Are accelerators akin to breweries or wineries? A Bayesian variance decomposition of accelerator and cohort effects

• 成果类型: Article
• 署名作者: Avnimelech, Gil; Dushnitsky, Gary; Ellsaesser, Florian; Fitza, Markus
• 署名单位: Ono Academic College; University of London; London Business School; Frankfurt School Finance & Management; University of Exeter
• 刊物名称: STRATEGIC MANAGEMENT JOURNAL
• ISSN/ISSBN: 0143-2095
• DOI: 10.1002/smj.3665
• 发表日期: 2025
• 页码: 534-579
• 关键词: accelerators; Bayesian hierarchical model; entrepreneurship; scale-ups; variance decomposition

摘要

Research SummaryThe literature on startup accelerators uncovers multiple factors associated with accelerators' advantages. Yet, we have a limited understanding of the relative magnitude of these factors. We ask: Are accelerators akin to breweries, where quality is mainly a function of the institution of origin (i.e., brewery for beer, accelerator for startups); or are they similar to wineries, where quality varies across cohorts (i.e., for a given winery, some vintages are of higher quality)? We explore this question using data from 1,350 tech-startups graduating from dozens of accelerators in a global technology hub. A Bayesian hierarchical variance decomposition approach is introduced to account for the highly-skewed zero-inflated distribution in startups' performance. We find that a notable fraction of startup performance is due to vintage; within-accelerator, cross-cohort variation.Managerial SummaryStartup accelerators (i.e., short-term programs designed to help startups grow) are highly popular, with dozens of accelerators operating around the globe. Our focus is on accelerator programs aimed at catapulting technology ventures towards high growth. We ask: Are accelerators akin to breweries, where quality is mainly a function of the institution of origin (i.e., brewery for beer, accelerator for startups); or are they similar to wineries, where quality also varies across cohorts (i.e., for a given winery, some vintages are of higher quality)? A Bayesian hierarchical variance decomposition approach isused to study data from a global technology hub, detailing the performance of hundreds of startups that graduated across multiple accelerators. We find that a significant portion of startup success is linked to cohort-specific factors within accelerators, highlighting the role of timing and dynamics of each accelerator cohort.