EFFECTS OF THE 1996 WELFARE REFORM ON IMMIGRANT POVERTY AND WORKFORCE PARTICIPATION IN THE UNITED STATES

Transcription

1 EFFECTS OF THE 1996 WELFARE REFORM ON IMMIGRANT POVERTY AND WORKFORCE PARTICIPATION IN THE UNITED STATES A Thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy By Kristen L. Johnson, B.A. Washington, DC April 11, 2012

2 Copyright 2012 by Kristen L. Johnson All Rights Reserved ii

3 EFFECTS OF THE 1996 WELFARE REFORM ON IMMIGRANT POVERTY AND WORKFORCE PARTICIPATION IN THE UNITED STATES Kristen L. Johnson Thesis Advisor: Gurkan Ay, Ph.D. ABSTRACT This study investigates how the 1996 welfare reform legislation, the United States Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA), impacted poverty and workforce participation outcomes for immigrants in the United States. Among other new restrictions for all individuals in the United States, PRWORA removed welfare eligibility for non-citizens for their first five years in the country and instituted work requirements for all citizens to receive welfare. This study uses a triple-difference linear probability model similar to Borjas (2003) to measure changes in the likelihood of being in poverty and participating in the workforce by citizenship status, state provision of welfare, and time. Results are calculated for two post-prwora time periods, and , and compared to the pre-prwora time period PRWORA is found to have increased the likelihood of being in poverty for non-citizens with children living in states not providing welfare to immigrants in the second post- PRWORA period. In states that chose not to provide welfare to immigrants, PRWORA increased the likelihood that naturalized citizens participate in the workforce in the first post-prwora period but decreased the likelihood that non-citizens in those states participated in the workforce in the second post-prwora period. These findings suggest that PRWORA negatively affected immigrants as a group, particularly noncitizens, and that the five-year bar on eligibility for welfare should be removed. iii

4 I want to thank my thesis advisor for all his patience, encouragement, and guidance throughout this process. I also want to thank my family and friends for their support and listening ear when I was not sure how I would get it all done. Most of all I want to thank Patrick, the love of my life, for giving up so much to make pursuing this degree possible. Kristen L. Johnson iv

5 TABLE OF CONTENTS 1 Introduction 1 2 Background and Literature Review 7 3 Conceptual Model 20 4 Data and Methodology 26 5 Empirical Results 56 6 Discussion and Conclusions 69 Appendices 73 Works Cited 98 v

6 CHAPTER 1. INTRODUCTION The United States Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996 ended welfare as we know it. 1 This law ended the federally administered Aid to Families with Dependent Children (AFDC) program and instituted the state-administered Temporary Assistance to Needy Families (TANF) programs with work requirements aimed to move individuals off the welfare rolls. One group was particularly affected by this legislation: non-citizen immigrants. Along with U.S. citizens, refugees were shifted to the TANF system 2 with added time limits for benefits while undocumented immigrants and temporary immigrants remained ineligible for nonemergency benefits and services. Non-citizen immigrants who arrived after the enactment of this legislation were barred from accessing TANF and Medicaid for five years and also barred from accessing Supplemental Security Income (SSI) and food stamps (now the Supplemental Nutrition Assistance Program, or SNAP) until citizenship was obtained (Primus 1996). Given concerns that immigrants were using public benefits at a disproportionate rate (Borjas 2002; Borjas 2003), this piece of legislation sought to limit any negative fiscal impact of immigrants on the U.S. government and reserve scarce resources for American citizens. In addition, some would argue that the overarching goal of barring immigrants from welfare programs was to discourage low-skilled immigrants from coming to the U.S. by eliminating welfare incentives (Borjas 2002). 1 President Bill Clinton, at the White House ceremony signing PRWORA into law in TANF is only available to qualifying refugee families with children under 18 while refugees who are blind, elderly, or disabled are eligible for SSI for a limited period of time. Those refugees who meet the income threshold for TANF eligibility but do not meet other requirements, such as those refugees without children, can receive federally funded Refugee Cash Assistance either through state administered programs, a public/private partnership, or a Wilson/Fish program (Office of Refugee Resettlement). 1

7 This paper examines the effects of the 1996 welfare reform on the incidence of poverty 3 and workforce participation for immigrants in the United States. The source of the data used in this study, the Current Population Survey, assigns poverty status to respondents based on their responses to survey questions about family income and family size. I employ a broad immigrant definition and classify all individuals living in the United States who were born outside of the United States as immigrants. Lawful permanent residents are ideally the primary sub-population of interest, since this group was arguably a target of welfare reform and experienced the greatest eligibility changes. Furthermore, because lawful permanent residents are essentially on a path to citizenship in the United States, focusing on this group allows for an analysis of changes in economic assimilation of future citizens since welfare reform. Given that available data does not include information about a foreign-born person s visa status, for the purpose of analysis this paper only separates immigrants as a whole by refugee and citizenship status. Since PRWORA was the first major change to welfare in more than 60 years, several studies examined various facets of the legislation s impacts on the behaviors and changing and socioeconomic outcomes of immigrants soon after the policy change. For example, Borjas (2002) focuses on the effect of welfare reform on immigrant participation in public assistance programs relative to natives and how welfare reform affected naturalization application rates. Another study, Borjas (2003), analyzes how PRWORA affected health insurance coverage, both private and public, of immigrants and accompanying labor supply changes among immigrants relative to the native population. 3 Poverty in this study is measured according to the United States government s official definition, which was developed in 1963 by Mollie Orshansky of the Social Security Administration and revised slightly since. The poverty threshold is updated yearly according to the Consumer Price Index. The limitations of this official measure and a discussion of alternative measures of poverty will be reviewed later in the paper. 2

8 In addition, Capps et al. (2002) perform an in-depth analysis on a range of socioeconomic outcomes for immigrant families in two cities with large immigrant populations, New York and Los Angeles, to explore the level of need these families have for public assistance. In addition to the above studies with immigrant-centric analyses, other studies shortly after PRWORA examined changes in the overall population as well as the immigrant population. One such study compares the use of public assistance by natives and immigrants of the same income level relative to the poverty line (Fix and Passel 1999). Another study takes a broader perspective and tracks the historical trends of the effect that U.S. immigration policy has had on the amount of poverty in the United States (Camarota 1999). Passel and Clark (1998) take a different approach and analyze taxes paid by immigrant subgroups as well as the changes in the levels of government (state vs. federal) receiving immigrant tax revenues compared to the levels of government providing benefits to immigrants during the years immediately following welfare reform. To my knowledge, my study is the first to focus specifically on PRWORA s effect on immigrant poverty in states that chose to not provide welfare benefits to post- PRWORA immigrants. I further extend my analysis within the same framework by investigating PRWORA s effect on immigrant workforce participation. Additionally, my study is one of the first in a decade since the initial post-prwora studies to examine the legislation s effects years on immigrants. Fix and Passel (1999) state that the complete effect of welfare reform on immigrants would not be felt for several years after reform. The major reason for this lag, they argue, is that immigrants who arrived before August 22, 1996 were in practice not 3

9 barred from receiving many public benefits. 4 Therefore, in the two to three years immediately following PRWORA s enactment a significant proportion of immigrants in the United States were technically still eligible for these benefits. In addition, others have noted that more than just a few years needed to pass for effects of welfare reform on economic assimilation to become accurately measurable and significant (Passel and Clark 1998). Economic assimilation for immigrants is often understood as reaching income parity with natives with similar personal characteristics. Whether as a result of PRWORA the proportion of immigrants in poverty increased or decreased relative to their immigrant and native counterparts is a proxy of economic assimilation. Additionally, the previous welfare regime had been in place for 60 years and states needed time to thoughtfully rework their policies. Changes so substantial that affected millions of individuals undoubtedly take more than two or three years to be felt. Schoeni and Blank (2000) note that [t]he earliest official implementation date was September 1996 and the latest was July 1997, when all states were required to begin operating under TANF. However, in some states the initial official plan was simply a placeholder, designed to allow the state to begin to draw down its TANF block grant, and some state policies were not changed until a later date (15). This paper, more than 15 years removed from welfare reform and several years past the publication of the majority of papers on related topics, in contrast to those papers examines poverty rates for a greater number of immigrants who came after PRWORA and how economic assimilation 4 PRWORA gave states the option to provide TANF and Medicaid to pre-prwora immigrants, which most did. Federal legislation reinstated other benefits to this population. First, the Balanced Budget Act of 1997 reversed PRWORA policy and allowed immigrants accessing SSI at the time of the legislation to continue receiving benefits and granted future access to pre-prwora immigrants who would qualify in the future. The Agricultural Research, Extension, and Education Reform Act of 1998 then expanded food stamp eligibility back to certain pre-reform immigrants (GAO 1998). 4

10 has changed since the enactment of the program. The results of this study will help clarify how immigrants are faring in post-prwora periods compared to the period before PRWORA, as well as how state choices have affected these outcomes. In many ways how a country chooses to spend its money demonstrates what it values. The Horatio Alger legend that if an individual works hard enough, he/she will achieve the American dream of prosperity on his/her own efforts alone continues to be an underlying tenet of social policies and deeply held values in the United States. PRWORA is another example of this. The legislation is grounded in economic theories that seek to mitigate moral hazard and adverse selection into public programs. However, some would argue that the new requirements and lifetime caps also at least partially stemmed from the perception that those on welfare are lazy, since some believe that motivated individuals would find work regardless of their circumstances and not need welfare. In political discourse welfare is often described less as a right in the United States than a shameful and short-term option of last-resort. With limited resources the government has also made the choice in PRWORA that immigrants do not have the same moral standing as citizens and cannot receive the same benefits. Since this legislation, immigrants have to prove themselves and survive economically in the United States for five years without expecting any help. That is a clear expression both of increasingly tight resources faced with growing demands and of the values and beliefs of those who crafted and passed the legislation. PRWORA s goals appear to be primarily about moving people off the welfare rolls and motivating them to work rather than improving economic outcomes for those individuals. Those choices have policy and ethical dimensions. Additionally, PRWORA is indicative of the recurrent belief that devolution to the states is the most efficient and 5

11 appropriate way of operating government programs. Whether devolution is the best modus operandi for all government programs continues to be debated and often depends on an individual s personal view of the role of government. Apart from the academic and political debates, though, is the question of what the impacts are of operating this way and leaving so much of the level of support an individual can receive up to geographical chance. This study probes these big picture questions through the lens of one program and one group living in the United States. In the following chapters I explore these issues in more depth. In Chapter Two I provide additional background on the policies and programs being examined in this study, as well as further insight on the work of previous studies in this field. In Chapter Three I discuss the econometrics model and my departures from previous models, in addition to economic theories that shaped PRWORA. In the following section, Chapter Four, I discuss the data I use in this study, my assumptions, and the model variables. I also provide descriptive statistics for those variables to lay the foundation for the model s interpretation. I then discuss the findings from my quantitative analysis in Chapter Five for both poverty and labor force participation outcomes. Last, in Chapter Six I place these findings in the policy context and discuss future areas of research. 6

12 CHAPTER 2. BACKGROUND AND LITERATURE REVIEW This chapter provides background information about the government programs and policies under analysis in this paper and presents findings from key papers that investigated related topics. The presented information will form a knowledge foundation to make the quantitative results more meaningful. The chapter begins with a discussion of how PRWORA changed welfare administration, requirements, and eligibility and then provides a brief history of the development of poverty guidelines in the United States. The second half of the chapter is dedicated to a survey of the seminal papers on the impacts of PRWORA on the general population and immigrants, as well as papers on the determinants of poverty and workforce participation. The paucity of recent literature on the questions investigated in this study reinforces the importance of my study in providing an up to date analysis of PRWORA s effects on immigrant poverty and workforce participation. Focal policies and programs Welfare reform legislation (PRWORA) introduced a new welfare program, Temporary Assistance to Needy Families (TANF). The impact of the denial of welfare to immigrants through this legislation is the main focus of this study. Eligibility is determined in large part by a household s income level relative to the federal poverty guidelines, hence the development of these guidelines is also discussed here. Temporary Assistance to Needy Families (TANF) Before 1996, the major welfare program (i.e. means-tested cash assistance) in the United States was a federally designed and administered program named Aid to Families with Dependent Children (AFDC). AFDC was established in 1935 as part of the New 7

13 Deal under President Franklin Delano Roosevelt. However, under President Bill Clinton, the United States Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996 fundamentally changed the U.S. means-tested public benefits system. President Clinton had made welfare reform a campaign promise (Vobejda 1996). Besides adding new restrictions and work requirements, the program shifted to a statedesigned and administered program named Temporary Assistance to Needy Families (TANF). Under this new system, state governments are given block grants from the federal government to administer this program but they have significant discretion over the design and eligibility standards of the program in their own states. This has resulted in a complex web of 51 different systems 5 for providing public means-tested benefits. Beyond the immigrant provisions, three major changes as a result of PRWORA 6 are the emphasis of transitioning from welfare to work, lifetime limits, and the additional state flexibility. First, TANF recipients are required to be working in the two years after they start receiving benefits. The law also sets proportional quotas for the number of participants needing to participate in work activities (i.e. job training, working, volunteer service, etc.) (OSCE 1996) in a state at any given time, provides additional funding for childcare, and allows for transitional Medicaid after families begin jobs that do not provide health insurance. Second, families cannot access TANF for more than five years cumulatively. States have the option to make this more flexible for a limited proportion of people. States were allowed to apply for waivers to experiment with AFDC as early as 1962 and this flexibility continued to increase; by the mid-1990s, 27 states had AFDC waivers 5 Programs in 50 states and the District of Columbia. 6 The bill s other provisions included reinforcing child support, combating teen pregnancy, and expanding healthcare coverage for welfare recipients. 8

14 (Schoeni and Blank 2000). However, PRWORA granted much more flexibility to states than these ad hoc waivers. While state programs vary, Blank (2000, 11) concludes that the common denominator has been that most states are trying much harder to increase work among existing recipients, to support work among ex-recipients, and to encourage applicants to move into work rather than receive public assistance. Decreasing poverty through work is an implicit policy goal of PRWORA, but reducing families on the welfare rolls (and thus the cost) and incentivizing work were the explicit goals. Changes to immigrant eligibility and benefits While PRWORA changed the mechanism and design of welfare distribution across the country, the immigrant provisions are the among the most onerous (Primus 1996, 17) and, with the exception of SSI where immigrants comprised about 50% of recipients in 1995, considered by some researchers to be unnecessary. Immigrants had previously been able to access federal benefits since the 1935 Social Security Act. Under TANF, however, states not only had the power to design TANF programs, but they also had discretion over whether to provide TANF to qualified aliens 7 during the five-year bar, and whether immigrants could access Medicaid at the state s expense. This devolution of power and discretion to the states is another one of the greatest impacts of this piece of legislation. States made different choices immediately following PRWORA, and in the years since, about eligibility rules for immigrants, which state-funded benefits immigrants can access, and which immigrant groups can qualify. At the time of PRWORA s signing, it was projected that the bill would save $54 billion over six years, 7 Legal permanent residents; refugees, asylees, and persons granted withholding of deportation; persons paroled by the INS into the country for at least one year; and certain battered women and children (Primus 1996, 14). Undocumented immigrants made up the majority but not the entire not-qualified aliens category. 9

15 primarily from the food stamp reductions and the immigrant provisions (Vobejda 1996). At that time, the Congressional Budget Office (CBO) estimated that the immigrant provisions constituted $23.8 billion of these projected savings (Primus 1996). Their predicted number of people to be affected was significant: CBO estimated that 500,000 immigrants would lose SSI and 1 million would lose food stamps. However, rather than being a pure cost-saver, PRWORA for immigrants was largely a cost-shifter from one level of government to another for those states that chose to provide some benefits to immigrants (Passel and Clark 1998). In addition, negative effects were predicted to disproportionately affect immigrants. A 1996 Urban Institute study (Zedlewski et al.) estimated that immigrants would constitute 46% of the 2.6 million individuals and 41% of the 1.1 million children in the United States who would fall below the poverty threshold as a result of PRWORA. Under the new legislation, most lawful permanent residents who arrived before August 22, 1996 were supposed to lose eligibility for federally funded benefits and services by January 1, This included not just TANF but food stamps (now SNAP), SSI, and Medicaid. Upon signing PRWORA into law, President Clinton promised to restore many of the benefits to this population (Vobejda 1996). As discussed in Chapter One, subsequent federal legislation combined with state action did in fact restore many benefits to this pre-prwora immigrant population. However, all immigrants who arrive after August 22, 1996 were barred from accessing TANF and Medicaid for their first five years here, and from SSI and food stamps until obtaining U.S. citizenship. 8 In addition, the income of an immigrant s sponsor was to be included in his/her application for public 8 Exceptions are: refugees and asylees during their first five years in the country, those who have served in the Armed Forces and their dependents, and those with substantial work histories- who can demonstrate 40 quarters of covered employment (Primus 1996, 15). 10

16 benefits for ten years, which was predicted to push many over the eligibility threshold. Federal eligibility for non-immigrants, undocumented immigrants, and refugees generally did not change, making the documented immigrant group who lost benefits the primary group of interest to this study. History of U.S. poverty guidelines This study uses the standard federal definition of the poverty threshold, as it is the measure used by the Census Bureau to provide statistics on poverty in the United States and is employed by state and federal agencies for determining eligibility for means-tested public benefits. However, the measure is not comprehensive and the basic premises behind the guidelines have not changed since they were articulated in Alternative measures have since been developed but are not widely employed for anything beyond research purposes. The official poverty measure used by the U.S. government was designed by Mollie Orshansky, who was working as an economist at the Social Security Administration when she wrote two articles on proposed poverty thresholds, a 1963 article that focused on a threshold for families with children and a 1965 article 9 which included two possible thresholds and measures for single individuals and two-person families (i.e. couples without children or single parents). Her thresholds built on four food plans developed by the Department of Agriculture: the liberal, moderate, and low cost plans developed in 1933 and the economy plan in 1961 following the 1955 Household Food Consumption Survey. Orshansky also used the conclusions of the 1955 Survey that for all families of three or more individuals in the United States, that the 9 These are the articles referenced by Fisher (1992): Orshansky, "Children of the Poor," Social Security Bulletin, Vol. 26, No. 7, July 1963, pp. 3-5; Orshansky, "Counting the Poor: Another Look at the Poverty Profile," Social Security Bulletin, Vol. 28, No. 1, January 1965, pp

17 average expenditure for all food used both inside and outside of the home during a week accounted for about one third of their average money income after taxes (Fisher 1992). Pairing this information with Engel's Law, an economic principle that the proportion of income allocated to necessaries, and in particular to food, is an indicator of economic well being (Fisher 1992), Orshansky theorized that the poverty threshold should be three times the dollar amount a family would spend on the economy food plan. Food expenses for two-person families were multiplied by 3.7 based on information from the 1955 Household Food Consumption Survey, and for individuals the threshold was set at 80% of the two-person family measure. In addition, Orshanky s original model included dozens of different calculated thresholds for different types and sizes of families living on farms and not on farms for both the low cost and economy food plans. A consolidated version of Orshansky s thresholds inadvertently ended up being how the United States measures poverty. In 1965, the federal government first began using the lower of the two thresholds Orshansky developed as a working or quasi official definition of poverty (Fisher 1992) and it became the official measure in It has remained the official measure. Fisher (1992) points out several problems with continuing to use Orshansky s model. One problem is that the Current Population Survey uses pre-tax income for its statistics while the food plans were constructed using information from post-tax income. Orshansky herself said this measurement difference would result in a conservative underestimate of poverty levels (Fisher 1992). Second, the economy food plan does not allow for eating out (common in the United States and for single, working parents), does not guarantee sufficient food levels for adequate nutrition, and was designed for 12

18 temporary or emergency use when funds are low." Next, although later surveys revealed that food expenses should be multiplied by a factor of more than three, the multiplier and the food price levels remained based on 1963 estimates. After 1969, Orshansky s 1963 poverty threshold would be adjusted for changes in overall price levels according to the Consumer Price Index, but not specifically for food prices and how they relate to overall income. A few minor changes were made in 1981, but despite numerous studies showing the deficiencies of Orshansky s measure for today s society, no substantial revisions have been made. Given the inadequacies of these measures, the National Academy of Sciences developed alternative methods for measure poverty in These thresholds included adjustments for median expenditures on food, clothing, shelter, and utilities as reported in the Consumer Expenditures Survey, taxes paid and tax breaks, medical expenses, work expenses such as childcare, non-work expenses such as household supplies, geographical adjustments, and price changes (Short et al. 1999). The results of using these alternative measures are mixed; compared to the official measure, accounting for in-kind transfers to the poor and tax benefits tends to lower the number of people in poverty while adding expenses besides food tends to increase the poverty rate significantly. The Census Bureau subsequently created 12 variables to measure some of the characteristics of these alternative poverty thresholds. These variables account for medical out of pocket (MOOP) costs in three different ways, updates the threshold using the standard Consumer Price Index method and using information from the Consumer Expenditure Survey, and provides geographically adjusted and non-geographically adjusted versions 13

19 of these measures (IPUMS-CPS 2011). These alternative measures have not influenced the official definition that affects standard statistics on poverty and benefits calculations. Expanding upon this work by the National Academy of Sciences, in November 2011 the Census Bureau released a new poverty measure named the Supplemental Poverty Measure (SPM). The goal of this new measure was to compile the components from those various alternative measures of poverty and update poverty calculations based on today s costs and resources, with the ultimate goal of eventually publishing this alternative measure simultaneously with the official measure every year (Short 2011). Rather than only looking at income from employment, the new measure includes government food, housing, and energy subsidies as well as tax credits in a household s resources and subtracts medical, childcare, and work-related costs as well as taxes. By including a range of resources and expenses rather than only food costs, which are only about a seventh rather than the estimated third of a household s income (Tavernise and Gebeloff 2011), the measure is a more realistic portrait of a household s expendable income. Some subsets of the population fared better with this new estimate and others fared worse. While the overall proportion of the population in poverty with this measure edged up to 16% in 2010 compared to the official 15.2%, the biggest increase was in the population near poverty. The number of households with incomes less than 50% above the poverty threshold under this new measure is 76% higher than the official measure, placing a third of the entire U.S. population in or close to poverty (DeParle, Gebeloff, and Tavernise 2011). While this measure is not available yet as a variable that could be analyzed with CPS data, it is interesting to note how the immigrant population, largely unable to access government benefits, fared under this new measure. Under the new 14

20 measure, 25.5% of the foreign-born population is in poverty, compared with 20% under the official measure (Short 2011). About a third (32.6%) of non-citizens are in poverty under the new measure, 5.7 percentage points higher than the official measure (Short 2011). Immigrant poverty under the official poverty guidelines is examined in this study, but does not provide the whole picture. Prior studies and research context Effectiveness of PRWORA The research is somewhat divided over how PRWORA affected benefits usage and workforce participation among immigrants. While policymakers could quickly point to decreases in welfare caseloads after PRWORA (50% between 1994 and 1999 and 47% from 1997 to 2008 (Kaminski 2011)), the key issue is how immigrant behaviors changed relative to natives since they were particularly targeted. First, studies on benefits usage compare the decline after PRWORA among immigrants to the decline among natives. As referenced previously, Fix and Passel (1999) found that before welfare reform all non-citizens with incomes below 200% of the federally established poverty level had similar rates of benefits usage as their citizen counterparts but after reform their benefits usage fell below that of citizens. The authors attribute the decline in benefits usage among still-eligible populations to a chilling effect, where benefits providers and immigrants themselves are not sure who qualifies and immigrants may be reluctant to apply. Whereas initially Borjas (2002) concludes that immigrant participation in welfare programs nationally declined after PRWORA, he finds that once California was removed from the sample, welfare participation did not change significantly. He attributes this lack of national decline in welfare participation among 15

21 immigrants to state provision of benefits and to increases in naturalization. Interestingly, according to Borjas analysis, California provided some of the most generous welfare benefits to immigrants after welfare reform, and yet California was also the driving factor in Borjas estimate of national decline. He does not find any concrete, empirical explanations for this phenomenon, but theorizes that anti-immigrant sentiment, manifested through California Proposition 187, played an important role, possibly by making immigrants afraid to apply for benefits they were eligible for or making service providers uncertain of the policy landscape. While Proposition 187 was overturned by the courts, it still later shaped the national immigrant restrictions on welfare in PRWORA. Conclusions regarding PRWORA s effects on workforce participation are similarly divided. Borjas (2003) finds that participation in Medicaid by non-citizen immigrants living in states that chose not to provide public assistance to immigrants declined relative to natives as a result of PRWORA, but that immigrants overall health insurance coverage increased due to more of this immigrant population receiving employersponsored health insurance. Borjas (2003) subsequently explores this uptake in employersponsored health insurance as it relates to the labor supply of immigrants and finds that the labor supply of non-citizen immigrant men increased in those states that chose not to provide benefits. He concludes that removing benefits is a motivator for increasing labor supply to compensate for those losses. Conversely, some other researchers (Blank 2000) find that overall decreases in benefits use and increases in immigrant labor force participation can be attributed to the other forces, including a robust economy supporting increased employment and reducing the need for welfare, and the introduction of the Earned Income Tax Credit (EITC) that 16

22 made work more profitable. More specifically, Blank and Schoeni (2000), citing a 1999 Council of Economic Advisors study, attribute only a third of the decline in welfare participation from to PRWORA. Blank and Schoeni (2000) also find that PRWORA had a statistically significant effect on welfare participation, decreasing participation by two percentage points, but no statistically significant effect on workforce indicators or earnings. Disputing Borjas findings, Blank and Schoeni (2000) attribute the increases in workforce participation of low-skilled women during this time period largely to economic growth. They also find important distributional effects from PRWORA: less skilled women as a whole experienced more of the positive effects of the policy (e.g. greater declines in poverty and increases in income) than more skilled women, but of these less skilled women, the women at the mean income level and higher appear to have experienced most of these positive effects. Determinants of poverty and workforce participation Various observed and unobserved personal characteristics can affect an individual s propensity to be in poverty. The differences in poverty levels according to education, race, and sex at different points in time in the United States are expanded upon through descriptive statistics in Chapter Four. While my econometrics model provides a tool for examining the relationship between an individual s immigration or citizenship status and poverty, the vehicle is how removal of previously conferred government benefits plays a role in this relationship. This study focuses not on an immigrant s personal characteristics, controlled for in the model, but the structural, government choices that affect socioeconomic outcomes like poverty and workforce participation. 17

23 The literature confirms the importance of the government s role in providing benefits that keep people out of poverty. Thus, for example, Brady and Kall (2008) focus on how macro-level sources of overall poverty (979) affect poverty outcomes for men and women in Western democracies differently. Their study identifies three predominant theories on these macro-level determinants: 1) liberal economic theories, which focus on the importance of economic growth, unemployment, productivity, and education; 2) structural theories, focusing on the industry composition of a country and demographic characteristics such as female labor-force participation, age and sex ratios, and single parenthood; and 3) power resources, including welfare benefits, public employment, and income redistribution (979). They conclude that social safety net spending, such as on retirement, welfare payments, unemployment insurance, and public health coverage, has the largest and most significant impact on reducing poverty for both men and women, with the magnitude of the impact being even greater for women. Similarly, Moller et al. (2003) view the reduction in poverty as driven by state action in redistributing income (23). They focus on various government policies, constitutional makeup, and political process that affect poverty reduction. In contrast to other researchers and economic theories that argue welfare reduces the incentive to work and will lead to poverty, they see welfare spending levels as coexistent with high spending on workforce development. Their study also concludes that more generous welfare states have significantly greater reductions in poverty (43). Two seminal studies, Pencavel (1986) and Killingsworth and Heckman (1986), have continued to influence how researchers think about the determinants of workforce 18

24 participation. Pencavel (1986) summarizes findings of the literature on various aspects of male workforce participation: the positive relationship between educational attainment and workforce participation, the greater likelihood for white men to participate than black men, the positive influence of being married on participation, the negative relationship between non-wage income and workforce participation, and the quadratic relationship between age and participation. Pencavel s study focuses on the effects that wage changes have on male workforce participation; he concludes that while wages and prices do impact behaviors, the overall impact is fairly small and is sometimes negative (i.e. higher wages resulting in lower participation) and thus the other determining factors should be explored in more depth. In the case of women, Killingsworth and Heckman (1986) identify marital status, family composition, and type of job as influencing workforce participation factors for women. My study is primarily a contribution to the exploration of macro-level government policies that affect poverty and workforce participation of a particular subset of the population, immigrants, while controlling for individual characteristics. 19

25 CHAPTER 3. CONCEPTUAL MODEL Research hypotheses My primary hypothesis tests the effect of the 1996 welfare reform on poverty for immigrants in the United States, particularly for non-citizens. I expected poverty to increase for immigrants as a result of this legislation, as a key social safety net in existence for 60 years prior to PRWORA was abruptly removed. As previously noted, prior research produced mixed findings on the effects of PRWORA immediately following the legislation. This study aims to provide a clearer perspective on welfare reform s effect on immigrants more than a decade years after the legislation was passed, as service providers and individuals have adapted. To accomplish this, I analyze how the incidence of poverty for immigrants changed across post-prwora periods and across states making different choices about providing welfare to lawful permanent residents during the five-year bar. My secondary hypothesis tests the effects of the 1996 welfare reform on workforce participation of immigrants. Specifically, I test whether immigrant workforce participation increased to compensate for a lack of the availability of cash benefits provided by the pre-prwora welfare programs. While I focus on a dichotomous variable representing whether an individual is in the workforce or not, I also briefly examine changes in the average number of hours worked per week. Analysis model Economic model One way of conceptualizing the theory behind how welfare reform was designed is through the lens of a labor-leisure model. Removing an individual s welfare benefits 20

26 means that leisure becomes more expensive, so a person is theoretically more motivated to work. In classical economics the supply of labor can be both an upward sloping and backward bending curve, such shown in figure 3.1: Wage (dollars per hour) Hours of work per day FIGURE 3.1. Supply of labor Source: Pindyck and Rubinfeld 2001, 511 As wages increase a person has a greater incentive to work more, substituting work for leisure (the substitution effect). However, when wages become high enough individuals do not need to work as many hours to sustain themselves and their consumption preferences (income effect) and the supply curve bends backwards. Since wages for the majority of low-income people will not increase enough to dramatically decrease the need to work, at least in the short-term (i.e. income effect will not dominate the substitution effect), most analysis related to the impacts of welfare focuses on the upward-sloping portion of the labor supply curve. As seen in figure 3.2, introducing TANF benefits to individuals theoretically decreases the number of hours they work and increases social inefficiencies: 21

27 $ of consumption per year $25,000 slope = wage = $12,500 X2 $0 X 1 Leisure FIGURE 3.2. Consumption vs. leisure with welfare Source: Gruber 2007, 490 Assuming that wages are constant, if the government guarantees a certain level of income (for example $12,500 per year in figure 3.2) through welfare, then those with strong preferences for leisure will have an incentive to stop working all together (such as person X moving from X 1 to X 2 ). Individuals with strong preferences for consumption above the $12,500 are not affected by the addition of a welfare benefit in this simple model. Designers of welfare programs sometimes experiment with the benefit reduction rate based on wages as a way to prevent the incentive to reduce workforce participation because of the presence of welfare. Since immigrants became ineligible for welfare presentation, the above graph shows a 100% benefit reduction rate. This theory forms the theoretical base of why workforce participation would increase as welfare was removed. Statistical Model The statistical model in this paper, a triple difference linear probability model, largely draws from the work done by Borjas (2003), seen in equation 1: 22

28 EQUATION Probability of healthcare coverage by immigration status and PRWORA y ij = X ij β + α 0 t ij + α 1 I ij + α 2 Gj + γ 0 (I ij t ij ) + γ 1 (I ij G j ) + γ 2 (G j t ij ) + θ(i ij G j t ij ) + ε ij, where Dependent variables y ij = three measures of health insurance coverage for an individual in a particular state or h ij = three measures of the labor supply of an individual in a particular state, Independent variables X= socioeconomic vector of refugee status, year of entry into the United States, age, gender, race, educational attainment, size of household, number of children, elderly, and disabled in the household, the state unemployment rate, and the state unemployment rate interacted with the immigration vector, t= post-prwora period, I = two dummy vector of citizenship status (naturalized citizens and non-citizens), and G = state s generosity towards immigrants dummy (set to unity if not classified as generous to post-prwora immigrants) A triple difference linear probability model adds another dimension to the traditional difference in difference model, making it a difference in difference in difference model. This model not only compares the outcomes of a group to a similar group by policy change but also across time, essentially a three-dimensional model. My model, outlined more thoroughly in the next chapter, examines changes in probability of being in poverty and workforce participation by citizenship and immigration status by state policy choices about providing welfare (i.e. TANF) to immigrants and by comparing one pre-prwora period to two post-prwora periods. As discussed in Chapters One and Two, Borjas (2003) examines how healthcare coverage of immigrants changed in the post-prwora period. Borjas (2003) uses the model in equation 1 to analyze the likelihood that naturalized citizens and non-citizens have health insurance coverage using 10 Borjas 2003,

29 three different specifications 11 in states that chose to provide a variety of means-tested benefits to immigrants during the five-year bar established by PRWORA. He then similarly analyzes changes in labor force participation for these groups using three specifications 12 with the same model. I closely follow the Borjas (2003) model for a similar analysis of socioeconomic changes (in this case poverty) and workforce participation after PRWORA of immigrants. Nonetheless, there are key differences between the models in Borjas (2003) and my own beyond my use of poverty rather than healthcare coverage as the independent variable. First, I do not include variables for the number of elderly and disabled persons in a household because these variables are not currently available in the Current Population Survey. I also exclude the family size variable as it is strongly correlated with the number of children in a household. Second, for both models I restrict the sample to persons years old for both models. Borjas (2003) restricts both models to persons under 65 but only restricts the labor supply model to those above 18. Third, I employ the Passel and Clark (1998) coding techniques, rather than the Borjas method, for determining refugee status. Fourth, rather than the multi-factor approach primarily used by Borjas (2003), I restrict my coding of state benefits provision to provision of TANF to immigrants during the five-year bar. Fifth, I also depart from Borjas (2003) by determining a person s nativity and citizenship status not by the status of the head of household but by their status in their own right. The age restrictions, individual citizenship status determination, and TANF specification are meant to make the effects measured more conservative and 11 The specifications are Medicaid coverage, any sort of health insurance coverage, and employersponsored insurance. 12 The specification are a dummy variable set to unity if the individual worked the previous week, the log of total hours worked in the previous year, and a dummy variable set to unity for working full-time in the previous year. 24

30 precise, focusing on a specific population and program; the other differences are minor. Additionally, changes in healthcare coverage are arguably more closely tied to workforce participation than poverty, since one can be working full-time and still be in poverty. 25

31 CHAPTER 4. DATA AND METHODOLOGY Data sets Primary data This paper uses data from the March Supplement of the Current Population Survey (CPS). The CPS survey, performed monthly by the Census Bureau on behalf of the U.S. Department of Labor, has been carried out in every state and Washington, D.C. for the last 50 years. In general, every month approximately 57,000 households are interviewed, totaling 112,000 individuals who are 15 years or older. The CPS also collects demographic data for children. The March Supplement in particular is widely used by researchers of interest in this study (Borjas 2002; Borjas 2003; Fix and Passel 1999) because it provides detailed, yearly data on workforce demographics, employment, and income. While the Census Bureau and the National Bureau for Economic Research (NBER) provide raw data CPS March supplement data directly, this paper uses the integrated data from IPUMS-CPS, a project of the Minnesota Population Center (King et al.). The Minnesota Population Center has extracted the raw CPS March Supplement data from , pooled the years, and generally harmonized the variables across years to account for changes in survey questions. IPUMS provides documentation for any differences that had to be accounted for across survey designs in the descriptions of the downloadable variables, supplies markers to match individual respondents with the household head, provides weight variables to correct for sub-population oversampling, and includes data quality flags for responses that have been edited because they were missing, illegible, or inconsistent. 26

32 CPS analysis sample My sample is restricted to calendar years and 1995 (just before the enactment of PRWORA on August 22, 1996), 1998 and 1999 (just after PRWORA), and 2008 and 2009 (more than a decade after PRWORA). The use of these pre- and the initial post-prwora dates are generally consistent with the initial PRWROA studies (Borjas 2002; Borjas 2003; Fix and Passel 1999). I have further restricted my sample to persons 18 to 64 years old to best reflect the working age population with the most agency over labor and economic outcomes. Additionally, persons below age15 are considered not in the universe for several key variables. 14 These restrictions dropped 381,866 observations from the sample but 582,009 observations remained. Complementary data After welfare reform, states had the option to provide state funded public benefits to immigrants who were barred from receiving federally funded benefits. There are 51 different TANF policies (for the 50 states and the District of Columbia), which can change yearly, with changes published in manuals for service providers and/or through policy memos and briefs. To keep track of these differences and annual changes, the Urban Institute manages a database, the Welfare Rules Database (WRD), which provides information on welfare eligibility, payments, and practices through a grant from the U.S. Department of Health and Human Services (The Urban Institute). The researchers at the Urban Institute update the database through analyzing state policy documents, compiling 13 CPS surveys ask about income, benefits receipts, and (generally) employment for the year prior to the survey. Demographic and other personal characteristic variables come from the current year but are assumed to have not changed dramatically from the previous year. Therefore, calendar years 1994, 1995, 1998, 1999, 2008, and 2009 correspond to the 1995, 1996, 1999, 2000, 2009, and 2010 CPS. The calendar year will be referred to in the rest of this study. 14 Weeks worked last year, usual hours worked last year, labor force participation, and highest level of educational attainment. 27