Research

Abstract: Firms increasingly compete for scientific talent to drive innovation. While individuals' productivity in an industrial setting is uncertain at the time of hire, firms can observe measures of academic productivity. However, because academia and industry operate under different institutional logics, the scientists most productive in academia may not be those most productive in industry. Using confidential administrative data on 40 cohorts of U.S. PhDs linked to wage and publication records, I estimate individuals' expected productivity in both academia and industry, regardless of their actual sector of employment. I do so using Marginal Treatment Effects and an instrument based on variation in firms' demand for PhDs upon graduation. I find that individuals' academic and industrial productivity are only moderately aligned (correlation=0.28), and that this alignment has declined over time. In fields with weaker alignment, firms are more likely to hire candidates who are subsequently laid off -- despite these individuals displaying strong academic productivity at graduation. I also find that individuals with higher industrial productivity require lower wage premiums to work in industry. Together, the results suggest that relying on academic productivity as a proxy for industrial potential can lead firms to hire individuals who both underperform in industry and are more costly to recruit. Overall, this study shows that evaluating scientists based on performance indicators developed in a different institutional context can result in costly hiring mismatches for firms.

Abstract: Should organizations concentrate scarce resources among a few knowledge workers or distribute them more broadly? The answer to this question depends on whether additional resources enable more ambitious work or simply allow lower-priority projects to move forward. We examine how the direction of innovative output changes as resource constraints are relaxed in the context of high-performance computing (HPC), a critical input to modern scientific research. Using data from XSEDE, an NSF-funded program that allocates supercomputing resources to researchers, we leverage system-wide capacity constraints to identify the causal impact of variation in resource constraints on scientific output. We find that relaxing constraints increases the number of publications and shifts the direction of research. Scientists pursue less popular and newer topics, explore areas beyond their prior expertise, and broaden the scope of their work. However, these directional shifts are associated with fewer citations, suggesting a trade-off between frontier-expanding innovation and impact. Our findings show that allocation strategies shape not just the volume but the trajectory of innovation, with direct implications for R&D managers and policymakers supporting innovation under resource constraints.

Abstract: In this paper, we examine how a startup’s knowledge foundations -- embedded in its core technology -- influence its performance in the exit market. Using a dataset of 1,006 biomedicine startups founded between 2005 and 2015, we focus on two key factors: (1) the degree of scientific specialization in the startup's core technology and (2) whether the technology's inventor is also the startup’s founder. Counterintuitively, we find that greater scientific specialization in a startup’s knowledge-base correlates with poorer exit market outcomes. Additional analyses suggest that this stems from such startups relying on narrower, less integrable technologies heavily dependent on tacit knowledge, which can hinder engagement with external stakeholders. However, the presence of an inventor-founder—an individual who invents the core technology and establishes the startup—moderates this relationship. When an inventor-founder is involved, the negative relationship with knowledge specialization is almost entirely mitigated. This suggests that inventor-founders may enhance the strategic value of specialized knowledge by making it more accessible to key stakeholders while also reinforcing its defensibility, counterbalancing its associated challenges. Interestingly, we also find that startups with an inventor-founder but without a specialized knowledge-base, or vice versa, perform worse on the exit market. These findings underscore the contingent value of knowledge-based resources in entrepreneurial contexts, emphasizing the importance of aligning knowledge characteristics with the founder to optimize firm outcomes. Our research highlights the nuanced relationship between knowledge specialization, founder roles, and startup performance, contributing to a deeper understanding of how knowledge-based resources shape firm success.