Printer Friendly

Invention quality and entrepreneurial earnings: the role of prior employment variety.

We use creativity theory to analyze the effects of occupational job variety and industry variety on invention quality, and entrepreneurial earnings. We test our ideas with survey data from 770 inventor-entrepreneurs who commercialized their own inventions. Results suggest that occupational and industry variety substitute for each other in positively affecting invention quality whereas a lack of industry variety is associated with greater entrepreneurial earnings. Results are consistent with the idea that high levels of both occupational and industry variety enables the generation and discovery of inventions, but these ideas are usually not technically feasible or financially viable.

Introduction

A central goal of entrepreneurship research is to identify the factors that predict entrepreneurial success (Shane & Venkataraman, 2000), particularly with respect to identifying new opportunities and successfully exploiting them for commercial profit (e.g., Arentz, Sautet, & Storr, 2013; Corbett, 2005). Much research has been dedicated to identifying a range of dispositional factors including risk-taking propensity and cognitive abilities (for a review, see Astebro, 2012). Alternatively, scholars have also sought to identify nondispositional factors, with a particular interest in human capital (for a review see Unger, Rauch, Frese, & Rosenvusch, 2011). The core idea of human capital theory is that investments in human capital (e.g., education or job skills) increase net lifetime earnings because of their value in most types of employment (Becker, 1964). In addition, Lazear (2005) posits that entrepreneurial success is contingent on acquiring all the relevant and diverse business skills required to start and run a new venture. However, while many conclude that human capital is essential to entrepreneurial success (e.g., Briiderl, Preisendorfer, & Ziegler, 1992; Cooper, Gimeno-Gascon, & Woo, 1994), some remain skeptical of its value for entrepreneurship (e.g., Baum & Silverman, 2004; Haber & Reichel, 2007).

To resolve this debate, Unger et al. (2011) conducted a meta-analytic review of 30 years of human capital research and found only a small positive effect of human capital on entrepreneurial success. They offer the explanation that human capital effects on entrepreneurial success depend on how relevant human capital is to the specific entrepreneurial tasks required for success. For example, Davidsson and Honig (2003) find human capital to be more strongly related to the task of starting a new business than the task of completing the first sale or turning a profit. Consequently, Unger et al. suggest the importance of addressing how human capital reflects learning from prior experiences to effectively execute the variety of entrepreneurial tasks required for success.

Considered together, human capital research on entrepreneurship provides two suggestions: entrepreneurship may require a variety of business skills (Lazear, 2005), and some business skills are more important for certain entrepreneurial activities whereas other business skills are more important for other entrepreneurial activities (Davidsson & Honig, 2003). In this article, we suggest that unpacking human capital into its component parts--be it knowledge and skills acquired through prior occupational experiences or prior industry experiences--and separating out the effects of these component parts on specific entrepreneurial tasks should provide new insights into the link between human capital and entrepreneurial success.

In reviewing the aforementioned research, we posit that distinguishing between occupational variety and industry variety (cf. Astebro & Thompson, 2011) and between two dependent variables that capture entrepreneurial success--i.e., invention quality and entrepreneurial earnings--makes a useful contribution to furthering our understanding of the role of human capital on entrepreneurial success. In contrast to Astebro and Thompson, we argue that knowledge and business skills associated with an occupation are different from the knowledge and business skills associated with an industry. In doing so, we use creativity theory to analyze the specific and joint effects of occupational variety and industry variety on invention quality and entrepreneurial earnings. We link both types of employment variety to invention quality in terms of creativity (cf. Fiet, 2002; Shane, 2000) and to entrepreneurial earnings in terms of acquiring relevant knowledge and business skills (cf. Astebro & Thompson; Astebro, Chen, & Thompson, 2011). (1) Specifically, given that research has consistently linked prior experience variety (of any sort) to creative quality and established a negative relationship between industry variety and entrepreneurial earnings (Astebro & Thompson), we expect that both occupational and industry variety will be related to invention quality, while occupational variety coupled with a lack of industry variety is most likely to maximize invention quality and entrepreneurial earnings.

To test these ideas, we examine the joint effects of occupational and industry variety on invention quality and entrepreneurial earnings. Our data set consists of a sample of 770 Canadian independent inventor-entrepreneurs (i.e., those that commercialize their inventions outside the confines of established organizations), examined before by Astebro and Thompson (2011) in terms of effects on household income. This sample is unique in that it allows us to test effects on entrepreneurial success comprising both invention quality and entrepreneurial earnings because the inventors personally face both challenges of invention and commercialization; other samples typically focus only on one of the two. The data set provides information on the extent to which inventors' prior employment experiences have spanned different occupations and industries, and records the creative quality of an invention, whether the invention reached the market, and the present value of future entrepreneurial earnings. Our empirical specification includes the interaction between occupational and industry variety--missing from existing research--as well as two new dependent variables: invention quality and the entrepreneurial earnings from commercially exploiting these inventions.

Our empirical results suggest that the effect of prior employment should be decomposed into occupational and industry variety for better understanding of its effects. We observe both occupational and industry variety to have positive relationships with invention quality, and to act as substitutes for each other in relation to the invention's creative quality. Furthermore, industry variety has a negative relationship with entrepreneurial earnings while occupational variety has no relationship. Thus, the entrepreneur who has worked in many different occupations within the same industry seems to stand the best chance of achieving entrepreneurial success.

Theory

We begin by acknowledging that prior experience comprises both employment-relevant job variety and personal-background diversity. (2) Employment variety refers to differences across individuals based on employment-related attributes such as profession, function, education, knowledge, skills, or expertise. There is evidence suggesting that most individuals identify new entrepreneurial opportunities from prior employment experiences (Bhide, 1994). Evidence also suggests that entrepreneurial success is determined by prior employment experiences (e.g., Briiderl et al., 1992; Gimeno, Folta, Cooper, & Woo, 1997). Thus, a focus on how the variety of individuals' prior employment experiences affects entrepreneurial success is warranted.

We focus on prior employment experiences by distinguishing between occupational and industry variety because they reflect differences in human capital and prior knowledge of new technologies, markets, ways to serve markets, and customer problems (Marvel & Lumpkin, 2007; Shane, 2000). Occupational variety is the number of different occupational fields of experience that an individual has worked in before. For example, at a general level, accounting, finance, and marketing represent different occupational fields. Industry variety is the number of different distinct industries that an individual has worked in before; for example, being an accountant in the pharmaceutical, computer software, and consumer electronics industries. We consider how these two types of prior experiences might independently and jointly affect entrepreneurial success with regard to two aspects of entrepreneurial success: invention quality and the entrepreneurial earnings from exploiting these inventions.

Prior Employment Variety and Invention Quality

Creative quality is an indication of invention quality (Fiet, 2002; Shane, 2000). Creative quality is defined as the extent to which an invention--the implemented idea borne of an identified opportunity--is both new and useful (Amabile, 1996; Ward, 2004). For instance, the United States Patent and Trademark Office grants patents to those who "invent or discover any new and useful process, machine, article of manufacture, or composition of matter, or any new and use improvement thereof' (U.S. Patent and Trademark Office, 2015). Similarly, the Canadian Intellectual Property Office grants patent protection to the first applicant who presents an invention that is "new--first in the world, useful--functional and operative, and inventive--showing ingenuity and not obvious to someone of average skill in the field of invention" (Canadian Intellectual Property Office, 2015). Newness represents the extent to which an invention is novel or different when compared to existing ideas and solutions. Usefulness represents the extent to which the invention is an appropriate response to a problem. Thus, creative quality requires that new inventions are useful in that they offer a better solution than existing solutions.

Given that prior employment experience characterized by variety is often--but not always--associated with creative quality (e.g., Taylor & Greve, 2006), we suggest distinguishing between occupational and industry variety to consider how their joint effects might be associated with invention quality. To this end, we consider four illustrative combinations: high occupational variety coupled with high industry variety, high occupational variety coupled with low industry variety, low occupational variety coupled with high industry variety, and low occupational variety coupled with low industry variety.

Since research has shown that those with a variety of employment experiences are capable of taking different perspectives to a problem and have a varied knowledge base to work with to discover new ideas (e.g., Amabile, 1996; Fleming, Mingo, & Chen, 2007), one might expect that both types of variety will interact in such a way that the positive association each type of variety has on invention quality will increase with increasing levels of the other type of variety. Switching between many occupations and industries requires learning a great variety of new and different knowledge, techniques, skill, and expertise (March, 1991). This endows a broad knowledge stock that enables a broader search for new ways to combine or reconfigure existing knowledge to create new inventions (Greve & Taylor, 2000). Specifically, occupational variety enables a search for ways to combine or reconfigure knowledge acquired from different occupational experiences whereas industry variety enables a search for ways to combine or reconfigure knowledge acquired from different industry experiences. Thus, one possibility is that combining ideas from various occupational skills presents the greatest chance of creating a new industrial application invention the more varied the range of industrial applications an individual has worked with. Alternatively, combining ideas from many different industries presents the greatest chance of creating a new industrial application invention if an individual knows all occupational functions of business. Either way, one might expect that the combination of high occupational variety and high industry variety will be associated with the highest invention quality.

However, the high occupational variety and high industry variety combination may be too much variety for invention quality because "too-diverse knowledge can result in unwieldy and impractical outputs... creativity requires the application of deep knowledge because individuals must understand a knowledge domain to push its boundaries with any nontrivial likelihood of success" (Taylor & Greve, 2006, pp. 723-724). In other words, while high levels of both occupational and industry variety certainly enable the generation and discovery of a great variety of new invention ideas, these ideas are usually not technically feasible or financially viable to implement. Indeed, invention quality requires both divergent and convergent thinking (Cropley, 2006; Guilford, 1950), where divergent thinking involves generating different possible ideas and convergent thinking involves analyzing and combining different ideas to find the best idea (Guilford, 1950). While variety is critical to divergent thinking, variety without knowledge depth hampers convergent thinking because knowledge depth involves knowing the relevant criteria of novelty and effectiveness and knowing how to turn novelty into effective solutions (Cropley). Thus, prior employment variety must produce the appropriate balance of knowledge breadth and depth to foster the divergent and convergent thinking necessary for invention.

The high occupational variety and high industry variety combination may also be too much variety in that this combination reflects an overinvestment in a varied employment experience. For instance, many pursue a varied employment experience because they have a taste for variety instead of optimizing for a future entrepreneurial career (Astebro & Thompson, 2011). Individuals motivated by a taste for variety change occupations and industries to satisfy this taste rather than to learn the knowledge and skills relevant to developing a new invention. These individuals do not discriminate between the value of occupational variety and the value of industry variety; either set of variety satisfies a taste for variety, and the investment level may therefore be too high from a learning perspective.

Taken together, our discussion suggests that prior employment variety will be associated with invention quality to the extent that the combination of occupational variety and industry variety maximizes divergent thinking, minimizes convergent thinking, and optimizes learning. Conversely, the combination of low occupational variety and industry variety is likely to have the opposite outcomes: while likely to be associated with the development of convergent thinking and optimized learning, divergent thinking is nonexistent. This leaves us with the remaining two possible combinations of high occupational variety coupled with low industry variety and low occupational variety coupled with high industry variety. The challenge is to identify which of the two combinations of occupation and industry variety is more likely to be associated with the appropriate development of divergent thinking, convergent thinking, and learning.

Following the same logic as before, high occupational variety coupled with low industry variety should enable an individual to draw from their varied occupational experiences to think divergently when generating a variety of new ideas while the lack of varied industry experiences should be associated with the development of in-depth industry knowledge that enables convergent thinking when selecting, developing, and implementing ideas into new inventions for a given industry. In contrast, low occupational variety coupled with high industry variety should enable an individual to draw from their varied industry experiences to think divergently but industry variety is also associated with a limited convergent thinking ability when selecting, developing, and implementing ideas into new inventions. Thus, it seems that prior employment experiences characterized by the combination of high occupational variety with a lack of industry variety may be associated with the highest quality inventions.

Prior Employment Variety and Entrepreneurial Earnings

Given that occupational variety combined with a lack of industry variety could be associated with the highest quality inventions relative to the other combinations of prior employment variety, the natural follow-up question is whether this combination might also be with the highest entrepreneurial earnings. We believe this will be so for two reasons.

First, the jack-of-all-trades theory posits that an individual must be sufficiently proficient in all of the functions of a business to effectively run it for economic profit (Lazear, 2005). Building on this theory, we reason that to the extent that shifting occupations provides cross-functional training across entrepreneurial-specific and managerial skills, an increase in occupational variety should be positively associated with entrepreneurial earnings. And while an individual can certainly outsource a certain task for which they may lack skills, successful outsourcing depends on the individual still knowing enough to judge the quality of applicants and suppliers across all functional tasks (Lazear).

Second, empirical evidence has consistently shown that extensive prior experience in the same industry--i.e., a lack of industry variety--is associated with greater entrepreneurial earnings (e.g., Briiderl et al., 1992; Ganotakis, 2012; Gimeno et al., 1997; Pennings, Lee, & van Witteloostuijn, 1998; Shane, 2000). For example, the literature on start-up performance shows that prior industry variety of the founder(s) significantly decreases the performance of the start-up (Briiderl et al.; Ganotakis; Gimeno et al.; Pennings et al.). In addition, the literature on industry dynamics has shown that withinindustry spinoffs have higher performance than diversifying start-ups; new firms with founders with in-depth experience of a particular industry perform better than those with founders with prior experience shaped by other industries (see Klepper & Thompson, 2010). This deepening-of-industry-knowledge effect occurs because a lack of industry variety conversely entails acquiring more knowledge of new technologies, markets, ways to serve markets, and customer problems for a specific industry (e.g., Marvel & Lumpkin, 2007; Shane). (3) Thus, industry variety should be negatively associated with entrepreneurial earnings.

Methods

Data and Sample

We use responses to a survey, conducted in February 2004, of 770 independent inventors self-identified through their use of the services of the Canadian Innovation Centre (CIC) sometime during the decade preceding the survey. The inventor sample began with a sampling frame consisting of 6,405 inventors that had asked the CIC to evaluate their inventions between 1994 and 2001. Of these 6,405, less than 10% were from repeat inventors. Because we wanted to have as much variance as possible in our measures of employment variety, we randomly removed all but one assessment from repeat inventors, as well as records with incomplete data and drew a random sample from the remaining frame. We continued by searching the yellow pages for current addresses and traced 1,770. These were contacted by surface mail stating we would be calling. We were then able to contact 934 by telephone, and from these we completed 830 telephone interviews of which 770 had complete information on the dependent variables referring to events up to 2003 (see the online Appendix available at http://papers.ssm.com/sol3/cf_dev/AbsByAuth.cfm?per_id=263222). The adjusted response rate was 61%.

To understand the composition of the inventor sample better, we further drew a comparison sample from the general Canadian population. Using random digit dialing we queried a sample of 300 Canadians from the general population based on sampling quotas for province, employment, and gender to reflect similarities in the aggregate with the inventors on these three variables. Comparisons were then made on background characteristics (see the online Appendix for details).

The combined samples from the general population matched with the inventors contains unusually high proportions reporting that they are or have been self-employed (63%), or have owned a business (60%). However, the rate of entrepreneurship is much higher for the inventor sample than for the general population sample. Inventors are also significantly more likely to come from an entrepreneurial family, and to have worked in a greater variety of occupations and industries. Fifty-four percent of the inventors have spent 6 or more years developing inventions, and 75% have worked on more than one invention. Thus, our sample contains inventors with extensive inventor and entrepreneurial experience, and prior experience comprising uncharacteristically high occupational and industry variety.

As with most surveys we expect sampling and response biases. We control for sampling bias by using a probit model of the probability of being able to trace an address. We also control for nonresponse bias by estimating the probability of response from the traceable sample. We multiply the probabilities of tracing and response and invert the product for use as selection weight in the analysis (Holt, Smith, & Winter, 1980). The results are qualitatively similar using the product of the sampling and response probabilities as when not using them. Results reported in the body of the text are without the sample selection corrections. Results with the sample selection corrections applied are available in the online Appendix.

Refusals to respond to survey questions vary by question, but are not extremely high for any particular question. Nevertheless, list-wise deletion of observations with item nonresponse will remove a large number of observations so we impute missing item data on independent variables using multiple imputations (see Little & Rubin, 1987). The number of missing values that are imputed (selected variables) vary from 53 (time to develop invention), 44 (investments), 16 (invention quality), 15 and 5 (measures of occupational and industry variety), to 5 and 2 (teamwork to invent and to commercialize). In particular, we impute missing values five times using switching regression conditional on observed data, where the imputed values are randomly drawn from the distribution of the prediction, rather than the point estimate (Van Buuren, Boshuizen, & Knook, 1999). Five complete data sets with imputed and nonimputed data are generated. Results are pooled across the five data set using equations 12.17-12.20 in Little and Rubin. This method appropriately takes into account the inherent uncertainty in the predictions of the missing values.

Dependent Variables. There are two dependent variables reflecting entrepreneurial success in this analysis: invention quality and earnings from commercially exploiting these inventions.

Invention Quality. Current thought in the creativity literature defines creative invention outcomes as novel and useful (Amabile, 1996), where usefulness in the entrepreneurial context refers to "valuable ideas for new goods and services that will appeal to some identifiable market" (Ward, 2004, p. 174). That is, invention quality is not limited to being "novel," "new," or "clever," but also potentially valuable to an identifiable customer base (Amabile; Ward). Thus, we measure "invention quality" as the commercial potential of a new invention. The measure was obtained from the records of the Inventor's Assistance Program (IAP) at the CIC in Waterloo, Canada. This program helps inventors to evaluate the potential commercial opportunity of an invention before significant R&D expenditures and commercialization efforts are made. The purpose is to advise inventorentrepreneurs on whether and how to continue efforts. IAP program evaluators assess a range of technological and economic variables related to the invention as such. (The online Appendix contains a list of all items rated for each invention.) The evaluations are based on a well-established assessment process that had been in use for over 20 years and had been applied to over 10,000 inventions (for reliability estimates, see Baker & Albaum, 1986). Because assessments occurred before commercialization, they avoid problems such as methods bias (Campbell & Fiske, 1959) and hindsight bias (Fischhoff, 1975). The IAP's evaluators were extensively trained by a chief evaluator, who ran the program consistently from 1981 to 2000, and a group meeting at the end of each review provides feedback to ensure appropriate measures for each invention. The IAP's evaluations have been found to successfully predict future revenues of commercialized inventions (Astebro, 2003). The evaluations are compiled into a report containing item scores together with a cover page containing a single indicator--the overall evaluation--of the commercial potential. This overall evaluation represents our measure of invention quality.

Entrepreneurial Earnings. This represents the sum of the present value of all future revenues stemming from commercially exploiting an invention. We ask detailed questions on the dollar value of revenues for each year of revenues, the end year of revenues, and other possible revenue streams, such as obtaining royalties and/or selling the intellectual property (IP). Only in five cases was the IP sold or licensed by the inventor to another party. Excluding these five inventions from the analysis does not change results. If future sales are truncated by survey date, we forecast remaining sales with a standard Bass diffusion model. (4) (Regression estimates were however consistent if we abstain from any type of estimate of future sales and use only the reported sales data. See the online Appendix.) Data are discounted to 2003 using the Canadian consumer price index. (Details of how we compute the discounted present value are reported in the online Appendix.)

It is possible that inventors are not able to provide accurate responses during our phone interviews. Indeed, some of these inventions were developed up to 10 years before the interview. Thus, there are likely to be measurement errors that can bias any regression estimates toward white noise, i.e., zero. Had we chosen to obtain more contemporary data we would likely reduce such noise, but on the other hand would have had to deal with a greater degree of truncation of data on entrepreneurial earnings. We chose to avoid as much as possible truncation of the dependent variable in favor of more noisy data. (Details of the method we use to deal with the right truncation of data on entrepreneurial earnings are reported in the online Appendix.)

Twenty-two percent of the inventions are rated as having high commercial potential by the IAP and given a positive recommendation for further development. The remaining 78% are deemed of low commercial potential. Furthermore, 11% of all inventions are commercialized. The average present value of revenues is $67,433. Average present value conditional on commercialization is $619,739. The two dependent variables invention quality and entrepreneurial earnings are positively correlated (r =.24, p <.01). As a reflection of this correlation, conditional on commercialization, entrepreneurial earnings are higher with a positive than negative assessment ($711,233 vs. $530,823).

Independent Variables. Occupational Variety. To assess occupational variety, we use the measure of occupational variety in Wagner (2003) verbatim, while adding the middle sentence with examples to clarify further: "In how many different occupational fields of experience have you been active? Accounting, farming, marketing, and plumbing would be examples. We are interested not in the number of specific jobs you have had but the number of past and present occupational fields of experience." Several articles examining the jack-of-all-trades hypothesis have used similar measures of the variety in occupational employment history, notably Astebro et al. (2011), Astebro and Thompson (2011), Silva (2007), and Wagner. Across these articles, the probability to become an entrepreneur invariably increases with occupational variety and the construct, thus, is quite robust. (5) In this data set, 25% of the respondents have worked in more than five occupational fields, the median being three.

Industry Variety. We also ask how many distinct industries the respondent had worked in. As reported in Table 1, there is considerable variation in the responses. Eighteen percent have worked in more than five industries, the median being three.

Each variety measure is coded as 1,2,3,4,5,6-10, more than 10 (see Table 1). Forexpositional purposes we linearize the two measures by taking the midpoint of the interval 6-10, and top-coding the upper open interval to 11. Eleven represents the lowest value of the open interval and a conservative estimate of the potential range of values for it. With these transformations we can compute the average number of industries worked in (3.90), which are less than the average number of occupations worked in (4.41), indicating that there are more individuals who shift occupations than there are individuals who shift across industries.

Notice that changing between two occupations may sometimes mean that the individual will, almost by default, change industries. For example, changing from being a lawyer to a software programmer, while being two distinct occupations, likely also implies changing industries. We are, however, interested in the marginal effect of each variable. Regressions, therefore, always contain both variety measures. Estimated coefficients, thus, measure the marginal impact of a change in one of the variety measures, holding the other constant. To give an idea of the independence of the two measures, we find that 60% of those working in one industry (16% of the sample) had worked in more than one occupation within that industry. Furthermore, 45% of those having worked in only one occupation (11% of the sample) had worked in more than one industry. There, thus, appears to be meaningful independent variation, as suggested also by the correlation between the two measures (r =.47), which is far from 1.00.

Table 1 reports few economically or statistically significant differences in inventor prior employment variety on the dependent variables. Differences may nevertheless appear once taking into account differences in other variables. We, therefore, include a range of control variables, including two forms of teamwork, whether the invention was developed as normal duties at work, inventor age, years of schooling, managerial experience, R&D investments (time and money), whether the invention was patented, and commercialization investments (time and money).

Descriptive statistics and all pairwise correlations are reported in Table 2. Some correlations of note are between invention quality and entrepreneurial earnings (r =.24), and between entrepreneurial earnings on the one hand, and R&D investments (time and money), commercialization investments (time and money), team to commercialize and invent, invention developed at work, and whether the invention was patented, on the other. The same variables are also correlated at about the same levels with invention quality. Further, working in a team to commercialize the idea is correlated with R&D investments (time and money), commercialization investments (time and money), invention developed at work, and whether the invention was patented. Finally, all four investment variables are strongly correlated with each other and with patenting.

Results

In this section, we report results of our multivariate tests. We begin with an examination of the effect of inventors' employment variety on invention quality. For this, we use a probit regression in which the dependent variable equals one if the IAP assessed the invention to have commercial potential, and zero otherwise. The regressions include controls for general human capital: age and work experience. Age and work experience are reported in categories in the survey, but for ease of presentation these are transformed into years of age and years of experience using the midpoints of the categories. Results using the ordinal categories are consistent with those presented here and are available on request. We try several additional inventor characteristics, including years of invention experience, managerial experience, business ownership experience, family business experience, education, gender, and marital status. These are statistically irrelevant and are left out of our main regressions as they did not affect results. The online Appendix reports regressions when these variables are included. We further control for field of application using a set of 10 dummy variables, (6) and for the year when the invention was developed by a set of 8 dummy variables. Estimates for these 18 dummies are suppressed in tables but available on request. We also control for details of the invention process: invention developed as part of normal duties at work, whether the idea was patented, and the amount of time and dollars spent on the idea prior to sending it to the IAP for review. As the dependent variable invention quality is measured precommercialization, we do not include any variables associated with commercialization efforts in this regression, as those are endogenous to (follows later in time to) the initial creative invention stage.

Column 1 of Table 3 contains the control variables. Among the control variables, developing the idea at work and the investment variables (time and money), are positively and significantly correlated with invention quality. The age of the inventor has a negative correlation. The effect of age is linear; a squared term added is not significant (results available on request). The main results in Column 2 indicate that both occupational and industry variety are positively correlated with invention quality. There is also a significant negative interaction between the two measures of variety. In Figure 1, we plot the joint effects of the two variety variables, holding all other variables constant at the sample mean. The figure illustrates that the largest probability of high invention quality occurs either where the number of industries is low and the number of occupational fields is high (p =.39), or where the number of industries is high and the number of occupational fields is low (p =.59). The mean probability of high invention quality is p =.22, and the minimum occurs at the opposite corners, with p =.12 (low, low), or p =.16 (high, high), respectively. The figure shows a clear substitution effect between occupation and industry variety on invention quality.

We now shift focus to the earnings from the invention. Column 3 of Table 3 reports the results of Tobit regressions of earnings on control variables. Developing the idea at work, and working with a team to commercialize the invention are both positively correlated with entrepreneurial earnings. Finally, the investments put in to commercialize the invention increases entrepreneurial earnings. Column 4 indicates that industry variety is significantly negatively correlated with entrepreneurial earnings. Conversely, occupational variety has no relation to entrepreneurial earnings. In the spirit of the test of substitutability between occupation and industry variety on invention quality, we also explore the interaction between the two variables on entrepreneurial earnings. The interaction is positive and significant. However, this interaction turns out to be picking up a nonlinearity in one of the main effects, rather than being a true interaction. This can happen when two highly correlated variables are interacted with each other, in particular when the main effect of one of them is insignificant (Cortina, 1993). After further robustness analysis (available in the online Appendix), we consider this interaction effect not to be particularly robust and we, therefore, do not put much faith in it. Excluding the interaction term, the main effect of number of industries is still negative and significant ([beta]= -1.186, p<. 01).

[FIGURE 1 OMITTED]

Our results suggest that while occupational and industry variety are both correlated with invention quality, these concepts are, together, less correlated with entrepreneurial earnings. Two joint F-tests show this rather clearly. The joint significance of occupational variety, industry variety, and their interaction in Column 2 of Table 3 is [chi square] (3) = 13.39, p <.01, while their joint significance in Column 4 of Table 3 is [chi square](3) = 3.26, p <.05. In addition, we compare the marginal effects of occupational and industry variety on the two outcome variables. Reviewing Figure 1, holding industry variety constant at 1, adding one additional occupation would typically increase the probability of obtaining high invention quality by three percentage points, for example shifting the probability from.24 to.27 when going from six to seven occupations. Instead, holding occupation variety constant at 1, adding one additional industry would typically increase the probability of high invention by four to five percentage points, for example, shifting the probability from.33 to.38 when going from six to seven industries. Conversely, the probability of obtaining any earnings is reduced by two percentage points when going from less than six industries to having worked in six or more industries. Variations across working in between one to five industries are insignificant. And as explained before, shifting occupations does not affect the probability of obtaining any earnings. It is, thus, evident that both the joint correlation between occupational and industry variety on the one hand and entrepreneurial earnings on the other, as well their marginal effects are smaller for determining earnings than for determining invention quality.

Robustness Analysis

We considered the impact of sampling and response bias. This cannot be performed under multiple imputation, so we select one sample from the five generated complete samples, and run regressions with the likelihood function weighted by the inverse of the product of the sampling and response probabilities, as recommended by Holt et al. (1980). The coefficient estimates reported in the online Appendix are consistent with those reported in Table 3.

Since forming a team is likely to be endogenous, we additionally examine this process using a two-stage approach. In the first stage, we model the probability of teamwork for commercialization as a function of predetermined variables. In the second stage, we weight the likelihood function with the inverse of the propensity to form a team estimated from the first stage. This selection-control method is described in Hirano, Imbens, and Ridder (2003). The resulting coefficient estimates reported in the online Appendix are consistent with those reported in Table 3.

As explained, we have two team-based variables which we include: whether the inventor invented together with others or not, and whether the inventor worked together with others to commercialize the invention or not. Other unmeasured team-based effects may occur on the margin for teams, which represent 21% of the sample, such as the impacts of the skills and backgrounds of team members, and the size of the team. We, therefore, explore whether results were sensitive to dropping all projects that were team-based in the commercialization phase. We then end up with 609 observations. Results for the DV = invention quality are similar as when including team-based projects (see online Appendix) while results with the DV = Log (entrepreneurial earnings) unfortunately do not converge and so the robustness of the latter specification to excluding teams is not clear.

We further explore whether results are sensitive to which type of resources the team members other than the inventor contributed. We have data on whether a team was formed to obtain "human capital" and "social capital," respectively. (7) We introduce two dummy variables. The first describes whether a partner joined the inventor to provide human capital, and the second whether a partner joined the inventor to provide social capital. This regression (see online Appendix) shows that the key teamwork effect is when the partner brings social capital. However, the results for the key independent variables regarding occupational and industry variety and their interaction do not change remarkably when including these new team control variables.

Another robustness analysis centers on whether the results are sensitive to the exclusion of some specific measures of general and specific human capital (Becker, 1964). For example, if the variety in occupational experience is positively correlated with education and education is positively correlated with entrepreneurial success but excluded, then the effect of occupational variety will be upwards biased. For general human capital, one usually measures age, education, and work experience, of which we currently include age and work experience as controls. For specific human capital, we note that there is no agreed on exact definition of specific human capital for entrepreneurship, but there is a variety of proxy measures. (8) In the online Appendix, we present regressions where we also include education and a range of additional variables to proxy for specific human capital. (9) The key issue is whether including these additional human capital proxies drastically reduces coefficient estimates for our variety measures, not the specific estimate of any single additional variable. The online Appendix reports that there are no remarkable changes to the estimates of occupational and industry variety when including these additional control variables.

Discussion

We seek to identify the association between occupational and industry variety and entrepreneurial success, which would be consistent with a theory of creativity. We show how occupation and industry variety have independent and joint correlations with invention quality and entrepreneurial earnings. Our findings contribute to understanding the effects of prior employment variety on entrepreneurial success in several ways.

First, we make a contribution by finding contrasting effects for different configurations of employment variety on entrepreneurial success. Specifically, we find both occupational and industry variety to positively relate to invention quality. We also find evidence of occupational and industry variety to be substitutes: high invention quality is most likely achieved by either high occupational variety coupled with an industry focus or high industry variety coupled with an occupational focus. That both types of employment variety can be substituted for each other is consistent with the argument that either type of employment variety is associated with divergent thinking. It is also consistent with the argument that high occupational variety coupled with high industry variety is too much variety for invention quality because of unwieldy and impractical ideas that are either not technically feasible or financially viable (e.g., Taylor & Greve, 2006). Similarly, we argue that invention quality is affected by too little or too much variety because both cases reflect a lack of learning different skills for building a commercially successful new venture. Whereas too little variety indicates a complete lack of interest in learning new skills, too much variety indicates a motivation toward satisfying a taste for variety.

Second, we found that industry variety negatively relates to entrepreneurial earnings. A simple explanation is that a lack of industry variety reflects deep within-industry knowledge that is more valuable to generating entrepreneurial earnings than the benefits of obtaining a variety of business knowledge from different industries. More precisely, while greater occupational or industry variety indicates greater invention quality, a lack of industry variety generates higher entrepreneurial earnings because it reflects more hours spent learning how a specific industry works. That is, an optimal investment strategy in occupational variety to foster invention quality coupled with a lack of industry variety may maximize new venture success. (10)

Alternatively, it is possible that the contrasting effects of the two types of employment variety on entrepreneurial earnings might be explained by occupational and industry variety being driven by a third, unobserved, variable: taste for variety. (11) Taste for variety functions like a consumption good: people with a high taste for variety are motivated to seek out a number of different job situations to satisfy this preference. Thus, those with a high taste for variety are more likely to become entrepreneurs because entrepreneurship offers an individual the opportunity to conduct a number of different tasks. Our regressions as well as additional data are consistent with this idea: the number of occupations and number of industries are high in our data set, they are also strongly positively correlated, and they are also positively correlated with becoming an inventor-entrepreneur.

Finally, we make another contribution by studying entrepreneurial success in terms of both invention quality and entrepreneurial earnings in the same data set. Specifically, we study entrepreneurial success by combining a subjective expert rating of invention quality with an objective monetary measure of entrepreneurial earnings, in contrast to previous studies that focus on one measure often to the exclusion of the other (Amabile & Mueller, 2007). For instance, researchers who employ experimental designs often use expert ratings to assess the creativity of new business ideas (e.g., Gielnik, Frese, Graf, & Kampschulte, 2012). This consensual assessment approach to measuring creativity is founded on the principle that an outcome is creative when experts agree that it is novel and useful (Amabile, 1996). However, such an approach is still susceptible to subjective bias since experts ultimately apply their own criteria to determine the novelty and usefulness of the product in question. Nevertheless, there are recent attempts that go beyond these subjective measures. For example, Taylor and Greve (2006) used collector market value of comics whereas Uzzi and Spiro (2005) used box-office revenues. By using a unique data set on inventors, we were able to study both the discovery of invention opportunities as reflected in the assessed invention quality, and the commercial exploitation aspects of entrepreneurial success to show the contrasting effects of occupational and industry variety.

Our study has some limitations. First, we do not have data on the number of years the inventors spent in each occupation or in each industry. This might have been helpful to further examine the relative contributions of both occupational and industry variety to entrepreneurial success. Second, the sample of inventors is drawn from administrative records of a particular agency servicing Canadian inventors. It is unclear how this sample represents all inventors because there is no omnibus of inventors for comparison. We instead compare our sample to a matched sample of Canadians and find some intriguing differences. Notably, our sampling strategy may be more comprehensive and inclusive of independent inventors than those basing their research on patent holders (e.g., Markman, Balkin, & Baron, 2002; Singh & Fleming, 2010) because patent holders do not represent all inventors. In fact, only 15% of inventions in our sample are patented, indicating that patent holders represent a small fraction of all inventors. With the kind assistance of the CIC, we were able to efficiently obtain a reasonably large sample that is likely to represent a fairly broad cross-section of inventor-entrepreneurs. Further, it provides us with the unique advantage of studying both the discovery of invention opportunities and the commercial exploitation of opportunities phases of entrepreneurship. Third, given the endogeneity of past experience it is difficult to make strong causal claims from this analysis. While timing of events helps to rule out reverse causality, unobserved heterogeneity is a primary concern since we use cross-sectional data. Further analysis of the two key variables introduced in this article--occupational and industry variety--using panel data could make a contribution. Finally, our analysis is limited to linear measures of occupational and industry variety.

Limitations notwithstanding, we recommend that inventor-entrepreneurs accumulate a substantial amount of prior experience by engaging in a number of different tasks, primarily by switching occupations within the same industry. More generally, we contribute to research focused on identifying the nondispositional factors associated with entrepreneurial success.

REFERENCES

Amabile, T.M. (1996). Creativity in context. Boulder, CO: Westview Press.

Amabile, T.M. & Mueller, J.S. (2007). Assessing creativity and its antecedents: An exploration of the componential theory of creativity. In J. Zhou & C.E. Shalley, (Eds.), Handbook of organizational creativity (pp. 31-63). Mahwah, NJ: Lawrence Erlbaum Associates.

Arentz, J., Sautet, F., & Storr, V. (2013). Prior-knowledge and opportunity identification. Small Business Economics, 41, 461-478.

Astebro, T. (2003). The return to independent invention: Evidence of risk seeking, extreme optimism or skewness-loving? Economic Journal, 113, 226-239.

Astebro, T. (2012). Returns to entrepreneurship. In D. Cummins (Ed.), Handbook of entrepreneurial finance (pp. 45-108). Oxford: Oxford University Press.

Astebro, T. & Bernhardt, I. (2003). Start-up financing, owner characteristics and survival. Journal of Economics and Business, 55(4), 303-320.

Astebro, T. & Bernhardt, I. (2005). The winner's curse of human capital. Small Business Economics, 24(1), 63-78.

Astebro, T., Chen, J., & Thompson, P. (2011). Stars and misfits: Self-employment and labor market frictions. Management Science, 57(11), 1999--2017.

Astebro, T. & Thompson, P. (2011). Entrepreneurs: Jacks of all trades or hobos? Research Policy, 40, 637-649.

Audia, P.G. & Goncalo, J.A. (2007). Past success and creativity over time: A study of inventors in the hard disk industry. Management Science, 53, 1-15.

Backes-Gellner, U., Tuor, S.N., & Wettstein, D. (2010). Differences in the educational paths of entrepreneurs and employees. Empirical Research in Vocational Education and Training, 2, 83-105.

Baker, K.G. & Albaum, G.S. (1986). Modeling new product screening decision. Journal of Product Innovation Management, 3, 32-39.

Bates, T. (1990). Entrepreneur human capital inputs and small business longevity. The Review of Economics and Statistics, 72(4), 551-559.

Baum, J.A.C. & Silverman, B.S. (2004). Picking winners or building them? Alliance, intellectual, and human capital as selection criteria in venture financing and performance biotechnology startups. Journal of Business Venturing, 19, 411 --436.

Becker, G.S. (1964). Human capital. Chicago, IL: University of Chicago Press.

Bhide, A. (1994). How entrepreneurs craft strategies that work. Harvard Business Review, 72, 150-161.

Bruderl, J., Preisendorfer, P., & Ziegler, R. (1992). Survival chances of newly founded business organizations. American Sociological Review, 57, 227-242.

Bublitz, E. & Noseleit, F. (2014). The skill balancing act: When does broad expertise pay off? Small Business Economics, 42(1), 17-32.

Campbell, D.T. & Fiske, D.W. (1959). Convergent and discriminant validation by the multi-trait-multimethod matrix. Psychological Bulletin, 56, 81-105.

Canadian Intellectual Property Office. (2015). What is intellectual property? Retrieved from http://www. ic.gc.ca/eic/site/cipointernet-internetopic.nsf/eng/wr03716.html?Open&wt_src=cipo-patent-main, accessed 4 November 2015.

Cooper, A.C., Gimeno-Gascon, J., & Woo, C. (1994). Initial human and financial capital as predictors of new venture performance. Journal of Business Venturing, 9, 371-395.

Corbett, A.C. (2005). Experiential learning within the process of opportunity identification and exploitation. Entrepreneurship Theory and Practice, 29(4), 473-491.

Cortina, J.M. (1993). Interaction, nonlinearity, and multicollinearity: Implications for multiple regression. Journal of Management, 19(4), 915-922.

Cressy, R. (1996). Are business startups debt-rationed? The Economic Journal, 106(438), 1253-1270.

Cropley, A. (2006). In praise of convergent thinking. Creativity Research Journal, 18, 391-404.

Davidsson, P. & Honig, B. (2003). The role of social and human capital among nascent entrepeneurs. Journal of Business Venturing, 18, 301-331.

Drescher, L., Thiele, S., & Weiss, C. (2008). The taste for variety: A hedonic analysis. Economics Letters, 10, 66--68.

Evans, D. & Leighton, L.S. (1989). Some empirical aspects of entrepreneurship. American Economic Review, 79, 519-535.

Fiet, J.O. (2002). The systematic search for entrepreneurial discoveries. Westport, CT: Quorum Books.

Fischhoff, B. (1975). Hindsight is not equal to foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1, 288-299.

Fleming, L., Mingo, S., & Chen, D. (2007). Collaborative brokerage, generating novel ideas, and creative success. Administrative Science Quarterly, 52, 443--475.

Ganotakis, P. (2012). Founders' human capital and the performance of UK new technology based firms. Small Business Economics, 39, 495-515.

Gielnik, M.M., Frese, M., Graf, J.M., & Kampschulte, A. (2012). Creativity in the opportunity identification process and the moderating effect of diversity of information. Journal of Business Venturing, 27, 559-576.

Gimeno, J., Folta, T., Cooper, A., & Woo, C. (1997). Survival of the fittest? Entrepreneurial human capital and the persistence of underperforming firms. Administrative Science Quarterly, 42, 750-783.

Greve, H.R. & Taylor, A. (2000). Innovations as catalysts for organizational change: Shifts in organizational cognition and search. Administrative Science Quarterly, 45, 54-80.

Gronau, R.G. & Hamermesh, D. (2008). The demand for variety: A household production function.

Review of Economics and Statistics, 90, 562-572.

Guilford, J.P. (1950). Creativity. American Psychologist, 5, 444-454.

Haber, S. & Reichel, A. (2007). The cumulative nature of the entrepreneurial process: The contribution of human capital, planning and environmental resources to small venture performance. Journal of Business Venturing, 22, 119-145.

Hartog, J., Van Praag, C.M., & Van der Sluis, J. (2010). If you are so smart why are you not an entrepreneur? Returns to cognitive and social ability: Entrepreneurs versus employees. Journal of Management Strategy, 19, 947-989.

Hirano, K., Imbens, G.W., & Ridder, G. (2003). Efficient estimation of average treatment effects using the estimated propensity score. Econometrica, 7/(4), 1161-1190.

Holt, D., Smith, T.M.F., & Winter, P.D. (1980). Regression analysis of data from complex surveys. Journal of the Royal Statistical Society, Series A, 143(Part 4), 474-487.

Klepper, S. & Thompson, P. (2010). Disagreements and intra-industry spinoffs. International Journal of Industrial Organization, 28(5), 526-538.

Lazear, E.P. (2005). Entrepreneurship. Journal of Labour Economics, 23, 649-680.

Little, R.J.A. & Rubin, D. (1987). Statistical analysis with missing data. New York: John Wiley & Sons.

March, J.G. (1991). Exploration and exploitation in organizational learning. Organization Science, 2, 71-87.

Markman, G.D., Balkin, D.B., & Baron, R.A. (2002). Inventors and new venture formation: The effects of general self-efficacy and regretful thinking. Entrepreneurship Theory and Practice, 27, 149-165.

Marvel, M.R. & Lumpkin, G.T. (2007). Technology entrepreneurs' human capital and it effects on innovation radicalness. Entrepreneurship Theory and Practice, 31, 807-828.

McAlister, L. & Pessemier, E. (1982). Variety seeking behavior: An interdisciplinary review. Journal of Consumer Research, 9, 311--322.

Oberschachtsiek, D. (2012). The experience of the founder and self-employment duration: A comparative advantage approach. Small Business Economics, 39(1), 1-17.

Parker, S. (2009). The economics of entrepreneurship. Cambridge: Cambridge University Press.

Pennings, L., Lee, K., & van Witteloostuijn, A. (1998). Human capital, social capital, and firm dissolution. Academy of Management Journal, 41, 425-440.

Shane, S. (2000). Prior knowledge and the discovery of entrepreneurial opportunities. Organization Science, 11, 448-469.

Shane, S. & Venkataraman, S. (2000). The promise of entrepreneurship as a field of research. Academy of Management Review, 25, 217-226.

Silva, O. (2007). The jack-of-all-trades entrepreneur: Innate talent or acquired skill? Economics Letters, 97, 118-121.

Singh, J. & Fleming, L. (2010). Lone inventors as sources of breakthroughs: Myth or reality? Management Science, 56, 41-56.

Stuetzer, M., Obschonka, M., & Schmitt-Rodermund, E. (2012). Balanced skills among nascent entrepreneurs. Small Business Economics, I, 1-22.

Taylor, A. & Greve, H.R. (2006). Superman or the fantastic four? Knowledge combination and experience in innovative groups. Academy of Management Journal, 49, 723-740.

Unger, J.M., Rauch, A., Frese, M., & Rosenvusch, N. (2011). Human capital and entrepreneurial success: A meta-analytic review. Journal of Business Venturing, 26, 341-358.

U.S. Patent and Trademark Office. (2015). General information concerning patents. Retrieved from http:// www.uspto.gov/patents-getting-started/general-information-concerning-patents#heading-2, accessed 4 November 2015.

Uzzi, B. & Spiro, J. (2005). Teamwork and creativity: The small world problem. American Journal of Sociology, 111, 447-504.

Van Buuren, S., Boshuizen, H.C., & Knook D.L. (1999). Multiple imputation of missing blood pressure covariates in survival analysis. Statistics in Medicine, 18, 681-694.

Wagner, J. (2003). Testing Lazear's jack-of-all-trades view of entrepreneurship with German microdata. Applied Economics Letters, 10, 687-689.

Ward, T.B. (2004). Cognition, creativity, and entrepreneurship. Journal of Business Venturing, 9, 173-188.

(1.) Others have studied an individual's educational variety and its impact on the willingness (Backes-Gellner, Tuor, & Wettstein, 2010) and probability (Lazear, 2005; Silva, 2007) to become an entrepreneur; the variety in cognitive abilities and the influence on entrepreneurial entry and earnings (Hartog, Van Praag, & Van der Sluis, 2010); and the impact of the number of functional areas worked in on the progress in the venture creation process (Stuetzer, Obschonka, & Schmitt-Rodermund, 2012). Further, Bublitz and Noseleit (2014) find that the number of expert skills currently used on the job positively affects earnings, both as an employee and as an entrepreneur, but that can be due to unobserved ability and does not capture the variety in skills obtained from prior employment experiences. Finally, Oberschachtsiek (2012) investigates those coming from unemployment and find no correlation between self-employment duration and prior employment variety.

(2.) Background diversity refers to differences in team composition based on personal attributes such as age, gender, and ethnicity. While age, gender, and ethnicity all affect the propensity to become an entrepreneur and in many cases entrepreneurial success (Parker, 2009), the effects of background diversity variables can only be studied in the context of teamwork as these variables cannot vary in diversity across individuals (e.g., age cannot be more or less diverse across individuals). While diversity in teams is frequently studied, the variety of an individual's employment-related activities has only recently been addressed.

(3.) There is a complementary literature on the role of specific human capital and its impact on entrepreneurial success. Investments in specific human capital for entrepreneurship has been measured by, inter alia, past entrepreneurial experience, and prior wage work experience in the industry for which the entrepreneur enters. For examinations of these effects, see for example, Astebro and Bernhardt (2003, 2005), Bates (1990), Briiderl et al. (1992), Cressy (1996), and Evans and Leighton (1989).

(4.) If sales data are truncated by survey date and data truncation is correlated with independent variables, coefficient estimates may be biased. We, therefore, forecast future sales conditional on truncation using exogenous parameters. This implies that the forecast is uncorrelated with covariates and should not produce biased regression parameters. In addition, the forecasts are likely better estimates of sales than assuming zero sales when data are truncated. We use the Bass diffusion with exogenous parameters to forecast sales if truncated.

(5.) Others have used different measures of occupational variety when testing the jack-of-all-trades theory. For example, variety in formal educational background (Backes-Gellner et al., 2010), the coefficient of variation in performance on items of a cognitive skills test taken when young (Hartog et al., 2010), the number of functional areas worked in (Stuetzer et al., 2012), and the variation in the number of different topics studied as an MBA (Lazear, 2005). These alternative measures produce less consistent results.

(6.) These are roughly equivalent to industries, but more narrowly defined by the Innovation Centre given their vast experience with the type of inventions they typically review. The categories are Environmental or Energy, Automotive, Sports or Leisure, Toys or Games, Medical or Health, Tools, Household or General Consumer Products, High Tech Equipment, Security or Safety, Industrial Equipment, and Other.

(7.) We ask if "You needed to have your skills complemented by their skills," "You needed the capital they provided," "They had contacts that were useful," "They had (nonfinancial) resources that were useful (land, equipment, plant)," and "other."

(8.) Prior studies, for example, have used whether the individual has had prior self-employment or business experience, whether the individual had worked in a business owned by family members, the length of prior work experience in the industry which the person entered as an entrepreneur, and vocational or trade studies (e.g., Astebro & Bernhardt, 2003, 2005; Bates, 1990; Briiderl et al., 1992; Cressy, 1996; Evans & Leighton, 1989).

(9.) We include measures of whether the idea for their invention was stimulated by something at work, whether they have had a vocational education (science or engineering, business degree versus arts or social sciences), whether they went to trade school, whether they had ever been self-employed, years of previous business ownership experience, the number of businesses owned, how long they had worked at developing invention (in general), how many inventions or ideas they had actively worked on in their lifetime, and whether they belonged to an entrepreneurial family (mother, father, or siblings having operated a business).

(10.) Related research shows a lack of industry variety decreased invention quality in the hard disk industry because over time, inventors' new products tend to be incremental variants of their earlier products rather than being fresh and novel (Audia & Goncalo, 2007). At the same time, comic-book artists who have extensive cross-occupational experience within the comic-book industry were better able to produce commercially successful new comics (Taylor & Greve, 2006).

(11.) There is considerable evidence documenting a general taste for variety (see for example, Drescher, Thiele, & Weiss, 2008; Gronau & Hamermesh, 2008; McAlister & Pessemier, 1982).

Thomas Astebro is professor of strategy and entrepreneurship in the Department of Strategy and Business Policy at HEC-Paris, 1 Rue de la Liberation, 78351 Jouy-en-Josas, France.

Kevyn Yong is associate professor of management and entrepreneurship in the Department of Management at ESSEC Business School, 2 One-North Gateway, Singapore 138502.

Astebro acknowledges financial support from the Canadian Social Sciences and Humanities Research Council and the Canadian Natural Sciences and Engineering Research Council for collecting the data and HEC Foundation and the HEC Leadership Center for support writing this article. We thank Elie Matta and Eric Uhlmann for comments.

Please send correspondence to: Thomas Astebro, email: astebro@hec.fr and to Kevyn Yong at yong@ essec.edu.

DOI: 10.1111/etap.12217

Caption: Figure 1: Probability of High Invention Quality
Table 1
Summary Statistics of Key Independent Variables: Inventor Prior
Employment Experience

                                  Fractions
                       Average     Invention     Commercial = 1
                                  quality = 1

Occupational fields

1                        .11         .08 *             .09
2 or 3                   .39          .38              .42
4 or 5                   .25          .23             .21 *
6-10                     .18         .22 *             .19
>10                      .07          .10              .09

Industries worked in

1                        .16          .14              .24
2 or 3                   .40          .40              .47
4 or 5                   .26          .26              .19
6-10                     .12          .14             .07 *
>10                      .06          .06              .03

N = 770. Two-tailed t-statistics test with equal group variances
for significant differences between groups (0 or 1 for each
outcome) are indicated with * p< .10. Invention quality = 1
indicates that the invention was assessed by the IAP at the CIC as
of high earnings potential prior to significant R&D expenditures
and prior to commercialization efforts. Commercial = 1 indicates
that there were positive earnings, whereas Commercial = 0 indicates
no earnings.

Table 2
Means, Standard Deviations, and Bivariate Correlations

                        Mean (SD)       l       2       3       4

1 Log (earnings)       1.01 (3.14)
2 Invention quality     .22 (.41)      .24
3 No. occ. fields      4.41 (2.84)     .03     .09
4 No. industries       3.90 (2.69)    -.04     .04     .47
5 Team invention        .08 (.27)      .13     .10    -.01    -.00
6 Team commerc.         .21 (.40)      .30     .17    -.03     .02
7 Inv. dev. dut.        .11 (.31)      .21     .24     .03    -.01
8 Age                  49.88 (9.88)    .01    -.09     .03     .03
9 Work, exp            19.10 (2.47)    .04    -.04     .10     .10
10 Log (R&D)           5.26 (3.40)     .30     .35     .09     .07
11 Log (time dev.)     4.35 (1.71)     .24     .29     .09     .04
12 Patent               .15 (.36)      .21     .19     .06     .01
13 Log (invest.)       3.48 (4.33)     .43     .42     .10     .08
14 Log (eff comm.)     2.59 (2.74)     .36     .33     .09     .04

                         5       6       7       8       9      10

1 Log (earnings)
2 Invention quality
3 No. occ. fields
4 No. industries
5 Team invention
6 Team commerc.         .12
7 Inv. dev. dut.        .09     .10
8 Age                  -.09    -.05     .02
9 Work, exp            -.06    -.02    -.01     .43
10 Log (R&D)            .12     .23     .19     .06     .10
11 Log (time dev.)      .10     .17     .10     .08     .08     .53
12 Patent               .02     .13     .06     .08     .03     .33
13 Log (invest.)        .18     .36     .18     .00     .02     .54
14 Log (eff comm.)      .09     .30     .14     .01     .02     .46

                        11      12      13

1 Log (earnings)
2 Invention quality
3 No. occ. fields
4 No. industries
5 Team invention
6 Team commerc.
7 Inv. dev. dut.
8 Age
9 Work, exp
10 Log (R&D)
11 Log (time dev.)
12 Patent               .34
13 Log (invest.)        .49     .40
14 Log (eff comm.)      .64     .39     .72

N = 770. All correlations greater or equal to .07 (Pearson r) are
significant at p < .05 or better. Means, standard deviations, and
correlations are drawn from one of five complete samples.

Table 3
Invention Quality and Entrepreneurial Earnings

                                  Invention quality
                                        (1)               (2)

Number of occupational fields                       .098 *** (.036)

Number of industries                                .109 ** (.049)

No. of occupational field *                         -.017 ** (.007)
 No. of industries

Team to develop the idea = 1        .177 (.215)       .179 (.218)

Invention developed as part of    .667 *** (.178)   .716 *** (.182)
 normal duties at work - 1

Age                               -.017 ** (.007)   -.018 ** (.007)

Years of work experience           -.008 (.027)      -.011 (.027)

Log R&D investments               .122 *** (.023)   .126 *** (.023)

Log time to develop invention     .098 ** (.044)     .094 * (.045)

Patent--1                           .135 (.167)       .158 (.169)

Invention quality = 1

Team to commercialize
invention = 1

Log commercialization
investments

Log effort to commercialize
invention

N                                       770               770

                                  Entrepreneurial earnings
                                         (3)                 (4)

Number of occupational fields                           -.453 (.493)

Number of industries                                  -2.072 ** (.852)

No. of occupational field *                             .169 * (.102)
 No. of industries

Team to develop the idea = 1        2.152 (2.699)       2.068 (2.666)

Invention developed as part of    6.161 *** (2.255)   5.714 ** (2.239)
 normal duties at work - 1

Age                                 -.081 (.105)        -.108 (.105)

Years of work experience             .624 (.491)        .867 * (.514)

Log R&D investments                  .404 (.348)         .423 (.346)

Log time to develop invention        .055 (.731)        -.037 (.722)

Patent--1                           1.194 (2.021)       .820 (2.001)

Invention quality = 1               2.177 (1.968)       2.451 (1.982)

Team to commercialize             7.072 *** (1.888)   6.567 *** (1.864)
invention = 1

Log commercialization             1.080 *** (.319)    1.105 *** (.317)
investments

Log effort to commercialize        1.030 * (.550)      1.120 ** (.551)
invention

N                                        770                 770

Results across five samples with complete data where item
nonresponses have been multiple imputed. Coefficient estimates and
standard errors in parentheses are computed using Little and
Rubin's (1987) formulae. Significance levels:
*** .01, ** .05, * .10. Regressions include 10 fields of
application dummies and 8 year dummies. Robust standard errors.
Invention quality is a 0/1 variable and results are from a Probit
model. Entrepreneurial earnings are the natural log of up to 10
years of future revenues from the commercialization of the
invention, discounted to 2003 Canadian dollars, and set to one
dollar (so that the log is zero) when there are no revenues. A
Tobit model is used for this outcome.
COPYRIGHT 2016 Sage Publications, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Astebro, Thomas; Yong, Kevyn
Publication:Entrepreneurship: Theory and Practice
Geographic Code:1USA
Date:Mar 1, 2016
Words:10362
Previous Article:Mobility and entrepreneurship: evaluating the scope of knowledge-based theories of entrepreneurship.
Next Article:Individual entrepreneurial exit and earnings in subsequent paid employment.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters