Jeffrey H. Peo is an alumnus of Trinity College, the University of Pennsylvania, and the University of Virginia. His startup, walkli, was part of the 2018 MassChallenge Boston cohort. More Much has been written in recent years about startup accelerators. The focus of most of this work has centered on the outcomes of participating firms or on the spillover effects an accelerator has on the local startup ecosystem1 Examples: Accelerating Entrepreneurs and Ecosystems: The Seed Accelerator Model, Yael V. Hochberg, Rice University, Massachusetts Institute of Technology, and NBER; Analysis of Accelerator Companies:An Exploratory Case Study of Their Programs, Processes, and Early Results, David Lynn Hoffman & Nina Radojevich-Kelley. How accelerators affect individual founders - and especially founders of ventures that went on to fail - is less well studied.
This paper will look at the career choices made by founders after their initial ventures failed. Specifically, we are interested in evaluating whether or not participating in an accelerator affects the likelihood of starting another new venture within 3 years of initial failure, which we define as resiliency.
The purpose of this outline is to propose a structured approach to analysis using a data set of companies that applied to and or participated in the MassChallenge accelerator program between 2010 and 2017.
Accelerators can take many forms, but there are several consistent themes that link their structures and differentiate them from other types of innovation vehicles, such as incubators. In general, accelerators are short-duration programs (typically 3-6 months) and participants are part of a cohort that share a physical workspace and complete educational programs together while building their early-stage companies. Most accelerators have a competitive application process and several invest in the startups they accept in exchange for equity. In addition, accelerators provide their participants with access to mentors, strategic partnerships, and outside investors through planned sessions and demo days. Since Y Combinator was established as the first accelerator in 2005, the model has been replicated and expanded upon3The pace of growth of accelerator programs has accelerated greatly during the past decade. By 2015 more than 30% of companies that had raised a Series A round had been through an accelerator program4One-thirst of U.S. startups that raised a Series A in 2015 went through an accelerator.
In theory, accelerators should have a positive effect on both the companies and entrepreneurs that participate in them. During the accelerator program, the skills, experiences, and networks that the entrepreneur develops should make them better equipped to start a new venture, even if the venture they entered the accelerator failed.
Hypotheses & Expectations
We are interested in evaluating the effect accelerators have on the resiliency of founders of a failed venture, where we define resiliency as starting a new venture within three years of the initial failed venture. This paper will focus on evaluating two hypothesis:
Hypothesis (H1) The proportion of founders of failed ventures who go on to start a new company will be different for founders who went through an accelerator and founders who did not.
Hypothesis (H2) Accelerators are not equally advantageous for producing resilient founders across role, gender, and industry.
For Hypothesis (H1) we do not expect the proportions to be the same for the two groups. If accelerators are successful at developing entrepreneurs, then we would expect the proportion of resilient founders to be higher in the accelerator group than the non-accelerator group. However, we also want to consider the scenario that participating in an accelerator leads to pessimistic views on entrepreneurship and as a consequence a lower proportion of resilient founders than the non-accelerator population. If our analysis finds this to be the case, then more work will be required to better understand why.
We also expect to find that accelerators do not have the same effect across populations of graduates of the program. We expect that technical founders benefit the most from accelerators, that women benefit more than men, and that companies that operate in the digital space benefit more than those that operate in the physical space, and that these differences in benefits will be reflected in the proportion of founders of failed ventures who go on to start a new company.
Data & Analysis Plan
This project will leverage a dataset of companies that participated in the MassChallenge accelerator. The majority of analysis will rely on data from the 2016 and 2017 Boston programs. Across those two years, we have profiles of 256 companies that participated and 449 that had applied but were not selected after the final round of judging. The first step of analysis is to determine the current operating status of each company.
A subset of the MassChallenge data, cleaned
Startup.Name
Program
Primary.Industry
Secondary.Industry
Status
3DFortify
Boston 2016 Accelerator
General
Materials
Active
ADAY
Boston 2016 Accelerator
General
Labor Sourcing
Not Active
Adhesys Medical
Boston 2016 Accelerator
Healthcare / Life Sciences
Patient Care
Acquired
Analytical Space
Boston 2016 Accelerator
General
Telecommunications / Mobile
Active
Angiex Inc
Boston 2016 Accelerator
Healthcare / Life Sciences
Medical Therapies
Active
CareHood
Boston 2016 Accelerator
Healthcare / Life Sciences
Digital Health
Not Active
Catalyst for World Water
Boston 2016 Accelerator
Energy / Clean Tech
Water: Filtration / Accesiblity / Treatment
Not Active
CLEO
Boston 2016 Accelerator
High Tech
Film / Video / Photography
Active
The second step of analysis will require building our own database of founders of the failed ventures and coding their experience before and after their failed venture. Fields of interest include: gender, years of prior work experience, role at the failed venture, and three years of post-failure professional activity. Information on founders will be pulled from LinkedIn, Crunchbase, and Angelist.
To evaluate Hypothesis (H1) we are interested in the difference in proportion of founders of failed ventures that go on to start another company within three years in the population of founders who have gone through an accelerator program with those who have not.
\(H_0: p_1 - p_2 = 0\)
\(H_a: p_1 - p_2 \neq 0\)
We expect from our data set that \(n_1\hat{p}\), \(n_1(1- \hat{p})\), \(n_2\hat{p}\), and \(n_2(1- \hat{p})\) will all be great enough to assume that our distribution is normal (at least 10). We will set \(\alpha = 0.05\) and perform a two sample proportion test.
To evaluate Hypothesis (H2) we will use logistic regression with independent variables for \(gender\), \(role\), and \(industry\) and a binary dependent variable where:
We will use \(\alpha = 0.05\) to determine if any of the independent variables are significant indicators of likelihood to start a new company. We will evaluate this both for founders that went through the MassChallenge accelerator and those who did not and compare results.
Finally, we will perform some general exploratory data analysis to try and paint a picture of the types of new roles founders of failed ventures go on to take and comment on any observed differences between those who did and did not go through the MassChallenge accelerator.
Limitations & Further Study
This project is based on data from the MassChallenge accelerator program, located in Boston, MA. While we believe that it is a reasonable representation of accelerators in general, there may be nuances to the program that are unique. If the results of this study show significance, it would be worthwhile to replicate the work with other accelerators. Moreover, the approach outlined will only inform us if there is a difference in resiliency of entrepreneurs that go through and do not go through accelerators. Additional qualitative work would then be required to better understand any causal relationship.
