassar (2006) US A 500 1 Human capital NE subgroups Individuals with higher levels of financial and human capital harbor greater growth intentions.
Cassar (2007) US E 170 4 Motivation NEs vs. GP Independence and self-realization are important motivations for entry. Financial motivations determine growth preferences, most strongly so for women.
Chandler et al. (2005) SE E 408 5 Social capital Team composition
over timeLarge NE teams are more likely than small NE teams to take on additional members, but not more likely to drop members.
Davidsson & Honig (2003) SE E 380 4 Social and
human capital NEs vs. GP Bridging and bonding social capital are robust predictors for engagement in NE-ship; as is general human capital and prior start-up experience.
Delmar et al. (2000) SE A 1113 1 Atheoretical NEs vs. GP Higher levels of HC are associated with NE status. Males, recent immigrants, and
those between 25-34 years, and those with role models are more likely to engage.
Diochon et al. (2005a) CA E 119 3 Eclectic Gender No gender or general human capital differences in likelihood that NEs would
abandon their venture.
Diochon et al. (2007) CA E 91 5 Attribution
theory NE subgroups NEs predominantly offer internal-stable attributions in describing positive situations.
Diochon et al. (2008) CA E 91 5 Human capital NE subgroups
Neither NEs with prior start-up experience, nor those with business education had discernibly more financial management capability than novices/no business education.
Gatewood et al. (1995) US* E 85 2 Attribution
theory GenderMale-female differences in how NEs attribute entry. Males are more likely cite external factors; females tend to be driven by internal factors, while both have stable cognitions.
Johnson et al. (2008) US E 1114 2 Cognitive style NEs vs. GP NEs are more likely to have an ‘innovator’ cognitive style than the GP. This cognitive
style is also associated with optimism as regards to sales.
Kim et al. (2006) US A 1050 1 Forms of capital NEs vs. GP Neither financial nor cultural capital discriminates NEs from the GP on entry. Human
capital is a significant advantage, especially education or managerial experience.
Liao & Welsch (2003) US A 462 1 Social capital High tech NEs vs.
othersStructural, relational and cognitive social capital leads to growth intentions. Structural social capitals' effect on aspiration is less for technology based NEs.
Liao & Welsch (2005) US A 544 1 Social capital
NEs vs. GP, high tech NEs vs. others
No quantitative differences in various dimensions of SC between NE and the GP. However patterns of association between social, relational and cognitive SC do differentiate NEs. Technology based NEs have higher relational capital.
Summary of findings on the characteristics of nascent entrepreneurs.
www.qut.edu.au
Queensland University of Technology
Antecedents and characteristics of the new venture creation process
ProcessB [9]
Explaining new venture creation process
outcomes
OutcomesC [13]
Characteristics of nascent entrepreneurs
PersonA [19]
Person & processD [7]
Person & outcomes
E [11]
Process & outcomes
F [10]
G [2]
www.qut.edu.au
Queensland University of Technology
Articles published by research topic and complexity
Publishing period
1992 - 1997 1998 - 2001 2002 - 2005 2006 - 2009
Arti
cles
pub
lishe
d
0
2
4
6
8
10
12
14
16
18
20
Ann
ual a
rticl
e ra
te b
y co
mpl
exity
0
1
2
3
4
5
6
PeopleProcessOutcomesSingle topicMultiple topics
www.qut.edu.au
Queensland University of Technology
FINDINGS REGARDING CHARACTERISTICS OF NASCENT ENTREPRENEURS (I)
• As regards resource endowments, results largely confirm prior research:– NEs have more Human Capital (S-E exp; Education)– Social capital more varied results. Close role model
no longer important in the US?– Household wealth; income has no or little effect (but
NEs with more money have higher aspirations) – Methods issue lurking!
www.qut.edu.au
Queensland University of Technology
FINDINGS REGARDING CHARACTERISTICS OF NASCENT ENTREPRENEURS (II)
• As regards motivation and cognition, the research stream has produced novel and surprising findings:– Little or unexpected difference from others in “career
motivation”– Strong evidence against push motivation (‘revenge’)– NEs are more risk averse than others!?– NEs have higher ‘Need for Closure’ – preference for
order and predictability!– Theory/representativeness issue lurking!
www.qut.edu.au
Queensland University of Technology
FINDINGS REGARDING CHARACTERISTICS OF NASCENT ENTREPRENEURS (III)
• The majority works in partnerships or teams• The majority of ‘teams’ is ‘romantic’• The majority of the remaining teams aren’t exactly
‘textbook’ teams, either• The functionally well-balanced team that was assembled
purely for business purposes is a rare occurrence• Theory/representativeness issues lurking!
www.qut.edu.au
Queensland University of Technology
Population prop. = 2/7 venturesSampling prop = 2/3 ventures(because 10/15 individuals
belong to teams)
www.qut.edu.au
Queensland University of Technology
Fundamental issue with investigating ‘Characteristics of Nascent Entrepreneurs’
• It’s a bit like comparing ‘holiday makers’ to other people…
• Some of whom will, of course, go on vacation next week…
• So perhaps we should be looking at ‘what does the NE experience do to the person’ rather than ‘what person characteristics make you become NE’?
www.qut.edu.au
Queensland University of Technology
Findings regarding new venture creation process (I): Discovery
• Relatively non-systematic search for opportunity, and processes triggered by a particular idea rather than by a wish to become a founder-manager seem to be relatively more common than systematic, textbook-like processes
• Importantly, however, this descriptive result does not necessarily have any prescriptive implications. Even if less common those ventures resulting from systematic search may achieve better outcomes
www.qut.edu.au
Queensland University of Technology
Findings regarding new venture creation process (I): Exploitation
• Enormous variation in duration• Enormous variation in sequence
– Liao, Welsh and Tan (2005): “firm gestation is a complex process that includes more than simple, unitary progressive paths” (2005: 15) and “a process where developmental stages are hardly identifiable” (2005: 13).
• However, there are systematic sub-group differences• Additional method challenges• Theoretical boundary challenges
www.qut.edu.au
Queensland University of TechnologyJan 1. Sep. 1May 1 Dec . 31
Sampling day
Annual prop. = 25/75Sampling prop = 50/50
www.qut.edu.au
Queensland University of TechnologyJan 1. Sep. 1May 1 Dec . 31
Sampling day
www.qut.edu.au
Queensland University of Technology
Methods Insights• Level of analysis issues: The sampling procedure yields a
representative sample of current nascent entrepreneurs, but a biased sample of emerging ventures or of entrepreneurial teams
• There is a problem with a substantial group of ‘dilettante dreamers’ who never put their effort to an ‘acid test’
• Another group start more ambitious ventures but take longer to get up and running – and may mistakenly appear less ‘successful’– CAUSEE: Higher tech; higher ambition take longer; retailing
and brick-and-mortar-only get up quicker (but retail have lower survival once they are up and running)
www.qut.edu.au
Queensland University of Technology
Outcome
Contrast
Outcome Driver
Operational
vs.
Terminated
Operational
vs.
Still Trying
Still Trying
vs.
Terminated
Characteristic 1 ++++ ++ ++++
Characteristic 2 none – – ++++
Characteristic 3 none + – – –
Characteristic 4 ++++ none none
Characteristic 5 none none ++++
Characteristic 6 none none none
CAUSEE Digression: Analysis Example
www.qut.edu.au
Queensland University of Technology
Methods insight:
In this type of research it is important to– Control for type of venture– Analyse different types of outcomes at different
points in time
Otherwise all sorts of confounding effects will blur our understanding
www.qut.edu.au
Queensland University of Technology
FINDINGS REGARDING DRIVERS OF OUTCOMES(I)
• Some 33-50% of nascent ventures ever reach an operational stage (self-perceived / regular sales)
• HC influences appear weak or inconsistent• Effects of financial endowment variables and social
capital indicators are also unimpressive…• And a cursory look at the results for motivation and
expectation isn’t very enlightening, either• …but there’s more to be said…• …and measures of actual investment of capital
(HC,FC,SC) tend to come out with positive effects
www.qut.edu.au
Queensland University of Technology
Why the weak overall results?• It may be overly simplistic to assume that effects on outcomes
are direct, linear and generalizable across all types of ventures, founders, and environments
• The venture and the individual are distinct levels of analysis. The respondents may not invest all their capital in the focal venture, and the venture may draw on capital from multiple individuals
• The opportunity cost structure needs to be considered when assessing the effects of human capital on outcomes
• Outcomes for independent ventures are hard to assess, predict and interpret – especially when the respondents try to start very different ventures
www.qut.edu.au
Queensland University of Technology
Human capital indicator
Study and effecta
TotalUSb Canadac Netherlandsd Norwayd Swedene
+ 0 - + 0 - + 0 - + 0 - + 0 - + 0 -
General human capitalf
Education 4 27 0 0 1 0 0 6 0 0 1 0 1 9 0 5 44 0Work experience 0 18 1 0 1 0 0 4 0 n/a n/a n/a 0 6 0 0 26 1Team vs. Solo/Team size
0 17 0 n/a n/a n/a 0 7 0 n/a n/a n/a 3 10 3 3 34 3
(Age) 6 23 0 n/a n/a n/a 0 3 1 1 0 0 1 13 1 8 37 2(Gender - female) 0 34 0 2 2 0 0 4 0 0 3 0 1 5 0 3 42 0(Ethnicity or minority status)
0 5 2 n/a n/a n/a n/a n/a n/a 0 1 0 n/a n/a n/a 0 6 2
Specific human capital
Management experience
1 12 3 n/a n/a n/a 1 3 0 n/a n/a n/a 1 5 0 3 22 3
Industry experience 5 11 0 0 1 0 0 4 0 n/a n/a n/a 4 19 0 9 35 0
Start-up experience 4 26 4 0 2 0 1 5 0 4 2 1 15 13 2 24 46 7
Business or start-up classes
1 9 0 0 2 0 n/a n/a n/a n/a n/a n/a 2 4 0 3 15 0
Other n/a n/a n/a 2 0 0 n/a n/a n/a n/a n/a n/a n/a n/a n/a 2 0 0
Summary of findings on outcome effects of human capital by countrya) + denotes a sig. positive effect; 0 denotes no significant effect; - denotes a sig. negative effect (p<.05); b) based on Brush et al. (2008a); Dimov (2009); Edelman et al. (2008); Liao and Gartner (2006); Liao et al. (forthcoming); Lichtenstein et al. (2007); Newbert (2005); Parker (forthcoming); Parker & Belghitar (2006); Tornikoski and Newbert (2007); Townsend et al. (in press); c) based on Diochon et al. (2005a); Menzies et al. (2006); d) based on van Gelderen et al. (2005); e) based on Alsos and Kolvereid (1998); Alsos and Ljunggren (1998); Rotefoss and Kolvereid (2005); e) based on Davidsson and Honig (2003); Delmar and Shane (2003); Delmar and Shane (2004);
Honig and Karlsson (2004); Eckhardt et al. (2006); Samuelsson and Davidsson (2009); Shane and Delmar (2004); f) We have reluctantly (hence the parentheses) followed the practice of including age, gender and ethnicity among the indicators of general HC.
Effect of Human Capital on Outcomes (operational-still trying-terminated)
www.qut.edu.au
Queensland University of Technology
Methods Insight
• We should stop looking for direct, linear and very broadly generalizable effects with respect to a single outcome variable
• We need to model contingent relationships and assess multiple outcome variables over time in order to disentangle the intricacies and arrive at sounder interpretations
www.qut.edu.au
Queensland University of Technology
www.qut.edu.au
Queensland University of Technology
www.qut.edu.au
Queensland University of Technology
Outcome measure and effecta
TotalMaking Progress
Continuation
vs. non-continuation
Reachingfirst sales
Operational vs. terminated
Operational vs. any other
statusReaching
profitability+ 0 - + 0 + + 0 - + 0 - + 0 - + 0 - + 0 -
- Completed business plan of any form
8 6 0 3 3 0 0 4 0 2 3 1 0 3 0 0 1 0 13 20 1
- Composite measure of extent of planning
2 0 0 1 1 0 n/a n/a n/a n/a n/a n/a 0 4 0 n/a n/a n/a 3 5 0
- Sequence measure of early planning
n/a n/a n/a 6 2 0 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a 6 2 0
RESULTS ON BUSINESS PLANNING AND OUTCOMES
www.qut.edu.au
Queensland University of Technology
FUTURE RESEARCH…Issue / Category
Users of existing data sets may want to consider…
Designers of new data sets may in addition want to consider…
Statistically representative and/or theoretically relevant sampling
Excluding ‘dilettante dreamers’
Ways to deal with ‘modest venture’ dominance
Testing for non-response and attrition biases
Correcting through weighting the cases (individual and team levels)
Correcting through weighting down or eliminating cases (already) of long duration (venture level)
Applying methods for obtaining high initial and continued response rates
Dealing with decline of landline phones
Obtaining a larger sample to allow more sub-sample analysis
Obtaining a narrower, more homogenous and/or higher-potential sample for stronger theory testing
Including additional categories (e.g., social enterprise)
www.qut.edu.au
Queensland University of Technology
FUTURE RESEARCH…Issue / Category
Users of existing data sets may want to consider…
Designers of new data sets may in addition want to consider…
Level of analysis issues
Consistently applying an explicit level of analysis from sample restriction through use of explanatory, control and outcome variables
Applying the hitherto under-utilized team level
Explicitly modeling influences or effects on different levels
Designing the entire project with (a) specific level(s) in mind
Basing the design on the realization that the venture may draw on resources from more than one individual and, conversely, that the founder(s) may invest their resources in endeavors other than the focal venture
Including more other-than-venture level outcome indicators
www.qut.edu.au
Queensland University of Technology
FUTURE RESEARCH…Issue / Category
Users of existing data sets may want to consider…
Designers of new data sets may in addition want to consider…
Dealing with process heterogeneity
Controlling for initial stage of development
Re-organize the data set based on project time line
Checking that assumed outcome ‘milestones’ really can be regarded process outcomes (i.e., occur late)
Applying higher level of abstraction to gestation activity patterns
Double-checking data on activities pre-dating the first interview
Further refining conceptualization and operationalization of gestation activities
www.qut.edu.au
Queensland University of Technology
FUTURE RESEARCH…Issue / Category
Users of existing data sets may want to consider…
Designers of new data sets may in addition want to consider…
Dealing with other heterogeneity issues
Using well thought through control variables
Applying sub-sample analysis
Modeling contingent effects (e.g., moderation; mediation)
Obtaining a larger sample to allow more sub-sample analysis (cf. above)
Obtaining a narrower, more homogenous and/or higher-potential sample for stronger theory testing (cf. above)
Including controls/moderators/mediators not available in existing data sets
www.qut.edu.au
Queensland University of Technology
FUTURE RESEARCH…Issue / Category
Users of existing data sets may want to consider…
Designers of new data sets may in addition want to consider…
Choice and interpretation of dependent variables
Using several, carefully selected outcome indicators
Refraining from unwarranted ‘success’ or ‘failure’ labeling of outcome indicators
Distinguishing among indicators of engagement, persistence, and success, respectively
Paying attention to heterogeneity in process duration
Including improved measures of engagement, persistence, and success (and duration)
Including more other-than-venture level outcome indicators (cf. above)
Probing further into reasons for and losses associated with termination (individual and venture level)
www.qut.edu.au
Queensland University of Technology
Panel Studies of New Venture Creation
• Are a Great Innovation!• Are Practically Feasible!• Entail Quite a Number of Conceptual and Methods
Related Challenges!
Thank You!