As a result, some new growth options may end up outside the firm boundary in startups founded by former employees.
Though there are many anecdotes of employee departures from innovative incumbents to entrepreneurship, it is not obvious that R&D will increase employee startups. Instead, R&D might lead the firm to grow internally or become a more interesting workplace, leading to greater employee retention. To our knowledge, this is the first paper to explore startup creation as a type of R&D spillover.
We ask how R&D affects employee departures to entrepreneurship using US census employer-employee matched panel data, or each public firm establishment year, we follow departing workers and examine whether they are on the founding team of a new firm – top five earners of a firm founded within three years of when R&D is measured.
Our baseline model shows that a 100% increase in firm R&D leads to a 8.4% increase in employee-founded startups. Over the course of the sample, above-relative to below-median R&D changes yield 10,541 additional employee-founded startups, which is 9.7% of all employee-founded startups in the data.
These R&D-induced startups are much more likely to be venture capital-backed than the average employee-founded startup, suggesting that high-risk high-reward growth options from R&D are more often reallocated.
The startup creation spillovers that we document could arise from two channels. In what we term the intellectual capital channel, R&D generates new ideas or technologies that are deployed by employees in new firms. The alternative channel is human capital, where learning-by-doing during the R&D process increases employees’ entrepreneurial skills.
The cross-sectional evidence is more consistent with intellectual capital. High-tech parents and those that conduct broader research have higher effects per dollar of R&D. Further, within the population of employee startups, higher parent R&D is strongly associated with venture capital backing. This associates the effect with new ideas.
The intellectual capital channel is consistent with R&D-induced employee entrepreneurship being a direct avenue for R&D spillovers. These spillovers are crucial to economic growth – they also seem to be large in magnitude. However, they are difficult to observe, and little is known about their transmission channel. We also do not know much about the identity of spillover recipients. The literature has typically assumed that potential recipients are close in technological or geographic space. Research at the individual level has focused on inventor networks, particularly in academia.
We extend this literature by documenting a specific transmission channel for R&D spillovers, and by focusing on R&D inputs and the firm boundary rather that patents.
The loss of the employee and R&D output may be costly to the parent, but several tests suggest that the costs are not extremely large. We expect that if the effect is very costly to the firm, it will be smaller in states that enforce non-competitive covenants. Instead, those states exhibit an effect similar to states that weakly enforce non-competitive covenants. We also expect that if the effect is very costly, it will be smaller in sectors where intellectual property is easier to protect. Instead, it is equally strong in these sectors. Further, R&D-induced startups are likely not to compete in product markets with their parents, because they tend to be in different industries.
Our results have two policy implications. First, the employee entrepreneurship effect of R&D implies greater corporate underinvestment in R&D relative to the social optimum than previously thought. Second, the presence of R&D spillovers are one motivation for offering firms tax credits that lower their cost of R&D investment. The employee entrepreneurship effect of R&D is much larger in the instrumental variables model than in the fixed effects regression. This suggests, albeit in a partial equilibrium sense, that R&D tax credits are effective in that they lead to greater R&D-induced employee entrepreneurship.
A final contribution of our paper is to offer another channel for the link between industrial agglomeration and R&D spillovers, which is often attributed in part to the importance of tacit information. R&D spillovers have been found to be quite local and to decline with distance. A remarkable 88% of employee-founded startups in our data are located in the same state as the parent.