WPS7636 Policy Research Working Paper 7636 Opportunity versus Necessity Understanding the Heterogeneity of Female Micro-Entrepreneurs Gabriela Calderon Leonardo Iacovone Laura Juarez Trade and Competitiveness Global Practice Group April 2016 Policy Research Working Paper 7636 Abstract Entrepreneurs that voluntarily choose to start a business have better performance and higher skills than necessity because they are able to identify a good business oppor- entrepreneurs. A discriminant analysis reveals that discrimi- tunity and act on it—opportunity entrepreneurs—might nation is difficult to achieve based on these observables, be different along various dimensions from those who are which suggests the existence of unobservables driving forced to become entrepreneurs because of lack of other both the decision to become an opportunity entrepre- alternatives—necessity entrepreneurs. To provide evidence neur and performance. Thus, an instrumental variables on these differences, this paper exploits a unique data set estimation is conducted, using state economic growth covering a wide array of characteristics, including cognitive in the year the business was set up as an instrument for skills, non-cognitive skills and managerial practices, for a opportunity, to confirm that opportunity entrepreneurs large sample of female entrepreneurs in Mexico. Descrip- have higher performance and better management practices. tive results show that on average opportunity entrepreneurs This paper is a product of the Trade and Competitiveness Global Practice Group. The paper has been prepared as a background research paper for the World Development Report 2016 “Digital Dividends.” It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at liacovone@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Opportunity versus Necessity: Understanding the Heterogeneity of Female Micro-Entrepreneurs∗ † ‡ Gabriela Calderon, Leonardo Iacovone, and Laura Juarez§ JEL Classification: D22, D24, L26, O12 Keywords: Entrepreneurship, Opportunity vs Necessity, Skills, Firm Performance ∗ We gratefully acknowledge funding from Instituto Nacional del Emprendedor (INADEM) at the Ministry of Econ- omy in Mexico, USAID, The World Bank and the Instituto Nacional de las Mujeres (INMUJERES). We thank Paulina L´opez and Sandra Aguilar for outstanding research assistance. The views expressed in this article are solely those of the authors and do not necessarily reflect those of Banco de Mexico or the World Bank. † Non-resident research fellow at Poverty and Governance, Stanford University. Email ad- dress:gabcal@alumni.stanford.edu ‡ The World Bank, 1818 H Street NW, Washington, DC. Email address: liacovone@worldbank.org § on General de Investigaci´ Direcci´ on Econ´ omica, Banco de M´ exico, Av. 5 de mayo 18, M´ exico, D.F. Email ad- dress:ljuarezg@banxico.org.mx 1 Introduction Support programs for micro-businesses have become increasingly common in developing coun- tries in recent years for at least two reasons: first, micro-enterprises employ a substantial fraction of individuals in these economies (about 47 percent in Mexico ); and second, despite their prevalence, the majority of these micro-enterprises tend to stay small and have low productivity. However, the impact of such programs —many of which provide business grants, training or a combination of both— has been mixed at best. This raises the question about whether these impacts depend on characteristics or attitudes of the entrepreneur, and therefore better targeting could significantly improve them. In fact, some recent evidence suggests that, even though the mean effects of busi- ness training might be small and not significant, greater returns are concentrated in high-potential entrepreneurs, who are the most likely to adopt better entrepreneurial practices and earn higher profits after training (Calderon et al. 2013, De Mel et al. 2012, Fafchamps et al. 2014). In this paper, we provide some novel evidence on the heterogeneity of female microentrepreneurs in urban Mexico by comparing those who started their business out of opportunity with those who did so out of necessity. Our data come from the baseline survey for the randomized evaluation of a large business training program for female entrepreneurs, funded by the National Institute of exico. This data set is Entrepreneur and implemented by the NGO CREA, Mujeres Moviendo M´ unique in that it provides detailed information on business outcomes, access to credit, cognitive and non-cognitive skills for a large sample of female micro-entrepreneurs in selected urban areas in Mexico. For the empirical analysis, we first show mean differences in business performance, character- istics and skills between the two groups of microentrepreneurs. Then, we use discriminant analysis to find the combination of observable characteristics that best distinguishes opportunity from ne- cessity entrepreneurs.1 Finally, we estimate the effect of being an opportunity entrepreneur on the profitability of the business and the quality of management practices, using state GDP growth at the time when the business was opened as an instrument for opportunity. Our results suggest that on average opportunity entrepreneurs have significantly higher profits, better management practices, and higher cognitive and selected non-cognitive skills. Our discrim- 1 Discriminant analysis is a technique of species classification that has been used before to study microentrepreneur- ship by de Mel, McKenzie and Woodruff (2010) for Sri Lanka; and by Bruhn (2013) for Mexico. 2 inant analysis shows that management practices, some business characteristics and skills of the entrepreneur can be used to separate both groups, but the distinction based on these observables is only partially successful, which suggests the existence of unobservables determining both the decision to set up a business, based on a perceived opportunity, and firm performance. Finally, our instrumental variables results confirm that opportunity entrepreneurs have higher performance measured as profits and management practices, even after controlling for the observable skills of the entrepreneur. The policy relevance of these findings is twofold. First, as mentioned before, identifying the observable characteristics that define opportunity entrepreneurs could potentially improve the tar- geting of business support programs, by concentrating them on the entrepreneurs with the highest growth potential. This by no means implies disregarding low-performing necessity entrepreneurs. On the contrary, the latter might be better served by interventions that help ease their transition to salaried employment. Second, our focus on female entrepreneurs is also of great relevance given that their businesses seem to be at a particular disadvantage when trying to scale-up. For instance, in Latin America, micro-firms led by women have been found to have an even smaller size and lower productivity, compared to those led by men (Bruhn, 2009). Thus, the evidence provided in this paper can be readily applied to those in more need of targeted support. 2 Data and descriptive analysis Our data come from a baseline survey conducted in 2014 in eight urban areas in Mexico: on, Mexico City, Quer´ Aguascalientes, Dolores Hidalgo, Irapuato, Le´ ıo and etaro, San Juan del R´ Toluca. For the sampling design, a number of census tracks (called AGEBs by the Mexican Sta- tistical Institute, INEGI) with relatively high concentration of commercial activity were chosen.2 The questionnaire was applied only to women who are business owners or partners, i.e. who take managerial, financial and marketing decisions, and the sample included firm with no more than 5 employees or sales that are less than 4 million Mexican pesos per year (which is the definition of a 2 These areas with higher concentration of commercial activity were identified relying on the 2009 Economic Census data. 3 micro-enterprise).3 Our survey has information on a uniquely broad range of variables: sociodemographic charac- teristics, current and initial business characteristics, business outcomes (sales, profits and costs), managerial practices, cognitive and non-cognitive skills of the entrepreneur, access to credit, growth expectations and obstacles, among others. The sample consists of 10,275 female micro-entrepreneurs. To our knowledge, our survey is the first one measuring both cognitive and non-cognitive skills, in addition to business characteristics and outcomes, for such a large sample of micro-entrepreneurs in Mexico. We classify the women in our sample as ”opportunity” or ”necessity” entrepreneurs according to the self-reported reason for opening their business. A respondent is classified as opportunity entrepreneur if she reports opening her business either because (i) she wanted to become inde- pendent, (ii) she had money and found a good business opportunity or (iii) she wanted to practice her profession or develop her career profile. Conversely, necessity entrepreneurs are those who started their business because they could not find a well-paid or suitable job and needed a source of income. After excluding women who started their business out of family tradition or other rea- sons, we are left with a sample of 8,949 entrepreneurs (21 percent classified as opportunity and 79 percent as necessity). In Table 1, we report the mean differences between opportunity and necessity entrepreneurs in the following characteristics: (i) Business performance measures (weekly profits, sales and sales per worker); (ii) business practices, as measured by a standardized Composite Business Practices score (CBP)4 ; (iii) characteristics of the business and the entrepreneur (age, proportion with at least one worker and monthly salary expenses); (iv) cognitive skills (education, standardized scores of Raven and digit span recall tests); and (v) non-cognitive skills (standardized measures of agree- ableness, conscientiousness, extraversion, intellect/ imagination and neuroticism5 , impulsiveness, locus of control, willingness to take risks, self-confidence, self-efficacy, optimism, self-satisfaction and trust). 3 Fixed and semi-fixed stands on the street, and dwellings in which the woman was found selling or producing something (or had a sign indicating so) also qualified as enterprises and were included in the survey. 4 The CBP score is an index that measures how well entrepreneurs in our sample manage their business and is constructed considering measures of marketing, keeping stock, record keeping and financial planning, following Fafchamps and Woodruff (2014). 5 These personality traits are referred to as ”the big five” in the literature. 4 Regarding business performance, Table 1 shows that mean weekly profits and sales are higher for opportunity than for necessity entrepreneurs and the difference is statistically significant. Mean weekly sales per worker are also higher for opportunity entrepreneurs, but the difference is not sta- tistically significant, probably because, as shown in the next panel, those entrepreneurs have a significantly larger number of workers. Opportunity entrepreneurs also have a significantly higher composite business practice score, compared to necessity ones. This descriptive evidence con- firms that opportunity entrepreneurs have both better business performance and are manage their businesses significantly better. Opportunity entrepreneurs are 4 years younger than necessity ones on average, and their businesses are about 13 months younger. About 32 percent of opportunity entrepreneurs have at least one worker or more, whereas only 23 percent of necessity ones do. As a result, opportunity entrepreneurs have about double the salary expenses per month compared to necessity ones. All these mean differences are statistically significant. The next panel in Table 1 shows that, compared to necessity entrepreneurs, opportunity ones also have significantly higher mean cognitive skills. They have significantly higher mean scores in the Raven and digit span recall tests, and 1.5 more years of schooling. The mean difference in schooling between groups corresponds roughly to that between having finished secondary school and proceeded to the first year of high school and not having finished secondary school. The last panel in Table 1 shows that opportunity entrepreneurs have statistically significant higher mean scores for locus of control, impulsiveness, self-confidence, self-satisfaction, willing- ness to take risks, optimism and their attitude towards business growth, compared to their ne- cessity counterparts, and the opposite is observed for extraversion, conscientiousness and self- efficacy. However, it is reasonable to think that at least three of those traits in which opportunity entrepreneurs score higher on average - namely willingness to take risks, optimism and their atti- tude towards business growth - could be correlated with better business outcomes. 3 Discriminant analysis To complement the descriptive analysis presented so far, we use discriminant analysis to ex- plore whether management practices, business and entrepreneur characteristics and skills can be used to distinguish opportunity from necessity entrepreneurs; and then use the estimates to predict 5 whether a given observation belongs to each group.6 Table 2 presents the results for our full sample of both necessity and opportunity entrepreneurs, and then for the opportunity group compared to the top and bottom profit quartiles of the necessity group. We vary the set of characteristics that are used to separate entrepreneurs into different species. For instance, in Panel A, for the full sample of entrepreneurs, using only measures of business performance would lead us to correctly classify 52 percent of entrepreneurs as opportunity and 63 percent as necessity. Using measures of business characteristics and skills seems to improve the classification of opportunity entrepreneurs, as measured with a larger proportion of correctly classified ones, but not necessarily so for necessity entrepreneurs. In the last row of Panel A, using all variables combined results in 60 and 63 percent of opportunity and necessity entrepreneurs to be correctly classified, respectively. This is about a 10 percentage point improvement from classifying entrepreneurs as opportunity or necessity over a random classification, i.e. one that would be obtained by simply flipping a coin, indicating that there might be other unobserved factors that explain that a women is a necessity or an opportunity entrepreneur. The last two columns in Panel A show that, when including all the explanatory variables to- gether, our model would classify 42 percent of the entrepreneurs as opportunity and 58 percent as necessity, whereas in our sample the proportions reported are 21 and 79 percent, respectively. This suggests that some entrepreneurs who report themselves as necessity ones, are in fact more similar to their opportunity counterparts, according to the discriminant analysis. Panels B and C show the results when including all the opportunity entrepreneurs in our sam- ple and only necessity entrepreneurs in the bottom and top quartiles of self-reported daily profits. The objective is to see whether high and low performing necessity entrepreneurs are in fact clas- sified as more or less similar to opportunity ones. For instance, Panel B shows that, when all variables are combined in the discriminant analysis, about 74 percent of low-performing necessity entrepreneurs are correctly classified as being so. In comparison, in Panel C, about 49 percent of high-performing entrepreneurs are correctly classified as being necessity ones. This suggests that some high-performing necessity entrepreneurs are more similar in the characteristics we use for the discriminant analysis to opportunity ones, as would be expected. The last column shows that, although in Panel B about 57 percent of entrepreneurs overall are classified as necessity, which is 6 For a description of this methodology, please refer to de Mel, McKenzie and Woodruff (2010). 6 not very different from the results for the overall sample in Panel A, in Panel C only 44 percent of high-performing entrepreneurs are classified as necessity ones, confirming that a higher proportion of high-performing necessity entrepreneurs in fact ”look like” their opportunity counterparts. In conclusion, the discriminant analysis suggests that necessity and opportunity entrepreneurs differ in key characteristics and abilities, but some among those starting their business out of necessity actually resemble their more able opportunity counterparts. Therefore classifying en- trepreneurs based on observable characteristics seems not to be very accurate. This suggests the existence of unobservable traits that drive both the decision to set up a business to pursue a good opportunity and the entrepreneur’s performance. Accordingly, in order to confirm our descriptive results suggesting that opportunity entrepreneurs have higher performance, we need to adopt an IV approach, which we discuss in the following section. 4 Instrumental Variables Approach Opportunity entrepreneurs appear to perform better than necessity ones. In order to estimate the partial effect of opening a business out of opportunity, controlling for the relevant observable characteristics of the firm and the entrepreneur, we estimate the following regression: yi = α + βopportunityi + ΘT Xi + ψi (1) Where yi is alternatively the log of self-reported weekly profits or CBP score observed in 2014. Our key independent variable is opportunity , a dummy indicating that the entrepreneur reported opening her business out of opportunity, defined as before, and Xi is a row vector that includes age of entrepreneur, Raven test score, span test score, years of schooling, locus of control, risk attitude, self satisfaction and optimism which were the observable characteristics of the entrepreneur and the firm that were statistical significant among all of the variables used in our descriptive analysis. Robust standard errors are estimated for this equation. Even after controlling for all the measures of skills that are available to us, opportunity might be endogenous in equation (1), because it might be correlated with unobservable characteristics of 7 the entrepreneur, like her social networks that contribute to have higher profits and are captured in ψi . To overcome this endogeneity, we use a two-stage least square (2SLS) estimation in which the first-stage equation is: opportunityi = α + βGDP growth t0s + ΓT Xi + i (2) where GDP growth t0s , the instrument, is the GDP growth observed in state s where the en- trepreneur i lives, in the year t0 she decides to open her business. Given that our survey measures profits and management performance in 2014, our assumption is that state GDP growth at the time when the business was set up is exogenous to profits (and management performance) various years later and only influences them through the choice of starting a business out of necessity or opportu- nity.7 When we use a 2SLS specification we include years of opening fixed effects in both stages, in this way, we compare firms that opened in the same year and have similar characteristics. Table 3 presents the estimation results. Panel A and B show the effect of the opportunity dummy on weekly profits and the CBP score, respectively. Panel C, shows the first stage of the 2SLS presented in column 2. Column 1 shows that when estimating equation (1) by OLS, the effect of opportunity on weekly profits is positive and significant at 1 percent in Panel A, and so is the effect of this same variable on the CBP score in Panel B. In column 2, controlling for the potential endogeneity of this variable yields a much larger positive coefficient for opportunity for both profits and the CBP score, significant at 10 percent for both. The estimate in Panel A implies that women who open their business out of opportunity in year t0 , induced by higher economic growth in their state in that same year, have 2.6 times higher weekly profits than necessity ones; and that part of this effect can be explained by a 2 times better CBP score, i.e. by better management practices. The partial F statistic shown in panel C has a value less than 10, but larger than 5, which means that the maximum IV bias we are tolerating is less than 20 percent (Staiger and Stock, 1997). The fact that the 2SLS estimates in column 2 are larger than the OLS ones in both panel A and B, might indicate that opportunity is measured with error. An alternative explanation is that our instrument is identifying the effect of opportunity for a specific subset of “compliers”, in our 7 Poschke (2013) presents a model in which in each period, individuals decide whether to work or to run a firm. Selection of agents into entrepreneurship comes from the heterogeneity of the relevant outside option and the hetero- geneity in firm’s performance. 8 case, women responding to become opportunity entrepreneurs when there is a higher state GDP growth in the year they opened their business or those who become necessity entrepreneurs when there is a low state GDP growth8 ; i.e. a local average treatment effect (LATE), which might not be representative of the average population of female entrepreneurs.9 The third column in Table 3 excludes the entrepreneurs who opened their businesses after 2011 from the estimation sample to address the concern that our instrument might directly affect profits observed in 2014, the year our survey was conducted, through a correlation with state GDP growth in that year. The estimate for opportunity in Panel A and the first-stage coefficient of our instrument in Panel C are very similar in magnitude to our estimates in column 2, which is reassuring, even though both of them lose statistical significance, probably due to the smaller number of observations. The effect of opportunity on the CBP score in column 3 is positive, larger than in column 2, and significant at 5 percent. Table 4 presents some robustness checks. Specifically, we use state GDP growth observed either the year before (t0 − 1, in column 1) or the year after (t0 + 1, in column 2) as instruments, instead of that observed during the same year (t0 ). In Panel C, the estimate for lagged state GDP growth is close to zero and insignificant, whereas the one for t0 +1 is positive, of similar magnitude compared to the one for t0 in Table 3, and significant at 1 percent. In column 2, Panels A and B show that the effects on profits and CBP score are positive, but only the latter is significant at 10 percent. Finally, column 3 includes the state GDP growth in t0 − 1, t0 and t0 + 1 together as instruments. In Panel A and B, the effects of opportunity on profits and the CBP score are positive and significant at 10 and 5 percent, respectively, confirming that becoming an opportunity entrepreneur, induced by our instruments, increases business profitability in part due to better management. In Panel C, only the state GDP growth in t0 and t0 + 1 have a positive and significant effect on the endogenous variable, but the first-stage partial F statistic for the three instruments is low (3.8). 8 Poschke(2013) argues that agents that have a low productivity in the labor market have lower performance as entrepreneurs. Therefore it is possible that this type of people are commonly laid off when there is low GDP growth and always have low entrepreneurial performance. 9 If performance and the instrument are related in a linear way then it is possible that our estimates are upward biased due to a LATE. See Angrist and Krueger (1994, 2001) for a discussion of both measurement error and the LATE interpretation of the IV estimate, among other issues. 9 5 Conclusions We provide new evidence suggesting that opportunity entrepreneurs have more profitable busi- nesses, manage them better and have higher measures of cognitive and of some non-cognitive skills, compared to necessity ones. Using a discriminant analysis, we show that combining these characteristics allows us to separate both types of entrepreneurs, although not perfectly, and iden- tify those that even though they started their business out of necessity are more similar in ob- servables to their opportunity counterparts. Given that the discriminant analysis only increases 10 percentage points of the prediction between necessity and opportunity entrepreneur compared to a random assignment, the characteristics of the firm and of the entrepreneur considered in the anal- ysis cannot fully explain when an entrepreneur is either an opportunity or a necessity one. In this way, we proceed to analyze this characteristic with an instrumental variable approach, which con- firms that businesses led by opportunity entrepreneurs are significantly more profitable than those led by necessity entrepreneurs, in part because of better management practices. These results are relevant for improving the targeting of business support programs. For instance, these programs could obtain larger impacts by concentrating on the entrepreneurs with the highest growth po- tential, whereas other types of interventions might help low-performing necessity entrepreneurs improve their prospects for salaried employment. References [1] Angrist, J. D. and A. B. Krueger (2001). “Instrumental Variables and the Search for Identifica- tion: From Supply and Demand to Natural Experiments.” Journal of Economic Perspectives, 13(2):69-85. [2] Angrist, J. D. and A. B. Krueger (1994). “Why Do World War II Veterans Earn More than Nonveterans?” Journal of Labor Economics, Vol 12, No.1 74-97. [3] Bruhn, M. (2009). “Female-owned Firms in Latin America: Characteristics, Performance, and Obstacles to Growth.” World Bank Policy Research Working Paper Series 5122. 10 [4] Bruhn, M. (2013). “A Tale of Two Species: Revisiting the Effect of Registration Reform on Informal business Owners in Mexico.” Journal of Development Economics, vol. 103(C), 275- 283.) [5] Calderon, G., J. M. Cunha and G. De Giorgi (2013). “Business Literacy and Development: Ev- idence From a Randomized Controlled Trial in Rural Mexico.” NBER Working Papers 19740. [6] De Mel, S., D. McKenzie and C. Woodruff (2010). “Who are the Microenterprise Owners? Evidence from Sri Lanka on Tokman versus De Soto,” NBER Chapters, in: International Dif- ferences in Entrepreneurship, National Bureau of Economic Research, 63-87. [7] De Mel, S., D. McKenzie and C. Woodruff (2012). “One-time Transfers of Cash or Capital have Long-lasting Effects on Microenterprises in Sri Lanka.” Science. vol. 335, no. 6071, 962- 966. [8] Fafchamps, M., and C. Woodruff (2014). “Identifying Gazelles: Expert Panels vs. Surveys as a Means to Identify Firms with Rapid Growth Potential” Competitive Advantage in the Global Economy (CAGE). mimeo. [9] Fafchamps, M., D. McKenzie, S. Quinn, and C. Woodruff (2014). “Microenterprise Growth and the Flypaper Effect: Evidence from a Randomized Experiment in Ghana. ” Journal of Development Economics, 106, 211-226. [10] Poschke, M. (2013).“Who Becomes an Entrepreneur? Labor Market Prospects and Occupa- tional Choice.” Journal of Economic Dynamics and Control, Elsevier, vol. 37(3), 693-710. [11] Staiger, D. and J. Stock (1997). “Instrumental Variables Regression with Weak Instruments.” Econometrica, Vol. 65, No.3, 557-586. 11 Table 1: Mean Difference between Opportunity and Necessity Entrepreneurs (full sample) Opportunity group mean Necessity group mean Difference Measures of performance Weekly profits (self reported) 1937.635 1349.177 588.4576∗∗∗ Weekly sales (self reported) 4507.258 3540.586 966.6726∗∗∗ Weekly sales/workers 4395.353 4302.254 93.09963 Composite Business Practice score (standarized) .3060325 -.0842768 .3903093∗∗∗ Business characteristics Age of entrepreneur 41.81658 45.92807 -4.111494∗∗∗ Age of business in months 90.59086 103.6411 -13.05026∗∗∗ Proportion with one worker or more .3185 .2278 .0907∗∗∗ Costs:monthly salary expenses 1025.357 498.751 526.6064∗∗∗ Cognitive skills Total score of raven test (standarized) .0827833 -.0211648 .1039482∗∗∗ Total score of digit span test (standarized) .1459293 -.0328984 .1788277∗∗∗ 12 Years of schooling 9.789416 8.251668 1.537748∗∗∗ Non-cognitive skills Extraversion (standarized) -.0365413 .0155396 -.0520809∗∗ Agreeableness (standarized) -.0093652 .0075027 -.0168679 Conscientousness (standarized) -.069074 .0121551 -.0812292∗∗∗ Neuroticism (standarized) -.0246642 .0130341 -.0376982 Intellect/imagination (standarized) .0074094 -.0046432 .0120526 Self efficacy (standarized) -.0957561 .0110521 -.1068082∗∗∗ Locus of control (standarized) .0954217 -.0202922 .1157139∗∗∗ Impulsiveness (standarized) .0480202 -.0170598 .06508∗∗ Self confidence (standarized) .1074716 -.0390423 .1465139∗∗∗ Attitude towards risk (standarized) .0975181 -.0298163 .1273345∗∗∗ Self satisfaction (standarized) .1163573 -.0443902 .1607475∗∗∗ Optimism (standarized) .138387 -.0400507 .1784377∗∗∗ Attitude towards trust (standarized) .0137271 -.0011845 .0149116 Attitude towards growth (standarized) .0621355 -.0119533 .0740888∗∗∗ Observations 8949 ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 Table 2: Discriminant Analysis Variable set used Opportunity Necessity Classified as Classified as in classification correctly classified (%) correctly classified (%) opportunity (%) necessity (%) Panel A: Full sample Measures of performance only (1) 52.37 62.85 40.37 59.63 Business characteristics only (2) 59.18 56.19 46.48 53.52 Cognitive skills only (3) 59.06 56.45 47.17 52.83 Non-cognitive skills only (4) 57.13 53.53 48.78 51.22 All variables combined 60 62.75 42.06 57.94 Panel B: Opportunity group and first quartile of self reported daily profits of Necessity group All variables combined 60.00 74.11 43.13 56.87 Panel C: Opportunity group and fourth quartile of self reported daily profits of Necessity group All variables combined 60.00 49.46 56.92 43.80 Measures of performance (1) include Composite Business Practice score; Business characteristics (2) include: age of entrepreneur, age of business, proportion with one worker or more and monthly salary expenses; Cognitive skills (3) include raven (standarized) test score, digit span (standarized) test score and years of schooling; Non-cognitive skills (4) include (standarized) measures of agreeableness, conscientousness, extraversion, in- tellect/imagination, neuroticism, impulsiveness, locus of control, risk attitude, self-efficacy, optimism, self-confidence, self-satisfaction, attitude towards trust and attitude towards growth. Table 3: OLS and 2SLS Estimations (1) (2) (3) OLS model 2SLS model 2SLS model Sample: year ≤ 2010 Panel A: Weekly profits (log (x + 1) transformation) Opportunity 0.269∗∗∗ 2.609∗ 2.452 [0.038] [1.332] [ 1.493] Panel B: Composite Business Practice score Opportunity 0.270∗∗∗ 2.067∗ 3.538∗∗ [0.036] [1.118] [1.783] Panel C: First stage Growth - 1.237∗∗∗ 1.236∗∗ [0.467] [0.542] F-statistic - 7.014 5.204 P rob > F - 0.008 0.023 Observations† 4549 4549 2728 Note: All specifications include as controls: age of entrepreneur, Raven test score, Span test score, years of schooling, locus of control, risk attitude, self satisfaction, optimism, and fixed-effects for years when the entrepreneurs opened their own businesses. Robust standard errors in brackets; ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 †Observations in top 1% of (transformed) weekly profits were not included. 13 Table 4: Robustness Checks (1) (2) (3) IV: Growtht−1 IV: Growtht+1 IV: All instruments Panel A: Weekly profits (log (x + 1) transformation) Opportunity 12.482 1.218 1.368∗ [29.381] [0.985] [0.806] Panel B: Composite Business Practice score Opportunity 7.584 1.785∗ 2.138∗∗ [17.986] [1.030] [0.881] Panel C: First stage Growth - - 1.183∗∗ [0.575] Growtht−1 0.200 - -0.286 [0.467] [0.494] Growtht+1 - 1.402∗∗∗ 1.097∗∗ [0.513] [0.553] F-statistic 0.184 7.476 3.856 P rob > F 0.668 0.006 0.009 Observations† 3531 3531 3531 Year of opening dummies Yes Yes Yes Robust standard errors in brackets; ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 †Observations in top 1% of weekly profits (log (x + 1) transformation) were not included. 14