Supporting Entrepreneurs at the Local Level IN FOCUS The Effect of Accelerators and FINANCE, COMPETITIVENESS & Mentors on Early-Stage Firms INNOVATION Kathy Qian, Victor Mulas, and Matt Lerner FIRM CAPABILITIES & INNOVATION © 2018 The World Bank Group 1818 H Street NW Washington, DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org All rights reserved. This volume is a product of the staff of the World Bank Group. The World Bank Group refers to the member institutions of the World Bank Group: The World Bank (International Bank for Reconstruction and Development); International Finance Corporation (IFC); and Multilateral Investment Guarantee Agency (MIGA), which are separate and distinct legal entities each organized under its respective Articles of Agreement. We encourage use for educational and non- commercial purposes. The findings, interpretations, and conclusions expressed in this volume do not necessarily reflect the views of the Directors or Executive Directors of the respective institutions of the World Bank Group or the governments they represent. The World Bank Group does not guarantee the accuracy of the data included in this work. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. All queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2422; e-mail: pubrights@worldbank.org. Design & Layout: Aichin Lim Jones Photo Credits: IFC and World Bank Photo Librares Table of Contents ABSTRACT III INTRODUCTION 1 DATA 2 METHODOLOGY 5 FINDINGS 8 CONCLUSION 10 SURVEY AND DATA SPECIFICATIONS 11 REFERENCES 13 SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | I II | SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS Abstract W e investigate the association between entrepreneurship support programs and the likelihood of receiving funding for early-stage firms. We use a novel database of 2,887 early-stage technology companies from nine local ecosystems in eight countries that includes data about the founders’ demographic characteristics, educational background, work experience, and entrepreneurial history; we also use data about the start-ups’ history and evolution that follow their progress through support programs and early-stage funding. We isolate two support interventions—acceleration and mentorship—that the literature has found to have a larger effect on a firm’s performance, and we test if such effect is supported from an ecosystem perspective. After accounting for variations in founder characteristics and business environment, we find a positive association between acceleration and mentorship by experienced founders and the likelihood of receiving funding, whereas other support programs, such as incubation, are negatively correlated with funding. We also find that some founders’ characteristics, such as increased education and experience, have a positive correlation with funding. Authors are with the International Bank for Reconstruction and Development, World Bank Group. Data for this paper were collected through internal collaborations with Abdalwahab Khatib, Megha Mukim, Cecilia Paradi-Guilford, and Stefanie Ridenour. We are thankful for Paulo Guilherme Correa, Marcio Cruz, Philip Auerswald, and Erick Ramos Murillo for their guidance, as well as Justin Hill, Rodrigo Wag- ner, and Shinji Ayuha for their comments. We also thank Scott Henry, Nga Phuong Nguyen, Elene Allende Letona, and Hallie Applebaum for their contributions in the data collection and cleaning efforts and Domoina Rambeloarison for her contributions to the policy research review. We are grateful to our external partners at the Global Entrepreneurship Research Network for making their data available SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | III IV | SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS Introduction E ntrepreneurship and firm creation is an exciting topic for policy makers who are interested in economic development. Ample research documents the correlation of local entrepreneurship and economic growth (Delgado, Porter, and Stern 2010a, 2010b; Gennaiolo and others 2013; Glaeser and others 1992; Glaeser, Kerr, and Kerr 2012; Rosenthal and Strange 2003), and a large set of policies is applied to support entrepreneurship and to foster economic growth at local levels. Efforts to increase rates of entrepreneurship have ranged from long-term investments in human capital pipelines, available monetary capital, business-creation protocols, tax structures, or intellectual property laws to short-term support schemes for firms such as their providing growing technology-based ventures, one of their hands-on training, partnership opportunities, seed key objectives is to match start-ups with funding, funding, office space, or mentorship. thereby ensuring the resources that are needed to grow to the next stage. As technology-based start-ups have emerged as a global urban phenomenon (Florida 2013; The prevalence of those programs has generated Florida and King 2016), entrepreneurship support a growing set of literature that focuses on specific programs have become popular among investors, accelerators and mentor-support programs. The entrepreneurs, and policy makers. Those programs studies find that both programs influence early- include accelerators (fixed-term cohort-based stage firm performance—where measures of programs that last longer than two weeks and that performance are quantified metrics of funding, exit, provide mentorship, education, and funding to or acquisition (Gonzalez-Uribe and Leatherbee participating entrepreneurs),1 incubators (office 2018; Hallen, Cohen, and Bingham 2017; Hoffman space and administrative support services provided and Radojevich-Kelley 2012; Roberts and others for entrepreneurs), and other support schemes such 2016; Winston Smith and Hannigan 2015; Yu 2016). as mentor-matching events, executive retreats, or Fewer studies, however, have attempted a broader entrepreneurial boot camps. The programs aim to analysis that looks beyond firm-level effect and support firm creation, primarily in their early stage. focuses on the results of accelerators on the local Because many such programs are focused on fast- economy. Fehder and Hochberg 2014 find that start- Cohen and Hochberg (2014) define accelerator programs as a “fixed-term, cohort-based program, including mentorship and edu- 1 cational components, that culminates in a public pitch event or demo-day.” We largely follow this definition and take a flexible ap- proach in terms of duration of such programs; moreover, we consider these programs from durations starting at two weeks, given the diversity on the different maturity of stages of the ecosystems where data were collected. There have been several studies about the characteristics of accelerators, their taxonomy, and their difference with incubators and other support programs. See Miller and Bound (2011); Dempwolf, Auer, and D’Ippiloto (2014); Hoffman and Radojevich-Kelley (2012); and Hochberg (2016) for examples of such studies. SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | 1 up funding activity is exhibited in regions after an by people without founding experience in relation accelerator is established there. Bokhari and others to a firm’s performance. (2018), who investigate the effect of accelerators (as urban entrepreneurial amenities) on early-stage Our contribution to the existing literature is that first firms’ private equity performance, find evidence of we analyze support programs from the perspective of correlation between start-up firm performance and local ecosystems while using a new dataset. Second, accelerator program activities. we conduct this analysis with data from nine different ecosystems across the world, thereby testing results This paper takes the broader perspective and over various country conditions and ecosystems’ investigates the effect of support programs for maturity stage. Finally, we analyze a larger body of the local entrepreneurial ecosystem. Instead of support programs that are beyond accelerators. looking at a particular program or firm, we take the perspective of the entrepreneurial ecosystem The remainder of this paper is structured as follows: and look across a large sample of firms in a region first, we describe the novel ecosystem-level database to analyze which type of support program (for that we use; second, we present the methodology we example, accelerators, incubators, mentor matching, apply for our analysis; third, we review the findings; and entrepreneurial boot camps) is more effective and fourth, we explain conclusions for an early-stage firm’s performance for the overall ecosystem. Because of our emphasis on the local Data context, we follow the definition by Audretsch and Data from early-stage firms are challenging. Few Belitski (2016) of the entrepreneurial ecosystem as relevant databases of start-ups are readily available. “a dynamic community of inter-dependent actors The fast-paced and multidimensional dynamics of (entrepreneurs, supplies, buyer, government, etc.) start-up ecosystems—with new ventures constantly and system-level institutional informational and (a) being created, failing, and being closed and (b) socioeconomic contexts.” This analysis is possible being bought or transformed and thus changing because we use a novel data sample of early-stage names or purpose—make accurate measurement technology firms that has been collected over nine over time inherently difficult. ecosystems in eight countries through the Ecosystem Some databases collect information about start-ups Mapping Project of the Global Entrepreneurship globally. The most complete databases are Pitchbook Research Network (GERN). This database provides and CBInsights. Moreover, access to such databases an overall perspective of relevant firms, support programs, and investors in those ecosystems. is limited, and they are more representative of start- ups in the later stages of development and those To conduct our assessment, we isolate two support that operate in China, the United States, and other interventions—acceleration and mentorship—that developed countries. Other open databases lack the literature has found to have a larger effect on accuracy or are limited to their affiliations. For firm performance, and we test if such an effect is instance, Crunchbase is a self-reported database supported from the perspective of an ecosystem. with little or no curation; AngelList is constrained Furthermore, we expand on the questions posed by the demand and supply of angel investment by the literature and probe whether acceleration in this platform. None of those global databases, is different from other types of support programs, however, collect data at ecosystem level, nor do such as incubation, and whether mentorship by they provide a relevant dataset of firms that also experienced founders is different from mentorship depicts ecosystems of different stages of maturity, particularly those from developing countries.2 Pitchbook (database), Pitchbook, Seattle, WA, https://pitchbook.com/; CBInsights (database), CBInsights, New York, https:// 2 www.cbinsights.com/; Crunchbase (database), Crunchbase, San Francisco, https://www.crunchbase.com/; AngelList (database), AngelList, https://angel.co/. 2 | SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS To combat this poor data availability, we developed angels), and 869 institutional investors (venture a novel cross-section data sample of early-stage capital firms) gathered through surveys of 3,353 technology companies. This database collects data entrepreneurs in nine ecosystems: New York through a standard questionnaire from start-up City (United States), Cairo (Arab Republic of founders at ecosystem level. Endeavor Insight (which Egypt), Medellín (Colombia), Bogotá (Colombia), is a research firm) or the World Bank conducted Singapore, Santiago (Chile), Beirut (Lebanon), Dar the questionnaire surveys, and they are part of the es Salaam (Tanzania), and West Bank and Gaza GERN’s Ecosystem Mapping Project. The survey between 2013 and 2017 (table 1). questionnaire details can be found in the Survey and Data Specifications section in this report. Our database is different from databases of other early-stage firms in several aspects. First, our database Start-ups are defined as early-stage firms that provides a relevant sample of early-stage firms from are business ventures or social enterprises with a the perspective of the local ecosystem. Second, the financial sustainability model—even if they are too database provides data about founders’ entrepreneurial young to make money. The firms use an innovative characteristics and history, as well as about start-ups’ and technology-enabled approach to the product or history and evolution, including their progression service that the firms provide to ensure high growth through support programs and investment. Third, it and scalability. Those start-ups are primarily in the provides data about early-stage firms in ecosystems software, internet, and mobile application markets. that vary in maturity across geographies. Nongovernmental organizations, research projects, and nonprofits were excluded, but small and medium Our dataset includes seven variables of binary enterprises (SMEs) that were once start-ups but funding and four variables related to participation in have already begun scaling up were included. All support programs. In addition, we identify 11 binary start-ups were founded in 2009 and later. and continuous variables related to the founder characteristics, the firm-level characteristics, and Our dataset includes 2,887 unique start-ups, 68 the regional business environments that are used as accelerators, 247 incubators and other support controls in our regressions. All binary variables are programs, 717 individual investors (known as dummy variables that take on either a value of zero Table 1. Survey Details Survey Survey Number of Ecosystem Start End Survey Owner Responses New York City May 2013 November 2014 Endeavor Insight 643 Cairo December 2014 March 2015 Endeavor Insight 227 Medellín February 2015 September 2015 Endeavor Insight / World Bank 1,228a Bogotá February 2015 September 2015 Endeavor Insight / World Bank 1,228a Singapore March 2015 June 2015 Endeavor Insight 246 Santiago April 2015 June 2015 Endeavor Insight 147 Beirut February 2016 August 2016 World Bank 218 Dar es Salaam July 2016 September 2016 World Bank 221 West Bank and Gaza November 2016 February 2017 World Bank 423 a Survey administered by Endeavor Insight under a World Bank activity. Endeavor Insight collected 1,228 responses from Colombia. For this paper, the authors use only relevant data from Medellín and Bogotá. SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | 3 or one. Although our data sample includes 2,887 resulting in a maximum of 2,904 observations. unique start-ups, start-ups that operated in more than Table 2 presents the definition of variables. Table 3 one region were counted more than once, thereby breaks out means for each variable by region. Table 2. Definition of Variables Variable Category Definition Received funding (binary) a Funding Firm received angel or VC funding of any type, at any point during its existence, that was not provided as part of an acceleration or support program. Received funding from cross- Funding Firm received funding from an investor that has funded firms in more than ecosystem investor (binary) one regional ecosystem in our dataset. Received funding from more than Funding Firm received funding from more than one investor. 1 investor (binary) Received funding for 2 Funding Firm received funding for 2 years in a row at any point during its existence. consecutive years (binary) Received funding not from Funding Firm received funding from an investor that was not follow-on funding from an accelerator (binary)a accelerator it was accelerated by. Received funding from angel (binary) Funding Firm received funding from a person, broadly interpreted as received funding from an angel investor. Received funding from VC Funding Firm received funding from an organization, broadly interpreted as received (binary) funding from a VC firm. Accelerated (binary) Support Firm participated in an accelerator program that was cohort based, that was programs longer than 2 weeks, and that provided funding to some or all participating entrepreneurs. Participated in incubator or other Support Firm participated in an entrepreneurship support program of any time support program (binary) programs or format that did not fall within the definition of acceleration (definitional discretion left to respondent). Mentored by founder in dataset Support Founder of firm was mentored by founder of another firm in the dataset, (binary)b programs although not necessarily in the same regional ecosystem. Mentored by other (binary)b Support Founder of firm was mentored by someone who was not a founder in the programs dataset. Average years of work Founder Information derived by taking the difference of the firm founding date and experience per founder characteristics each founder’s earliest work experience start date, summing those figures and dividing by the number of founders. Has at least 1 founder with past Founder Firm has at least 1 founder who had founded a firm before founding the founding experience (binary) characteristics current firm. Has at least 1 founder with Founder Firm has at least 1 founder who had previously worked as a manager, managerial experience (binary) characteristics director, or C-level executive at a firm he or she did not found. Has at least 1 founder with Founder Firm has at least 1 founder who has received a professional, master’s, or postgraduate degree (binary) characteristics doctorate degree. No technical degree (binary) Founder Firm has no founders who studied computer science, engineering, characteristics information technology, mathematics, statistics, science, or medicine. No business degree (binary) Founder Firm has no founders who studied business or economics. characteristics 4 | SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS Average founder age Founder Information derived by summing founder ages and dividing by the number characteristics of founders. All founders are female (binary) Founder Firm has only female founders. characteristics Number of founders Firm-level Represents the number of founders associated with firm. characteristics Years of existence Firm-level Represents the total number of years the firm has been operational. characteristics Ease of Doing Business Business Represents conversion of the World Bank’s Ease of Doing Business overall percentilec environment ranking to integer percentile. Note: C-level = High-ranking officials of companies, including but not limited to Chief Executive Officer (CEO), Chief Financial Officer (CFO), Chief Information Officer (CIO), or Chief Operating Officer (COO; VC = venture capitalist. a. For the “Received funding (binary)” variable, we include follow-on funding by the accelerator. If Firm X was accelerated by Accelerator Y, received $50,000 as part of the acceleration program, and later received $250,000 from the accelerator after the acceleration program ended, we count the $250,000 from Accelerator Y as an instance of received investment. For the “Received investment not from accelera- tor (binary)” variable, we do not include the $250,000 from Accelerator Y as an instance of received investment. b. Ideally, we would be able to flag all mentors who were founders. However, because of the large number of mentors reported in the dataset across multiple regions, such identification was not possible to achieve accurately unless the mentors were founders in our dataset. Be- cause the unit of data collection was the ecosystem, we took great care to identify founders in the critical mass of start-ups at the center of a region, and, as such, this variable can be interpreted to be the effect of mentorship by an influential founder central to the ecosystem. c. We use a percentile transformation of the inverse of the 2017 Distance to Frontier metric from the World Bank Doing Business indicators (accessed March 2018), http://www.doingbusiness.org/. The “frontier” is the maximum aggregate Doing Business score across all coun- tries. For our regressions, higher percentiles mean that a country is closer to the frontier. The Doing Business indicators rank regulatory practices such as starting a business, getting electricity, registering property, getting credit, protecting minority investors, paying taxes, enforcing contracts, resolving insolvency, dealing with construction permits, and trading across borders. Methodology the right-hand side of equation corresponds to an explanatory variable (nonfunding variables). We use a logit model to test the seven funding We account for mean shifts in funding outcomes variables (dependent variables) against firm by using some specifications with regional and participation in support programs (acceleration, founding year fixed effects. Because of potential incubation, mentorship) and to account for variations correlations between errors at the ecosystem level, in the founder characteristics (experience, education, we also clustered standard errors by region. We run demographics), the firm-level characteristics a separate set of regressions for each of the seven (number of founders, years of existence), and outcome variables, and details on the specifications the regional business environment using seven can be found in table 4. regression specifications. Our choice of a logit model over a linear probability model is informed by P our hypothesis that the true probabilities for funding ln (1-P) = β0 + β1 X1 + β2 X2 + ... + βk Xk + ε can be extreme and that the effect of marginal interventions likely diminishes. Additionally, we There are some limitations to our approach. First, choose the logit over the probit model for potential because our dataset is based on a survey, it is easy transformation of log-odds into odds ratios subject to sampling biases, survivorship biases, where necessary. and errors inherent in self-reported datasets. Those biases are likely to inflate the likelihood of funding Our logit regression is represented in equation reported in the data. Second, findings may suffer below. The left-hand side of equation is the logit from endogeneity caused by the unclear causal transformation of the dependent variable, where relationship between the inputs and unobserved p is the probability that each dependent variable characteristics of the firm, and there are omitted (funding variables) is equal to one. Each X_n on variables in the analysis. SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | 5 Table 3. Means by Region Dar New West es York Bank and Variable Beirut Bogotá Cairo Salaam Medellín City Santiago Singapore Gaza Funding (binary) 0.39 0.44 0.24 0.25 0.18 0.75 0.43 0.41 0.32 Funding from cross- 0.01 0.13 0.08 0.02 0.13 0.15 0.07 0.16 0.04 ecosystem investor (binary) Funding from more than 1 0.08 0.02 0.08 0.01 0.05 0.51 0.11 0.26 0.02 investor (binary) Funding for 2 consecutive 0.06 0.02 0.02 0.00 0.02 0.33 0.06 0.08 0.06 years (binary) Funding not from 0.17 0.44 0.22 0.07 0.16 0.73 0.37 0.40 0.17 accelerator (binary) Funding from angel (binary) 0.10 0.40 0.17 0.06 0.11 0.34 0.09 0.30 0.14 Funding from VC (binary) 0.15 0.06 0.12 0.02 0.10 0.46 0.34 0.26 0.04 Accelerated (binary) 0.10 0.04 0.24 0.18 0.36 0.14 0.18 0.17 0.15 Participated in incubator or 0.06 0.06 0.12 0.02 0.31 0.14 0.18 0.22 0.30 other support program (binary) Mentored by founder in 0.05 0.02 0.44 0.22 0.25 0.19 0.14 0.29 0.26 dataset (binary) Mentored by other (binary) 0.24 0.10 0.49 0.43 0.44 0.19 0.21 0.32 0.59 Average years of work 5.40 1.30 3.40 2.60 3.40 4.90 2.00 4.40 3.70 experience per founder Has at least 1 founder with 0.41 0.12 0.44 0.21 0.44 0.29 0.34 0.23 0.36 past founding experience (binary) Has at least 1 founder with 0.42 0.29 0.36 0.15 0.43 0.48 0.27 0.42 0.38 managerial experience (binary) Has at least 1 founder with 0.50 0.13 0.13 0.12 0.22 0.32 0.12 0.16 0.26 postgraduate degree (binary) No technical degree (binary) 0.48 0.87 0.60 0.41 0.58 0.71 0.75 0.69 0.40 No business degree (binary) 0.71 0.87 0.81 0.82 0.67 0.69 0.86 0.73 0.68 Average founder age 30 28 29 24 29 31 30 30 28 All founders are female (binary) 0.08 0.04 0.02 0.10 0.05 0.07 0.00 0.08 0.15 Number of founders 1.80 1.40 2.10 1.70 2.00 2.00 1.60 1.80 1.80 Years of existence 2.90 2.80 2.10 2.10 2.70 2.50 1.90 2.50 2.30 Ease of Doing Business 31 70 37 29 70 97 72 100 31 percentile Note: VC = venture capitalist. 6 | SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS Table 4. Summary of Regression Specifications 1 2 3 4 5 6 7 Accelerated (binary) X X X X X X X Participated in incubator or other support program (binary) X X X X X X X Mentored by founder in dataset (binary) X X X X X X Mentored by other (binary) X X X X X X Average years of work experience per founder X X X X X At least 1 founder with past founding experience (binary) X X X X X At least 1 founder with managerial experience (binary) X X X X X At least 1 founder with postgraduate degree (binary) X X X X No technical degree (binary) X X X X No business degree (binary) X X X X Average founder age X X X All founders are female (binary) X X X Accelerated (binary) * participated in incubator or other support X X program (binary) Accelerated (binary) * mentored by founder in dataset (binary) X X Participated in incubator or other support program (binary)a X X mentored by founder in dataset (binary) Accelerated (binary) * mentored by other (binary) X X Participated in incubator or other support program (binary) * X X mentored by other (binary) Number of founders X X X X X X X Years of existence X X X X X X X Ease of Doing Business percentile X X X X X X Surveyed by Endeavor Insight (binary) X X X X X X Regional dummies X Founding year dummies X Note: “X” = where calculations were conducted. Details of the outcomes of these regressions can be accessed at http://bit.ly/2wShoxl For example, owing to lack of data availability, causal relationship between mentorship and funding we do not consider the quality of the team beyond and acceleration and mentorship is less clear and the founders or the quality of the business pitch. cannot be determined through this exercise. As such, Whereas acceleration generally occurs before we urge caution in interpreting the results, which we funding and it is likely reasonable to interpret this view as correlational and not causal. particular relationship as causal, the direction of the SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | 7 Nevertheless, we think that the analysis of this and funding (ranging from 5.0 to 8.6 percentage dataset is a unique contribution to the study of points), but we do not find a significant correlation entrepreneurship in light of the general difficulty for mentorship by mentors who were not known of obtaining firm-level data from early-stage to be founders. This finding suggests that practical companies with the depth and breadth necessary to operational experience may be more important compare outcomes across multiple programs. than technical knowledge, because mentors who are not founders but rather professors or corporate Findings professionals can impart the latter but not the former. Our analysis finds multiple statistically significant Third, by dissecting the data further through correlations between the funding and support interactions between acceleration and mentorship, programs, the founder characteristics, the we see that interactions take the sign of the firm characteristics, and the regional business mentorship variable, thus suggesting that the environment across a total of seven specifications effects of mentorship may be more dominant. for seven dependent variables. We find a positive interaction for firms that were both accelerated and mentored by an experienced Table 5 summarizes the average marginal effect of founder (ranging from 8.9 to 10.4 percentage each explanatory variable on the likelihood of funding. points), a negative interaction for firms that were The coefficient we present in each cell is the average accelerated and mentored by mentors not known to of every marginal effect that is significant at the p > be founders (ranging from –7.2 to –10.4 percentage 0.05 level for that combination of dependent and points), and a negative interaction for firms that independent variables. This approach means that each participated in nonacceleration support programs average marginal effect is the average of across up to and were mentored by mentors not known to be seven specifications. We report the marginal effect for founders (approximately –3.9 percentage points). each variable only if it is significant at the p > 0.05 We did not find a significant interaction effect for level for > = 50% of specifications it is entered in. firms that participated in nonacceleration support programs and were mentored by founders. We find several strong correlations between acceleration, mentorship, and funding for start-ups, even after Fourth, the magnitude of the marginal effect of controlling for the founder characteristics, the firm acceleration (ranging from 4.5 to 30.0 percentage characteristics, and the regional business environment. points) is on average greater than that of mentorship by an experienced founder (ranging from 5.0 to First, we find a large positive marginal effect 8.6 percentage points), indicating that acceleration of acceleration on funding (ranging from 4.5 to may have more effect on funding. However, in the 30.0 percentage points) but a negative effect for case of receiving funding from an angel investor, incubation and other nonacceleration support mentorship by an experienced founder has a greater programs (ranging from –7.16 to –13.3 percentage marginal effect than acceleration, suggesting that points). This finding can be interpreted in multiple angel investors may depend more heavily on personal ways. First, it is possible that those negative effects networks for sourcing potential investments or that are causal, although we cannot make this assertion the angels themselves are more likely to be mentors. within the limits of the current analysis. Second, it is also possible that this result is evidence that In addition to such findings related to acceleration accelerators are simply better at screening for the and mentorship, we also find statistically significant best start-ups, thereby leaving less stellar start-ups correlations for founder and firm characteristics. to attend other support programs. First, we find that increased education and Second, there is a positive and significant correlation experience has a positive and significant correlation between mentorship by an experienced founder with funding. Teams with at least one postgraduate 8 | DATA SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS Table 5. Summary of Marginal Effects Dependent Variables Funding Funding from Cross- from More Funding for 2 Funding Funding ecosystem than 1 Consecutive not from from Funding Funding Investor Investor Years Accelerator Angel from V Accelerated (binary) 0.181 0.045 0.136 0.103 0.101 0.071 0.300 Participated in incubator or other –0.133 –0.071 –0.129 –0.082 support program (binary) Accelerated (binary) * participated in incubator or other support program (binary) Mentored by founder in dataset 0.074 0.050 0.062 0.086 0.072 (binary) Mentored by other (binary) Accelerated (binary) * mentored by 0.089 0.104 founder in dataset (binary) Accelerated (binary) * –0.072 –0.104 mentored by other (binary) Participated in incubator or other — support program (binary) * mentored by founder in dataset (binary) Participated in incubator or –0.039 other support program (binary) * mentored by other (binary) Average years of work 0.004 0.009 experience per founder At least 1 founder with past 0.032 founding experience (binary) At least 1 founder with managerial 0.044 0.051 experience (binary) At least 1 founder with 0.129 0.091 0.067 0.128 0.120 postgraduate degree (binary) No technical degree (binary) 0.045 0.027 0.094 No business degree (binary) –0.044 Average founder age –0.004 All founders are female (binary) 0.036 Number of founders 0.094 0.015 0.066 0.030 0.101 0.043 0.044 Years of existence 0.097 0.037 0.094 0.067 0.101 0.049 0.123 Ease of Doing Business percentile 0.010 0.001 0.010 0.006 0.010 0.004 0.007 SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | 9 degree have a positive correlation with funding Conclusion (ranging from 6.7 to 12.9 percentage points). Similarly, teams with more senior experience are This paper is one of the first to look across multiple more likely to correlate with funding. We find the support programs and to analyze the role of largest marginal effects for founding teams with accelerators and mentors for funding of early-stage previous managerial experience (ranging from 4.4 to firms from the perspective of local ecosystems. On 5.1 percentage points), followed by founding teams the one hand, our findings suggest that acceleration with previous founding experience (approximately and mentorship by experienced founders likely do 3.2 percentage points), and finally founders with play a role in connecting early-stage technology any additional years of work experience (ranging companies to funding, both as independent and from 0.4 to 0.9 percentage point per year). combined interventions. On the other hand, we find that the influence of other support programs such We also find that each additional year of operation as incubators and mentorship by people who have for the business correlates positively with funding not been founders may not have an effect. Although (ranging from 3.7 to 12.3 percentage points per those findings do not provide clarity about the long- year). This finding is likely driven by a combination term effect on support programs beyond early-stage of effects, including the accumulation of experience firms, this evidence suggests that those interventions by the team. Given the same likelihood of raising do play some role in creating critical masses of funding each year, additional years will also mean funded start-ups. additional chances at a positive outcome, resulting in a higher cumulative effect. Additionally, founders Our findings about accelerators and mentorship are learning from attempts to secure funding, are consistent with other studies about such topics resulting in a higher likelihood of a positive outcome and support existing evidence between accelerators over the years. Finally, given the same firm quality, and mentors and early-stage firm performance— companies that have survived for a longer period understood as funding. Bokhari and others 2018 may look more attractive to investors. complement the findings by providing evidence of accelerators’ correlation with cumulative funding Second, we find positive and significant correlations and by suggesting that accelerators may also have a between team expertise and demographics and lasting effect over the long term. funding. We detect a positive correlation between all-business founding teams without technical Findings regarding the effect of higher education degrees and with funding (ranging from 2.7 to 9.4 and increased experience for founders are consistent percentage points) but a negative effect for all- with survey results from Wadhwa and others 2009. technical founding teams without business degrees Findings regarding the demographics of founders are (approximately –4.4 percentage points). We also also interesting in the context of existing literature. find a positive correlation between founding teams Robb and Coleman (2009) find that entrepreneurs’ with only women and funding for two consecutive university coursework and job history reveal that years (approximately 3.6 percentage points) and they are likely to be jacks-of-all-trades rather than between founding teams with additional founders specialists, which is consistent with our finding and funding (ranging from 1.5 to 10.1 percentage that all-business founding teams have a funding points per additional founder). advantage while a negative effect was found for all- technical founding teams. Furthermore, Robb and Finally, we find a small but significant effect of the Coleman (2009) find that women are less likely to regional business environment on successful funding receive funding, which is the opposite of our findings. outcomes (ranging from 0.1 to 1.0 percentage points for each percentile increase in the World Bank’s However, we are unable to make further statements Doing Business overall ranking). regarding causality or the mechanisms through which those correlations occur without additional research. 10 | METHODOLOGY SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS Are such results driven through a knowledge transfer are newly emerging, thereby encompassing also effect, in which accelerators and mentors impart SMEs that were once start-ups and that have reached practical knowledge to entrepreneurs? Is it through the scaling-up phase. This definition allowed us a network effect, in which accelerators and mentors to collect data that described the evolution of directly introduce promising entrepreneurs to valuable technology start-ups over time. connections regardless of any knowledge transfer? Or is it through a filtering effect, in which accelerators To identify the initial universe of technology start- and mentors are using rigorous acceptance criteria ups, we used multiple data sources, including and selecting more successful start-ups regardless databases from government, intermediaries (for of the incremental added value of the intervention example, accelerators), and founders (for example, itself? Further research will be needed to analyze the venture capital firms). Additionally, data sources causality of those presented correlations. such as Crunchbase, AngelList, and LinkedIn were used to accumulate initial lists of founders and Finally, we are continuing to expand the number of companies for outreach by contracted survey firms. ecosystems included in this dataset in the coming As the survey developed, the universe of technology years in partnership with the GERN. As the number start-ups was expanded through a snowball effect of regions covered reaches 20 to 30, we hope to be following the connections reported by founders. To better able to distinguish effects between regions increase outreach of the survey, we partnered with and to tease out interaction effects between regional multiple local intermediaries, founders, and public macroeconomic trends and firm-level data about and private entrepreneurship support organizations start-up creation. in each ecosystem. Nodes without location data and locations without Survey and Data Specifications geocode were passed to the Google Maps API to The standard questionnaire developed under obtain standard location data wherever possible. the Global Entrepreneurship Research Network This new dataset was de-duplicated using a process (GERN) Ecosystem Mapping Project includes the that marked similarities between names, email following categories of questions: (a) demographics addresses, URLs, and dates. Entities that were and educational history (including vocational, determined to be likely duplicates were then merged, boot camps, and certificate programs), (b) maintaining all existing data and privileging more employment history, (c) founding history (serial recent data in the event of conflict. For this process, entrepreneurship), (d) support programs history (for we used a combination of machine learning and example, acceleration, incubation, and so forth), manual methods. Education degrees and job titles (e) connections with mentors and mentees, and (f) provided by respondents were also classified into funding (angel and institutional). buckets using machine-learning methods. We then further reduced this cleaned dataset for analysis by Surveys followed a mixed distribution strategy that focusing on start-ups that met our narrow definition. included telephone, email, and in-person surveys. As such, the number of start-ups analyzed in this Entrepreneurs filled out an online survey or were paper may not be an exact match to other papers interviewed by Endeavor Insight, a survey firm, or written using the same datasets. The sample size a local partner. Technology start-ups are defined was selected in relation to the size of the sector in as for-profit business ventures that (a) have a the region, although the sample size in Santiago is financial model targeting high growth and (b) use likely somewhat low. The number of responses is an innovative and technology-enabled approach to not equal to the number of start-ups. the product or service that they provide to ensure scalability. So it could capture the whole technology Details on each individual survey structure and start-up ecosystem, the definition of start-up was collection methods can be found in Endeavor expanded beyond the phase in which such ventures Insight (2014) for New York City; Endeavor Insight SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | 11 and MC Egypt (2015) for Cairo; Endeavor Insight, Endeavor Insight, FOMIN, and J. P. Morgan (2016) FOMIN, and J. P. Morgan (2015) for Medellín; for Santiago; World Bank (2017a) for Dar es Endeavor Insight, MINTIC, and World Bank Salaam; World Bank (2017b) for Beirut; and World (2015) for Bogotá; Endeavor Insight and National Bank (2018) for West Bank and Gaza. University of Singapore (2015) for Singapore; 12 | SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS References Audretsch, David B., and Maksim Belitski. 2016. “Entrepreneurial Ecosystems in Cities: Establishing the Framework Conditions.” The Journal of Technology Transfer 42 (5): 1030–51. https://doi.org/10.1007/ s10961-016-9473-8. Bokhari, Sheharyar, Andrea Chegut, Dennis Frenchman, and Isabel Tausendschoen. Draft. 2018. “Is Innovation Really in a Place? Accelerator Program Impacts on Firm Performance.” MIT Center for Real Estate’s Real Estate Innovation Lab, Cambridge, MA. https://mitcre.mit.edu/wp-content/uploads/2018/03/ Is-Innovation-Really-in-a-Place_Accelerator-Impacts-on-Firm-Performance_030518.pdf. Cohen, Susan, and Yael V. Hochberg. 2014. “Accelerating Startups: The Seed Accelerator Phenomenon.” https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2418000 Delgado, Mercedes, Michael Porter, and Scott Stern. 2010a. “Clusters and Entrepreneurship.” Journal of Economic Geography 10 (4): 495–518. ———. 2010b. “Clusters, Convergence and Economic Performance.” U.S. Census Bureau Center for Economic Studies Working Paper 10-34, Washington, DC. Dempwolf, C. Scott, Jennifer Auer, and Michelle D’Ippolito. 2014. “Innovation Accelerators: Defining Characteristics among Startups Assistance Organizations.” Research paper, U.S. Small Business Administration, Washington, DC. https://www.sba.gov/sites/default/files/rs425-Innovation-Accelerators- Report-FINAL.pdf. Endeavor Insight. 2014. “The Power of Entrepreneur Networks: How New York City Became the Role Model for Other Urban Tech Hubs.” http://www.nyctechmap.com/nycTechReport.pdf. Endeavor Insight, FOMIN, and J. P. Morgan. 2015. “Creando el Ecosistema de Emprendimiento en el Sector Tech de Medellín: Tres Lecciones para el Crecimiento de un Sector de Alto Impacto en una Ciudad Innovadora.” http://www.medellintechmap.com/medellinReport.pdf. ———. 2016. “El Ecosistema del Sector Tech en Santiago: Una Evaluación del Ecosistema de Emprendimiento de la Tecnología y del Software en Santiago.” http://www.mapeosantiago.com/wp-content/uploads/2016/05/ EcosistemaDelSectorTechEnSantiago.pdf. Endeavor Insight and MC Egypt. 2015. “Network on the Nile: The Multiplier Effect at Work in Cairo’s Tech Sector.” http://www.cairotechmap.com/networkonthenile.pdf. Endeavor Insight, MINTIC, and World Bank. 2015. “Colombia Tech Map.” http://www.colombiatechmap. com/. SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | 13 Endeavor Insight and National University of Singapore. 2015. “Tech Founders in Asia’s Model City: Lessons from Mapping the Singapore Tech Ecosystem.” http://www.singaporetechmap.com/. Fehder, Daniel C., and Yael V. Hochberg. 2014. “Accelerators and the Regional Supply of Venture Capital.” http://dx.doi.org/10.2139/ssrn.2518668. Florida, Richard. 2013. “Why Today’s Start-Ups Are Choosing Urban Lofts over Suburban Office Parks.” CityLab, September 4. https://www.citylab.com/life/2013/09/why-todays-startups-are-choosing-urban-lofts- over-suburban-office-parks/6311/. Florida, Richard, and Karen M. King. 2016. “Rise of the Global Startup City: The Geography of Venture Capital Investment in Cities and Metros across the Globe.” Report, Martin Prosperity Institute, University of Toronto. http://martinprosperity.org/media/Rise-of-the-Global-Startup-City.pdf. Glaeser, Edward, Hedi Kallal, José Scheinkman, and Andrei Schleifer. 1992. “Growth in Cities.” Journal of Political Economy, 100 (6): 1126–52. Glaeser, Edward, Sari Kerr, and William Kerr. 2012. “Entrepreneurship and Urban Growth: An Empirical Assessment with Historical Mines.” National Bureau of Economic Research (NBER) Working Paper 18333, NBER, Cambridge, MA. Gonzalez-Uribe, Juanita, and Michael Leatherbee. 2018. “The Effects of Business Accelerators on Venture Performance: Evidence from Start-Up Chile.” Review of Financial Studies 31 (4): 1566–603. http:doi. org/10.1093/rfs/hhx103/4104437. Hallen, Benjamin L., Susan Cohen, and Christopher Bingham. 2017. “Do Accelerators Accelerate? If So, How? The Impact of Intensive Learning from Others on New Venture Development.” https://ssrn.com/ abstract=2719810. Hochberg, Yael. 2016. “Accelerating Entrepreneurs and Ecosystems: The Seed Accelerator Model.” In Innovation Policy and Economy, Vol. 16, edited by Josh Lerner and Scott Stern, 25–51. Chicago: University of Chicago Press. http://www.nber.org/chapters/c13584. Hoffman, David L., and Nina Radojevich-Kelley. 2012. “Analysis of Accelerator Companies: An Exploratory Case Study of Their Programs, Processes, and Early Results.” Small Business Institute Journal 8 (2): 54–70. https://www.sbij.org/index.php/SBIJ/article/download/136/93 Miller, Paul, and Kirsten Bound. 2011. “Startup Factories: The Rise of Accelerator Programs to Support New Technology Ventures.” Discussion paper, Nesta, London. https://media.nesta.org.uk/documents/the_startup_ factories_0.pdf. Robb, Alicia, and Susan Coleman. 2009. “Sources of Financing for New Technology Firms: A Comparison by Gender.” Kauffman Firm Survey Series report, Ewing Marion Kauffman Foundation, Kansas City, MO. http://www.kauffman.org/what-we-do/research/kauffman-firm-survey-series/sources-of-financing-for-new- technology-firms-a-comparison-by-gender. Roberts, Peter, Saurabh Lall, Ross Baird, Emily Eastman, Abigayle Davidson, and Amanda Jacobson. 2016. “What’s Working in Startup Acceleration: Insights from Fifteen Village Capital Programs.” Report, Global Accelerator Learning Initiative (GALI), Washington, DC. https://www.galidata.org/assets/report/pdf/GALI_ digital_041816.pdf. 14 | SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS Rosenthal, Stuart, and William Strange. 2003. “Geography, Industrial Organization and Agglomeration.” Review of Economics and Statistics 85 (2): 377–93. Wadhwa, Vivek, Krisztina Holly, Raj Aggarwal, and Alex Salkever. 2009. “Anatomy of an Entrepreneur: Family Background and Motivation.” Kauffman Foundation Small Research Projects Research, Ewing Marion Kauffman Foundation, Kansas City, MO. http://dx.doi.org/10.2139/ssrn.1431263. Winston Smith, Sheryl, Thomas Hannigan. 2015 “Swinging for the Fences: How Do Top Accelerators Impact the Trajectories of New Ventures?” Paper presented at DRUID15, Rome, June 15–17. http://conference. druid.dk/acc_papers/5ntuo6s1r5dvrpf032x24x5on5lq.pdf. World Bank. 2017a. “Tech Start-Up Ecosystem in Dar es Salaam: Findings and Recommendations.” Washington, DC: World Bank Group. http://documents.worldbank.org/curated/en/905101503998069531/ Tech-start-up-ecosystem-in-Dar-es-Salaam-findings-and-recommendations. ———. 2017b. “Tech Start-Up Ecosystem in Beirut: Findings and Recommendations.” Washington, DC: World Bank Group. http://documents.worldbank.org/curated/en/702081504876957236/Tech-start-up- ecosystem-in-Beirut-findings-and-recommendations. ———. 2018. “Tech Start-Up Ecosystem in West Bank and Gaza: Findings and Recommendations.” Washington, DC: World Bank Group.” http://documents.worldbank.org/curated/en/715581526049753145/ Tech-startup-ecosystem-in-West-Bank-and-Gaza-findings-and-recommendations. Yu, Sandy. 2016. “How Do Accelerators Impact High-Technology Ventures.” https://papers.ssrn.com/sol3/ papers.cfm?abstract_id=2503510. SUPPORTING ENTREPRENEURS AT THE LOCAL LEVEL: THE EFFECT OF ACCELERATORS AND MENTORS ON EARLY-STAGE FIRMS | 15