Welfare Reform and Older Immigrant Adults’ Medicaid and Health Insurance Coverage

Abstract

Objectives: To examine how federal noncitizen Medicaid eligibility restriction and generous state policy affect Medicaid and health insurance coverage among older adults with different citizenship status. Method: This study uses an older adult sample (65 years or older) from the Current Population Survey (CPS) and state data and employed a triple difference-in-differences approach to incorporate variations in citizenship status, time, and state eligibility. Results: Findings show that Medicaid coverage significantly declined among older noncitizens but increased among older naturalized citizens after Welfare Reform. Findings also show that the differences in older noncitizens’ health insurance coverage changes were significant between generous and nongenerous states. Discussion: Medicaid eligibility affects older immigrant adults’ Medicaid and health insurance coverage. Findings support the “protective citizenship” hypothesis but not the “chilling effect” and “labor market condition” hypotheses. Opposite patterns of change in Medicaid coverage between naturalized citizens and noncitizens raise doubt about the effectiveness of eligibility restrictions in reducing government spending.

Keywords

immigration policy evaluation federal and state policy health policy

Medicaid has played a critical role in protecting low-income individuals’ access to preventive and needed health care services by covering them with government-funded health insurance (Auchincloss, van Nostrand, & Ronsaville, 2001; Ayanian, Weissman, Schneider, Ginsburg, & Zaslavsky, 2000; Berk & Schur, 1998). Older immigrant adults are one of the major beneficiary groups of Medicaid. A high proportion of older immigrant adults participate in Medicaid: 22% of older immigrant adults were covered by Medicaid in the early 2000s, much higher than older native citizens whose coverage rate was only 8% during the same period (Nam, 2008). In addition, Medicaid is often the only health insurance option for older immigrant adults since a substantial proportion of them do not have Medicare, a common health insurance among older native citizens (Ku, 2009a).

Despite Medicaid’s crucial role in fulfilling the health care needs of older immigrant adults, access to Medicaid for this group of individuals has become constrained since the passage of Welfare Reform in 1996. Welfare Reform imposed eligibility restrictions on noncitizens. Under the current eligibility rules, noncitizens are no longer eligible for federally funded Medicaid if they immigrated to the United States after Welfare Reform and have not lived in the country for 5 years or longer. As a result, noncitizens have access to Medicaid for the initial 5-year period only if they live in generous states in which state-funded Medicaid programs cover those ineligible under the federal rule (Nam, 2011a; National Immigration Law Center, 2002; Zimmermann & Tumlin, 1999).

Although more than 10 years have passed since Welfare Reform, little is known about the impact of the new eligibility rules on Medicaid and health insurance coverage among older immigrant adults. The majority of existing studies investigate the impacts of Welfare Reform on immigrant children and working-aged immigrants (Borjas, 2003; Haider, Schoeni, Bao, & Danielson, 2004; Kaushal & Kaestner, 2005). Among the small number of studies on older immigrant adults, most rely on descriptive statistics (Fix & Passel, 1999; Ku, 2009b). Furthermore, these studies compared Medicaid and health insurance coverage between pre– and post–Welfare Reform periods without taking into account state variations in Medicaid eligibility (Fix & Passel, 1999; Ku, 2009a; Nam, 2008). One recent study included state variations in Medicaid eligibility in analyses but did not differentiate older naturalized citizens from noncitizens, two groups with distinct eligibility (Nam, 2011b). As a result, these studies are unable to fully estimate the impact of Medicaid eligibility changes on older immigrants by overlooking potential differences between generous and nongenerous states or the distinction between naturalized citizens and noncitizens.

To fill gaps in the existing literature, this study investigated whether and how Medicaid and health insurance coverage of older immigrant adults have changed following the passage of Welfare Reform, using a nationally representative sample of older adults from the Current Population Survey (CPS) and state-level data collected by the author. Results of this study contribute to the literature by focusing on older immigrant adults, a vulnerable but understudied population (Mold, Fryer, & Thomas, 2004; Nam, 2008). In addition, this study assessed the impacts of Medicaid eligibility while considering both federal and state eligibility rules. This study also investigated whether changes in Medicaid eligibility had different effects on older immigrant adults with different citizenship status: naturalized citizens and noncitizens. In this way, the study tested the three competing hypotheses on the effects of Welfare Reform on immigrants: the “chilling effect” of Welfare Reform, the emergence of “protective citizenship,” and the distinct effects of “labor market conditions” by citizenship status.

Background

The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (hereafter called “Welfare Reform”) has constrained immigrants’ access to public benefits by imposing restrictive eligibility rules on noncitizens. The aim of this policy change was to reduce the government’s expenditures by shifting the financial responsibility of supporting poor immigrants from the government to individuals and families (Congressional Budget Office, 1996). To achieve this goal, Welfare Reform imposed restrictive noncitizen eligibility rules on various public assistance programs, including Medicaid, Supplemental Security Income (SSI), and Temporary Assistance to Needy Families (TANF). For example, Welfare Reform made noncitizens ineligible for SSI unless they were already on the program before Welfare Reform (Gerst & Burr, 2011; Nam, 2011a). At the same time, Welfare Reform barred state governments from covering recent immigrants (noncitizens who immigrated to the United States after Welfare Reform and lived in the country for less than 5 years) with federally funded Medicaid. Welfare Reform also permitted state governments not to cover noncitizens with Medicaid after the 5-year bar period. Responding to shifts in federal eligibility rules, a small number of states established state-funded Medicaid programs to cover noncitizens ineligible under the federal rule. As a result, recent immigrants without citizenship do not have access to Medicaid unless they live in generous states with state-funded Medicaid. These new eligibility rules are fundamentally different from those in the pre–Welfare Reform period during which noncitizens possessed the same rights to public benefits as native citizens regardless of their states of residence as long as they were legal permanent residents (Nam, 2011a; National Immigration Law Center, 2002; Zimmermann & Tumlin, 1999).

Despite dramatic shifts in public assistance programs after Welfare Reform, we know little about the impact of these policy changes on immigrants (Blank, 2002). Although existing studies agree that noncitizens’ Medicaid coverage rates declined after Welfare Reform, they have not reached a consensus on why this happened (Borjas, 2003; Fix, Capps, & Kausbal, 2009; Kandula, Grogan, Rathouz, & Lauderdale, 2004; Kaushal & Kaestner, 2005). Some explain the decreased Medicaid coverage rates as the “chilling effect” of Welfare Reform. According to this hypothesis, immigrants, even eligible immigrants, voluntarily withdrew from public benefit programs out of fear of potential disadvantages: They were afraid that they would not be able to sponsor their family members’ immigration to the United States or that they might be deported if they received public benefits (Fix & Passel, 1999; Kandula et al., 2004; Ku & Matani, 2001). Others attribute the decrease in noncitizens’ Medicaid coverage to the emergence of “protective citizenship.” According to this hypothesis, noncitizens’ Medicaid coverage rates decreased following Welfare Reform simply because those in need acquired citizenship to ensure their access to public benefits. That is to say, the same group of immigrants may have participated in public assistance programs, but their citizenship status may have changed from noncitizen to naturalized citizen after Welfare Reform (Gilbertson & Singer, 2003; Ku, 2009a; Nam & Kim, in press; Van Hook, 2003). Another hypothesis explains decreased Medicaid coverage observed among noncitizens with labor market conditions by showing that the effects of Welfare Reform lose their statistical significance after controlling for labor market conditions (estimated with unemployment rates) and its interaction term with citizenship status. Based on this finding, the “labor market condition” hypothesis claims that shifts in noncitizens’ public assistance program participation rates are attributed to labor market conditions, not to eligibility restrictions: Noncitizens’ Medicaid coverage rates declined following Welfare Reform because the robust economy during the mid- to late-1990s reduced their economic need to a greater extent than it did for native citizens (Haider et al., 2004; Lofstrom & Bean, 2002).

In comparison to Medicaid, we know even less about the impact of Welfare Reform on overall health insurance coverage. Borjas (2003) argues that Welfare Reform made little impact on health insurance coverage rates among noncitizens based on the finding that their employment-based insurance coverage increased while their Medicaid coverage decreased significantly after Welfare Reform. Owing to the offset between Medicaid and employment-based insurance coverage, noncitizens’ overall health insurance coverage remained stable between pre- and post-reform periods. A study on a more vulnerable immigrant population, however, tells a different story. Using a sample consisting of children of single immigrant mothers, Kaushal and Kaestner (2005) suggest that Welfare Reform had negative effects on health insurance coverage among this disadvantaged group: An increase in employment-based insurance coverage was not large enough to compensate for a decrease in Medicaid coverage.

In addition, there is less research on older immigrant adults compared to younger immigrants. Only a small number of empirical studies report separate analysis results from older immigrant adult samples (Fix & Passel, 1999; Ku, 2009b; Nam, 2008). By comparing pre- and post-reform period statistics, Fix and Passel show that older immigrant adults’ Medicaid coverage did not decline shortly after Welfare Reform: It remained stable at 28.2% between 1994 and 1997 among older noncitizens but increased from 11.1% to 14.9% among older naturalized citizens. Using more recent data, Ku (2009a) found that the percentage of older noncitizens without health insurance increased following reform. Nam’s (2008) regression analyses show that changes in Medicaid coverage differed by citizenship status: It significantly decreased among older noncitizens but significantly increased among older naturalized citizens. Nam (2008) indicates that overall health insurance coverage, however, remained stable between pre- and post-reform periods among both older naturalized citizens and noncitizens. Another study by Nam (2011b) examined the overall impact of federal and state eligibility rules on older immigrants without differentiating noncitizens from naturalized citizens. Nam’s (2011b) multiple regression analyses showed that the older immigrant adults’ Medicaid coverage significantly declined after Welfare Reform, but generous state rules significantly reduced the negative impacts of the federal eligibility rules on older immigrant adults.

Valuable as they are, the existing studies have limitations. No known study considers state variations in noncitizen eligibility and citizenship status at the same time in investigating the effects of Welfare Reform on older immigrant adults’ Medicaid and health insurance coverage. A small number of studies investigated the impacts of Welfare Reform on older immigrant adults’ participation in Food Stamps and SSI while considering state eligibility rules and citizenship status (Gerst & Burr, 2011; Nam & Jung, 2008). These studies, however, are unable to inform us on how federal and state eligibility rules affect older immigrant adults’ Medicaid and health insurance coverage. Most existing studies on older immigrant adults’ Medicaid and health insurance coverage incorporated only the federal-level eligibility changes but not the state-level variations in their analyses (Fix & Passel, 1999; Ku, 2009a; Nam, 2008). The lack of attention to state-level variations may limit capturing the effects of policy changes fully. Numerous studies on other policy areas clearly indicate that state-level policies do affect target populations (Bansak, 2006; Nam & Jung, 2008; Powers, 1998), including one study focusing on older immigrant adults’ Food Stamp program participation (Nam & Jung, 2008). Although Nam (2011b) included state generosity in her analyses, her study did not differentiate naturalized citizens from noncitizens despite that impacts of Welfare Reform on immigrants were estimated to differ by citizenship status (Borjas, 2003; Nam, 2008; Nam & Jung, 2008).

To fill the gaps in the existing literature, this study examined whether and how Medicaid and health insurance coverage changed among older immigrant adults after Welfare Reform while taking into account variations in federal and state eligibility rules and older immigrant adults’ citizenship status. At the same time, this study tested the three competing hypotheses suggested in previous studies: the “chilling effects” of Welfare Reform, the emergence of “protective citizenship,” and the effects of distinct labor market conditions. The chilling-effect hypothesis predicted that Medicaid coverage rates would decline after Welfare Reform not only among noncitizens but also among naturalized citizens. Although their eligibility remained unchanged after Welfare Reform, naturalized citizens may have dropped out of the Medicaid program because they were afraid of potential disadvantages from public assistance program participation (Kandula et al., 2004). In contrast, emergence of the “protective citizenship” hypothesis predicted that the Medicaid coverage rate would increase among naturalized citizens but decrease among noncitizens because eligibility restrictions would motivate immigrants to acquire citizenship to ensure access to public benefits. The “labor condition” hypothesis predicted Medicaid coverage of noncitizens and naturalized citizens to remain stable when the differential effects of labor market conditions by citizenship status were controlled for.

Method

Data and Sample

This study used individual-level data from the Annual Social and Economic Supplement (ASEC Supplement) of the CPS and state-level data collected by the author. The CPS collects information on demographic, social, and economic characteristics from a nationally representative sample of the noninstitutionalized civilian population on a monthly basis, using a rotating panel design. The CPS is assessed to collect accurate and reliable data on the immigrant population as well as the native-born population (Schmidley & Robinson, 2003). The ASEC Supplement provides information on health insurance and public assistance program participation in greater detail than what is available in the CPS (U.S. Census Bureau, 2008).

This study utilized CPS data from the pre- and post-reform periods (1994-1996 and 2003-2008). It excluded data collected between 1997 and 2002 since some states made changes to the noncitizens’ Medicaid eligibility rules during this period. For instance, New York reinstated noncitizen’s Medicaid eligibility in 2002 after the New York Court ruled that the state government cannot deny legal immigrants’ rights to state-funded Medicaid programs, and Washington stopped covering noncitizens with state-funded Medicaid in 2002 (Fremstad & Cox, 2004).

In addition to individual-level data, this study used state-level data compiled by the author. First, the study used existing literature (National Immigration Law Center, 2002; Zimmermann & Tumlin, 1999) to get information on state Medicaid eligibility rules between 1996 and the early 2000s. Second, the study downloaded state Medicaid manuals from 50 states’ websites to collect information on noncitizen eligibility in 2008. Third, the study conducted telephone interviews with state Medicaid directors and policy analysts to obtain additional information. Through telephone interviews, it verified the accuracy of the data collected using the existing literature and Medicaid manuals, and gathered information otherwise unattainable, such as exactly in what year state noncitizen eligibility had changed. As such, this study used more updated data on state Medicaid eligibility rules than existing studies (National Immigration Law Center, 2002; Zimmermann & Tumlin, 1999). In addition to state Medicaid eligibility rules, this study included state-level unemployment rates in analyses. The annual state unemployment rates used in this study were obtained from the U.S. Bureau of Labor Statistics.

The sample of this study consisted of older adults: 65 years or older at the time of the interview. This study does not include Native Americans (N = 778) in the sample because CPS provides inconsistent data on Indian Health Service coverage across the observation period of this study, which could affect the estimation of health insurance coverage of Native Americans (State Health Access Data Assistance Center, 2006). The sample of this study included only the first observation for each individual because multiple observations from the same individuals are likely to underestimate standard errors (Greene, 2003). The final sample included 105,873 older adults.

Measures

The dependent variables for this study consisted of two dichotomous variables: Medicaid and health insurance coverage. Since the CPS added a health insurance verification question in 2002, which was estimated to raise health insurance coverage rates by 8 percentage points (Nelson & Mills, 2001), this study treated those identified as insured only through the verification question as “uninsured” to generate consistent measures across the observation period. For the Medicaid coverage variable, a value of “1” was assigned to those covered with Medicaid and “0” to others. The health insurance variable was assigned the value of “1” for those covered with any of the following types of health insurance: Medicaid, Medicare, private insurance, and military insurance. For this variable, the value “0” was allocated to others.

The citizenship variable was composed of three groups: native citizens (those born in the United States or to a citizen parent if born outside of the U.S. territories), naturalized citizens (foreign-born individuals who became citizens through naturalization), and noncitizen (foreign-born individuals who are not naturalized). The Welfare Reform indicator separated postreform observations from prereform ones: Older adults included in the data collected between 1994 and 1996 had the value of “0” in the Welfare Reform indicator and those between 2003 and 2008 had “1.”

The state generosity variable was created based on each state’s noncitizen Medicaid eligibility rules adopted after Welfare Reform. This study defined a generous state as one with a state-funded Medicaid program that covers noncitizens ineligible under the federal rule. Nine states fit such definition of generous states: California, Delaware, Hawaii, Maine, Massachusetts, Minnesota, Nebraska, New York, and Pennsylvania. Although Massachusetts reduced its noncitizen eligibility from full to partial coverage in 2003 it was identified as generous because of its broader coverage: Massachusetts’s state Medicaid program covers disabled and long-term unemployed noncitizens as well as pregnant women and children. Supplemental analyses with the sample excluding Massachusetts residents produced substantively identical results as those reported here. In creating the state generosity variable, this study assigned the same value to both pre- and post-reform periods to each state despite that most states have different Medicaid eligibility rules between the two periods: Every state was required to cover noncitizens under the same rules as citizens before Welfare Reform while each state developed its own rules after the reform. A difference in noncitizen eligibility rules between the two periods in each state is to be caught in the interaction term between state generosity and reform indicator.

This study used a continuous measure of labor market condition indicator: annual unemployment rate from the Bureau of Labor Statistics. Previous studies employed the same variable in assessing the effects of local labor market conditions (Haider et al., 2004; Lofstrom & Bean, 2002).

This study included demographic, socioeconomic, and household characteristics as control variables. The race and Hispanic-origin variable had five categories: non-Hispanic White (White hereafter), non-Hispanic Black (Black hereafter), Hispanic, non-Hispanic Asian (Asian hereafter), and other non-Hispanic group (other hereafter). The age variable consisted of four groups (65 to 69, 70 to 74, 75 to 79, and 80 or older). This study used a categorical variable of age instead of a continuous one to maintain consistency across survey years: CPS set upper limits for age variable inconsistently during the observation period of this study (80, 85, or 90, depending on survey years). Gender, disability status (e.g., having a health problem or a disability that prevents or limits work), and marital status were dichotomous variables (male or female, disabled or not, married or not). The education variable had three categories: no diploma, high school diploma, and college education. Per capita household-income variable was generated by dividing total household income by the number of household members, then inflation-adjusted 2008 dollar value using the Consumer Price Index. Household size was a categorical variable consisting of three groups (1, 2, and 3 or more people). The homeownership variable consisted of two categories (homeowners and non-homeowners). The asset income variable indicated whether an older adult’s household had income from assets (e.g., interests and dividends).

Analytical Strategy: Triple Difference-in-Differences Approach

This study used a difference-in-differences approach, which is an analytical strategy frequently used in policy impact assessment (Bertrand, Duflo, & Mullainathan, 2004; Blank, 2002). This approach addresses the issue of unobserved variable bias by comparing changes in outcomes after policy adoption between the target and nontarget populations. Since the nontarget population (native citizens in this study) is not expected to be affected by policy shifts (e.g., noncitizen eligibility changes), any changes observed in this group would likely arise from other unobserved changes that may have taken place concurrently with the adoption of the policy of interest (e.g., increased premium for private health insurance). Accordingly, an observed gap between the target and nontarget populations in terms of changes in an outcome measure indicates the impact of the policy shift of interest (Borjas, 2003; Khandker, Koolwal, & Samad, 2010; Nam & Jung, 2008).

This study employed a triple difference-in-differences approach that incorporates three independent variables: citizenship, time (Welfare Reform indicator), and state generosity. Analytical strategy is summarized in Equation (1):

Y_{ist} = α + β_{1} * C_{ist} + β_{2} * G_{is} + β_{3} * (C_{ist} * G_{is}) + β_{4} * R_{it} + β_{5} * (G_{is} * R_{it}) + β_{6} * (C_{ist} * R_{it}) + β_{7} * (C_{ist} * G_{is} * R_{it}) + β_{8} * U_{ist} + β_{9} * (C_{ist} * U_{ist}) + β_{10} * X_{ist} + ε_{ist}

where Y_ist is Medicaid or health insurance coverage for an older adult i in state s at time t, C_ist is a vector of citizenship status dummies of an older adult i (naturalized citizens and noncitizens in comparison with native citizens [reference group]), G_is denotes the generosity of state s in which an older adult i lives, R_it is the Welfare Reform indicator, U_ist indicates the unemployment rate of state s where an older adult i lives at time t, X_ist denotes a vector of control variables (listed in the Measures section and reported in Table 1), and e_ist represents a random error.

Table 1.

Sample Characteristics by Citizenship Status and by Time (Weighted)

	Native citizens		Naturalized citizens		Noncitizens
Characteristics	Prereform	Postreform	Prereform	Postreform	Prereform	Postreform
Percentage in sample	91.44	88.92**	5.32	7.84**	3.23	3.25
Race and Hispanic origin (%)
White	88.34	85.99**	70.38	48.96**	34.61	32.20**
Black	8.47	8.69	2.15	4.51	3.87	6.21
Hispanic	2.57	3.58	17.33	25.14	42.16	40.06
Asian	0.57	0.95	9.91	21.30	18.89	21.51
Other	0.05	0.78	0.23	0.08	0.47	0.03
Age (%)
65-69	31.26	28.16**	26.14	30.97**	33.46	35.08**
70-74	27.47	23.97	26.59	23.91	27.44	27.62
75-79	19.99	21.63	16.27	20.22	15.02	18.43
80+	21.29	26.24	31.01	24.91	24.08	18.87
Female (%)	58.29	57.21**	59.59	57.89	61.58	59.14
Education (%)
Less than high school	35.39	24.01**	38.84	35.52**	62.00	54.57**
High school diploma	34.87	37.70	30.41	28.76	20.78	22.34
College	29.74	38.30	30.75	35.72	17.22	23.09
Married (%)	56.06	55.94	53.41	58.48**	47.66	52.35*
Disabled (%)	28.32	23.21**	22.74	16.98**	21.64	18.38*
Household size (%)
1 person	31.18	31.69**	31.40	24.32**	17.09	16.12
2 people	54.63	55.05	48.09	47.55	36.56	33.83
3 or more	14.18	13.26	20.51	28.13	46.35	50.05
Household income (US$)	22,707	25,920 **	21,192	24,064**	16,228	18,514**
Homeowner (%)	82.27	84.38**	75.07	75.00	54.25	61.74**
Asset income (%)	72.89	60.91**	71.32	52.20**	50.69	38.77**
Unemployment rate	6.34	5.34**	6.98	5.41**	7.36	5.51**
Generous state (%)	27.72	26.06**	48.13	49.41	55.11	48.19**
N (unweighted)	32,050	62,345	2,002	5,703	1,260	2,513

Data source: Current Population Survey (CPS), 1994 to 1996 and 2003 to 2008.

Note. N = 105,873. For statistical tests, this study uses the Pearson χ² statistic for categorical variables (e.g., race and Hispanic origin, education) and the Wald test for continuous variables (e.g., income). Statistical test results on differences between pre- and post-reform periods are indicated as p values below.

p < .05. **p < .01.

In Equation 1, the two coefficients of citizenship variables (b₁) assess differences in Medicaid (or health insurance) coverage between the target populations (naturalized citizens and noncitizens) and the nontarget population (native citizens, reference group) in nongenerous states prior to Welfare Reform. The coefficient of state generosity (b₂) estimates differences in an outcome measure among native citizens between generous and nongenerous states before reform. The coefficients of interaction terms between citizenship status and generosity (b₃) indicate whether gaps in Medicaid coverage between noncitizens (or naturalized citizens) and native citizens (reference group) differ between generous and nongenerous states before reform.

The parameters of interest are b₆ and b₇. These two sets of coefficients test the “chilling effects,” “protective citizenship,” and “labor market condition” hypotheses. The coefficients of interaction terms between citizenship status and the reform indicator (b₆) estimate the impact of federal eligibility by showing changes in Medicaid coverage after Welfare Reform among noncitizens (and naturalized citizens) in comparison to native citizens. If the coefficient between noncitizen and reform indicator is significantly negative, we can conclude that older noncitizen’s Medicaid coverage declined after reform significantly more sharply than their native counterparts, suggesting that the federal Medicaid eligibility restriction had a negative impact on older noncitizens. Both the “chilling effect” and “protective citizenship” hypotheses predict the interaction term between the reform indicator and noncitizen to be significantly negative. The “labor market” condition hypothesis, however, predicts the interaction term not to be statistically significant after controlling for unemployment rate and its interaction term with noncitizen. At the same time, the “chilling effect” and “protective citizenship” hypotheses predict different results on the interaction term between the reform indicator and naturalized citizen. The “chilling effect” hypothesis predicts the interaction term coefficient to be negative since the federal eligibility restriction is expected to have a negative impact even on eligible immigrants (naturalized citizens). The “protective citizenship” hypothesis, however, predicts the opposite result since Welfare Reform may motivate naturalization among immigrants as a way to ensure access to public benefits; therefore, the percentage of Medicaid recipients is hypothesized to rise among naturalized citizens. As in the noncitizen case, the “labor market condition” hypothesis predicts little impact of Welfare Reform on naturalized citizens.

The coefficients of triple interaction terms (b₇) show the effects of state generosity by indicating differences between generous and nongenerous states in the magnitudes of changes in Medicaid coverage among older noncitizens (or older naturalized citizens) compared to older native citizens. Both the chilling effect and protective citizenship hypotheses predict the coefficient of the triple interaction term among noncitizen, Welfare Reform indicator, and state generosity to be positive, since nongenerous states impose barriers to Medicaid on noncitizens whereas there are no such obstacles in generous states. The two hypotheses’ predictions on the triple interaction term among naturalized citizen and the other two variables, however, are not identical. The “chilling effect” hypothesis predicts the triple interaction term with the naturalized citizen indicator to be positive since the impact of the eligibility rule change on naturalized citizens is supposed to be identical to that on noncitizens. It is not clear, however, what the predicted outcome of the “protective citizenship” hypothesis will be. On the one hand, older immigrants are faced with greater incentives for naturalization in nongenerous states where access is guaranteed only through citizenship. Under this scenario, the coefficient of the triple interaction term is expected to be negative, suggesting a lower rate of naturalization (and a lower rate of Medicaid coverage among naturalized citizens) in generous states. On the other hand, naturalization may be deemed easier in generous states since these states have more incentives to facilitate naturalization in order to reduce state spending. For this reason, these states are more likely to provide services for naturalization (e.g., free English classes for those preparing for citizenship examination). It is also plausible that stronger immigrant rights movements in generous states were successful in mobilizing older immigrants for naturalization and public assistance program participation (Fujiwara, 2005). If this is the case, the triple interaction term with the naturalized citizen indicator is expected to have a positive coefficient. The “labor market condition” predicts that the coefficients of the triple interaction terms are expected not to be statistically significant when controlling for the state unemployment rate and its interaction terms with citizenship status.

In running the triple difference-in-differences analyses, this study used logit regressions because the dependent variables (Medicaid and health insurance coverage) were dichotomous. This study clustered standard errors at the household level in consideration that some households include multiple older adults (Greene, 2003). This study used weighted data for both descriptive and multivariate analyses, using a weight variable generated by the CPS in consideration of its sampling design and nonresponse bias (U.S. Census Bureau, 2008).

In addition to the main analyses reported in this article, this study ran supplementary analyses to check the robustness of the findings. First, the study ran a regression with two additional variables on state characteristics: the percentage of foreign-born people in the state of residence in 1990 and the rate of change of this percentage between 1990 and 2000. These two state characteristic variables were created using the 1990 and 2000 census data. Second, it ran a supplemental analysis with a sample that excluded potential refugees who entered the United States after Welfare Reform. Medicaid eligibility for refugees differs from that of other immigrants in that refugees remain eligible for the first 7 years in the United States (Nam, 2011a; Zimmermann & Tumlin, 1999). This study identified potential refuges as those from refugee-sending countries based on Passel and Clark (1998). Third, the study ran an additional regression with a sample containing no Massachusetts residents. Fourth, regressions were also run with a sample excluding California residents. Borjas (2002) shows that analysis results differ when California residents are excluded from the sample in estimating the effects of noncitizen eligibility restrictions on immigrants’ public assistance program participation. He suggests that it probably happened because California is a large state with a high percentage of immigrants in state population. Fifth, the study conducted analyses after excluding observations collected after 2006 because Medicaid started requiring applicants to submit legal documentation of their immigration status beginning in 2006. This procedural change may have affected older immigrant adults’ Medicaid coverage (Kaiser Commission of Medicaid and the Uninsured, 2006). Last, this study ran regressions without the household size variable. Since the main analysis model uses the income variable that takes into account household size (per capita household income), it may overcontrol household size.

Results from these supplementary analyses do not differ substantively from those reported here, except for minor differences in the analyses using the sample without California residents. Although the coefficients of all major variables (interaction terms between citizenship and reform indicator and triple-interaction terms) have the same sign as those in the main analysis model, two interaction terms loses statistical significance at the .05 level: the interaction term between naturalized citizen and reform indicator in Medicaid regression and the triple-interaction term among noncitizen, reform indicator, and state generosity in health insurance regression (p value of these coefficients are .28 and .27, respectively). Different results between the main and the supplementary analyses may be explained by the fact that the latter has a smaller sample than the former since California residents consist of 10% of the sample (27% of immigrants and 9% of native citizens). (Results from supplementary analyses are available from the author.)

Results

Table 1 shows older adults’ characteristics by Welfare Reform indicator and by citizenship status. As expected, older noncitizens’ socioeconomic conditions were worse than the other two groups during both pre- and post-reform periods, as indicated with lower educational attainment, income, and homeownership rates. In the older adult sample, the percentage of native citizens significantly decreased, the proportion of naturalized citizens significantly rose, and that of noncitizens remained stable between pre- and post-reform periods. The percentage of the youngest group (65 to 69 years old) decreased among native citizens but increased among the two immigrant groups. Homeownership rates significantly increased among native citizens and noncitizens but remained stable among naturalized citizens. These results suggest that older adults with distinct citizenship status experienced different patterns of changes in their demographic and socioeconomic characteristics.

Table 2 reports Medicaid and health insurance coverage of older adults by citizenship status, state generosity, and time (pre- vs. post-reform period). As indicated in Table 2, the patterns of change in Medicaid and health insurance coverage differ by citizenship status. Medicaid coverage remained stable among native citizens between the two periods, increased among naturalized citizens, but decreased among noncitizens in both generous and nongenerous states. Table 2 also shows the impact of state generosity: Naturalized citizens’ Medicaid coverage increased by only 4 percentage points in nongenerous states while the corresponding statistics in generous states rose by 13 percentage points; among noncitizens, a decline in Medicaid coverage was slightly higher in nongenerous states than in generous states (9 percentage points vs. 7 percentage points). At the same time, health insurance coverage rates indicate that noncitizens in nongenerous states were hit hardest after Welfare Reform: Their health insurance coverage declined by 7 percentage points and those of other groups remained stable or only slightly declined.

Table 2.

Health Insurance Coverage Rates by Citizenship Status and by Time

		Nongenerous states		Generous states
		Prereform	Postreform	Prereform	Postreform
Medicaid	Natives	7.46	7.54	8.23	8.12
	Naturalized citizens	8.30	11.81**	13.66	26.23**
	Noncitizens	30.23	21.19**	40.33	33.06**
Health insurance	Natives	99.33	99.00**	99.30	98.74**
	Naturalized citizens	98.72	97.37*	98.03	97.72
	Noncitizens	89.56	82.31**	88.63	87.40

Note. N = 105,873. This study uses the Pearson χ² statistic to test differences between pre- and post-reform periods. Statistical test results on differences between pre- and post-reform periods are indicated as p values below.

p < .05. **p < .01.

Table 3 presents the results of the triple difference-in-differences analyses on Medicaid and health insurance coverage. It reports results only on major variables: citizenship status, Welfare Reform indicator, and state generosity; interaction terms and triple interaction terms among these three variables; and state unemployment rate and its interaction terms with citizenship status. Results on other variables are as expected. Blacks, Hispanics, and Asians are more likely to have Medicaid and less likely to have health insurance than Whites. Higher income and homeownership are negatively associated with Medicaid coverage but positively associated with health insurance coverage. (Full results are available from the author.)

Table 3.

Logit Regression Results on Medicaid and Health Insurance Coverage

	Medicaid	Health insurance
Naturalized citizen (NatC)	−0.795 (0.311)*	−0.900 (0.653)
Noncitizen (NonC)	0.828 (0.360)*	−2.131 (0.529)**
Generous state (GenS)	0.197 (0.059)**	−0.013 (0.178)
NatC*GenS	0.032 (0.191)	−0.245 (0.464)
NonC*GenS	0.046 (0.190)	0.021 (0.321)
Welfare Reform (Ref)	0.110 (0.042)**	−0.366 (0.125)**
Ref*GenS	−0.094 (0.074)	−0.173 (0.207)
Ref* NatC	0.333 (0.165)*	0.116 (0.397)
Ref*NonC	−0.550 (0.170)**	−0.194 (0.286)
RefNatCGenS	0.661 (0.216)**	0.764 (0.506)
RefNonC GenS	0.304 (0.226)	0.769 (0.368)*
State unemployment (Unemp)	0.012 (0.014)	−0.022 (0.038)
NatC*Unemp	0.129 (0.042)**	0.071 (0.083)
Nonc*Unemp	0.031 (0.051)	0.061 (0.070)
Pseudo R²	0.183	0.193

Note. N = 105,873. Logit regressions control for demographic, socioeconomic, and household characteristics as described in the Measures section.

p < .05. **p < .01.

The first column in Table 3 reports the results on Medicaid coverage. The significantly negative coefficient of the naturalized citizen indicator shows that this group is less likely to have had Medicaid coverage than older native citizens (reference group) with comparable characteristics prior to Welfare Reform. At the same time, the positive coefficient of the noncitizen indicator suggests that older noncitizens are more likely to have had Medicaid coverage than their native counterparts prior to Welfare Reform, after controlling for demographic and socioeconomic factors. The Welfare Reform indicator has a significantly positive coefficient, implying that Medicaid coverage for native citizens increased after Welfare Reform.

As described in the Method section, the interaction terms between reform indicator and citizenship status (b₆) estimate the impact of federal eligibility restrictions. The interaction term between reform indicator and naturalized citizens has a significantly positive coefficient, suggesting that their Medicaid coverage significantly increased after Welfare Reform in comparison with native citizens. This finding is consistent with the prediction of the “protective citizenship” hypothesis. The interaction term between reform indicator and noncitizen is significantly negative as anticipated: Noncitizens’ Medicaid coverage significantly decreased after Welfare Reform.

The triple interaction terms (b₇) assess the effects of state generosity. The triple interaction term with the naturalized citizen indicator has a significantly positive coefficient, indicating that an increase in Medicaid coverage is significantly higher in generous states than in nongenerous states. The coefficient of the triple interaction term with the noncitizen indicator is positive but not statistically significant, suggesting that generous state eligibility does not have a significant impact on noncitizens.

It is noticeable that the interaction terms between reform indicator and citizenship status and triple interactions have statistically significant coefficients despite the fact that this study controlled for the state unemployment rate and its interaction with citizenship status. These findings suggest that the federal-level noncitizen eligibility restriction after Welfare Reform and state variation in Medicaid eligibility have significant effects on older immigrants’ Medicaid coverage even when labor market conditions are considered, contrary to the predictions of the “labor market condition” hypothesis.

The second column in Table 3 provides analysis results on health insurance coverage. The coefficient of reform indicator is significantly negative, suggesting that health insurance coverage declined among older native citizens after Welfare Reform. None of the interaction terms is statistically significant, suggesting that the federal eligibility shift has no impact on health insurance coverage either among naturalized citizens or among noncitizens. The triple interaction term with the naturalized citizen indicator is not statistically significant. The coefficient of triple interaction term with the noncitizen indicator is large and significantly positive, suggesting that the generous state eligibility rule has a positive effect on noncitizens’ health insurance coverage.

Predicted Probabilities of Medicaid and Health Insurance

Figure 1 presents predicted probabilities of Medicaid and health insurance coverage by citizenship status, Welfare Reform indicator, and state generosity. Predicted probabilities are estimated based on regression results in Table 3 for a typical low-income older adult in the sample: a White married woman without disability who is aged between 65 and 69 years, who does not have a high school diploma and lives in a household consisting of two members with an income of US$5,000 per person, whose household owns neither a house nor income-generating assets, and who lives in a state with an unemployment rate of 6%. Predicted probabilities show the effects of citizenship, the federal Medicaid eligibility restriction, and the generous state eligibility on older adults more clearly than do the descriptive statistics in Table 2 because these results are estimated while holding other factors constant.

Figure 1.

Predicted probabilities of Medicaid health insurance coverage

Figure 1 clearly shows the roles of citizenship status on older immigrant adults’ Medicaid coverage. Although Medicaid coverage shows little change among older native citizens in both generous and nongenerous states, it increased among older naturalized citizens and decreased among older noncitizens. Figure 1 also indicates differences by state generosity: Older naturalized citizens’ Medicaid coverage increased by 6 percentage points in nongenerous states while it rose by 19 percentage points in generous states; older noncitizens’ Medicaid coverage declined by 9 percentage points and by 5 percentage points, respectively.

Results on the predicted probabilities of health insurance coverage also show the importance of state generosity among older immigrant adults. Although the patterns of change among older native citizens do not differ by state generosity, older naturalized citizen’s health insurance coverage declined by 1 percentage point in nongenerous states and increased by the same amount in generous states. At the same time, noncitizens’ coverage decreased by 5 percentage points in nongenerous states but remained stable in generous states.

Discussion

This study examined changes in Medicaid and health insurance coverage among older adults after Welfare Reform imposed restrictive noncitizen eligibility restrictions. Using newly collected state Medicaid eligibility data, this study investigated the impact of state-level Medicaid eligibility on older immigrant adults as well as that of federal-level policy change. In addition, this study tested the competing “chilling effect” and “protective citizenship” hypotheses by looking at the effects of policy changes by citizenship status. This study also included the unemployment rate and its interaction with citizenship status in the analyses to test the “labor market condition” hypothesis.

Analysis results of the older adult sample from the CPS data clearly show the impacts of Welfare Reform on older immigrant adults. The Medicaid coverage rate significantly declined among older noncitizens after Welfare Reform, indicating that the federal eligibility restrictions constrained this group’s access to Medicaid. Older naturalized citizens’ Medicaid coverage, however, increased after Welfare Reform, showing opposite patterns of changes between the two older immigrant groups with different citizenship status. These findings support the “protective citizenship” hypothesis, which predicts that the noncitizen eligibility restriction would have motivated naturalization among immigrants in need, and therefore, raise naturalized citizens’ Medicaid coverage rate. Analysis results are inconsistent with what were expected by the “chilling effect” and the “labor market condition” hypotheses.

This study also shows the roles of generous state eligibility rules in protecting older immigrants’ access to Medicaid and health insurance. Analysis results show that older naturalized citizens’ Medicaid coverage increased in generous states at a significantly higher rate than in nongenerous states. This finding suggests that an increase in older immigrant adults’ naturalization rates were higher in generous states than in nongenerous states. Descriptive statistics using the CPS data confirm this: Naturalization rates rose from 59% to 71% in generous states and from 66% to 70% in nongenerous states.

It is also of interest that generous state eligibility rules protect older noncitizens’ health insurance coverage from the negative impacts of the federal eligibility restriction. Predicted probability estimation shows that the chances for noncitizens’ obtaining health insurance are stable in generous states, whereas they declined by 5 percentage points in nongenerous states. These findings contrast with Borjas’s (2003) finding that state eligibility rules had no significant impact on younger noncitizens’ health insurance coverage. Differences, however, may be explained by older immigrant adults’ unique conditions: They are more likely to have health problems due to aging and have a stronger desire to seek health insurance coverage than younger immigrants. Further study is warranted to enhance our understanding on different policy impacts by age, and divergent findings between this study and Borjas justify separate analyses of immigrants at different life stages.

This study has the following limitations. First, this study is unable to rule out the possibility that the changes in Medicaid and health insurance coverage observed in this study may have been caused, in whole or in part, by other immigration policies or socioeconomic changes that occurred concurrently with Welfare Reform. For example, the Illegal Immigration Reform and Immigrant Responsibility Act of 1996 made it more difficult for noncitizens to sponsor immigration of their relatives (Balistreri & Van Hook, 2004). This policy change may have influenced older immigrant adults’ decision to naturalize and to participate in public assistance programs. Second, the CPS does not collect data on each immigrant’s legal status and does not allow for differentiation of legal permanent residents from undocumented immigrants. Considering that these immigrant groups’ positions in the health care and welfare systems and labor market are fundamentally different from each other, the inability to separate them out from each other may have biased the analysis results of this study. Third, this study is unable to investigate whether and how changing racial and ethnic compositions of immigrants affected their Medicaid and health insurance coverage. The majority of immigrants now come from Latin America and Asia, whereas most immigrants came from European countries in the past (Congressional Budget Office, 2004). Considering that poverty rates among older immigrant adults from Latin America and Asia are much higher than those from Europe (Gerst & Burr, 2011), immigrants’ compositional changes may have affected their public assistance program participation and health insurance coverage. Further research is called for to ascertain the impacts of noncitizen eligibility on older immigrant adults from different regions and countries. Last, the state unemployment rate variable in this study may not have captured local labor market conditions fully. Although it is an economic indicator used in existing studies on the topic (Haider et al., 2004; Lofstrom & Bean, 2002), unemployment rates in a smaller geographic area (e.g., county or metropolitan area) may reflect local labor market conditions more accurately than the state level. Economic indicators other than unemployment (e.g., median wage of low-educated workers) may also enhance our understanding of labor market conditions.

The findings of this study have the following implications for future research and policy development. In terms of research implications, this study demonstrates the need for special attention to older immigrant adults. As noted previously, the results of this study are different from those found in Borjas (2003) on younger immigrants: In this study, the variations in state eligibility rules had significant effects on older noncitizens’ health insurance coverage, whereas Borjas’ study showed little impact. Since older immigrants’ position in the health care and welfare systems differs from that of younger immigrants, it is not surprising that the same policy yielded distinct outcomes for different subsets of the immigrant population at different life stages.

As for policy implications, findings in this study cast doubt on the effectiveness of noncitizen eligibility restrictions in reducing government spending on public assistance programs. The opposite patterns of changes in Medicaid coverage between older naturalized citizens and noncitizens suggest that older immigrant adults may have acquired citizenship to ensure their access to Medicaid. That is to say, cost savings that may have been realized from decreased Medicaid coverage among older noncitizens are likely offset by increased coverage for older naturalized citizens. Furthermore, restricting older immigrants’ Medicaid coverage likely raises long-term health care costs even if it were to succeed in excluding immigrants from Medicaid coverage in the short term. Evidence shows that a lack of health insurance limits access to preventive and needed medical care (Ayanian et al., 2000), which causes high long-term costs to society by increasing the use of costly emergency room care (Scarpaci & Kearns, 1997). Accordingly, it is not surprising that emergency Medicaid expenditures for older immigrant adults ineligible for regular Medicaid program have rapidly grown in recent years (DuBard & Massing, 2007). Given this study’s findings and other empirical evidence on the negative consequences of limited access to medical care, policy makers should reconsider the current policy of restricting Medicaid eligibility of noncitizens.

Footnotes

Acknowledgements

The author is grateful to Hyo Jin Jung, Charles Jarrett, Susanna Johnson, Carissa Clark, and Amanda Brower for their wonderful research assistance and to Karen Dodson for editorial help.

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

The author received the following financial support for the research, authorship, and/or publication of this article: This study was supported in part by grants from the John A. Hartford Foundation’s Geriatric Social Work Faculty Scholars program, the Louis and Samuel Silberman Fund faculty grant program, and the Center for Aging at Washington University in St. Louis.

References

Auchincloss

A. H.

van Nostrand

J. F.

Ronsaville

(2001). Access to health care for older persons in the United States: Personal, structural, and neighborhood characteristics. Journal of Aging and Health, 13, 329-354.

Ayanian

J. Z.

Weissman

J. S.

Schneider

E. C.

Ginsburg

J. A.

Zaslavsky

A. M.

(2000). Unmet health needs of uninsured adults in the United States. JAMA-Journal of the American Medical Association, 284, 2061-2069.

Balistreri

K. S.

Van Hook

(2004). The more things change the more they stay the same: Mexican naturalization before and after Welfare Reform. International Migration Review, 38, 113-130.

Bansak

Raphael

(2006). The effects of state policy design features on take-up and crowd-out rates for the state children’s health insurance program. Journal of Policy Analysis and Management, 26, 149-175.

Berk

M. L.

Schur

C. L.

(1998). Access to care: How much difference does Medicaid make? Health Affair, 17, 169-180.

Bertrand

Duflo

Mullainathan

(2004). How much should we trust differences-in-differences estimates? Quarterly Journal of Economics, 119, 249-275.

Blank

R. M.

(2002). Evaluating Welfare Reform in the United States. Journal of Economic Literature, 40, 1105-1166.

Borjas

G. J.

(2002). Welfare Reform and immigrant participation in welfare programs. International Migration Review, 36, 1093-1123.

Borjas

G. J.

(2003). Welfare Reform, labor supply, and health insurance in the immigrant population. Journal of Health Economics, 22, 933-958.

10.

Congressional Budget Office. (1996). Federal bedgetary implications of the Personal Responsibility and Work Opportunity Act Reconciliation Act of 1996 CBO Memorandum. Washington, DC: Author.

11.

Congressional Budget Office. (2004). A description of the immigrant population. Washington, DC: Author.

12.

DuBard

C. A.

Massing

M. W.

(2007). Trends in emergency Medicaid expenditures for recent and undocumented immigrants. JAMA-Journal of the American Medical Association, 297, 1085-1092.

13.

Fix

M. E.

Capps

Kausbal

(2009). Immigrants and welfare: Overview. In Fix

M. E.

(Ed.), Immigrants and welfare (pp. 1-36). New York, NY: Russell Sage.

14.

Fix

M. E.

Passel

J. S.

(1999). Trends in noncitizens’ and citizens’ use of public benefits following Welfare Reform. Washington, DC: The Urban Institute.

15.

Fremstad

Cox

(2004). Covering new Americans: A review of federal and state policies related to immigrants’ eligibility and access to publicly funded health insurance. Washington, DC: Kaiser Family Foundation.

16.

Fujiwara

L. H.

(2005). Immigrant rights are human rights: The reframing of immigrant entitlement and welfare. Social Problems, 52, 79-101.

17.

Gerst

Burr

J. A.

(2011). Welfare use among older Latin American immigrants: The effects of federal and state policy. Population Research and Policy Review, 30, 129-150.

18.

Gilbertson

Singer

(2003). The emergence of protective citizenship in the USA: Naturalization among Dominican immigrants in the post-1996 Welfare Reform era. Ethic and Racial Studies, 26, 25-51.

19.

Greene

W. H.

(2003). Econometric analysis (5th ed.). Upper Saddle River, NJ: Prentice Hall.

20.

Haider

S. J.

Schoeni

R. F.

Bao

Danielson

(2004). Immigrants, Welfare Reform, and the economy. Journal of Policy Analysis and Management, 23, 745-764.

21.

Kaiser Commission of Medicaid and the Uninsured. (2006). New requirements for citizenship documentation in Medicaid. Washington, DC: Author.

22.

Kandula

N. R.

Grogan

C. M.

Rathouz

P. J.

Lauderdale

D. S.

(2004). The unintended impact of Welfare Reform on the Medicaid enrollment of eligible immigrants. Health Services Research, 39, 1509-1526.

23.

Kaushal

Kaestner

(2005). Welfare Reform and health insurance of immigrants. Health Services Research, 40, 697-721.

24.

Khandker

S. R.

Koolwal

G. B.

Samad

H. A.

(2010). Handbook on impact evaluation: Quantitative methods and practices. Washington, DC: The World Bank.

25.

(2009a). Changes in immigrants’ use of Medicaid and Food Stamps: The role of eligibility and other factors. In Fix

M. E.

(Ed.), Immigrants and welfare (pp. 153-192). New York, NY: Russel Sage.

26.

(2009b). Health insurance coverage and medical expenditures of immigrants and native-born citizens in the United States. American Jounral of Public Health, 99, 1322-1328.

27.

Matani

(2001). Left out: Immigrants’ access to health care and insurance. Health Affair, 20, 247-256.

28.

Lofstrom

Bean

F. D.

(2002). Assessing immigrant policy options: Labor market conditions and postreform declines in immigrants’ receipt of welfare. Demography, 39, 617-637.

29.

Mold

J. W.

Fryer

G. E.

Thomas

C. H.

(2004). Who are the uninsured elderly in the United States? Journal of the American Geriatrics Society, 52, 601-606.

30.

Nam

(2008). Welfare Reform and older immigrants’ Medicaid participation and health insurance coverage. American Journal of Public Health, 98, 2029-2034.

31.

Nam

(2011a). Welfare Reform and immigrants: Non-citizen eligibility restrictions, vulnerable immigrants, and the social service providers. Journal of Immigrant & Refugee Studies, 9, 5-19.

32.

Nam

(2011b). Welfare Reform and elderly immigrants’ health insurance coverage: The roles of federal and state Medicaid eligibility rules. Journal of Gerontological Social Work, 54, 819-836.

33.

Nam

Jung

H. J.

(2008). Welfare Reform and older immigrants: Food Stamp program participation and food insecurity. The Gerontologist, 48, 42-50.

34.

Nam

Kim

(in press). Welfare Reform and elderly immigrants’ naturalization: Access to public benefits as an incentive for naturalization. International Migration Review.

35.

National Immigration Law Center. (2002). Guide to immigrant eligibility for federal programs (4th ed.). Los Angeles, CA: Author.

36.

Nelson

C. T.

Mills

R. J.

(2001). The March CPS health insurance verification and its effect on estimates of the uninsured. Washington, DC: U.S. Census Bureau.

37.

Passel

J. S.

Clark

R. L.

(1998). Immigrants in New York: Their legal status, income, and taxes. Washington, DC: Urban Institute.

38.

Powers

E. T.

(1998). Does means-testing welfare discourage saving? Evidence from a change in AFDC policy in the United States. Journal of Public Economics, 68, 33-53.

39.

Scarpaci

J. L.

Kearns

R. A.

(1997). Locational attributes of health behavior. In Gochman

D. S.

(Ed.), Handbook of health behavior research II: Provider determinants (pp. 81-102). New York, NY: Plenum.

40.

Schmidley

A. D.

Robinson

J. G.

(2003). Measuring the foreign-born population in the United States with the current population survey (1994-2002 Population Division Working Paper, pp. 1-22). Washington, DC: U.S. Department of Commerce.

41.

State Health Access Data Assistance Center. (2006). Historical changes in the CPS health insurance coverage items, 1989-2005 survey years. Washington, DC: Author.

42.

U.S. Census Bureau. (2008). Current population survey (2008 ASEC Technical Documentation Vol. 2008). Retrieved from http://www.census/gov/apsd/techdoc/cps/cpsmar08.pdf

43.

Van Hook

(2003). Welfare Reform’s chilling effects on noncitizens: Changes in noncitizen welfare recipiency or shifts in citizenship status? Social Science Quarterly, 84, 613-631.

44.

Zimmermann

Tumlin

K. C.

(1999). Patchwork policies: State assistance for immigrants under Welfare Reform. Washington, DC: Urban Institute.