98685 August 2015 · Number 8 MIGRANT REMITTANCES, CAPITAL CONSTRAINTS AND NEW BUSINESS STARTS IN DEVELOPING COUNTRIES Marek Hanusch and Paul M. Vaaler1 According to the World Bank (2013), migrant on one entrepreneurial activity vital to private remittances to developing countries in 2013 sector-led growth in developing countries: new totaled US$ 414 billion, quadruple the business starts. Analyses of annual new approximately US$ 100 billion remitted to business starts in 47 developing countries developing countries in 2000. The sheer size and observed from 2002-2007 suggest that substantial growth of remittances to developing remittances have significantly positive effects, countries in the 2000s compel attention but only for countries in the lowest quintile of regarding how remittances are used, with past local capital access. Remittances matter as early- research summarized by Brown (2006) stage capital for business creation associated indicating that they generally finance household with private sector-led economic growth, but consumption. On the other hand, Yang’s (2011) such entrepreneurial use diminish with increase review of the same literature notes to even moderate levels of local capital access. circumstances when remittances may also support business investment. Woodruff and Our findings contribute to research on Zenteno (2007), for example, report evidence remittances and entrepreneurship in developing from survey data indicating that remittance countries. First, to the extent remittances help flows from the US through migrant networks finance development (Ratha, 2003), our study alleviate capital constraints for microenterprises guides understanding about when such finance in Mexico. is more likely to affect entrepreneurs leading business creation. Second, we take a cross- We reconcile these differing research country approach to investigating one specific perspectives by proposing that remittances to entrepreneurial process by which remittances developing countries generally finance likely affect economic growth and development, household purchases, but broaden financing a complement to other recent cross-country scope to entrepreneurial activities when local studies relating remittances to broader capital constraints are substantial. Our empirical macroeconomic indicators of the same trends study builds on this proposition. We ask how (e.g., Guiliano and Ruiz-Arranz, 2009). We constrained local capital access is before elaborate on these points below. remittances have a significantly positive impact 1 This MFM Practice Note was cleared by Seynabou Sakho, Practice Manager (GMFDR). Empirical Methodology substantial capital constraints, we follow Model Specification and tests. To test our methods previously used by Vaaler (2011), proposition that remittances boost new business starting with a statistical model explaining starts when developing countries have annual new business starts closely following his: New Business Starts ijt    1 Remittance sijt 1   2 Capital Access ijt 1   3 Remittance s * Capital Access ijt 1  a 1 Controls ijt 1  b1 Regions j c 1Years t   ijt a 5 b 5 c 5 In (1), the dependent variable, New Business Consistent with our research proposition, Starts is count of new businesses entered into Remittances should be positively related to New official registries of a migrant’s home country i Business Starts (β1 > 0) as should Capital Access (β2 (in geographic region j) during year t. This > 0) however scored. The interaction term, measure likely understates the actual number of Remittances*Capital Access, tests whether and new businesses created to the extent that they how quickly the impact of remittances on new are founded and operate in the country’s Business starts in the recipient country informal economy. diminishes with better local capital access. This interaction term should be negatively related to Key right-hand side terms in (1) vary by country New Business Starts (β3 < 0). i, region j, and year t. Remittances comprises total remittances in US dollars divided by the We include five additional country- and year- recipient country population. We again follow varying Controls with the expected impact on Vaaler (2011) by measuring Capital Access New Business Starts in parentheses: the natural different ways. Our principal measure, General log of home-country GDP (Log GDP) (+); the Capital Access varies from 0 (low capital access) natural log of home-country GDP per capita (Log to 10 (high capital access),2 and scores the ability Per Capita) (+); the percentage of home-country of entrepreneurs to gain access to financial GDP growth (GDP Growth) (+); the percentage capital in countries around the world. It is based of home-country GDP comprised of state- on the Milken Institute’s expert annual owned enterprises and government) (Percent assessment of several country components, State GDP) (-); and the US dollar value (in including an ‘alternative sources of capital’ from billions) of home-country inward foreign direct private placements of debt, credit cards, investment (Inward FDI) (+). Finally, we include personal savings, family and friends. An year (Year) and region (Region) dummies to alternative Venture Capital Access measure is also capture other unspecified effects on New a 0-10 score but is based exclusively on Business Starts tied to time and geographic assessment of the alternative sources of capital location. component. Another alternative Bank Capital Access measure varies from 1 (very difficult) to 7 Data and Sampling. Data for terms in (1) come (very easy) based on responses by business from different sources. Remittances data are from executives to an annual survey question from the World Bank’s Word Development Indicators the World Economic Forum’s Global as are data for Controls. General Capital Access Competitiveness Report asking how easy it is in a and Venture Capital Access data are from the country to obtain a bank loan with only a good Milken Institute (2002-2007), while Bank Capital business plan and no collateral. Access data are from the World Economic Forum 2 General Capital Access (and Venture Capital for 2002-2003 and a 0-10 scale after 2003. We re- Access described below) is measured on a 0-7 scale scale the 2002-2003 measures to the 0-10 scale. August 2015 · Number 8 · 2 (2002-2007). Our base sample using General New Business Starts is significant across Columns Capital Access comprises 195 observations from 2-5 at the 1% level. 47 developing countries for 2002-2007. Somewhat sparser data availability when using Together, these results support our research Venture Capital Access and Bank Capital Access proposition that remittances boost decreases the sample to 164 and 166 entrepreneurial activity in developing countries, observations respectively. Column 1 of Table 1 but the boost diminishes with better capital presents means and standard deviations for all access. Results in Columns 2-3 using the General terms in (1). Capital Access score, suggest a broad base for this support. Negative binomial estimates best Estimation Strategy. New Business Starts is a suited for count data of New Business Starts yield count variable exhibiting substantial dispersion, expected signs at significant levels for thus, we rely primarily on cross-sectional Remittances, Capital Access and their interaction negative binomial regression with robust in Column 2. Panel GMM estimates of logged standard errors. To reduce the possibility of New Business Starts in Column 3 are less efficient estimation bias related to omitted variables and than negative binomial estimates of unlogged or reverse causation, we also take the natural log counts, but also less vulnerable to bias given the of New Business Starts, add a lagged value of this lagged dependent variable and instruments that transformed dependent variable to the right- Hansen and Arellano-Bond tests suggest are hand side of (1), and then use a dynamic panel properly applied. Signs on all three terms in Generalized Method of Moments (GMM) Column 3 again exhibit the expected sign with estimator initially developed by Arellano and significance at the 1% level for Remittances and Bond (1991). This GMM panel estimator the Remittances*Capital Access interaction. generates plausibly exogenous instruments in both levels and differences in levels of the The positive impact of remittances on new lagged dependent variable and all other right- business starts diminishes quickly. Simulation hand side terms. Table 1 reports the number of results suggest that a one-standard deviation instruments generated, and diagnostic tests increase in Remittances from 0.120 ($120 per commonly used to verify that GMM panel person in the recipient country) to 0.325 ($325 estimation assumptions are met. per person) raises New Business Starts by nearly 5,000 annually for the developing country with Results the lowest General Capital Access score, Madagascar with a 2.36 score in 2005. When In Columns 2-5 of Table 1 we observe a positive General Capital Access scores exceed relationship between Remittances and New approximately 3.75, the cut-off for the lowest Business Starts across negative binomial and quintile of countries sampled, the positive panel GMM estimations and alternative Capital impact of Remittances on New Business Starts Access scoring measures, as expected. The loses significance, even at the 10% level. positive relationship is significant at the 1% level (except in Column 4). Capital Access, however Conclusion scored, also exhibits the expected positive relationship with New Business Starts across Research on remittances in developing countries Columns 2-5 but is significant at commonly- suggests that they generally finance household accepted levels only after negative binomial purchases, but that substantial capital estimation with General Capital Access in Column constraints in recipient countries may broaden 2 and Venture Capital Access in Column 4. The uses of remittances for entrepreneurial expected negative relationship between the activities. Consistent with that proposition, we interaction term, Remittances*Capital Access and showed that remittances boost new business starts in developing countries with substantial August 2015 · Number 8 · 3 capital constraints. Migrant remittances can About the authors: play an important role as early-stage capital for Marek Hanusch, Economist, World Bank’s business creation associated with private sector- Macroeconomics & Fiscal Management Global led economic growth and development, but that Practice (GMFDR) role fades quickly as local capital access email: mhanusch@worldbank.org increases from the poorest to even moderate Paul M. Vaaler, Professor, Law School & levels. Our findings set the stage for future Carlson School of Management, University of research investigating other new roles for Minnesota remittances as the economic development process advances. TABLE 1 Results from Regression Analyses of New Business Starts on Remittances, Capital Access and Related Terms, 2002-2007 Descriptive Statistics, (1) (2) (3) (4) (5) Estimators and Capital Means Negative Sys-Diff Negative Negative Access Measure→ (Std. Dev.) Binomial GMM Binomial Binomial Variables ↓ GCA GCA VCA BCA 34,993 New Business Starts (Yijt) (88,342) 8.982 0.952** Log New Business Starts (Yijt-1) (1.801) (0.073) 0.120 4.057** 5.426** 1.202 3.993** Remittances (β1) (0.215) (1.493) (0.810) (0.751) (1.578) 4.509 0.145† 0.085 0.100* 0.130 Capital Access (β2) (0.925) (0.087) (0.137) (0.046) (0.121) Remittances* Capital Access 0.560 -0.910** -1.100** -0.866** -1.897** (β3) (1.082) (0.282) (0.167) (0.257) (0.560) 24.310 0.754** 0.147 0.681** 0.714** Log GDP (γ1) (1.550) (0.060) (0.098) (0.062) (0.065) 7.441 -0.056 -0.118 -0.089 0.103 Log Per Capita (γ2) (1.030) (0.091) (0.185) (0.084) (0.096) 5.103 0.015 -0.003 -0.022 -0.025 GDP Growth (γ3) (3.212) (0.021) (0.022) (0.026) (0.022) 14.359 0.020 -0.019 0.009 0.022 Percent State GDP (γ4) (4.277) (0.016) (0.039) (0.015) (0.070) 2.560 0.074** -0.008 0.075** 0.067** Inward FDI (γ5) (4.457) (0.019) (0.009) (0.018) (0.020) -9.715** -2.296 -7.055* -9.285** Constant (α) (1.471) (1.514) (1.369) (1.683) Year (ϕ1-6) Yes Yes Yes Yes Region (χ1-5) Yes No Yes Yes Instruments Generated 48 Hansen 2 Test 23.80 (p < 0.69) Arellano-Bond AR(2) Z Test 1.34 (p < 0.18) N 195 195 185 164 166 Wald 2 (R2) 1069** 5870** 1108** 879** Means and standard deviations appear in Column 1. Coefficient estimates and robust standard errors appear in Columns 2-5. The dependent variable in Columns 2 and 4-5 is New Business Starts (mean = 37,539, std dev = 93,115). The estimator is negative binomial regression. The dependent variable in Columns 3 is the natural log of New Business Starts (mean = 9.057, std dev = 1.798) The estimator is panel dynamic system and difference Generalized Method of Moments regression. Capital Access descriptive statistics above are for General Capital Access (GCA). Results where Capital Access is Venture Capital Access (VCA) (mean = 3.125, std dev = 1.647) appear in Column 4. Results where Capital Access is Bank Capital Access (BCA) (mean = 2.940, std dev = 0.684) appear in Column 5. The base sample of 195 country-year observations from 2002-2007 (Columns 1-3) includes 47 countries grouped alphabetically by six geographic regions: East Asia and Pacific (4): Indonesia, Malaysia, Philippines and Thailand; Europe and Central Asia (9): Armenia, Croatia, Latvia, Lithuania, Moldova, Romania, Russia, Turkey and Ukraine; Latin America and Caribbean (14): Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Guatemala, Haiti, Jamaica, Mexico, Nicaragua and Peru; Middle East and North Africa (7): Egypt, Jordan, Lebanon, Morocco, Oman, Tunisia and Yeman; South Asia (4): Bangladesh, India, Pakistan and Sri Lanka; and Sub-Saharan Africa (9): Botswana, Ghana, Kenya, Madagascar, Malawi, Senegal, South Africa, Tanzania and Uganda. Difference GMM regression results in Column 3 are based on a sub-sample of 185 country-year observations from 2002-2007 and 44 countries –dropping Ecuador, Egypt and Senegal. Negative binomial regression results in Column 4 (Column 5) are based on a sub-sample of 164 (166) country-year observations from 2002-2007 and 42 countries –dropping Botswana, Haiti, Madagascar, Senegal and Yemen. Significance: † p < 0.10, * p < 0.05, ** p < 0.01. ___________________________________________________________________________________________________________________________ This note series is intended to summarize good practices and key policy findings on MFM-related topics. The view expressed in the notes are those of the authors and do not necessarily reflect those of the World Bank, its board or its member countries. Copies of these notes series are available on the MFM Web site (http://worldbank.org/macroeconomics) August 2015 · Number 8 · 4