Signalling vs. Support: The Strategic Design of Start-up Accelerators

Abstract

Start-up accelerators play an important role in providing resources to new ventures through mentoring, education, peer interaction, certification, networking, and (in-kind) financing. Numerous studies have described different design choices of accelerators and empirically evaluated their effectiveness on start-up performance, e.g., revenue growth and funding success. However, the determinants and strategies shaping an accelerator’s design have been underexplored, as existing studies predominantly treat the design as exogenous. This paper addresses this gap by developing a formal model that conceptualises accelerator design as an endogenous, strategic response to environmental parameters such as the quality and number of applying ventures. The model considers an accelerator that selects a cohort from a pool of heterogeneous start-ups with unobservable quality and supports the selected start-ups through training. The accelerator can exert effort to improve the precision of noisy quality signals (screening) and allocate resources to post-entry support to selected ventures. Both activities incur costs and are constrained by a fixed budget. The accelerator maximises the expected post-program cohort quality, taking into account the trade-off between improving selection accuracy (which increases the perceived quality of start-ups through signalling) and directly supporting selected start-ups (which directly improves the quality of participants), yielding closed-form expressions for the optimal accelerator design. Solving the model for the Subgame Perfect Bayesian Equilibrium, we show that the optimal design critically depends on the entrepreneurial context, such as the share of high-quality start-ups and the selectiveness of the program (cohort size relative to applicant pool size). We find that when high-quality applicants are scarce, screening effort declines and more resources are allocated to post-entry support, as the returns to screening are relatively low. Conversely, the model shows that as program selectiveness increases, accelerators invest more in screening to extract maximum value from limited cohort slots. These comparative statics yield empirically testable predictions and offer practical implications for accelerator managers, suggesting that design should adapt dynamically to changes in the applicant pool and program maturity. The paper contributes to the accelerator literature by providing a strategic explanation for observed variation in program design and resulting effectiveness. Thus, it lays a foundation for future theoretical extensions and empirical validation aimed at refining our understanding of entrepreneurship support organisations.