Business owners, employees, and firm performance

The novel Finnish Longitudinal OWNer-Employer-Employee (FLOWN) database was used to analyze how the characteristics of owners and employees relate to firm performance as determined by labor productivity, survival, and employment growth. Focusing on the role of the employment history, the results show that previous experience in a high-productivity firm strongly predicts high productivity and probability of survival for the entrepreneur’s new firm. This can be interpreted as evidence of knowledge spillovers through labor mobility of both the owners and the employees. The results also show that the owner’s high education in a technical field is positively related to firm performance. Different findings for owner-entrepreneurs and pure owners suggest that the definition of entrepreneurship matters.


Introduction
Entrepreneurship has been the subject of intensive research and public policy debate for some decades. Recent evidence supports the established view that new businesses create a disproportionate share of new jobs (Haltiwanger et al. 2013). However, the associated mechanisms involve intensive market experimentation through entries and exits (Kerr et al. 2014), with job creation at some firms accompanied by job destruction at others (Carree and Klomp 1996;Davis et al. 1996;Davis and Haltiwanger 1999). Very few new firms achieve the sustained high growth required to make the transition from a small to a large enterprise.
An international comparison shows that a small number of rapidly growing firms account for differences in employment growth between countries (Anyadike- Danes et al. 2015). There is evidence that (young) firms that achieve high productivity can survive and grow, contributing to economic growth through productivityenhancing firm-level restructuring, but significant effects seem to take at least a decade to emerge (Hyytinen and Maliranta 2013;Dumont et al. 2016;Haltiwanger et al. 2017;Maliranta and Hurri 2018). In general, earlier business dynamics research has shown that entrepreneurship can be expected to have a sustained impact on economic growth when based on a firm's ability to achieve high productivity. In addition, the origins of any noteworthy contribution to economic growth may lie far in the firm's past.
Focusing on Btrue^entrepreneurs who run limited liability companies (rather than the self-employed), we find support for the economic impact of productivityenhancing entrepreneurship (Levine and Rubinstein 2016). Using the novel Finnish Longitudinal Owner-Employer-Employee (FLOWN) data, this study presents new evidence of how the skills of entrepreneurs contribute to company performance in terms of productivity, survival, and growth. More specifically, it will be shown that a formal university education and earlier work experience in a high-productivity firm are positively related to the performance of the entrepreneur's own business. This is especially the case when the owner herself works in the firm-that is, the finding pertains especially to entrepreneur-owner firms. With regard to the role of the owner's past work experience, our findings support the view that the importance of entrepreneurs as mediators of economic growth is based on the transfer of knowledge spillovers from high-performing incumbent firms to new (or older) businesses.
The rest of the paper is organized as follows: Section 2 provides a brief review of the literature. The data are described in Section 3. Section 4 reports the results of our analysis, and Section 5 concludes the paper.

Literature
The present analysis draws on the literatures related to entrepreneurship, human capital, business dynamics, and economic growth to investigate empirically how performance is explained by the human capital of the entrepreneur and her employees, along with multiple other characteristics of the firm and the prevailing economic conditions when the business started. This analysis illuminates the factors underlying firms' productivity, survival, and employment growth as the essential elements of the innovation-led micro-level dynamics of productivity and economic growth emphasized in current Schumpeterian growth theory (Aghion et al. 2014).
There is a large literature on factors explaining entrepreneurial choice (Hyytinen et al. 2014;Lofstrom et al. 2014;Sahaym et al. 2016) and examining where entrepreneurs come from (Hyytinen and Maliranta 2008;Elfenbein et al. 2010;Nanda and Sørensen 2010;Andersson et al. 2012). According to Lazear (2005), rather than specializing in any one skill, entrepreneurs are multi-talented jacks-of-all-trades, and this theory is supported by empirical evidence of the positive relationship between a varied work and educational background and the likelihood of starting one's own business (Stuetzer et al. 2013). There is extensive evidence that the founder's human capital is a crucial determinant of business success (e.g., Bates 1990;Cooper et al. 1994;Dahlqvist et al. 2000;Shane and Stuart 2002;Colombo and Grilli 2005), and there is also large literature on the role of personnel characteristics in firm performance Ilmakunnas and Maliranta 2005;Haltiwanger et al. 2007). One of the contributions of the present paper is that it grounds these ideas in an exceptionally diverse range of data.
In terms of economic growth, entrepreneurial activity of particular interest can be found where the business owner provides labor input with one or more employees, as such firms are more likely to exhibit growth intentions and an ability to create Bgood^high productivity jobs. A number of studies have examined how prior work experience explains entrepreneurial success. At the same time, a growing literature in labor economics shows how employee mobility can contribute to the productivity of the destination firm (Maliranta et al. 2009;Balsvik 2011;Stoyanov and Zubanov 2012;Ilmakunnas and Maliranta 2016).
One deficiency of the literature on the effects of human capital on firm performance is that the roles of entrepreneurs and their employees have typically been examined in isolation. The need for a parallel analysis is highlighted by the fact that the skills of entrepreneurs and their staff may be complementary (as suggested for instance by Lazear's (2005) theory). The absence of any such synthesis reflects a lack of data linking business owners, firms, and employees over time (Goetz et al. 2017). To our knowledge, only Rocha et al. (2016) have studied the human capital of both founder team and company personnel. However, while they analyze the effect of founders' management and entrepreneurial experience, they fail to provide a broader view encompassing the work experience of entrepreneurs and employees.
In Nordic countries, comprehensive tax-based register data offer new opportunities for empirical analyses of entrepreneurship. Linking data on business owners to their past experience, both in the labor market and as owners of earlier businesses, allows us to study the dynamics of entrepreneurship, business formation, and firm performance from a totally new and wider perspective. For example, in Norway, Berglann et al. (2011) constructed a more extensive picture of entrepreneurship on the basis of linked register data on ownership.
In most countries, concentrated firm ownership with a single controlling owner has been found to predominate (Claessens et al. 2000;Faccio and Lang 2002). In this one-shareholder context, ownership and management are also closely connected, and the influence of managers on firm performance is well documented (Bloom et al. 2012). Based on the universe of limited liability companies, the present study concentrates on owners with a majority share, who have a strong motivation to maximize the value of their firm and to control the use of its assets. In particular, we focus here on business owners who are also on the firm's payroll and who are themselves typically involved directly in production, characterized with good reason as entrepreneurs. For the purpose of comparison, separate analyses were also performed on firms where the main owner is not working principally at the end of the year-the so-called pure owner firms. Drawing on novel linked data on business owners, this analysis supplements earlier Finnish studies of entrepreneurship, which include Pajarinen et al. (2006, 2011), Ali-Yrkkö et al. (2007, Hyytinen and Maliranta (2008), Murray et al. (2009), Tourunen andLaaksonen (2009), and Kontinen and Ojala (2010).

Data
The Finnish Longitudinal Employer-Employee Data (FLEED) have been used in the analysis of entrepreneurship. The FLEED data are constructed by linking various administrative registers, such as Employment Statistics, Business Register and Financial Statements Statistics of Statistics Finland. The basic unit in this dataset is an individual who belongs to the workingage population of Finland and who can be linked to the company and plant in which she works. The data have three characteristics that are particularly useful in the entrepreneurial analysis: the data allow following basically the entire working-age population of Finland over time starting from 1988, include a rich information content on the individuals and their jobs, and make it possible to trace individuals' labor market transitions (Ilmakunnas et al. 2001;Hyytinen and Maliranta 2008). However, the lack of data linking limited liability companies to owner entrepreneurs has represented a serious limitation so far (Maliranta and Nurmi 2004). Fortunately, data on the enterprise links of persons defined as entrepreneurs have improved in recent years thanks to the Self-Employed Persons' Pension Act (YEL) and the Farmers' Pension Act (MYEL). Legislation prior to 2011 made pensions obligatory for persons owning more than half of their company stock or voting rights, either alone or with family members. At the beginning of 2011, this threshold was lowered to 30% for personal ownership (50% with family members). In addition, that person must be in a leading position and working in the company. Beyond this, detailed information on ownership was not available.
The present study utilizes new register data on ownership from the Finnish Tax Administration. Since 2006, tax returns for business activities have included information on the principal owners of corporate entities. Limited liability companies are obliged to report the personal information of 1-10 shareholders; if there are more than 10, the rule applies to those who own at least 10% of the capital stock. In addition, every person in receipt of a shareholder loan must also be reported.
By linking information on owners, employers, and employees, we have created a new Finnish Longitudinal OWNer-Employer-Employee (FLOWN) database. 1 For present purposes, this database has been used to focus on Btrue^entrepreneurs only, defined as the primary owners of limited liability companies, and the data were extended by new variables describing the histories of owners and personnel. This rich dataset includes the following distinct dimensions.
Owner data includes the personal characteristics and history of the principal owner of each limited liability company; enterprise-level measures of ownership structure were also calculated. (source: Tax register data, FLEED, Statistics Finland) Enterprise data includes information on the demographic status and performance of each limited liability company. (source: Business Register and Financial Statements data, Statistics Finland) Employee data includes information on the characteristics of personnel in each enterprise and their previous employment relationships. (source: FLEED) This linking process was complicated, as enterprises may have multiple person-id or business-id owners, as well as owners with multiple enterprise linkages. We traced ultimate owners through one or two company linkages-for example, if company A is partly (0.5) owned by company B, which is in turn partly (0.5) owned by person 1, we define person 1 as one owner of company A with the corresponding share of ownership (0.5 × 0.5 = 0.25). As we were unable to differentiate between cash flow rights and voting rights, these were assumed to be equal.
Our target population was Finnish business sector enterprises in 2008 with at least one employee, measured on full-time equivalent units (according to the Business Register). The study examined the performance of manufacturing industries (NACE 2: 15-22, 24-37), construction (45), and services (50-52, 55, 60-64, 70-74) in the period 2009-2013. To measure firm age irrespective of mergers and acquisitions, the age of the oldest establishment was used to create birth cohorts. Information on establishments belonging to the firm is available annually from 1988. The study concentrates on firms with highly concentrated ownership, where it was possible to identify one principal individual owner with a share of more than 50%. These owners were divided into (a) entrepreneurowners (working in the firm that they own) and (b) pure owners with no employee status in their firm. Employment status was defined according to end of year data. Analysis is performed separately for young firms (established in 2003-2008) and for all cohorts where younger and older firms are pooled. Three distinct least squares models of firm performance were estimated to explain labor productivity (log) levels in 2009 and 2013 and employment growth (log-difference) between the years 2009 and 2013. Labor productivity is computed as value added per the number of full-time equivalent persons in the firm (negative values are excluded for technical reasons, see Table 1 for information on the percentages of the negative value added originally). Firm survival/exit up to 2013 was estimated using a simple logit model, with separate estimates for entrepreneurowner and pure owner firms. Regressors were measured in 2008 and can be divided into firm attributes (firm size, capital intensity, birth year effects, two-digit industry); owner attributes (education, general experience/age, work experience in the firm, share of women, previous experience/spillover from another employer); and personnel attributes (similar to those of the owners). The owner's other principal ownerships were also controlled for.
Similarly to the study by Ilmakunnas and Maliranta (2005), education is classified into four levels: 0 (basic comprehensive school education); 1 (vocational or upper secondary); 2 (university of applied sciences or lower university), and 3 (higher university). The three highest levels are classified further into two broad fields of studies: Btechnical^(natural sciences or technology) and Bnon-technical^(other fields). So, eventually education is defined in seven categories in terms of level and field. Linkages to the previous employer (i.e., the latest other employment relationship found in [2001][2002][2003][2004][2005][2006][2007] were used to evaluate the role of spillovers. Previous work experience was quantified as length of employment relationship and quality of the firm in terms of relative productivity in the industry (two-digit NACE 2), measured as productivity quartiles.
Estimations were based on the shares approach from the productivity analysis literature, where the shares of personnel with respect to different attributes (e.g., education and experience) are used as explanatory variables. Ilmakunnas and Maliranta (2005) used a similar approach to study labor characteristics and wage-productivity gaps. This approach offers a flexible way to describe the personnel structures of the firm. For the owners, the nature of the data (only one primary owner per firm) meant that shares equaled 0 or 1. Even though they may have employee status, owners were not included in personnel, enabling stricter comparison between the two groups.
Following the examples of Ilmakunnas and Maliranta (2005) and Davis et al. (2014), our baseline analysis is performed by employment-weighted regressions (firm employment in 2008 was used as weight). In so doing, our results reflect better the business sector as an aggregate whole, ensuring that the large number of small enterprises would not dominate the results. Unweighted results are pretty similar to weighted results for younger firms, where the size distribution is not unduly skewed. However, nor does the weighting has an effect on our main findings with pooled data. In what follows, we focus on commenting weighted regressions but we also report important estimates obtained with unweighted regression in footnotes.

Empirical analysis
4.1 Owners, employees, and the success of young firms The first part of the analysis began by focusing on young firms established in the period 2003-2008 and their subsequent performance. To properly identify a firm's operational start-up year, plant-level information on entry was used to determine birth cohorts. The findings suggest a definite association between owner and employee history and firm performance. In the second part of the analysis (Section 4.2), we pooled the data for young and older cohorts to investigate any possible differences in the results.
To begin, prime owner firms were defined as those limited liability companies for which one individual owner owning more than 50% of company stock could be identified. These prime owner firms constitute about half of all the limited liability companies in the Finnish private business sector. These firms were further divided into entrepreneur-owner and pure owner, according to the employer of their prime owner at the end of the year. Owner employees were defined as those having an employment relationship or an entrepreneur status with the firm they primarily owned; pure owners were defined as those who could not be connected to the firm as an employed person. No account was taken of ownership changes between birth and the period of analysis. In some cases, the primary owner may have changed during the firm's life cycle, but we were especially interested in the influence of characteristics of the owner closest to production in the period before firm performance was measured. Basic characteristics of the 2008 data (both for all birth cohorts and for young firms only) are described in Table 1.
The summary shows that a prime owner is involved in production in about two thirds of prime owner firms while the prime owner is connected to the firm mainly through cash flows in only a third. For young firms, the situation is more balanced. Somewhat surprisingly, the differences in firm attributes are not striking; pure owner firms are on average slightly larger, but their labor productivity and capital structure remain similar to entrepreneur-owner firms. Up to 2013, young firms were smaller and less likely to survive than the average. However, the probability of survival was on average 10 percentage points higher for entrepreneur-owner firms than for pure owner firms.
A comparison of staff personal characteristics with those of prime owners reveals some distinct differences. Owners are more highly educated than their employees, especially in the case of pure owners of young firms, where 22.5% (= 4.2% + 7.2% + 7.1% + 4.0%) of owners have university of applied sciences or higher university degrees while only 12.0% (= 4.2% + 2.1% + 4.2% + 1.5%) of staff are highly educated. Across all firms, entrepreneur-owners more often have a technical vocational education (27.3%) than pure owners (22.6%).
Owners were found to be older and more experienced than their personnel. A great majority of owners are 40 or older while a majority of the employees are 40 or younger. In young firms, owners and employees are clearly younger; half of entrepreneur-owners are older than 40, and almost a half of their employees are younger than 30. While most entrepreneur-owners have worked for the firm for 1-10 years, most of their employees have worked there for less than 5 years. Differences in experience between pure owners of all firms and their employees are even greater, with 20.9% of pure owners close to retirement or retired and 3.8% of employees. Only 15.0% of prime entrepreneur-owners and 19.0% of pure owners are women. For younger cohorts, these shares are slightly higher.
We were particularly interested in how owners' and employees' previous experience influenced the performance of their current firm, based on two aspects of that previous experience: length and quality (where quality is defined as relative productivity level within that industry for their last previous employment during the period 2001-2007). Working history was traceable for about half of the owners and employees. Unsurprisingly, previous employment spells can more often be traced for owners and employees of younger firms; of the remainder, those with no experience in another firm may be working in the same firm as in 2008, or defined as self-employed, unemployed, or students. As expected, a small share (3.8%) of owners in young entrepreneur-owner firms had at least 15 years' experience in the previous firm while a relatively large percentage (40.9%) of owners in entrepreneur-owner firms had 1-5 years' experience in the previous firm.
Quality of previous experience was defined as the relative productivity of the previous employer by industry-specific productivity quartile 2 ; missing links or missing information meant that this could not be specified for all firms. Labor productivity was calculated only for firms with at least one employee, and some production data were based on information from the previous or following year. In some cases, the previous employer might be a public sector organization with no value added or a small Bnatural person^firm. For young firms, the performance of the previous employer can be defined for two thirds of the entrepreneur-owners and half of the pure owners. For the pooled sample, previous employer productivity can be determined only for every fourth (employee) owner. For employees, the figures are person-weighted averages. Information about previous experience could be traced for every second employee. For 80% of firms, some information could be found on employees' previous experience, and for 75% of firms, information on past productivity was available for at least one employee (not shown in the table). More interesting observations can be made about patterns related to the productivity of the previous firm/employer. The largest percentage of owners come from the highest quartile firms, and the fewest come from the lowest quartile firms; these patterns are consistent for both entrepreneur-owner firms and pure owner firms and among both young firms and all firms, suggesting mobility among owners away from high-productivity firms. These dynamics are consistent with the view that owner mobility is potentially an important mechanism of productivity spillover between firms, including spillover from older to newer production. The comparison with corresponding patterns for employees is very interesting, showing that a significant percentage of employees had worked in a relatively low-productivity firm, suggesting employee mobility away from low-productivity firms. These patterns are consistent with the view that employee mobility is one element of productivity-enhancing restructuring (i.e., creative destruction).
The results of the econometric models for young entrepreneur-owner firms are presented in Table 2 and for young pure owners in Table 3. We compare the importance of owner and employee characteristics for firm performance, measured in terms of labor productivity (in 2009 and 2013), probability of survival (2009)(2010)(2011)(2012)(2013), and employment growth (2009)(2010)(2011)(2012)(2013). In addition to personal characteristics, we controlled for firm size, age, number of primary ownerships (> 1), capital structure, and sector in 2008.
Estimates based on educational attainment reveal that higher university level education of the prime owner in 2008 is strongly positively related to firm labor productivity in 2009 (and in 2013). For example, the results suggest that a firm whose entrepreneur-owner has a technical higher university education is 25.7% 3,4 more productive than an otherwise similar firm whose entrepreneur-owner has only a basic education. Interestingly, there is no equivalent effect for employees. However, when the variables controlling education and other owner characteristics are omitted from the model, the effect of employee technical higher university education becomes economically (= 28.4% = exp.(0.250)-1) and statistically (p = 0.05) significant (results are not reported here), confirming a need to control owner and employee effects together. 5 The results also suggest that owner technical education increases survival probability, but owner education appears unrelated to firm employment growth.
Estimates of the effects of general experience (age) suggest a hump-shaped pattern in effect on survival, peaking in the 41-50 age group for both owners and employees. The same pattern was found using a continuous age variable (results are not reported here). Generally, however, age structure has little predictive power for performance among young firms. While owner gender seems to have no effect on productivity, there is some indication that percentage of female employees is negatively related to (current) productivity level but positively related to survival.
Perhaps the most interesting results pertain to the effects of previous experience (in terms of length and quality). As Table 2 indicates, length of previous experience of entrepreneur-owners is generally unrelated to firm performance. However, there is a strong positive relationship between the productivity level of the previous firm and firm performance in terms of productivity and survival (but not employment growth). Our estimates suggest that a new firm whose entrepreneurowner has had experience at a low-productivity firm is 11.5% (= exp(0.109)-1) 6 less productive in 2009 (12.2% = exp(0.115)-1 in 2013) 7 and has a 7.5% points lower probability of survival than an otherwise similar firm whose entrepreneur-owner has had experience at a high-productivity firm. However, no significant difference was found between these firms in terms of employment growth. So, in addition to the stronger mobility of owners away from high-productivity firms than from lower-productivity firms (see Table 1), these results suggest that the flow of entrepreneur-owners from high-productivity firms is reflected in the better performance of young destination firms.
Corresponding results for employees of the entrepreneur-owner firms suggest a substantially stronger relationship between the performance of previous and current firm. According to the estimates, a new firm  whose employees all come from a low-productivity firm is 27.0% (= exp(0.239)-1) 8 less productive and has a survival probability 8.6% points lower than an otherwise similar firm whose employees all come from a highproductivity firm. While we interpret these results as indicative of spillover, it is important to acknowledge that other mechanisms may also be at play here. It can be argued, for example, that the matching of high-skilled employees with high-technology firms may be reflected in the transitions of employees from high-(or low-)productivity firms to other high-(or low-)productivity firms (Ehrl 2014). Nevertheless, it is notable that owners of new high-productivity firms often come from older high-productivity firms where they previously worked. Table 3 reports the results for young pure owner firms; some key findings will be briefly reported here. Generally, characteristics of pure owners are only weakly related to firm performance. However, a pure owner with a technical higher university education has a significant positive effect on productivity in 2009 once controls for employee characteristics are dropped. One interesting exception is the strong negative relationship between pure owner age (general experience) and productivity level in 2013. Another exception is that high productivity is positively related to relatively long experience (10-15 years) at the previous firm. Contrary to our findings for entrepreneur-owner firms, no relationship was found between the productivity level of the previous firm and the productivity level or survival of the pure owner's firm.
Additionally, the coefficients of the model suggest that, among employees of pure owner firms, previous employer performance is particularly important in explaining productivity differences. The results indicate that a young firm whose employees all come from a low-productivity firm is at least 46.7% (= exp(0.383)-1) 9 less productive in 2009 (51.3% = exp(0.414)-1 in 2013) 10 and has a 13.3% points lower survival probability than an otherwise similar firm where all employees come from a highproductivity previous employer. The effects of employee characteristics for entrepreneur-owner and pure owner firms are more similar, though somewhat less precise for pure owner firms, perhaps in part because of a smaller sample size. While productivity level is negatively related to percentage of female employees, previous employer productivity level is strongly positively related to current employer productivity level in 2009 and 2013.
These findings broadly correspond to those of earlier studies; for example, Ilmakunnas and Maliranta (2005) reported the positive effect on productivity of education and the negative effect of younger and less experienced staff. While Rocha et al. (2016) found that firm performance benefits from the quantity and quality of founders' human capital, they also found that owner and workforce quality had stronger effects on employment growth. 8 The estimate from the unweighted regression is 29.2% = exp(0.256)-1. 9 But the estimate from the unweighted regression is only 30.2% = exp(0.264)-1. 10 But the estimate from the unweighted regression is only 33.9% = exp(0.292)-1.  In the second part of the analysis, we pooled young and older firms into the same sample. This has two advantages. First, a larger sample provided more degrees of freedom, yielding statistically more precise estimates. Second, a comparison of the results of our previous analysis of only young firms with pooled data results enabled the identification of any differences in effects between younger and older firms. Results from the pooled data are reported in the Appendix Tables 4 and 5.
The birth year of a firm (or firm cohort) was determined on the basis of the firm's oldest establishment. By linking firm-and establishment-level data, we were able to trace firm birth year back as far as 1989. Our regression models of firm performance included a set of dummy variables for firm birth year, using firms born in 1988 or earlier as a reference group. As well as controls for firm birth year, previous models for the pooled data were estimated by adding length of seniority of owners and employees to measure the effects of firmspecific work experience. The larger sample enabled more precise estimation of the effects of entrepreneurowner and employee characteristics, providing a clearer indication that education level is strongly related to firm productivity. However, in economic terms, the effects are generally smaller for entrepreneur-owners than for employees. In particular, there is a very strong relationship between the educational level of employees (measured in 2008) and productivity in 2013 (and less so in 2009). One interpretation of this difference is that the productivity effect of a higher education takes time to emerge because the mechanism involves timeconsuming innovation efforts. There is also evidence that higher education among both entrepreneur-owners and employees increases survival probability.
There is no indication that productivity increases with entrepreneur-owner or employee age, but employee seniority was found to be positively related to productivity in 2013. According to the estimates, female entrepreneur-owners' firms have lower productivity levels than those of males, but there is no difference in survival probability or employment growth. We also found that the effect of gender on productivity is stronger for employees than for entrepreneur-owners.
Length of experience of entrepreneur-owners in the previous firm does not affect productivity (as was true of young firms only). In the pooled data, no relationship was found between employees' length of experience in the previous firm and firm productivity level. This contrasts with our earlier finding for young firms that low productivity level in 2009 is associated with a high percentage of employees with little previous experience (less than 1 year) or none. However, this gap vanishes for productivity in 2013. One possible explanation for this is that the experience obtained in the current firm compensates for employees' initial lack of experience-based human capital. As shown in Tables 1, 72.7% (= 100.0% − 27.3%) of employees in young entrepreneur-owner firms had at least some previous experience in another firm. The corresponding number for all entrepreneur-owner firms is 54.3% (= 100.0% − 45.7%). Overall, the employees of young firms have more previous experience in another firm than the employees of older firms.
Results from the pooled data (reported in the Appendix Tables 4 and 5) confirm the earlier findings for young firms that previous employer productivity is strongly positively related to current productivity level and survival probability for entrepreneur-owners and employees. This suggests that the role of productivity spillover is not confined to young firms. These effects are less clear for pure owners, suggesting that owner involvement in production is important for transfer of information and implementation of new technologies. However, larger standard errors prevent any strong conclusions (which is at least partly a result of having now a smaller sample size).
As a final step, the analysis was extended to investigate more thoroughly the birth cohort effects from previous models for entrepreneur-owner firms. Figure 1a shows estimated cohort effects on productivity in 2009 and 2013, and Fig. 1b shows cohort effects on survival probability and employment growth. Figure 1a suggests, for example, that when controlling for a wide array of variables in our regression model, firms established in 1995 were 3.1% (=exp(0.031)-1) more productive in 2009 than firms established in 1988 or earlier. Figure 1b shows that firms established in 1995 had 2.8% points lower survival probability and 1.9% (=exp(0.019)-1) lower employment growth from 2009 to 2013 than firms established in 1988 or earlier. While there are considerable short-term fluctuations in cohort effects, these patterns generally indicate that younger firms exhibit lower productivity (especially when measured in 2009) and lower survival probability; on the other hand, surviving (very) young firms exhibit higher employment growth when compared to older firms, broadly aligning with the evidence from the firm growth literature (Haltiwanger et al. 2013)

Further robustness checks
A number of robustness checks were performed in the course of the analysis. Beyond those mentioned above, this section summarizes some of the most interesting findings. As a natural extension, measuring firm performance by average wages per employee rather than labor productivity does not markedly alter the main findings, with similar or even stronger effects of both entrepreneur-owner and employee technical university education on firm performance for all firms and young firms. While spillover from previous firms resembles productivity estimations in 2009, both owner and employee quality effects for wage rate in 2013 are mostly statistically insignificant. For pure owner firms, there is some evidence that employee mobility from highproductivity firms is also reflected in 2013 wage level. As expected, there is a clear positive connection between firm size and level of wages.
With regard to the identification of owners, primary owners identified in the tax registers were compared to the FLEED definition of entrepreneurs, based on the Self-Employed Persons' Pension Act. This information is restricted to the owners who have an employment relationship with the firm. Based on complementary calculations, results using the FLEED information resemble our findings for entrepreneur-owners using FLOWN data. It should be noted that Statistics Finland revised its system for production of business statistics in 2013. Given these changes in the definition of statistical units and the harmonization of some firm classifications, this may result in time series breaks. However, measuring productivity or employment growth to 2012 rather than 2013 does not markedly alter our main conclusions.
We have performed OLS estimations with a Stata procedure that weights each observation by employment (measured by full-time equivalent units). More precisely, we use analytic weights that weight observations as if each observation is a mean computed from a sample of size w, where w is the weight variable. 11 There are two main reasons why we prefer using employment-weighted estimation, as in Ilmakunnas and Maliranta (2005), Davis et al. (2014), or Ilmakunnas and Maliranta (2016), for example. Firstly, we are ultimately concerned with the effects at the aggregate industry level and weighted regressions are more compatible with those than unweighted ones (aggregate labor productivity is an employment weighted average labor productivity of the firms, see Van Biesebroeck (2003, footnote 37)). Second, weighted estimation provides us with a more efficient procedure in the presence of heteroscedasticity. We have presented the key findings as unweighted in footnotes. The main conclusion is that for the findings regarding young firms, the weighting does not make much difference. For all firms, the results on productivity spillovers through the entrepreneur-owner are even stronger if unweighted. Our all unweighted regression results are available upon request.
The effects of previous experience of those owners and employees that have no previous employment status may also entail important policy insights. It would be interesting to investigate in more detail the role of owners and workers coming from unemployment or studies. Previous employment status could also be divided according to the private or public sector or entrepreneurship status. We have performed some estimations for young firms by using a variable that indicates the main type of activity status in 2002 separately for the owner and the employees. The variable classifies previous experience in private sector employment, public sector employment, entrepreneur, unemployed, student, or the rest. 12 There are some very interesting and partly peculiar remarks (only statistically significant results are reported here): For young owner-entrepreneur firms, the owner's previous entrepreneurial experience seems to affect productivity and survival negatively so that compared to owner background in the private sector employment, the firm is 11.2% (= exp(0.106)-1) less productive in 2009 and has a 5.2% points lower probability of survival. Firms with owner background in the public sector have 6.2% (= exp(0.0601)-1) higher employment growth than firms with the owner-entrepreneur coming from the private sector. For pure owners, firms with the owner coming from unemployment have 18.8% (= exp(0.172)-1) lower productivity and from student background 15.5% (=exp (0.144)-1) lower productivity 11 Using analytic weights is equivalent to estimating a model multiplied by the square root of w where w is the weight variable. Instead of using individual-level data, our data presents enterprise-level averages so the coefficients and covariance matrix from the transformed regression are obtained using this procedure. See more information: https://www.stata.com/support/faqs/statistics/analytical-weights-withlinear-regression/ 12 We thank one of our anonymous referees for this suggestion. than firms with the owner coming from the private sector. Pure owner firms with previous entrepreneurs as owners have 6.0% points higher survival chances. For personnel history, the findings are less noteworthy. These findings deserve more attention in the future work but clearly go beyond the scope of this paper where the focus is on productivity spillovers between firms. However, inclusion of this variable in our model did not change the main results of this paper.

Concluding remarks
How the characteristics of owner entrepreneurs and their employees affect firm performance and economic growth remains poorly understood in the literature. By creating a new dataset with significant research potential linking owner, employer, and employee, this paper opens up novel perspectives illuminating the impact of entrepreneurial skills on economic growth, so informing policy debate on the effects of ownership, education, and economic conditions on firm performance.
The diversified paths of primary owners and their employees are reflected in future company performance. Previous employer quality, measured in terms of relative productivity, is transferred through owners and employees as knowledge spillover related, for example, to technology or management. High-quality owners create firms capable of achieving and maintaining sustained high performance in terms of productivity, survival, and employment growth.
Our results lend support to the view that employees' entrepreneurial skills nurtured in highproductivity firms can be transferred to achieve higher productivity, especially in entrepreneurowner firms. First, there is a strong positive relationship between the productivity level of the firm where the owner previously worked and the productivity level of the firm where the owner now works. Second, there is evidence of considerable employee mobility from high-productivity firms to ownership of a new firm (where the owner also works). These findings are consistent with the view that the transition of employees from highproductivity firms to entrepreneurship is an important business dynamic, driving knowledge spillover in the economy. Our results also indicate intensive employee mobility from low-productivity firms toward new and young firms, representing an important element of creative destruction. The reallocation of employees in creative destruction means that a greater share of the employees provides labor inputs to productively managed firms (Bloom et al. 2016). The present analysis also demonstrates that identification of owners is crucial for understanding company life cycles. Young firms have relatively younger and more educated human capital but are more dependent on the inflow of know-how from more experienced firms.
Our results demonstrate the importance of considering the independent effects of the owner and employee characteristics in parallel manner in any analysis of firm performance, as owner and employee background and skills may play different roles in the development of employment and productivity. In addition, this analysis indicates a need to deal separately with entrepreneur-owner and pure owner firms. In entrepreneur-owner firms, an owner's technically orientated education was found to impact positively on productivity performance and survival probability, but no such relationship was found in pure owner firms. One explanation for this difference is that closer owner links to production are needed to successfully exploit technical education and previous experience. In contrast, the potential contribution of pure owners pertains to factors that cannot be captured by measures of education and experience.
This also relates to definitions of entrepreneurs and entrepreneurship. The present analysis illuminates entrepreneurial performance when we rely on a narrower definition of entrepreneur as requiring both strict (majority) control of ownership and direct involvement in production as an employee of the firm. For entrepreneurs who meet only the broader definition of having majority control but no direct involvement through employment, we were unable to establish any clear empirical relationship with owner characteristics. The analyses did however reveal a number of important empirical relationships concerning the characteristics of employees of pure owner firms. These findings are unsurprising in cases where entrepreneurs are multi-talented jack-of-all-trades with no single specialized skill and so hire professionals with specialized education and/or experience, as described in Lazear's (2005) theory.
The present study provides a starting point for closer analysis of owner entrepreneurs and their importance for the wider economy. Policy makers should devote special attention to the role of labor mobility in entrepreneurial activity, as it seems to be an important channel of knowledge spillover from high-performing incumbent firms to new (or older) businesses. In other words, the effect of the prior experience of entrepreneurs and their employees should be considered more broadly. In future work, we aim to identify truly entrepreneurial start-ups with the greatest growth potential. We are also interested in the productivity effects of a more diversified ownership structure and multiple owners. Finally, future research should devote more attention to the underlying mechanisms connecting initial business conditions and creative destruction to firms' long-run productivity.   Pseudo-R-squared 0.0536 ***p < 0.01; **p < 0.05; *p < 0.1   Pseudo-R-squared 0.0685 ***p < 0.01; **p < 0.05; *p < 0.1