83341 AUTHOR ACCEPTED MANUSCRIPT FINAL PUBLICATION INFORMATION Remittances and Financial Inclusion : Evidence from El Salvador The definitive version of the text was subsequently published in World Development, 54(Feb.2014), 2013-10-15 Published by Elsevier THE FINAL PUBLISHED VERSION OF THIS ARTICLE IS AVAILABLE ON THE PUBLISHER’S PLATFORM This Author Accepted Manuscript is copyrighted by the World Bank and published by Elsevier. It is posted here by agreement between them. Changes resulting from the publishing process—such as editing, corrections, structural formatting, and other quality control mechanisms—may not be reflected in this version of the text. You may download, copy, and distribute this Author Accepted Manuscript for noncommercial purposes. Your license is limited by the following restrictions: (1) You may use this Author Accepted Manuscript for noncommercial purposes only under a CC BY-NC-ND 3.0 Unported license http://creativecommons.org/licenses/by-nc-nd/3.0/. (2) The integrity of the work and identification of the author, copyright owner, and publisher must be preserved in any copy. (3) You must attribute this Author Accepted Manuscript in the following format: This is an Author Accepted Manuscript of an Article by Anzoategui, Diego; Demirgüç-Kunt, Asli; Martínez Pería, María Soledad Remittances and Financial Inclusion : Evidence from El Salvador © World Bank, published in the World Development54(Feb.2014) 2013-10-15 http://creativecommons.org/licenses/by-nc-nd/3.0/ © 2013 The World Bank REMITTANCES AND FINANCIAL INCLUSION: EVIDENCE FROM EL SALVADOR Diego Anzoategui, Asli Demirgüç-Kunt, and María Soledad Martínez Pería ∗ Abstract: This paper investigates the impact of remittances on financial inclusion. Using household-level survey data for El Salvador, we examine whether remittances affect households’ use of savings and credit instruments from formal financial institutions. We find that although remittances have a positive impact on financial inclusion by promoting the use of deposit accounts, they do not have a significant and robust effect on the demand for and use of credit from formal institutions. If anything, by relaxing credit constraints, remittances might reduce the need for external financing from financial institutions, while at the same time increasing the demand for savings instruments. JEL classification: F37, G21, 016 Keywords: remittances, financial inclusion ∗ Diego Anzoategui is a Ph.D. student at New York University and a consultant in the Finance and Private Sector Development Research Group of The World Bank. Asli Demirgüç-Kunt is Director of the Development Research Department in the Development Economics Vice Presidency of The World Bank. María Soledad Martínez Pería is Research Manager of the Finance and Private Sector Development Research Group of The World Bank. We thank David McKenzie for helpful comments and discussions. Florencia Moizeszowicz and Noemi Soledad Lopez provided excellent research assistance. The views and opinions expressed in the paper are those of the authors and do not reflect those of the World Bank or its Executive Directors. Corresponding author: María Soledad Martínez Pería, Development Research Group, The World Bank, 1818 H St., N.W., MSN MC - 3-307, Washington, DC 20433. mmartinezperia@worldbank.org. I. Introduction Remittances have become a significant source of external financing for developing countries. They reached US$401 billion in 2012, more than three times the amount of official development assistance and over one half of the private capital flows received by developing countries in that year. Remittances are especially significant for small developing countries neighboring large rich economies. For example, remittances account for approximately 16.4 percent of GDP in the case of El Salvador and they represent the second most important source of external flows after exports. There is an extensive literature on the effects of remittances on growth, investment in microenterprises, poverty, inequality, health, and education 1. However, the literature has so far ignored the impact of international remittances on financial inclusion - households’ access to and use of financial services. This question is important because there is a growing body of research showing that promoting financial inclusion can have significant beneficial effects for households. Based on a randomized evaluation of a microcredit program in India, Banerjee et al. (2010) find that access to microcredit leads to greater investment in business durables, increases the number of businesses started, and improves the profitability of existing ones. Karlan and Zinman (2010) conduct a field experiment in which a finance company randomly liberalized screening criteria on consumer loans in South Africa and find significant positive effects of access to credit on 1 For a thorough review of the literature on the economics of remittances see Rapoport and Docquier (2006). Studies investigating the macreoeconomic effects of remittances include: Giuliano and Ruiz-Arranz (2009), Mundaca (2009), and Chami and Fullenkamp (2012). This last study argues that the potential effect of remittances on economic growth is ambiguous. Massey and Parrado (1998), Woodruff and Zenteno (2001 and 2007), Dustmann and Kirchkamp (2002), Mesnard (2004), Yang (2006), and Woodruff (2007) examine the impact of remittances on investments in microenterprises. The literature on the effects of remittances on poverty includes: Adams (2004 and 2006), Adams and Page (2005), Taylor et al. (2005), Yang and Martinez (2005), and Acosta et al. (2007). Studies on the impact of remittances on inequality include Adams (1992), Taylor and Wyatt (1996), and McKenzie and Rapoport (2007). Kanaiaupuni and Donato (1999), Hildebrandt and McKenzie (2005), Lopez Cordova et al. (2005), and Amuedo-Dorantes et al. (2007) analyze the effects of remittances on health and mortality, while Cox-Edwards and Ureta (2003), Hanson and Woodruff (2003), Lopez Cordova et al. (2005), Yang (2008), and Acosta et al. (2007) investigate the impact on education. consumption, economic self-sufficiency (measured by employment status and income), and some aspects of mental health and outlook. 2 Studies on the impact of accessing and using savings products also find positive effects. In particular, the literature has found that providing individuals access to savings instruments increases savings (Aportela, 1999, Ashraf et al., 2010a 3 ), female empowerment (Ashraf et al. 2010b), productive investment (Dupas and Robinson, 2009), and consumption (Dupas and Robinson, 2009 and Ashraf et al. 2010b). Using household-level survey data for El Salvador for the period 1995-2001, this study investigates the impact of remittances on financial inclusion. In particular, we focus on whether remittances promote the use of deposit accounts and credit by examining the effect of remittances on the likelihood that households have a deposit account, apply for loans, and receive loans from formal financial institutions (commercial and saving banks, credit unions, credit cooperatives, etc). There are several ways in which remittances could potentially affect financial inclusion. First, remittances might increase the demand for savings instruments. The fixed costs of sending remittances make the flows lumpy, providing households with excess cash for some period of time. This might potentially increase their demands for deposit accounts, since financial institutions offer households a safe place to store this temporary excess cash. Second, remittances might increase household’s likelihood of obtaining a loan. Processing remittances flows provides financial institutions with information on the income of recipient households. This information might make financial institutions better willing and able to extend loans to otherwise opaque borrowers. On the other hand, since remittances might help relax households’ 2 Other studies that find that microcredit increases consumption or income include Pitt and Khandker (1998) and Khandker (2005). However, Roodman and Morduch (2009) are unable to reproduce those results. 3 Ashraf et al. (2010a) show that giving migrants from El Salvador access to bank accounts that allow them to have control over the funds they remit results in an increase in savings. 2 financing constraints (Giuliano and Ruiz-Arranz, 2009, Chami and Fullenkamp, 2012), the demand for credit might fall as remittances increase. We conduct estimations of the likelihood of using deposit and credit services, allowing for municipality-, or household-level fixed effects to control for factors other than remittances that might affect financial inclusion at the household level. 4 Also, to deal with the potential endogeneity of remittances, we conduct three different set of instrumental variables regressions, using economic conditions in migration recipient US states, municipality measures of the importance of remittances , and lagged values of remittances, respectively, as instruments. Our estimations consistently show that households that receive remittances are more likely to have a deposit account at a financial institution. Our most conservative estimates indicate that receiving remittances increases the likelihood of having an account by at least 11 percentage points and an additional colon per household member in remittances raises this probability by 8 percentage points. These effects are sizeable given that on average only 19 percent of households have an account. However, remittance-recipient households are not more prone to request or receive a loan. This suggests that though remittances might have the potential to encourage the use of savings instruments, they do not necessarily foster the demand for and use of credit, perhaps because they help to relax credit constraints. Our paper is related to studies that examine the impact of remittances on the development of the domestic financial system. 5 Aggarwal et al. (2011) use balance of payments data on remittances and analyze how these flows affect bank credit and deposit amounts for 109 developing countries over the period 1975-2007. They find strong evidence indicating that 4 El Salvador is divided into 262 municipalities. 5 This question is important given the literature has shown that financial development can have a significant impact on economic growth and poverty reduction (King and Levine, 1993; Beck et al., 2000; and Beck et al., 2004). 3 remittances promote financial development as measured by the ratio of bank deposits and of bank credit to GDP. Gupta et al. (2009) use a similar methodology as Aggarwal et al. (2011) and find analogous results for a sample of Sub-Saharan African countries. Focusing exclusively on Mexico and using municipality-level data, Demirgüç-Kunt et al. (2011) find that municipalities where a higher proportion of households receive remittances have a higher number of bank branches and accounts per capita, and larger shares of deposits to GDP. Although these papers find evidence of a positive impact of remittance flows on financial depth, they provide little information regarding the effect of remittances on financial inclusion. Financial inclusion is not the same as financial depth. Financial depth might be high and financial inclusion might be low if large amount of credits are assigned to few households. For example, the fact that larger inflows of remittances increase financial depth at the country or at the municipality level does not necessarily mean that those households that receive remittances are more likely to be financially included. In particular, the increase in credit at the macro level due to higher remittances flows might only be channeled to non-remittance recipient households, keeping poorer remittance recipient households out of the financial system. Similarly, deposits might increase in the economy even if remittance recipients themselves do not open accounts, as the remittances flowing into the economy are spent and end up in the deposit accounts of non- remittance recipients. Our paper is also complementary to a literature that addresses the impact of remittances on economic development and their interaction with the financial sector. In particular, Giuliano and Ruiz-Arranz (2009) use balance of payments data and find evidence indicating that remittances constitute an alternative way of financing investment, especially in shallow financial systems. Hence, the authors suggest that instead of boosting credit markets, remittances act as a 4 direct source of funds to financially constrained households. This paper directly tests the mentioned hypothesis by analyzing the effect of remittances on credit demand. Our paper contributes to the study of the impact of remittances on the financial sector in several ways. First, to our knowledge, this is the first paper to analyze the impact of remittances on financial inclusion directly. Second, unlike the previous literature that focuses only on the banking system, this paper also considers other formal financial institutions in El Salvador such as cooperatives, credit unions, and financieras. Third, by looking at whether households apply for loans, this study is able to examine the impact of remittances on the demand for credit and not purely on credit outcomes. This is useful because it can allow us to begin to assess to what extent remittances might relax credit constraints. Fourth, by using survey as opposed to balance of payments data, this study can circumvent some of the limitations of the previous studies. In particular, the survey data used in this study has the potential to capture remittances flows received through formal and informal channels, minimizing concerns about measurement error in remittances. 6 Also, because we use household-level panel data we are able to control for unobserved household characteristics that can affect both remittances and financial inclusion, reducing concerns about endogeneity. Finally, our study offers evidence on the impact of remittances on financial inclusion for a new country - El Salvador – for which remittances represent a very significant share of GDP. The rest of the paper is organized as follows. Section II characterizes the Salvadoran financial system during our period of study. Section III describes the data and the empirical methodology we employ. Section IV presents the results of our estimations. Section V concludes. 6 This is a significant concern in the study by Aggarwal et al. (2011) that uses balance of payment data on remittances. 5 II. The Salvadoran financial system during the late 1990s El Salvador’s financial system witnessed a number of reforms during the 1990s. In the late 1980s, during the final years of the civil war, the system became largely insolvent. Trying to overcome that situation, in 1989, the Salvadoran government adopted a reform plan with the aim of increasing competition and efficiency. In this process, several laws were passed allowing the privatization of banks, removing restrictions on foreign bank entry, and establishing new supervision rules. As a consequence of these reforms and due to good macroeconomic performance, there was a considerable increase in the depth and the size of the financial system (see Fuentes 2001). As shown in Table 1, from 1991 to 2001, the ratio of bank and non-bank credit to the private sector expressed as a share of GDP increased from 23 percent to 39 percent, while the ratio of demand and term deposits to GDP rose from 24 percent to 36 percent. The number of banks and non-bank institutions did not change significantly during the period: there were 14 institutions that were allowed to collect deposits in 2001, only one more than in 1991. The Salvadoran financial sector became one of the deepest in Central America 7 by the end of the 1990s (see Figure 1). However, despite the financial reforms undertaken during this decade, financial inclusion remained low in El Salvador. Based on a survey of rural households, conducted by FUSADES (Salvadoran Foundation for Economic and Social Development), 8.7 percent of households had a deposit account as of 1995 and 6.8 percent had a loan from a formal financial institution. III. Data and empirical methodology 7 Central America in this case includes the following countries: Guatemala, Honduras, El Salvador, Nicaragua, and Costa Rica. 6 The household-level data we use in this study come from the National Rural Household Survey carried out by FUSADES in 1996, 1998, 2000, and 2002. All the information gathered by each survey refers to the previous calendar year. The survey uses a questionnaire adapted from the World Bank’s Living Standards Measurement Survey and covers a stratified, nationally representative, random sample of rural households. The four waves contain information for about 937 households dispersed across the 14 departments that make up El Salvador. Within our sample, 719 households are present in more than 1 wave and 451 households have data for the four waves, allowing us to construct a panel. The survey includes data on demographic characteristics, education, employment, economic activities, wealth, and income of households. With respect to migration, the survey has information on the number of members per household that have migrated to other countries, on whether the individuals that migrated sent remittances in a given year, and on the amount sent. The survey also contains information on households’ use of financial services offered by formal financial institutions (i.e., commercial banks, cooperatives, credit unions, and other financial institutions), development institutions (i.e., NGOs), and informal sources (i.e. stores, friends, employers). In this study, we focus on formal financial institutions in order to analyze the impact of remittances on access to financial services such as loans, savings, and checking accounts. Table 2 shows the summary statistics and descriptions of the variables we employ in our analysis. Our empirical analysis of the impact of remittances on financial inclusion is based on the literature analyzing the determinants of households’ use of financial services. Theoretically, Zeller (1995) shows that access to credit and savings increases households’ utility by allowing for consumption smoothing. It also shows that the demand for savings and credit is a function of 7 household income. Empirically, studies such as Kochar (1997), Pal (2002) and Barslund and Tarp (2008) study the factors driving the demand for formal and informal loans, while Kiiza and Pederson (2002) focus on the use of deposit accounts. Overall, these studies highlight the role of household size, education, age, and gender composition as key determinants of the demand for financial services. These variables are intended to capture household income as well as the ability to understand the benefits of using financial products. 8 Given the literature discussed above, in order to investigate the impact of remittances on financial inclusion, we estimate the following model: Financial inclusionit= α+β1Remittancesit + β2Educationit + β3Ageit + β4 Number of adultsit + β5Share of dependentsit + β6Share of femalesit + εit (1) Where i denotes households, t denotes time. Financial inclusion refers to three different alternative dependent dummy variables indicating: (i) whether the household has a deposit account at a formal financial institution; (ii) whether any member of the household has applied for a loan from a financial institution; and (iii) whether the household has received a loan from a financial institution. Considering the whole sample, on average, 19 percent of the households have a deposit account, whereas 7 percent of households asked for a loan from a formal financial institution, and 6 percent received a loan. Depending on the specification, Remittances refers either to a dummy variable indicating whether the household received remittances from abroad in a given year or to the real amount of remittances per household member that a household received in a given year. In the survey, information on remittances sent is reported for each member of the household living abroad. For 8 Given the potential for reverse causality going from the use of financial services to income, we include the determinants of income in equation (1), rather than include income directly. We thank an anonymous referee for this suggestion. 8 those households with more than one migrant, we compute the total amount of remittances by summing the remittances sent by all the members residing overseas. We then divide this sum by the number of household members to get the per household member data. The remittances dummy variable equals one when the sum of remittances flows to a household is different from zero in a given year. On average, 20 percent of households received remittances in our sample and the average annual real amount of remittances received per household member is 340 colones or 39 dollars (in 1995 prices). Ex-ante, as discussed in the introduction, we expect remittances to increase the use of deposit accounts. On the other hand, the impact of remittances on credit demand and use is a priori ambiguous. Education denotes the average number of years of education completed by adult household members. We consider as adult members individuals that are 18 years old or older. This variable is included as a determinant of the household’s earnings (following Mincer, 1974) and can also capture the extent to which households can understand the features and usefulness of saving and credit instruments. Ex-ante, we expect education to be positively related to the use of financial services. Taking the whole sample, the adults’ average number of years of education completed is 3.51. Age refers to the average age of the adults in the household and is intended to capture the potential experience of the household, a major determinant of household’s earnings (Mincer, 1974). We expect this variable to be positively correlated with remittances. In our sample, the average age of adults in the household is 40. The number of adults in the household is a measure of household size and proxy for household income potential. Hence, we expect this variable to be positively correlated with the use of financial products. The mean number of adults is 3. 9 The share of dependents refers to the ratio of household members below 18 years of age. We assume that individuals below 18 years of age are less likely to work and, hence, households that have a higher percentage of dependents will have lower incomes, less ability to save but, potentially, a greater need for financing for things like education and health. The share of dependents averages 0.49 in our sample. Finally, the share of female adults is likely to be negatively correlated with household income because generally female earn less than men 9 and from that perspective we expect this variable to also be negatively correlated with the use of financial services. On the other hand, to the extent that females are more risk averse than men it is possible that they save more than men and borrow less 10, in which case we would expect the share of females to be positively correlated with the use of deposits and negatively correlated with credit. In our sample, the share of female adults averages 0.5. An important concern regarding the interpretation of the results of equation 1 is that estimates could be affected by endogeneity as a result of omitted factors or reverse causation. Our estimates could be biased if there are some omitted variables that affect both the likelihood that a household receives remittances and the probability that the household has a deposit account, applies for a loan, or has credit outstanding. We address this concern by running fixed effect estimations either at the municipality or household level. There are at least two reasons why reverse causation could also be a problem. First, financial inclusion might reduce the costs of sending and receiving remittances and, hence, might 9 For seminal studies on the male-female earnings gap see, among others, Mincer and Polacheck (1974) and Goldin and Polacheck (1987). 10 Analyzing student attitudes to student debt, Davies and Lea (1995) find that men are more likely to be in debt than women. Using both historical data and experiments, Griskevicius et al. (2012) examined how sex ratio influences saving, borrowing, and spending in the United States. They find that male-biased sex ratios decreased men's desire to save for the future and increased their willingness to incur debt for immediate expenditures. 10 make migrants more prone to send and households to receive remittances. Recent data on the costs of sending remittances to El Salvador collected by The World Bank does not suggest that the costs are lower for migrants using deposit taking financial institutions as opposed to money transfer operators. 11 Hence, a priori it is not clear that financial inclusion leads to lower costs of remittances. Second, financial institutions could finance migration, and, consequently, increase the remittance flow towards households with access to credit. We think formal financial institutions are not likely to finance migrants since these are typically poor individuals with no collateral. Even though we see less merit in the concerns about reverse causation, we nonetheless run three sets of instrumental variables regressions to deal with potential biases. In our first and preferred set, we use municipality-level weighted average measures of economic conditions in US states where Salvadoran migrants reside as instruments. 12 In particular, we focus on real GDP per capita and household median income in US states during 1995, 1997, 1999, and 2001. For each municipality in El Salvador, we identify the main US destinations for Salvadoran migrants using data on migration patterns for Salvadorans from the North American Integration and Development Center (NAID). 13 We then construct, for each Salvadoran municipality, a weighted average of economic conditions in US states where migrants from that municipality reside. We weight every state in the US according to the share of migrants from that municipality 11 See data on the cost of sending remittances from the US to El Salvador from 2008 (the closest year to our sample period for which data is available) at http://remittanceprices.worldbank.org. 12 We follow McKenzie and Rapoport (2007), who in their study of the impact of migration on income inequality in Mexico, use economic conditions in the US as one of their instruments for migration from Mexico. A similar approach is pursued by Yang (2008), Amuedo-Dorantes and Pozo (2010) and Orrenius, Zavodny, Cañas, and Coronado (2010). 13 The mentioned database is available at http://gis.ats.ucla.edu/naid/. The NAID gathered information on the main destinations in the US of Salvadoran migrants during the last five years. We assume that migration patterns during the last five years are similar to those at the time of our household surveys. In the case of Mexico, Hildebrandt and McKenzie (2005) show that the historic migration rates are a strong predictor of current migration rates. 11 in El Salvador that reside in each US state. Hence, for every municipality j and year t the instruments are constructed as follows: I = � wji RealGDPpcit i I ℎ = � wji Household median Incomeit i where: i represents every US state, RealGDPpcit and Household median incomeit are the real GDP per capita and real household median income in US state i at year t, respectively. wji is the share of migrants from municipality j in El Salvador (out of the total migrants from that municipality) residing in state i in the US. The basic intuition for why we use these variables as instruments is that we think that economic conditions in US states will affect the ability of migrants to send remittances back to their communities in El Salvador, but these variables are unlikely to directly influence the use of financial services by households back in El Salvador. We consider the possibility that US economic conditions can affect household’s income in El Salvador via channels other than remittances but find that US real GDP per capita and household median income across US states are not significantly correlated with Salvadoran households’ non-remittances income. The correlation between non-remittances household income in El Salvador and the median household income in the US is 0.01 and has a p-value of 0.58. The correlation between non-remittances household income in El Salvador and the real GDP per capita across US states is -0.02, with a p-value of 0.37. Our second set of instrumental variable estimations consider the municipality percentage of households that receive remittances and the annual average amount received per municipality 12 as instruments for whether a household receives remittances. 14 The intuition here is that there are network effects 15 that will make some municipalities more prone to migration and to receiving remittances (i.e., those municipalities from which individuals have migrated in the past and which have formed a network of migrants in the destination country) and that this will affect the likelihood that a given household receives remittances. But, at the same time, we do not expect these municipality-level variables to affect the household’s use of financial products, especially after including municipality-level fixed effects. Finally, we conduct estimations where we use lagged values of regressors to instrument for the current values of the remittances. These estimations involve a smaller number of observations, since we focus on households that are present in at least 3 surveys and, hence, have at least 2 lagged values of remittances that we can use as instruments. IV. Results Table 3 presents the estimation results for the likelihood of having a deposit account, receiving a loan, and applying for a loan. 16 The table shows two sets of estimations: including the remittances dummy variable (Panel A) and, separately, the annual real amount of remittances per household member (Panel B). For each measure of remittances, we conduct different estimations, depending on whether we include fixed effects and, if so, at what level of aggregation. Results in columns (1), (4) and (7) of each panel do not include any fixed effects, those in columns (2), (5) and (8) control for municipality fixed effects, and those in columns (3), (6) and (9) allow for household fixed effects. We believe that estimations including municipality 14 A similar approach is pursued by Stephenson (2011). 15 For a discussion of the importance of networks of migrants see Munshi (2003). . 16 We estimate a linear probability model to avoid the incidental parameters problem associated with probit estimations. 13 and households effects help mitigate concerns about biases resulting from omitted variables driving the association between remittances and the use of deposits and loans. Columns (1)-(3) in both panels show the estimation results for the impact of remittances on the likelihood that a household has a deposit account. The results indicate that in all model specifications both the remittances dummy and the annual amount of remittances per household member are statistically significant and positive. Furthermore, both variables have economically large effects. Based on column (3), we find that receiving remittances increases the probability that the household will have a deposit account by 11 percentage points. Furthermore, an additional colon per household member in annual remittances raises the likelihood of having an account by close to 9 percentage points. These effects are sizeable considering that in the sample 19 percent of households have deposit accounts. As expected, Adults average education and average age of adults have a positive and significant impact on the likelihood that the household owns a bank account across specifications. On the other hand, the number of adults and the percentage of female members are primarily significant once we include household fixed effects. The results for the likelihood of receiving a loan from a formal financial institution are shown in columns (4)-(6) in both panels. In contrast with the previous results, the remittances variables are generally not significant. These results could be driven by the fact that there are potentially two opposing effects of remittances on credit: (1) remittances can be used as an informal type of collateral or as a means to provide information on households’ income that might make financial institutions more willing to provide credit; and (2) remittances may help relax households’ credit constraints and, hence, could reduce their demand for credit. Thus, these 14 two effects may well be counteracting each other, causing remittances to be generally insignificant. We find evidence in indicating that Adults average education has a positive and significant influence on the probability of receiving a loan. This result holds in all model specifications, except the one that includes household fixed effects. The share of female adults tends to have negative effect on the likelihood of receiving a loan. On the other hand, the higher share of dependents (i.e., members below 18 years of age), the higher the likelihood of receiving a loan. Columns (7)-(9) in both panels of Table 3 show that remittances tend to reduce the demand for credit from financial institutions. This is consistent with the notion that remittances might help relax households’ credit constrains, lowering their demand for loans. However, the coefficients on the remittances dummy and on the annual amount of remittances received are not significant. Adults average education and the share of dependents typically have a positive and significant effect on the likelihood that the household requested a loan. On the other hand, the percentage of female adults has a negative impact on this probability. Even though estimations in Table 3 try to address concerns about endogeneity by including fixed effects to control for omitted variables, as mentioned above, these results could still be biased due to reverse causation (i.e., the use of banking services driving the likelihood that a household receives remittances). The presence of financial institutions may cause higher remittance flows, either because financial inclusion allows people to finance migration, and hence increases migration flows and remittances, or because financial inclusion is associated with lower costs of receiving remittances and, hence, a greater propensity to do so. Neither of these seems to be a first order concern. Financial institutions in El Salvador are an unlikely 15 source of credit to finance migration. While access to financial services might facilitate receipt of remittances, the primary channel is from migration to financial inclusion. 17 Nevertheless, to try to assuage concerns about endogeneity, we run 3 sets of instrumental variables (IV) regressions. First, we use as instruments municipality-level weighted averages of economic conditions in US states where Salvadoran migrants reside. In particular, we focus on the real GDP per capita and the median household income across the US states that are destinations for El Salvador migrants. Second, we consider the municipality percentage of households that receive remittances and the annual average amount received per household member at the municipality as instruments for whether a household receives remittances. Finally, we report estimations where we use lagged values of remittances as instruments for the current value of this variable. For all IV estimations, we report results focusing on the remittances dummy variable as opposed to the remittances amount per household member for two reasons. First, we want to reduce the number of tables we report and we have found that the results are qualitatively the same regardless of which remittances variable we focus on. Second, we expect the remittances amount per household variable to be more prone to measurement error which could lead to estimation biases. In other words, we expect households to adequately report whether they receive remittances or not, but we anticipate more measurement error when it comes to questions about remittances amounts. Nevertheless, all IV results using the remittance amount variable are available upon request. Table 4, Panel A presents the results of the first stage estimations where the likelihood of receiving remittances is instrumented with the weighted average of real GDP per capita and/or 17 We checked this by running a regression where we replace the remittance dummy for a dummy which equals one for households with at least a member that is residing overseas. The results are nearly identical when we include the dummy for whether the household has a migrant member in place of the dummy for whether it receives remittances. Thus, the effect on the use of deposit accounts appears to be driven by migration, which causes remittance flows. These results are available upon request. 16 the real median household income for US states where Salvadoran migrants reside. We use these instruments separately and jointly. In this table, the size of the Cragg-Donald F-statistics exceeds the Stock and Yogo (2005) critical values, indicating that we do not have a weak instruments problem. We find that GDP per capita in US states has a positive effect on the likelihood that the household receives remittances (Table 4. Panel A columns (1), (4) and (7)) when this variable is included on its own. Furthermore, the median household income across US states has a positive impact on the likelihood that the household receives remittances (Table 4 Panel A columns (2), (5), and (8)). Table 4 Panel B show the second stage estimations where we instrument the likelihood of receiving remittances. We use the municipality level weighted GDP per capita for US states where migrants reside and the corresponding median household income individually and jointly as instruments. Because the instruments are defined at the municipality level, we control for municipality fixed effects to isolate the effect of the instruments from other omitted municipality characteristics. The overidentification tests reported in both panels suggest that economic conditions in US states where Salvadoran migrants reside are valid instruments (i.e., we cannot reject exogeneity of the other instrument conditional on one instrument being exogenous). We find that our main results are robust to controlling for the potential endogeneity of remittances. Namely, while remittances have a positive effect on the likelihood that a household has a deposit account, it does have a statistically robust impact on the likelihood of applying for and receiving a loan. In Table 5, we present results from IV estimations using the percentage of household that receive remittances and the annual average amount received per household member at the 17 municipality level as instrument for whether a household receive remittances. Panel A show first-stage estimations and Panel B show the second stage results. The Craig-Donald tests in Panel A allow us to reject the null that the proposed instruments are weak. The second stage overidentification tests reported in panel B indicate that our second stage is overidentified. Consistent with results reported so far, Table 5, panel B results indicate that remittances have a positive impact on the likelihood that the household has a deposit account but have no effect on the loan variables. Finally, in Table 6 we show IV results where we use first and second lagged values of remittances to instrument for this variable. In this estimations, we focs only on households that appear in at least 3 of the four surveys included in our dataset, hence the number of observations is smaller. Panel A, first stage results indicate that lagged values of remittances are good predictors of future values of this variable. At the same time, Panel B second stage results are consistent with previous findings that remittances are positively correlated with the likelihood that the household has a deposit account but are not correlated with loan outcomes for the household. V. Conclusions As the importance of remittances for developing countries has grown, a sizeable literature has flourished examining the impact of remittances on various aspects of countries’ development. An issue which has received little attention is the effect of remittances on financial inclusion. This issue is important given the growing evidence that financial inclusion has many beneficial effects for households. 18 This study examined the impact of remittances on financial inclusion using data from a four-wave rural household survey for El Salvador. In particular, we analyzed the impact of remittances on the likelihood that households use financial services such as deposit accounts and loans. Overall, we find that remittances have a positive impact on financial inclusion by promoting the use of deposit accounts. These results hold controlling for unobserved household characteristics and using instrumental variables regressions to correct for the potential endogeneity of remittances. On the other hand, remittances do not have a significant effect on credit from formal financing institutions. Hence, we conjecture that by relaxing credit constraints, remittances might reduce the need for external financing by financial institutions, while at the same time increasing the demand for savings instruments. There are a number of potential avenues for future research. First, it would be interesting to analyze the extent to which remittance recipients that have accounts, actively use these accounts to save and manage their daily transactions. Second, it would be important to go deeper into the reasons why those that receive remittances do not seem to have a higher demand for credit. In particular, it would be useful to analyze whether indeed this is due to the fact that remittances relax credit constraints or because the credit products offered to remittance recipients are not considered adequate by this population. We leave these research questions for future work. 19 References Acosta, P., Fajnzylber, P., Lopez, H., (2007). The impact of remittances on poverty and human capital: Evidence from Latin American household surveys. International remittances and the household: Analysis and review of global evidence. World Bank Policy Research Working Paper 4247. Adams, R., (1992). The impact of migration and remittances on inequality in rural Pakistan. Pakistan Development Review 31, 1189-203. Adams, R., (2004). Remittances and poverty in Guatemala. World Bank Policy Research Paper 3418. Adams, R., (2006). Remittances and poverty in Ghana. World Bank Policy Research Paper 3838. Adams, R., Page, J., (2005). Do international migration and remittances reduce poverty in developing countries? World Development 32, 1645-1669. Aggarwal, R., Demirgüç-Kunt, A., Martínez Pería, M.S., (2011). Do remittances promote financial development? Journal of Development Economics 96 (2), 255-264. Amuedo-Dorantes, C., Pozo, S., (2010). When do remittances facilitate asset accumulation? The importance of remittance income uncertainty. Mimeo. http://homepages.wmich.edu/~pozo/research%20papers/Asset_accumulation_May%207_ %202010.pdf Amuedo-Dorantes, C., Sainz, T., Pozo, S., (2007). Remittances and healthcare expenditure patterns of population in origin communities: Evidence from Mexico. INTAL-ITD Working paper 25. Aportela, F., (1999). Effects of financial access on savings by low-income people. MIT Department of Economics Dissertation Chapter 1. 20 Ashraf N., Aycinena, C., Martinez, A., Yang, D., (2010a). Remittances and the problem of control: A field experiment among migrants from El Salvador. Mimeo. Ashraf, N., Karlan, D., Yin, W., (2010b). Female empowerment: Further evidence from a commitment savings product in the Philippines. World Development 28 (3), 333-344. Banerjee, A., Duflo, E., Glennerster, R., Kinnan, C., (2010). The miracle of microfinance? Evidence from a ramdomized evaluation. MIT.Bureau for Research and Economic Analysis of Development working paper 278. Barslund, M., Tarp, F., (2008). Formal and informal rural credit in four provinces of Vietnam. Journal of Development Studies 44, 485-503. Beck, T., Demirgüç-Kunt, A., Levine, R., (2004). Finance, inequality and poverty: Cross-country Evidence. NBER Working Paper 10979. Beck, T., Levine, R., Loayza, N., (2000). Financial intermediation and growth: Causality and causes. Journal of Monetary Economics 46, 31-77. Chami, R., Fullenkamp, C., (2012). Workers' remittances and economic development: realities and possibilities, in Maximizing the Development Impact of Remittances, New York: United Nations Conference on Trade and Development. Cox-Edwards, A., Ureta, M., (2003). International migration, remittances, and schooling: Evidence from El Salvador. Journal of Development Economics 72(2), 429-461. Davies, E., Lea, S., (1995). Student attitudes to student debt. Journal of Economic Psychology 16 (4), 663-679. Demirgüç-Kunt, A., López Córdova, E., Martínez Pería, M.S., Woodruff, C., (2011). Remittances and banking sector breadth and depth: Evidence from Mexico. Journal of Development Economics 95(2), 229-241. 21 Dupas, P., Robinson, J., (2009). Savings constraints and microenterprise development: Evidence from a field Experiment in Kenya. National Bureau of Economic Research Working Paper 14693. Dustmann, C., Kirchkamp, O., (2002). The optimal migration duration and activity choice after remigration. Journal of Development Economics 67(2), 351-72. Fuentes, J., (2001). Estructura competitiva del mercado bancario Salvadoreño en la década de los noventa. Reserve Central Bank of El Salvador, Economic and Financial Research Department. Giuliano, P., Ruiz-Arranz, M., (2009). Remittances, financial development, and growth. Journal of Development Economics 90, 144-152. Goldin, C., Polacheck, S., (1987). Residual Differences by Sex: Perspectives on the Gender Gap in Earnings. American Economic Review 77, 143-51. Griskevicius, V., Tybur, J., Ackerman, J., Delton, A., Robertson, T., White, A., (2012). The financial consequences of too many men: Sex ratio effects on saving, borrowing and spending. Journal of Personality and Social Psychology 102(1), 69-80. Gupta, S., Pattillo, C., Wagh, S., (2009). Effect of remittances on poverty and financial development in Sub-Saharan Africa. World Development 37(1), 104-115. Hanson, G., Woodruff C., (2003). Emigration and educational attainment in Mexico. Mimeo, University of California at San Diego. Hildebrandt, N., McKenzie, D., (2005). The effects of migration on child health in Mexico. Economia: Journal of the Latin American and Caribbean Economic Association 6, 257- 289. 22 Kanaiaupuni, S., Donato, K., (1999). Migradollars and mortality: The effects of migration on infant survival in Mexico. Demography 36 (3), 339-353. Karlan, D., Zinman, J., (2010). Expanding credit access: Using randomized supply decisions to estimate the impacts. Review of Financial Studies 23 (1), 433-464. Khandker, S., (2005). Microfinance and poverty: Evidence using panel data from Bangladesh. World Bank Economic Review 19(2), 263-286. Kiiza, B., Pederson, G., (2002). Household financial savings mobilization: Empirical evidence from Uganda. Journal of African Economies 10(4), 390-409. King, R., Levine, R., (1993). Finance and growth: Schumpeter might be right. Quarterly Journal of Economics 108, 717-737. Kochar, A. (1997). An empirical investigation of rationing constraints in rural credit markets in India. Journal of Development Economics 53(2), 339–371 Lopez Cordova, E., Tokman R. A.,Verhoogen E.A., (2005). Globalization, migration, and development: The role of remittances. Economía 6 (1), 217-256. Massey, D., Parrado, E., (1998). International migration and business formation in Mexico. Social Science Quarterly 79 (1), March. McKenzie, D., Rapoport, H., (2007). Network effects and the dynamics of migration and inequality: Theory and evidence from Mexico. Journal of Development Economics 84(1), 1-24. Mesnard, A., (2004). Temporary migration and capital market imperfections. Oxford Economic Papers 56, 242-62. Mincer, J., (1974). Schooling, Experience, and Earnings. Columbia University Press: New York. 23 Mincer, J., Polacheck, S., (1974). Family investments in human capital: earnings of women. Journal of Political Economy 82, 76-108. Mundaca, G., (2009). Remittances, financial markets development and economic growth: The case of Latin America and Caribbean. Review of Development Economics 13 (2), 288- 303. Munshi, K., (2003). Networks in the modern economy: Mexican migrants in the U.S. labor market. Quarterly Journal of Economics 118 (2), 549–599. Orrenius,P., Zavodny, M., Cañas, J., Coronado, R. (2010). Do remittances boost economic development? Evidence from Mexican States. Research Department Working Paper 1007. Federal Reserve Bank of Dallas. http://www.dallasfed.org/assets/documents/research/papers/2010/wp1007.pdf Pal, S. (2002). Household sectoral choice and effective demand for rural credit in India. Applied Economics, 34(14), 1743-1755. Pitt, M., Khandker, S., (1998). The impact of group-based credit on poor households in Bangladesh: Does the gender of participants matter? Journal of Political Economy 106(5), 958–96. Rapoport, H., Docquier, D., (2006). The economics of migrants' remittances. In S. Kolm and J. Mercier Ythier, eds. Handbook of the Economics of Giving, Altruism and Reciprocity, Amsterdam: North Holland, Vol. 2, Chap. 17. Roodman, D., Morduch, J., (2009). The impact of microcredit on the poor in Bangladesh: Revisiting the evidence. Center for Global Development Working Paper Number, August. http://www.cgdev.org/files/1422302_file_Roodman_Morduch_Bangladesh.pdf 24 Stephenson, A.V., (2011). The effects of outside income on household behavior: The case of Remittances in Jamaica. Economics Dissertations. Paper 79. Georgia State University. Stock, J., Yogo, M., (2005). Testing for weak instruments in linear IV regression, In J.H. Stock and D.W.K. Andrews (eds), Identification and Inference for Econometric Models: Essays in Honor of Thomas J. Rothenberg, Cambridge University Press. Taylor, J.E., Wyatt, T.J., (1996). The shadow value of migrant remittances, income, and inequality in a household-farm economy. Journal of Development Studies 32, 899-912. Taylor, J.E., Mora, J., Adams, R., Lopez-Feldman, A., (2005). Remittances, inequality, and poverty: Evidence from rural Mexico. Mimeo, University of California. Woodruff, C., (2007). Mexican microenterprise investment and employment: The role of remittances. Integration and Trade 27, July-December. Woodruff, C., Zenteno, R., (2001). Remittances and micro-enterprises in Mexico. Mimeo. University of California San Diego. Woodruff, C., Zenteno, R., (2007). Migration networks and microenterprises in Mexico. Journal of Development Economics 82, 509-528. Yang, D., (2006). Why do migrants return to poor countries? Evidence from Philippine migrants’ responses to exchange rate shocks. Review of Economics and Statistics 88, 715-735. Yang, D., (2008). International migration, remittances, and household investment: Evidence from Philippine migrants’ exchange rate shocks, Economics Journal 118, pp. 591-630. Yang, D., Martinez, C., (2005). Remittances and poverty in migrants’ home areas: Evidence from the Philippines. In C. Ozden and M. Schiff, eds. International Migration, Remittances, and the Brain Drain, World Bank. Zeller, M., (1995). The demand for financial services by rural households: conceptual 25 framework and empirical findings. Quarterly Journal of International Agriculture 34(2): 149-170. 26 Table 1: Financial system indicators for El Salvador Year Credit to private sector Demand and term Number of to GDP (%) deposits to GDP (%) institutions 1991 23.4 23.8 13 1992 27.5 25.7 13 1993 27.7 28.4 13 1994 30.9 31.1 17 1995 34.8 29.9 21 1996 36.9 31.3 21 1997 40.3 32.6 18 1998 41.9 33.2 17 1999 43.8 34.4 16 2000 41.4 34.0 15 2001 39.0 36.0 14 Source: Central Bank of El Salvador and Financial System Superintendency of El Salvador. 27 Table 2: Variable description and summary statistics Variable Description Obs. Mean Sd. Deposit account Dummy variable. Equals one if the household has a 2503 0.19 0.39 checking or savings account in a formal financial organization. Loan received Dummy variable. Equals one if the household received a 2708 0.06 0.25 loan from a formal financial institution in a given year. Loan requested Dummy variable. Equals one if the household requested 2717 0.07 0.26 a loan from a formal financial institution in given year. Remittances dummy Dummy variable. Equals one if the household received 2741 0.20 0.40 remittances from overseas in a given year. Remittances amount Value of annual remittances per household member in 2708 0.34 1.03 real terms (thousands of colones). Adults average education Average number of years of education of adults. 2742 3.51 2.73 Share of female adults Percentage of female adults in the household. 2748 0.50 0.18 Number of adult members Number of adult members in the household. 2730 3.17 1.40 Average age of adults Average age of adults in the household 2699 40.09 10.29 Share of dependents Percentage of household members below 18 years of age. 2281 0.49 0.17 Real GDP pc in US Weighted average of real GDP per capita in US states 2578 33.90 2.92 remittance-source states where Salvadoran migrants reside. For every municipality j and year t the weighted GDP per capita is constructed as follows: ∑ , where i represents US states and wji is the percentage of migrants from municipality j that lives in US state i. Real median household Weighted real median income in US states where 2578 40.55 2.57 income in US remittance- Salvadoran migrants reside. For every municipality j and source states year t the weighted household income is constructed as follows: ∑ , where i represents US states and wji is the percentage of migrants from municipality j that lives in US state i. % of households receiving Percentage of household that receive remittances from 2621 0.20 0.24 remittances per overseas in a given year per municipality. municipality Municipality avg. per HH This variable shows the average by municipality and year 2621 0.45 0.82 member annual real of the per household member annual real amount of amount of remittances. remittances (in thousands). 28 Table 3. Estimations for the likelihood that the household has a deposit account, has or requested a loan from a financial institution. This table shows the results from the estimations of a linear probability model. In both Panel A and B, the dependent variables are Deposit account in columns (1)-(3), Loan received in columns (4)-(6), and Loan requested in columns (7)-(9). All dependent and independent variables are defined in Table 2. In Panel A, the variable Remittances is dummy for whether the household receives remittances from overseas, whereas in Panel B Remittances is the annual amounts per household member received in remittances from overseas. Robust t-statistics are shown in brackets. *,**, *** denote significance at 10, 5 and 1 percent, respectively. Panel A. Remittances: dummy for whether the household receives remittances from overseas in a given year Deposit account Loan received Loan requested (1) (2) (3) (4) (5) (6) (7) (8) (9) Remittances 0.120 0.108 0.111 0.009 0.002 -0.003 0.005 -0.000 -0.003 [4.96]*** [3.97]*** [3.30]*** [0.65] [0.14] [-0.16] [0.36] [-0.03] [-0.13] Adults avg. education 0.045 0.044 0.021 0.009 0.009 0.004 0.009 0.007 0.005 [12.53]*** [10.24]*** [2.58]*** [3.62]*** [3.16]*** [0.73] [3.53]*** [2.64]*** [0.92] Share of female adults 0.070 0.093 0.180 -0.112 -0.066 -0.046 -0.119 -0.067 -0.048 [1.44] [1.63] [2.52]** [-3.77]*** [-2.08]** [-0.88] [-3.90]*** [-2.06]** [-0.93] Number of adult members -0.000 0.003 0.035 0.007 0.010 0.019 0.006 0.010 0.018 [-0.00] [0.42] [2.72]*** [1.60] [1.92]* [2.19]** [1.16] [1.78]* [2.08]** Average age of adults 0.004 0.004 0.006 -0.001 -0.000 -0.001 -0.001 -0.001 -0.001 [4.41]*** [3.55]*** [3.56]*** [-1.36] [-0.40] [-0.53] [-1.56] [-0.88] [-0.68] Share of dependents -0.022 -0.049 -0.084 0.072 0.072 0.197 0.053 0.053 0.186 [-0.39] [-0.75] [-0.79] [1.92]* [1.81]* [2.38]** [1.33] [1.28] [2.18]** Constant -0.189 -0.183 -0.311 0.062 0.016 -0.053 0.093 0.052 -0.033 [-2.82]*** [-2.32]** [-2.62]*** [1.42] [0.32] [-0.58] [1.97]** [0.97] [-0.36] Observations 2025 1924 1954 2200 2099 1964 2207 2106 1970 Adj-R2 0.110 0.159 0.059 0.015 0.056 0.033 0.014 0.059 0.004 Municipalities fixed effects No Yes No No Yes No No Yes No Household fixed effects No No Yes No No Yes No No Yes Panel B. Remittances: annual amount of remittances per HH member received from overseas (in real terms) Deposit account Loan received Loan requested (1) (2) (3) (4) (5) (6) (7) (8) (9) Remittances 0.078 0.079 0.091 0.001 0.001 -0.006 -0.001 -0.002 -0.009 [6.80]*** [6.07]*** [6.05]*** [0.13] [0.12] [-0.87] [-0.24] [-0.36] [-1.35] Adults avg. education 0.044 0.042 0.020 0.009 0.009 0.004 0.009 0.008 0.005 [12.16]*** [9.83]*** [2.55]** [3.66]*** [3.21]*** [0.70] [3.56]*** [2.72]*** [0.92] Share of female adults 0.033 0.047 0.132 -0.109 -0.065 -0.024 -0.115 -0.064 -0.022 [0.69] [0.85] [1.91]* [-3.60]*** [-1.97]** [-0.45] [-3.71]*** [-1.91]* [-0.42] Number of adult members 0.004 0.007 0.038 0.007 0.010 0.019 0.006 0.010 0.018 [0.56] [0.90] [2.98]*** [1.59] [1.95]* [2.18]** [1.13] [1.80]* [2.03]** Average age of adults 0.004 0.004 0.006 -0.001 -0.000 -0.001 -0.001 -0.001 -0.001 [4.41]*** [3.60]*** [3.57]*** [-1.25] [-0.40] [-0.48] [-1.47] [-0.88] [-0.59] Share of dependents -0.000 -0.017 -0.083 0.071 0.072 0.197 0.052 0.052 0.185 [-0.00] [-0.27] [-0.81] [1.90]* [1.82]* [2.37]** [1.29] [1.26] [2.16]** Constant -0.186 -0.183 -0.298 0.059 0.015 -0.066 0.091 0.050 -0.048 [-2.82]*** [-2.36]** [-2.55]** [1.35] [0.28] [-0.72] [1.91]* [0.93] [-0.52] Observations 2009 1909 1936 2183 2083 1946 2190 2090 1952 Adj-R2 0.125 0.175 0.003 0.015 0.056 0.061 0.014 0.059 0.004 Municipalities fixed effects No Yes No No Yes No No Yes No Household fixed effects No No Yes No No Yes No No Yes 29 Table 4: Estimations instrumenting remittances with economic conditions in US states This table shows the results of regressions instrumenting the likelihood of receiving remittances in a given year. We show the first stage regressions in Panel A and the second stage in Panel B. The instruments used are Real GDP per capita in US states in columns (1), (3), (4), (6), (7) and (9) and Real median household income in US states in columns (2), (3), (5), (6), (8) and (9), both defined in Table 2. Columns (1)-(3) are the regressions for the likelihood that the household has a deposit account. Columns (4)-(6) are the regressions for the likelihood of receiving a loan, while columns (7)-(9) show the regressions for the likelihood of applying for a loan. All independent variables are defined in Table 2. Municipality fixed effects are estimated but not reported. Robust t-statistics are shown in brackets. *,**, and *** denote significance at 10, 5 and 1 percent, respectively. Panel A. First Stage Regressions. Dependent Variable: dummy for whether household receives remittances from overseas (1) (2) (3) (4) (5) (6) (7) (8) (9) Real GDP pc in US states 0.022 -0.013 0.019 -0.014 0.020 -0.016 [6.53]*** [-0.48] [6.17]*** [-0.52] [6.24]*** [-0.59] Real median HH income in US states 0.024 0.038 0.021 0.036 0.021 0.038 [6.76]*** [1.30] [6.38]*** [1.27] [6.45]*** [1.35] Adults avg. education 0.010 0.010 0.010 0.012 0.012 0.012 0.011 0.011 0.011 [2.56]** [2.52]** [2.50]** [3.10]*** [3.06]*** [3.05]*** [2.88]*** [2.84]*** [2.82]*** Share of female adults 0.326 0.325 0.325 0.306 0.306 0.305 0.301 0.300 0.300 [5.57]*** [5.57]*** [5.57]*** [5.50]*** [5.50]*** [5.50]*** [5.40]*** [5.39]*** [5.39]*** Number of adult members 0.005 0.005 0.005 0.004 0.004 0.005 0.004 0.004 0.004 [0.66] [0.68] [0.71] [0.65] [0.66] [0.69] [0.56] [0.58] [0.61] Average age of adults 0.007 0.007 0.007 0.007 0.007 0.007 0.007 0.007 0.007 [5.87]*** [5.88]*** [5.88]*** [6.45]*** [6.45]*** [6.46]*** [6.36]*** [6.37]*** [6.37]*** Share of dependents -0.017 -0.015 -0.014 -0.060 -0.058 -0.057 -0.063 -0.062 -0.060 [-0.26] [-0.23] [-0.22] [-0.96] [-0.94] [-0.93] [-1.03] [-1.00] [-0.98] Constant -1.039 -1.260 -1.375 -0.930 -1.129 -1.256 -0.922 -1.124 -1.269 [-7.83]*** [-8.05]*** [-4.94]*** [-7.40]*** [-7.61]*** [-4.58]*** [-7.37]*** [-7.59]*** [-4.63]*** Observations 1895 1895 1895 2065 2065 2065 2072 2072 2072 R2 0.287 0.288 0.288 0.273 0.274 0.274 0.272 0.273 0.273 Cragg-Donald weak inst. F-stat 42.68 45.68 23.30 38.03 40.65 20.90 38.88 41.63 21.45 Panel B. Second Stage Regressions. Deposit account Loan received Loan requested (1) (2) (3) (4) (5) (6) (7) (8) (9) Remittances 1.131 1.131 1.131 0.054 0.043 0.036 0.059 0.051 0.045 [5.46]*** [5.58]*** [5.59]*** [0.49] [0.40] [0.34] [0.54] [0.47] [0.42] Adults avg. education 0.033 0.033 0.033 0.008 0.008 0.008 0.007 0.007 0.007 [5.33]*** [5.35]*** [5.35]*** [2.65]*** [2.70]*** [2.73]*** [2.16]** [2.20]** [2.22]** Share of female adults -0.267 -0.267 -0.267 -0.077 -0.074 -0.071 -0.081 -0.078 -0.076 [-2.52]** [-2.54]** [-2.55]** [-1.64] [-1.57] [-1.51] [-1.71]* [-1.66]* [-1.61] Number of adult members -0.006 -0.006 -0.006 0.010 0.010 0.010 0.009 0.009 0.010 [-0.55] [-0.55] [-0.55] [1.92]* [1.93]* [1.94]* [1.78]* [1.79]* [1.80]* Average age of adults -0.004 -0.004 -0.004 -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 [-1.67]* [-1.69]* [-1.69]* [-0.59] [-0.52] [-0.47] [-0.97] [-0.93] [-0.88] Share of dependents -0.023 -0.023 -0.023 0.075 0.074 0.074 0.056 0.056 0.055 [-0.25] [-0.25] [-0.25] [1.88]* [1.88]* [1.87]* [1.34] [1.34] [1.33] Observations 1895 1895 1895 2065 2065 2065 2072 2072 2072 Log likelihood -1305 -1305 -1305 36.26 38.25 39.36 -34.86 -33.11 -31.94 Over-identification test chi2 - - 1.50e-05 - - 0.406 - - 0.239 Over-identification test p-value - - 0.997 - - 0.524 - - 0.625 30 Table 5: Estimations instrumenting remittances using percentage of households that receive remittances and the annual average amount received per municipality This table shows the results of the regressions instrumenting the likelihood of receiving remittances in given year with our second set of instruments. We show the first stage regressions in Panel A and the second stage in Panel B. The instruments used are % of households receiving remittances per municipality (columns (1), (3), (4), (6), (7) and (9)) and Annual avg. amount received per municipality (columns (2), (3), (5), (6), (8) and (9)), both defined in Table 2. Columns (1)-(3) are the regressions for the likelihood that the household has a deposit account. Columns (4)-(6) are the regressions for the likelihood of receiving a loan, while columns (7)-(9) show the regressions for the likelihood of applying for a loan. All independent variables are defined in Table 2. Municipality fixed effects are estimated but not reported. Robust t-statistics are shown in brackets. *,**, and *** denote significance at 10, 5 and 1 percent, respectively. Panel A. First Stage Regressions. Dependent Variable: dummy for whether household receives remittances from overseas (1) (2) (3) (4) (5) (6) (7) (8) (9) % of HH receiving remittances 0.972 1.019 0.957 1.006 0.960 1.008 per municipality [18.98]*** [15.89]*** [19.51]*** [16.55]*** [19.57]*** [16.58]*** Annual avg. amount received 0.138 -0.022 0.134 -0.022 0.134 -0.022 per municipality [7.48]*** [-1.18] [8.09]*** [-1.28] [8.11]*** [-1.27] Adults avg. education 0.007 0.009 0.007 0.008 0.010 0.008 0.007 0.010 0.007 [1.95]* [2.22]** [1.96]** [2.38]** [2.81]*** [2.40]** [2.19]** [2.62]*** [2.20]** Share of female adults 0.240 0.300 0.241 0.229 0.276 0.231 0.224 0.271 0.225 [4.39]*** [5.28]*** [4.40]*** [4.44]*** [5.10]*** [4.47]*** [4.33]*** [5.00]*** [4.36]*** Number of adult members 0.002 0.006 0.002 0.005 0.009 0.004 0.004 0.008 0.004 [0.28] [0.93] [0.25] [0.73] [1.29] [0.68] [0.68] [1.22] [0.62] Average age of adults 0.006 0.007 0.006 0.006 0.007 0.006 0.006 0.007 0.006 [5.65]*** [5.86]*** [5.67]*** [6.25]*** [6.46]*** [6.28]*** [6.21]*** [6.41]*** [6.23]*** Share of dependents -0.024 -0.035 -0.026 -0.052 -0.061 -0.053 -0.054 -0.064 -0.056 [-0.42] [-0.56] [-0.44] [-0.94] [-1.01] [-0.96] [-0.98] [-1.07] [-1.01] Constant -0.371 -0.317 -0.371 -0.371 -0.314 -0.372 -0.360 -0.301 -0.360 [-5.19]*** [-4.03]*** [-5.19]*** [-5.43]*** [-4.19]*** [-5.44]*** [-5.30]*** [-4.05]*** [-5.31]*** Observations 1924 1924 1924 2099 2099 2099 2106 2106 2106 R2 0.394 0.301 0.395 0.384 0.292 0.385 0.383 0.290 0.384 Cragg-Donald weak inst. F-stat 360.1 55.98 181.6 380.7 65.39 192.9 382.8 65.75 193.9 Panel B. Second Stage Regressions. Deposit account Loan received Loan requested (1) (2) (3) (4) (5) (6) (7) (8) (9) Remittances 0.247 0.359 0.242 -0.003 0.044 -0.005 -0.012 0.039 -0.015 [4.45]*** [3.34]*** [4.39]*** [-0.08] [0.76] [-0.15] [-0.34] [0.66] [-0.40] Adults avg. education 0.042 0.041 0.042 0.009 0.008 0.009 0.008 0.007 0.008 [10.25]*** [9.52]*** [10.28]*** [3.33]*** [3.02]*** [3.34]*** [2.81]*** [2.52]** [2.82]*** Share of female adults 0.047 0.010 0.049 -0.065 -0.079 -0.064 -0.063 -0.079 -0.063 [0.83] [0.15] [0.86] [-2.00]** [-2.26]** [-1.97]** [-1.92]* [-2.22]** [-1.89]* Number of adult members 0.002 0.002 0.002 0.010 0.010 0.010 0.010 0.009 0.010 [0.30] [0.20] [0.31] [2.01]** [1.95]* [2.01]** [1.88]* [1.82]* [1.88]* Average age of adults 0.003 0.002 0.003 -0.000 -0.001 -0.000 -0.001 -0.001 -0.001 [2.61]*** [1.59] [2.65]*** [-0.34] [-0.73] [-0.31] [-0.75] [-1.14] [-0.73] Share of dependents -0.041 -0.035 -0.042 0.071 0.075 0.071 0.052 0.056 0.052 [-0.66] [-0.55] [-0.66] [1.89]* [1.96]** [1.88]* [1.31] [1.40] [1.30] Observations 1924 1924 1924 2099 2099 2099 2106 2106 2106 Log likelihood -646.8 -687.6 -645.5 47.14 43.52 47.09 -23.03 -25.85 -23.15 Over-identification test chi2 - - 1.725 - - 1.126 - - 1.317 Over-identification test p-value - - 0.189 - - 0.289 - - 0.251 31 Table 6: Estimations instrumenting remittances using lagged values of remittances This table shows the results of the regressions instrumenting the likelihood of receiving remittances in given year with our third set of instruments. We show the first stage regressions in Panel A and the second stage in Panel B. We use two lags of the variable Remittances (dummy for whether household receives remittances from overseas in a given year) as instruments. Column (1) shows the results of the regression for the likelihood that the household has a deposit account. Column (2) presents the regressions for the likelihood of receiving a loan, while column (3) shows the regressions for the likelihood of applying for a loan. All independent variables are defined in Table 2. Municipality fixed effects are estimated but not reported. Robust t- statistics are shown in brackets. *,**, and *** denote significance at 10, 5 and 1 percent, respectively Panel A. First Stage Regressions Dependent variable. Remittances: dummy for whether household receives remittances from overseas in a given year. (1) (2) (3) Remittances (t-1) 0.357 0.355 0.357 [6.64]*** [6.56]*** [6.64]*** Remittances (t-2) 0.212 0.215 0.212 [3.45]*** [3.45]*** [3.45]*** Adults avg. education 0.006 0.007 0.006 [0.84] [0.97] [0.84] Share of female adults 0.282 0.292 0.282 [2.96]*** [3.08]*** [2.96]*** Number of adult members 0.003 0.006 0.003 [0.22] [0.43] [0.22] Average age of adults 0.005 0.005 0.005 [2.59]*** [2.69]*** [2.59]*** Share of dependents -0.097 -0.093 -0.097 [-0.84] [-0.81] [-0.84] Constant -0.158 -0.188 -0.158 [-1.03] [-1.22] [-1.03] Observations 749 746 749 R-squared 0.458 0.461 0.458 Cragg-Donald Weak Intruments F-Stat 36.06 35.12 36.06 Panel B. Second Stage Regressions Deposit Loan received Loan requested account (1) (2) (3) Remittances 0.279 -0.067 -0.078 [2.54]** [-1.29] [-1.51] Adults avg. education 0.046 0.018 0.015 [6.11]*** [3.52]*** [2.84]*** Share of female adults 0.089 -0.054 -0.033 [0.89] [-0.85] [-0.52] Number of adult members 0.004 0.003 0.002 [0.29] [0.31] [0.18] Average age of adults -0.002 -0.001 -0.001 [-0.68] [-0.68] [-0.84] Share of dependents 0.002 0.072 0.026 [0.01] [1.03] [0.35] Observations 749 746 749 Log likelihood -327.3 38.73 -2.190 Over-identification test chi2 0.0408 0.671 1.043 Over-identification test p-value 0.840 0.413 0.307 32 Figure 1: Private credit to GDP in Central America (average 1995-2001) 45% 40% 40% 34% 35% 30% 28% 25% 20% 17% 17% 15% 10% 5% 0% El Salvador Costa Rica Guatemala Honduras Nicaragua Source: Central American Executive Secretariat of the Monetary Council and Central Bank of El Salvador. 33