Since the foundation of Y Combinator in 2005 and Techstars in 2007, startup accelerators have rapidly emerged as a new form of organization in the startup ecosystem. In Korea, Primer was founded in 2010 as the first startup accelerator and was then followed by other early accelerators. The recent government certification of accelerators has since triggered the rapid spread of accelerators. However, despite this rapid diffusion of accelerators, there is a paucity of research. This paper seeks to offer a conceptual model as to how accelerators add value. It also provides suggestions for accelerator design, public policy and future research. Accelerators can mitigate institutional voids that hinder the creation and success of startups. They also assist startups to gaining access to human, social, and financial capital by offering education programs, mentoring service, networking opportunities, and demo days. In addition, mentoring helps startups to test key assumptions and receive feedback on their ideas and business models. In doing so, startups can evaluate their ideas and business models more objectively and from diverse perspectives, mitigating problems stemming from the overconfidence and inexperience of entrepreneurs. After a brief review of the history of accelerators, the paper delineates the sources of high failure rates of startups, explaining how accelerators add value by mitigating the difficulties associated with startups and how the characteristics of entrepreneurs, startups and ecosystems affect the value-added roles of accelerators. The paper concludes by offering implications for theory, practice, and public policy and by providing future research suggestions.