A Little Bit of Knowledge Is a Dangerous Thing

2352-6734/ © 2016 Elsevier Inc. All rights reserved. Available online 18 October 2016 cross small, cross-sectional samples. Ten studies showed insignificant effects between experience and the likelihood of disengagement, seven showed a negative relationship, two had mixed findings, and three revealed a positive relationship. None of the studies presented or reported findings of curvilinear results. As such, we propose a framework that revisits the fundamental assumption about experience and its influence on entrepreneurial trajectories. We turned to research demonstrating the non-linear influence of experience. A closer examination of cognition, strategic management, and entrepreneurship studies reveals that novices, the moderately experienced, and experts all leverage their experience differently in their pursuits (Cormier and Hagman, 1987; Haleblian and Finkelstein, 1999; Toft-Kehler et al., 2014). This research reveals that “a little bit of experience can be a dangerous thing” – at low levels, actors inappropriately apply experience to seemingly similar, yet inherently different, tasks. We argue that the differential effects of experience may also produce a non-linear relationship on the decision to disengage. 3. Methods and data To answer our research question we created a longitudinal dataset with information about founders and their ventures. We constructed our sample using two databases maintained by Statistics Sweden: RAMS (yearly data on all firms) and LISA (yearly data on all Swedish inhabitants from 1989). From RAMS we sampled three full cohorts of firms started 1994, 1995, and 1996, followed until 2002. From LISA we created experience variables for all prior venturing activities from 1989 to 1993 and used National Tax Board data to gather financial information. To decrease industry heterogeneity, we limited our sample to firms in knowledgeintensive sectors based on OECD classifications (Götzfried, 2004). As a result, our sample consists of the full population of 29,338 new knowledge-intensive ventures founded in Sweden between 1994–1996. 3.1. Dependent variable The dependent variable, likelihood of disengagement, is based on a yearly indicator of whether an individual is still working fulltime in their venture (1= disengaged, 0=working full-time). The founder disengages when they begin work on or at another firm. We highlight two other decisions related to this variable: We retained bankrupt firms in our sample, since this is a viable pathway for disengagement, and bankruptcy and liquidation are rare (Thorburn, 2000). However, we dropped firms that experienced a trade sale which is generally considered an exit-outcome which does not reflect disengagement due to unmet expectations (Arora and Nandkumar, 2011). 3.2. Independent variable Our independent variable is entrepreneurial experience, defined as years of full-time involvement as a founder or co-founder in a prior venture. To reduce the effect of individuals arbitrarily entering and exiting (e.g. tax speculation or engaging in “portfolio entrepreneurship” (Westhead and Wright, 1998)), we required a two-year gap between ventures in the same industry and location to be considered a separate venturing activity. 3.3. Control variables To address alternative explanations, we included several control variables. We control for founders’ basic demographic characteristics: gender (1=male), age, and number of children living at home (updated annually). To account for entrepreneurs’ ability to support the firm, we included household wealth based on equity reported to tax authorities, which tracks wealth excess of 800,000 SEK (~100,000 USD). Firms in our data are either incorporated (limited liability) or unincorporated (partnerships and proprietorships with unlimited liability). Since financial liability may affect the likelihood of disengagement, we controlled for legal form (1= incorporations). To account for additional investments in underperforming firms, we included a new investments variable based on yearly equity injections (Wennberg et al., 2010). We controlled for entrepreneurial earnings as a measure of performance. We used firm-level performance variables from RAMS and individual-level data from LISA to calculate earnings based on Hamilton's (2000) definition [revenues – expenses = money taken out + entrepreneurial earnings] and used its natural log value to correct for skewness. Education was measured in years. We also controlled for other types of experience:Management experience was based on a “personnel responsibility” categorical variable in the 1990 census (0=no experience, 1=some experience, and 2=extensive experience). Industry experience was a count of years within the focal industry. Venture similarity experience was based on Finkelstein and Haleblian's (1999) measure of comparing industry affiliations of prior ventures. Last, we controlled for industry (at the SIC-2 level) and time-varying effects (year dummies). 1 To minimize right censoring and incorporate up to five years of data prior to the focal venture. 2 In Sweden, 35% of all new firms belong to these sectors, including information technology, chemicals, medicine, telecom, finance, business services, education and research (Folta et al., 2010). 3 We deleted 1102 trade sales (representing 1.2% of disengaged founders) from the dataset. One limitation of our sample is that we are unable to directly measure the number of bankruptcies. However, according to the Swedish census data, the recent average number of annual bankruptcies is 0.19% of all privately held firms (from 2010 to 2015). R.V. Toft-Kehler et al. Journal of Business Venturing Insights 6 (2016) 36–46