The Financial Assimilation of Immigrant Families: Intergeneration and Legal Differences DISSERTATION

The Financial Assimilation of Immigrant Families: Intergeneration and Legal Differences DISSERTATION Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University By Qifan Wang Graduate Program in Economics The Ohio State University 2012 Dissertation Committee: Lucia Dunn, Advisor Stephen Cosslett Sherman Hanna

Copyrighted by Qifan Wang 2012

Abstract This dissertation empirically examines economic issues related to the behavior of immigrants in the areas of debt, default and holding of various types of assets. It is divided into three main parts using two different new sources of data related to immigrants. In the first part, using the Consumer Finance Monthly (CFM), I am able to determine if an immigrant has been both born and raised abroad (referred to as a real immigrant ); or has been born abroad but raised from the age of 12 in the U.S. (referred to as the 1.5 generation ); or has been born and raised in the U.S. by parents who were both born abroad (referred to as second generation). My major research questions or hypotheses address whether the different immigrant and nonimmigrant groups differ in (1) type and amount of debt holdings (i.e., credit card, other consumer loans, payday loans, mortgage debt, etc.); (2) default history on the different types of debt; and (3) amount of asset holdings in various categories (savings accounts, stocks, bonds, liquid vs. non-liquid, etc.). The important independent variables are the institutional settings in their home countries, which let us understand if differences in home legal environments affect the adjustment of immigrants to the average non-immigrant behaviors. Another hypothesis in this part of the dissertation is whether the institutional settings of the home countries impact economic behaviors in the U.S. and the effect change between generations. ii

The second and third parts of the dissertation examine whether the debt and asset behaviors, as well as the homeownership rate, differ by immigrants legal status. These two parts of the dissertation have important policy implications about illegal immigrants and the consequences of economics field, using the New Immigrant Survey (NIS). My major research questions or hypotheses address whether immigrants from different parts of the world and entering the U.S. under various visa statuses differ in (1) credit card debt and mortgage debt holdings (these are the categories available in the NIS); (2) amount and type of assets (savings accounts, stocks, bonds, liquid vs. non-liquid, etc.); and (3) their home buying behaviors. I find that illegal immigrants who become legalized through amnesty programs do not behave differently in debt, default and bankruptcy behaviors, and are only slightly less likely to become homeowners when compared with legal immigrants. The results suggest that the amnesty law successfully assimilate the illegal immigrants in their financial behaviors. Since the purpose of amnesty programs is to provide an opportunity for the once illegal immigrants to get assimilated to the society, my results provide some evidence in the economic field to support the success of the amnesty programs. iii

Dedication Dedicated to my wife and my parents iv

Vita 1985... Born in Jianhu, China 2007... B.S. Economics, Minor in Applied Mathematics, Renmin University of China 2008... M.A. Economics, The Ohio State University 2008 to present... Graduate Teaching Associate, Department of Economics, The Ohio State University Fields of Study Major Field: Economics v

Table of Contents Abstract... ii Dedication... iv Vita... v List of Tables... ix List of Figures... xi Chapter1. the U.S The first, the second, and the one in between: The debt behavior of Immigrants in 1.1 Introduction... 1 1.1.1 Background... 1 1.1.2 Three groups of immigrants... 2 1.1.3 Research questions... 3 1.1.4 Outline... 4 1.2 Literature review... 5 1.2.1 Immigration literature... 5 1.3 Model and estimation... 6 1.3.1 Theoretical Model... 6 1.3.2 Empirical Model... 11 1.3.3 Estimation strategy... 12 1.4 Data... 14 1.4.1 Comparison of immigrant-relevant data... 14 1.4.2 The CFM data... 14 1.4.3 Descriptive statistics... 16 1.5 Results... 16 1.5.1Results on default behavior... 16 1.5.2 Results on access to credit... 19 1.6 Conclusions... 21 1.6.1 The conservative Real immigrants... 21 1.6.2 The Americanized 2 nd generation... 21 1.6.3 The myth of the 1.5 th generation... 22 Chapter2. Debt Behavior of Immigrants in the U.S: The Impact of Legal Status upon Entry vi

2.1 Introduction... 33 2.1.1 Illegal immigrants in the U.S... 33 2.1.2 Definition of New Immigrants... 33 2.1.3 Outline... 34 2.2 Empirical Model... 35 2.3 Data... 36 2.4 Descriptive Statistics... 37 2.5 Results... 44 2.6 Conclusion and Discussion... 45 Chapter3. upon Entry The Homeownership Rate among Immigrants in the U.S: The Impact of Legal Status 3.1 Introduction... 55 3.1.1 Background... 55 3.1.2 Outline... 57 3.2 Theories and Literature Review... 57 3.2.1 General homeownership literature... 57 3.2.2 The Institutional Setting argument: Theory and Literature... 58 3.2.3 English Language Literature... 61 3.2.4 Commitment to the country... 62 3.3 Model and Estimation Strategy... 64 3.3.1 Theoretical Model... 64 3.3.2 Empirical Model... 68 3.3.3 Econometric model... 70 3.4 Data... 72 3.5 Descriptive Statistics... 73 3.6 Results and Discussion... 77 3.6.1 Demographics... 77 3.6.2 Income... 78 3.6.3 English Proficiency... 78 3.6.4 Education... 79 3.6.5 States... 80 3.6.6 Robustness check of the model... 83 3.7 Conclusions... 84 References... 102 vii

Appendix A Description of the dummy variables for visa status in this paper... 118 Appendix B Definition of the Variables used in this paper... 119 Appendix C Description of The Property Rights Index... 121 Appendix D Description of the Physical Property Rights (PPR) in the International Property Index... 122 viii

List of Tables Table 1.1 Sample Size... 23 Table 1.2 Descriptive Statistics... 24 Table 1.3 Regressions... 25 Table 1.4 Regressions... 26 Table 1.5 Regressions... 27 Table 1.6 Regressions... 28 Table 1.7 Regressions... 29 Table 1.8 Regressions... 30 Table 1.9 Regressions... 31 Table 1.10 Regressions... 32 Table 2.1 Country profiles... 47 Table 2.2 Descriptive Statistics by country of origin... 48 Table 2.3 Entry Visa Status by Country of Origin... 49 Table 2.4 Education by Country of Origin... 50 Table 2.5 Sample Profile by State in which the Respondents live during the interview... 51 Table 2.6 Regressions... 52 Table 3.1 Income and Net Worth Statistics by Country of Origin (NIS data)... 88 Table 3.2 Visa Status and homeownership rate... 89 Table 3.3 Living arrangements by country of origin (%)... 90 Table 3.4 Sample Profile by State in which the Respondents live during the interview... 91 ix

Table 3.5 Property Rights Protection and Access to Finance Indexes... 92 Table 3.6 Logistic regression on dummy variable for homeowners.... 93 Table 3.7 Logistic Regression on the likelihood of being homeowners... 95 Table 3.8 Logistic regression on dummy variable for homeowners.... 98 x

List of Figures Figure 1 Plot of the relationship between homeownership rate of a country and this country s Private Credit ratio.... 100 Figure 2 Plot of the relationship between homeownership rate of a country and this country s Domestic Credit ratio.... 100 Figure 3 Plot of the relationship between homeownership rate of a country and this country s Property Rights Index... 101 Figure 4 Plot of the relationship between homeownership rate of a country and this country s Property Right Protection Level.... 101 xi

Chapter1. The first, the second, and the one in between: The debt behavior of Immigrants in the U.S 1.1 Introduction 1.1.1 Background By the end of 2010, it was estimated that 40 million foreign-born people were living in the United States, which accounts for about 13% of its total population (2010 Census). An estimated additional one million immigrants come to the U.S every year. The United States has been one of the most popular destinations of immigrants from all around the world. As of 2010, the top contributing originating countries of the immigrants to the U.S included Mexico, China, Philippines, India, Vietnam, Cuba, El Salvador, Dominican Republic, Canada and South Korea. As the size of the immigrant population grows, various economic issues relevant to this group have emerged and some of them have even become the center of policy debates. So far, most of the immigrant-relevant research has focused on their labor market participation and homeownership rate. Little research has been done to explore the debt behaviors of the immigrants. There are several reasons why immigrants may behave differently from non-immigrant Americans in their debt behaviors. First, immigrants from different countries may have different attitudes towards carrying debts. Some countries might have more conservative debt attitudes than others. This can affect the amount of debt immigrants would like to carry. Second, different countries have different levels of development of consumer finance markets. Taking this into account, immigrants might be less familiar with the 1

financial products in the U.S. This may have two effects on their debt behaviors. On one hand, immigrants might be less likely to take debts due to their lack of knowledge about the debt products. On the other hand, immigrants lack of financial knowledge and familiarity with financial markets could mean a higher search cost, which make them less likely to get a better deal in borrowing, thus having a higher borrowing costs like interest rate. Third, immigrants can have different costs from non-immigrant Americans when they default or file for bankruptcy, due to lack of family support or perceived risk of deportation. If they default or file for bankruptcy, immigrants are less likely than non-immigrant Americans to rely on family members for support, and thus they may have less incentives to engage in behaviors that might lead to default. In addition, for illegal immigrants, there is a risk of being discovered and deported from the U.S. For this reason, illegal immigrants might have less incentives to default or file for bankruptcy. 1.1.2 Three groups of immigrants In this study, one of my research questions focuses on the household debt behaviors of immigrants. CFM has information of respondents country of birth. I define immigrants as those who are born outside the United States. In addition, with the information of country in which the respondents stayed when they were between 12 and 18 years old, I further categorize the immigrants into two subgroups: 1.5 generation immigrants (who were born abroad and grown up in the U.S between age 12 and 18) and the real immigrants (those who were both born and raised abroad). 2

Moreover, the CFM also has the information of the country of birth respondents parents. Using that information, I define the 2 nd generation immigrants as those whose both parents were born abroad. The 1.5 generation immigrants and the 2 nd generation immigrants could be different in two key aspects from the real immigrants as well as the non-immigrant Americans. On one hand, 1.5 generation and 2 nd generation immigrants are more familiar with the American society and are more assimilated than the real immigrants, thus the arguments I used in the previous section to explain the possible reasons why immigrants might be different from non-immigrant Americans might not play well in explaining their financial behaviors. On the other hand, the 1.5 th and 2 nd generation immigrants, however assimilated they are, are still immigrants or immediate descendants of immigrants and thus may behave differently from the nonimmigrant Americans. For example, since they have longer tenure in the U.S, they may face better access to credit compared with real immigrants. But being raised in immigrant families and influenced by home country culture, the 1.5 th and 2 nd generation immigrants may still carry on part of the old credit habits thus behave differently from nonimmigrants in terms of default and bankruptcy behaviors. 1.1.3 Research questions In this paper, I am going to focus on two following questions: 1: Is there a difference in the access to credit between non-immigrant Americans and immigrants who (a) came to the U.S. as adults, (b) were born abroad but raised from 3

age 12 onward in the U.S., and (c) are born in the U.S. but having immigrant parents? Specifically, I use two variables to measure the access to credit a: The likelihood of having a specific debt, like credit card. b: The interest rate they pay on the debts they carry. 2: Is there a difference in the default behavior between non-immigrant Americans and the three immigrant groups? Specifically, I use three variables to measure the default behaviors: a: The likelihood of filing for bankruptcy b: The likelihood of being late for debt payments for at least 60 days. c: The number of late payments in the last 6 months. Using the debt information in the CFM, I am going to investigate the following debts in my paper, both separately and in a more combined fashion: 1. Credit card debt. 2. Installment debt. 3. Mortgage and other collateralized debt, like Heloc. 4 Educational loans. 5 Payday loans. 1.1.4 Outline 4

The rest part of the chapter is organized as the following. In part 1.2, I am going to have a literature review. In part 1.3, I discuss a model along with some estimation issues. Part 1.4 presents main results and interpretations. In part 1.5, I will conclude and discuss future research plans. 1.2 Literature review 1.2.1 Immigration literature Immigrant status is an identity and can potentially change economic outcomes in many areas. Akerlof and Kranton (2000) first incorporated the concept of identity into economic research. They proposed that identity is associated with different social categories and how people in these categories should behave and showed that the inclusion of identity substantively change several economic outcomes. Recently, economists have slowly use culture as a possible determinant of economic phenomena. Immigrants come from different countries with different culture background. Guiso, Sapienza and Zingales (2006) discussed the possible effects which culture might have on various economic outcomes. Carroll et al (1994) tested the hypothesis that cultural factors influence saving by comparing saving patterns of immigrants to Canada from different cultures. Using data from the Canadian Survey of Family Expenditures, they find little evidence of cultural effects on saving. Using NLSY 79 data, Amuedo-Dorantes and Pozo (2002) explored the wealth accumulation patterns of younger cohorts as well as immigrants' and non-immigrants' 5

precautionary savings in response to income uncertainty. They find that immigrants accumulate less wealth than do non-immigrants. Additionally, they found that young non-immigrants' wealth accumulation patterns appear more responsive to income uncertainty than those of their immigrant counterparts. 1.3 Model and estimation 1.3.1 Theoretical Model Following the model by Lawrence in the Journal of Money, Credit and Banking (1995), I will extend that model in the following way. In this model, I focus on the consumer debt, including debt on credit cards, payday loans and other bank loans. For the current stage, I exclude the mortgage debt because while the consumers take the debt on credit cards to smooth their consumption, taking mortgages to buy houses is a more complicated debt because buying a house can be viewed as both consumption and investment behaviors. I will extend my discussion to include mortgage in the future. The assumption is that if consumers application for credit gets denied, they can always shop around in more places and finally obtain the credit they want at a higher price. In other words, I assume that the different levels of access to credit can be solely described by different interest rates, under the assumption that there is no private market for insurance against income loss. Consider first the standard life cycle model with perfect capital market. In a two-period model, consumers maximize expected lifetime utility with preferences for consumption described by 6

( ) ( ) [ ( )] Where: is the consumption in period 1 is the consumption in period 2 is the subjective rate of time preference is the one-period, constant relative risk aversion (CRRA) utility function with properties that,, and ( ) is infinite. I assume that consumption comes from two sources: Income and borrowing/saving: I will analyze the default or bankruptcy cases when the parameters are such that households that borrowed in the first period necessarily default in the low income state in the 2 nd period but it is not optimal to strategically default in the high-income state. I start with this most tractable case now and I will analyze other special cases in the future. Substitute into the utility function, I have the utility function as following: 7

( ( )) ( ) { [ ( ) ( ) ( ( )) ( )]} { ( ) { [ ( ) ( ( )) ( ( )) ( )]} Subject to { ( ( )) ( ) ( ) ( ) Where is the income in period 1 is the saving in period 1 if it is negative or the borrowing in the 1 st period when it is positive A saver gives up, units of period-one consumption in return for units of additional period two consumption, where equals ( ( ) ( )). Similarly, borrowers can increase period-one consumption by, units by giving up units of period-two consumption (with certainty) where, again, ( ( ) ( )). ( ) is the interest rate for saving Where ( ) is a 8

variable measuring individual s risk level, which is a function of individual s age, gender, marital status, education, income, race and notably immigration status etc. for borrowers in the 1 st period and 0 for savers in the 1 st period ( ) is the interest premium for borrowing, which is also a function of risk variable. ( ) is the probability of low-income state in the 2 nd period, which is a function of individual s risk level. is the income earned in low income state ( ) is the probability of high-income state in the 2 nd period. is the income earned in the high income state is penalty of default or bankruptcy, which is a function of risk factors, including the immigration status. Assume that banks can only claim income in excess of and that there is some additional cost of default or bankruptcy to the borrower. There are several reasons underlying this cost. First, a record of default or bankruptcy could hurt the credit score, which subsequently could raise the borrowing cost of the households in the future. Although the penalty is more likely to happen in the future, in my 2-period model, I assume that the households immediately are affected in the period. Second, there is a literature arguing that there is a stigma effect of default or bankruptcy for households, thus adding another dimension of the cost of default. In addition, I 9

assume that there is an additional penalty for immigrant in the case of default. Immigrants suffer more from higher borrowing costs and more limited access to credit as a result of default and bankruptcy because they have fewer family and friends in the U.S from whom they can borrow money from. A competitive, risk neutral bank charges the borrower a rate at which expected profits equal zero, or ( ( ) ( )) ( ( )). Banks are willing to lend at up to the maximum loan size ( ) ( ) ( ) ( ) ( ) which a borrower who receives in period two can repay. Specifically for immigrants, in this model I speculate that immigrants face higher borrowing cost and lower return on investments if they save. This could be due to several reasons. First, their English proficiency level, lack of familiarity of the financial markets in the U.S and lack of financial knowledge could make their search cost for both borrowing and investment opportunities higher when compared with nonimmigrants. Second, their relatively short tenure in the U.S may imply a relatively limited access to credit. In the empirical research, I are going to examine the hypothesis that immigrants face higher borrowing costs. For savers, the Lagrangian is ( ) { [ ( ) ( ) ( ( )) ( )]} ( ( ( )) ) 10

For borrowers, the Lagrangian is ( ) { [ ( ) ( ( )) ( ( )) ( )]} ( ( ( ) ( )) ( ( )))) This optimization yields two first-order conditions. 1 st equation defines optimal saving and is the same as in the no default case. The expression shown in the 2 nd equation defines the desired loan size of a borrower who faces a q percent chance of default. For Savers, ( ) ( ) ( ) ( ( )) ( ) ( ) ( ) For borrowers ( ) ( ( )) ( ) ( ) ( ) ( ) ( ) This model is not analytically solvable. Without an explicit solution of in this model, it is hard to draw specific conclusions. Thus in the future, simulation methods will be needed. 1.3.2 Empirical Model In the estimation, I adopt the standard basic model used in previous literature to examine the determinants of the default behavior among immigrants. In particular, 11

consider estimating the following linear probability model from the cross-sectional data: Where is the dummy for defaulting on the debt, is the dummy for education achievement, is the household income, abroad), is the dummy for Real 1 st generation immigrants (born and raised is the dummy for 1.5 generation immigrants, is the dummy for 2 nd generation immigrants, is the dummy for race 1.3.3 Estimation strategy In the estimation, I need to take account for the selection problems. For example, when I study the default in credit card problem, the respondents need to have a credit card first in order for them to show up in my data set. This problem makes my data sets being examined nonrandom. Heckman (1979) proposed a two-stage estimation 12

method to address this problem, where in the first stage probit estimation is used to control the selection bias. This study uses a modified version of Heckman s two-step procedure that accommodates a logistic in the first stage following Lee (2009) and Golder (2000). Specifically, in the first stage, I am going to run a logistic regression on the dummy on having a specific debt and run the regression on the dependent variable of interest in the second stage. By doing this, the problem of selection can be addressed and solved. In my tables in the paper, I am only going to present the estimations in the second stage, while I am going to report any significant results in the first stage in the results part. One caution here is the multicollinearity problem in the regression. Multicollinearity is a condition where independent variables are strongly correlated with each other. The independent variables we use include education and income, which could be correlated and thus impose a multicollinearity problem. I use several methods to try to reduce the impact of multicollinearity. First, the data I use has a relatively large sample size (6792 observations), which can relieve the multicollinearity problems. Secondly, I try to follow the previous literatures and only include explanatory variables which have been commonly used before in my analysis. I have tried changing the specification by dropping several dummies from the regression, and the results do not change much, which suggests that the multicollinearity problem in this model is not severe. In addition, I need to point out that although dropping correlated dummy variables from the regression may be able to relive the multicollinearity, it 13

may lead to misspecification problem of the model, which is even more serious than multicollinearity. 1.4 Data 1.4.1 Comparison of immigrant-relevant data There are several data sets which have relevant immigrant information. The best known is the Survey of Consumer Finance (SCF), which has detailed information on households assets and liabilities. But the SCF does not have the information of country of birth of respondents, thus cannot be used to explore immigrant-relevant issues. In 1979 Youth Cohort of the National Longitudinal Surveys (NLSY 79), it is possible to get the respondents immigrant status and being a panel data set, it has its advantage over other cross-sectional data sets. But since the NLSY 79 data started following the respondents about 30 years ago, when the respondents were at least 18, the respondents in the data today are at least around 50 years old, thus making them less representative of the whole immigrant population. Health and Retirement Study has similar problem with NLSY 79, since it mainly focuses on relatively older population and neglecting younger population. The Panel Study of Income Dynamics (PSID) has an immigrant sample addition of 1997/1999. I do not choose PSID data in my research because the focus of PSID is on the household expenditure pattern, which is different from my focus in this paper, which is the household debt and default behavior. 1.4.2 The CFM data 14

I use the Consumer Finance Monthly (CFM) data in this paper. This survey started from February 2005 and is currently being conducted by the Center for Human Resource Research (CHRR) at Ohio State University. Data for this survey is gathered on a monthly basis in a telephone survey from a random national-level sample using Random-Digit-Dialing techniques. The data set used in this paper contains the survey data collected from January, 2008 to December, 2010, since January 2008 is the first month when the country of birth question was introduced. This data provide the most recent and comprehensive information about credit card history, track of bill payment, measures of debt stress, household assets and liabilities and so on. In the data set, the question of country of birth is used to identify respondents immigration status. Immigrants are defined as those who are born outside United States. In addition, another question In what country and state did you live the most between the ages of 12 and 18? is asked in the CFM survey. With this information, I further categorize the immigrant group into two sub-groups. For those who were both born and lived abroad between age 12 and 18, they are defined as real first generation immigrants. For those who were born abroad but lived mostly in the U.S when they were between age 12 and 18, I define them as 1.5 generation immigrants. In addition, I also have the information of the country birth of the parents of respondents. Using that information, along with the information of country of birth of the respondents, 2 nd generation immigrants are defined as those who were born in the U.S while both of their parents were born abroad. In my following analysis, I will use 15

two control groups: (1) immigrants vs non-immigrants, and (2) real immigrants vs 1.5 generation immigrants vs 2 nd generation immigrants. 1.4.3 Descriptive statistics Table 1.1 and table 1.2 present weighted descriptive statistics of the CFM data. I compare the country profile of the immigrants with the census data and make a new weight in order to make the data set more representative of the immigrant population. I do not see much difference in the access to credit among different groups, including the share of whom have credit cards, mortgage debt etc. In terms of the total amount of the outstanding debt, either non-mortgage or mortgage, the immigrants, especially the 1.5 th generation of immigrants, have an average higher debts than the nonimmigrants. As for the education achievements, the immigrant population also appears to be better educated than non-immigrants. This can serve as additional evidence that due to the telephone interview nature of the survey, the CFM data set mainly capture the financial information of the relatively better-off immigrants, who are different from some people s stereotype about immigrants, especially those from Latino countries like Mexico or El Salvador. 1.5 Results 1.5.1Results on default behavior In Table 1.3, I used different specifications to examine whether there is a different likelihood of default in mortgage between immigrant and non-immigrant groups. The households are defined as defaulters if they are late in their mortgage payments for 60 days or more. In the regressions, I tried two sets of specifications: 1: a single dummy 16

for immigrants; 2: dummies for real immigrants (born and raised abroad), 1.5 generation immigrants and 2 nd generation immigrants. I use two-stage Heckman selection method to estimate my regressions, with the first stage being logistic regression on the likelihood of having outstanding mortgage debt. My results find that in the first stage, there is no difference in the likelihood of having a mortgage between immigrants and non-immigrants; in the second stage, being an immigrant, however real, 1.5 generation, or 2 nd generation, does not imply a different likelihood of default on their mortgages, compared with non-immigrants. I find that while immigrant status does not matter. When the mortgage is underwater (outstanding mortgage debt greater than the current house price), the homeowner is more likely to default on the mortgage. In model 1 and model 4, I introduce the interaction term between immigration status and being underwater. Again, no significant estimates are found, which imply that immigrants do not behave differently from non-immigrants in making default decisions when their houses are underwater. In Table 1.4, I present two sets of results. In model 1 and model 2, the dependent variable is the dummy variable for being late for payments for 60 days in non-credit card and non-mortgage debts, using the Heckman selection methods. The results in model 2 suggest that first-generation immigrants (including both real and 1.5 generation) are more likely to default in those debts, with the results in the 1 st stage suggesting that immigrants are less likely to have outstanding debts. Model 1 breaks down the immigrant group into two subgroups. The results in model 1 show that the 17

1.5 generation immigrants are more likely to default on these debts while no significant estimates are found on the real immigrants. Moreover, the first stage estimation finds that real immigrants are less likely to carry the debt. Comparing results from model 1 and model 2, I can conclude that while the 1.5 generation immigrants do not behave differently in carrying out these debts and are more likely to default on them compared with non-immigrants, the real immigrants are more conservative in that they are less likely to carry debts and are less likely to default if they do take the debts. The results here draw a picture of localization : while the real immigrants tend to carry more conservative attitudes towards taking and managing debts, the 1.5 generation immigrants, being raised in the U.S, are more open to the idea of having debt and less careful in managing their debts. Model 3 and 4 in Table 1.4 use the dummy for being late for credit card payments for 60 days or more as the dependent variable, with the first stage of the Heckman being the likelihood of having a credit card. In the first stage, I do not find any significant difference between immigrants and non-immigrants in the likelihood of having a credit card as well as the likelihood of default on the credit cards debt. I also control the credit card limit in the regression and the results suggests that the higher the credit card limit, the less likely that the cardholder is going to default. A higher credit card limit is always associated with better credit score, implying a higher income. This shows that the income is the main factor in determining the default rate. Model 3 through model 6 in table 1.5 use a continuous variable of number of late credit card payments as the dependent variable in the second stage, which is different 18

from the dummy variable approach in table 4. I use Heckman selection model to estimate the regressions with the first stage being having outstanding balance on credit cards, the second stage being Tobit regression on the number of later payments on credit cards. I used two criterions to capture the number of late payments in the past 6 months: 6 and 12. My results find that the 1.5 generation immigrants are more likely to have credit cards, and conditional on having a credit card, they have more late payments on credit cards than non-immigrants. These findings here again support my previous finding that the 1.5 generation immigrants are more likely to default on their debts. Model 1 and model 2 in table 1.5 use the dummy for being late for 60 days in any loans (including mortgage and credit cards). Table 1.6, instead, replaces the dependent variable with the number of late payments in the last 6 months in noncredit card loans and in all kinds of debts. No significant differences are found between immigrants and non-immigrants. Model 1 and Model 2 in table 1.7 are logistic regressions on the likelihood of filing for bankruptcy. The results show that the immigrant group is less likely to file for bankruptcy because the real immigrants are less likely to file for bankruptcy while the 1.5 generation of immigrants are not significantly different from non-immigrants. In addition, the 2 nd generation immigrants also show no difference in the likelihood of filing for bankruptcy compared with non-immigrants. 1.5.2 Results on access to credit Another focus of my study is the difference in the access to credit between immigrants and non-immigrants, if there is any. I use several different variables to 19

measure the access to credit: the likelihood of having a credit card, credit limit on credit cards, interest rate on the debt etc. In model 5 and 6, which are logistic regressions with the dependent variable being the dummy of having credit cards, I found that the real immigrants are more likely to have credit cards. Model 3 and model 4 suggest that, conditioning on having a credit card, there is no difference in the credit limit granted between immigrants and nonimmigrants. Table 1.8 and table 1.9 report the results on interest rates on various debts. All the models use a two-stage Heckman selection method with the first stage being having the specific debt. For example, in model 1 in table 8, the first stage is logistic regression on having a mortgage, and the second stage is a tobit regression on the interest rate on mortgage. I do this exercise on interest rate on mortgage, credit card, payday loan, installment debt, bank loan, education loan and Heloc. Generally, I find little evidence about the difference between immigrants and non-immigrants in debts. The only exception is that the real immigrants enjoy a lower interest rate on bank loans than non-immigrants. This result, along with earlier finding that the immigrants are more likely to own credit cards, shows that the immigrants, somehow surprisingly, enjoy a relatively fair access to credit compared with non-immigrants. My last exercise reported in table 1.10 includes the tobit regression on the total amount of debt held by the households. My results show that the 1.5 generation immigrants take more debts than non-immigrants, while no significant results are 20

found on real immigrants and 2 nd generation immigrants. Recall that in my earlier results, I find that 1.5 generation immigrants are more likely to default. Based on my results, the 1.5 generation immigrants can be portrayed as a more reckless group in the U.S: take more debt, and more likely to default. On the other hand, the similarity between immigrants and non-immigrants shows a significant degree of economic assimilation of immigrants and their acceptance of the debt culture in the U.S. On the other hand, the similarity can also be partly explained by the nature of the survey: The CFM data is a landline telephone interview so that the respondents, including the immigrants, are more likely to have a home, be better off economically, and older. All these characteristics suggest a relatively longer tenure in the U.S of those immigrants, thus a higher degree of economic assimilation. 1.6 Conclusions 1.6.1 The conservative Real immigrants These results demonstrate a picture of conservative real 1 st generation immigrants: borrow less, do not behave differently in default behaviors and be less likely to file for bankruptcy. These findings are consistent from my previous predictions that due to lack of informal credit sources of credit, immigrants face higher cost of default and thus have more incentives to behave responsibly and try to avoid default or bankruptcy. 1.6.2 The Americanized 2 nd generation On the other extreme, the 2 nd generation immigrants, who are the offsprings of the 1 st generation immigrants, do not behave differently from the non-immigrant Americans. 21

It is not surprising as the 2 nd generation immigrants, after all, are not immigrants at all. Their behavior difference, if there is any, can be better understood in a study of race and ethnicity, rather than a study of immigration. 1.6.3 The myth of the 1.5 th generation I have made a handful of findings by comparing the immigrant with the nonimmigrant population. The most noticeable findings come from my effort to further divide the immigrant population into more detailed subgroups. Taking advantage of the unique data set I have, I discover the heterogeneity disguised before. The results turn out to show us that though in most occasions, I tend to treat the immigrant population as one group, great differences exist between the real immigrant and the 1.5 th generation immigrants. The results show that the 1 st generation immigrants who grew up abroad are less likely to file for bankruptcy and do not behave differently compared with non-immigrants in terms of default behaviors, those 1.5 th generation immigrants who grew up in the U.S, however, are on average taking more debts, more likely to default and have more average number of defaults than nonimmigrants. These findings partly tell us a story of a very active and fast process of being overly Americanized. But if I compare this reckless 1.5 th generation with the 2 nd generation, the story gets more complicated. Why do 2 nd generation immigrants, who arguably have immersed themselves even longer than the 1.5 th generation immigrants, have less trouble making their payments and behave more responsibility when they carry out debts? These questions still remain to be answered in the future. 22

Table 1.1 Sample Size Sample size Percentage All 1st generation immigrant 468 6.89 Real Immigrant 287 4.23 1.5 generation immigrant 181 2.66 2nd generation immigrant 173 2.55 Everyone else 6151 90.56 Total 6792 100 23

Real Immigrant 1 Table 1.2 Descriptive Statistics 1.5th generation immigrant 2 2nd generation 3 Nonimmigrant All age 52.69 48.36 63.34 56.99 56.72 married 0.77 0.71 0.58 0.65 0.65 household size 4.4 4.57 4.51 4.36 4.38 Less than high school 0.02 0.09 0.08 0.05 0.06 High School 0.2 0.17 0.21 0.24 0.24 Some college 0.22 0.32 0.25 0.29 0.29 College degree 0.28 0.24 0.24 0.2 0.21 Graduate degree 0.24 0.18 0.18 0.18 0.18 Have mortgage 0.6 0.52 0.36 0.47 0.48 Have Heloc 0.14 0.17 0.15 0.16 0.16 Have education loan 0.13 0.17 0.1 0.11 0.11 Have installment 0.26 0.28 0.23 0.3 0.29 debt Have bankloan 0.08 0.04 0.04 0.05 0.05 Have payday loan 0.01 0.02 0.02 0.01 0.01 Have other debt 0.04 0.03 0.02 0.03 0.03 Have credit card 0.79 0.76 0.78 0.75 0.75 Household income 76386.67 73330.27 65199.93 67986.97 68409.98 Credit card debt 4194.68 5328.2 1429.08 2506.99 2634.08 Non mortgage debt 9263.83 10574.48 10145.8 8260.52 8422.66 Total debt 121382.45 116223.46 63802.66 73316.19 76271.34 1 Born and raised abroad. 2 Born abroad, but lived in the U.S when they were between 12 to 18. 3 Born in the U.S, parents of whom born abroad. 24

Table 1.3 4 Regressions Model 1 SE Model 2 SE Model 3 SE Model 4 SE age 0 0.01 0 0 0.01 0 0.01 married -0.36** 0.18-0.36** 0.01-0.36** 0.18-0.36** 0.18 northeast -0.25 0.15-0.25 0.18-0.25 0.15-0.24 0.15 south -0.01 0.12-0.01 0.15-0.01 0.12-0.01 0.12 west 0 0.13-0.01 0.12-0.01 0.13 0 0.13 Household income -1.33** 0.57-1.24** 0.13-1.24** 0.57-1.26** 0.57 (millions) year2009 0.34*** 0.11 0.34*** 0.57 0.34*** 0.11 0.34*** 0.11 year2010 0.08 0.13 0.08 0.11 0.07 0.13 0.07 0.13 Real immigrant -0.14 0.24-0.04 0.13 1.5 generation -0.04 0.29-0.08 2 nd generation -0.09 0.37-0.12 1 st generation 0.20-0.05 0.19-0.11 0.2 black 0.27 0.18 0.26 0.18 0.26 0.18 0.25 0.18 Hispanic 0.51** 0.22 0.54** 0.21 0.52** 0.21 0.51** 0.21 Native american -0.38 0.40-0.4 0.40-0.4 0.4-0.38 0.4 Asian 0.13 0.31 0.26 0.30 0.25 0.29 0.19 0.3 Other race 0.27 0.33 0.27 0.33 0.27 0.33 0.27 0.33 Less than high school -0.11 0.33-0.1 0.33-0.1 0.32-0.09 0.33 High school 0.13 0.16 0.14 0.16 0.14 0.16 0.14 0.16 Some college 0.27** 0.13 0.27** 0.13 0.27** 0.13 0.27** 0.13 Graduate degree 0 0.14 0 0.14 0.01 0.14 0.01 0.14 Underwater 0.37** 0.15 0.42*** 0.15 0.42*** 0.14 0.37** 0.15 Real immigrant 1.44 0.70 *underwater 1.5 generation -0.06 0.65 *underwater 2 nd generation *underwater -0.28 1.77 1 st generation *underwater 0.60 0.46 0.6 0.46 No of observations 1791 1791 1791 1791 Log Likelihood -1491-1493 -1494-1493 4 All regressions use Heckman 2-stage selection model. Only the results on the 2nd stage is reported here. The dependent variable for the 1st stage is dummy for having mortgage, and the dependent variable for 2nd stage is being late in mortgage 25

Table 1.4 5 Regressions Model 1 Model 2 Model 3 Model 4 age 0.00 0.01 0.00 0.01-0.01 0.00-0.01 0.00 married -0.33 0.22-0.32 0.24-0.08 0.12-0.08 0.12 northeast 0.09 0.15 0.08 0.15-0.01 0.12-0.02 0.12 south 0.28* 0.15 0.27* 0.16-0.13 0.11-0.13 0.11 west 0.30** 0.14 0.29** 0.15-0.07 0.11-0.07 0.11 Household income -1.70** 0.72-1.80** 0.73 0.29 0.52 0.30 0.52 (millions) year2009 0.39*** 0.11 0.39*** 0.11 0.01 0.10 0.00 0.10 year2010 0.48*** 0.13 0.49*** 0.12-0.08 0.11-0.08 0.11 Real immigrant 0.43* 0.22 0.15 0.15 1.5 generation 0.09 0.37 0.12 0.19 2 nd generation 0.64*** 0.23 0.16 0.21 1 st generation -0.44 0.48-0.14 0.28 black 0.19 0.31 0.24 0.27 0.38** 0.16 0.38** 0.16 Hispanic -0.31 0.29-0.33 0.26 0.04 0.16 0.06 0.17 Native american 0.61 0.44 0.56 0.48 0.00 0.32 0.00 0.32 Asian -0.17 0.31-0.27 0.31-0.12 0.25-0.09 0.26 Other race -0.15 0.31-0.13 0.32 0.17 0.25 0.17 0.25 Less than high 0.43 0.45 0.37 0.48 0.19 0.29 0.18 0.29 school High school 0.32** 0.14 0.34** 0.15 0.05 0.14 0.05 0.14 Some college 0.39* 0.23 0.44** 0.18 0.35*** 0.12 0.35*** 0.12 Graduate degree 0.20 0.20 0.24 0.18 0.24** 0.12 0.24* 0.12 Credit card limit -3.68** 1.68-3.68** 1.68 No of observations 1874 1874 1691 1691 Log Likelihood -1676-1676 -1403-1403 5 All regressions use Heckman 2-stage selection model. Dependent variable for the 1 st stage of model 1 and 2 is dummy of having non-mortgage non-credit card loans, and being late for 60 days in these debts for the 2 nd stage. For model 3 and 4, the dependent variable in the 1 st stage of the Heckman selection model is dummy of having a credit card, and being late for credit card payment for at least 60 days being the dependent variable in the 2 nd stage. 26

Table 1.5 6 Regressions Model 1 SE Model 2 SE Model 3 Model 4 SE Model 5 SE Mode l 6 SE age 0.00 0.00 0.00 0.00-0.04 0.02-0.04* 0.02-0.03 0.02-0.03 0.02 married -0.11 0.11-0.11 0.11-0.27 0.61-0.28 0.60-0.34 0.60-0.34 0.60 northeast -0.06 0.11-0.06 0.11 0.04 0.70 0.05 0.70-0.01 0.70-0.01 0.70 south -0.01 0.09-0.01 0.09-0.27 0.57-0.27 0.57-0.28 0.56-0.29 0.56 west -0.10 0.10-0.10 0.09 0.15 0.59 0.15 0.59 0.21 0.59 0.20 0.59 Household income -0.80* 0.46-0.80* 0.45 3.09 2.68 3.10 2.68 2.92 2.66 2.93 2.66 year2009 0.01 0.08 0.01 0.08 0.78 0.51 0.79 0.51 0.71 0.50 0.72 0.50 year2010-0.14 0.09-0.14 0.09 0.34 0.59 0.34 0.58 0.23 0.58 0.25 0.58 1 st generation 0.06 0.13 1.61** 0.80 1.52* 0.80 Real immigrant 0.03 0.17 1.32 1.11 1.24 1.11 1.5 generation 0.07 0.19 1.90* 1.05 1.83* 1.04 2 nd generation -0.12 0.25 0.37 1.48 0.47 1.47 black 0.35*** 0.12 0.35*** 0.12 1.12 0.79 1.07 0.79 1.05 0.78 1.03 0.78 hispanic 0.13 0.14 0.12 0.13-0.44 0.90-0.34 0.85-0.47 0.89-0.33 0.85 Native American 0.02 0.29 0.01 0.29 0.54 1.63 0.47 1.62 0.50 1.62 0.45 1.62 asian -0.06 0.23-0.08 0.22-2.31 1.71-2.41 1.68-2.30 1.70-2.34 1.67 others 0.00 0.22 0.00 0.22 1.19 1.48 1.24 1.48 1.49 1.44 1.48 1.45 Less than High School 0.12 0.24 0.12 0.24-0.41 1.41-0.44 1.40-0.46 1.38-0.49 1.38 High school 0.19* 0.12 0.19* 0.11 0.12 0.70 0.11 0.70 0.03 0.69 0.02 0.69 Some college 0.42*** 0.10 0.42*** 0.09 1.63*** 0.60 1.64*** 0.60 1.57*** 0.60 1.58* ** 0.60 Graduate degree 0.22** 0.11 0.23** 0.11 0.53 0.70 0.54 0.70 0.60 0.70 0.60 0.70 Credit card - -14.12 9.40-14.27 9.39-13.46 9.34 limit 13.63 9.33 No of observations 1874 1874 1384 1384 1384 1384 Log Likelihood -1325-1326 -1496-1496 -1526-1526 6 The dependent variable of Model 1 and Model 2 is being late for any debt payment for at least 60 days. For model 3-6, the dependent variable is number of later payments in credit card (censored at 6 for model 3-4 and censored at 12 for model 5-6). All models use Heckman 2-stage regression, with the first stage being the logistic regression on the likelihood of having any debt (1&2) or credit card(3-6). 27

Table 1.6 7 Regressions Model 1 Model 2 Model 3 Model 4 age -0.01 0.04-0.07** 0.03-0.04** 0.02-0.04** 0.02 married -2.01** 0.95-0.72 0.75-0.50 0.59-0.54 0.59 northeast 0.39 0.88 0.09 0.87-0.33 0.64-0.31 0.64 south 0.73 0.78 0.05 0.71 0.05 0.51 0.10 0.51 west 0.59 0.81 0.03 0.77-0.48 0.56-0.44 0.56 Household income -7.41* 4.07-7.14* 4.07-4.20 2.65-4.32 2.65 year2009-0.66 0.66-0.82 0.65 0.44 0.46 0.46 0.46 year2010-0.35 0.78-0.67 0.76-0.31 0.55-0.29 0.55 1 st generation 0.17 1.12 0.38 0.78 Real immigrant -1.83 1.82-0.69 1.04 1.5 generation 2.57* 1.46 1.28 1.07 2 nd generation -4.64 3.03-2.07 1.59 black 1.62 1.01 2.48*** 0.92 1.61** 0.70 1.53** 0.70 hispanic -1.51 1.31-1.82 1.22 0.06 0.82-0.03 0.79 Native american 1.11 2.73-0.61 2.44 1.09 1.63 0.93 1.63 asian -0.32 1.99-1.02 1.87-0.63 1.39-1.16 1.37 others -0.10 1.76 0.39 1.75 0.55 1.27 0.57 1.27 Less than high school 3.31 2.03 1.03 1.75 0.63 1.29 0.65 1.29 High school 1.92** 0.89 1.63* 0.87 1.16* 0.65 1.25* 0.64 Some college 2.24*** 0.79 2.69*** 0.78 2.29*** 0.56 2.33*** 0.56 Graduate degree 1.39 0.91 1.84** 0.90 1.11* 0.64 1.20* 0.64 No of observations 1864 1864 1863 1863 Log Likelihood -2033-2187 -2214-2216 7 All regressions use Heckman 2-stage selection model. Only the results on the 2 nd stage are reported here. In model 1 and model 2, dependent variable for the 1 st stage is dummy of having any nonmortgage and noncredit card loans, and number of late payments in nonmortgage and noncredit card loans in the last 6 months in the second stage. In model 3 and model 4, dependent variable in the 1 st stage is the dummy for having any debt, and number of late payments in any debt in the last 6 months in the 2 nd stage 28

29 Table 1.7 8 Regressions Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 age 0.02*** 0.01 0.02*** 0.01 0.15 0.07 0.14* 0.07 0.02*** 0.01 0.02*** 0.01 married 0.27 0.19 0.26 0.19 3.64 2.29 3.63 2.29 0.63*** 0.15 0.63*** 0.15 northeast -0.42* 0.24-0.42* 0.24 2.32 2.26 2.31 2.26-0.24 0.20-0.24 0.20 south -0.19 0.17-0.18 0.17 1.33 1.91 1.34 1.91-0.65*** 0.15-0.65*** 0.15 west 0.14 0.19 0.14 0.19 1.83 2.00 1.89 2.00-0.21 0.16-0.22 0.16 Household income -6.40*** 1.55-6.42*** 1.55 53.10 8.76 53.16* 8.77 0.01*** 0.00 0.01*** 0.00 year2009-0.04 0.16-0.03 0.16-0.62 1.77-0.55 1.77-0.39*** 0.13-0.39*** 0.13 year2010-0.76*** 0.23-0.75*** 0.23-2.80 1.95-2.74 1.95-0.06 0.17-0.06 0.17 1 st generation -0.66* 0.36-2.46 2.83 0.56** 0.26 Real immigrant -1.14** 0.54-1.60 3.52 0.68* 0.35 1.5 generation -0.25 0.47-4.36 4.04 0.43 0.34 2 nd generation -1.00 0.75-3.10 5.03-0.08 0.44 black 0.00 0.25-0.01 0.25-7.05 2.99-7.00** 3.00-0.62*** 0.20-0.61*** 0.20 hispanic -0.45 0.33-0.50 0.32 2.70 3.01 3.39 3.12-0.06 0.24-0.08 0.23 nativeamerican -0.46 0.63-0.48 0.63 5.98 6.15 5.88 6.15-0.60 0.45-0.60 0.45 asian -0.67 0.78-0.80 0.77 4.49 4.45 4.66 4.57-0.02 0.47 0.00 0.46 others -0.47 0.50-0.47 0.49-1.50 4.72-1.45 4.72-0.07 0.36-0.07 0.36 Less than high - 0.62* 0.36 0.61* 0.36-14.54 5.24 school 14.51*** 5.24-1.51*** 0.29-1.50*** 0.29 High school 0.76*** 0.22 0.78*** 0.22-3.84 2.36-3.85 2.36-1.11*** 0.18-1.11*** 0.18 Some college 0.76*** 0.21 0.76*** 0.21-3.86 1.98-3.80* 1.98-0.59*** 0.17-0.59*** 0.17 Graduate degree -0.33 0.30-0.32 0.29-0.75 2.13-0.80 2.13 0.08 0.23 0.08 0.23 household size 0.01 0.05 0.01 0.05 0.53 0.51 0.52 0.51-0.10** 0.04-0.10** 0.04 No of observations 1872 1872 1691 1691 1874 1874 Log Likelihood 119.46 115.39-6423 -6422 378.04 377.7 8 Model 1 and 2 are logistic regressions with the dummy of having filed for bankruptcy before as the dependent variable. Model 3 and 4 are using Heckman 2- stage regression, with logistic regression using dummy for having a credit card in the 1 st stage and tobit regression using the credit card limit being the dependent variable in the 2 nd stage. Model 5 and 6 are logistic regressions using the dummy of having a credit card as the dependent variable.

30 Table 1.8 9 Regressions Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 age 0.01** 0.01 0.01** 0.01-0.02 0.02-0.02 0.02 0.05 0.06 0.05 0.05 married 0.72*** 0.17 0.73*** 0.17 0.08 0.53 0.08 0.53-0.72 0.67-0.75 0.66 northeast -0.31* 0.18-0.32* 0.18-0.59 0.46-0.59 0.46 0.83 0.70 0.78 0.68 south -0.17 0.14-0.18 0.14-0.58 0.43-0.56 0.43-0.32 0.71-0.33 0.69 west -0.24 0.15-0.24 0.15-0.35 0.44-0.36 0.44-0.45 0.64-0.48 0.63 Income 0.22 0.64 0.25 0.64-1.76 1.85-1.84 1.86 0.55 2.61 0.43 2.67 year2009-0.40*** 0.13-0.40*** 0.13-1.06** 0.46-1.05** 0.46-0.22 0.53-0.31 0.53 year2010-0.60*** 0.15-0.60*** 0.15-1.27*** 0.42-1.26*** 0.42-0.48 0.57-0.72 0.58 1 st generation -0.04 0.22 0.33 0.62-0.01 0.93 Real immigrant -0.13 0.28 0.07 0.79-0.80 1.28 1.5 generation -0.01 0.31 0.61 0.90 0.64 1.10 2 nd generation -0.60 0.44-0.29 0.98 2.79 1.85 black 0.23 0.22 0.24 0.22 0.57 0.87 0.57 0.87-0.65 0.99-0.76 0.96 hispanic -0.20 0.24-0.16 0.25 0.25 0.84 0.20 0.80-0.45 0.89-0.46 0.88 nativeamerican -0.01 0.49 0.02 0.49 1.07 1.98 1.09 1.97 1.65 2.53 1.76 2.49 asian -0.36 0.36-0.27 0.36-1.08 1.07-1.21 1.04 0.52 1.41 0.73 1.40 others -0.08 0.36-0.08 0.36 0.72 1.07 0.73 1.07-0.25 1.75-0.21 1.73 Less than high school -0.16 0.37-0.17 0.37-2.83** 1.36-2.81** 1.36-1.57 4.18-1.58 4.02 High school -0.07 0.17-0.09 0.17-0.35 0.46-0.34 0.46 0.46 1.37 0.49 1.29 Some college 0.17 0.15 0.17 0.15 1.23*** 0.41 1.23*** 0.41-0.64 0.66-0.59 0.64 Graduate degree -0.03 0.17-0.04 0.17 0.40 0.45 0.42 0.45 0.04 0.60 0.12 0.61 No. of Obs 1707 1707 1790 1790 1756 1756 Log Likelihood -2958-2955 -1492-1490 -1275-1272 9 All regressions use Heckman 2-stage selection model. Only the results on the 2 nd stage are reported here. In model 1 and model 2, dependent variable for the 1 st stage is dummy of having a mortgage, and the interest rate on the mortgage in the 2 nd stage. For model 3 and model 4, the dependent variable is dummy for having a Heloc, and the interest rate of Heloc in the 2 nd stage. For model 5 and model 6, the dependent variable is dummy for having an education loan, and the interest rate of education loan in the 2 nd stage.