Money transfers among banked and unbanked Mexican immigrants.

1. Introduction

The recognition of the matricula consular as an acceptable alien identification by financial institutions has received considerable attention in the media. (1) Mexican consulates around the world have issued the matricula consular since 1871, and its recent acceptance has allowed Mexican nationals to more easily enter the financial mainstream regardless of immigration status. (2) The Mexican government has advocated its approval, and the U.S. Treasury currently allows the recognition of the matricula consular as a means to grant Mexican immigrants access to the U.S. banking system. In addition to providing lower transaction costs, U.S. banks may encourage immigrants to save and there may be an overall increase in the flow of funds to Mexico.3 In this paper, we test the hypothesis that access to banks helps boost the level of money transfers sent to Mexico by Mexican immigrants in the United States, whether in the form of periodic remittances or money carried personally by Mexican immigrants when returning home. Specifically, we examine (1) trends in the use of banking services by Mexican immigrants, (2) determinants of the use of banking services, and (3) the implications of access to the U.S. banking system on money-transferring practices of Mexican immigrants according to their legal status.

Using data from the Mexican Migration Project from 1970 to 2004, we first document trends in the use of U.S. banking services by Mexican immigrants according to their age and decade of last visit. Given the low usage of financial services in Latin American countries, it is not surprising that Mexican immigrants, and in particular undocumented immigrants, have a relatively low level of interaction with U.S. banks. (4) Nonetheless, it is important to document the portion of immigrants who utilize U.S. banking services, because increasing access to banking services by Mexican immigrants can potentially affect the volume of money transfers by lowering the cost of sending money and by helping immigrants save.

In the second part of the paper, we look at the linkages between banking, remittances, and lump sums brought back home by Mexican immigrants. According to the U.S. State Department and the Central Bank of Mexico, remittances by Mexican nationals to relatives and friends were approximately $10 billion in 2002, or roughly 1.7% of Mexico's GNP. Mexican immigrants without bank accounts often rely on money-wiring agencies that charge more than 10% to transfer funds, while banks charge considerably less (Orozco 2002; Bills 2004). (5) This suggests that, without banks, Mexican families may spend about $1 billion per year to send their remittances. Hence, through their lower transaction costs, improved access to banking services could potentially increase the amount remitted by Mexican immigrants to their home communities. In addition to reducing transaction costs, U.S. banks may encourage Mexican immigrants to save and, potentially, remit more to Mexico. After addressing the potential endogeneity of being banked (having a U.S. bank account) and the fact that a sizable number of immigrants choose not to send money home, we test whether usage of U.S. banking services is associated with differential levels of transfers to Mexico, with a special focus on immigrant legal status.

Our analysis sheds light on the potential effects of the matricula and other identification cards in increasing access to banks on immigrants' remittance flows and on the lump sums they take back to their families and communities at the end of their migration spells. In particular, we find that banking among Mexican immigrants in our sample is limited. Overall, only 9% of our sample had a bank account while living in the United States on his/her latest trip. Undocumented immigrants are significantly less likely to be banked, whereas immigrants who speak English, earn more, stay for longer periods of time in the United States, and bring their spouses along with them to the United States are significantly more likely to be banked. Additionally, institutional and community of origin characteristics seem to play a significant role in the use of banking services by Mexican immigrants, suggesting that the availability of a banking infrastructure back home and immigrants' familiarity with the banking system play an important role in immigrants' decision to be banked while in the United States.

Finally, access to banking appears to have different consequences on the dollar amount remitted on a periodic basis compared to the lump sum immigrants bring back home at the end of their migration spells, with these results varying by immigrants' legal status. While banking does not appear to significantly raise monthly remittance flows by Mexican migrants eventually returning to Mexico--whether temporarily or permanently it helps boost the amount they bring back home. Whereas banked legal immigrants brought back about $1790 more than unbanked legal immigrants, the differential was over $5200 for banked undocumented immigrants relative to unbanked undocumented immigrants.

Our paper is organized as follows. Section 2 discusses the importance of immigrant banking and money transfers and briefly reviews the literature on this subject. In section 3, we describe our data set derived from the Mexican Migration Project survey. Trends in banking, saving, and remitting among Mexican immigrants are documented in section 4. The descriptive analysis motivates our theoretical framework, explained in section 5, as well as our empirical methodology in section 6 of the paper. We discuss the results of our money transfer models in section 7, placing special emphasis on immigrants' money transfer practices according to their legal status. Finally, section 8 concludes the paper.

2. Importance of Banking Access by Immigrants and Review of the Literature

Given the extent of Mexican migration (legal and undocumented) to the United States, (6) learning about the use of banking services by Mexican immigrants and the implications of access to U.S. banks on money transferring practices has become increasingly important for various reasons. First, bringing the "unbanked" into the U.S. banking system allows for a more efficient regulation of currency in the United States and, in the case of undocumented immigrants, lets the government know about the number of undocumented immigrants and their financial activities. Hence, there is a need, first and foremost, to learn about the extent of the "unbanked" population among legal and undocumented Mexican immigrants.

Secondly, and as pointed out earlier, immigrants' access to the U.S. banking system may increase the flow of funds remitted to Mexico by helping immigrants save and by lowering the cost of wiring and transferring money back home. Assessing whether this is the case is of interest given the uses of remittances back in Mexico. Remittances are often sent to meet a variety of needs of non-migrating family members back in Mexico, such as food, rent, health, and educational expenses (Brown and Ahlburg 1999; Cox Edwards, and Ureta 2003). In addition, remittances often serve an economic development purpose by financing the purchase of productive assets in home communities, such as land and businesses (Durand et al. 1996; Belo 2001; Woodruff and Zenteno 2001; Amuedo-Dorantes, Bansak, and Pozo 2005). As a result, policymakers have promoted lower-cost methods of money transfers given the importance of these remittances for economic development. (7) Despite the increasingly competitive market for money transfer services (Handlin, Krontoft, and Testa 2002; Lowell and de la Garza 2002), the costs of remittances can vary widely by type of method, with banks offering a low-cost alternative for money transfers (Orozco 2002). Therefore, integrating the unbanked remitters into the mainstream financial system may have a significant impact on remittances.

Third, immigrants who previously relied heavily on informal remitting methods are now able to access banking services that provide a safer transmission method (Amuedo-Dorantes and Pozo 2005). Bank accounts do not require immigrants to carry large sums of cash (immigrants carrying large sums of cash are often victims of crimes), guarantee a safe transfer of funds to their families back home, and facilitate bill paying through their wide range of wiring, checking, and bill payment services.

Finally, learning about immigrants' banking use is particularly important given banks' ability to engage in financial intermediation. Through their capacity to take deposits, banks not only raise immigrants' ability to save, but aid in the development of a credit history and increase the likelihood that immigrants' savings are channeled into productive investments. In addition, banks may be able to alleviate credit constraints through credit cards, mortgages, and other loans. In this respect, the role played by financial intermediation on economic development cannot be overstated and, consequently, it appears beneficial to enhance immigrants' accessibility to the banking system.

Despite the importance of remittances for numerous receiving countries, there has been little research on the transmission mechanism of remittances (Amuedo-Dorantes and Pozo 2005) and, to our knowledge, no one has directly examined the relationship between banks and the level of remittances. (8) The possible lack of research on the effect of banking on the level of remittances may be due to the relatively low use of banks by those who remit. (9) Alternatively, it may be due to the fact that few data sets include detailed information about Mexican immigrants, their legal status, their banking usage, and level of money transfers.

Given the lack of research on banking and remittances, we examine (1) trends in the use of banking services by Mexican immigrants, (2) determinants of their use of banking services, and (3) the implications of access to the U.S. banking system on their money transferring practices according to their legal status.

3. Data

In order to carry out the study, we use data from the Mexican Migration Project (MMP). The MMP database is the result of a multidisciplinary study of Mexican migration to the United States. Currently, the MMP database includes detailed social, demographic, and economic information from approximately 16,000 households in 107 representative communities in 17 Mexican states. (10) The MMP survey has been carried out annually in the winter months of 1982-1983 and 1987-2004. (11) Two to five Mexican communities are surveyed each year and a sample of approximately 200 households is randomly selected in each community. For each household, a complete life history is gathered for the household head, which includes detailed information on past migration experiences in the United States. After gathering detailed information on these households, interviewers travel to the destination areas in the United States to administer identical questionnaires to households from the same communities in Mexico who have settled in the United States and no longer return home. Altogether, the MMP provides reasonably representative data on authorized and unauthorized Mexican immigrants in the United States (Massey and Zenteno 2000; Munshi 2003; Amuedo-Dorantes, Bansak, and Pozo 2005).

The MMP is one of the richest data sets available for studying Mexican migration to the United States and contains important information on immigrants banking and remittance behavior, in addition to legal status at the time of migration. (12) For the purpose of this study, we use the information collected from more than 4,400 migrating household heads who were interviewed upon their return to Mexico. We choose this sample due to the fact that a large share of Mexican immigrants, particularly those who remit, return to Mexico (Lowell 1992; Lindstrom 1996; Reyes 1997; Orrenius 1999). In addition, the sample design of the MMP is such that return migrants interviewed in Mexico are intended to be a representative group of such immigrants. (13) Lastly, in an effort to minimize recall bias and ensure greater accuracy of responses, we further restrict our sample to those whose last trip to the United States occurred since 1970. (14)

The following three variables constitute the primary focus of our analysis: 1) the funds remitted each month to families back in Mexico, 2) the lump sum taken back home to Mexico, and 3) the ownership of a bank account during an immigrant's most recent trip to the United States. In addition, we account for a variety of personal and household characteristics affecting remitting behavior, regardless of whether it takes place periodically (monthly remittances) or at the end of their migration experience (a lump sum brought back to Mexico). Some of these characteristics include immigrants' personal characteristics (age and gender), immigrants' ability to save and remit money home (as captured by their educational attainment, ability to speak English, monthly earnings, and time in the United States), and the economic needs of family members back home (as captured by whether the presence of a spouse and the percent of household members of non-working age in Mexico). (15)

Finally, we construct variables on immigrants' community of origin and their state of residency while in the United States. We use this information to assess the role that regional development, the existence of a banking network, and the banking culture, among other factors, might play in their banking, saving, and remitting behavior while in the United States. (16) A detailed description of the variables used in our analysis, as well as their means and standard deviations, are included in Appendix A.

An average of 76% of immigrants remitted approximately $190 home on a monthly basis, in real terms. (17) In addition to these monthly remittance payments, about 73% of immigrants in our sample saved and took some money with them back to Mexico at the end of their last U.S. trip. The average amount being brought home equaled $1155 in constant dollars. Immigrants' limited banking use may have affected their overall remitting behavior. Only 9% of our sample had a bank account during their last U.S. trip. This low percentage may be, in part, due to immigrants' undocumented status. Only 31% of immigrants in the MMP were documented during their last U.S. trip. Other interesting demographic characteristics include gender, age, and human capital. Ninety-five percent of our sample is male. On average, immigrants in the sample were 35 years old when they last migrated to the United States, and possessed limited human capital. Average educational attainment is five years of schooling, and only 24% of immigrants spoke English during their last U.S. trip. Additionally, the vast majority of our sample (94%) worked while in the United States, with 84% leaving their spouses behind with a family comprised of mostly non-working age dependents (65%). The average trip duration was two years. In what follows, we examine Mexican immigrants' banking trends and remitting patterns.

4. Banking Trends, Saving, and Remitting Behavior of Mexican Immigrants

To motivate our analysis, we first provide descriptive evidence of trends in the use of banking services by Mexican immigrants. Subsequently, we examine monthly remittances and lump sum transfers by immigrants' banking status.

Trends in the Use of Banking Services

As mentioned earlier in the paper, the share of Mexican immigrants with bank accounts while living in the United States has been limited but is increasing over time. As can be seen in Table 1, only 9% of our sample of Mexican migrant household heads had a bank account during their most recent trip to the United States. However, the use of banking services has increased steadily during the past several decades from 3% (as a share of immigrants) in the 1970s, to 8.5% in the 1980s, to almost 10% in the 1990s. In the period 2000-2004, the share reached close to 20%, although this sample is relatively small. To address potential cohort effects, Table 1 also details the bank use by age groups within any given arrival decade, and the pattern of increased banking usage can still be observed over time. Banking also varies over time by legal status. While not monotonic, the share banked of both undocumented and legal immigrants rose over this 35-year time period. In each decade, the share of legal immigrants with bank accounts was more than four times that of undocumented Mexican immigrants.

Remittances and Lump Sum Transfers by Banking Status

To assess the general patterns of transfers by immigrants to their families in Mexico, we present the average remittances and lump sum transfers by banking status in Table 2. We distinguish between remittances repatriated while in the United States (which we call "remittances") and lump sums brought back to Mexico (which we call "savings returned to Mexico").

For our sample of Mexican immigrants, average remittances are larger among the banked than the unbanked, but this difference is not statistically significant. However, when conditioning on remitting (e.g., remittances > 0), banked return migrants remit $71 more per month than unbanked return migrants, and this difference is statistically different from zero. Turning to the lump sums brought back to Mexico, banked return migrants bring back significantly more than their unbanked counterparts. The difference between the banked and unbanked is $2328, and $3063 when conditioning on the amount brought back to Mexico being nonzero. Thus, there is evidence that, among returnees, banked immigrants transfer more money back to Mexico than unbanked immigrants.

Overall, these figures, while solely differences in means by banking status, motivate the need to consider immigrants' banking access as a primary determinant in their money-transferring patterns.

5. Theoretical Framework

Our goal is to empirically examine whether access to banking is associated with a higher volume of immigrants' monthly remittances and/or a larger one-time lump sum brought back home by immigrants at the end of the migration spell. To illustrate the various ways in which access to banking may affect immigrants' periodic remittances as well as one-time transfers, we use a simple two-period model in which immigrants altruistically care for their families. Specifically, we assume that immigrants' utility depends on their leisure and consumption of a purchased market good, as well as on their families' consumption of goods and leisure. (18) Hence, immigrants' objective function can be described as the sum of their utilities in each period, with [delta] representing their discount rate as follows:

max V = U ([c.sup.1.sub.i],[c.sup.1.sub.f]); [l.sup.1.sub.i],[l.sup.1.sub.f]) + [delta]U ([c.sup.2.sub.i],[c.sup.2.sub.f];[l.sup.2.sub.i],[l.sup.2.sub.f])

The vectors [c.sub.i] and [l.sub.i] stand for immigrants' consumption and leisure, whereas [c.sub.f] and [l.sub.f] represent the consumption and leisure choices of immigrants' families. The utility function has the standard general properties (i.e., increasing in each argument, with negative second partial derivatives). Immigrants maximize this objective function subject to budget and time constraints in each period. In period 1, immigrants choose whether to remit money to their families back home ([R.sub.i]) so as to contribute to their consumption of goods and leisure. For simplicity, we assume that credit constraints prevent immigrants from transferring resources from the second to the first period via borrowing. However, saving ([s.sub.i]) is possible. In period 2, the principal plus its return (i.e., (1 + r)*[s.sub.i]) may be used for consumption or, if not utilized, it may be taken back home as a lump sum ([LS.sub.i]). (19) Finally, immigrants are endowed with T units of time in each period that they can use for leisure or work, with w representing their ongoing market wage. Therefore, normalizing the price of purchased market goods to 1, we can write immigrants' constraints as:

[c.sup.1.sub.i] + (1 + [p.sub.r]) * [R.sup.1.sub.i] + [s.sub.i] = w * (T - [l.sup.1.sub.i])

in period 1, and

[c.sup.2.sub.i] + [LS.sub.i] = w * (T- [l.sup.2.sub.i]) + (1 + r) * [s.sub.i]

in period 2, with: r = r([b.sub.i]) and [p.sub.r] = [p.sub.r]([b.sub.i]).

The price of remitting funds to Mexico, including wiring fees, bank charges, time taken to remit money home, and risk of being a victim of theft if carrying large sums of cash, is captured by [p.sub.r]. The price of remitting funds to Mexico ([p.sub.r]) as well as immigrants' savings ([s.sub.i]) are likely to be affected by whether or not immigrants have access to the U.S. banking system ([b.sub.i]). For instance, banking access may lower the cost of remitting funds to Mexico and/or provide immigrants with access to a variety of savings and investment instruments that may result in greater wealth accumulation, (20) possibly affecting immigrants' periodic remitting patterns as well as the amounts they take back home at the end of their migration spells. Specifically, the problem previously described suggests the following reduced form expressions for immigrants' periodic and one-time transfers to their families:

[R.sup.*.sub.i] = f([p.sub.r]([b.sub.i]),r([b.sub.i]),[delta]w) and [LS.sup.*.sub.i] = g([p.sub.r]([b.sub.i]),r([b.sub.i]),[delta],w). (1)

In this paper, we are interested in empirically assessing the signs of [partial derivative][R.sup.*.sub.i]/[partial derivative][b.sub.i] and [partial derivative][LS.sup.*.sub.i]/[partial derivative][b.sub.i], both of which are, a priori, ambiguous. For instance, focusing on immigrants' periodic remitting patterns, the possibility exists that access to banking lowers the cost of remitting funds to Mexico (i.e., [partial derivative][p.sub.r]/[partial derivative][b.sub.i] < 0). In that event, there would be an income effect predicting that immigrants (now enjoying a larger disposable income) will remit more money to their families. Furthermore, there would be a substitution effect suggesting a substitution away from consumption towards remitting due to the lower cost of remitting funds to Mexico, therefore enhancing the consumption and leisure of immigrants' families. Jointly, both effects would predict that: [partial derivative][R.sup.*.sub.i]/[partial derivative][p.sub.r] * [partial derivative][p.sub.r]/[partial derivative][b.sub.i] > 0.

However, access to banking may not only lower the cost of remitting, but also increase immigrants' return to their savings (i.e., [partial derivative]r/[partial derivative][b.sub.i] > 0). Once more, we would have an income effect predicting an increase in immigrants' remittances following their augmented disposable incomes. Nonetheless, the higher return to immigrants' savings would also raise the opportunity cost of remitting money on a periodic basis, pushing immigrants towards saving and taking back home the principal plus the accumulated return at the end of the migration spell. This negative substitution effect may be particularly strong among Mexican immigrants who go back home on a frequent basis, and consequently may be able to wait to bring their accumulated savings to their families back home with them upon their return. Therefore, it is unclear whether immigrants' periodic remittances should increase with immigrants' access to banking (that is; [partial derivative][R.sup.*.sub.i]/[partial derivative][b.sub.i] > or < 0 since: [partial derivative][R.sup.*.sub.i]/[partial derivative]r * [partial derivative]r/[partial derivative][b.sub.i] > or < 0).

By the same token, without further assumptions, it is unclear whether immigrants' access to banking will raise or reduce the lump sum they bring back home at the end of their migration spells. Given the potential ambiguity in the signs of the effects of banking on immigrants' monthly remittances and one-time lump sum transfers to their families, we empirically assess their signs in the following sections.

6. Empirical Methodology

Before examining the impact that banking may have on our sample of Mexican immigrants, it is worth noting that our inferences about the effect of banking on money transfers to Mexico are derived from the group of Mexican immigrants who return to Mexico at some point in their lives, whether temporarily or permanently. As noted by others in the literature (Gubert 2002), this could be an issue if our intent were to make inferences about the effect of banking on the money transfers of the overall Mexican population. However, to the extent that we are interested in learning about the impact of access to banking on Mexican immigrants, modeling the migration decision is not crucial. Moreover, since returnees are the ones most likely to engage in transferring money back to Mexico and, furthermore, constitute the vast majority of Mexican immigrants to the United States due to the proximity between the two countries, focusing on this population should provide us with useful insights about the potential effect of banking on money transfers from Mexican immigrants in the United States (Lowell 1992; Lindstrom 1996; Reyes 1997; Orrenius 1999).

Given that a large share of immigrants in our sample (approximately 25%) do not send money home, we propose the following Tobit model for estimating the effects of banking (b) on Mexican immigrants' money transfers to their families (denoted by T). This model is applied whether these refer to monthly payments (R) or to the lump sum brought back home at the end of the migratory experience (LS):

[T.sup.*.sub.ifc] = [beta][b.sub.ifc] + [X'.sub.ifc][phi] + [[epsilon].sub.ifc],

with:

[T.sub.ifc] = 0 if [T.sup.*.sub.ifc] [less than or equal to] 0, [T.sub.ifc] = 0 if [T.sup.*.sub.ifc] < 0, (2)

and where [[epsilon].sub.ifc] ~ N(0, [[sigma].sup.2]), i = migrating household head, f = family in Mexico, and c = community in Mexico. In addition to immigrants' access to banking ([b.sub.ifc]), the model in Equation 2 accounts for a variety of immigrants' personal, family, and community-of-origin characteristics (included in vector [X.sub.ifc] and discussed later) possibly affecting their remittances as well as lump-sum payments taken back home at the end of their migration spells.

Challenges in the Empirical Analysis

A couple of issues regarding the model in Equation 2 are worth noting at this point.

Censoring. First, as noted in Equation 2, some of the values for our dependent variable are observed (i.e., those cases in which Mexican immigrants remitted or took money back home), whereas the remaining values (i.e., those instances in which Mexican immigrants do not remit nor take money back home) are censored or unobserved. When some of the data are censored, the distribution that applies to the sample data is a mixture of discrete and continuous distributions, rendering the use of OLS inappropriate.

Two general types of estimation methods are typically used under such circumstances. One of them is a two-part selection model, where the likelihood to remit or bring money back home is first modeled using a probit model. The second stage is estimated by OLS and uses the predicted values from the first-stage probit to correct for the ongoing selection. The advantage of the two-part selection model is that its allows for the regressors to have a different effect on the likelihood of transferring money versus on the dollar amount finally remitted or brought back home. The major disadvantage is that the results tend to be quite sensitive to the identification exclusions, which are often disputable given the difficulty of finding factors that affect the decision to remit or bring money back home, but do not influence the dollar amount transferred home by the immigrant.

An alternative often used when modeling immigrants' money-transferring practices is the Tobit model (Brown 1997; Ravallion and Dearden 1988; Schrieder and Knerr 2000). The Tobit takes into account the diverse nature of the distribution of immigrants' money transfers by modeling the likelihood of remitting or bringing money back home and the dollar amount finally transferred as a function of the same covariates. A potential disadvantage of the Tobit model is that a change in any regressor will have the same effect (same sign) on both the probability of remitting or bringing money back home and on the dollar amount finally transferred home. (21) Nonetheless, recognizing (a) the difficulty of conceiving appropriate identifiers that affect the decision to remit or bring money back home without influencing the dollar amount transferred home by the immigrant, and (b) the sensitivity of the findings to the choice of identifiers, we view the estimation of Equation 2 as a Tobit model as preferable.

Endogeneity. The possibility exists that immigrants' banking status in Equation 2 is endogenous to their money-transferring behavior. The endogeneity of immigrants' banking status may be originating from various sources. First, omitted variables affecting Mexican immigrants' money-transferring practices, such as precise information on their wealth, may be correlated with immigrants' banking status in the United States As a result, our estimates of the effect of banking on immigrants' money-transferring practices are likely to be biased. A second source of endogeneity of banking is its potential simultaneity of banking with remitting as well as with saving (to bring money back home). Mexican immigrants planning to remit or bring money back home may choose to become banked during their stay in the United States given the advantages of remitting through banks, such as greater safety and lower transferring costs, and the saving and wealth accumulation process often facilitated through banking.

Final Empirical Specification

We take into account the two econometric challenges previously discussed. To first address the potential endogeneity of immigrants' banking status, we rely on the method of instrumental variables and estimate the predicted probability of being banked using the following probit model:

P([b.sub.ifc] = 1/[Z.sub.ifc]) = [PHI]([z'.sub.ifc][gamma]) + [u.sub.ifc], where [u.sub.ifc]|[Z.sub.ifc] ~ N(0,1) (3)

and where the vector [Z.sub.ifc]. includes immigrants' personal, family, and community-of-origin characteristics in vector [X.sub.ifc]. Additionally, the vector [Z.sub.ifc]. includes information on other factors possibly affecting immigrants' accessibility to banks, such as whether immigrants were familiar with banking before migrating to the United States, as well as two sets of dummy variables, indicating their states of origin in Mexico as well as the state in the United States where they resided during their last U.S. trip. These dummy variables are chosen to account for differences in the institutional and economic environment in which immigrants operate, both at home in Mexico as well as in the United States, as typified by the banks' outreach efforts to the unbanked, banking fees, and any bank-related use taxes, among many others. In this regard, we check the significance of the first-stage effects of our selected instruments and find that these regional and banking dummies are jointly significant in explaining migrants' likelihood of being banked during the last visit to the United States. (22)

To help identify Equation 3, information on whether immigrants were familiar with banking before they migrated and the two sets of dummies indicative of immigrants' state of origin and state of residence while in the United States are excluded from Equation 2. These variables are chosen as factors that likely affect immigrants' money transfers only through their effect on the probability of having a bank account in the United States, which we account for when modeling migrants' money transfer practices through their predicted probability of being banked. The intuition behind our choice of our instruments is as follows: Familiarity with banking before migrating to the United States increases the immigrant's likelihood of opening a bank account in the United States. Having a bank account in the United States will then increase both the probability of accumulating more wealth and the likelihood that immigrants will use banks to remit money back home. In addition, being banked may also raise the dollar amount sent back or brought back to Mexico, given the lower cost of remitting money through banks and the ongoing accumulation of savings in U.S. banks.

Something similar can be said with respect to the use of state dummies as instruments. However, to the extent that these state dummies may be indicative of potential regional differences in income, wealth, and other characteristics possibly influencing immigrants' money-transferring practices (in addition to their likelihood of being banked), we include measures of the standard of living (a proxy for the economic development) of immigrants' states of origin in Mexico. Additionally, we examine whether remittances are correlated with per capita personal income by U.S. state. (23) We are unable to find any significant relationship between these two variables. Therefore, given their correlation with banking, we use this set of U.S. state dummies as a reasonable choice for instrumenting immigrants' banking likelihood along with immigrants' familiarity with banking. At any rate, given the potential sensitivity of instrumental variable regression analysis to the variables chosen as instruments of the endogenous regressors, we carry out the analysis with and without instrumenting for migrants' banking use while in the United States to assess the robustness of our results to the various specifications.

After estimating the probit model specified in Equation 3, we then estimate the following Tobit model:

[T.sup.*.sub.ifc] = [beta][P.sub.Banked] + [X'.sub.ifc] + [[epsilon].sub.ifc], with [T.sub.ifc] = max (0, [T.sup.*.sub.ifc] and [[epsilon].sub.ifc] | [P.sub.Banked] [X'.sub.ifc] ~ N(0, [[sigma].sup.2])

where [P.sub.Banked] represents immigrants' predicted probability of having a bank account during their last trip to the United States. (24) As noted earlier, the goal of our empirical work is to estimate the effect of banking on the amount transferred back home. The overall banking effect results both from banks being able to offer lower-price remitting services ([p.sub.r](b)) and through banks' role in the facilitation of the accumulation of savings (s(b)). Under such circumstances, the use of banking services may affect both periodic remitting patterns and the accumulated savings brought back home by immigrants upon their return to their origin communities.

What other factors may affect immigrants' remittances and their one-time transfers at the close of their migration experiences? We consider a variety of immigrants' personal, family, and community-of-origin characteristics possibly correlated with immigrants' banking. These characteristics are included in the vector [X.sub.ifc] so as to better isolate the effect of banking on immigrants' remitting patterns. For instance, immigrants' money transfers are likely to depend on their income and ability to send funds abroad--as proxied by w in our reduced form expressions (Equation 1). As a proxy for income, we include real monthly earnings while in the United States, educational attainment, and ability to speak English. To capture immigrants' ability to send money to their families, we include information on their legal status while in the United States. Undocumented immigrants might encounter greater difficulties in finding employment than their legal counterparts and, even when employed, they are subject to a greater economic uncertainty than legal immigrants. Hence, they might experience greater difficulty in remitting or taking a lump sum back home than their documented counterparts.

Immigrants' remitting patterns and their decision to take money back home is also likely to be affected by their families' income back home. We use two variables to capture economic need as a proxy for immigrants' family income back in Mexico. First, we introduce a dichotomous variable indicative of whether an immigrant's spouse stayed back in Mexico. Secondly, we control for the percent of non-working age family members left back home as a measure of economic dependence. For both variables, we expect to find a positive relationship with the likelihood and the dollar amount remitted or taken back home by the migrant in the United States.

Finally, we include a variety of demographic, migratory, community, and economic characteristics possibly affecting immigrants' money-transferring behavior. Gender and age are important demographic controls, since men of working age are still more likely to be the main breadwinners in immigrants' families, making them more likely to remit and remit larger sums of money back home than, for example, female immigrants of non-working age. Additionally, we account for the duration of immigrants' trip to the United States to address the potential remittance decay that takes place as immigrants become more assimilated to and settled in the host country. A variety of community characteristics likely to affect immigrants' money transfers to their families are also included in the analysis. These characteristics include the population density of immigrants' communities of origin as well as two dichotomous variables indicative of the standard of living in immigrants' states of origin. The population density captures the rural versus urban nature of immigrants' origin communities. In turn, the dummy variables indicative of the standard of living in immigrants' states of origin allow us to control for unobserved income, wealth, and other macroeconomic characteristics that may be influencing immigrants' money-transferring practices. Lastly, we include information on the bilateral Mexican U.S. exchange rate as an economic characteristic possibly affecting immigrants' remitting behavior. Indeed, when an immigrant remits money home via a bank or money transfer firm, s/he sends U.S. dollars and the recipient collects Mexican pesos. As such, the exchange rate may play a significant role in the migrant's decision to remit or bring money back home.

7. Mexican Immigrants' Remitting Behavior

Table 3 displays the results from estimating a Tobit model of the dollar amount remitted by Mexican immigrants on a monthly basis. The model is estimated both with and without instrumenting for immigrants' banked status while in the United States. The instrumental variable results make use of the predicted probability of being banked derived from a probit model estimating the likelihood of having a bank account during their last U.S. trip. According to the probit estimates in Appendix B, undocumented immigrants are approximately 7.3 percentage points less likely to be banked than their legal counterparts. Similarly, Mexican immigrants who migrate and leave their spouses behind appear to be four percentage points less likely to be banked than those who did not (e.g., their single counterparts and their married counterparts who bring their spouses to the United States). This result may be simply signaling these immigrants' intent to return back to Mexico in the near future and their lack of interest in familiarizing themselves with the U.S. banking system. In contrast, Mexican immigrants who speak English, enjoy higher monthly earnings, and stay longer periods of time in the United States are significantly more likely to be banked.

Finally, institutional and community-of-origin characteristics seem to play a significant role in the use of banking by Mexican immigrants. The two sets of dummy variables indicative of immigrants' state of origin and of their state of residence in the United States demonstrate the importance of regional characteristics influencing immigrants' decision to open a bank account during their last U.S. trip. (25) While the coefficients on these individual dummy variables have been suppressed in the table due to space limitations, tests of joint significance indicate that these dummy variables help explain the probability of being banked. Furthermore, Mexican immigrants originating from communities familiar with banks before they migrated to the United States are 3.5 percentage points more likely to be banked in the United States. Familiarity with banks is defined as a dichotomous variable equal to one if the immigrant's community of origin (before the individual migrated to the United States) had more than the average number of banks of all communities in our sample.

These results shed light on the important role played by the availability of a banking infrastructure back home--and, therefore, immigrants' familiarity with the banking system--in immigrants' decisions to be banked while in the United States. Overall, the probit model for immigrants' likelihood of being banked during their last U.S. stay is identified through (a) its distinct functional form with respect to the Tobit models in Table 3 and Table 4, and (b) the use of statistically different from zero information on the familiarity with banking in their communities of origin before migrating to the United States, as well as on other information characteristic of their communities of origin and their U.S. states of residence.

Mexican Immigrants' Monthly Remitting Patterns

Table 3 shows the results from estimating a Tobit model examining the determinants of Mexican immigrants' monthly remittance volume with and without instrumenting for immigrants' bank use while in the United States. Note that the coefficients in the Tobit model measure the partial effects of changes in our regressors on the expected value of the latent variable ([T.sup.*.sub.ifc]). However, the variable we wish to understand better is the observed income transfer ([T.sub.ifc]). Therefore, in addition to the estimated coefficient and the standard errors, Table 3 includes information on the effect of our independent variables on the probability of remitting on a periodic basis, as well as information on the sensitivity of monthly remittances to changes in the regressors.

Some of the general factors positively linked to immigrants' periodic remitting patterns are worth noting. For instance, regardless of whether we carry out the analysis with or without instrumenting for immigrants' banking use while in the United States, male immigrants and immigrants leaving spouses behind are up to 31 and 18 percentage points more likely to remit monthly to their families back in Mexico, respectively, than female immigrants and immigrants who do not leave their spouses behind. In particular, male immigrants and immigrants who leave spouses in Mexico remit an average of $91-$93 and $60-$64 more, respectively, on a monthly basis in real terms than their female and either single or married counterparts migrating jointly with their spouses. Similarly, immigrants' earnings potential, as proxied by their ability to speak English and reflected by their monthly earnings, is also directly linked to their monthly remittances. Indeed, immigrants' monthly earnings significantly raise their remittances. Yet the significance of these two factors in shaping Mexican migrants' remitting patterns disappears once we instrument for their banking use in the United States. In fact, migrants' educational attainment appears to dampen their remittances. This finding may be explained by immigrants' initial motivation for remitting money home. For instance, more educated migrants may come from families with lesser economic need and, as such, may remit less if they are doing so for altruistic reasons. Alternatively, more educated migrants may endure less economic uncertainty while in the United States and, consequently, may remit less to their families if their primary reason to remit money home is for insurance purposes (e.g., in case their migration experience does not go well). Finally, in consonance with the earlier literature on remittances (Zarate-Hoyos 2004) as well as journalistic sources (Kraul 2001), we observe lower remittances being sent to family members in urban areas in Mexico (as proxied by the population density in the community of origin) regardless of whether or not we instrument for immigrants' banking use. Likewise, the two models in Table 3 reveal that Mexican immigrants remit more when the U.S. dollar strengthens, as indicated by a higher bilateral exchange rate.

Turning our attention to the effect of banking, Table 4 displays the effect that immigrants' banking and legal status may have on their likelihood to transfer money to their families and, in that event, on the amount finally sent home with and without instrumenting for immigrants' banking use during their last U.S. trip (models 1 and 2, respectively). These effects are computed using the coefficients, as well as the effects of immigrants' banking and legal status on their probabilities of transferring money back home and on the dollar amount finally transferred, in Table 3. Additionally, Table 4 displays the corresponding joint significance tests. Overall, the results in Table 4 show the lack of statistical significance of Mexican immigrants' banking and legal status on their monthly remitting practices, whether we instrument for their banking usage or not. In particular, according to the figures in rows 3 and 4, Mexican immigrants in our sample with access to banking do not appear to be more likely to remit, nor do they seem to remit a greater amount than their unbanked counterparts. In what follows, we examine the possibility that access to banking instead results in an increase in the dollar amount saved and brought back home by immigrants at the end of their migration spells.

Lump Sums Brought Back by Mexican Immigrants at the End of their U.S. Migration

Focusing on a second type of income repatriation, Table 5 displays the results from examining the determinants of the level of funds taken back home by Mexican immigrants at the end of their migration experiences. Some of the determinants of the dollar amount brought back home by immigrants at the end of their migration spell, whether or not we address the endogeneity of their banking status, include their age, whether or not they left a spouse behind, population density in their states of origin, and the Mexican-U.S. bilateral exchange rate. In particular, older Mexican immigrants, immigrants who leave their spouse in Mexico, and those coming from relatively more urban areas (as captured by the population density in the origin) are significantly less likely to bring money back home than their younger counterparts, Mexican immigrants who do not leave spouses behind, and Mexican immigrants from more rural communities, respectively. The magnitude of these effects on the dollar amount brought back home is relatively large in the case of Mexican immigrants who leave spouses behind, who bring back an average of $317 less (or $448 if we do not address the endogeneity of their banking status) than those who do not leave their spouses in Mexico. As indicated by the figures in Table 3, return migrants leaving spouses behind instead appear to remit on a periodic basis. Additionally, the possibility exists that immigrants with spouses left back in Mexico may be more likely to cross back and forth between Mexico and the United States with a greater frequency than those who are single or migrate with their spouses to the United States. As a result, they may have accumulated less money and, hence, bring less money back home. Likewise, as in Table 3, return migrants continue to be more likely to take money back home when their communities are rural. However, unlike for monthly remittances, an increase in the bilateral foreign exchange rate hampers Mexican immigrants' money repatriation practices and reduces the dollar amount saved and brought back home at the end of the U.S. migration experience. The possibility exists that, when the U.S. dollar is stronger, immigrants opt to send money home on a monthly basis given the still-reasonable money sum received by their families back home after the monetary conversion. As such, they have fewer savings to bring back home.

Other determinants of the dollar amount brought back home by Mexican immigrants at the end of their migration spells lose their statistical significance as we address the endogeneity of immigrants' banking status while in the United States. This is the case with Mexican immigrants' English proficiency and with the duration of their last U.S. stay, both of which appear to raise the real dollar amount brought back home by approximately $191 if they speak English, and by $9 per additional month in the United States in the model without IVs. When we instrument for immigrants' banking use, English proficiency and duration are not significant in the Tobit model. This implies that their effects on lump sum transfers are solely through their effect on the probability of banking, which is captured in the first stage probit.

Turning to the impact of banking on immigrants' likelihood to bring part of their savings back home as well as on the dollar amount finally repatriated, the figures in Table 6 show the generally robust significance of banking on immigrants' money transfer practices, whether we use an instrumental variable approach or not. Access to banking increases the lump sum brought back home by undocumented (relative to legal) immigrants at the end of their migration spells. According to the figures from model (1), the undocumented do not appear to bring back home more than their legal counterparts when both are unbanked (row 1). However, this sum rises to a statistically different-from-zero sum of $552 when we compare banked and undocumented return migrants to banked and legal return migrants (row 2). The aforementioned differences between the lump sum transfers of undocumented and legal immigrants are further accentuated as we instrument for immigrants' banking use (model 2, row 2). In that case, whereas undocumented and legal immigrants do not appear to bring back home different sums when unbanked, their money-transferring practices significantly differ if they are both banked, with the undocumented bringing back home, on average, $3,392 more than their legal counterparts.

The importance of access to banking among the undocumented is once more highlighted when comparing the lump sums brought back home by banked versus unbanked undocumented return migrants (rows 3). For instance, when we instrument for immigrants' banking use, banked and undocumented immigrants are not only twice as likely to save and take back home some of their savings relative to their unbanked counterparts, but they also repatriate an average of $5,222 more than their unbanked equivalents in constant dollars (model 2, row 3). A similar effect is found among legal immigrants, who increase their likelihood of bringing money back home at the end of their migration spell when banked relative to when they are unbanked by up to 38 percentage points (model 2, row 4). Correspondingly, banked and legal immigrants appear to bring back home up to $1790 more than their unbanked and legal counterparts. In sum, while undocumented return migrants with access to banking do not appear to remit more on a periodic basis than their unbanked counterparts, they do seem to bring back home a larger dollar sum at the end of their migration experiences than unbanked and undocumented return migrants. Hence, the possibility exists that undocumented return migrants with access to banking forgo their periodic remittances and, instead, choose to save and bring back a lump sum upon their return home.

8. Conclusions

The recognition of the matricula consular as an acceptable alien identification by financial institutions has been surrounded by substantial debate. If accepted widely, the matricula may reduce the difficulties of living in the United States for undocumented Mexican immigrants to a large degree by facilitating access to the U.S. banking system. Once banked, undocumented immigrants may be more likely to save as well as remit part of their savings home, whether on a monthly basis through access to the low-cost wiring and money-transfer services offered by banks, or as a lump-sum payment taken back home in person upon return to Mexico. The proponents of the card's acceptance seek to boost the remittances sent home by helping immigrants save and by providing them with a safer transmission mechanism. However, the possibility also exists that undocumented immigrants redirect their savings and increase their accumulated savings brought back home or to boost their consumption while in the United States.

In this paper, we use data on more than 4,400 migrating household heads interviewed upon their return to Mexico from the Mexican Migration Project (MMP). We choose this sample due to the fact that a large share of Mexican immigrants, particularly those who remit, return to Mexico, whether temporarily or permanently at some point in their lives (Lowell 1992; Lindstrom 1996; Reyes 1997; Orrenius 1999). Furthermore, the sample design of the MMP is such that return migrants interviewed in Mexico are intended to be a representative group of such immigrants. Focusing on this group of Mexican immigrants, we examine the hypothesis that access to banking services increases the level of funds sent back home by Mexican immigrants while in the United States.

We find that despite a steady increase over the past 30 years, bank usage by Mexican immigrants during their U.S. stay remains fairly low, particularly among those who return to Mexico within a short period of time. Low banking use by Mexican immigrants may be due to their undocumented status, low earnings, and/or distrust toward the financial system. In particular, we find that undocumented immigrants are less likely, by almost eight percentage points, to be banked than documented Mexican immigrants in our sample. Therefore, the acceptance of the matricula consular by financial institutions may potentially bring in a large number of unbanked (the undocumented) into the U.S. banking system, and this development may have large effects on immigrants' money-repatriating practices.

Finally, access to banking has very different consequences on the dollar amount remitted home on a periodic basis and on the lump sum brought back home at the end of the migration spell by immigrants, depending on their legal status. Specifically, while banking use does not seem to significantly raise monthly remittance flows regardless of immigration status, it boosts undocumented and legal immigrants' savings and the real dollar amount brought back home upon their return by over $5200 and by approximately $1790, respectively. That is, immigrants appear to forgo their periodic remittances and, instead, significantly increase their savings and the lump sum repatriated at the end of their migration spells.

In sum, the findings suggest that facilitating undocumented immigrants' access and use of the banking system may increase the future flow of funds brought back by immigrants to their communities at the end of their migration spells. Given the potential role played by remittances in financing productive investments in immigrants' origin communities, marketing and financial literacy efforts targeted at income-repatriating immigrants may result in greater comfort with and usage of low-cost financial services in the United States and in Latin America. These efforts may prove valuable in fostering economic growth in Mexico.

Appendix A
Description of Variables Used in the Analysis

Variables                         Definition

Dependent variables
  Likelihood of remitting on      Likelihood of remitting money to
    a monthly basis                 Mexico on a monthly basis
                                    during their last U.S. trip

  Dollar amount remitted          The real dollar amount remitted
    monthly                         on a monthly basis to Mexico
                                    during their last U.S. trip
                                    (includes nonremitters)

  Likelihood of saving and        Likelihood of saving and bringing
    bringing money back home        money to Mexico at the end of
                                    their last U.S. trip

  Dollar amount saved and         The real dollar amount saved and
    brought back home               brought back to Mexico at the
                                    end of their last U.S. trip
                                    (includes those who do not
                                    bring back savings)

Independent variables
  Banked                          Dummy equal to 1 if migrant had
                                    a bank account during the last
                                    U.S. trip

  Undocumented                    Dummy equal to 1 if migrant
                                    lacked proper documentation at
                                    time of last entry

  Male                            Gender dummy

  Age                             Age at time of last migration to
                                    the United States

  Years of education              Years of educational attainment

  Speaks English                  Dummy equal to 1 if migrant
                                    spoke English during last U.S.
                                    trip

  Monthly earnings in the         Monthly earnings during their last
    United States                   U.S. trip in real dollars
                                    (adjusted for inflation using
                                    CPI-U(82-84))

  Duration of last trip to the    Duration of last U.S. trip in
    United States                   months

  Left spouse in Mexico           Dummy variable for leaving
                                    a spouse in Mexico

  Dependents in Mexico            Percent of HH members of non-
                                    working age back in Mexico

  Community population            Number of people in migrant's
                                    origin community

  Lowest standard of living       Standard of Living dummy
    in origin                       created using the classification
                                    presented by INEGI at http://
                                    www.inegi.gob.mx/prod serv/
                                    contenidos/espanol/niveles/
                                    The lowest level includes the
                                    states of. Guerrero and Oaxaca.

  Second lowest standard of       Standard of Living dummy for
    living in origin                the states of. Hidalgo, Puebla,
                                    and San Luis de Potosi

  Exchange rate                   Mexican peso per U.S. dollar
                                    foreign exchange rate: Federal
                                    Reserve Bank (Series: FRB:
                                    Exchange Rate G.5)

  Bank familiarity                Dummy equal to 1 if there was
                                    a greater than average number
                                    of banks in the community of
                                    origin before migrating to the
                                    United States

  Employed while in United        Dummy variable for working
    States                          while in United States

  Aguascalientes                  Mexican State dummy
  Baja California del Norte       Mexican State dummy
  Colima                          Mexican State dummy
  Chihuahua                       Mexican State dummy
  Durango                         Mexican State dummy
  Guanajuato                      Mexican State dummy
  Guerrero                        Mexican State dummy
  Hidalgo                         Mexican State dummy
  Jalisco                         Mexican State dummy
  Michoacan                       Mexican State dummy
  Nayarit                         Mexican State dummy
  Nuevo Leon                      Mexican State dummy
  Oaxaca                          Mexican State dummy
  Puebla                          Mexican State dummy
  San Luis Potosi                 Mexican State dummy
  Sinaloa                         Mexican State dummy
  Zacatecas                       Mexican State dummy
  Alabama                         U.S. state dummy
  Alaska                          U.S. state dummy
  Arizona                         U.S. state dummy
  Arkansas                        U.S. state dummy
  California                      U.S. state dummy
  Colorado                        U.S. state dummy
  Connecticut                     U.S. state dummy
  Delaware                        U.S. state dummy
  District of Columbia            U.S. state dummy
  Florida                         U.S. state dummy
  Georgia                         U.S. state dummy
  Idaho                           U.S. state dummy
  Illinois                        U.S. state dummy
  Indiana                         U.S. state dummy
  Iowa                            U.S. state dummy
  Kansas                          U.S. state dummy
  Kentucky                        U.S. state dummy
  Louisiana                       U.S. state dummy
  Maryland                        U.S. state dummy
  Massachusetts                   U.S. state dummy
  Michigan                        U.S. state dummy
  Minnesota                       U.S. state dummy
  Mississippi                     U.S. state dummy
  Missouri                        U.S. state dummy
  Montana                         U.S. state dummy
  Nebraska                        U.S. state dummy
  Nevada                          U.S. state dummy
  New Jersey                      U.S. state dummy
  New Mexico                      U.S. state dummy
  New York                        U.S. state dummy
  North Carolina                  U.S. state dummy
  Ohio                            U.S. state dummy
  Oklahoma                        U.S. state dummy
  Oregon                          U.S. state dummy
  Pennsylvania                    U.S. state dummy
  South Carolina                  U.S. state dummy
  Tennessee                       U.S. state dummy
  Texas                           U.S. state dummy
  Utah                            U.S. state dummy
  Virginia                        U.S. state dummy
  Washington                      U.S. state dummy
  Wisconsin                       U.S. state dummy
  Wyoming                         U.S. state dummy
  Puerto Rico                     U.S. state dummy
  United States, state unknown    U.S. state dummy

Variables                              Mean             SD

Dependent variables
  Likelihood of remitting on          0.7611         0.4265
    a monthly basis

  Dollar amount remitted            193.1972       266.3643
    monthly

  Likelihood of saving and            0.7288         0.4446
    bringing money back home

  Dollar amount saved and           1154.849       3989.131
    brought back home

Independent variables
  Banked                              0.0887         0.2843

  Undocumented                        0.6872         0.4637

  Male                                0.9496         0.2189

  Age                                34.9353        12.2692

  Years of education                  5.3493         3.9145

  Speaks English                      0.2412         0.4279

  Monthly earnings in the           936.3478       2952.868
    United States

  Duration of last trip to the       26.4992        55.0836
    United States

  Left spouse in Mexico               0.8411         0.3656

  Dependents in Mexico                0.6497         0.2155

  Community population              77102.88      209638.50

  Lowest standard of living           0.0359         0.1861
    in origin

  Second lowest standard of           0.1234         0.3290
    living in origin

  Exchange rate                       2.9973         3.2147

  Bank familiarity                    0.1562         0.3631

  Employed while in United            0.9425         0.2327
    States
  Aguascalientes                      0.0268         0.1616
  Baja California del Norte           0.0352         0.1843
  Colima                              0.0266         0.1608
  Chihuahua                           0.0539         0.2258
  Durango                             0.0517         0.2214
  Guanajuato                          0.1781         0.3826
  Guerrero                            0.0204         0.1414
  Hidalgo                             0.0145         0.1196
  Jalisco                             0.1319         0.3384
  Michoacan                           0.1164         0.3207
  Nayarit                             0.0305         0.1720
  Nuevo Leon                          0.0143         0.1186
  Oaxaca                              0.0189         0.1363
  Puebla                              0.0207         0.1423
  San Luis Potosi                     0.1001         0.3002
  Sinaloa                             0.0175         0.1310
  Zacatecas                           0.1427         0.3498
  Alabama                             0.0007         0.0260
  Alaska                              0.0002         0.0150
  Arizona                             0.0212         0.1441
  Arkansas                            0.0011         0.0336
  California                          0.5249         0.4994
  Colorado                            0.0129         0.1127
  Connecticut                         0.0005         0.0212
  Delaware                            0.0002         0.0150
  District of Columbia                0.0020         0.0450
  Florida                             0.0277         0.1643
  Georgia                             0.0122         0.1097
  Idaho                               0.0187         0.1356
  Illinois                            0.0778         0.2679
  Indiana                             0.0056         0.0749
  Iowa                                0.0002         0.0150
  Kansas                              0.0083         0.0910
  Kentucky                            0.0011         0.0336
  Louisiana                           0.0005         0.0212
  Maryland                            0.0005         0.0212
  Massachusetts                       0.0005         0.0212
  Michigan                            0.0032         0.0561
  Minnesota                           0.0007         0.0260
  Mississippi                         0.0011         0.0336
  Missouri                            0.0023         0.0474
  Montana                             0.0007         0.0260
  Nebraska                            0.0016         0.0397
  Nevada                              0.0165         0.1273
  New Jersey                          0.0027         0.0520
  New Mexico                          0.0086         0.0920
  New York                            0.0178         0.1323
  North Carolina                      0.0120         0.1087
  Ohio                                0.0016         0.0368
  Oklahoma                            0.0092         0.0957
  Oregon                              0.0099         0.0991
  Pennsylvania                        0.0120         0.1087
  South Carolina                      0.0043         0.0653
  Tennessee                           0.0020         0.0450
  Texas                               0.1511         0.3582
  Utah                                0.0032         0.0561
  Virginia                            0.0025         0.0498
  Washington                          0.0083         0.0910
  Wisconsin                           0.0011         0.0336
  Wyoming                             0.0072         0.0847
  Puerto Rico                         0.0007         0.0260
  United States, state unknown        0.0032         0.0561

Source: Mexican Migration Project (MMP107).

Appendix B
Probit Model for the Likelihood of Being Banked (a)

Variables                     Coefficient     SE         Marginal
                                                         Effect

Undocumented                  -0.5646 ***    0.0870     -0.0730
Male                          -0.0003        0.1773     -2.82E-05
Age                            0.0004        0.0037      5.47E-05
Years of education             0.0158        0.0108      0.0017
Speaks English                 0.5427 ***    0.0833      0.0757
Monthly earnings in
  the United States            1.75E-05 *    9.14E-06    1.89E-06
Duration of last trip          0.0112 ***    0.0014      0.0012
Duration of last
  trip squared                -1.66E-05 ***  3.59E-06   -1.79E-06
Left spouse in Mexico         -0.2861 ***    0.0885     -0.0359
Dependents in Mexico           0.0992        0.1703      0.0107
Size of population in
  community of origin         -2.13E-07      2.47E-07   -2.30E-08
Lowest level standard
  of living in origin          0.0065        0.3016      0.0007
Second lowest level
  standard of living in
  origin                      -0.1072        0.4032     -0.0108
Banking in community
  of origin                    0.2794 **     0.1283      0.0349
Exchange rate
  (peso/dollar)                0.0338 **     0.0140      0.0037

Regression fit statistics
  Number of observations            2978
  LR chi-square (48)                 509.16
  Prob>chi-square                      0.0000
  Log likelihood                    -702.92615
  Percent banked                       0.0887
  Predicted probability
    of being banked                    0.0990

The regressions include a constant. The reference categories
for the standard of living dummies are Mexican states with
standards of livings included in the third, fourth or fifth
category of living standards as specified in Appendix A. The
reference category for the state of interview in Mexico is
Zacatecas and for the U.S. state where they lived during
their last trip to the United States is California.

* Signifies statistically different from zero at the 10%
level or better.

** Signifies statistically different from zero at the 5%
level or better.

*** Signifies statistically different from zero at the 1%
level or better.

We thank Ed Balsdon, David Bjerk, Eric Brunner, Donna Ginther, Richard Fry, Shumalit Kahn, Terra Mckinnish, Susan Pozo, Steven Raphael, Sarah Senesky, Betsey Stevenson, Madeline Zavodny, and participants at the 2004 ASSA and SOLE meetings, as well as attendees at a Federal Reserve Bank of Atlanta seminar for their helpful comments on previous drafts of this paper. This project was supported by a Research, Scholarship, and Creative Activity grant from San Diego State University.

Received July 2005; accepted February 2006.

References

Amuedo-Dorantes, Catalina, Cynthia Bansak, and Susan Pozo. 2005. On the remitting pattern of immigrants: Evidence from Mexican survey data. Economic Review Q1:37-58.

Amuedo-Dorantes, Catalina, and Susan Pozo. 2005. On the use of differing money transmission methods by Mexican immigrants. International Migration Review 39(3):554-77.

Bean, Frank D., Rodolfo Corona, Rodolfo Tuiran, Karen A. Woodrow-Lafield, and Jennifer Van Hook. 2001. Circular, invisible, and ambiguous immigrants: Components of differences in estimates of the number of unauthorized Mexican immigrants in the United States. Demography 38:411-22.

Becket, Gary S. 1981. A Treatise of the Family. enlarged edition. Cambridge: Harvard University Press.

Belo, Pedro J. 2001. "Comments at Remittances as a Development Tool: A Regional Conference Inter-American Development Bank." Accessed 2 April 2004. Available at http://www.iadb.org/mif/website/static/en/remit.asp. Bills, Steve. 2004. Wells plans even more remittance expansion. American Banker, 3(23) December 3, p. 1.

Brown, Richard P. C. 1997. Estimating remittance functions for Pacific Island migrants. World Development 25(4):613-26.

Brown, Richard P. C. and Dennis A. Ahlburg. 1999. Remittances in the South Pacific. International Journal of the Social Economics 26:325-14.

Chiuri, Maria Concetta. 2000. Individual decisions and household demand for consumption and leisure. Research in Economics 54:277-324.

Cox Edwards, Alejandra, and Manuelita Ureta. 2003. Income transfers and children's schooling: Evidence from El Salvador. Journal of Development Economics, Special Issue 72(2):429-61.

Dinerstein, Marti. 2003. IDs for illegals: The 'matricula consular' advances Mexico's immigration agenda. Center for Immigration Studies, Backgrounder, pp. 1-11.

Durand, Jorge, William Kandel, Emilio A. Parrado, and Douglas S. Massey. 1996. International migration and development in Mexican communities. Demography 33(2):249-64.

Gubert, Flore. 2002. Do migrants insure those who stay behind? Evidence from the Kayes area (western Mali). Oxford Development Studies 30(3):267-87.

Hall, P., and S. R. Wilson. 1991. Two guidelines for bootstrap hypothesis testing. Biometrics 47:757-62.

Handlin, Elizabeth, Margrethe Krontoft, and William Testa. 2002. Remittances and the unbanked. Chicago Fed Letter, Number 175a, Special Issue, pp. 1-4.

Inter American Development Bank. 2002. "Survey of Remittance Senders: U.S. to Latin America Second Regional Conference on Impact of Remittances as a Development Tool." Accessed 11 April 2004. Available at http:// www.iadb.org/mif/website/static/en/remit03.asp.

Kraul, Chris. 2001. Latino immigrants sending less money home. Los Angeles Times, 17 December, p. C1.

Lindstrom, David P. 1996. Economic opportunity in Mexico and return migration from the United States. Demography 33(3):357-74.

Lowell, Lindsay B. 1992. Circular mobility, migrant communities, and policy restrictions: Unauthorized flows from Mexico. In Migration, population structure, and redistribution policies, edited by C. Goldscheider. Boulder, CO: Westview, pp. 137-57.

Lowell, Lindsay B., and Rodolfo O. de la Garza. 2002. The development role of remittances in U.S. Latino communities and in Latin American countries. In Sending money home, edited by R. de la Garza and B. L. Lowell. Oxford, England: Rowman and Littlefield Publishers, pp. 3-27.

Massey, Douglas S., and Rene Zenteno. 2000. A validation of the ethnosurveys: The case of Mexico-U.S. migration. International Migration Review 34(3):766-93.

Munshi, Kaivan. 2003. Networks in the modern economy: Mexican migrants in the U.S. labor market. The Quarterly Journal of Economics 118(2):549-99.

Orozco, Manuel. 2002. Remittances, costs and market competition. Unpublished paper, Inter-American Dialogue.

Orrenius, Pia M. 1999. Return migration from Mexico: Theory and evidence. Ph.D. dissertation, University of California, Los Angeles.

Perez, Evan. 2003. North Carolina credit union banks on Latino immigrants. The Wall Street Journal, 9 May, p. B1.

Porter, Eduardo. 2003. Banks can accept foreign IDs. The Wall Street Journal, 19 September, p. A12.

Ravallion, Martin, and Lorraine Dearden. 1988. Social Security in a 'moral economy': An empirical analysis for Java. Review of Economics and Statistics 70(1):36-44.

Reyes, Belinda. 1997. Dynamics of immigration: Return migration to western Mexico. San Francisco, CA: Public Policy Institute of California.

Schrieder, Gertrud, and Beatrice Knerr. 2000. Labour migration as a Social Security mechanism for smallholder households in sub-Saharan Africa: The case of Cameroon. Oxford Development Studies 28(2):223-36.

Sherraden, Michael, Mark Schreiner, and Sondra Beverly. 2003. Income, institutions, and saving performance in individual development accounts. Economic Development Quarterly 17(1):95-112.

U.S. Treasury. 2003a. Fact sheet: Final regulations implementing consumer identity verification requirements under Section 326 of the U.S. Patriot Act. Washington, DC: Office of Public Affairs.

U.S. Treasury. 2003b. Fact sheet: Results of the notice of inquiry on final regulations implementing consumer identity verification requirements under Section 326 of the U.S. Patriot Act. Washington, DC: Office of Public Affairs.

Woodruff, Christopher, and Rene Zenteno. 2001. Remittances and microenterprises in Mexico. UCSD, Graduate School of International Relations and Pacific Studies Working Paper.

Wooldridge, Jeffrey M. 2003. Introductory econometrics. 2nd edition. Cincinnati, OH: Thomson South-Western.

Zarate-Hoyos, German A. 2004. Consumption and remittances in migrant households: Towards a productive use of remittances. Contemporary Economic Policy 22(4):555-65.

(1) In July 2002, the U.S. Treasury issued a notice of proposed rulemaking to implement the requirements of section 326 of the PATRIOT Act. The rule requires (among other things) that financial institutions develop a Customer Identification Program (CIP) that implements procedures to collect identifying information about customers opening an account (U.S. Treasury, 2003a, 2003b). This rule allows financial institutions to accept, in particular, the matricula consular, which is often used by undocumented immigrants to open bank accounts (Porter 2003). Even prior to the acceptance of the matricula consular, however, banks did not need a social security number or tax ID for an individual to open a bank account, as long as it was not interest bearing (Handlin, Krontoft, and Testa 2002).

(2) To obtain the card, a Mexican national living abroad needs to supply proof of Mexican citizenship (such as a certified copy of a birth certificate or another official ID issued by any Mexican or U.S. authority) to a Mexican consulate, which will issue a matricula consular valid for 5 years for approximately $28 (Dinerstein 2003). As of September 2003, the matricula consular card was accepted by 280 financial institutions, over 1000 local law enforcement agencies, and more than a dozen state motor vehicle agencies (Porter 2003).

(3) The U.S. Treasury, along with the international aid community, has been interested in finding ways to channel the sizable remittance funds to promote economic development in Mexico (Dinerstein 2003).

(4) Due in part to a lack of deposit insurance, currency devaluations and bank failures, only about 5%-30% of households in Latin America holds a basic checking account (Inter-American Development Bank 2002).

(5) For example, the Latino Community Credit Union in Durham, North Carolina, charges only $10 to wire $1000 (Perez 2003).

(6) As of March 2000, there were an estimated 8.39 million Mexican-born persons living in the United States, of whom approximately 3.9 million were unauthorized (Bean et al. 2001).

(7) For example, the IADB supports increasing the level of remittances to underdeveloped countries by reducing the cost of transfers, given that remittance flows to Latin America are substantially higher than the total of official development assistance to the region (Handlin, Krontoft, and Testa 2002). Specifically, the IADB suggests that costs can be reduced by increasing competition among providers of money transfer services and promoting the use of formal financial systems, thereby decreasing the cost of sending remittances. The latter is analyzed in this paper.

(8) In their paper, Amuedo-Dorantes and Pozo (2005) explore the choice of transmission mechanism and find that undocumented immigrants are less likely to use banking services. In contrast, more educated and more skilled immigrants are more likely to use U.S. banks for repatriating earnings.

(9) Researchers have identified several difficulties in promoting formal financial systems as a method of remittance transmission. First, the recipients of remittances are often reluctant to deal with banks in Latin America because the banking system is seen as corrupt (Handlin, Krontoft, and Testa 2002). Second, there is confusion over the proper documents needed to open an account. Contrary to widespread belief, banks do not need a social security number or tax ID for an individual to open a bank account, as long as it is not interest bearing (Handlin, Krontoft, and Testa 2002). Lastly, additional barriers, such as the cost of closing an account in the United States and the lack of communication between U.S. and Mexican banks, may also hinder the level of banking among Mexican migrants.

(10) As of the MMP 107, the sample covers communities in the states of Aguascalientes, Baja California Norte, Chihuahua, Colima, Durango, Guanajuato, Guerrero, Hidalgo, Jalisco, Michoacan, Nayarit, Nuevo Leon, Oaxaca, Puebla, San Luis Potosi Sinaloa, and Zacatecas.

(11) The MMP interviews were conducted in communities of various size, ethnic composition, and economic development that are typical source regions for migrants bound for the United States. In addition, the sample expands over time to incorporate communities in newer sending states.

(12) It should be noted that the remitting patterns of those who do not return or "stayers" would be quite different from those who return to Mexico. However, due to the lack of representativeness of stayers in the MMP, we do not include them and the results of our study do not reflect the behavior of this group of immigrants.

(13) For further information regarding the methodology and survey design of the MMP107, see http://mmp.opr.princeton.edu/databases/studydesign-en.aspx.

(14) Given that the survey has been conducted since 1982, it is not surprising that migrants' most recent trips have largely been taken over the past 30 years. According to the authors' calculations, approximately 80% of those in the MMP107 sample who migrated to the United States did so between 1970 and 2002.

(15) Most of the variables for the study came from the MMP107 Migration, Community and Household files. These datasets and more information on the Mexican Migration Project can he found at http://mmp.opr.princeton.edu.

(16) Unfortunately, we are unable to directly measure the level of communication between U.S. banks and those in place in migrants' communities of origin in Mexico.

(17) All transfer and earnings data have been inflation-adjusted using the CPI-U (82-84).

(18) This is one of the ways in which altruism can be modeled. That is, migrants may be assumed to care about their families' consumption of goods and services, independently of their families' preferences (Chiuri 2000). This form of altruism is called "paternalism." Other models of altruism consider that individuals derive utility from their families' utilities, even if their consumption of goods and services is not what the individuals might have desired for them. This alternative form of altruism is called "caring" (Becker 1981).

(19) We assume that migrants only remit during the first period to allow for a higher discounting of the lump sums taken back home by migrants at the end of their migration experience relative to monies sent on a periodic basis to their families.

(20) While neoclassical theory does not predict that an increase in access to an interest-bearing financial instrument will necessarily lead to increased savings, the possibility exists that increases in access, information, and facilitation increase savings among immigrants as was found in the case of IDAs (individual deposit accounts) for the poor (Sherraden, Schreiner, and Beverly 2003).

(21) A second potential disadvantage of the Tobit and two-part selection models is their reliance on normality and homoscedasticity in the latent variables. However, as noted by Wooldridge (2003), neither conditional normality nor heteroskedasticity affect the unbiasedness or consistency of the OLS estimates and, as a result, for reasonable deviations from these assumptions, the Tobit model still provides good estimates.

(22) In particular, the joint significance test of the instruments in the banking equation yield the following chi- square statistic: Chi-square(38) = 68.57, with: Prob > chi-square = 0.0017.

(23) For this analysis, state-level income data were extracted from the Bureau of Economic Analysis (Series: CA1-3 for 1990) and state-level remittances were constructed from the MMPI07 to correspond to the state where the immigrant resided during their last U.S. trip.

(24) Standard errors of the estimates are computed using the bootstrap procedure suggested by Hall and Wilson (1991).

(25) These results are not shown due to space limitations, but are available from the authors upon request.

Catalina Amuedo-Dorantes, San Diego State University, Department of Economics, San Diego, CA 92182-4485, USA; E-mail: camuedod@ mail.sdsu.edu; corresponding author.

Cynthia Bansak, San Diego State University, Department of Economics, San Diego, CA 92182-4485, USA; E-mail: cbansak@ mail.sdsu.edu.

Table 1. Trends in the Use of Banking Services by Immigrants (a)

                                             Standard
                                     Mean    Deviation

Overall banked                       0.089     0.284
Decade last visited United States
1970s                                0.030     0.172
  Undocumented                       0.016     0.125
  Legal                              0.093     0.292
  Younger than 20                    0.027     0.164
  20s                                0.034     0.182
  30s                                0.041     0.198
  40s                                0.017     0.128
  50s and older                      0.016     0.125
1980s                                0.086     0.280
  Undocumented                       0.051     0.221
  Legal                              0.209     0.407
  Younger than 20                    0.089     0.287
  20s                                0.110     0.314
  30s                                0.068     0.252
  40s                                0.073     0.261
  50s and older                      0.079     0.270
1990s                                0.098     0.297
  Undocumented                       0.036     0.187
  Legal                              0.181     0.385
  Younger than 20                    0.183     0.389
  20s                                0.082     0.275
  30s                                0.105     0.307
  40s                                0.108     0.310
  50s and older                      0.079     0.270
2000-2004                            0.187     0.391
  Undocumented                       0.074     0.263
  Legal                              0.303     0.463
  Younger than 20                    0.400     0.548
  20s                                0.128     0.337
  30s                                0.200     0.405
  40s                                0.280     0.458
  50s and older                      0.083     0.289

(a) Authors' tabulations using the MMP107.

Table 2. Remittances and Lump Sum Transfers of
Migrants by Banking Status (a)

                                     Unconditional

Migrant                                                t
Characteristics       N      Mean     Difference   Statistic

All migrants
Remittances
  Unbanked         3290      192.93
  Banked            318      200.69      -7.76       -0.48
Savings
returned
to Mexico
  Unbanked         2984      955.25
  Banked            285     3283.47   -2328.22       -4.38 ***

                  Conditional on Remitting (Remittances > 0)

Migrant                                           t
Characteristics      N        Mean   Difference   Statistic

All migrants
Remittances
  Unbanked         2545      249.4
  Banked            199      320.7       -71.30   -3.39 ***
Savings
returned
to Mexico
  Unbanked         2177    1309.35
  Banked            214    4372.84     -3063.49   -4.42 ***

(a) Authors' tabulations using the MMP107. Remittances and savings
returned to Mexico are inflation-adjusted.

*** Signifies statistically different from zero at the 1% level or
better.

Table 3. Tobit Model for the Dollar Amount Remitted Monthly (a)

                            Model (1)--Without IVs

                                         Standard
                                         Error the
Variables              Coefficient           SE

Dollar amount
  remitted monthly

Banked                   -3.8726          29.1125
Undocumented             14.4282          16.1677
Banked *
  undocumented           22.2580          45.5837
Male                  228.6299 ***        33.0794
Age                      0.4840            0.6347
Years of
  education              -3.2624           1.9393
Speaks English         50.5977 ***        16.2965
Monthly
  earnings in
  the United
  States               0.0054 ***          0.0020
Duration of last
  trip                   -0.2538           0.3215
Duration of last
  trip (2)               0.0002            0.0010
Left spouse in
  Mexico              136.2646 ***        17.6645
Dependents in
  Mexico                 31.4126          29.2347
Population
  density in
  origin               -0.0002 ***         0.0000
Lowest living
  standard in
  origin                -39.0107          30.4231
Second-lowest
  living
  standard in
  origin                 -27.347          18.9838
Exchange rate          5.5355 ***          2.0393

Regression Fit Statistics

Number of
  observations            2928
(Left) censored
  observations             690
LR chi-square
  (16)                   174.57
Prob>chi-
  square                 0.0000
Log likelihood          -16641.48

                             Model (1)--Without IVs

                        Partial           Partial
                        Effect on         Effect
                      Probability          on the
                        of Being        Conditional
Variables              Uncensored       Expectation

Dollar amount
  remitted monthly

Banked                   -0.0044          -1.8426
Undocumented             0.0162            6.8509
Banked *
  undocumented           0.0246           10.8155
Male                     0.2771           90.8476
Age                      0.0005            0.2309
Years of
  education              -0.0037          -1.5565
Speaks English           0.0554           24.7642
Monthly
  earnings in
  the United
  States                6.07E-06           0.0026
Duration of last
  trip                   -0.0003          -0.1211
Duration of last
  trip (2)              1.71E-07           0.0001
Left spouse in
  Mexico                 0.1602           60.2638
Dependents in
  Mexico                 0.0352           14.9872
Population
  density in
  origin                -1.84E-07         -0.0001
Lowest living
  standard in
  origin                 -0.0448          -18.0371
Second-lowest
  living
  standard in
  origin                 -0.0311          -12.8123
Exchange rate            0.0062            2.6410

Regression Fit Statistics

Number of
  observations
(Left) censored
  observations
LR chi-square
  (16)
Prob>chi-
  square
Log likelihood

                              Model (2)--With Ivs

                                         Bootstrap
Variables              Coefficient           SE

Dollar amount
  remitted monthly

Banked                  128.3099           97.474
Undocumented             1.7420           20.2420
Banked *
  undocumented          171.9979          161.2104
Male                  233.3662 ***        29.2460
Age                      0.4026            0.5736
Years of
  education            -3.9525 ***         1.9015
Speaks English           19.0567          17.3452
Monthly
  earnings in
  the United
  States                 0.0041            0.0070
Duration of last
  trip                 -0.8693 **          0.3916
Duration of last
  trip (2)               0.0013            0.0012
Left spouse in
  Mexico              141.4107 ***        19.2781
Dependents in
  Mexico                 13.7824          31.7137
Population
  density in
  origin               -0.0002 ***        3.32E-05
Lowest living
  standard in
  origin                -29.4325          20.6252
Second-lowest
  living
  standard in
  origin                -23.3387          17.0686
Exchange rate           3.9167 ***         1.9516

Regression Fit Statistics

Number of
  observations            2801
(Left) censored
  observations             668
LR chi-square
  (16)                   199.29
Prob>chi-
  square                 0.0000
Log likelihood         -15700.881

                              Model (2)--With Ivs

                        Partial           Partial
                        Effect on         Effect
                     the Probability      on the
                        of Being        Conditional
Variables              Uncensored       Expectation

Dollar amount
  remitted monthly

Banked                   0.1530           62.6679
Undocumented             0.0021            0.8503
Banked *
  undocumented           0.2051           84.0056
Male                     0.3053           93.2533
Age                      0.0005            0.1965
Years of
  education              -0.0047          -1.9304
Speaks English           0.0225            9.4040
Monthly
  earnings in
  the United
  States                4.86E-06           0.0020
Duration of last
  trip                   -0.0010          -0.4246
Duration of last
  trip (2)              1.54E-06           0.0006
Left spouse in
  Mexico                 0.1785           63.5323
Dependents in
  Mexico                 0.0164            6.7315
Population
  density in
  origin                -2.12E-07         -0.0001
Lowest living
  standard in
  origin                 -0.0359          -14.0150
Second-lowest
  living
  standard in
  origin                 -0.0282          -11.207
Exchange rate            0.0047            1.9130

Regression Fit Statistics

Number of
  observations
(Left) censored
  observations
LR chi-square
  (16)
Prob>chi-
  square
Log likelihood

(a) We refer to the conditional expectation as the expectation
conditional on both remitting back home and other regressors. The
regressions include a constant, and the three upper living standard
categories are used as reference given the small number of
observations in our sample that fall within each category. Remittances
and monthly earnings are inflation-adjusted. *** Signifies
statistically different from zero at the 1 % level or better, ** at
the 5% level or better, and * at the 10% level or better.

Table 4. The Effects of Being Banked on the Dollar Amount
Remitted Monthly by Migrant Legal Status'

                                                       Joint
                                                    Significance
                                                    (Chi-Square
Group                Computation      Coefficient    Statistic)

Model (1)--Without Ivs

1 Unbanked:
  undocumented
  vs. legal        [[beta].sub.2]       14.4282         0.80

2 Banked:
  undocumented    ([[beta].sub.2] +
  vs. legal        [[beta].sub.3])      36.6862         0.71

3 Undocumented:
  banked vs.      ([[beta].sub.1] +
  unbanked         [[beta].sub.3])      18.3854         0.14

4 Legal: banked
  vs. unbanked     [[beta].sub.1]       -3.8726         0.02

Model (2)--With IVs

1 Unbanked:
  undocumented
  vs. legal        [[beta].sub.2]       1.7420          0.01

2 Banked:
  undocumented    ([[beta].sub.2] +
  vs. legal        [[beta].sub.3])     173.7399         1.61

3 Undocumented:
  banked vs.      ([[beta].sub.1] +
  unbanked         [[beta].sub.3])     300.3078         2.89

4 Legal: banked
  vs. unbanked     [[beta].sub.1]      128.3099         1.73

                      Partial
                       Effect          Partial
                       on the           Effect
                    Probability         on the
                      of Being        Conditional
Group                Uncensored       Expectation

Model (1)--Without Ivs

1 Unbanked:
  undocumented
  vs. legal            0.0162           6.8509

2 Banked:
  undocumented
  vs. legal            0.0408           17.6664

3 Undocumented:
  banked vs.
  unbanked             0.0202           8.9729

4 Legal: banked
  vs. unbanked         -0.0044          -1.8426

Model (2)--With IVs

1 Unbanked:
  undocumented
  vs. legal            0.0021           0.8503

2 Banked:
  undocumented
  vs. legal            0.2072           84.5859

3 Undocumented:
  banked vs.
  unbanked             0.3581          146.6735

4 Legal: banked
  vs. unbanked         0.1530           62.6679

(a) (T = [[beta].sub.1]Banked + [[beta].sub.2]Undocumented +
[[beta].sub.3]Banked * Undocumented + X' [phi] + [epsilon]) +
[epsilon]). We refer to the 'conditional' expectation as the
expectation conditional on both remitting back home and other
regressors. Remittances and savings returned to Mexico are
inflation-adjusted.

*** Signifies statistically different from zero at the 1% level
or better, ** at the 5% level or better, and * at the 10% level
or better.

Table 5. Tobit Model for the Dollar Amount Saved
and Brought Back Home (a)

                                Model (1)--Without IVs

Variables                      Coefficient         SE

Dollar Amount Saved
   and Brought Back Home
  Banked                      1850.6920 ***       448.7753
  Undocumented                 244.0318           254.2401
  Banked * undocumented       1133.9970 *         705.5931
  Male                         477.2218           498.5768
  Age                          -19.9377 **         10.0721
  Years of education            12.2346            29.8124
  Speaks English               494.4999 **        256.1718
  Monthly earnings
    in the United
    States                       0.0378             0.0303
  Duration of last trip         25.0305 ***         4.8768
  Duration of
    last trip (2)               -0.0572 ***         0.0143
  Left spouse in Mexico      -1128.8980 ***       271.5064
  Dependents in Mexico         453.5924           456.8454
  Population density
    in origin                   -0.0038 ***         0.0006
  Lowest living
    standard in
    origin                      65.4781           467.2145
  Second-lowest
    living standard
    in origin                  -36.7404           293.1653
  Exchange rate               -114.4684 ***        32.4508

Regression Fit Statistics
  Number of observations                      2729
  (Left) censored
    observations                               633
  LR chi-square (16)                           199.77
  Prob>chi-square                                0.0000
  Log likelihood                            -21196.002

                                   Model (1)--Without IVs

                                   Partial
                                   Effect        Partial
                                    on the       Effect
                              Probability         on the
                                  of Being    Conditional
Variables                       Uncensored    Expectation

Dollar Amount Saved
   and Brought Back Home
  Banked                             0.1482      771.2048
  Undocumented                       0.0201       92.0867
  Banked * undocumented              0.0920      459.7592
  Male                               0.0394      176.3275
  Age                               -0.0016       -7.5664
  Years of education                 0.0010        4.6431
  Speaks English                     0.0406      191.0300
  Monthly earnings
    in the United
    States                         3.11E-06        0.0143
  Duration of last trip              0.0021        9.4992
  Duration of
    last trip (2)                 -4.72E-06       -0.0217
  Left spouse in Mexico             -0.0921      -448.396
  Dependents in Mexico               0.0374      172.1398
  Population density
    in origin                     -3.14E-07       -0.0014
  Lowest living
    standard in
    origin                           0.0054       24.9410
  Second-lowest
    living standard
    in origin                        -0.003      -13.9194
  Exchange rate                     -0.0094      -43.4411

Regression Fit Statistics
  Number of observations
  (Left) censored
     observations
  LR chi-square (16)
  Prob>chi-square
  Log likelihood

                                Model (2)--Without IVs

                                               Bootstrap
Variables                     Coefficient         SE

Dollar Amount Saved
   and Brought Back Home
  Banked                      4746.262 ***       2247.523
  Undocumented                -105.581            443.6802
  Banked * undocumented       9098.211 **        5468.917
  Male                         493.7221           780.622
  Age                          -23.30263 **         9.0815
  Years of education            -5.3752            31.4662
  Speaks English               -72.5584           375.3437
  Monthly earnings
    in the United
    States                       0.0013             0.0960
  Duration of last trip          9.6958             7.1716
  Duration of
    last trip (2)               -0.0329 **          0.02411
  Left spouse in Mexico       -813.1941 ***       365.019
  Dependents in Mexico         335.0528           427.2412
  Population density
    in origin                   -0.0038 ***         0.0008
  Lowest living
    standard in
    origin                     296.5602           229.4889
  Second-lowest
    living standard
    in origin                  255.8215           270.4204
  Exchange rate               -134.416 ***         39.2254

Regression Fit Statistics
  Number of observations                     2601
  (Left) censored
    observations                              619
  LR chi-square (16)                          184.42
  Prob>chi-square                               0.0000
  Log likelihood                           -20110.332

                                 Model (2)--Without IVs

                                Partial
                                Effect          Partial
                                on the          Effect
                             Probability        on the
                               of Being       Conditional
Variables                     Uncensored      Expectation

Dollar Amount Saved
   and Brought Back Home
  Banked                             0.3809     1790.3652
  Undocumented                      -0.0085      -39.9212
  Banked * undocumented              0.7301     3431.9893
  Male                               0.0397      181.3002
  Age                               -0.0019       -8.7901
  Years of education                -0.0004       -2.0276
  Speaks English                    -0.0058      -27.3011
  Monthly earnings
    in the United
    States                         1.00E-07        0.0005
  Duration of last trip              0.0008        3.6574
  Duration of
    last trip (2)                 -2.72E-06       -0.0128
  Left spouse in Mexico             -0.0649     -316.5849
  Dependents in Mexico               0.0269      126.3872
  Population density
    in origin                     -3.02E-07       -0.0014
  Lowest living
    standard in
    origin                           0.0237      113.6961
  Second-lowest
    living standard
    in origin                        0.0205       97.6466
  Exchange rate                     -0.0108      -50.7038

Regression Fit Statistics
  Number of observations
  (Left) censored
     observations
  LR chi-square (16)
  Prob>chi-square
  Log likelihood

(a) We refer to the 'conditional' expectation as the expectation
conditional on both remitting back home and other regressors. The
regressions include a constant and the three upper living standard
categories are used as reference given the small number of
observations in our sample that fall within each category.
Savings returned to Mexico and monthly earnings are
inflation-adjusted. *** Signifies statistically different
from zero at the 1% level or better, ** at the 5% level or
better, and * at the 10% level or better.

Table 6. The Effects of Being Banked on the Dollar Amount
Saved and Brought Back Home by Migrant Legal Status (a)

                                                          Joint
                                                       Significance
                                                       (Chi-square
Group                 Computation       Coefficient     Statistic)

Model (1)--Without IVs

1 Unbanked:
  undocumented
  vs. legal         [[beta].sub.2]       244.0318          0.92

2 Banked:
  undocumented     ([[beta].sub.2] +
  vs. legal         [[beta].sub.3]      1378.0288 *        2.47

3 Undocumented:
  banked vs.       ([[beta].sub.1] +
  unbanked          [[beta].sub.3])    2984.689 ***       20.97

4 Legal: banked
  vs. unbanked                         1850.6920 ***      17.01

Model (2)--With Ivs

1 Unbanked:
  undocumented
  vs. legal         [[beta].sub.2]       -105.581          0.06

2 Banked:
  undocumented     ([[beta].sub.2] +
  vs. legal         [[beta].sub.3])     8992.63 **         6.99

3 Undocumented:
  banked vs.       ([[beta].sub.1] +
  unbanked          [[beta].sub.3])    13844.473 ***      13.15

4 Legal: banked
  vs. unbanked      [[beta].sub.1]      4746.262 **        4.46

                        Partial           Partial
                     Effect on the     Effect on the
                      Probability       Conditional
                       of Being         Expectation
Group                 Uncesourced

Model (1)--Without IVs

1 Unbanked:
  undocumented
  vs. legal             0.0201            92.0867

2 Banked:
  undocumented
  vs. legal             0.1121           551.8459

3 Undocumented:
  banked vs.
  unbanked              0.2402           1230.9640

4 Legal: banked
  vs. unbanked          0.1482           771.2048

Model (2)--With Ivs

1 Unbanked:
  undocumented
  vs. legal             -0.0085          -39.9212

2 Banked:
  undocumented
  vs. legal             0.7216           3392.0681

3 Undocumented:
  banked vs.
  unbanked              1.1110           5222.3545

4 Legal: banked
  vs. unbanked          0.3809           1790.3652

(a) T=[[beta].sub.1]Banked + [[beta].sub.2] Undocumented +
[[beta].sub.3] Banked * Undocumented + X' [phi] + [epsilon].
We refer to the 'conditional' expectation as the expectation
conditional on both remitting back home and other regressors.

*** Signifies statistically different from zero at the 1%
level or better, ** at the 5% level or better, and
* at the 10% level or better. Remittances and savings returned
to Mexico are inflation-adjusted.