WPS6382 Policy Research Working Paper 6382 Impact Evaluation Series No. 86 SME Registration Evidence from a Randomized Controlled Trial in Bangladesh Giacomo De Giorgi Aminur Rahman The World Bank International Finance Corporation Investment Climate Department Business Regulation Unit March 2013 Policy Research Working Paper 6382 Abstract Informality is pervasive in developing countries. In registration. They find that the treatment made firms Bangladesh, the majority of firms are informal and as more aware of the procedures, but had no impact on such they might not have access to prime markets, while actual registration. The results point toward potentially lowering the tax base. The authors implemented an low benefits and high indirect costs of registration as information campaign on registration, including both the the main barriers to formality (e.g. access to markets, step-by-step procedures and the potential benefits from taxation, labor and product regulations). This paper is a product of the Business Regulation Unit, Investment Climate Department, International Finance Corporation. 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 degiorgi@stanford.edu and arahman@ifc.org. The Impact Evaluation Series has been established in recognition of the importance of impact evaluation studies for World Bank operations and for development in general. The series serves as a vehicle for the dissemination of findings of those studies. Papers in this series are part of the Bank’s Policy Research Working Paper Series. 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 SME REGISTRATION EVIDENCE FROM A RANDOMIZED CONTROLLED TRIAL IN BANGLADESH∗ Giacomo De Giorgi† Aminur Rahman‡ ∗ We are grateful to PEDL initiative for the partial funding. Thanks to seminar participants at the SME Initiative workshop at the MIT and at World Bank Group workshops in Washington and Singapore. We thank Najy Benhassine, Andrei Mikhnev, Massimilano Santini for their comments, Syed Estem Dadul Islam, Farin Islam, Rotbah Nitia, Ismail Hossain, and Belayet Hossain for survey and data collection assistance, and Isaac Opper for excellent research assistance. Corresponding author: Giacomo De Giorgi (degiorgi@stanford.edu) Economics Department, Landau Economics Building, 579 Serra Mall, Stanford, CA, 94305-6072. † Stanford University, BREAD, CEPR and NBER ‡ World Bank Group 1 1 Introduction Firms’ informality is pervasive in developing countries. Often �rms are not registered with business registration or tax authorities. The 2010 census of 55,000 small and medium enterprises (SMEs) in 19 major district towns in Bangladesh suggests that about 70% of �rms are informal, according to either their business or tax registration status. It is widely known that informality decreases with the size and performance of �rms, which suggests that formality might cause better economic performance. Clearly this might not be the case; better performing �rms might simply decide to be formal. This still leads to the question of why so many �rms remain informal. A 14-country survey of informal �rms by the World Bank suggests that the lack of information about the registration process and the time it takes to register a business are two leading causes for �rms to operate informally. A 2008 International Finance Corporation (IFC) survey of SMEs in Bangladesh suggests that while the majority of the businesses believe that it is better to register their business with the registration authority, named the Registrar of Joint Stock Companies and Firms (RJSC), and operate formally, the cost to register a business, the corruption and the complicated registration processes keep many from becoming formal. In contrast to these survey �ndings, however, there is a growing body of evidence suggesting that although easier, less costly, and speedier registration procedures cause a modest increase in registration, direct incentives are more effective (McKenzie and Sakho, 2010; Bruhn, 2011; Kaplan, Piedra, and Seira, 2011; de Mel, McKenzie, and Woodruff, 2012). Building on a major business registration reform in Bangladesh that substantially reduces the time, complexity, and hidden costs of registering a business, our intervention aims to encourage SME registration through a targeted information campaign. The information campaign is meant to raise awareness of the potential bene�ts of registration and clarify the registration procedures implemented with the recent registration reforms. There are clear legal bene�ts of registration, starting from name protection to easier access to legal authorities in case of controversies. In terms of potential economic bene�ts, one can imagine that registered �rms have better access to markets, both �nancial and credit ones, can stipulate formal contracts and be better protected against fraudulent behavior of third parties. On the other hand, one needs to consider that associated with registration there are both direct and indirect costs. Direct costs are often small as in our current case post-reform, while indirect costs can be substantial due to the higher monitoring by the relevant institutions, including the tax authority, which would inevitably raise the (expected) tax burden on the registered �rm. If such indirect costs are large compared to direct costs and potential bene�ts, a reduction in the direct costs of registration will have a negligible effect on actual registration. At the same time the potential economic bene�ts of formality might be small or too uncertain. One needs to consider that the low performing �rms might disappear if registration were to be mandatory and severely enforced, however losing the left tail of the distribution of �rms in terms of pro�tability might be a positive outcome in the longer term as the more productive, and larger �rms, already registered or at the margin of such choice would not have to face “unfair� competition from informal �rms with low productivity. Another important point to take into account is that the government also loses a large share of tax revenues due to unregistered economic activity. Our intervention aims at understanding whether the lack of information on the process and the potential bene�ts of registration constitute a signi�cant barrier to formalization. We therefore implement a randomized controlled trial in Bangladesh where we provide face-to-face information to randomly chosen �rms. The intervention consists of one visit by a facilitator to the informal �rms, on which we have previously collected baseline information. Our analysis focuses �rst on whether the information campaign was somewhat successful in terms of exits from informality; if such a �rst stage appears signi�cant, then one can investigate the causal relationship between registration and economic performance. Unfortunately, our results are discouraging given that our information campaign had a positive and signi�cant effect on the knowledge of the registration procedures but no effect (both economic and statistically) on actual business registration. Very few �rms, below 5%, register both in the treatment and control group, and we estimate a “precise� zero effect. Although disappointing, we take this result as consistent with previous evidence and once again suggestive that the actual barriers to registration have little to do with the direct costs of emerging from informality and most likely have to do with the indirect costs associated with it, in particular in an environment where the density of small/low productivity �rms seems rather high, and/or with low bene�ts of formalization. The rest of the paper proceeds as follows: in Section 2 we sketch a conceptual framework on the decision to register; Section 3 describes the institutional context, while in Section 4 we present our intervention. Section 6 presents our empirical analysis; and �nally Section 7 concludes. 2 A Narrative on the Decision to Register A �rm’s decision to register rests upon the comparison of costs and bene�ts. On the one hand, registration can be bene�cial as it might allow better access to markets, both goods and �nancial/credit markets. For example, a registered �rm might be able to borrow from a formal lender at a lower cost of capital, access customers in distant markets, and write long-term contracts. On the other hand, there are both direct and indirect costs of registration. Direct costs may include registration fees (monetary contribution), time to register, and the cost of acquiring information about the registration process. Indirect costs could be the costs that the �rm has to incur once registered, for example compliance with the labor and goods regulations as well as taxation. We can then summarize the decision in an expected present value framework as follows: ΠR NR f − Πf > F EES + T IM E + IN F O + COM P LIAN CE + T AXAT ION . (1) Direct Costs Indirect Costs Firm f will register if the discounted present value of pro�ts in the registered state ΠR minus pro�ts in the unregistered state ΠN R is larger than the costs, both direct and indirect, e.g. monetary, time and information costs as well as taxation and compliance. Clearly, the difference in pro�ts between the two states accounts for the lower cost of capital, when registered, and the ability of a registered business to access additional markets and a potentially different (and larger) set of players. If �rms are credit constrained, then the monetary F EES can contribute substantially to the inability to register. We should expect, other things equal, less constrained �rms to register; a similar argument would apply to time constraints, due to the size or business process of the �rm. However, we also need to consider that the availability of information may or may not be highly correlated with the degree of credit to which a �rm has access. The intervention we implement in this work aims mainly at increasing the amount of information on registration available to the �rm. At the same time, one can think, that by informing �rms’ owners about the process, the information also modi�es the expectations regarding all the components of costs on the right-hand-side of the inequality. There are other reasons why the �rst three terms on the right-hand-side of the inequality might be affected by the intervention; in particular F EES are going to be reduced because an intermediary will not be needed, and better knowledge of the actual registration process would also reduce the time it would take to register (T IM E ). Notice that the registration reform implemented by the Bangladeshi authority, discussed below, is orthogonal to our intervention. The effects of such intervention need not be homoge- neous in the population of �rms. Indeed, one can conjecture that small-scale production �rms will probably not register as the �xed cost of registration will outweigh the bene�ts, however bigger (unregistered) �rms oriented toward bigger markets might be affected positively by the intervention. It is however an empirical issue as there are a large number of conflicting forces at play. 3 Context: The Business Registration Reform in Bangladesh In 2009-2010, Bangladesh implemented a major information and communication technology (ICT) based business registration reform initiative.1 Prior to this reform, registering a business required on average 42 days (against the international best practice of 1 day) to complete an 18 steps process of registration. This would require on average 4-5 visits to the agency, sometimes 10 visits, and a 4-5 week payment process of registration fees through a stamp duty 1 For the details of the reform see World Bank Group (2013). procedure. This latter step often involved arti�cial stamp shortages and stamp price hikes by the officials, and stamp forgery (causing government revenue leakages). Further impediments included hassle and harassment by officials and the need for middlemen at every step to navigate the bureaucratic maze of the registration process and side payments and bribes. Due to the ICT led reform, the registration time has been reduced to 1 day, and one �nal visit to pick up the registration certi�cate. The payment of fees can be completed in 15 minutes through leveraging one of the country’s leading bank networks (BRAC bank). The applicant can access a transparent online system in which a business can check its status without the need to visit the agency. Overall, it seems safe to assume a reduced scope of corruption due to limited interactions with the officials. 4 Intervention 4.1 Information campaign We implemented an information experiment around this business registration reform to test if a simpli�ed registration regime would encourage informal �rms to register and become formal once they learn how to register and about the potential bene�ts of registration. We extracted a sample of informal �rms (3,000) from two waves of the IFCs quarterly Business Con�dence Surveys, BCS (Q1 and Q2, 2009) and IFCs Informality Survey of 2010.2 These are all representative surveys of the businesses operating in Bangladesh. The data from these surveys form our baseline sample with information regarding busi- ness registration as well as several variables concerning business performance and composition. We then randomly assign these �rms to a treatment and control group. The assignment is implemented at the �rm level. We then had members of our staff visit the “treated� �rms and provide information on the process and potential bene�ts of registration. In practice the staff members presented a set of potential legal and economic bene�ts of the registration and the step-by-step registra- tion process. They were equipped with two booklets containing all the information in easy to understand (Bangla) language. The bene�ts booklet also contained virtual stories of differ- ent entrepreneurs, so that the treatment group could relate their personal situations to these �ctitious characters. The potential legal and economic bene�ts of the registration are as fol- lows: protection of business name and goodwill, greater access to bank loans, limited liabilities, continuity of businesses, better business con�dence, raising �nance, greater ownership rights, and enhancement of social status. The visits were conducted in two waves, according to the original spacing of the BCS and informality surveys. The �rst set of �rms were visited in March-June of 2010, while the remaining �rms were visited during January-February of 2011. We then conducted follow-ups (phone) interviews in April-July 2011 for the �rst batch of �rms 2 Details of the surveys are available from the authors. and May-July 2012 for the second one. In practice, this was done so that all the follow-up interviews occurred about one year after treatment. Notice that our intervention was orthogonal to the simpli�cation and informatization pro- cedure implemented by the RJSC, the new registration procedure was available to all �rms irrespective of their treatment status, and at baseline �rms indicated that they were unaware of this reform initiative. 5 Treatment and Control Selection The procedure to randomly select the receivers of the information visits on registration is based on pure random assignment at the �rm level, where conditional on the predicted value of an outcome index, i.e. pro�ts in this case, 50% of the subjects are randomized in and the remaining are randomized out. The procedure consists of running a regression to estimate pro�ts: Πi = f (xi ) + ui , where Πi are the recorded pro�ts of �rm i, the x s include a large series of input in a �rm production function, e.g. physical and human capital, labor, credit and so on. We then construct the predicted value of the pro�ts Πi , as a dimensionality reduction index (a single index rather than a very large set of x s), and based on that index we form pairs of comparable �rms and assign one to the treatment and one to the control groups. We repeat such assignment mechanisms several hundreds of times in order to minimize the sum of squares difference between each pair of treatment and controls in the predicted index. 6 Empirical Analysis 6.1 Registration Given the randomized controlled trial nature of our intervention the identi�cation of the pa- rameter of interests rests on the validity of the randomization procedure. We estimate both simple linear probability and probit models of the registration decision according to whether the �rm was assigned to a treatment or control group. The results on registration are somewhat discouraging, but consistent with de Mel, McKenzie, and Woodruff (2012), and summarized in Table 2. In fact the treatment effect is essentially zero; we �nd no evidence that providing detailed information on the registration procedures and on the potential bene�ts increases the proba- bility of registration. In particular the number of registered �rms, post treatment, is very low both in the treatment and control group and clearly not differential across the two groups. This result is robust to several speci�cations, for example adding baseline controls as in column 2-4 as well as to a probit model in columns 5 and 6 or using a propensity score matching method to deal with the sample attrition. We match �rms based on their baseline characteristics so that only matched treated/control pairs contribute to the estimation of the parameters in column 7 of the table. It is also interesting to note that although the treatment had no effect on actual registration behavior, we �nd that �rms that were treated, i.e. visited by the staff member, are more likely to know (self-declared) how to register (Table 3). When asked why they did not register, the treated �rms are about 6-9 percentage points less likely to declare that they do not know how to register. This is a 20 percent difference over the control �rms. At the same time, when asked about whether they were visited by a staff member who explained the new registration procedures, treated �rms are signi�cantly more likely to respond affirmatively to such a question as shown in Table 4. Ultimately, it appears that treated �rms indeed received the treatment but simply did not act upon it. These results lead us to believe that the impediments to registration are not due to the lack of information but rather to other constraints. In particular, given that very few �rms register both in the treatment and control group, it seems that the direct costs of registration are not the main issue. It seems plausible that the larger probability of taxation might be the binding constraint. If direct costs were the major constraints then both treatment and control �rms should register after the implementation of the RJSC reform as the direct costs went down dramatically. At the same time, it is also possible that the actual or perceived bene�ts of formalization are simply too low or uncertain for �rms to register. 7 Conclusions Given the prevalence of informality among �rms in less developed countries, we implemented a randomized controlled trial to investigate the effects of a face-to-face information campaign about the potential legal and economic bene�ts of registration, and a step-by-step demonstra- tion on how to register. Our treatment followed the informatization reform of the registration system of RJSC in Bangladesh. The reform replaced lengthy and costly registration procedures with online speedy procedures, causing the number of days required for registering a company to decrease from 42 to just one day. We randomly selected a large number of �rms to be visited by our team members. About one year after the visit, we re-interviewed the �rms. Although our treatment seems to have affected self-declared knowledge of the registration process, it did not affect registration behavior. We �nd no evidence that information constraints are the main barrier to registration for informal �rms. At the same time, given the overall low registration rate among treatment and control �rms, we believe that the main barriers to registration are due to the indirect costs and/or the low perceived bene�ts of registration. In particular, one needs to consider the higher taxes, and possibly stringent regulations, to which a registered �rm would be subject. References Bruhn, M. (2011): “License to Sell: The Effect of Business Registration Reform on En- trepreneurial Activity in Mexico,� The Review of Economics and Statistics, 93(1), 382–386. de Mel, S., D. McKenzie, and C. Woodruff (2012): “The demand for, and consequences of, formalization among informal �rms in Sri Lanka,� Policy Research Working Paper Series 5991, The World Bank. Kaplan, D. S., E. Piedra, and E. Seira (2011): “Entry regulation and business start-ups: Evidence from Mexico,� Journal of Public Economics, 95(11-12), 1501 – 1515. McKenzie, D., and Y. S. Sakho (2010): “Does it pay �rms to register for taxes? The impact of formality on �rm pro�tability,� Journal of Development Economics, 91(1), 15–24. World Bank Group (2013): Reforming Business Registration: A Toolkit for the Practition- ers. Investment Climate, Washington D.C. Table 1: Balance table (1) (2) (3) (4) (5) (6) Control Treated Number With Number of Firms Non-missing Standard With Non- Standard VARIABLES Values Mean Deviation missing Values Mean Deviation Year that Business Began 1,448 1995 14.41 1,444 1995 14.84 Number of Employees 1,453 26.19 80.89 1,448 22.39 74.05 Percent of Goods Sold Domestically 1,445 99.10** 7.040 1,438 99.58** 5.161 Monthly Revenue 1,440 2.271e+06 1.589e+07 1,427 2.054e+06 1.421e+07 Monthly Profit 1,390 107,564 541,972 1,391 103,997 528,668 Percent that have Invested 1,102 0.407 0.491 1,098 0.438 0.496 Percent in Services 1,453 0.407 0.491 1,448 0.395 0.489 Percent in Dhaka 1,453 0.380 0.486 1,448 0.363 0.481 Percent in Chittagong 1,453 0.191 0.393 1,448 0.193 0.395 Percent in Rajshahi 1,453 0.142 0.349 1,448 0.148 0.356 Percent that are Sole Proprietorships 1,453 0.818 0.386 1,448 0.829 0.376 Percent that are Partnerships 1,453 0.149 0.356 1,448 0.144 0.351 Percent that Import Directly 1,070 0.0327 0.178 1,066 0.0356 0.185 Number of Firms in the Informality survey 351 n/a 350 n/a *** p<0.01, ** p<0.05, * p<0.1 All variables come from BCS2, BC3, and Informality surveys and thus are the "Baseline" version of the question. "Monthly Profit" and "Monthly Revenue" are answers to the question "How much revenue did you make last month?" Table 2: The effect on registration (1) (2) (3) (4) (5) (6) Linear Linear Linear Propensity Regression Regression Regression Probit Probit Score Matching VARIABLES Registered? Registered? Registered? Registered? Registered? Registered? Treated? -0.00620 -0.00657 -0.0115 -0.00679 -0.00699 -0.00615 (0.00759) (0.00776) (0.00994) (0.00948) (0.0123) (0.00760) Number of Employees 0.000165** 0.000122 0.000156 0.000149 (7.65e-05) (8.84e-05) (0.000113) (0.000154) Is it in Dhaka? 0.0184* 0.00627 0.0256 0.00958 (0.00991) (0.0131) (0.0177) (0.0141) Is it in Chittagong? -0.00811 -0.0254 -0.00710 (0.0125) (0.0168) (0.0115) Is it in Rajshahi? -0.00446 -0.0147 -0.00388 -0.0128 (0.0123) (0.0149) (0.0132) (0.0129) Is it a Sole Proprietorship? -0.147*** -0.161*** -0.225 -0.360 (0.0484) (0.0515) (0.326) (0.400) Is it a Partnership? -0.135*** -0.156*** -0.0253 -0.0302 (0.0492) (0.0526) (0.0171) (0.0204) Percent of Goods Sold Domestically 0.000858 0.000834 (0.000949) (0.000997) Does the Firm Import? 0.0218 0.0317 (0.0265) (0.0546) Did it Invest from Oct - Dec '08? 0.0130 0.0171 (0.0100) (0.0150) BCS 3 0.0196** 0.0175* 0.0219** 0.0290* 0.0318* (0.00876) (0.00902) (0.00990) (0.0168) (0.0190) Informality 0.00447 0.0125 0.0268 (0.00937) (0.0105) (0.0258) Constant 0.0113 0.0608 0.0853 0.0196*** (0.00785) (0.107) (0.113) (0.00585) Observations 1,133 1,124 791 580 323 1,128 R-squared 0.005 0.105 0.132 0.001 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 There are "Year Founded" fixed effects in every column but (1). The probit results display marginal effects, evaluated at the means of the continuous variables. The propensity score matching used nearest neighbor matching with replacement. Columns (3) and (5) uses data from BCS2 and BCS3; the rest use data from BCS2, BCS3 and Informality. Table 3: The effect on awareness (1) (2) (3) (4) (5) (6) (7) Linear Linear Linear Propensity Score Regression Regression Regression Probit Probit Probit Matching VARIABLES Don't Know Don't Know Don't Know Don't Know Don't Know Don't Know Don't Know Treated? -0.0806*** -0.0293 -0.0859*** -0.0866*** -0.0318 -0.0914*** -0.0661** (0.0293) (0.0373) (0.0304) (0.0317) (0.0398) (0.0321) (0.0298) Number of Employees -0.000767** -0.000671** -0.00157** -0.00155** (0.000328) (0.000298) (0.000688) (0.000638) Percent of Goods Sold Domestically 0.00116 0.000941 0.00145 0.00158 (0.00377) (0.00373) (0.00415) (0.00416) Is it in Dhaka? -0.0746 -0.112*** -0.0806 -0.122*** (0.0488) (0.0387) (0.0517) (0.0400) Is it in Chittagong? -0.0133 -0.0424 -0.00970 -0.0453 (0.0623) (0.0485) (0.0665) (0.0506) Is it in Rajshahi? -0.0216 -0.0350 -0.0249 -0.0425 (0.0557) (0.0483) (0.0594) (0.0501) Is it a Sole Proprietorship? 0.316 0.338* 0.369** 0.365** (0.194) (0.190) (0.179) (0.159) Is it a Partnership? 0.302 0.306 0.389* 0.392* (0.199) (0.193) (0.209) (0.214) Does the Firm Import? -0.145 -0.174* (0.102) (0.102) Did it Invest from Oct - Dec '08? -0.0210 -0.0227 (0.0379) (0.0402) BCS3 -0.197*** -0.194*** -0.200*** -0.211*** -0.196*** -0.191*** (0.0338) (0.0372) (0.0351) (0.0337) (0.0393) (0.0355) Informality -0.200*** -0.242*** -0.200*** -0.242*** (0.0365) (0.0410) (0.0366) (0.0385) Constant 0.602*** 0.614 0.668 0.469*** (0.0303) (0.648) (0.641) (0.0229) Observations 1,143 801 1,134 1,115 776 1,110 1,138 R-squared 0.041 0.128 0.109 0.004 The dependent variable is equal to one iff the individuals report that they do not register either because they do not know how or do not know how to online. Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 There are "Year Founded" fixed effects in every column but (1), (4), and (7). The probit results display marginal effects, evaluated at the means of the continuous variables. The propensity score matching used nearest neighbor matching with replacement. Columns (3) and (6) uses data from BCS2 and BCS3; the rest use data from BCS2, BCS3 and Informality. Table 4: The effect on awareness (1) (2) (3) (4) (5) (6) (7) Linear Linear Linear Propensity Score Regression Regression Regression Probit Probit Probit Matching VARIABLES info1 info1 info1 info1 info1 info1 info1 Treated? 0.382*** 0.336*** 0.397*** 0.393*** 0.364*** 0.432*** 0.372*** (0.0289) (0.0357) (0.0296) (0.0277) (0.0341) (0.0289) (0.0298) Number of Employees 0.000199 9.21e-05 0.000317 0.000176 (0.000292) (0.000272) (0.000316) (0.000340) Percent of Goods Sold Domestically -0.00132 -0.00147 -0.00110 -0.00140 (0.00329) (0.00328) (0.00373) (0.00397) Is it in Dhaka? -0.108** -0.0863** -0.138*** -0.116*** (0.0471) (0.0379) (0.0520) (0.0436) Is it in Chittagong? -0.107* -0.160*** -0.116** -0.175*** (0.0600) (0.0472) (0.0578) (0.0449) Is it in Rajshahi? 0.0592 0.0659 0.0627 0.0789 (0.0535) (0.0469) (0.0641) (0.0586) Is it a Sole Proprietorship? 0.282 0.327 0.684*** 0.743*** (0.221) (0.219) (0.0788) (0.0886) Is it a Partnership? 0.227 0.330 0.944*** 0.943*** (0.225) (0.222) (0.0249) (0.0317) Does the Firm Import? -0.00696 -0.0192 (0.0961) (0.109) Did it Invest from Oct - Dec '08? 0.0702* 0.0719* (0.0361) (0.0428) BCS3 0.189*** 0.213*** 0.189*** 0.227*** 0.240*** 0.237*** (0.0331) (0.0354) (0.0338) (0.0397) (0.0398) (0.0429) Informality 0.256*** 0.228*** 0.309*** 0.295*** (0.0362) (0.0405) (0.0445) (0.0524) Constant -0.00596 -0.490 -0.579 0.139*** (0.0296) (0.583) (0.582) (0.0230) Observations 931 662 924 931 639 899 928 R-squared 0.194 0.272 0.277 0.144 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 There are "Year Founded" fixed effects in every column but (1), (4), and (7). The probit results display marginal effects, evaluated at the means of the continuous variables. The propensity score matching used nearest neighbor matching with replacement. Columns (3) and (6) uses data from BCS2 and BCS3; the rest use data from BCS2, BCS3 and Informality.