Trends in Noncitizens' and Citizens' Use of Public Benefits Following Welfare Reform: 1994-97

Background

With the enactment of the 1996 welfare reform act,⁽¹⁾ Congress imposed broad new restrictions on legal immigrants' access to public benefits, set new time limits on refugees' eligibility for many federal benefits, and introduced new bars on the access of "unqualified immigrants" to services.⁽²⁾ But perhaps more important than these changes in eligibility are welfare reform's chilling effects which may discourage immigrants from using health, nutrition, or other types of benefits, despite the fact that many remain eligible. These effects originate, among other things, in confusion on the part of immigrants and providers about who is eligible for benefits and in fears relating to the application of the public charge doctrine.⁽³⁾

An earlier study by the Urban Institute found evidence of such chilling effects in Los Angeles County.⁽⁴⁾ In that study, approved applications of legal noncitizen families for Medi-Cal and Temporary Assistance for Needy Families (TANF) fell 71 percent between January 1996 and January 1998, while there was no decline among citizens. The drop occurred even though there was no change in legal immigrants' eligibility for these programs in California and denial rates in the county remained steady during the period examined.

In this brief report we use the Census Bureau's Current Population Survey (CPS) to document national trends in immigrants' use of public benefits in the period following welfare reform. Specifically, we examine changes in participation between 1994 and 1997 reflected by the March CPS.⁽⁵⁾ During 1994 changes in welfare rules were just beginning to be broadly debated. By the end of 1997, welfare reform had been in place for a year and a half, although full implementation was not complete. In addition, the CPS for both years provides comparable data on benefit use for the entire nation.

The current analysis builds on methods developed by the Urban Institute over the past decade that permit us to distinguish refugees, naturalized citizens, and temporary immigrants from other legally present immigrants.⁽⁶⁾ Such distinctions are important for two reasons. First, conventional comparisons between the benefit use rates of natives and the foreign born mask substantial variation in rates and trends among substantively different segments of the foreign-born population. Second, following welfare reform, citizenship status has become an increasingly important determinant of eligibility for public benefits.

We should emphasize that most legal immigrants and refugees remained eligible for welfare and Medicaid benefits throughout the period examined (1994 through 1997).⁽⁷⁾ The same cannot be said for federal food stamps, however: many legal immigrants' eligibility was supposed to end as of September 1997, while new noncitizen applicants became ineligible in October 1996. Finally, while most immigrants arriving after welfare reform's enactment are barred from federal means-tested public benefits for at least five years,⁽⁸⁾ these "future" immigrants represented a small share of the noncitizen population at the time of the March 1998 CPS.

Principal Findings

• When viewed against the backdrop of overall declines in welfare receipt for all households, use of public benefits among noncitizen households ⁽⁹⁾ fell more sharply (35 percent) between 1994 and 1997 than among citizen households (14 percent). These patterns hold for welfare (defined here as TANF, SSI, and General Assistance), food stamps, and Medicaid.
• Refugees experienced declines (33 percent) that were at least as steep as those within the noncitizen population  despite the protections for refugees incorporated into welfare reform and the fact that few refugees had lost their eligibility for benefits by March 1998.⁽¹⁰⁾
• For low-income populations (i.e., with incomes below 200 percent of poverty), program usage also fell faster for noncitizen than citizen households.
• Welfare use in noncitizen households with children also fell faster (36 percent) than in households with children where all adults are citizens (23 percent).
• One result of these trends is that noncitizens accounted for a disproportionately large share of the overall decline in welfare caseloads that occurred between 1994 and 1997. While 23 percent of the drop in welfare caseloads can be ascribed to noncitizens, they represented only 9 percent of households receiving welfare in 1994.
• Welfare use among elderly immigrants and naturalized citizens did not appear to change between 1994 and 1997.
• When welfare use among all households is examined, noncitizen participation levels were higher than citizens' in both 1994 and 1997, despite rapid declines in noncitizen use rates. But when we look at poor households (i.e., with incomes under 200 percent of poverty), noncitizens' participation rates in 1994 were no different from those of citizens; by 1997, however, levels had declined so that noncitizens had lower participation rates than citizens (14.5 versus 17.9 percent). When we examine poor households with children, noncitizen rates were lower for both 1994 and 1997  falling to almost half of the level of citizens in 1997 (14.0 versus 25.8 percent).
• Neither naturalization nor rising incomes accounted for a significant share of noncitizens' exits from public benefit use.

In the following section we examine patterns of benefit use in three different ways. First, we examine benefit use by household, disaggregating by all households, by households with incomes below 200 percent of poverty, and by those containing children. We then present findings for individuals, distinguishing use patterns for working-age adults (age 18 to 64) and the elderly (age 65 and over). We conclude the section by disaggregating trends by legal status, most notably program participation by refugees and naturalized citizens. In each instance, differing units of observation reveal different relative levels and trends in benefit use by citizens and noncitizens.

The analysis examines the use of welfare, food stamps, and Medicaid. While complete results are set out in the figures and tables included in this report, selected outcomes are highlighted in the narrative below. Since trends in food stamp and Medicaid use generally parallel those for welfare, we usually report results for only welfare.

DETAILED FINDINGS

A. Household-Level Analyses

Relative benefit use rates for citizens and noncitizens differ greatly depending on whether we focus on all households, poor households, or households with children. The reasons for these differences are straightforward. Poor households are far more likely to be eligible for and use benefits, and noncitizens are more likely to be poor. Fifty-four percent of noncitizen households have incomes below 200 percent of poverty, compared with 31 percent of citizen households (Table 4).

While controls for poverty have occasionally been taken into account in discussions of immigrant welfare use,⁽¹¹⁾ differences between households with and without children have been less frequently invoked. Yet such differences are significant because households with children are considerably more likely to use benefits, and immigrant households are more likely to contain children. Of households headed by noncitizens of working age, 55 percent include children, compared with 35 percent of comparable citizen households. As we report below, when we separately control for poverty and the presence of children, differences in program use rates between citizens and noncitizens diminish and, in some instances, disappear altogether. When we control for both poverty and the presence of children, noncitizen use of benefits is consistently lower than that of citizens, both before and after welfare reform.

All Households. Welfare receipt by noncitizen households fell much faster (35 percent) than citizens' receipt (14 percent) between 1994 and 1997. However, despite these steeper declines, noncitizen use of welfare remained higher than citizens' in 1997  9.0 versus 6.7 percent (Table 1 and Figure 1). By 1997, far more immigrants had lost their eligibility for food stamps than welfare. But our data show that noncitizens' participation in each program declined at roughly the same rate, with a marginally faster decline in welfare than in food stamps (35 percent versus 30 percent).

Households below 200 Percent of Poverty. The picture of higher welfare use by noncitizens shifts significantly when we control for poverty. By 1997, noncitizens with incomes below 200 percent of poverty had use rates that were significantly lower than citizens' rates  14.5 versus 17.9 percent (Chart A). Here again, noncitizen participation rates dropped faster than citizens' between 1994 and 1997 (33 versus 10 percent). For both food stamps and Medicaid, noncitizens' use also fell faster than citizens' during the same period, so by 1997 relative participation levels within each program were effectively the same for both groups (Table 1 and Figure 1).

Households with Children. When we control for the presence of children, we also see declines to the point where noncitizen welfare usage rates in households with children are not significantly different from citizens'rates (8.9 versus 9.6 percent, in Table 3). Rates for both declined rapidly between 1994 and 1997  35 percent for noncitizens and 23 percent for citizens  but the difference in the rate of decline is not statistically significant (Table 3 and Figure 3).

In households with noncitizen adults, many of the children are citizens, in most cases because they were born in the United States. In fact, there is at least one citizen child in 85 percent of noncitizen households with children. These mixed-status households are of substantial demographic importance in the United States, as about one in 10 American children lives in a household where one or more of the parents is a noncitizen and one or more of the children is a citizen.⁽¹²⁾

Households with Children and with Incomes below 200 Percent of Poverty. Given the large share of noncitizen households that are poor and contain children, one approach to assessing relative benefits is to control for both poverty and the presence of children.When we do so, we see much lower use among noncitizen households, both before and after welfare reform. With the rapid declines that occurred for both groups, noncitizen welfare use in 1997 is about half the rate for citizens  14.0 versus 25.8 percent. (See Chart B, Table 3, and Figure 3.)

B. Individual-Level Analyses

Individual-level analyses of welfare use among citizens and noncitizens produce results that differ from results of aggregate household-level analyses because of patterns of welfare reporting in the CPS, differences in welfare use, and structural differences in the populations.⁽¹³⁾ "Welfare use" for an individual in the CPS is defined as having income from TANF, General Assistance, or SSI; a household "uses welfare" if anyone in the household has welfare income. One issue in reconciling individual and household use rates, as well as comparing survey with administrative data, is the fact that income data in the CPS is collected only for persons age 15 and over. Thus, if a child is receiving public assistance income, the income will either be ascribed to the parent or missed.

This income-based measure of welfare use means that most households report welfare participation as though there were only a single welfare recipient. For example, a single mother with two children receiving TANF income is only counted as one welfare unit or recipient, not three. The individual-level analysis differs from the household approach by ascribing use to only the reported welfare recipient, not to other household members who are not reported as receiving welfare. One reason it is important to examine individual level benefit use is that noncitizen households are significantly larger than citizen households.

Because there are almost twice as many adults (197 million, see detailed table B) as households (103 million, see Detailed Table A), welfare use rates for individuals should be lower than for households. In addition, noncitizen households are larger than citizen households and are more likely to contain children. Thus, we would expect larger differences in usage rates between households and individuals among noncitizens; the data support this.

Our analysis below focuses on two important subpopulations: working-age adults, age 18 to 64, and the elderly, age 65 and over.

Working-Age Adults. Working-age noncitizens' use of welfare fell roughly three times faster than citizens' between 1994 and 1997  41 versus 15 percent (Table 2). A similar pattern is evident for Medicaid.⁽¹⁴⁾ By 1997, there is no statistical difference between citizen and noncitizen participation rates for welfare (4.0 versus 3.3 percent) or Medicaid (6.7 versus 7.2 percent). See Table 2 and Figure 2.

Elderly Immigrants. In sharp contrast to most of the other components of the analysis reported here, we find no statistically significant decline in either welfare or Medicaid use on the part of elderly noncitizens. In fact, we find no significant change in welfare receipt among the elderly overall, regardless of citizenship status (Table 2 and Figure 2).

Elderly noncitizen use of welfare and other benefits is much higher than is the case for citizens. In 1997, only 3.7 percent of elderly citizens used welfare, compared to 19.0 percent of noncitizens (Table 2 and Figure 2). Higher use of welfare (primarily SSI) and Medicaid among elderly noncitizens can be attributed to the fact that many have not worked in the United States long enough to qualify for Social Security or Medicare. Moreover, the absence of a decline in usage since 1994 may be explained, at least in part, by the restoration of SSI benefits to pre-enactment immigrants.

One result that emerges from the analysis is an apparent rise between 1994 and 1997 in the number and share of naturalized elderly receiving welfare benefits  from 99,000 or 5.9 percent, to 167,000 or 9.0 percent (Detailed Table B). During the same period, there is a commensurate decline in the number of noncitizen elderly receiving benefits (from 213,000 to 163,000). One hypothesis is that the decline in noncitizen participation for this subpopulation may be attributable in part to naturalization. However, these numerical changes are not statistically significant, so the results cannot be treated as definitive.⁽¹⁵⁾

C. Benefit Use by Immigrant Status

Historically, immigration status has been a strong predictor of immigrant use of public benefits. This is partly because the foreign-born population consists of groups with varied eligibility for benefits. While naturalized citizens and refugees are eligible for benefits on the same terms as native-born citizens, legal permanent residents' use of benefits has been conditioned by deeming and public charge restrictions; temporary immigrants and the undocumented are largely barred from services. Further, the socioeconomic characteristics of the groups differ substantially, resulting in different needs.⁽¹⁶⁾ The importance of disaggregating the foreign-born population by status can be seen in the analysis of household use rates (Table 1).

Foreign-Born Population. Between 1994 and 1997, welfare use in households headed by all foreign-born persons fell by 21 percent or 2.5 percentage points (from 11.7 percent to 9.2 percent). However, this general trend masks very different levels and trends among the various immigrant groups.

Refugees. Refugees, who have historically had the highest levels of public benefit use among the foreign born, account for 8 percent of immigrant-headed households, but for 21 percent of immigrants' welfare use (Detailed Table A). Their use rate remained high  24.5 percent in 1997  but even this level represented a decline of 8.8 percentage points from the pre-reform level of 33 percent (Chart C).

Naturalized Citizens. Naturalized citizens, who have historically had the lowest levels of public benefit use among legally present immigrant populations, represent the other extreme. While naturalized citizens make up 41 percent of immigrant households, they account for only 31 percent of immigrants' welfare use. Their use of benefits was virtually identical to that of native citizens and did not change significantly between 1994 and 1997.

Noncitizens. The residual foreign-born subpopulation, noncitizens, had a large decline of 4.9 percentage points to 9.0 percent in 1997, but their use rate remained somewhat higher than that of citizens.

General Observations

Chilling Effects versus Eligibility Changes

• Because comparatively few legal immigrants were ineligible for public benefits as of December 1997, it appears that the steeper declines in noncitizens' than citizens' use of welfare, food stamps, and Medicaid owe more to the "chilling effect" of welfare reform and other policy changes than they do to actual eligibility changes. In addition, the fact that welfare use among noncitizens dropped as steeply as food stamp use (where new restrictions extended far more broadly) suggests that eligibility changes in one program may chill noncitizens' use of other programs. Over time, eligibility changes will become more important as most immigrants admitted after August 22, 1996, will be ineligible for most means-tested public benefits for at least five years after their entry to the country.

Noncitizens Do Not Appear to Be Naturalizing to Retain Benefits

• The consistently low share and number of naturalized immigrants who receive benefits indicate that few immigrants are becoming citizens in order to retain benefits. If most immigrant benefit recipients were naturalizing to retain benefits, the number of naturalized citizens receiving benefits would have grown substantially more than it did.⁽¹⁷⁾

Rising Incomes Do Not Explain Lower Program Participation Rates among Noncitizens

• One possible explanation for the faster declines in program participation among noncitizens than among citizens could be that incomes are rising faster for noncitizens. To address this issue, we use demographic standardization techniques (described below) to partition changes in program participation over the 1994­97 period into the share attributable to changes in income and the share attributable to changes in income-specific participation rates. Our analysis finds that most of the change is not due to rising incomes. Only 6 percent of the decline in welfare use among noncitizens, versus 30 percent of the decline among citizens, can be explained by rising incomes; similar results hold for food stamps and Medicaid (Table 4).

Reduced Use of Health and Other Benefits among Populations Not the Focus of Welfare Reform

• While the apparent decline in welfare use among noncitizens of working age may be an intended and positive policy outcome, these data also reveal sharp declines in the use of health, nutritional, and cash assistance within populations that are thought to be more vulnerable and were not a focus of welfare reform. These vulnerable populations include refugees and the citizen children of noncitizens. As the data presented above indicate, there have been sharp declines in benefits use for both populations. At the same time, health uninsurance rates for noncitizens (46 percent) remain much higher than among citizens (16 percent).⁽¹⁸⁾ Persistently high uninsurance rates, coupled with the decline in program participation documented here, have broad implications for targeting outreach efforts to expand enrollment in programs like California's Healthy Families initiative and for other state efforts to expand enrollment in Medicaid and the Child Health Insurance Program.⁽¹⁹⁾

Noncitizens Do Not Demonstrate a Greater Propensity to Receive Benefits

• The data indicate that higher benefit use rates on the part of noncitizen versus citizen households are due to the fact that immigrant households are poorer and more likely to contain children, not because noncitizens have a greater disposition towards receiving benefits. In fact, among poor households with children, immigrants have lower use rates for welfare, food stamps, and Medicaid both before and after welfare reform.

Notes on the Analysis

Legal Status Distinction. This analysis employs imputation techniques developed by the Urban Institute that make it possible to assign some legal statuses (notably refugee and nonimmigrant status) to foreign-born persons included in the Current Population Survey. These techniques enable us to disaggregate changes in immigrants' benefit use by legal status  despite the fact that the CPS only distinguishes citizens from noncitizens. Disaggregating the immigrant population in this manner is important because usage patterns vary considerably by legal status and, following welfare reform, legal status has become an increasingly important determinant of immigrants' eligibility for public benefits. Specifically, the rules below have applied following welfare reform:

• Naturalized citizens remain eligible for public benefits on the same terms as native-born citizens.
• Noncitizen refugees retain eligibility for means-tested federal benefits, including food stamps, Child Health Insurance Program (CHIP), Medicaid, TANF, and SSI, for five to seven years following their entry into the United States.
• Legal immigrants entering before August 22, 1996, retain eligibility for several major means-tested public benefit programs: Medicaid, TANF, and the Child Health Insurance Program. Each is jointly funded by the states and the federal government. Welfare reform gave the states the option to extend TANF and Medicaid to pre-enactment legal immigrants and virtually all states have chosen to do so.

Pre-enactment legal immigrants' eligibility for federal food stamps remains limited to the elderly, disabled, and children. California and other states have extended state-funded food stamps to working-age immigrants left out by the federal program.

While pre-enactment immigrants' eligibility for SSI was largely eliminated by the 1996 legislation, it has been restored to disabled or elderly immigrants who were receiving SSI when welfare reform passed or who subsequently become disabled.

• Legal immigrants entering after August 22, 1996, have been barred from receiving federal means-tested benefits, including TANF, Medicaid, CHIP, SSI, and food stamps, for at least five years after entry and effectively until they naturalize.
• Undocumented immigrants remain generally ineligible for most major public benefits.

Partitioning the Change in Welfare Use. The percentage of a group (e.g., citizens, noncitizens, natives) using welfare can be thought of as the product of two sets of percentages or rates: the income distribution of the group and the percentage of each specific income category who receive welfare. To cite a specific, but simplified, example using data from tables 1 and 4, 6.7 percent of citizen households received welfare in 1997. We arrived at this percentage from the following calculation: 17.9 percent of citizen households with incomes below 200 percent of poverty received welfare, and these households represent 30.7 percent of citizen households; of the 69.3 percent of citizen households with incomes above 200 percent of poverty, only 1.7 percent receive welfare. Thus, 6.7 = 17.9 x 0.307  +  1.7 x 0.693. For noncitizen households, 9.0 percent received welfare in 1997 (9.0 = 14.5 x 0.541  +  2.6 x 0.459). In other words, poor noncitizen households are less likely to use welfare than poor citizen households (14.5 percent versus 17.9 percent), but poor households are much more common among noncitizens than citizens (54.1 percent versus 30.7 percent), so overall noncitizen households are more likely to use welfare than are citizen households.

We can think of two extreme explanations for the change in welfare use between 1994 and 1997. At one extreme, the 35 percent reduction in use for noncitizens (from 13.9 to 9.0 percent) could occur because the rate of welfare use at every income level for noncitizens fell by 35 percent; were this to occur, the overall rate decline would be explained completely by changes in usage rates. On the other hand, the rate of welfare use could stay constant for each income group of noncitizens, but incomes could rise so that more of the noncitizen population fell into higher income groups which use less welfare. In this case, the overall rate decline would be due entirely to changes in income level. In practice, neither extreme occurs; a combination of the two accounts for the change.

A demographic technique called standardization permits us to partition the overall change in welfare use into a portion attributable to changes in usage rates and a portion attributable to changes in income distribution. The standardization requires four sets of percentages, two for each year: the percentage of the population falling into each income category in 1994 and 1997 (i.e., the income distributions), and the percentage of each income category receiving welfare in 1994 and 1997 (i.e., the use rates). For the partition results shown in Table 4, we use eight household income categories: incomes less than 50% of poverty, 50­74%, 75­99%, 100­124%, 125­149%, 150­174%, 175­199%, and 200% of poverty or more.

The 1997 income distribution multiplied by the 1994 detailed use rates gives the percentage that would have received welfare in 1997 if use rates had not changed (i.e., only incomes had shifted). Subtracting this hypothetical rate from the actual 1997 overall use rate gives a measure of change attributable to income changes. We can calculate another measure of the income effect by subtracting the actual 1994 overall use rate with the hypothetical rate computed with the 1994 income distribution and the 1997 detailed use rates. The average of these two estimates is the amount of change between 1994 and 1997 attributable to changes in income. Any remaining change is the share attributable to changes in welfare usage patterns.⁽²⁰⁾

Coverage of welfare use in the CPS. Finally, we should note that the CPS data on benefits use employed in this analysis were reported by the Census Bureau. Both welfare use and welfare income are known to be underreported, possibly substantially, in the CPS. We do not correct for either type of underreporting in our analysis. Further, the data have not been adjusted to take into account program eligibility rules or the misreporting of public benefit use on the part of immigrants and native citizens. Our uncorrected comparisons assume, in effect, that reporting patterns did not change between the 1995 and 1998 CPS. Despite these limitations, CPS data are conventionally used to characterize trends in benefit use.⁽²¹⁾ There is no reason to believe that the trends documented in this report are biased or otherwise invalid.

Tables and Figures

BIBLIOGRAPHY

Das Gupta, Prithwis. 1993. Standardization and Decomposition of Rates: A User's Manual. U.S. Bureau of the Census, Current Population Reports, Series P23-186. Washington, D.C.: U.S. Government Printing Office. September.

Ellwood, Marilyn R., and Leighton Ku. 1998. "Welfare and Immigration Reforms: Unintended Side Effects for Medicaid." Health Affairs 17 (3): 137­51. May/June.

Fix, Michael, and Wendy Zimmermann. 1998. "The Legacies of Welfare Reform's Immigrant Restrictions." Interpreter Releases. November 16.

Fronstin, P. 1998. "Sources of Health Insurance and Characteristics of the Uninsured: An Analysis of the March 1998 Current Population Survey." EBRI Issue Brief 204. December.

National Research Council. 1998. From Generation to Generation, The Health and Well-Being of Children in Immigrant Families. Washington, D.C.: Institute of Medicine, National Academy Press.

Passel, Jeffrey S., and Rebecca Clark. 1998. "Immigrants in New York: Their Legal Status, Incomes, and Taxes." Washington, D.C.: The Urban Institute. April.

Van Hook, Jennifer, Jennifer E. Glick, and Frank D. Bean. 1999. "Public Assistance Receipt Among Immigrants and Natives: How the Unit of Analysis Affects Research Findings." Demography 36 (1): 111­20. February.

Zimmermann, Wendy, and Michael Fix. 1998. "Declining Immigrant Applications for Medi-Cal and Welfare Benefits in Los Angeles County." Washington, D.C.: The Urban Institute. July.

Endnotes

1.  The Personal Responsibility and Work Opportunity Reconciliation Act, Pub. L. 104-193 (1996).

2.  While most of these unqualified immigrants are undocumented immigrants, many unqualified immigrants are legally present in the United States and have work authorization. See, generally, Michael Fix and Wendy Zimmermann, "The Legacies of Welfare Reform's Immigrant Restrictions," Interpreter Releases, November 16, 1998.

3.  "Public charge" is a term used by the Immigration and Naturalization Service (INS) and the State Department to describe someone who is, or is likely to become, dependent on public benefits. Public charge considerations have historically been a factor in the admissibility of aliens (i.e., grant of a green card) and, only rarely, in the deportation of aliens who have been in the United States less than five years.

In the past several years, public charge has been inappropriately invoked in some instances where noncitizens have attempted to reenter the United States and where immigrants have sought to naturalize. In some cases, noncitizens seeking to adjust status, naturalize, or reenter the country have been asked to repay public benefits. The legality of compelling repayment in these contexts is suspect.

4.  See Wendy Zimmermann and Michael Fix, "Declining Immigrant Applications for Medi-Cal and Welfare Benefits in Los Angeles County," The Urban Institute, July 1998.

5.  The CPS collects information on program use and income in March for the preceding calendar year. Thus, the information collected in the March 1995 CPS pertains to calendar year 1994; and the March 1998 CPS, to calendar year 1997. Throughout this document, we use the CPS data from March 1995 and 1998 for reference years 1994 and 1997. The data from the March 1995 CPS have been reweighted to correct for an error in the official weights (Jeffrey S. Passel and Rebecca Clark, "Immigrants in New York: Their Legal Status, Incomes, and Taxes," The Urban Institute, April 1998). The CPS data used here are not corrected for underreporting of welfare use or welfare income. (See "Coverage of Welfare Use in the CPS.")

6.  See Passel and Clark, 1998.

7.  Welfare reform gave the states the option of barring legal immigrants in the United States before August 22, 1996, from TANF and Medicaid. However, virtually all states extended benefits to these pre-enactment immigrants. See Fix and Zimmermann, 1998.

8.  "Federal means-tested public benefits" have been determined to be Temporary Assistance for Needy Families (TANF), Medicaid, the Child Health Insurance Program (CHIP), Supplemental Security Income (SSI), and food stamps.

9.  Unless otherwise noted, the comparisons of citizens and noncitizens exclude refugees and temporary immigrants (i.e. "nonimmigrants" according to immigration law), which are treated separately. Citizens include natives, persons born in Puerto Rico and other outlying areas, and immigrants who have acquired citizenship through naturalization. Noncitizens include aliens admitted as lawful permanent residents and undocumented immigrants.

10.  A refugee is defined legally as a person outside his/her country of nationality who is unable to return because of a well-founded fear of persecution. Because their departure from their home country is involuntary and unplanned and because many suffer physical or mental trauma, refugees have been made eligible for most public benefits from the date of their arrival.

We assign refugee status based on country of birth and period of entry to the United States. For persons entering after 1980, we define a "refugee country" as one where refugees and asylees account for more than 40 percent of total admissions of legal permanent residents, refugees, and asylees during any two-year period. See Passel and Clark, 1998.

11.  See, for example, From Generation to Generation, The Health and Well-Being of Children in Immigrant Families, National Research Council, Institute of Medicine, National Academy Press, Washington D.C., 1998.

12.  See Fix and Zimmermann, 1998.

13.  See Jennifer Van Hook, Jennifer E. Glick, and Frank D. Bean, "Public Assistance Receipt Among Immigrants and Natives: How the Unit of Analysis Affects Research Findings," Demography 36 (1, February 1999): 111­20.

14.  In the CPS, food stamp usage is a household-level variable, so we do not report individual usage patterns.

15.  They are, however, consistent with administrative data, specifically the Food Stamp Program Quality Control data for fiscal years 1994 and 1997.

16.  See, for example, Passel and Clark, 1998.

17.  The data in Detailed Table B show that the number of naturalized citizens age 18 and over reporting welfare income rose by only 75,000 between 1994 and 1997, while the number of naturalized citizens increased by 1,419,000. During that period, the number of noncitizens reporting welfare use fell by 316,000.

18.  P. Fronstin, "Sources of Health Insurance and Characteristics of the Uninsured: An Analysis of the March 1998 Current Population Survey," EBRI Issue Brief 204, December 1998.

19.  We should note that declines in Medicaid caseloads affect more populations than just noncitizens and can also be viewed as an unintended effect of welfare reform. See Marilyn R. Ellwood and Leighton Ku, "Welfare and Immigration Reforms: Unintended Side Effects for Medicaid," Health Affairs 17 (3, May/June 1998): 137­51.

20.  This second difference can be shown to be equivalent algebraically to the average of two separate estimates of change attributable to difference in usage rates. One measure compares the actual 1997 overall use rate with a hypothetical rate computed as the 1997 detailed use rates multiplied by the 1994 income distribution. The other subtracts actual 1994 overall use rate from the hypothetical rate based on the 1994 detailed use rates and the 1997 income distribution. For more information on standardization and partition, see Prithwis Das Gupta, Standardization and Decomposition of Rates: A User's Manual, U.S. Bureau of the Census, Current Population Reports, Series P23-186, Washington, D.C.: U.S. Government Printing Office, September 1993.

21.  See, for example, Fronstin, 1998. Planned Urban Institute reports that take into account patterns of underreporting and program eligibility rules are expected in the summer of 1999.