Prepared for a conference sponsored by the International Center for Migration, Ethnicity and Citizenship of the New School for Social Research, and the Luce Foundation: "New Immigrants in New York: The Incorporation of Recent Immigrants in New York City." New York City, December 7-8, 2000.  We would like to thank Makada Henry for research assistance.

I.  National Legislation
    Welfare reform marked the end of the entitlement to cash assistance. State discretion in determining eligibility and benefits was increased.  In ending the entitlement, PRWORA decoupled eligibility for cash and in-kind assistance.  The stated intention was to reduce access to cash assistance, but, at least for citizens, not to reduce Food Stamps or Medicaid.  Welfare reform also included a substantial devolution from the federal government to the states of the responsibility for deciding on eligibility and benefits for noncitizens.  One of the goals of welfare reform was to restrict access to welfare for immigrants.  The strongest restrictions were imposed on noncitizen access to the federal Food Stamp program, but states have the option of replacing federal Food Stamps with their own programs.  Hence, variation in the treatment of immigrants across states is likely to have increased after welfare reform.
    Historically, naturalized citizens and refugees have been eligible for the same benefits as native-born citizens; but legal permanent residents have been subject to "deeming" and "public-charge" restrictions, and temporary and undocumented immigrants have been ineligible for benefits.  "Deeming" means that the sponsor's income is deemed to be available to an immigrant for a certain number of years after arrival -- three for TANF and Food Stamps, and five for SSI.  If enforced, this would effectively prevent an immigrant from using welfare for three to five years. According to the law, an immigrant can be deported if s/he becomes a public charge within five years, but this has rarely been enforced.
    Table 1 summarizes the effects of PRWORA on citizens and non-citizens in New York State.  Undocumented immigrants and those on temporary visas remain ineligible for benefits (other than Medicaid emergency services).  Legal immigrants arriving after August 22, 1996 were barred from all Federal means-tested benefits (other than Medicaid emergency services) for at least five years, and effectively until they naturalize.  For legal immigrants who were in the United States before August 22, 1996, the sponsor-income deeming period was extended for up to ten years for most types of benefits.  Noncitizen refugees' and asylees' period of eligibility for means-tested federal benefits (TANF, SSI, Food Stamps, Medicaid, Child Health Insurance Program) was limited to 5-7 years after entry.
    PRWORA also barred noncitizen immigrants who were in the United States before August 22, 1996 from Food Stamps, unless they were disabled, under 18, or over 64, and already enrolled in the Food Stamps program on that date.  New noncitizen applicants were to become ineligible in October 1996, and many legal immigrants' eligibility was supposed to end in September 1997, but their eligibility was restored by the Balanced Budget Act (BBA) of 1997.  As shown in Table 1, in the case of Food Stamps, New York has allowed counties the option of a partial replacement of federal Food Stamps for noncitizens by a state-county program.  Eligibility is limited to those less than 18 or greater than 59, and to the disabled.  New York City is among those counties choosing to participate.
    States were given the option to bar legal immigrants who were in the United States before August 22, 1996 from TANF and Medicaid (other than SSI-derived benefits), but nearly all states (including New York) kept them eligible.  PRWORA also made these legal immigrants who were in the United States before August 22, 1996 ineligible for SSI, but the 1997 BBA restored eligibility for disabled or elderly immigrants who were receiving SSI on that date or who later became disabled.
    As shown in Table 1, General Assistance, previously known in New York as Home Relief and now as Safety Net Assistance, was restricted for both citizens and noncitizens.  Cash assistance was limited to 24 months for all recipients of Safety Net Assistance.  Except for asylees, benefits for new noncitizens and other new state residents are equal to only 50 percent of the state benefit level.
    Another Federal law enacted in 1996, the Illegal Immigration Reform and Immigrant Responsibility Act (IIRAIRA), tightened the requirements for sponsors to support immigrants.  It requires sponsors to sign a legally enforceable affidavit to support the immigrant if necessary, and authorized government-funded agencies to sue for reimbursement of means-tested benefits.

II.  Public Assistance Receipt in New York City
    New York City, like most other states and cities, has experienced a dramatic drop in public assistance rolls since 1994.  Between 1994 and 2000 the welfare caseload has fallen by more than 500,000 persons.  The decline has brought an unambiguous fiscal benefit to New York City, with direct savings of almost $500 million per year.  A number of national studies have explained the caseload decline as partly resulting from the strong economy and partly from specific policy actions, including active diversion of applicants (Figlio and Ziliak, 1999; Wallace and Blank, 1999).
    The decline in welfare caseloads in New York City has been controversial.  There have been allegations that an important reason for the drop in welfare rolls has been administrative diversion - making it tougher to get welfare by stricter administrative procedures (Sengupta, 2000).  The case for diversion is supported by a sharp rise in the number of applicants who were rejected for public assistance, from 26 percent to 56 percent, and a 77 percent increase between 1993 and 1998 in the number of fair hearings complaints by applicants who were denied access to public assistance.  In the vast majority of these hearings, the city’s actions have been overturned, and applicants declared eligible for public assistance (Chernick and Reschovsky, 2000).  The city has also been accused of illegally denying applicants for Food Stamps and Medicaid, and the courts have prevented the city from converting its welfare intake centers into so-called "job centers" (Welfare Law Center, 2000)
    If administrative diversion has been an important factor in reducing welfare caseloads, then such procedures may have a differential impact on citizens vs. noncitizens.  For example, if the city is taking advantage of language barriers as a way to discourage applicants, non-natives are likely to be affected more than natives.
    Given the changes in the law, and the increased administrative hurdles which the city has raised to getting public assistance, our expectation was that New York City would show a reduction in the number of families getting both public assistance and Food Stamps.  Nationally, the intent of the law was to reduce the receipt of public assistance, with less reduction in Food Stamps, and perhaps an expansion in Medicaid coverage.  However, Food Stamps might be expected to decline more in New York than nationally, because many new immigrants arrived in New York after 1996 and most of them are ineligible for Food Stamps until they become citizens.

Other Factors Affecting Program Participation
    In addition to the "push" provided by PRWORA and city administrative policies, "pull" factors were also operating to reduce the public benefit rolls during the mid-1990s.  Cities around the country have benefited from the strong economic growth in the 1990's.  The most recent data show that for the nation as a whole, between 1998 and 1999 the number of central city residents in poverty fell by 1.8 million, and household income of central city residents, though still substantially lower than in the rest of the country, grew faster than elsewhere (U.S. Census Bureau, 2000).  Job growth has also been strong in New York City in this period, actually surpassing the national rate in the most recent years.  From 1997 to 1999, New York City job growth exceeded two percent each year, outperforming any equal span of time during the last three decades.  The expanding New York economy has increased demand and possibly wages for low-skilled workers.  Increases in the Earned Income Tax Credit (EITC) and the minimum wage have also made work more attractive to low-skilled individuals in recent years.  New York State supplements the national EITC with its own refundable credit.  The New York State credit was expanded after 1997, so that it now equals 22 percent of the federal EITC.
    Systematic differences in employability and earnings capacity between citizens and noncitizens could also have an effect on the rates of decline in public assistance.  If the pull factor of economic growth was more powerful for those with greater earnings capacity, and non citizens tend to have less education and higher incidence of female headship, then one would expect noncitizens to move off the welfare rolls at a slower rate than citizens.
    In addition to the general "pull" of a stronger labor market, the mere passage of time might make those immigrants who were already in the U.S. in 1994-95 less likely to need assistance by 1997-98.  The longer they are in the U.S., the more their English, education, location-specific skills, and knowledge of the job market would be expected to improve.  Therefore, they might be more likely to find a job and to earn more at the later date.  Of course, with the passage of time immigrants also acquire eligibility for aid and learn more about the New York City welfare system.  This might make some immigrants more likely to be on the rolls at the later date.  Because of these offsetting forces, the net effect on program participation over time is unpredictable.
    It is difficult to disentangle the effects of welfare reform from the influence of these other factors on welfare recipiency in a single city.  Moreover, without longitudinal data, it is not possible to trace the flows off of and onto benefit programs in detail.  We can only observe net changes in program recipiency. Our goal in this paper is therefore more modest: to compare public transfer program participation among citizens and noncitizens in New York City before and after the welfare reform act of 1996.

III.  Findings of Other Studies
    An influential 1999 report issued by the Urban Institute (Fix and Passel, 1999) found a "chilling effect" of welfare reform on the receipt of cash assistance by noncitizens.  They combined all cash assistance (AFDC/TANF, General Assistance, and SSI) into one category, and examined the receipt of cash assistance, Food Stamps, and Medicaid nationwide in 1994 and 1997.  They found that receipt of cash assistance fell from 13.9 percent to 9.0 percent of households headed by noncitizens, but only from 7.8 percent to 6.7 percent of households headed by citizens.  A similar pattern was found for Food Stamps and Medicaid.  The proportion of noncitizen-headed households receiving Food Stamps fell from 15.4 percent to 10.8 percent, while that of citizen-headed households fell only from 8.5 percent to 6.8 percent.  Medicaid receipt fell from 26.5 percent to 20.8 percent of noncitizen households, but only from 13.5 percent to 12.6 percent of citizens.  Noncitizens’ receipt was higher than citizens’ because they had lower incomes and were more likely to have children.  When these factors were controlled for, noncitizens were less likely to receive benefits than citizens.
    The Urban Institute's 1999 report also examined the receipt of cash assistance and Medicaid by individuals.  Among adults age 18-64, noncitizens' cash assistance receipt fell 41 percent from 1994 to 1997, while citizens' fell only 15 percent.  Working-age noncitizens' Medicaid receipt fell thirty percent, while citizens' fell only ten percent.  Among the elderly, noncitizens are much more likely than citizens to be on SSI and Medicaid because they are less likely to qualify for Social Security or Medicare, but there were no declines in welfare (i.e., SSI) or Medicaid receipt for either citizens or noncitizens.  The report concluded that the larger declines for noncitizens than citizens must be due to a "chilling effect" of the policy changes, not to actual changes in eligibility, because eligibility had changed for only a few immigrants by 1997.  They also concluded that few noncitizens are naturalizing to retain benefits, as the number of naturalized citizens reporting welfare receipt rose by less than a quarter of the reduction in the number of noncitizens reporting welfare receipt.
    Research by Borjas (1999) corroborates the Urban Institute's findings with respect to Food Stamps.  He studied the trends in use of Food Stamps by immigrants from 1969 to 1997.  He found that immigrants and natives had virtually the same rate of Food Stamp receipt in 1969, but between 1969 and 1994 immigrants' use increased faster than natives'.  Since 1994, however, the gap has narrowed as the decline has been steeper for immigrants than natives.  All of the immigrant-native gap is due to differences in socioeconomic characteristics, especially education.   In another paper, Borjas (2000) followed up the Urban Institute's study with his own analysis of benefit receipt in 1994-1998.  He found that the excess decline in noncitizens' receipt of benefits was almost entirely a California phenomenon.  In California, the all-noncitizen households' benefit-participation rates (for cash benefits, Food Stamps, and Medicaid combined) dropped from 27.7 percent in 1994 to 14.0 percent in 1998, while that of all-citizen households dropped by only 1.6 points, to 13.6 percent.  In the rest of the country, the drop in benefit receipt was the same for all-noncitizen and all-citizen households, just 2 percentage points.  Households containing both citizens and noncitizens (primarily noncitizen parents with children born in the U.S.) lost benefits to the same degree, about five percentage points, in California and the rest of the country.  In California, but not elsewhere, these "child-only" welfare cases retained benefits to a greater degree than all-noncitizen households.  If we combine these mixed households with the others headed by noncitizens, the drop in benefit receipt would be only a little bigger for noncitizen- than for citizen-headed households outside California.
    Borjas' findings should make us cautious about inferring local patterns from national data.  Since California has a large proportion of the nation's immigrants, whatever occurs there heavily influences the statistics for immigrants in the U.S. as a whole.  Historically New York City -- now the second-largest immigrant destination -- has had a more pro-immigrant ethos than California.  This paper is designed to find out whether welfare reform had a "chilling effect" on immigrants' public benefit receipt in New York City, as it did in California, or whether New York City differs from California as much in this respect as it does in other ways.
 
IV.  Data
    Our data source for analyzing changes in participation in welfare programs among immigrants is the March supplement to the Current Population Survey (CPS) for 1995, 1996, 1998, and 1999.  The CPS is a monthly survey of about 50,000 U.S. households.  The March CPS asks questions about family income and demographic characteristics, and about participation in certain means-tested public programs: public assistance ("welfare," including AFDC/TANF and General Assistance), Supplemental Security Income (SSI), Food Stamps, Medicaid, housing subsidies (Section 8 vouchers or public housing), energy assistance, free or reduced price school lunch, and Earned Income Tax Credit (EITC).  For most types of benefit, the CPS reports only whether anyone in the household received it.  Receipt of benefits by individuals is recorded only for public assistance, Medicaid, and EITC.
    Beginning in 1994, the March CPS has asked questions regarding place of birth, date of arrival, and citizenship status.  Since the changes in the law relate to citizenship status and date of arrival, this distinction is crucial to our analysis.  Passel (1996) has identified an error in the 1995 CPS weights, which results in undercounting Asians/Pacific Islanders and American Indians/Eskimos/Aleuts.  We use his corrected weights and race codes for 1995.
    The New York City sample of the March CPS consists of 2123 households in 1995, 1579 in 1996, 1586 in 1998, and 1568 in 1999, with about 4000 persons in each of the latter three years.  To increase our sample sizes before and after welfare reform, we pooled 1995 and 1996 ("before"), and 1998 and 1999 ("after").  This gives us 3702 households in 1995-96, and 3154 households in 1998-99, with about 9000 persons in the former and 8000 in the latter period.1    Because the March CPS asks about income and program participation in the previous year, we refer to the "before" period as 1994-95, and the "after"  period as 1997-98.  Because we are dealing with the low-income population, we ignore the topcoding of income data in the CPS.

Methods
    Our focus in this paper is on the major benefit programs that were affected by the 1996 welfare reform: public assistance (AFDC/TANF and General Assistance), SSI, Food Stamps, and Medicaid, as well as housing vouchers and public housing.  In New York City, before welfare reform public assistance was called AFDC and Home Relief; now those programs are called Family Assistance and Safety Net Assistance.  To analyze program participation by noncitizens, we divide the sample into three groups: 1) citizens; 2) noncitizens who entered the United States before 1996; and 3) noncitizens who entered the United States between 1996 and 1999.  The  distinction between earlier and later arriving immigrants is based on the fact that, with the exception of refugees and asylees, all noncitizens arriving after August 22, 1996 are ineligible for Federal benefit programs for at least five years.2   Those already in the U.S. in 1996 were mostly "grandfathered."  Under state option, most states chose to retain eligibility for TANF and Medicaid for noncitizens who were in the U.S. on August 22, 1996.  Moreover, the Balanced Budget Act of 1997 restored eligibility for SSI and Food Stamps to many noncitizens who were in the U.S. in 1996.   Our goal, therefore, is to determine whether there have been differences in the change in participation in the major public programs as between citizens and noncitizens who arrived before 1996.  For ease of exposition, we shall refer to noncitizens who arrived before 1996 simply as "noncitizens."
    We use a standard difference-in-difference approach, comparing the change in program  participation rates for citizens between 1994-95 and 1997-98 to the change in participation rates for noncitizens.  The difference-in-difference is computed as:  (noncitizen rate1998-99 - noncitizen rate1994-95) - (citizen rate1998-99 - citizen rate1994-95).  Hence a negative number means that the benefit receipt rate has declined more for noncitizens than for citizens, while a positive number means a bigger drop for citizens.  We compute a t-statistic for the difference-in-differences.  Given the small sample size, we will consider t-statistics greater than 1.28 (a prob-value of .20 in the two tails) as statistically meaningful.
    We compute participation rates on both a household and a person basis.  For most types of benefit, the CPS reports only whether anyone in the household received it.   We compute percentages both of households and of persons having access to the benefit in this way.  Personal receipt is recorded only for PA and Medicaid, and we report those percentages, too.  For housing subsidies and Food Stamps, the household is the most useful unit of analysis, since eligibility is based on household income.  Also, a reduction in housing costs through housing subsidies clearly benefits everyone living in the household.  The household basis also allows us to compare our results to studies of immigrant participation in other states, since most other analyses have been of household receipt of benefits.  It is also useful to compare the person-level analysis to the household-level analysis because household size may vary across citizenship and benefit receipt status.

V.  Results
    The results are presented in Figures 1 through 9.  The data on which these figures are based are reported in appendix tables A-1 through A-5.  Results are presented using both household and person weights.  Note that the person-weighted version is not an analysis of whether an individual personally received a program’s benefits.  A household or a person is counted as receiving a benefit if anyone in the household, whether child or adult, got the benefit.  In the household-level tables, households are classified in terms of the citizenship status of the household head, while in the person level tables, individuals are classified according to their own citizenship status.  Changing the citizenship classification from the household to person level is potentially important because many households with adults who are noncitizens may have children born in the United States.  In the household-level tables, each household is counted only once, while in the person-level tables, each person is counted (including children).  Hence, differences in size of household with citizen and noncitizen heads could have an impact on the person-weighted results.
    To make these definitions clear, consider the third row, first column of Tables A-1a and A-1b.  Table A-1a indicates that, in 1994-95, 13.7 percent of households headed by a noncitizen had at least one member of the household who got public assistance.  Table A-1b indicates that 12.0 percent of individuals who were noncitizens resided in households where at least one person got public assistance in 1994-95.
    Tables A-1 and A-2 show receipt of cash assistance -- AFDC/Family Assistance and Home Relief/Safety Net Assistance in Table A-1, and SSI in Table A-2.   Table A-1 shows a drop in receipt of public assistance of 2.5 percentage points for citizen households, and 3.3 percentage points for noncitizen households.  The difference in differences between citizens and noncitizens is insignificant.
    Figure 1 shows the household-level change in public assistance receipt, and Figure 2  shows the person-weighted change in public assistance.  In contrast to Figure 1, Figure 2 shows both a higher rate of welfare receipt for citizens than for noncitizens, and a greater rate of decline.  The difference-in-differences is statistically significant at the 13% level (see Table A-1b).  What is the source of the difference between households and persons?  Recall that Figure 2 is based on assigning weights to each person in the household, and classifying individuals by their own citizenship status, rather than that of the household head.  If a citizen lives in a household whose head is a noncitizen, and at least one household member gets public assistance, that person will be counted as a citizen getting public assistance in Figure 2.
    First we consider why the rate of public assistance receipt among citizens rises when we weight by persons and classify by the citizenship status of the person, as opposed to the household.  Nearly a quarter of the children in our sample are citizens living in households where the head is a noncitizen.  Children in these "mixed" households had a high rate of welfare receipt -- thirty percent in 1994-95 -- relative to persons in the other types of mixed household, where the head is a citizen and other adults are noncitizens, or vice versa.  When we weight by persons, citizen children in mixed households are reclassified into the citizen category, raising the rate of welfare receipt among citizens.  In contrast, the rate falls among noncitizens, because noncitizen members of citizen-headed households are reclassified into the noncitizen category, and this group has a relatively low rate of welfare receipt.
    However, this reclassification story does not explain why the drop in household public assistance receipt is greater for citizen persons than noncitizen persons.  One possibility is that  children who are citizens living in mixed households were more likely to lose benefits, perhaps because their  noncitizen parent(s) were especially likely to be cut off from welfare after welfare reform.  This conjecture turns out to be false.  When we look at changes in welfare receipt for mixed households, as opposed to households where all of the members are either citizens or noncitizens, we find hardly any change among the mixed category, while there is a substantial drop among the homogeneous households.  This finding by itself would lead to a lower rate of decline among citizens when person weights are used instead of household weights.
    This result for New York City is different from recent caseload changes in California, where the child-only caseload has increased sharply since 1994.  In California, these child-only cases are predominantly citizen children with undocumented alien parents (Macurdy et al, 2000) According to the Human Resources Department of the City of New York, New York City has relatively few child-only cases.3
    Instead, the reversal in the rate of decline turns out to hinge on household size.  In 1994-95 household size was about the same for citizens and noncitizens among public assistance  recipients.  However, by 1997-98 average household size among noncitizens on welfare had risen by 23 percent, from 3.3 to 4.1 persons.  Among citizens on welfare, household size remained unchanged.  Thus among noncitizen households, smaller families left the rolls at a faster rate than larger families.  Though noncitizen households lost benefits at about the same rate as citizen households, the drop in the proportion of persons receiving benefits is smaller for noncitizens than citizens, because smaller families left the rolls at disproportionate rates.  Why this differential change by household size occurred remains to be explained.
    To summarize our discussion, the household-weighted results indicate no significant difference between citizen and noncitizen households in rates of welfare decline.  This would suggest that the impact of changes in welfare policy are neutral as between immigrant and non-immigrant households.   However, there appears to be a differential impact in terms of household size.  Among noncitizens, smaller households are much more likely to leave the rolls than larger families.  Among citizens, there is no such pattern.
    In other research (Chernick and Reimers, 2000), we have investigated the changes in public assistance receipt by ethnicity as well as citizenship.  The results shed some light on this apparent neutrality of the impact of welfare reform on noncitizens and citizens in New York City.  We estimated a set of linear probability models of public assistance receipt, which include ethnicity/citizenship effects, period ("after" welfare reform; i.e., 1997-98) effects, and interaction terms between period and ethnicity/citizenship, plus various combinations of demographic controls and interaction terms between the period and control variables.  The ethnicity/citizenship categories are: non-Hispanic blacks, Puerto Ricans, other Hispanic citizens, Hispanic noncitizens, and non-Hispanic whites and Asians (the reference group). The demographic controls are dummy variables for female headship, presence of children under age 18, whether the household head is under age 65, and whether s/he lacks a high school diploma.
    The coefficients of the period effect and the interaction term between period and ethnicity are summarized in Table 3. The coefficient on the interaction term, say for black non-Hispanics, measures the "difference in differences"; that is, the change in the probability of welfare receipt for non-Hispanic blacks from 1994-95 to 1997-98, minus the change in the probability of welfare receipt for the reference group (non-Hispanic whites and Asians).  The t-statistic provides a statistical test of whether the drop in receipt is significantly greater among non-Hispanic blacks than these others.
    The regression shows no significant change in the rate of welfare receipt among blacks, regardless of whether controls are introduced.   Among whites and Asians, the decline is only about one percentage point, which is at or close to statistical significance, when the interactions of age and education with period are omitted.  When they are included, the period effect for whites and Asians is always insignificant.  This last result indicates that, once we have controlled for the effect of the head’s age and education on the change in the probability of household welfare receipt in the later year, there is no additional change in the rate of receipt by whites and Asians.
    However, Puerto Ricans suffered a highly significant decline in welfare receipt — 12.5 percentage points without controls, and 9.1 percentage points with the full set of controls.  Among other Hispanics, the estimated declines are only about half as large as for Puerto Ricans, but the decline is measured more precisely for noncitizens than citizens.  In fact, for Hispanic citizens the decline between 1994-95 and 1997-98 is not significantly greater than for whites and Asians.  For noncitizen Hispanics, the decline is significantly greater at the 6.5 percent level, even when we control for family structure, age and education.  However, when we control for the differential effect of family structure in the later year (columns 5 and 7 of Table 3), the decline for Hispanic noncitizens also becomes insignificant.  This insignificance indicates that, if a household is at risk of welfare receipt in 1997-98 because it is headed by a female, then for Hispanics other than Puerto Ricans there is no additional likelihood that they lost public assistance.  Thus, once we introduce controls for the characteristics that put families at risk of receiving public assistance, the greater decline for Hispanics seems to have occurred mainly among Puerto Ricans.
    These results indicate that one group of citizens in New York City — Puerto Ricans — were disproportionately affected by the combination of welfare reform and the expanding economy.  The greater decline in welfare receipt for this group of citizens made the overall declines for citizens and noncitizens more similar than they would otherwise have been.
Moreover, among other Hispanics, citizenship does not seem to have had any impact on the rate of decline in welfare receipt.

Educational Attainment and Public Assistance Receipt
    We hypothesized that differences in the earnings potential of citizens and noncitizens could also help to explain differences in the rate of decline in public assistance.  Lack of a high school diploma is a high risk factor for welfare receipt (Klawitter et al, 2000).  Therefore, we divide households into two groups: those headed by high school dropouts, and those whose head has at least a high school diploma.   Table 2 shows that in New York City in 1997-98, the rate of welfare receipt was 17.8 percent among households headed by a high school dropout, as opposed to five percent in more educated households.  Lack of a high school diploma is much more common among noncitizens than citizens, with 36 percent of noncitizen households headed by a high school dropout, as compared to 23 percent of citizen households.  Hence, based on educational attainment alone, noncitizens are at greater risk of needing welfare than citizens.
    We first consider levels of welfare receipt by education and citizenship status.  Among those who have not graduated from high school, we find that noncitizens are about four percentage points less likely to get public assistance than citizens.  However, among the more educated, noncitizens have a three percentage point higher rate of welfare receipt.
    We next look at changes in the rate of welfare receipt for high school dropouts and for those with high school or more.  New York experienced vigorous job growth in the period from 1995 to 1998.  High school graduates were somewhat more likely to leave the rolls than dropouts, presumably because the former were better able to take advantage of the increase in employment opportunities in this period.  There was a 31 percent drop in the public assistance receipt rate for households headed by a high school graduate, and a 23 percent drop for those headed by a dropout.  However, because the level of welfare participation is so much higher among dropouts, their percentage-point decline is more than twice as large as that of graduates.
    Controlling for education, we find virtually no difference between citizens and noncitizens in the decline in welfare receipt.  Among high school dropouts the decrease was 5.1 percentage points for citizen households, and 5.7 points for noncitizen households.   Among those with more education, the decrease was 2.1 percentage points for citizens, and 2.2 points for noncitizens.  Thus, whatever the combination of expanding job opportunities and tougher administrative screening procedures which have combined to reduce welfare rolls in New York City, the impact on citizen and noncitizen households has been similar.

SSI
    As shown in Figure 3, SSI receipt shows a slight increase for citizen households, and a somewhat bigger increase among noncitizen households.   For citizens, participation goes from 8.4 percent to 8.7 percent, while for noncitizens it increases from 9.2 percent to 11.1 percent (Table A-2a).  Neither of these individual changes nor the difference-in-differences is statistically significant. Unlike public assistance, weighting by persons leaves the results unchanged (see Figure 4).  The increase in SSI receipt among noncitizens is entirely among those age 65 and over. Between 1994-95 and 1997-98 the participation rate in SSI for noncitizen elderly increased by a statistically significant 8 percentage points, from 33 percent to 41 percent.  By contrast, the rate for citizen elderly was only 12 percent in 1997-98.  The reason for the big difference between citizens and noncitizens is that most elderly citizens have worked in the United States long enough to qualify for Social Security benefits that exceed the SSI benefit, whereas many noncitizens have not.
    The 1996 welfare reform act was designed to reduce noncitizen participation in SSI. As shown in Table 1, people who were already on SSI retained their eligibility.  With the exception of refugees and asylees, legal noncitizens must have worked 40 quarters to become eligible.  Hence, legal immigrants must reside in the U.S. for a minimum of 10 years before becoming eligible for SSI.  Refugees and asylees are eligible for SSI for seven years after the date of admission to the United States.  The increase in SSI participation among noncitizens suggests that the welfare reform law, and the fear that noncitizens would be cut off, may have induced some noncitizens living in New York at the time of welfare reform to apply for and receive benefits.  Over time, one would expect the proportion of noncitizen elderly getting SSI to decline, as more of them will have entered the U.S. after 1996 and therefore will be ineligible for SSI if they have not worked in the U.S. for at least ten years.
 
Food Stamps
    As shown in Figure 5 (Table A-3a), participation in the Food Stamp program declined by 3.7 percentage points among noncitizen households, as compared to 1.3 percentage points among citizens.  While not statistically significant, the difference-in-differences is larger than its standard error.  Weighting by persons and classifying by citizenship of the person, rather than the household, moves citizens who live in households headed by a noncitizen to citizenshi/p status and  gives more weight to large households.  These changes raise the Food Stamp participation rate for citizens and lower the rate for noncitizens (Figure 6).  They also enlarge the percentage-point decline, reegardless of citizenship status.  However, in contrast to public assistance, even on a person-weighted basis the decline in participation rates is greater for noncitizens than for citizens.
    The drop in Food Stamp receipt reflects three factors: the pull of the strong local economy reducing need for Food Stamps, changes in legal eligibility as part of the welfare reform bill, and administrative changes which, intentionally or not, have had the effect of denying Food Stamps to many eligible households.  The changes in Food Stamp eligibility are discussed above, and summarize in Table 1. As in a number of other states, New York provides partial state replacement for federal Food Stamps.  The funds for this replacement come from the TANF block grant, and do not represent a commitment of new state or city dollars to Food Stamps.
 
Medicaid
    As shown in Figure 7, in 1994-95 the proportion of households getting Medicaid was about 8 percentage points greater for noncitizen heads than for citizens.  However, the proportion of persons living in Medicaid households is about the same for citizens and noncitizens (Figure 8).  This difference indicates that among citizens, Medicaid households are larger than non-Medicaid households, while among noncitizens there is no difference in household size.
    The data show that there was basically no change in the rate of Medicaid receipt between 1994-95 and 1997-98.   However, as shown in Figures 7 and 8, the overall lack of change masks the fact that Medicaid receipt went down slightly among citizens, but increased among noncitizens.  The "diff-in-diff" between the two groups is not statistically significant.  The increase in Medicaid among noncitizens is linked to the increase in SSI.  SSI recipients -- whether citizens or not -- are categorically eligible for Medicaid.
    In other research, we have found that at least 30 percent of households who lost public assistance in New York City during this time period also lost Medicaid benefits (Chernick and Reimers, 2000).  This loss of Medicaid by public-assistance losers is offset by the increase in SSI recipients (who automatically get Medicaid); by Medicaid expansions, particularly among women giving birth; and by the implementation of New York’s version of the Child Health Insurance Program, Child Health Plus.  Because New York City is required to pay 25 percent of the cost of Medicaid, the fiscal cost of rising recipiency rates for SSI is important.  This is particularly so since expenditures per person on elderly Medicaid recipients are much higher than expenditures on families with children.

Subsidized Housing
    Figure 9 shows rates of housing subsidy.  The subsidies include any program under which the household is paying a reduced rent; that is, both public housing and vouchers.  Surprisingly, the results indicate that among both citizens and noncitizens, the rate of housing subsidy increased between 1994-95 and 1997-98.  The increase is bigger for noncitizens than for citizens, and is statistically significant (see Table A-5a).  The difference-in-differences between citizens and noncitizens is also significant at the 10 percent level.
    The overall increase in rates of housing subsidy indicated by the CPS is puzzling, as is the differential between citizens and noncitizens.  The administrative data from New York City’s Department of Housing Preservation and Development (HPD) do not indicate an increase in the number of public housing units or subsidized units during this time period.  Other experts on housing in New York City also say there has been no increase in public housing units or in vouchers during this period.4  We investigated the possibility that the greater increase for noncitizens reflected the success of certain refugee groups, particularly from the former Soviet Union, in gaining access to subsidized housing in New York City.  However, data on housing subsidy rates by country of origin and citizenship status do not support this conjecture.  Another possibility is that the CPS sampling frame for New York City changed during the period of analysis.  However, experts at the Bureau of Labor Statistics say that the changes in the CPS sample in New York were normal ones that are unlikely to cause an increase in reported housing subsidies.5  Hence, we treat the observed increase with some skepticism.

VI.  Summary and Conclusions
    The 1996 welfare reform bill  (PRWORA), while restricting access to public assistance for all residents, included additional restrictions on noncitizens’ access to public benefit programs, particularly for the Food Stamp program.  New York has chosen to replace some of the federal programs with its own state programs.  These changes in the law occurred while New York City was in the midst of a vigorous program to reduce the number of people on public assistance.  Because New York State did not implement its own welfare reform plan until 1999, we cannot yet assess the full effects of welfare reform in New York City.  However, we can determine the early effect of these changes in law.
    To do so, we used the Current Population Survey to compare changes in the rate of receipt of the major public benefit programs for citizens and noncitizens in New York City, between the years 1994-95 and 1997-98.  The programs examined include public assistance/welfare, SSI, Medicaid, Food Stamps, and subsidized housing.  We find that in general, noncitizens do not show greater rates of decline in benefit receipt than citizens.  The exception is the Food Stamp program, where the percentage-point decline for noncitizens was almost three times as high as the decline for citizens, 3.7 as compared to 1.3 percentage points.
    Our results, while consistent with the generally pro-immigrant stance of New York public policy, appear to differ from national studies that find a "chilling effect" of welfare reform on noncitizens.  However, our findings are consistent with a recent paper by Borjas that argues that the national pattern of greater reductions among noncitizens is explained mainly by the results in California.  Together, these studies indicate that one consequence of welfare reform is that paarticular policy choices at the state and local level have become even more important than before in determining immigrant access to public benefit programs.
    We do find some intriguing differences when we measure household public assistance receipt by individuals rather than households, classifying individuals by their own citizenship rather than that of their household head.  In that case, public assistance receipt declines more among citizens than noncitizens, and Medicaid receipt increases for noncitizens while remaining constant for citizens.  These results may reflect "mixed-citizenship" households -- for example, noncitizen parents with citizen children born in this country -- or differences in household size between immigrants and citizens.
    We expected that tougher administrative hurdles to qualifying for public assistance would potentially have a greater impact on noncitizens than citizens, due to greater language barriers and less ability to understand the "system."  What we found would seem to support this hypothesis, in that public assistance receipt dropped by almost 10 percentage points for Hispanics, but remained essentially unchanged for blacks.  However, when we divided Hispanics into Puerto Ricans, non-Puerto Rican citizens, and noncitizens (see Table 3), we found that the Hispanic-black difference was not particularly a reflection of citizenship status, in that the decrease in public assistance receipt was twice as great for Puerto Ricans, who of course are citizens, as for other Hispanics, whether citizen or not.  Since language barriers would presumably be lower among Puerto Ricans than among other Hispanics, this result does not provide strong support for the importance of language barriers in cutting noncitizens off of public assistance.  Overall, the black-Hispanic differentials we find may be characterized as "gap-closing."  Hispanic rates of receipt of public assistance were greater than rates for other groups in 1994-95, but the rates had been equalized by 1997-98.  This helps to explain why the drop was particularly sharp for Puerto Ricans, since they had the highest rates of receipt among our three groups of Hispanics prior to welfare reform.

References

Borjas, George. 1999. Immigration and the Food Stamp Program. Joint Center for Poverty Research Working Paper No. 121, Northwestern University (Sept).

Borjas, George. 2000. Welfare Reform and Immigration. Xerox, Harvard University (June).

Chernick, Howard, and Andrew Reschovsky.  2000.  "Devolution and the Fiscal Health of Cities," mimeo. November.

Chernick, Howard, and Cordelia Reimers.  2000.  "Welfare Reform and New York City’s Low-Income Population."  Unpublished paper, Hunter College, November 8.

Figlio, David N., and James P. Ziliak.  1999.  "Welfare Reform, the Business Cycle, and the Decline in AFDC Caseloads." in Sheldon H. Danziger, ed., Economic Conditions and Welfare Reform.  Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, pp. 17-48.
 
Fix, Michael and Jeffrey S. Passel. 1999. Trends in Noncitizens' and Citizens' Use of Public Benefits Following Welfare Reform: 1994-97. Washington, DC: Urban Institute (March).

Klawitter, Marieka, Robert D. Plotnick, and Mark Evan Edwards. 2000.  "Determinants of Initial Entry onto Welfare by Young Women," Journal of Policy Analysis and Management, 19 (4), Summer. Pp. 527-546.

Macurdy, Thomas, David Mancuso, and Margaret O’Brien-Strain.  2000.  "The Rise and Fall of California’s Welfare Caseload: Types and Regions, 1980-1999."  Public Policy Institute of California.

Passel, Jeffrey.  1996.  "Problem with March 1994 and 1995 CPS Weighting." Memorandum, Urban Institute, November 12.

Sengupta, Somini.  2000.  "At One Center, A Study in Welfare Cuts."  New York Times, June 27.

U.S. Bureau of the Census.  2000.  "Money Income in the U.S., 1999."  Current Population Reports, Series P60-209.  September.

Wallace, Geoffrey and Rebecca M. Blank. 1999. "What Goes Up Must Come Down?", in Sheldon H. Danziger, ed., Economic Conditions and Welfare Reform.  Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, pp. 49-90.
 
Welfare Law Center, November 2000.  Available at www.welfarelaw.org