You are on page 1of 47

Interstate Migration of the US Poverty Population: Immigration "Pushes" and Welfare Magnet "Pulls" Author(s): William H.

Frey, Kao-Lee Liaw, Yu Xie, Marcia J. Carlson Source: Population and Environment, Vol. 17, No. 6 (Jul., 1996), pp. 491-536 Published by: Springer Stable URL: http://www.jstor.org/stable/27503494 . Accessed: 12/09/2011 09:21
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Population and Environment.

http://www.jstor.org

Interstate Migration of the US Poverty Population: Immigration "Pushes" and Welfare Magnet "Pulls"
William H. Frey The University of Michigan
Kao-Lee Liaw McMaster University

YuXie The University of Michigan


Marcia J. Carlson

The University of Michigan

This study evaluates


ing the 1985-90 period

the social and demographic


based on an analysis of

structure of poverty migration dur


recent census data/Particular atten

tion is given to the roles of two policy-relevant factors that are proposed to be linked to poverty migration. The first of these is the role of immigration from abroad
and its effect on the net out-migration of longer-term residents with below-poverty

incomes, from States receiving the highest volume of immigrants. Such a response, it is argued, could result from job competition or other economic and social costs associated with immigration. The second involves the poverty population "magnet" effect associated with State welfare benefits (AFDC and Food Stamp payments) which has come under renewed scrutiny in light of the impending reform of the
This research on Pov is supported of Wisconsin Institute for Research by the University data for The migration erty Small Grants Program and by NICHD grant No. ROI HD29725. at the Population this paper were from 1990 Studies Center, University of Michigan prepared US Census files. The authors acknowledge assistance, Cathy Sun for computer programming and Ron Lue-Sang for preparing maps and graphics. to Dr. Frey, Population Please address of Studies Center, correspondence University 1225 South University Ann Arbor, Ml 48104-2590. Avenue, Michigan, A Journal of Interdisciplinary and Environment: Population Volume 17, Number 6, July 1996 ? 1996 Human Sciences 491 Press, Inc. Studies

492 AND ENVIRONMENT POPULATION

federal welfare program. The impact of both of these factors on interstate poverty migration is evaluated in a broader context that takes cognizance of other socio demographic subgroups, and State-level attributes that are known to be relevant in internal migration. This research employs an exceptionally rich data explaining base of aggregate migration flows, specially tabulated from the full migration sam ple of the 1990 US census (based on the "residence 5 years ago" question). It also employs an analysis technique, the nested logit model, which identifies separately the "push" and "pull" effects of immigration, welfare benefits, and other State attri butes on the migration process. Our findings are fairly clear. The high volume of immigration to selected US States does affect a selective out-migration of the poverty population, which is stronger for whites, Blacks and other non-Asian minorities as well as the least-educated. These results are consistent with arguments that internal migrants are responding to labor market competition from similarly educated immi
grants. "push" we Moreover, than a rather found reduced that the "pull." of impact In contrast, occurs primarily immigration exert State welfare benefits as only a

as effects on the interstate migration of the poverty population?either "pulls" or "pushes," although some demographic segments of that population are more prone to respond than others. Inaddition to these findings, our results reveal the strong impact that a State's racial and ethnic composition exerts in both retaining and attracting migrants of like race and ethnic groups. This suggests the potential for a greater cross-state division in the US poverty population, by race and ethnic status. Data Used: 1990 US census tabulations of full migration ("residence 5 years minimal
ago") sample. Note: Detailed 1990 census statistics on migration of the poverty and

William H. Frey "Im nonpoverty populations for individual states can be found in: migration and InternalMigration for US States: 1990 Census Findings by Poverty Status and Race," Population Studies Center Research Report No. 94-320.

INTRODUCTION
This research evaluates the erty migration during the 1985-90 Particular attention populations. factors that have been linked to 1. Immigration from abroad. tion from abroad and its effect Americans from States social and demographic structure of pov period and its impact on States' poverty is given to the roles of two policy-relevant in recent debates: inter-State migration The first of these is the role of immigra on the net out-migration of native-born

the highest volume of immigrants (e.g., receiving Both descriptive statistics and analyses California). migration the 1990 census 1995a; 1995b; The Washington (Frey, 1994b; 1993) and of the 1980 census Post, (Walker, Ellis & Barff, 1992; Filer, & Hunter, low 1993) suggest that it is the least educated, 1992; White income and poverty residents who are leading the way out, apparently, in to competition from low-skilled While response employment immigrants. based admissions have been given higher priority in the Immigration Act of especially based on

493
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

still constitute only a small part of the overall im admissions flow and there is the possibility of further adjustment (Fix & Pas migration im of the most heavily 1993). Moreover, sel, 1994; Martin, governors to call for more drastic State and federal-level States are beginning pacted levels, legislation to reduce the incentives for continued high immigration (The New York Times, 1993; 1994). 2. Welfare benefits. The second policy-relevant factor to be empha sized involves the poverty population effect that has often been "magnet" linked to a State's welfare with those associated payments, particularly 1990, these

AFDC (Aid to Familieswith Dependent Children). This linkage has come


under renewed benefits and be scrutiny in States with generous welfare cause of the impending reform of the federal welfare program. Consider able research has examined the linkage between welfare benefits and mi 1981; Gr?mlich gration (Cebula, 1979; Cebula & Belton, 1994; Southwick, & Laren, 1984; Blank, 1988; Clark, 1991; Peterson & Rom, 1990; Voss, Corbett & Randell, & Hartman, 1992; Cushing, 1993; Hanson 1994; Walker, 1994; Schr?m & Krueger, 1994; see also review inMoffitt, 1992). recent of these, employing The most limited sample survey data or with How coverage, geographic suggest that this linkage is relatively modest. in these studies were too limited to make ever, the kinds of data used inferences about the aggregate redistribution impacts for States associated with migration flows of detailed sociodemographic that are subpopulations known to respond differently to State area-level and "pulls" in "pushes" addition to the effects of welfare benefits. These limitations were, in part, a of the unavailability of the full census sample migration ma consequence is required for such an analysis. The present study employs trix, which a data base specially tabulated from the 1990 US census. such This study is divided into two parts. The first section reviews detailed census findings on the patterns of State gains and losses in poverty 1990 The questions here are: Which addressed States gain largest populations. net numbers of poverty migrants via internal migration? Which States lose largest net numbers of poverty migrants? Are these State patterns of net associated with the policy from factors?recent migration immigration abroad, and high welfare benefits? The second section of the analysis goes beyond a description of State net migration of the migration patterns toward an examination process. two separate parts of Focussing on the most mobile age groups, it evaluates the migration affect the magnitude of State out-mi process: factors which affect the drawing power of destination gration flows, and factors which States. For each part of the migration the relative process, we evaluate of State immigration importance levels, and State welfare benefits, vis-?-vis

494 POPULATIONAND ENVIRONMENT

in order to determine if these other factors known to influence migration, streams. factors exert separate "pushes" or "pulls" on inter-State migration our analysis technique, Moreover the nested allows us to ex logit model, to impact upon key amine the extent to which these factors are more likely US nativity?in affect race, gender, education, population subgroups?by inter-State migration of the US poverty population. Are welfare benefits ing most streams of poor women? Are high likely to exert a "pull" on migration to be "pushed" away from high immigration school dropouts most likely are among those addressed States? These questions in our examination of the migration process. our findings show that immigration to a few port-of-entry Overall, States exerts a far larger impact on the internal migration of the nation's as a "push," than do State welfare poverty population?acting benefits, of the migration pro acting as "pulls." This finding, from our examination cess, is consistent with net out-migration large reflects some response patterns of State net losses and gains. The of poverty migrants from High Immigration States to immigration, of other influences on independent is especially evident among whites, process. This response In contrast, our results show that State wel the least educated. exert only minimal ef (AFDC and Food Stamps, combined) as inter-State migration of the US poverty population?either observed

although some demographic segments of that popula tion are more prone to respond than others. These patterns are discussed more fully in the second part of the analysis.

the migration Blacks, and fare benefits fects on the "pulls" or "pushes,"

MIGRATION DATA
The migration data for this study are drawn exclusively from tabula tions of the fixed interval 5-year migration question of the 1990 US decen nial census. This is the only nation-wide data set available that is appropri ate for examining stream processes with associated aggregate migration detailed regions and population subgroups. The main State origin-to-desti nation matrix for 1985-90 migration stream and non-migrant populations is by age (5-year age groups), gender disaggregated tion attainment (less than high school, high school status (non-Hispanic race-ethnic college graduate), American Indians),1 nativity (US-born, Hispanics,
1 Because cross-classification the census that were made tabulations available of the separate race and Hispanic variables,

(males, females), educa graduate, some college, Whites, Blacks, Asians, and pov foreign-born),
a for this study did not provide a pro origin, we developed

495
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

income) where most of income, above poverty erty status (below poverty on the poverty population. the present study will focus exclusively Despite and the obvious strengths of using census data for this aggregate migration a well-known of pop is the unavailability weakness redistribution analysis, at the beginning of the 1985-90 period, since only ulation characteristics This time are available. could be identified at census that characteristics in defined for the poverty population, limitation is particularly noteworthy the poverty popula the 1990 census on the basis of 1989 income. Hence, that may the poverty population tion as defined here only approximates have existed at any point over the 1985-90 period.

OF POVERTYMIGRATION STATEPATTERNS
the gain and loss patterns of recent poverty migration, Before discussing are particularly relevant to our policy focus first on those States which Benefit States (see Table factors: High Immigration States and High Welfare 1). The High Immigration States are identified elsewhere (Frey, 1994a) and

we

include the six States with greatest numeric immigration gains. They include: (Note: California, New York, Texas, New Jersey, Illinois and Massachusetts. Florida also attracts a large number of immigrants, our earlier typol While status (Frey, 1994a) classes it as a High Internal ogy of States by migration rather than immigration domi internal migration State because Migration nates its population Benefit States represent those gain.) The High Welfare with greatest combined AFDC and Food Stamp benefits, when adjusted for and State B for definition State cost-of-living differences (see Appendix and New two High include values). They Immigration States, California as well as Vermont, Wisconsin, Washington, Minnesota, Oregon, Utah, York, Kansas and Rhode Island. (Alaska and Hawaii are excluded because compa rable welfare benefit information was unavailable.) The data in Table 1 show that the observed net internal migration rates

to collapse all into a single one. This was done by, first, classifying cedure these two variables Indian (includ of the races, Black, Asian Pacific Islanders), and American persons (including to their actual reported races. Persons reporting a Hispanic ing Eskimos and Aleuts) according esti as Hispanics. white The non-Hispanic classed was, therefore, origin were population as Hispanics on the Hispanic identified themselves mated persons who origin by subtracting or as "other" on the race as either white themselves item, from the total of persons identifying are: Blacks, Asians, American item. The resulting, mutually Indians, His exclusive, categories in whites This procedure tends to understate whites and non-Hispanic non-Hispanic panics, States where more than a minimal this paper will refer to non-Hispanic number of Hispanics as simply whites are Black "whites." or Asian. For convenience,

C co o C w ?- CO eu ? ?a do

<NN. O O m O m li^oO fN K(N O r- iA r- o m CO (NOnV" ,?" "TOS ?*oto <o t

NrnnrnNOOv?O OvDf-OmneOK f? rorOrro vOOSt? ro co" lC cn O 0> r-^ cn t? CM r- CN

I I I I I

?2

C o 3 o. o Q > O a.

un C c .2

c J2

O D CL CL

f E
? *?

s -o S<

o 2

r>* ro o m o m NNNr-OO oo cn -^ rs^ o^ <? cn rCcc?ro o o

^?rSC^C/lr-00C/>N 0^rtn?vr\i\LncN in as m *? os m en ro r-N in CS rf N N vu r- CN r- I

C O
as

l\ cn rv cn

?2&

r^ en vu ^r

co m o co *t m cn tt "3*co ^r c? cn in ro" t-^

^ I I

romcNcNroomrs. r-or^cacNr-^ros \j3rococNvOmmo ro o cn TT in rC cd" co" ro m cn ro r c^ o

? %
52 ^ c O?C? i? in ^

ra

E * ? c
c sO .2 "S D?

> O a. c o Z

hv cq r- *r r>s cn ^T fN r-^ CN CN ?

I I I I I

OtNls:t--pv?rf^ rs ?-' ^' o n

< g
mV o <-o

III

U? ??
o a. c U

o?

in vO ro ??3| ro ro r^ ^ m' ci in o

o a. a; > o a. c o Z a; > o

I l IT I I I

riri <-;q oq q sO^vOrOOvO?cN

in o^ Q oo

03 ?S

<S y? ? o? zi r Mr(^ O VU rt ro f-^ CN i-^ CN rvincov?^ro^aq OOr-O?? ?? *? t? - s? xc? 5 .2P r> * := Q. m "EX

*
c o -^

o 2h oo

D? O

g? .2 ? 'M .z:

? ?
<-? tt t-^ r?. vo TT ^' N ^ N rf o? r^fNinr?^Ttrnis

i"* o

?8

Si m c v. a> ** c o c c o ?JD E ?

<Un? no ?

< s? u*

-S
PO ?o

??g
?S to Z O !< or U ?-? r u X o c c D "> *" u. LU Z D? O -C

<?i <9 Q> is> OS ??=2 T3 ,- <?o Ci < c c O ? *? *?? "Sb o *T3 c

0-2$. ou

?S !Z . c o

S c c S E S-S co ? t/> ^ > 3 > 2O .b UZpZz:2l>>>JOD^?i

0) c O

?^ s ^

?? ifl S S ?i ^J ? o S iS co ^? ^ D? * "^ C?* # # ? * w

497
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

in most of these States are for the poverty and nonpoverty populations That is, for four of the High Immigration consistent with prior expectations. are for the poverty population States, rates of net internal out-migration (For New York, both higher than those for the non-poverty population. rates are relatively high though slightly higher for the nonpoverty popula Benefit States (which tion.) By the same token, each of the High Welfare are not also High Immigration States) show higher rates of net internal in for the poverty population than for the nonpoverty migration population. In both the conventional these patterns counter wisdom. instances, The conventional in the migration literature pattern, well-established to be most on positive selective demo (Long, 1988), shows migration oc incomes, high educations, graphic characteristics?high higher-skilled are it is these segments of the labor force which cupations. This is because most to national income and employment be responsive opportunities cause of their valued human capital attributes, and the more specialized nature of their occupations. The redistribution process, then, might be as a "circulation of elites" (Frey, 1979). The process typically characterized sees employment-gaining States as attracting higher rates of nonpoverty, than rates of poverty or lower-skilled migrants. college graduate migrants, the same token, migration-losing States show greater rates of outflow By here, States, differ 1994b), and benefit effects Another among shown their and college The patterns nonpoverty, graduate migrants. for High Benefit Immigration States and for High Welfare from those conventional patterns shown for other States (Frey,

suggests that there may be unique immigration and welfare on the migration of the poverty population. noteworthy aspect of the redistribution pattern for High Immi inter gration States, is the net impact of immigration from abroad vis-?-vis nal migration on changes in the overall poverty population for the State. As Table 1 shows, each of the High Immigration States shows a negative net over the 1985-90 internal migration of the poverty population period. this net loss among internal poverty migrants ismore than com However, via immigration from abroad. pensated by gains in the poverty population in Table 1.) In contrast, the migration (See right-hand panel dynamics within the High Welfare Benefit States is influenced much more strongly by the internal migration process, so that the poverty gains attributed to inter over the 1985-90 period either dominate or largely contrib nal migration ute to the overall poverty migration gains for the State. The relative of immigration and internal migration for poverty population gains at the conclusion as it relates discussed of this paper, especially the main focus of this investigation Immigration States. However, and determinants of internal migration of the patterns evaluating across States. population impact will be to High involves poverty

498 AND ENVIRONMENT POPULATION

Patterns of Migration Gains and Losses


on poverty migration the above discussion focussed patterns for the High States and High Welfare Benefit Immigration specifically we will now turn to an examination of overall State patterns of gains States, and losses with respect to internal migration of the poverty population. Are the greatest poverty migrant gaining States also the High Welfare Benefit States? Are the greatest poverty migration States also the High Immi losing sub gration States? How do these patterns differ across key demographic are addressed here. These questions groups? in Table 2 lists the greatest net migration gaining and The first column over the 1985-90 period. The States, among the poverty population, losing While is that the largest net out-migra noticeable aspect of these patterns are associated tions of poverty populations with five High Immigration which all other States in the magnitudes dominate of their poverty States, to -48,000 from -92,000 for New York, Illinois, out-migrations?ranging New Jersey and California. Texas, In contrast, the greatest poverty net in-migration States are not domi nated by those with high welfare benefits. Rather, Florida leads the list by a is also the case for the nonpoverty fairly wide margin, which population which is not shown here. Three High Welfare Benefit States, Washington, are among the ten top poverty gainers?all and Wisconsin, Oregon, higher in rank than they appear on the similar list for the nonpoverty population. are a mix among those States that gain in poverty population Nonetheless, with a relatively high growth in service industry employment opportunities retiree magnets (Florida, North Carolina, Georgia, Tennessee), (Florida, Ari Benefit States. An additional zona), as well as some of the High Welfare to nearby States of factor, suggested by this list, is a spillover movement and Arizona) (forWashington, poverty migrants leaving California Oregon or Illinois (forWisconsin). A comparison of the poverty gainers and losers with a corresponding list for the nonpoverty population (Appendix A) indicates one important both gainers and losers for the nonpoverty are more difference: population For example, apt to have growing and declining economies, respectively. located in the prosperous South Atlantic Virginia and Maryland, region, among the top ten nonpoverty magnets, State of Georgia next to ranks second from other States. Moreover, poverty populations States appear on the list of migration declining These include Michigan and Ohio population. rust belt, Oklahoma, from the oil patch region, appear namic and the economically dy in attracting non Florida additional economically losers for the nonpoverty from the deindustrializing and Iowa which witnessed most

Oi-vnrs.co.-GorN.oco

.O^OOOOr-r-sOtN ^ rn .? sO ir>" lo LO t/> <Nr- r- l i I III

I I 1 I

3 32
m ?2-0 t"-?.2 S ?

00

> oc-?" 1o S ?S-5 ? ? u S-B -c 'S S c ? u te c o u u < V) W)


o z g I g < z O m 3 .c g o ^ Ns?TTNM(N?-(NOCri

^?

*?3*

g3i |

,2 .2 :? "p

? J2 ?<

- ^ rO? 00 n m rn?-_ W 00 vO vO N c.? (*?(N (N P? t-* r." I I ! I i I i ! I I

o <
g < z 0*

S ?2 S c o m rt *= ro d> xz Tro>oOMNMfn<T\0^r ? V?lflrr-inCnOCOvOf aivOi-cnooioeo^ooooo .??CT?" C?" vO CNr-"r-? | rC rf t/?<N<N | | | | | |

u5?iz

zz<

? ?!o

? -1 ^
O
?r

vcnrnNtO^'fifnN n.inrN.T?-O^?NlOCOvJDCOCO r- O ?-vIfl N r- \fl (N fflk ^ M rirMcrTco'rCsOrnrT'cNiN \J^

?N CO ?s
I

O J5 hLU D ? Z < o

Z _

-* u

z 9 'S *? R- S ozir??>5< c .2 Su 8 is c : 1 B LU D ?

C?

J? ?o> 50 C k. c CD Z

<* 5

lois

S 2 ? g ^ S ^ * & s -r= O * 5 c * C > 3~ C ?% ?/i v? > = Z jU?USi^I Z

> S E ? g s

ti =

-_

ojtOmi-cniOWiNN ?jNOsqqiN^qm?N^

mmcovomv?vornOO N OJ in Ifl \? 00 \? c? in r- 00^ .? Por^rCcN'cn^OTfOCT?'vO

U
S J^? *?m? O-cji C-Q c r 2 ? ?^ *C rt?hj ?"C C^O<SZ^'<<D comm,? iNfMOfocrtcr? rtjfsOr-?-inmvov?\o^

I I I I I I I I

.2 St-?

2*

_ cocNCNr>.intNcrttN.f-co T? <t> vOL?*ir-Ov?r-0>VOO> .? (N^orneotCvucrfCtn I I 1 1 I I I I

0? IS CD O flj-?

00 if

1lfJiflJ.il Z

z ^ n r-' (N ^ m' von c? o? d

500 POPULATIONAND ENVIRONMENT

in farming during this period. Although downturns five High Immigration States also appear on this list?including the three top losers, New York, Texas and Illinois?the of nonpoverty is not com out-migration population from High Immigration States. ing primarily This relates to an important contrast between inter-State migration pat terns for the poverty population and the nonpoverty (see Map population from the poverty population is heavily focused on 1). The net out-migration a few origin States, dominated the High primarily by Immigration States. The number of net in-migration States is relatively large and much more diffuse. This represents a "push-oriented" In contrast, process. migration the net migration of nonpoverty tends to be more population "pull-ori in that the number of net in-migration States ismuch ented" smaller, and on the coastal States located mainly represent economically prosperous of the US. It is interesting to note that California parts represents an out State for poverty migrants, and an in-migration State for non migration to the fact that it is largely the migrants. This may be attributable and less-skilled segment of the California population that is com with in the labor market. The nonpoverty as peting immigrants population, with the college graduate population in (see Frey, 1995b) may be operating a somewhat different labor market where the effect of recent immigration poverty poverty rather than compete with their employment may actually complement op (White & Hunter, 1993). portunities The data in Table 2 also break down States gaining and losing poverty for whites and Blacks, States migrants by race and ethnicity. Particularly with are most represented by High Im greatest poverty net out-migration States. Among whites, the same five High States migration Immigration dominate as was the case with the total population. poverty out-migration, These States, along with Massachusetts, dominate most of losing States for poverty whites, much more so than they do for the nonpoverty white popu lation (see Appendix Illinois rivals New York as the A). Among Blacks, net exporter of poverty out-migrants, unlike the case with the non largest New and Texas, poverty Black population. Moreover, Jersey, California each on the list of greatest Black poverty net exporting States are not on the list for nonpoverty Blacks. The out-migration of nonpoverty comparable Blacks appears to be more accentuated in economically struggling States (Louisiana, Mississippi, Pennsylvania, Michigan). While there is some similarity in the largest net out-migration States for white and Black poverty migrants, this is less the case for their net in destination States (see Map 2). This reflects, in part, the different migration concentrations of the two races as well as their historic roots geographic 1987; Johnson & Roseman, 1990). Poverty Blacks, (Long, 1988; McHugh,

501
W. FREY, K.-L LIAW, Y. XIE, AND M. CARLSON

MAPI.

Net

Internal

Migration

Poverty Population

Non

Poverty Population

5Greatest Losing

5 Greatest Gaining

OtherLosing

K5|

Other Gaining

are more apt to relocate in the South Atlantic, especially to for example, one and North Carolina. For whites, Florida remains the number Georgia gainer followed by States that lie nearby California. While High Welfare as Benefit States do not dominate each list, they are more prominent for poverty whites and Blacks than they are for these races' non gainers

502 POPULATIONAND ENVIRONMENT

MAP

2.

Net

Internal

Migration

Poverty Whites

Poverty Blacks

5 Greatest Losing

5 Greatest Gaining

OtherLosing

RJ9

OtherGaining

and A). This is the case forWashington are and Minnesota Blacks, Wisconsin Oregon, and do not for the poverty population, among appear among the top ten for the non-poverty population. The greatest gaining and losing States for poverty Asians and His poverty (see Appendix counterparts among whites. Among the top six gaining States

503
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

MAP

3.

Net

Internal

Migration

Poverty Hispanics

Poverty Asians

5Greatest Losing OtherLosing

j_|

5Greatest Gaining

OtherGaining j5s?sj

in Table 2. However, these geographic patterns are panics are also shown not distinctly different from those of these groups7 nonpoverty counterparts in Hawaii, Is highest (see Appendix A). Asian poverty net out-migration same four States which dominate Asian Illinois and Texas?the New York, net out-migration ranks fourth rather than nonpoverty (although Hawaii

504 AND ENVIRONMENT POPULATION

New York, California and Texas constitute the largest first). For Hispanics, States. These same three States dominate the list poverty net out-migration of nonpoverty movement away from Cali Hispanic out-migrants, although fornia is less prominent among this group. Finally, nonpoverty Asians and are overwhelmingly attracted to single-destination States?Cali Hispanics fornia for the former, and Florida for the latter. These attractions appear to than traditional ethnic ties or estab have less to do with welfare benefits & Levin, lished chain migration 1993; patterns (Barringer, Gardner both Asian and Hispanic 1989; Frey, 1995a). Hence, poverty McHugh, patterns, across States, do not appear to reflect the influences of either benefits or immigration "pushes," as much as whites welfare and Blacks. show greatest out-migration from sev poor Asians and Hispanics Although eral of the High Immigration States, these patterns do not differ significantly from their nonpoverty is the some counterparts. One exception population re what higher out-migration for poor Hispanics from California, perhaps to increased employment with competition sponding immigrants.

State Attributes and Net Poverty Migration


to evaluate our two policy the relationships between way variables levels and State welfare (State immigration benefits) and internal for States is a multivariate poverty migration analysis approach. We present in Tables 3 and 4 where a State's poverty net migration such analyses level for 1985-90 is regressed on a series of State attributes, including our two as measured variables from abroad, (level) policy by: immigration and the combined AFDC and Food Stamp benefits level (average 1985-90; of annual 1985 and 1988 values, adjusted for State cost of living varia in the analyses included tions). The other State attributes represent eco are known to affect migration in nomic factors which (percent of change in service em of change 1985-89; percent manufacturing employment, Another with State cost per capita income, 1985-1989, the violent crime rate, rate, 1985), adjustments; unemployment over a geographic of States 1985-89, regional classification for the Northeast variables the South the Midwest region, region, the Mountain and the Pacific division, where division division, are not included the South, which in the South Atlantic division, the omitted category) and the log of the State's 1985 population represent size (controlling for scale). Each of the equations in Tables 3 and 4 pertain to net migration for a specific demographic subgroup. This permits us to evaluate the significance of immigration vis-?-vis and welfare benefits in affecting State internal migration other State attributes for different de mographic categories of the poverty population. ployment, of living averaged (dummy Atlantic parts of 1985-89; average

00 <

00 <
I I I I

4?

0)

O0 <

oo <

(N vD ?N CN(

T " I l"

00 <

O
I5?c ? <

' o -913 a? aa. 2 ? 00 jQ il .y "5

EU O? c< m
et 85|

~ ?"<-r

.=

8g

*? *

?s
to -? a.

.2 "S

a. c*

o 00 as
O) u co CO 2 CO e a; LU IDN r-O t- ^ O O t O W r- O CN

CO t- m Ch ro CO vO cN cN CN r- m o m

3 < % </S c o o O? 00 a> ?? i_ QJ "O >-n


C

co

t? o m ix o m m ro ix o O cN r- o

CN r- f? o

-3- t? m

I I

00 "5 U E o LO 2

0> s?> 00 -3" i? CN i? r-(Nr-0(N(^r

r~ *? CN t? CN O

vO

CO

rx cn o vo o m o O m O t~ r- Tf T

CN m

CN vO ro ro vO

2
o o ro co i? in CN ^ O CN CN O CN co i? O i? O CN in r- o O CN CN ro CN i? O v?> 0) ^c a? Z o ON l? 00 en

O
^ 12

c
.2? ^

2?
* LU < 3.2 "O -"

f? c ?2 av aj

u en -C 00 CO NNCOm^?vO?N Mr-t-r-r-DOO ro o^ ro CO ro O ^ CN CN CN CN O m O

o to

c o co ^ m r- vo co oo CN r- o o o m o Tt CN CN rx o? <*- ?n r- CN CO f? r- O ^T Q. O Q. J3 3 5 JC ai Ja " S-^ u

? E a> a> 0) "D


-2 W ? .2

52 N
CN^ <
a> ro op-oc </?

1-5 lo
_qj ob if _0) CO 2 CN ro CO vO rs| 00 r CN m i? Tf o o o ro ro ro CN (NO "sf

g?

o 3 a. o CL O) > o CL

u "= & > "* o ?. c

a. ? ?S ?o

c"S ? ? *= "^ ;?s 0) 3 m ? ^ U?^ g g .? .2? 03 "?5

"" 1> o 00 ?- a? in i -O ?5 C u.

C mo TS DO

(H C h. ]S Z 3

O 00 t? COC a? U

il

i 8?E

en

11 ?U OQ vfc LU c< .2-0 a. o _OU 5 ? CD i Q. 2.S 00 ja o .2 ~~ O <D ooz2 u c a; Eo CD -U

00 o c CO N

go|?

"a; "O?o? v2 u x "S

? SR
V2 u .! (y ?5 -s s ?

\<B c
JO

S ? s - o?
"c 00 ?n * o * ^. m S^ ? *

? en

5 (2 ?. sr a. ck

507
W. FREY, K.-L LIAW, Y. XIE, AND M. CARLSON

The first two columns of Table 3 present the results of this analysis for the total poverty populations of males and females, respectively. The gen are undertaken because we anticipate that migration der-specific analyses of poor women will be most responsive to a State's welfare (and children) benefits. However, the results show clearly that of the two policy variables, alone shows a significant and strong negative immigration impact on State nor females does the welfare internal migration levels. For neither males benefit factor show up to be strong. The only other State attribute which shows a relatively for both males large effect on State poverty migration, of the standardized (as measured by the magnitude regression is the service employment coefficient), growth variable. The analyses of the remaining age groups on Table 3 show that these same variables are dominant for the primary labor force age categories, and 45-54, as well as for ages 5-14 which can represent the 25-34, 35-44, children of people in the former ages. The net migration of both males and females in these age categories shows a relocation away from High immi in their service States, toward States with growth gration industries, espe cially those located in the South Atlantic region. The impact of State wel fare benefits appears to be almost negligible for the poor people in these
age categories.

and females

three age categories where immigration does not show a strong on poverty migration are those which do not include the primary impact labor force years: ages 15-24, ages 55-64, and ages 65 and above. The migration of the former age group, over the 1985-90 period, really pertains to movement away from the parental home, to college or the military, and to a first but not a career job. The latter two age groups are comprised, to an increasing extent, of retirees or near-retirees. Factors which do show up to be significant for the migration of the 15-24 age group are a State's service employment rate. These indi growth level, and its unemployment viduals also are exceptionally likely to leave the Northeast region. Further, in this group show the strongest positive relationship for women migration to State welfare benefits, of any of the age groups examined. Yet, the rela is still relatively small and statistically tionship insignificant. The net migration of poor people in the elderly age categories appears to be associated with States that have high service employment growth? the need for service to cater to the reflecting, perhaps, industry workers in these States. Not unexpectedly, both the pre-elderly and elderly elderly show a strong tendency toward net out-migration from the populations "snow belt" States in the Northeast census and Midwest regions. These findings for the elderly, pre-elderly and post-teenage populations which are not responsive to immigration from abroad, stand in contrast to the strong

The

508 POPULATIONAND ENVIRONMENT

in the pri impact immigration appears to exert on the population negative to the argument that poor labor force ages. It adds further evidence mary are moving in response to competition from immigrants for employ people ment in these areas. Further evidence, fo this view, appears on Table 4 which supporting on the poverty population in the age 25-34 age group, specific only to different levels of education. These analyses make clear that immigration from abroad holds its greatest negative of men impact on the net migration with and women less than high school, high school and some college. In the migration of poverty men and women with bachelor's contrast, degrees to immigration. is not responsive It is the former group which is perhaps In competing with most vulnerable Two immigrants for new employment. other contrasts can be made between and college gradu the less-educated cuses ate poverty populations shown here. It is that the former groups appear to to manufacturing be more positively responsive growth op employment whereas the latter are strongly drawn to States with high levels portunities, of service employment the former groups are more growth. Secondly, drawn to States in the South Atlantic the latter are strongly region, whereas more greatly attracted to Pacific-region States. These data, coupled with the age-specific reviewed above, point up the importance of a analyses levels toward of its poverty immigration inducing out-migration labor force among the less-skilled within primary population, particularly In contrast, none of the analyses shown above suggests that State ages. welfare benefits are important in affecting State net migration levels for the
poor.

State's

STATEATTRIBUTES AND THE MIGRATION PROCESS


The previous section has presented an overview of State net migration which gains and losses along with a multivariate suggests that im analysis but not welfare benefits are important in effecting those gain and migration loss patterns. The analysis of net migration for individual States represents an assessment of the outcome of more complicated processes? migration streams to and from each of the States. While useful in involving migration the outcomes of the migration anal the net migration process, evaluating not provide information about the relative of ysis does importance "pushes" and "pulls" of specific migration streams, and the importance of our policy and welfare this benefits?in factors?immigration affecting more complex It is in this section that we address the process of process. and the roles of these two policy factors vis-?-vis other migra migration,

509
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

from a tion determinants in affecting the magnitude of out-migration State?as distinct from what the "drawing power" they exert in attracting in-m ?grants from other States. The ability to distinguish an attribute's "push" from its "pull" effects on if it streams has important policy For example, implications. migration were found that the magnitude of State welfare benefits had more "pull" not necessarily than "push" effects, then lowering those benefits would even though raising induce an out-migration of the poverty population, attract poor migrants from other States. Or if immigration ex them would erted stronger "push" than "pull" effects on a State's internal migration, serve then legislation which would restrict immigration to that State would to directly reduce internal out-migration to other States. To assess both the "push" and "pull" effects of State attributes on the we employ inter-State migration the process of the poverty population, nested model which has been popularized Liaw (Liaw & Bartels, by logit 1991). 1982; Liaw & Ledent, 1987, 1988; Liaw, 1990; Liaw & Ottomo, this approach, the "push" effects are assessed via a Departure Using are to destinations) and the "pull" effects (allocation of migrants Model, via a Destination Selection Model assessed (see Figure 1). In these an alyses, we evaluate Departure Model Specific attention rate), immigration model focuses on the State attributes, shown in Figure 2, as origins in the in the Destination and as destinations Selection Model. is given to immigration from abroad (defined here as the Each and State welfare benefits (as measured above). two age groups: age 25-29, and 30-34 in order to exam

FIGURE 1. Nested

logit model

of

interstate migration.

DESTINATION MODEL (FOR MIGRANTS)


X ?49 ORIGIN STATES 48 DESTINATIONSTATES ?STATE ATTRIBUTES ?POPULATION SUBGROUPS

DEPARTURE MODEL (FORRESIDENTS)


?49 ORIGIN STATES ?STATE ATTRIBUTES ?POPULATION SUBGROUPS

510 POPULATIONAND ENVIRONMENT

FIGURE 2. Attributes

of states.

RATE IMMIGRATION STATE AFDC & FOOD STAMPBENEFITS RACIALSIMILARITY GROWTH TOTALEMPLOYMENT MFG EMPLOYMENT GROWTH SERVICE EMPLOYMENT GROWTH PERCAPITA INCOME UNEMPLOYMENT COLD CLIMATE HOT CLIMATE

ine precisely how these State attributes will affect the migration process for In addition, each model these primary labor force age categories. incorpo rates interactions with key population subgroups classed by race-ethnicity, and nativity. (See Figure 3.) The findings, attainment, gender, education in described below, are the result of extensive analysis which preliminary different combinations of State attributes, and their interactions vestigated with population (Note: Both "first" and "best" estimates of each subgroups. model are displayed below.)2

2 a between interactions for these models involved Preliminary analyses investigating attributes and all relevant demographic range of relevant State origin or destination sub-popu se lations toward explaining rates of departure, resident and migrant patterns of destination lection. The original list of State attributes was more extensive than that listed in Figure 2 and included an assessment of State employment growth by industrial sector, and the State's vio rate. Similarly, a variety of interactions were lent crime with the population sub employed made census at our disposal. These cross-tabulation groups, possible by the detailed took the form of cross-tabulations, and multivariate specifications. in Tables 5 and 7 include two models shown "first" for each age group: and "best" specification. was The "first" specification arrived at after we had specification, our preliminary our tentative model, conducted the most and represents analyses including State attributes The and their interactions with relevant important population subgroups. "best" specification this tentative model, those variables involved rerunning only including and interactions which of We showed statistical both versions greatest present significance. to point up the lack of significance the model with associated factors several (in the first were which or policy and relevant on theoretical specification) initially considered grounds, preliminary analyses The final results also showed promise in our preliminary analyses.

511
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

FIGURE 3. Key population

subgroups

(within poverty

population).

AGE
25-29, 30-34

RACE-ETHNICITY AMERICAN WHITES, BLACKS, ASIANS, HISPANICS, INDIANS

GENDER
MALE, FEMALE

EDUCATION
THAN HIGH SCHOOL, HIGH SCHOOLGRADUATE LESS COLLEGE SOMECOLLEGE, GRADUATE

NATIVITY
NATIVE-BORN, FOREIGN-BORN

Statistical Model
For a particular subpopulation that the t, let p?j,t denote the probability in state i in 1985 had moved to state j by 1990. Although population living our theoretical interest centers on the determinants of pjj,t, latent proba our database contains population-specific is unobserved. Instead, bility pj,t estimates of p rates, which can be viewed as sample-analog out-migration we con the definitions of marginal and conditional ?j,t. By probabilities, ceptually decompose p?j,t into (D P?j,t = Pi+,t Pji?,t/

where p?+ ,t is the marginal probability of departure from state i, and Pjii,t is the conditional state j as the destination given probability of choosing framework, we separately departure from state i. In our nested logit model and the model p?+ ,t and Pjli,t and label them as the Departure Model we will discuss the Destination Model. First, for the sake of convenience, Destination and then the Departure Model. The Destination Model Model a logit specification with predictors is essentially that are population-spe cific as well as destination-specific. That is, (2) Pjiu = exp(?' yu#t) / 2 exp(?' y?M),
k ?* i

512 POPULATIONAND ENVIRONMENT

where y?j,t is a column vector of explanatory variables; ?' is a row vector of unknown coefficients; in the denominator and the summation is over all With similar notations, we destinations. the Departure possible specify as Model (3) pi + ft= where natural exp(a' xu)/ [1 + exp(ot' xi/t)], l?,t, the including selection model:

variables x?,t is a column vector of explanatory of the destination logarithm of the denominator

(4) li/t= log

[2exp(?'yik,t)J.
k * \

the drawing power of the rest li,t is called the inclusive variable measuring of the system on the potential migrant out of area i. a' is a row vector of coefficients. unknown are to be estimated by the maximum The unknown parameters quasi likelihood method the least squares (see Liaw & Ledent, 1987). Unlike this method does not depend on the logarithms of the observed method, and hence is not constrained As a frequencies by many zero frequencies. the two analyses can incorporate migration streams across consequence, all 49 States (continental US, including District of Columbia).

DESTINATION SELECTION PROCESS


the model for the destination is estimated selection process the departure process, we first present an inventory of the results of before the Destination Choice Model (shown in Table 5), followed by a summary of the overall contributions to the total explanation for major factors in the model (Table 6). The model results, in Table 5, are useful because they interactions between State attributes and population point up significant to explanation the contributions (Table 6) provide the subgroups. However, best gauge as to the overall impacts State attributes exert on the destination selection of the migration process. Because

Immigration To study the effects of foreign-born vari immigrants, our explanatory able is the immigration rate of a State which is the potential destination for interstate migration. The rate is defined the immigration by dividing 1985-90 foreign-born immigrants of the State (aged 5 and over in 1990) by

t- CN co d I

CM I I

c^ <n^ en en tn oo e? cnd

vO <N I

O en

d d

O O

o d

O cn

m en

il

r- k rs crin O) oo CO vO rr

vjd vO

r? lo

en

i-^r-^ m' d

il

cn tT en v?> I

co I

oq I

rs[ I

r- r- en VU O CN ^ en r-* LO

CN I

d I

<n un d CN

0> o d d d I I
m oo o o CN LO vO^ tn vO I

CN co d d l I
vO v? I vO I

rx m t?- o vu 00 00 00 vD co t-^ t-"d m' d

c* vO d I

o I

o I

CN (N r- O^ t? en O CN O O O O

o^ rv? rf r^ CN CT? vO CN vO 00

o> lo tn

cK rr ^

m O

*? r

d d

rsi d I CN v? I

r? CT> r? 00 00 O rs. CN vjOt en r^ t? en d

*? m o CN <? CN O I

odd
CN O en d fN crj "?Tvo O o? TT TT

vDd

vD I

in vD I

O) (Nr CN rr d

00 r*? l?

in ?-^

do

irirt-

O ,?

CT> m rp

Tf CN

d ! I

i? o> O 00 O r? ?*0t? CT?vO Or-nOn

O I

O O

vO O

l\ en O (N O ?N *- fN

d d o d

I I
?' ?

* I 1 -i cu u C Q. es ccj 03 ? t?
< I c O 00

_J je

?j

.t? c c E < cv_ O 00 ?'

<U vU " ?? u- u_ c u C3 "?Z Q. OJ Is oo < X <

I EE

ro .2 -S? cu cu ? -o <S *co* 75 ,<u ?

Si c

00 *~ CU ?j X

E O c/)

'

cd E

cd E

_cu cd E (U

O 00

CQ

o I C/>
z w o *

O CQ i </> 3

ci_ O Q? c? D

c?

CD C? CD <? <L> cecee <? OJ <? <U QJ - rf> ^v>00 q0 f>f\ ULiaaaaa

ISS3 ?JJ5 Q 2
uj fe
CO ? fc'E

H3 CE}Cd CS CQ o; "o oo in C? Cn Cn CO u CO mU ? v T3"D ~OTZ "O O O O O O o -c o-2 O ? O O O o.oooS o .2 CO *3 13 13 cu 2 c II2 - ?2 o^ o? o3 oo 2 2 u ob? oo ? .5? E oo m 00 rd?UUUUU O P ? :?????L?.U.LL u- f-1 H- Jr1 c^ U. ^^ LL. u u u E^ O. LL E I" .E < C? C? Q? < < < < < < CU . cu . <y o A3 "? ca"0

EEEEE

c 1 cu cu . CO , a?. a

Eo E

g E

l^c^S

cu cu c c cu cu u CO CO cd OL Cl C?< E E cd ?3 . co ?? "O 3-O "? "" O "O O " ~^= cd o^= O tu O cu .i .i .i c/i Lo lo

?cS?

LO fN? r?"a*

ro LO

cq rs. lo o Tf 00 ^? ro I

?-; I

rs. cq lo lo

r-. cq rs. O) rx co co m o^ ro

U ?> Q. Q. 3 <-n ~ O CU co < o o ro d 1

O O u p ?

O *- CN CO Nt-OK O rsi ??* O

<N

O O ro *? ^r lo rsj t? jv. r

O? O ? ?
I I I ro I hv I ro h?. I

CN LO

vu lo rs. >? t? *-* ,

CN ro O O) ro 00 ""tf* ro

r-. rp cq cq cq' ^ r>! rf lo ro ? CN j | CN

o u

r? ro O

TT

odd
lo ro

p d

lo <n O) cq ?fN?d

N-i? rv

OOrorsjrrO ioNt-Nr-o

^ fpj

do
rv. tt r-' r\ ro co

? ? ? ? ? ?
l It id vjOi-; lo cq o LO LOO*"CNrf ro *? i ro )I )

\r

IS r- lOOi s? O LO rO*

3 tO O ~ u CO

rv-

rr ro

d
ro O CN r\rn O ? I

d
ro T?*

??.?. **:?) O O CN *-"

^f rr? rs.

co O N. rv lo r-r rr ro r- o

CO

OO
I

? ? ? ? ?
lli r^ cq o vD cN cq o d c* cN r*-" t?" CN i? i CN I I CL O . -o 2 3 ^ 2 o? O 3 -2 c 3 O ? CL O o?3

< c
a> o CN~

oq CN l

IN. r- lO\O) vO O iO ro

r-

rM

O U

CO r- O r- LO O

r? o I

LO

t? o

CN o I i

?? fN.

odd

d d

co t? co vO rv. "3 o> ??O vO rj- ro ??o ,_ o o o o o o III d

O o

O o JZ u m X ? c o

0) CO

c o 00 3 -a c o

c E
o

0J 5 u

O O

op I c

O c O ' w-

^ ?-? ?? ? eu LU ?; W? TO CU c 3 ? _o ^ .- C? Q) CD c H LO U S< ni 3 cr t? Q. OU(D Su ?8 3 O LO #

.2 If c u. ?E m .y r Q- 2 "tn i/> C X < _j >-> >*

3 ? .E . ~0 .? u? t? lu Q> 00 _?D O w j=.c -r: O u o o u o 3 -S "O c? C

SI U U LO 3 ?o 00^ -5P xo r- Q j~ ?2 o?

= ?

*i:

v>

s^

_0) JO

.2 co >>. fi

5 < #

SOU

O c Q.

iO?11 -S
m U o?

7= c r= lu E-9 .2 E -2 E .?? co L?" Lo LoO g g m 3 3-5 U U i-S U -g 5-g LU LU LU U f0 c? o? 3.

3 ^> c CD lu a c o aa a? LU LU aiE.2 * 11 E OOlu ? c? LU <u d,~ CD <UU ? U U <L)OJ O h- LO (T)

See

c o CU CD

U w c CD

CD -} vu

"M

CD CD <D ilu "c "c "c ? o ?'Q

S??

Ses CD

w 5o

? D ?

c c
o o

515
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

TABLE 6
Factors from the Best Destination Effects of Deleting Selected for Interstate Poverty Migrants Choice Model in the 25-29 and 30-34 Age Groups Decrease Deleted Explanatory Factor Rho-square
Rho-square

in

Quasi

Log

Likelihood
-1096941 -1096978 -1109335 -1098814 -1100594 -1138504 -1109927 -1102675 -1171711 -844249 - 844033 -858472 -845519 -849539 -875680 -853644 -849305 -903341

0.1461 Immigration AFDC & Food Stamp Benefit Racial Similarity 0.1365 Income, Employment Growth & Unemployment Coldness at Destination Population Size at Destination 0.1360 Distance 0.1416 Contiguity Distance & Contiguity 0.0879 0.1516 Immigration AFDC & Food Stamp Benefit Racial Similarity 0.1373 income, Employment Growth & Unemployment Coldness at Destination Population Size at Destination Distance 0.1422 0.1465 Contiguity Distance & Contiguity 0.0922
Notes: Statistics

0.1461 0.1447 0.1433 0.1138

Ages 25-29 0.0007 0.0008 0.0104 0.0022 0.0036 0.0331 0.0108 0.0052 0.0589 Ages 30-34 0.0008 0.0006 0.0151 0.0021 0.0061 0.0324 0.0102 0.0059 0.0602

0.1518 0.1504 0.1463 0.1200

for Ages 25-29: Likelihood of Null Mdl Quasi-Log Likelihood of Best Mdl Quasi-Log of Best Model Rho-square Statistics for Ages 30-34:

-1284639 -1095997 0.1468

Quasi-Log Likelihood of Null Mdl


Likelihood of Best Mdl Quasi-Log of Best Model Rho-square

-995139
-843477 0.1524

the 1985 population size of the State (also aged 5 and over in 1990). The unit is percent per 5 years. For the poor interstate migrants of both 25-29 and 30-34 age groups, the immigration rate had both positive and negative effects. The positive

516 POPULATIONAND ENVIRONMENT

for migrants who were better educated (with at least some The negative effect occurred to the U.S.-born less well college education). or less) migrants. The negative effect of school graduation educated (high of the less well edu choice immigration on the destination propensities was stronger for to all races, except for Asians. cated migrants occurred It found and American Indians than for Non-Hispanic Whites and Blacks. Hispanics to Asians, With the effect was statistically respect insignificant. The negative effects for a State's immigration rate, interacting with the less well-educated is consistent with segments of the migrant population, the net migration observed earlier. They suggest that at least part of patterns the negative net internal migration for noncollege in High Immi graduates States might be attributed to a reduced "pull." However, as our gration later results show, the increased "push" away from these States exerts a more dominant process. impact on the migration

effect was

Welfare Benefits
To study the potential attractions of welfare benefits to the poor inters our explanatory tate migrants, is the real annual AFDC and food variable stamp benefit (AFDCFSB) per recipient family. The variable was created from the observed nominal values of the State in question for 1985 and 1988. The nominal values of the two years were first adjusted by the CPI to the 1992 dollar value. Then, the 1985 values of all States were adjusted by cost of living indices (McMahon & Chang, the 1985 State-specific 1991). Due to data limitations, the 1989 cost of living indices were used to adjust the 1988 values. the Finally, the real values were obtained by averaging 1985 and 1988 adjusted values. The unit is $10,000 per family. We found that AFDCFSB had a positive (attractive) effect on the poor female migrants of all races. The positive effect was stronger on American and Hispanics. Whites Indians, Asians and Blacks than on Non-Hispanic With respect to education selectivity, we found that poor female interstate least likely to be attracted migrants with the lowest level of education were It appears as if the least educated by AFDCFSB. poor migrants were much less knowledgeable about interstate variation in social welfare, and intends to counter the stereotype that the least educated poor being drawn to "wel fare magnets." the significant interactions that are shown between Despite State welfare benefits and select population in Table 5, the subgroups, overall contribution in explaining of these benefits the destination choices of poor migrants isminimal. This is discussed in our review of the below Table 6 results.

517
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

Other State Attributes at Destination


Racial Similarity. That the United States is a melting pot is more an ideal than a reality. We expect that interstate migrants are more prone to a destination with a familiar racial milieu. To capture this effect, choose our explanatory in the following way. variable is racial similarity defined For migrants of a given race, we first find the proportion of the 1985 popu lation of a potential destination in question belonging to the same race, to the proportion. and then apply logit transformation For our analysis, we use five mutually racial categories: exclusive (1) Non-Hispanic (2) White, Asian (including Pacific Islander), (4) Hispanic, and (5) American Black, (3) we call the last four categories Indian. For convenience, the ''minority
races."

We found that racial similarity had highly significant positive effect on of the poor migrants of every race. In the destination choice propensities to the mi relative to the Non-Hispanic those belonging Whites, general, races were more strongly attracted by racial similarity. This was par nority Indians and Asians. ticularly true for poor American To see if the least educated poor migrants were especially subject to the positive effect of racial similarity, we into the destination introduced choice interactions racial model racial similarity, three-way involving and the lowest level of education (less than high school grad background,
uation).

terms showed status signifi that the low education of Asian and Hispanic the propensities poor migrants cantly strengthened to select destinations with a higher concentration mi of the corresponding norities. However, in the 30-34 age group, Black high school dropouts were found to be significantly less attracted by racial similarity than their better educated counterparts. The interaction

la Our analysis employs Labor Market Variables. the conventional bor market variables: income per capita, (2) employment (1) growth rate, rate of potential destination. and (3) unemployment in For a State as a potential destination, income (per capita) is defined the following way. We first adjust the 1985 and 1989 nominal per capita incomes of the State by the corresponding cost of living indices of the same & Chang, The 1985 and 1989 adjusted values are 1991). years (McMahon then averaged. The unit is $10,000 per person. We found that for poor migrants, the expected positive effect of desti to those with the highest (college) nation income only occurred level of
education.

518 AND ENVIRONMENT POPULATION

subject to the pull of total the pull of ser employment growth rate at potential destination. However, vice employment limited to the less well educated (less than growth was some college education) poor migrants. Those with college education were unaffected It seems by the pull of service employment practically growth. that the "pull effects" of service employment in the late 1980s was growth limited to the low-skill jobs that did not need college level education. mostly We use unemployment rate as a proxy for the difficulty in finding and that among the conventional labor market holding a job. Since we believe rate ismost to be affected by the feedback variables, unemployment likely effect of immigration, our unemployment rate is the 1985 unemployment rate (rather than the average of the 1985 and 1989 unemployment rates). We found that among poor migrants, educa only those with college tion were subject to the negative effect of the unemployment rate at poten tial destination. For the 25-29 age group, this negative effect was not even statistically significant. Our overall about the effects of the conventional labor impression is that the destination market variables choice behaviors were more likely to result in the improvement of incomes and employment for better edu cated poor migrants. Their less educated at counterparts, being mainly tracted by destinations with less likely to improve jobs, were low-quality their economic situation with migration. We use the coldness of winter to represent the poorness Climate. of at potential climate destination. The coldness of a State is defined as of cities with records from average of the heating degree-days weighted 1951 to 1980, using city populations as the weights. The unit is 1000 de grees-days. We found that coldness at destination had a highly significant negative effect on the destination choice propensities of the poor migrants. The neg ative effect was somewhat stronger for the 30-34 age group than for the 25-29 age group. Size. The size effect of potential destination is repre Population sented by the natural log of its population. As expected, this variable has a of choice propensities highly significant positive effect on the destination the poor migrants.

As proxies for the employment opportunities we use two measures of employment growth: rate and service employment growth rate. Both the 1985-1989 dividing employment growth size. The unit is "proportion per 4 years." We found that poor migrants were clearly

at potential destination, total employment growth variables are computed by by the 1985 employment

519
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

Geographic Attributes
The Destination also includes two geographic attri Selection Model Distance and Contiguity, which butes, Decay, represent aspects of geo selection process of mi graphic structure known to affect the destination are Their effects, and interactions with key population grants. subgroups, discussed below. To study the negative effect of distance, our explan Distance Decay. is the natural log of the weighted State distance between atory variable centers. The unit is "log of miles." population We found that distance had a highly significant negative effect, and that the effect was the strongest for the least educated. The effect was also than for those with at least some stronger for the high school graduates college education. We expect that migrants are more prone to select a neigh Contiguity. in addition to having a short distance, the neighbor State, because boring State functions as an "intervening opportunity" for migrants who could ing have gone to a non-adjacent State. To capture the effect of intervening our explanatory is a dummy vari is contiguity which variable opportunity, able assuming the value of 1when the potential destination shares a com mon border with the origin. We found that the positive effect of contiguity was highly significant for the poor migrants. We also found that the contiguity effect is somewhat that those with stronger for those with some college education, suggesting better education were more in "interstate suburbaniza likely to participate tion" (e.g., the migration from New York and Philadelphia metropolitan centers to their suburban areas in New Jersey).

Relative Importance of Immigration and Welfare Benefits


As indicated earlier, our evaluations to of State attribute contributions the total explanation of destination selection provides the best assessment of their overall process. To evaluate impacts on this part of the migration the relative importance of an explanatory we delete the factor from factor, the "best" specification and observe in the model's the resulting decrease in Rho-square, the greater the decrease power explanatory (Rho-square): the more important the deleted factor (see Table 6). The findings show clearly that both immigration and welfare benefits were much less important than the set of conventional labor market vari ables in affecting the destination choice behaviors of poor migrants. The

520 POPULATION AND ENVIRONMENT

effects of the negative effect of cold winter was greater than the combined conventional labor market variables. However, the effect of racial similarity was much greater than any of these factors. Spatial proximity (represented and, to a lesser extent, destination by both distance and contiguity) popula tion size were factors. However, by far the most important explanatory these latter factors might be thought of as geographic "controls." An impor tant conclusion nor from these comparisons is that neither immigration welfare benefits affect racial similarity or the the destination choice labor market variables. process nearly as much as

DEPARTURE PROCESS
This section discusses the results of the Departure Model, representing the "push" effects of the various State attributes and their interactions with summarize the most key population important aspects of subgroups. We as shown in Table 7. We then discuss the overall the Model, impact of our two policy factors, and other State attributes, in contributing to the overall of the Model (shown in Table 8). explanation Immigration State
on

we found that the immigration In the Departure Model, rate of the of origin had significantly effects on Whites, positive (repulsive) and American Blacks, Hispanics Indians, but significantly negative effects
Asians.

For some races, the effects of foreign-born immigrants turned out to be selective with respect to the level of education. For Whites, the repulsive effect of immigration was particular strong on those who were least edu cated (less than high school education). Blacks, the By contrast, among turned out to be less affected than their better educated the repulsive effect of immigration. For Asians, the comple counterparts by mentary effect of immigration was reduced to near zero in the 30-34 age group, and was turned to a repulsive effect in the 25-29 age group. None in concert with viewed theless, when immigration's overall effect on mi these results indicate that poor less-edu (discussed below), grant departure are affected by a strong "push" associated cated internal migrants with immigration. least educated

Welfare Benefits
Do welfare grant benefits interact with With the exception departures? in their effects on State mi gender some confirmatory of Whites, evi

q en cn 1

O) t r^ rt LH en 00 r-^ uS Tf Tt LH

CN I

(N

in ?N

sO

I (N Tf ^ co

7 ' 77
O CN en cn O O

I vO t- (N en *~ en

cn t? t? en r- oCN r^ or- co Oco CN ? ? ? ? ? ? I m cn rn rs. cn -^rt rr to 6 ^

O O i
m CN cn CN

en O I
oqoin^^irjino^oqcMNnfn (N^^^lfiddfNN^nr^fn

(Nr-r -m <?*

I I

o ? ? I I

en

I 1 II I I I I7 I I I

en CN I

t? O* en o co oo r- r- r- r- i- O O O O ? O O

m o o i

ONCNMOr-n^MNcOMr ?? CN r- OCNCN*? CN O "^

?????????????

"^mr

I I l I I l I I I I
>?N 0)0 *r CN cni ? O^ c? os in ^ m' r-: in

CN I

n in voco oq r c? CJ?cn en e?i *? I

cq

in

i i i

7T
d o o 1 I

sO r- N K O q o*- 6r- ? cn I
rn ?-7

OO O CO O r- O

o 6 6
rs? cn v?

o o I

en O

en o

cn cr> vu i? -* en CN r- cn cn O O O O O

1 l I I I
qviDCjsrvrnrv.cNror-^tncN^Tr vod'~*--in?N<Ncncr?d^?:T?tn

cq Ln in q oo o? i-^^

rn I

I I I I i I I I Y Y

CT> CNN. cn 00 vO O^ r- i- r- O i- O q CN O O O O O O I

O o I

mrovoor?m^crvinoo??^co n O r- PO '? m r- i-

(NO

Tf LT)r

ddddddddddddd

I I 1 I I I l I l l

o o jC
00 if c

O O

00 *X c C? -C wo

O O JE ? on -c 00 ir c ro ?n

c o O) a. o _Q? OO rfl

s? S

tU
C CO w. ~<U c? Q. tt) > ^co <x<$

c? E E ? OJ

oll

<u 5? J2 Si "5 u <u E C ca?

__ -u __. ? C CO ? u .?2 CO 9-c ?* > ' C O . 'S a) 2 c oo o C? *p f? cri CQ <X<

I? s ? 2 "P^s E*

,_?r ll, c

8
*
c

X < <J2^ 5 C ? c

ai c

u co

'C *c *?z *c

oooo
a> c tu C? <? o; tu a> c c c c tu cu OJ CU > > C? CQ CQ?i CQ cn tn tn tn tn co co o 2 o o ."t? ."ti ."ti ."t? "?Z"C *-I *?z co co cO cO

?J 5? D II 3) - - - ?J rt (? (Q (d (Q (0 O. Q? Q? C-. C C C C C C O O O O O O rOcocOcococo D? 00 DO 00 &

Q) "^ ca c o

G> "^ ca c o ro .DO ,

so SB uj

cq DO .00

00
.Ec? Eu

EE? E O ft EEEEEE

E *E S E2 E2
E E

i3?LL. QJ <

oct o? O^Joo lo In ?? i? c? i>> U t? ? CO CO CQ CO ? ??? u u u u u u LL. 03 CO ?Q CO CQ CQ < << < C-. Q? ?- C? ?. ?

EEEE

co *u s. i
O CD 00 < ro C O " , CD <L> O u

O CN Tt CO ro* ro

IN

I I i

CN I

Oo IN ro r \Q <J0*

V?NNC? o rx* ?-* r- t?* CN I

CN CO O r- r LO ro LO CN TT

77

*~

in cn in in cn ^t"

r- ro ro ro r- CN O O

I I
ro CN t? Tt ro* ro*

lo vO O I

ro ro

CO LTJt? Or-M

d d d

Mff\^0 CN \D LO ^ O O r-' r-'

?? CN cr> ro O LOQ) LT)LOO} i?' CN O T-* CN

O O

I I
^o^TrTrLnrNvqpcN^ in ^' irl d o a* o in ?-' cn ~ T 'T '

odd I
fNOr^r O ro ro* in

O m LO o d

I I I

tN ro in ro CN i? CN CN ^ *-" vO*vO*f?* O

I I

CL LO

CD O u

rrroro * ' ^ ?" ?' *~; : O O *-'

III
r^ o^ rv. vo in* "^r*

r-rfmrocOLoo r? ^ ?fN ?? *?* o ? ?in ^. ? ? O* O* i-*o' o* o o'

II

ro tr> cn co cr? *? *? o cNrru^rorooo^inino> t?* T-* r-' cn o d d

*? cr> i-" *-

lili
cn in vo cn i-" O* T- CO i-" T~ CN I <3) p cq cq cq Tj* ro in* in in*

t? CO CO CO ?N pppi-^ ?? d ? do o

II

I 1 I
V , CL D to o*- a> 00 CD 00 < CN C LO cd u u a. co CD o u IN T?- vO CN CN (Ti o* ^

IN popLO*Tt LO* I

77
o m cn rr co tn cq o O) cq f-" CN O ro* O

O O CN v? "?* IN I

00 vO

I LO i- Tf o* o* o*

I I I

cn ar\ CTv CN ro lo IN <^ o o i-" d

I I CN c? I
CN VO IN -<t ?1* ro i? KincNKt^cqoQr vorTroindocoiniNrr

co I COr-V? CN lOr O O O o

CNcq ro ro in ^

I I I
, CD o u i? CN 00 O O t-*

1-* i? <*rco

I I I

in i mot N IOr-CN ?-* odd

r^inror^OroOi?^cn cn|rtLncNinroinvoo^(N ddd^dr-*cNodro

in *? ro i? O O r- y- rO O

I I J* u co
c o 3 CL <u o ?o c tn .00

o o o d l

o o

-1 c

o o E
COOOJS C Cr OO 00 vi ?D-O o o ? ?g u- u- * * * c E E ? '00*00,2 "i_ *c ^

c '5b z

SC? o C? u 3 LU &? CE
J?.5 S 2

T3

CD "? 00 _ JZ ?

00* JO .2 i o C? c JS Q. X

? loo r ou s _i? .E-C -^j u c ^ S -S? ss-l J? J? J? ? ^ g ? 3C JO JO JO JO 00 "? e > = =0 CT3 C0 C0 CX? C0 CTJj? S S 1-2 0 - C ? S O Lu *-* -*- r-? 2 ^-ST3 o 5 l2 S im O^ .? .? .2 ?COO ?-LU 2 rial V) -o^ lo lo c .y S 9 CD CD L^'?oO ? ?90 CL> C U C g S g v? en t/i w> -. O? fy _5? .2 .2 S ra >^ U U i= uj O 0 SOS .2 s- g a ? < ? j5-2 U j3-EU co nj 03 U U

s? ^

O) *D .a O LO c^ c o O *3 C0 c.O O QQ J3 U ro U 3 c D"0 O -O UJ -?3

m ro <L> U_ LO U

co cr lo A _c

C* 0? U

IinUm

<

X <

u-

5:

je j= je ?

523
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

TABLE 8
Factors from the Best Departure Effects of Selected Choice Model Interstate Poverty Migrants in the 25-29 and 30-34 Age Groups Decrease Deleted Explanatory Factor Rho-square Rho-square Ages 25-29 0.0027 0.0004 0.0072 0.0006 0.0003 0.0019 0.0112 0.0007 Ages 30-34 0.0035 0.0004 0.0097 0.0012 0.0015 0.0013 0.0112 0.0004 in Quasi for

Log

Likelihood
-887425 -885293 -891749 -885453 -885137 - 886672 895549 -885568 -754279 -751789 -759148 -752433 - 752637 - 752490 -895549 -751776

0.0671 Immigration AFDC & Food Stamp Benefit Racial Similarity 0.0626 Income, Employment Growth & at Origin Unemployment Coldness & Hotness at Origin Population Size at Origin Education (population subgroups) Inclusive Variable 0.0691 0.0514 Immigration AFDC & Food Stamp Benefit Racial Similarity 0.0453 Income, Employment Growth & Unemployment at Origin Coldness & Hotness at Origin Population Size at Origin Education (population subgroups) Inclusive Variable 0.0546
Notes: Statistics for Ages 25-29:

0.0694 0.0692 0.0695 0.0679 0.0586

0.0546 0.0538 0.0535 0.0537 0.0586

Quasi-Log Likelihood of Null Mdl Quasi-Log Likelihood of BestMdl


of Best Model Rho-square Statistics for Ages 30-34:

-951299 - 884879
0.0698

Quasi-Log Likelihood of Null Mdl


Likelihood of Best Mdl Quasi-Log of Best Model Rho-square

- 795177
751464 0.0550

and 30-34 age groups. In the 25-29 age group, the negative effect of AFDCFSB was also significant on the poor female American Indians. We found very little evidence that the retention effect of AFDCFSB was selective with re In the 25-29 age group, poor females with some col spect to education. lege education were shown to be somewhat more subject to this retention

is provided in Table 7. However, dence only the negative (retention) effect of AFDCFSB

for poor female Hispanics in both 25-29 significant

was

524 POPULATIONAND ENVIRONMENT

effect. While in informative, these findings are put into broader perspective our analyses show that State welfare benefits at origin con (below) which to the departure of poor migrants. tribute little to the overall explanation

Other State Attributes of Origin


Racial Similarity. Except for Asians, the poor adults of every race were to the retention effect of racial similarity to the origin population. subject This retention effect was particularly strong on Hispanics In and American dians. For Asians, With respect the retention effect is not significantly different from zero. to place of birth, the retention effect of racial similarity to in general was weaker for the foreign-borns than for the origin population for those with Asian or Hispanic U.S.-boms, except However, background. for Asians and Hispanics, this difference by place of birth was either rela tively weak or even reversed. We also found that the least educated Asians were particularly subject to the retention effect of California. A probable reason for this is the rela tive proximity of California to Asian countries. The income Variables. level at origin turned out to retention effect. This effect was significant only on the best-educated poor in the 30-34 age group. With respect to employment the retention effect was and clear. We opportunities, relatively general found that in both 25-29 and 30-34 age groups, the poor were subject to a strong retention effect of the total employment growth at origin. However, the service employment separate compo growth at origin as an additional nent did not show any significant retention effect. With respect to the un rate at origin, we found that it did not have any repulsive employment have Labor Market very limited effect on the poor adults. to the coldness of winter, State, in addition variable the hotness of summer (in 1000 as a weighted is also computed This variable average cooling degree-days). of the annual data of cities with records from 1951 to 1980. We found that both coldness and hotness had significant positive (re pulsive) effects on the poor adults. Size at Origin. We use the natural log of the origin pop Population ulation size to represent the size effect. In light of the huge interstate varia tion in population size, it is not surprising that this variable had a highly (retention) effect. significant negative For each origin Climate. also use as an explanatory

we

525
W. FREY K.-L. LIAW, Y XIE, AND M. CARLSON

Population Subgroups
In both 25-29 and 30-34 age groups, Education. the higher the departure the level of eduction, higher other variables are controlled. we that the when propensities, found

to minority Races. the poor belonging Except for Asians, Minority races had lower departure than their Non-Hispanic White propensities true for American This was especially this is Indians. While counterparts. also true among Blacks, foreign-born than the Blacks were more migratory In terms of departure propensities, U.S.-bom. Asians were not significantly different from Whites.

Other in Effects. Two other State attributes are included Population the analysis as "compositional" variables: Person Born inOther States, and Armed Forces Share of Origin Population. The first variable ismotivated by earlier migration studies that show the residents, who are not born in a is State, are most prone to move out (Long, 1988). The second variable in that Armed Forces personnel may motivated similar considerations by be resident in the State over a transitory period. Unlike the other Popula tion Subgroup variables, which pertain to disaggregations of the study pop these two attributes pertain to the origin State's composition in ulation, 1985. The Proportion Born in the State exhibits a significant positive effect on departure rates for both 25-29 and 30-34 age groups. The Armed Forces' share of Origin Population the Black subpopulation. is significant only when interacted with

Drawing

Effects

of

Inclusive

Variable

The inclusive variable of the attractiveness represents the perceived rest of the United States associated with the Destination and Choice Model should exert a positive (drawing) effect on the departure propensities. its explanatory because However, power overlaps substantially with that of the education of the de factor, its positive effect in the best specification can only be revealed for those with at least some college parture model For females with education. less than some college education, we found the curious result that the States with higher inclusive values turned out to have a lower departure rate. When the education factor is removed from the departure model, the inclusive variable turned out to have a theoretically meaningful coefficient (0.2391) and a large t-ratio (22.3), implying that the States that were more

526 POPULATIONAND ENVIRONMENT

accessible
ture rates.

to attractive

destination

were

more

likely to have

higher

depar

Relative

Importance

of

Immigration

and Welfare

Benefits

its deletion from the best speci fication: the greater the decrease, the more the deleted factor. important We found that the retention effect of State Welfare Benefits (AFDCFSB) was much less important than the effects of most other factors, whereas the effect of foreign immigration was more than the com repulsive important bined effect of the conventional labor market variables. This is a significant finding and suggests that the strong immigration effect on net internal mi gration of the poor is attributable primarily to its "push" effects. Also signif icant is the strong effect that racial similarity exerts in retaining migrants in our States with similar racial profiles. Still, when this effect is controlled, increases the departure of Whites, findings show that high immigration Blacks and other non-Asian minorities. Compared with other factors, the inclusive variable played a relatively minor role in the departure model. In other words, in depar the variation ture rates did not depend on the variation in the drawing power strongly from the rest of the country. This is consistent with an earlier observation (based on Map 1) that poverty migration responds largely to "push" effects at origin, and is directed to a fairly diffuse array of destinations.

As in our evaluation of the Destination assessment in explaining of State attributes on the relative contributions of each factor Table 8). Again, the importance of a factor in Rho-square due to uated by the decrease

Choice process, our overall the departure process will rely to the overall explanation (see in the Departure Model is eval

CONCLUSION
This study has analyzed detailed 1990 census migration data to assess the impact of two policy-alterable State attributes?immigration levels and State welfare benefits?for the inter-State migration of the nation's poverty In our descriptive trends and corre population. analyses of net migration lates, as well as our more process, in-depth investigation of the migration our does results are fairly clear. High levels of immigration to selected US States affect a selective out-migration of the poverty population, when other relevant factors are taken into account. This impact tends to be stronger for as well as for the least edu Blacks, and other non-Asian minorities These results are consistent with arguments that internal migrants

Whites, cated.

527
W. FREY, K.-L. LIAW, Y XIE, AND M. CARLSON

are responding to labor market competition immi from similarly educated account This phenomenon for the small displacement effects may grants.3 in studies which the labor displacement observed examine impacts of im on the native population but do not take selective native out migrants into account 1994). Immigra migration (Borjas, 1994; Martin & Midgely, tion's impact on the internal migration process acts as a stronger "push" toward increasing the out-migration of the resident poor, than as a reduced to discourage further poverty "pull" in-migration from other States. Our results were equally as clear in assessing the effects of State wel fare benefits on internal migration of the poverty population, by showing our early descriptive these effects to be very small. Although (in analysis Table 1) showed that High Welfare Benefit States had greater rates of net than for the nonpoverty population, in-migration for the poverty population our multivariate to welfare indicates that this is not attributable analysis benefits per se when other relevant factors are taken into account. More over, our analysis of the migration process shows that State welfare benefits exert similarly small effects on both the departure and destination selection of inter-State poverty migrants. The importance of immigration for the redistribution of poverty popu this paper has lation, therefore, has both direct and indirect effects. While a secondary examined the indirect effects of immigration, as it precipitates across States, the direct contribution internal migration of immigration to State poverty gains are substantial. As Table 9 shows, three of the top four in poverty migration States gaining (from all migration sources)?Califor these gains from immigration alone. nia, Texas, and New York?achieve Over the 1985-90 period, 34 States received more poor migrants from im than they did from net internal migration. However, the impact migration is particularly in the High of immigration Immigration States as is heavy evidenced from the "turnover" in poverty population for California (Table is noteworthy since the immigrant population 10). This dynamic appears to rely just as much on welfare benefits as the native-born (Borjas, 1994; The New York Times, 1995).
to a net with other explanations findings are also consistent linking immigration of longer-term and native-born In addition US residents. to labor market compe out-migration taxes and social expenditures tition, immigrants may also exert indirect pressure via increased on an area's residents, or by posing in the housing market. Some of the out competition on a on race, may that selective reflect the kinds of tastes or prejudices movement, especially broader geographic It scale that had previously suburban motivated white flight (Frey, 1994a). is not possible with the present net to disentangle for the observed research these motivations re the demographic with the out-migration associated However, out-migration. selectivity to immigration, as documented under be expected sponse here, is consistent with what would conditions of labor market competition with low-skilled immigrants {Borjas, 1994). * These

c o *S 4 00 2

NrTf

r-oico^cncNinsjocoeoroiNCNO^t? rCO mrr

t-

t?

T-coomiN r-CNCNr-T-i?

1*5 c CO

00 .E

?i 00 r T-TrcNCOOO^Or-^lNCOOCNcoCT>CNC?ro?--TrrOvO rCNCNr-r-COCNCO CNi-rCNr-rOr?

2
00 c m U c

3 a. o a.

t? CNco^invOINCOO^Or-CNro^Ln^OlNCOO^O?? r-*? r- r- i? r? *? f- r-

CN r- CN CN CN

> O CL 0> 55 C? < c a> c o a. E o U c o 00 2 o m co o

c co O * ?= C i_ co

OCOCNNMrU^rOCNCNCOrOCOOrOO^CNLnCOCJ'iOOrO miNcocNi?r-mOLnmo^vorovoovoo??o^rrcoiN ONt-inrommcNnv?^roo^Or'in^ONv?o^oo CJD (N O" N VO" r-" CN r-" OO" ?" in 00 r-" 0> CN \? ^ "^ r-S TT 0"?" C^ *? t? r- t? t? t?? i mcNfN ^ KOivOCNCN

I I

C ceu T3 o CL r- "3 ?2 E o 2< U 00 r c .2


*co v_ 00

2 ?

Ntno^NcnOMnr-incNO?-c?iNOMooocoin NCNNKinOOm^NCN^MCNN^teO^ONr IN cn oo cr> lo ^ ^f ?^o o^ ^ ^ o m i? m en a?"o" rr n r-" oo" in cn n o" o" vD cn tn ^ i? i-r-fN^rmOmCNCNCNT <?* ,- r- r

in rs vo in rr C; vu^ od" in n ^f t- t?

H? oo

INCNi? CNvOCNCOOCTiOCO rOCNCOCOCNlNrOCOLnCO?? CNOO>itK?-^vOffvrs|t-.OOCOOOv?^O>*v?^r-00 n in n in cn vu co en en vun rn o^ rn cn rq C; cq_ o^ ^ q^ CN r-"o" O oC N en o" M""CN (N CNr-" o" O0m" CN d" 0> ^t ^ ^ OeovOvOininTf^fnrnrofnfNfNCNCNCNCNCNr-r-f rr T

^ h. S

cS

5 "?? O u

.2 COC CD > L? 0) <= >. c u vJ , .2 % ?S g r? </?O C ?J oo c u = c CO ?/i f^j tn "C ?y c *j?? D? ^ .E s5. .= ? x E .2

ts

.E

o 2 --

? 2

co c 00 c

-* u

o S 2 S

S?

oZ?jzB 5 Z .y "co?S g

vDTj-r-^COinO^KrOMONfM^^fv?O'-rOin,<tnOiON,1'OOOv(D

COa^inr-OOO^iNNNr-^v?mr-LnO^rnfNOMnKv?OQOWNO^ m^r^ ^ enT-encs\enencsTtcs,itcnum)Ttrf-^-,--+'-*'^i'-t'

t^(N

t'

.- -*.

~>

CO en

fOTMn>DNOOffiOT-(Nfn,tin^ONCO(T\Or-fNn'?lfl*X)NCO(^Or CviCNP^CNCNCNjCNrnror^rocnrnf^ror^ro^^^Tr^^^^^^^^

ONtmK?Nffinu'ioa?tinfnNr-ooo OMy1KC^?-NCO,t?NNr-in'?tOOOOO^NMm'-^^OONNKCN vOCNvDT-^-rni-i-tnnr-ONOOr-mr-^CNVOr-V?r-T-O^CyivOr-yO en in ?-* (n r-s in n er d ^f r-^ rn | no cn *-^

I rC cn viTu^ o

i-^ ro rC

i ii r

r^51?^ I I I I

.._, . , . V?m O NMnoO^N i^N^OOO^O^N^KNOCOPnMnOOiN '? '?D u j u, >j i? i? ?^ u j Mw v -*w i ?? -v >? ?? >^ -s -m i? v -?v o^ en ey\ r? cn oo O ininTri-r-MinoorNiOi-N^rr-o^^f-fN^c^no^r-cN _ m or? nr* ?^ ifi ?J^rvi <"*> ^n iy\ ^ fvi rr\ rs. f?5 (NfnNOnrnrnfov?^iNOr-vomfNMmiflinNi-oo^OiNOOui rvi <^jrr\ lt\ lt\ fs^ *_ r~i C"^ i? CON VO r-^ CN *? t-^ t-^ fCr-^LO cnCsTlOOr^Tv^l^inC^OCN^i-^Tf ?? *? r^r en

3 s 2

Si -Q o^i-NtNNinr-Tj-ooflO^rncNp^o^^fnfnpnooffitnrninoomo iniJiKr-0^'?troiOiN(N?-Chfnr-TtLr)NO?-roN,tcOv<OCifNino OMTiOfNfNv?iOi^r-OfnOcONv?mrnfNCO^vOl'Or-pncOTr | cNrscNt^T^ co^fCrClC^O^rornr^

-----

rn Os T-^r-^inChrnu-T^

,,,iTT7

g < E g c o ? ea </>-? 2 .2 O v) O > 3 c S<D ? ;= -y w O <*?

-a c ? J5 c 2 fo
>\2 o C? C O' "2 i

o u

E j3 o u

.2 'E

<y$ ^

s?

OT.2

o. ? ^ o U " X 2 J .y _2 .E ?= Etn.?.2

-c m O CL c ?? r?

?~ 2 ?

~" ni _ -1__ ? i2 ~ ri CO ? ^ ? c

<5?u?552z505?>5Z?z55izu^<

?5"SO?

Si o Z

530 AND ENVIRONMENT POPULATION

TABLE 10
Immigration and Internai Migration Components Poverty Population for

California's

1985-90 Migration Components Demographic


Categories

Rates per 1990 Population Immigration from Abroad


14.1 5.4 1.8 20.2 30.2 15.9 12.6 14.4

Immigration from Abroad


Total* 450,777

Net

Internal

Net

Internal

Migration
-48,050 -42,575 -6,522 -15,040 16,117 -22,379 -25,671 -10,733 -13,988 8.7 -12,536 8.3 19.5 -368 -11,693 20.7 1,268 -12,822 -9,340 -4.790 -5,276 -5,397 8.0

Migration
-1.5 -3.9 -1.8 -1.1 4.5 -1.5 -1.5 -1.3 -4.0 -3.8 -0.2 -1.6 0.2 -2.0 -2.3 -2.4 -3.3 -2.4

Race-Ethnicity* Whites 58,588 Blacks 6,836 276,479 Asians 108,874 Males 230,861 Females Education** Less than High School High School Craduates Some College Craduates College Age 5-1481,689 15-24 164,260 25-34 112,649 35-44 47,505 45-54 22,056 55-64 12,775 65+9,843 Ages 5 and above in 1990 **Ages 25 and above in 1990 11.0 17.1 11.7 11.0 4.3 219,916 117,672 30,673 27,055 29,428

Latinos Gender*

losses for these States' gains and internal migration not only increase their numbers of poor but change poverty populations their characteristics. The poor populations in of these States will become and Asian, with lower educations and younger Hispanic age creasingly structures. The fact that net out-migration of the resident poor selects on somewhat for race, suggests that in characteristics, comparable except immigration divisions sharp demographic creasingly by income class will exist within these high immigration areas (Frey, 1995b). Moreover, the importance of a State's racial and ethnic similarity as both a "pull" and reduced "push" for in of greater cross-State divisions poverty migration suggests the possibility race and ethnicity. This would consistent with be poverty populations, by

The

531
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

the larger demographic which appears to be emerging across balkanization areas (Frey, 1995a), associated with widening US States and metropolitan by race and class. spatial distinctions

REFERENCES
& Levin, Michael Robert W., Herbert and Pacific R., Gardner, J. (1993). Asians Barringer, in the United States. A 1980 Census Monograph. New York: Russell Sage. Islanders of and wage levels on the location decisions Blank, Rebecca M. (1988). The effect of welfare female-headed households. Journal of Urban Economics 24, 186-211. in the post-war J. (1992). National Borjas, George origin and the skills of immigrants period. In George Freeman and the work force. Chi (Eds.). Immigration J. Borjas and Richard of Chicago Press. cago: University of immigration. Literature 32 J. (1994). The economics Journal of Economic Borjas, George 1667-1717. (December 1994), Richard of state and local J. (1979). A survey of the literature on the migration-impact Cebula, Public Finance 34, (1), 39-84. policies. of Richard J. & Belton, Willie J. (1994). Voting with one's feet: An empirical Cebula, analysis and migration of the American American Indian, 1985-1990. Journal of public welfare and Sociology Economics 53(3), 273-280. Institute. Rebecca. affect migration? Washington, DC: The Urban (1991). Does welfare in Brian J. (1993). The effect of the social welfare system on metropolitan migration Cushing, the U.S., by income group, gender and family structure. Urban Studies 30(2), 325-338. Filer, Randall K. (1992). The effect of immigrant arrivals on migratory patterns of native work ers. In George force and the work (Eds.). Immigration J. Borjas and Richard B. Freeman Clark, of Chicago National Bureau of Economic Research. pp. 245-269. University Chicago: Press. & Passel, Jeffrey. (1994). and immigrants' Fix, Michael setting the record straight. Immigration DC: The Urban Institute. Washington, on the population H. (1979). The changing Frey, William impact of white migration composi areas. Demography tions of origin and destination 16(2), 219-238. metropolitan H. (1994a). The new white Frey, William flight. American Demographics (April), 40-48. H. (1994b). for US states: and internal migration 1990 census Frey, William Immigration Ann Arbor, Ml: Popula findings by poverty status and race. Research Report No. 94-320. tion Studies Center, University of Michigan. H. (1995a). The new geography of US population shifts: Trends toward balkan Frey, William ization. New In Reynolds Farley (Ed.). The State of the Union: Social Trends pp. 271-336. Sage. case study. (1995b). Immigration and internal migration 'flight': A California and Environment 353-375. 16, 4(March) Population areas: H. (1995c). and internal migration Frey, William Immigration 'flight7 from US metro Toward a new demographic balkanisation. Urban Studies 32(May):733-757. in H. & Speare, Alden, and metropolitan Jr. (1988). Regional Frey, William growth and decline the United New York: Russell Sage. States. A 1980 Census Monograph. re Edward M. & Laren, Deborah S.. (1984). Migration redistribution and income Gr?mlich, Resources Journal of Human 19, 489-511. sponsibilities. L. & Hartman, Russell attract? Institute for Re Hanson, John T. (1994). Do welfare magnets search on Poverty discussion of Wisconsin-Madison. Paper no. 1028-94, University Frey, William Johnson, Curtis. from Los black outmigration James H., Jr. & Roseman, (1990). Increasing The role of household of the American and kinship systems. Annals Angeles. dynamics Association of American 80(2), 205-222. Geographers York: Russell H.

532 AND ENVIRONMENT POPULATION

on the inter factors and ecological variables (1990). Joint effects of personal A nested Geo in Canada: adults pattern of young logit analysis. migration provincial 189-208. 22(3), Analysis graphical and interpretation of a nonlinear migra (1982). Estimation Liaw, Kao Lee & Bartels, Cornelius. tion model. 14, 229-245. Analysis Geographical and ecological factors on Joint effects of personal (1988). Liaw, Kao Lee & Ledent, Jacques. in Canada. Science Canadian 11, Journal of Regional interprovincial migration elderly Liaw, Kao-Lee. 77-100. and maximum likelihood (1987). Nested Liaw, Kao Lee & Ledent, Jacques. quasi logit model A flexible methodology for analyzing method: patterns. interregional migration Regional and Urban Economics Science 17, 67-88. in A. (1991). adults Liaw, Kao Lee & Ottomo, patterns of young migration Inter-prefectoral Studies Journal of Population Japan: An explanation using a nested (Japan), logit model. 14, 1-20. in the United States. A 1980 Census and residential mobility Long, Larry. (1988). Migration New York: Russell Sage. Monograph. to an to the United States: Journey Elizabeth. (1994). Martin, Philip & Midgley, Immigration uncertain Bureau. Martin, Susan XXI, destination. Population Bulletin 49, (2).Washington, DC: Population Reference

on immigration Vol. Forbes. reform. Migration (1993). The commission World, (2 2/3), 43-44. cost of living differences: Walter W. & Chang, (1991). Geographical MacMahon, Shao-Chung. 20. Illinois State 1991. MacArthur/Spencer Series Number Interstate and intrastate update for the Study of Educational Finance. Center University: Review in the United Kevin E. (1987). Black migration reversal States. Geographical McHugh, 77, 171-187. in the United Kevin E. (1989). Hispanic redistribution and population McHugh, migration States. Moffitt, Professional (1992). 1-61. Robert. ture 30(1), 41(4), 429-439. Geographer Incentive effects of the US welfare system. Journal of Economic Litera

York Times. (1993). Florida wants US aid for immigrants, December 21, p. A7. York Times. (1994). The unfair immigration burden, January 11, p. A14 a retirement For elderly in the US, April York Times. 16, (1995). plan immigrants, 1995, p. A1. DC: The Brookings Paul E. & Rom, Mark C Peterson, (1990). Welfare magnets. Washington, Institution. in welfare F. & Krueger, Gary. and the Sanford Interstate variation benefits (1994). Schr?m, concerns Institute for Re of the poor: Substantive and symbolic responses. migration of Wisconsin-Madison. search on Poverty Discussion University Paper no. 1032-94, The New The New The New Growth Public welfare and recipient migration. Jr. (1981). Southwick, Lawrence, programs and Change 22-32. (October), Paul R., Corbett, & Randell, and public Thomas Interstate migration Richard. Voss, (1992). In Patrick C welfare: The migration of a low income population. Jobes, decision-making F. Stinner William & John M. Wardwell and Migration: (Eds.). Community, Society Press of in America, Non-economic pp. 111-147. Lanham, MD: University Migration America. Linked migration Walker, Robert, Ellis, Mark & Barff, Richard. (1992). systems: Immigration in the United and internal labor flows States. Economic 68, 234-248. Geography the witch doc low-income households: James R. (1994). Migration Walker, among Helping tors reach consensus. on Poverty Discussion Institute for Research Paper no. 1031-94, of Wisconsin-Madison. University The Washington are seen fleeing Post. some Poor Americans states as immigrants (1993). move in, September 12, p. A3. to Lori. (1993). The migratory Michael workers of native-born White, J. & Hunter, response

533
W. FREY, K.-L. LIAW, Y. XIE, AND M. CARLSON

in the labor market. at 1993 Meetings the presence of immigrants Presented of the Popu lation Association of America, Cincinnati (April). inter the composition-dependent Xie, Yu. (1989). An alternative purging method: Controlling action in an analysis of rates. Demography 26, 711-716. Xie, Yu. (1992). The Sociological layer model log-multiplicative 57 (3), 380-395. Review, for comparing mobility tables. American

.. ^rmrnOr-Ofsr^oOvO ^ *"! 2j "^ ""C R ?l ?1 ?X ?R. lo o r- *- r

""C

COOcNcnrorot^rsiiot ?. t-cocy?co"9,mrs?fNT-f O) fN cn rr_ c? ro ca^co ooq n ? rCocKcdoo""iorr'^:'^:'fN ? * *? ! I ' i I l I I

0? o ?3 ?o ?? -o ! 45*9*5 g > oc o *2n o o .^ ? -5 "w?'S \ oz CZ>?^<*>c CO^OICNfO'^'t^'^'CNfN ?n vO ^ ^ :tt LO>-o i vu t m r? i-' f."

E ?
o E ? *5 !5 > > o w jo *5 OO

Z ? 3 = 1^3 U
O)

pnooN?orNCOfOcoifl <UvO*OOeOC^O(Nt-vOUl I ) I I i I I I I I

3 o <
ig n

Si

z o ? c? g ?I < z

O < z c? LU h Z r LU
o-? I- _2 UJ co ? CO Z <

55 E
Z?T >U^2">

? Si "DJ2

?=

? 8 ? ira? 5 M ?
g-*

?JChfONfNNONr-fNO

rM^tftoocyitn^?--*-tn <D n q in fNv to ?a^i*^ ^ N vi vo 1?' t? ?N O m m NCNr- O ON00N i O " | i j I I i I J )

PS 3 'S ?

5 ? (5
Nn(N(NC0NO>nmN C^fNNnif|r-N^(*iN cuinOt-^vOvOTcofNyu o

m U C 3 iAc U -*?_2 C ? ? S < < *

TTNohOmtoortMO COlTlCriO>CNOWrOCO' ^niNOlfNfNOlAr-fn *M oC <<r vO iflO N N i-" N r-" ^rsOc?vo?-'-f-oo^o>> co I I I I t I I I I

5 r* ro f? -? -r? = C *? ^ o o>o -^ -5 o? . ZL?2<525u

B2^IJi^

? "75o? o

.2 ?

?2 O > x

re

Oi-OOf-Mf/ifn^v?N tNs?WNOONv?INCO^ N O CK m"W v? CN00 vu CN O ^Ov?oeoooNMnt-o 00 5 <u 00 __"=. > *? o ss.

r-iigSf-vUNrttCrXN N tK rN ri co oC CO ffv ^* Tf O |??Ov?Tt^fMi~f-00>0> i i w'vONfNfNi-r.pTI i I I I t I t

^-s C g^t? sac?is-s. n o o <u -c ? d?^: o


-5 .E* ?ZCZ<^>U2*>l ^? Z ^ en .-' r>? ^

Ci . rop

euc O .2 m O ob ? Z 5i

2?2=500->->

Z lo g?K ob" ? ?S ^i^fNf?t'invONeoc?O

APPENDIX B
State Welfare State
Alabama Arizona Arkansas

Benefits

Used

in This Study* Annual Benefit $5,458 $7,351 $6,678 $9,221 $7,623 $8,462 $6,716 $6,403 $7,041 $6,938 $8,010 $7,404 $7,087 $8,222 $8,616 $6,682 $6,439 $8,110 $7,497 $7,791 $8,389 $9,381 $5,390 $7,024 $8,457 $8,043 $7,503 $8,292 $7,184 $7,361 $8,694 $6,860 $8,478 $7,281 $7,591 $9,051 $7,916 $8,508 $6,635 $8,347 $6,029 $6,179 $8,884 Level

California
Colorado Connecticut Delaware

District of Columbia Florida Georgia Idaho Illinois


Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts

Michigan
Minnesota

Mississippi
Missouri Montana Nebraska Nevada

New Hampshire New Jersey


New Mexico

New York North Carolina North Dakota Ohio Oklahoma


Oregon Pennsylvania

Rhode Island South Carolina South Dakota


Tennessee Texas

Utah

APPENDIX B Continued
State Vermont Virginia $9,384 Washington West Virginia $7,185 Wisconsin Wyoming $10,359 $7,220 Annual Benefit Level

$9,628 $8,244

of combined *Benefits the average AFDC and Food Stamp Levels represent (assuming maxi mum AFDC values. for State) for years 1985 and 1988, adjusted by the CPI to 1992 Dollar in Cost of Living from 1985 and 1989 esti for State variations Values were further adjusted mates and Chang (1991) by McMahon Source for Combined of Entitlement Levels: Overview AFDC/Food Stamp Benefit Programs: on Ways 1993 Green of Representatives, Committee and Means, Washing Book, US House 1993. ton, DC: US Government Printing Office,