Dense urban environments increase entrepreneurial exit—the closure of new ventures. In this study, we argue that regional features that favor adaptation will offset the effect of urban densities on exits. Specifically, we focus on the effects of (1) knowledge spillovers and (2) regional industrial diversity.

The effect of new knowledge on entrepreneurial exit has received little scholarly attention. Building on the concept of knowledge spillovers, we argue that new venture adaptation is facilitated, and the pressure to close is eased, by a region’s stock of new knowledge. Greater knowledge affords new ventures the ability to modify existing products and processes in response to market shocks. Similarly, the level of industrial diversity helps reduce the chances of exit by presenting a venture the opportunity to apply their products and expertise in a variety of downstream industries (Rosenthal and Strange, 2003). Finally, because the level of local knowledge is often determined by the level of industry diversity, we expect the effect of knowledge spillovers on exit is moderated by the level of diversity in the region. We test our hypotheses using a panel dataset of 356 United States metropolitan statistical areas (MSA) covering an eight-year period from 2003 to 2010. Using US Census data, we define the local rate of entrepreneurial exit as the number of young firms (age 5 years or younger) that close in year t divided by the total number of firms in the region in year t-1.

We find support for our hypothesis that greater industry diversity reduces entrepreneurial exit, but, contrary to prediction, we find that more new knowledge increases entrepreneurial exit. We also find that the effect of knowledge spillovers on entrepreneurial exit is indeed moderated by industry diversity. Specifically, we find that knowledge spillovers reduce exit when industry diversity is low and increase exit when industry diversity is high. Indeed, we find that the exit rate is lowest when knowledge levels are high and industry diversity is low. In addition to practice, policy and future research, our study also informs the ongoing diversity vs. specialization debate in the regional science literature (Beaudry & Schiffauerova, 2009). In essence, the indications are that diversity (specialization) is best when knowledge levels are low (high).