 |
Immigration: The Demographic and Economic Facts6. Receipt of Welfare and Other Government ExpendituresPublished by the Cato Institute and the National Immigration Forum
This chapter and the next bring data to bear on whether
immigrants are a burden on, or a benefit to, natives by way of
the public coffers. This chapter estimates the outflows of
expenditures for public services used by immigrants; Chapter 7
then estimates the inflows of taxes from immigrants and finally
considers the net balance between inflows and outflows.
The available data that bear upon the subject fall into
three categories: (a) crude and partial, (b) somewhat aged, and
(c) Canadian. None of the three gives a completely satisfactory
answer in itself. But the conclusions reached with all three
types of data are similar, so there is considerable assurance
that the overall conclusion is reliable. If one prefers not to
rely on the crude and partial data, which embody some very rough
assumptions, then one is left with the 1975 data, and there are
no other data invalidating the 1975 survey.
The Crude and Partial Data
The appropriate mode of analysis of the effect of immigrants
on the public purse is a present-value assessment based on the
lifetime experience of the relevant cohorts of immigrants. This
sort of analysis examines each of the past cohorts as a separate
observation. If the analysis is to be meaningful, the cohorts
must be much the same from period to period relative to natives;
this requirement was met reasonably well in the United States
data through the 1970s. Unfortunately, there are now no suitable
data available for life-cycle analysis. Therefore, various
researchers have recently attempted to make do with data on the
various entry cohorts, mostly considering them lumped together.
There is also interest in the partial calculation of the
flows of funds to and from states and local government alone,
leaving out the flows with respect to the federal government.
Such a partial analysis certainly is a legitimate interest for
the nonfederal governments, because it measures the effects on
their own coffers; a negative flow to them, even if there is a
positive flow to the federal government, certainly suggests an
inquiry into the controlling federal policies, though from the
point of view of the public as a whole, a negative effect on
nonfederal governments alone would not constitute sound grounds
to reduce immigration.
This section pulls together what may be learned from this
body of work based on messy recent data. (In the appendix there
is also discussion of some widely promulgated but entirely flawed
papers that purport to bear on this subject.)
The bare facts are particularly illuminating with respect to
this topic. There is wide belief in these two assertions about
current immigration: (1) Federal expenditures on immigrants have
been increasing. (2) Government expenditures on recent immigrants
are greater than expenditures on natives. Both of these
assertions are false. Expenditures considered to be "welfare" in
discussions of immigration (though there are many other expensive
means-tested programs for which data on immigrants and natives
have never been shown) now total about $404 per person annually
for immigrants, and about $260 for natives in about 1992. But
these sums are a drop in the bucket of total government social
outlays on immigrants and natives. As we shall see, the relevant
total is roughly $3,800 for natives; no meaningful total can be
computed for immigrants because of differing lengths of time in
the country. These magnitudes invalidate the two common beliefs
noted above.
Estimating Narrowly Defined Welfare Expenditures
From census and administrative records, Rebecca Clark (1994)
calculated 1990s' expenditures for immigrants and natives on Aid
to Families with Dependent Children (AFDC), food stamps,
Supplemental Security Income (SSI), and General Assistance.
Foreign-born persons taken altogether have perhaps a 10-20
percent higher probability (as distinct from dollar amounts) of
obtaining these goods and services than do natives. From her data
I estimate that federal expenditures total an average of $404 per
year per immigrant, while the average native receives $260 (in
about 1992). For those who wish more detail, there is a brief
appendix at the end of this chapter.
Immigrants who arrived between 1970 and 1990 average
receipts of slightly more than those who arrived before 1970. But
this could well be the result of age and time in country rather
than any differences in types of persons in the two cohorts
(computed from Clark 1994, Table A2).
Nonrefugees who enter legally through quotas are not
permitted to receive public welfare assistance for three years.
Hence, they cannot have a high rate of use of these programs.
Illegals are often afraid to seek such assistance. Refugees,
however, are entitled to such assistance immediately upon entry,
which (together with their needy circumstances) explains a much
higher rate of use of welfare soon after arrival than for
natives.
There is some slight evidence in Clark's data (Table B2)
that post-1970 entrants use more of these welfare programs than
do pre-1970 entrants (but lesser use by 1987 to 1990 entrants
than by 1970 to 1986 entrants; recent entrants require some time
to learn about such programs).
In earlier decades, too, there was greater use of welfare by
immigrants than by natives. As of 1975, the Census Bureau's
authoritative Survey of Income and Education (Simon, 1981; 1984;
1989, Chapter 5) showed that immigrants had higher welfare
receipts than natives. And it was also true at the turn of the
century, according to the report of the 1911 immigration
commission (cited without specific quote by Wall Street Journal,
April 25, 1995, 1).
Among foreign-born persons aged 65 years and over, a
meaningfully greater--and growing--proportion receive welfare
than among natives. This is the result of some immigrants having
arrived too late to qualify for Social Security benefits; the
welfare payments are a substitute for Social Security.
These data on narrowly defined welfare usage do not accord
with the common belief that more recent immigrants are
substantially different from previous immigrants with respect to
welfare usage.
Use by Illegal Aliens
Illegal aliens usually do not obtain public services lest
they be detected and expelled by authorities. The INS-Westat
study of formerly illegal aliens legalized in 1986 under the IRCA
program (U. S. Department of Justice 1992) shows substantially
lower rate of use of welfare among this group than among natives.
But because interviews are the source of the immigrant data,
while official records are the source of the native data, it is
quite probable that the rates of use by immigrants are
understated.
Another survey (DaVanzo et al. 1994, 46) of an undocumented
immigrant group probably as needy as any--Salvadorans--shows much
higher rates of use: AFDC, 14 percent; food stamps, 22 percent;
WIC (women, infants, and children), 33 percent; unemployment
compensation, 8 percent; worker's compensation, 4 percent. For
the undocumented population as a whole, however, in which half or
more of the recent cohorts are more middle-class immigrants who
are visa-overstayers rather than border-jumpers, it seems
reasonable to expect perhaps half these rates.
For reasons to be discussed below when we return to the
subject of undocumented aliens, there is no need for us to sort
out these conflicting assessments of narrowly defined welfare
use.
Relevant Government Expenditures Other Than Welfare
The data on the four federal "welfare" programs discussed
above do not include most government payments to the elderly or
expenditures for local public schooling. We now proceed to
include them so as to assess total government expenditures on
various cohorts of immigrants and natives.
Calculations about Social Security and Medicare, by far the
most expensive government transfer programs, pertain more to
natives than to immigrants. This is because immigrants typically
arrive when they are young and healthy, and because older recent
immigrants do not qualify for Social Security. And expenditures
on long-time immigrant residents must be thought of differently
than recent immigrants, as will be explained below.
Expenditures on immigrants for Social Security and Medicare
are particularly difficult to estimate because the payments
differ very much among age groups. And the sizes of the various
age groups of foreign-born residents differ greatly because of
the deaths of older immigrants and the increasing rates of
immigration in recent years. Nevertheless, I shall make rough
estimates for arrivals since about 1970. Then I will explain why
data for aged immigrants who arrived earlier are not relevant
here. No total calculations will be made from these current
incomplete data for any groups of immigrants, however, because
the totals would be difficult to interpret and inevitably
controversial.
Working from the Statistical Abstract of the United States,
total federal expenditures of $305 billion in 1992 for Social
Security and $133 billion for Medicare indicate (by dividing by
the total population) expenditures per native of $1,305 and $566,
respectively, for the two programs. The 1975 Survey of Income and
Education (SIE) data (discussed below) suggest that the average
receipt per immigrant who arrived within the past 25 years was
(and probably still is) less than a fifth of the average
expenditure per native--say $261 and $113 for 1992 for the two
programs, respectively, for argument. (Some allowance for the
public support of the recently arrived immigrant aged is embodied
in the relatively heavy SSI payments that substitute for Social
Security.)
Schooling costs estimated by Ms. Clark imply $536 per capita
for immigrants, and $923 per capita for natives. The expenditures
are lower for the immigrant population because the proportion of
children among the total immigrant population is smaller than
among the total native population, according to her data
(telephone conversation). (Separate data for recent cohorts
probably would show higher immigrant expenditures than the $552
estimated for all; see age distribution of immigrants in Figure
3.1.) For perspective, the dollar gap in schooling between
natives and immigrants is several times as large as the gap in
the public welfare discussed above, running the other way.
Regarding unemployment compensation, we can safely assume
similar expenditures of $138 per capita for immigrants and
natives, based on experience with unemployment compensation in
the 1970s (Statistical Abstract); the possibility of important
error is small here. For Medicaid it is reasonable to assume
higher expenditures for immigrants than for natives in the same
proportion as for the welfare programs above; both programs must
reflect immigrants being somewhat poorer on average than natives.
Federal and state Medicaid expenditures are about $90 billion and
$70 billion, respectively, in the early 1990s, so expenditures
per person are about $627 for natives and $752 for immigrants.
(Contrary to the assumption above that immigrants use more
of welfare programs than do natives, Clark [1994] notes that a
careful survey of food stamp use by the Food and Nutrition
Service found that immigrants use less food stamps than do
natives. She suggests that the FNS data are better than the
census data, on which other welfare estimates are based, and that
"[t]he findings from the food stamp quality control data suggest
that if there is bias in the estimates for the other public
assistance programs, the census-based estimates are more likely
to overstate welfare costs associated with immigrants than to
understate them" [p. 18].)
Total Government Expenditures Properly Defined
Now we can add together all transfer payments plus schooling
costs. This is the appropriate measure of government expenditures
to use in any assessment of the costs and benefits of
immigration. The expenditures on natives per capita are much
greater than the expenditures on welfare alone for either natives
or immigrants.
Accurate but Aged Census Bureau Survey Data
The only source of data for the United States that is
entirely competent to produce a satisfactory overall answer from
a conceptual point of view--the massive and authoritative 1975
SIE--is two decades old. No U.S. survey since then has gathered
data in a fashion that will enable us to make a satisfactory
lifetime estimate of the tax-and-transfer effects of immigrants,
and only such a lifetime assessment is a proper basis for an
assessment of this matter. Therefore, recent auxiliary data and
studies that bear upon projecting those earlier estimates onto
the present situation also will be examined.
First will come the data and results from the SIE. Next, I
will compare Akbari's analyses of recent Canadian surveys to the
SIE results. Then come recent partial studies of welfare use and
taxes paid. Last to be considered is whether the "quality" levels
of immigrants have changed during recent decades in such manner
as to make the earlier assessment inappropriate to the present
situation.
The Basic SIE Study
In 1976 the Census Bureau surveyed 156,000 U.S. households
(including about 15,000 mostly legal immigrant families) to learn
about 1975 family income and welfare services. From this
sample--by far the best for its purpose that has ever been
assembled anywhere--I constructed a picture of lifetime economic
behavior by assuming that the information on immigrants who had
been here (say) 2 years as of 1975 describes the representative
immigrant family after 2 years, those here 10 years in 1975 stand
for the 10th year in the United States of a representative
family, and so on.
Outlays for Other Than Old-Age Benefits
Data on welfare and Supplemental Security payments,
unemployment compensation, AFDC, and food stamps in 1975 are
found in columns 1-4 in Table 6.1. The average native-born U.S.
family received $498 from these programs in 1975. (The
calculations include families getting no assistance.) There is
not much variation among immigrant families that arrived between
1950 and 1974; the average was $548. There is not much difference
between natives and immigrants.
Providing school for immigrant children during an average
family's first five years in the United States cost slightly less
than the $859 spent for the average native family, because
immigrants tend to come before family completion. After that,
expenditures on immigrants were higher, rising from $1,068 to
$1,237 during the next 15 years. (The difference is not because
immigrants have many children, but because the average native
family is older, with a larger proportion of children no longer
in school.)
Expenditures on the Elderly
Retirement programs bulk much larger than any other
welfare-type costs for natives and in the economy generally. The
1975 data from the SIE showed that native U.S. families on
average received $735 for Social Security, $167 for Medicare, and
$20 for Medicaid, a total of $922. Immigrant families received a
total of $92 during the first five years in the United States,
$227 the second five years, $435 the third five years, and $520
the fourth and fifth five-year periods. The difference was large.
Immigrants receive less Social Security and Medicare simply
because of their youthfulness. As with groups of immigrants
arriving in all countries throughout history, those coming to the
United States tend to be young, strong, and often unmarried.
Data on the Canadian Experience
Ather Akbari (1989; 1995) conducted two similar studies for
Canada, one using 1981 data and the other using 1991 data, from
two separate sources (census data for 1981 and Survey of Consumer
Finances for 1991). With respect to government expenditures on
immigrants and natives, the results of his study of 1981 data are
entirely consistent with my study for the United States using
1970s data, and his study of 1991 data shows no major differences
from the study of the 1981 data.
Trends in Welfare Use by Immigrants
If there have been substantial changes over the decades in
the propensity of immigrants to receive transfer payments, a
synthetic analysis considering the various cohorts as parts of
the same life cycle would be invalidated. Therefore, data on the
changes over time in welfare use by immigrants must be examined;
see Table 6.2.
To avoid confusion: Some have noted that, as Borjas says,
"the probability of participation in public assistance increases
the longer the immigrants reside in the United States" (Milken
Institute 1994, 22). But this observation is neither new nor
meaningful. Income and taxes typically increase with length of
residence in the United Stated, too--a natural function of age
and experience.
Also, please keep in mind that the proportions of persons
receiving welfare are not fundamental; it is the dollars received
that matter.
Table 6.2 shows that both native and foreign-born groups had
barely perceptible declines from 1979 to 1989 in proportions
receiving welfare. Those immigrants who entered 1980-1990 had a
slightly lower rate of receipt than those who entered 1970-1979,
giving no evidence of an increase (though it sometimes takes a
while before immigrants learn to use the welfare system).
Among foreign-born persons aged 65 years and over, a more
meaningful and growing proportion receive welfare than that among
natives. As noted above, this is due to many immigrants having
arrived too late to accumulate enough work time to earn Social
Security benefits; the welfare is a substitute for Social
Security. Clark sums up this section as follows:
Among immigrants, high rates of welfare use are limited to one
group of immigrants--those who entered as refugees--and one type
of welfare--SSI. For other types of welfare, immigrants who did
not enter as refugees are no more likely to use welfare than
natives. In fact, for food stamps, immigrants who are not
refugees or asylees are substantially less likely to use benefits
than natives (Clark 1994, 18).
A reminder, however: The subject of this section--welfare
narrowly defined--is relatively unimportant in the overall
picture of government expenditures on immigrants and natives, as
the first section showed
Expenditures for Illegal Aliens
Earlier in the chapter, divergent estimates were shown for
the extent of use of narrowly defined welfare services by
undocumented aliens. But even if use is perhaps double that of
legal immigrants instead of less than theirs--most unlikely
indeed, given that few illegals would qualify for SSI, a key item
for legal immigrants--it would not have a dominant effect on the
overall balance. An expenditure of $600 per year per person would
seem to be a high-side guesstimate.
Clark et al. (1995) estimated expenditures for 1994 federal
incarcerations, schooling, and providing emergency medical
services in connection with the populations of undocumented
aliens in the seven states in which reside about 86 percent (2.9
million of the 3.4 million) of the total population of illegal
aliens. The total estimates are as follows: incarceration, $474
million; Medicaid, $445 million; schooling of children, $3.1
billion (executive summary). The schooling expenses are
considerably larger than for legal immigrants and exceed somewhat
the expenditures for natives--say $1,070 per year per immigrant
for schooling--and $320 per person for the total of incarceration
and emergency medical care. Social Security and Medicare must be
nearly zero because illegal aliens do not qualify.*
Adding these categories, on average, the total expenditure
for illegal aliens is $1,690 per person. This amount may be
compared with the total of perhaps $2,200 for legal immigrants.
The estimate for illegal aliens is about 45 percent of the $3,800
expenditures per capita on natives (and of course taxes paid by
illegals are lower, as shown above).
(Please notice that no allowance is made for incarceration
costs in above discussions of legal immigrants and natives, which
inflates the previously cited cost estimates for illegals
relative to others. The incarceration costs are included
nevertheless, because some persons may wish to have this
information.)
Immigrants versus Natives with Similar Background Characteristics
Behavioral scientists habitually do not stop with
considering aggregate behavior; nor do they typically focus upon
the variables relevant for policy issues. Instead, they tend to
ask about the effect of a given characteristic when as many other
characteristics as possible are held constant econometrically.
This implies inquiry into the effect of whether or not a given
person is an immigrant while as many other life characteristics
as possible are being "controlled." Many such studies have been
carried out, and I report them here, even though--I repeat--they
are not relevant to policy issues, simply because to leave them
out would be to seem incomplete.
Blau (1984) inquired into the propensities of immigrants and
natives in the United States with otherwise-similar
characteristics to receive welfare (other than Social Security,
not the amounts received) and answered that there are few
important differences between the groups. Baker and Benjamin
(1993) obtained similar results for Canada. Tienda and Jensen
(1986; 1988) and Jensen (1988) also find for the United States,
as Baker and Benjamin (1993) find for Canada, that the
probability of receipt of public assistance, and the amount of
receipts, is no greater for immigrants than for natives when a
variety of demographic characteristics is held constant.
Using different (and, we think, more appropriate) methods,
Akbari and I (forthcoming) tackle much the same question and
arrive at much the same answer about amounts of receipts. Using
data for both Canada and the United States, and spanning the
years 1975 to 1990, we find a common pattern between the
countries and over time, as follows: (1) When other
characteristics are held constant, being an immigrant does not
affect use of welfare services. (2) The overwhelming determinants
of welfare services (other than for the aged) are family
structure and education. (3) These findings are invariant over
time and between the United States and Canada.
Family structure--and especially the presence of
female-headed families with two or more children--is the most
important determinant of the rate of use of transfer payments
(aside from Social Security, Medicare, and other retirement
transfers).
The fact of being an immigrant has little or no effect on
welfare receipt when the other relevant variables are held
constant.
Akbari and I also answer another question: Even if there is
no difference between tightly defined demographic classes of
immigrants and natives, are immigrants more likely than natives
to be found in the demographic groups that receive more welfare?
The result we find: Immigrant families are represented roughly
proportionally (relative to natives) in the high-welfare-use
categories.
The fact that the results are almost indistinguishable for
the United States and Canada, despite the differences in decades,
despite the difference in benefits paid for children and the
difference in systems of selection of immigrants, suggests that
there is an underlying general principle of human behavior at
work that governs migration across time and space: the persons
who are the most economically productive are the likeliest to
move.
Differences among Groups
Behavioral scientists also tend to distinguish among the
behavior of ethnic groups, and among immigrants by country of
origin. But the distinctions are seldom (if ever) relevant for
policy purposes. One may also find differences between natives
and either (a) all immigrants taken together, or (b) some groups
of immigrants, within particular states, as is the case for California.
Such data may be relevant to the situations of particular states,
especially with respect to their financial arrangements with the
federal government. It would be misleading, however, to use such
data as a representative sample for the United States as a whole,
because the flows of immigrants to particular states are often not
representative of the overall flow of immigrants.
|
