Before returning to the central the

Before returning to the central themes of this article, we wish to remark briefly on its general relevance to research on entrepreneurship. One of the perennial obstacles to empirical research on the transition to entrepreneurship is the difficulty of obtaining suitable data. Acquisition of appropriate data can be especially challenging if the researcher is interested in studying the formation of a particular type of organization, inwhich case entrepreneurial events may be sufficiently rare that even a large random sample would not contain enough information to support statistical inference. Compounding this problem, it is sometimes difficult even to specify the boundaries of the population—or risk set—of individuals who might reasonably be expected to become entrepreneurs of a particular type. The upshot of these difficulties is that much of the literature on entrepreneurship can be aptly criticized for sampling on the dependent variable, or drawing conclusions from samples that only include principals of actual start-up companies (Carroll and Mosakowski1987). The research design we have employed circumvents this problem. By limiting the focus to academic founders of biotechnology companies, we are able to identify the population of individuals who are “at risk” of becoming academic entrepreneurs. And by selecting all events, we escape the problem of small numbers. More generally, we believe that the useof case-cohort data structures should enable researchers to begin to assemble data sets that eschew the methodological shortcomings that have rendered inference questionable in many studies in the entrepreneurship literature.