The 1990 PUMS is used to analyze the relationship between immigration and outmigration of the native born in New York City. The study population is limited to native born males who lived in the five boroughs in 1985. The relationship between immigration and the probability of various kinds of moves is assessed using logistic regression. Results suggest that immigration has an insignificant effect on migratory behavior, with the exception of inter-borough migration. Unlike prior work, this study examines a single metro area, and does not limit itself to inter-state migration. These results are consistent with more recent work (Card, 2001; Kritz et al, 2001) which has failed to find a positive labor market level effect of immigration on native migratory behavior. However, the inter-borough finding is consistent with voluntary residential segregation within the city.

I. Introduction

The relationship between immigration and internal migration of prior residents is poorly understood, yet has important implications for our understanding of the effect of immigration on regional labor and housing markets. In New York City, for example, recent population growth has been characterized by an increase in the foreign born population and a decrease in the native born population. For example, results from the 2000 Census suggest that between 1990 and 2000 the city’s population grew to over 8 million, an increase largely fueled by immigration . However it is not clear whether, or how, these population flows are related (Frey, 1995).

Previous studies have focussed on the relationship between immigration and inter-state migration, or on net migration from metropolitan areas. The strikingly asymmetrical spatial distribution of immigration increases interest in examining particular metropolitan areas, such as New York City or Los Angeles, separately. Yet it is difficult to design an appropriate empirical test at this level. This study uses the 1990 public use census micro sample and I.N.S. data on immigration to examine the relationship between immigration and the probability of migration from a New York City borough between 1985 and 1990. The results suggest that immigration is not positively related to outmigration of native born males from the five boroughs, but is positively related to the probability of inter-borough migration. The next section provides some background information. The theory is described in section III. The fourth section contains data and methods. Results are presented in section V. The last section concludes.

II. Background

One interesting question about the location and mobility of immigrants is its possible relationship to the migration patterns among the native born. This issue has been a source of disagreement among researchers, yet potentially has important implications for our understanding of the demographic and economic effects of immigration. There is a wealth of descriptive information that suggests that immigrants tend to locate in areas of declining native population, but it is very difficult to tell whether the underlying process is one of "replacement" or of "displacement".

Several studies of both inter-state and metro area migration have found a statistically significant negative association between immigration and the migration of native born whites and blacks. This perspective is most closely associated with the work of William Frey (Frey, 1999; Frey, 1995a; Frey 1995b; Filer 1992; White and Hunter 1993). The negative correlation between immigration and migration, and the tendency for native born outmigrants from high immigration areas to have relatively low levels of educational attainment has led Frey to the conclusion that the relationship between immigration and migration is causal, and that native born whites and blacks with little human capital are essentially being pushed from areas of high immigration. This relationship has been attributed to either labor market competition or a desire for ethnic segregation on the part of the native-born. In either case the result has been described as demographic "balkanization" (Frey, 1995).

Yet this point of view has not gone uncontested. Inter-state migration has been criticized as an inappropriate measure in a test of labor market displacement, due to the lack of correspondence between states and labor markets. An analysis of the one hundred largest metropolitan areas found no support for an immigration-migration relationship, in a specification which used a labor market share rather than share of the population as the measure of immigration (Wright, Ellis and Rebel, 1997). Several other studies of metro area patterns similarly found a lack of evidence for labor market displacement (Card, 1997; 2000; White and Imai, 1998). A recent study of metro areas Card (2001) found no effect on inter-city mobility, but some evidence that wages and employment in certain occupations are lower in high immigration cities. A recent analysis of interstate migration using the 1990 public use micro sample similarly found no relationship between immigration and native migration (Kritz and Gurak, 2001). Furthermore, the emphasis on the trend toward ethnic segregation, or "balkanization" has been perceived by some to be anti-immigrant in tone (Ellis and Wright, 1998).

However, there has been some evidence which suggests that metro areas which are major immigrant destinations such as New York City and Los Angeles may constitute exceptions to overall national patterns (Butcher and Card, 1991). Wright, Ellis and Reibel find that their results were sensitive to the inclusion or exclusion of New York and Los Angeles. Card (2001) notes that the downward pressure on wages and employment would be experienced primarily in "traditional gateway cities like Miami and Los Angeles." Due to the extremely high concentration of immigration in a relatively small number if metropolitan areas, an analysis of the relationship between immigration and migration in a single metropolitan area would be highly desirable.

New York City has both high immigration and a declining native population. Between 1980 and 1990, for example, despite the entry of over 850,000 immigrants, net migration was negative. William Frey uses information about net migration in New York City and other areas to support his hypothesis about immigration and "push" migration. Between 1990 and 1996, for example, international migration at 634,502 and net domestic migration was –929,541. While the gross figures are suggestive, this relationship is not well understood (Frey, 1995b). Between 1990 and 2000 the non-hispanic white population declined by approximately ten percent. The New York City Department of City Planning noted that, "Immigration played a crucial role in the population increase over the decade, with nearly 1.2 million immigrants admitted to New York City in the 1990s. This high level of immigration has, to a large extent, countered a substantial net outflow of residents to other parts of the nation." (New York City Department of City Planning, 2001)

One difficulty with analyzing a single metropolitan area is finding a way to introduce sufficient variation into the estimation. One approach is to analyze very small geographic areas. One preliminary analysis utilized zip code level census data from 1980 and 1990 and I.N.S. data on legal admissions (Hempstead, 2001). A census survival methodology was used to calculate zip code level net migration rates for the native born population, and these were regressed on immigration and other socio-economic characteristics of zip codes. While a significant negative relationship was found, causality could not be determined. The fact that the relationship was shown to exist at a small geographical level suggests but does not establish that some process either in addition to or instead of labor market competition is at work. Both the scenario of natives choosing not to co-reside with immigrants and the scenario of immigrants moving into neighborhoods which natives were vacating for unrelated reasons are consistent with these findings, but these two versions of events have very different implications.

III. Theory

There are three general ideas which have been advanced to explain the relationship between immigration and migration. These theories suggest different empirical tests, and have quite different implications. The first is the notion of "displacement" resulting from labor market competition. This was the theory investigated by Card(1991) in his initial analysis of the Miami labor market after Mariel. If immigrants compete with prior residents in a given labor market, one might then expect some outmigration from the labor market, which in this case might be proxied by the metropolitan statistical area. Yet under this scenario one would not necessarily expect to find a relationship between immigration and net domestic migration at very small levels of geographical aggregation, although if housing markets were extremely tight this could occur.

This might be modeled as follows:

Pm_ijt = ?+ ?₁X₁ + ?₂X₂ + ?₃imm_i(t-n) + ?

where Pm_ijt is the probability of migration from labor market i to some labor market j in time t where j ? i. The vector X₁ contains labor market characteristics, X₂ contains information about the human capital of the individual, and imm_i(t-n)measures immigration to labor market i at some time t-n . This is a theory about migratory behavior among those in the labor force, and would be expected to be more powerful among those with human capital characteristics similar to those of arriving immigrants.

A second possible explanation for a relationship between immigration and displacement could be described as ethnic displacement, or native born avoidance. If the relationship between immigration and the net migration of prior residents is based on the unwillingness of the latter to co-reside with the former, one then might expect to see a negative relationship between immigration and net domestic migration at smaller levels of geographical aggregation – i.e. zip codes or census tracts, rather than at the level of labor market, since domestic out migrants would not necessarily leave the metro area. Under this version of events the primary motivation for moving is not competition from new arrivals in the labor market.

This scenario would look much the same,

Pm_ikt = ?+ ?₁X₁ + ?₂X₂ + ?₃imm_i(t-n) + ?

although the migration measured in the dependent variable in this model is the probability of migration from some small residential area or neighborhood i to some other neighborhood k at time t, in which k may be either inside or outside of labor market j but where k ? i. In this specification the vector X₁ measures characteristics of the neighborhood, the vector X₂ represents socioeconomic or demographic characteristics of the native born population while imm_i(t-n) measures immigration to neighborhood i at some time t-n .

Finally, if the relationship between immigration and net migration of prior residents is one of replacement, in which immigrants move into neighborhoods that prior residents have already vacated for some other reason, such as population aging, one may find a relationship between immigration and net migration at small geographical levels, such as zip codes or census tracts. However unlike the previous case, the true relationship is between immigration and lagged domestic outmigration, where immigration is the dependent variable.

This situation could be represented by:

imm_it = ?+ ?₁X₁ + ?₂X₂ + ?₃nmig_i(t-n) + ?

where imm_it measures immigration into neighborhood i at time t. The vector X₁ contains characteristics of neighborhood i , while X₂ measures characteristics of immigrants. In this specification nmig_i(t-n) measures net migration for neighborhood i by the native born at some time t-n.

The dependent variable for the first two models was represented as a probability, and depending on the specification, microdata on individuals or aggregate data in the form of rates may be used to measure native migration, or migration of prior residents. In the third model, net migration rates for small areas would be used as a repressor. Immigration is measured in the aggregate in all of these models. While these three theories have been described separately, they are not necessarily mutually exclusive. "Replacement", for example, can precede "native born avoidance", as immigrants can move into a neighborhood which has been vacated by some segment of the native born population, (who may be retiring), and this influx in turn leads to the exodus of others. Meanwhile native-born residents in another section of the metropolitan area may be choosing to exit because of labor market competition with these same immigrants, who do not happen to be their neighbors. Theories about simultaneous processes are more complex to test empirically.

The different separate scenarios are summarized in the table below.
 
 

Theory	Dependent variable	Independent variable	Possible to analyze For single  Metro area	Estimating equation
Labor Market Displacement	Migration from labor market	Immigration into labor market	No	Pmijt = ?+ ?1X1 + ?2X2 + ?3immi(t-n) + ?
Native Born Avoidance	Migration from neighborhood	Immigration into neighborhood	Yes	Pmikt = ?+ ?1X1 + ?2X2 + ?3immi(t-n)+ ?
Replacement	Immigration Into neighborhood	Migration out of neighborhood.	Yes, in Theory	immit = ?+ ?1X1 + ?2X2 + ?3nmigi(t-n) + ?

There are constraints which make it very difficult to estimate either the first or third model in a single metropolitan area. For example, in a study of one metro area it is not really possible to test the labor market displacement theory, since there would be only one measure of immigration to the New York City metropolitan area, which would be a proxy for one labor market. Similarly, the third model is hard to estimate due to the lack of high quality small area migration estimates, although the zip code level work by Hempstead (2001) represents one crude attempt to create such estimates.

This study will attempt to test the second hypothesis, whether immigration is significantly related to outmigration among the native born from particular areas of New York City. For the city as a whole, it is known that net internal migration has been negative while net international migration has been positive. Results from a preliminary zip code level analysis suggests that some type of negative relationship between immigration exists at a smaller geographic level, but it is not possible to separate the second and third explanations (Hempstead, 2001). However, micro data from the PUMS provides the opportunity to make a more refined test.

IV. Data and methods

The public use micro sample (1%) of the 1990 census is the primary dataset used in this analysis. These data include information about residence in both 1990 and 1985. Residence information is provided to the geographical level of a census variable called a PUMA (public use micro area). Unfortunately, detail about the 1985 PUMA is quite limited; only the county of former residence is known.

Data on immigrants comes from two sources. The census provides information on the "stock" of foreign born residents, with information about country of origin, year of arrival, educational attainment and occupation. The Immigration and Naturalization Service provides annual data on legal immigrants and intended residence. While the Census data theoretically includes both legal and illegal immigrants, as well as those who are in this country temporarily, it is known to under measure illegal immigrants in particular, and immigrants from Latin America and Asia in general, due in part to the inadequacies of the Census Bureau’s address list in certain high immigration places such as New York City. Another source of data which might be considered is the New York City Housing and Vacancy Survey, which includes micro data on immigrants and natives, including prior residence data for those who have moved in the last five years. However, this survey only includes current residents of New York City, which makes it of relatively little use for a study of outmigration. The PUMS data, on the hand, provide information on current and prior residents.

Data from the I.N.S. provide a measure of the annual flow of immigrants. Only those given legal permanent status are included, and recent backlogs in the legalization process have made these data an increasingly poor measure of annual flow, since many of he immigrants who have "adjusted" their status in a given year have actually been residing in the country for some time. While this is the major source of data about visa status and intended residence, the educational attainment and occupation data are often missing.

This analysis used the PUMS as a basic source of data on the mobility and other characteristics of the native born, while the INS data were used to get estimates of immigration. The study was restricted to native born males aged 18-64 who resided in one of the 5 boroughs of New York City in 1985. There were approximately 18,000 of these males in the PUMS. Due to the lack of detail about residence in 1985, it was necessary to find another way to introduce variation in immigration into the model, since there were only five different aggregate measures of immigration corresponding to the five boroughs. While the zip code level study cited previously (Hempstead, 2001) used zip-code level information on immigration, it was also based on measures of net migration rates rather than the micro data employed here.

To compensate for the lack of detail about place of prior residence, the measurement of immigration is included in a ratio form. Each native-born male in the sample is assigned to a relatively broad demographic category based on his age and race. The aggregate population of these different demographics brackets in 1985 by race and nativity was obtained from the Census Bureau. The ratio of immigration between 1983-1989 in this age group to the native born population in the various age-race brackets is calculated. For convenience I will refer to this measure as the INR – immigrant-native ratio.

This measure is somewhat unusual, as it is the ratio of a recent flow of immigrants to a stock of a particular type of native population. But while primarily filling a practical role in allowing a multivariate analysis of a single metro area where there is relatively little geographical detail about the borough of prior residence, it also makes analytic sense, as it provides a certain measure of the intensity of immigration to particular demographic groups. It is reasonable to assume that the effect of immigration on migratory behavior might depend on the relative size of critical aspects of the immigrant population – a native born male of working age might be expected to have less of a reaction to a very elderly immigrant population with whom he would have little or no interaction than to a population of his own age with whom he may compete for public space, housing and potentially jobs.

Unfortunately, ethnic and racial information by nativity and borough from the inter-censal population estimate is a bit sparse. Data are broken down into only three racial groups - white, black and other, and there are no separate figures available for Hispanics. Since the "other" race population is quite small, the highest values of the ratio measure are for this group. Since little is known about the ethnic and racial characteristics of those reporting "other", the model is estimated on blacks and whites only. Since there are 5 New York City boroughs, and 3 age groups, and 2 race groups used in this model, there are ultimately 30 different values of the INR. The distribution ranges from .015 to 3.35, with a mean of .428.

V. Results

Table 1 provides some descriptive information about mobility among men aged 20-64 years in 1990. As can be seen in these unweighted frequencies, nearly 18,000 males from the PUMS sample lived in one of the five boroughs in 1985. Of these, nearly one half had moved since 1985, but about forty percent of all moves were to another location within the same borough. About fifteen percent of movers moved to a different borough, nearly twenty percent moved to a suburban location within the New York City MSA, and the rest (approximately 25% of movers) relocated either to "upstate" New York, outside of the MSA, or to some location outside the MSA. As can be seen in Table 2, mobility differed for native and foreign-born men, with the latter group being more likely to move, but less likely to leave the city.

Table 1: Type of move between 1985 and 1990, males aged 20-64 years residing in New York City in 1985
 

Type of move	Number	Percent
Did not move	9736	54.50
Moved within borough	3638	20.36
Moved to different borough	1067	5.67
Moved out of city, within MSA	1380	7.73
Moved out of MSA	2043	11.44
Total	17864	100.00

Source: 1990 Public Use Microdata, 1% Sample, Bureau of the Census

Table 2: Type of move by nativity, males aged 18-64 years residing in

New York City in 1985
 

Type of move	Foreign-born (%)	Native-born (%)
Did not move	50.52	56.81
Moved within borough	28.17	20.01
Moved to different borough	7.73	6.51
Moved out of city, within MSA	5.04	6.40
Moved out of MSA	8.53	10.26
Total	100.00	100.00

Source: 1990 Public Use Microdata, 1% Sample, Bureau of the Census

Among native born males, age, educational attainment and are related to mobility, as can be seen in Table 3. The effect of age is not surprising. Although retirement migration was explicitly excluded from this model it can be seen that while the oldest age group are less likely to move, when they do move they are more likely to leave the MSA. Similarly, the youngest group is relatively more likely to move around within the five boroughs. The effect of educational attainment on mobility is also interesting. The most educated groups are most likely to move, and are most likely to move relatively far. Those with a high school degree or less are less likely to move, but if they do move those without a high school degree are more likely to leave the MSA than are those with a high school degree.

Table 3: Mobility by age group and educational attainment, native born males, 1990
 

Type of move since 1985	Age (years)			Educational Attainment
	18-34	35-44	35-64	<H.S.	H.S.	< B.A.	B.A.	> B.A.
None	46.50	57.99	77.96	60.38	60.48	54.46	51.72	52.39
Intra-borough	23.96	20.21	10.79	21.07	20.08	19.94	19.37	18.47
Inter-borough	8.14	6.30	3.24	7.22	5.26	6.37	7.33	6.91
Suburbs of MSA	7.21	7.35	2.96	2.57	5.01	7.66	10.11	10.57
Out of MSA	8.53	8.16	5.05	8.77	4.05	11.57	11.48	11.66

Source: Public Use Micro Sample (1%) of 1990 Census, native-born males living in NYC in 1985

The overarching question was whether, after controlling for these important individual characteristics, immigration had a significant effect on migratory behavior. We used logistic regression to estimate the probability of moving in a model which included individual-level characteristics, the INR, and fixed effects for boroughs. While there surely are other neighborhood level characteristics which it would be desirable to include, there is no sub-borough information about prior residence. Aside from the INR, other neighborhood effects are incorporated into the model through the use of borough dummies. Another variable representing the rest of the immigrant population (i.e. exclusive of the particular age group measured in the INR) relative to the subject’s demographic bracket is also included in most specifications and labeled RINR. The ratio of the entire immigrant population to the demographic bracket was also used. Results were quite similar for models with two measures of immigration versus models using only one.

Since there is no sub-borough information about prior residence available for this study, it seems appropriate to only model mobility at the inter-borough or greater level. Separate logistic regressions estimated the probability of moving to another borough, to the suburbs, or out of the metro area altogether. The overall finding was that the INR and RINR were positively and significantly related to the probability of moving to another borough, but did hot have a significant effect on the probability of moving further. The effect of the INR was greater than was that of the RINR.

Results from a logistic regression are presented in Table 4, where in addition to the coefficient and standard error, the upper and lower bounds of the odds ratios calculated from the exponentiated coefficient are shown as measures of both significance and magnitude of the effect. The dependent variable in this case is the probability of moving to another borough or beyond and the excluded categories are the youngest age group, the highest educational attainment group, the "white" race group, and 1985 residence in Queens.

	Probability of moving to another borough or beyond
				Odds ratio	95% Confidence limits
Variable	Coefficient	Standard Error		Point Estimate	Lower Bound	Upper Bound
Age 35-44 years	-.4210	.0690***		.656	.573	.752
Age 45-64 years	-1.0502	.0912***		.350	.293	.418
< H.S.	-.4077	.0797***		.665	.569	.778
H.S.	-.5177	.0767***		.596	.513	.693
H.S. < B.A.	-.2492	.0736***		.779	.675	.900
B.A.	-.0686	.0762		.934	.804	1.084
Black	-.2854	.0963**		.752	.622	.908
Public Assistance	-.0820	.1206		.921	.727	1.167
Married	.2162	.0471***		1.241	1.132	1.361
Owns home	.0947	.0463*		1.099	1.004	1.204
Bronx	.2092	.0850*		1.233	1.043	1.456
Manhattan	.2876	.0656***		1.333	1.172	1.516
Staten Island	.0127	.1088		1.013	.818	1.254
Brooklyn	-.0328	.0619		.968	.857	1.093
INR	.7310	.2748**		2.077	1.212	3.560
RINR	.0420	.0354		1.043	.973	1.118

*** = < .001, ** = < .01, * = <.05

It is worth noting that the human capital variables perform as expected. Age is negatively and significantly related to mobility, while the effect of educational attainment is positive. Blacks were significantly less likely to move than were whites. Home ownership and marriage were positively related to the probability of having moved during the past 5 years. Not surprisingly, those who lived in Manhattan in 1985 were the most likely to have moved. Note that blacks were less likely to have moved than the excluded white category. The measure of immigration is positive and significant, suggesting the ratio of immigrants in the subject’s age group to their own race and age group increases the probability of native inter-borough migration. The next set of results decomposes that aggregate category into several component parts.

In Table 5, odds ratios are presented from several models of more particular types of moves. The first model represents the probability of making an inter-borough move. The second estimates the probability of moving out of the city, but remaining within the metro area. This category consists moves to the suburbs of the New York City MSA - which may be in located in New York, New Jersey or Connecticut. The dependent variable in the final model is the probability of a move out of the MSA – to either a location within New York State or to another state. It appears that the measure of immigration is significantly related to the probability of an inter-borough move but not to the others.

*** = < .001, ** = < .01, * = <.05

This general finding is repeated in a variety of specifications, with slightly different measurements of the INR, and the inclusion of different variables. When estimated separately by borough of residence in 1985, however, results varied significantly. For those living in Bronx and Brooklyn in 1985, the effect of the INR on the probability of migration to another borough was positive and significant, while for Manhattan the situation was reversed. For Queens the immigration coefficient was not significant, and for Staten Island there were too few cases to estimate.

An examination of the pattern of inter-borough migration reveals that much of it originated from Brooklyn and Queens and was directed to Staten Island and the Bronx. This can be seen in Table 6.

Table 6: Inter-borough migration, 1985-1990
 

Borough of Residence	1990
1985	Bronx	Manhattan	Staten Island	Brooklyn	Queens
Bronx		7616	588	2758	4242
Manhattan	10402		1330	9478	8344
Staten Island	532	966		1680	672
Brooklyn	3458	11116	9072		12502
Queens	5026	9058	1442	5908

Note: Weighted frequencies, 1990 public use micro sample, boldface indicates net gain for borough listed in column.

For example, there were negative net exchanges between Brooklyn and all of the other boroughs. Queens only gained migrants from Brooklyn. Manhattan gained from Brooklyn and Queens, and lost to the Bronx and Staten Island, while the Bronx only lost residents to Staten Island. Staten Island gained from all boroughs. The pattern for whites looks much like this overall pattern. For blacks, Manhattan plays the role of Brooklyn, in that it loses residents to all of the other boroughs, while the Bronx plays the role of Staten Island, and gains from all boroughs. But it is the case for both blacks and whites that boroughs with relatively few immigrants (the Bronx, Staten Island) gained inter-borough native-born migrants from boroughs with more immigrants (Brooklyn, Queens, Manhattan). However it is also the case that Brooklyn, Queens and Manhattan are the three most populous boroughs, making it relatively unsurprising that their net exchanges with Bronx and Staten Island would be negative. This is particularly true in the case of Staten Island, which had fewer than 400, 000 residents in 1990, next to more than 1,200,000 in the Bronx, the second smallest borough. However, the overall attraction of Staten Island is real - between 1990 and 2000 Staten Island grew by 17%, a considerably greater percentage increase than that experienced by any other borough.

Another way to think about the relationship between immigration and inter-borough migration is to consider the ratio of the INR in the borough of residence in 1985 to that in the borough of residence in 1990. This number will be greater than one for moves to places with fewer immigrants, and will be less than one for moves to boroughs with more immigrants. A move from Staten Island to Brooklyn, for example, would certainly yield an INR ratio of less than one. For all interborough moves, the average INR ratio is 1.6, suggesting that moves away from immigrants outnumbered moves towards immigrants. The median was approximately 1.2.

These results suggest that for native born black and white males, the ratio of the number of male immigrants their own age to the number of native born males their own race and age positively and significantly affects the probability that they will move to another borough, but does not significantly affect their likelihood of leaving the city altogether. In the context of the theories of immigration-native migration relationships discussed previously, this finding suggests perhaps a desire on the part of these native born males to avoid co-residing closely with immigrants rather than a labor market competition scenario.

VI. Conclusions

The basic result of this study is consistent with several other recent analyses of immigration and migration, in that in general no significant relationship is found between immigration and the probability of outmigration among the native born. However, there is the finding of a positive and significant relationship between immigration and the probability of an inter-borough move. The fact that immigration does not seem to have any effect on the probability of moving to the suburbs and beyond supports recent studies which have found no evidence of a labor market "push" effect (Card, 2001; Walker, Ellis and Wright, 1997). However the effect on inter-borough migration suggests perhaps a desire for residential segregation on the part of the native born population (Frey, 1999). Yet while these results are suggestive, it is difficult to infer causality since information about exact location of prior residence is lacking, meaning that the actual proximity of immigrants to the native born within boroughs is not known.

Inter-borough migration among foreign born residents who moved somewhere within New York City between 1985 and 1990 was quite different than that described here for native born whites and blacks. While the majority of foreign born residents of all boroughs who moved somewhere within the city stayed in their borough of residence in 1985, it is clear that when inter-borough migration occurred, the majority of it went to Queens. Nearly half of those moving out of Manhattan, for example, moved to Queens. Additionally, nearly three quarters of inter-borough moves by foreign born residents of Brooklyn in 1985 were to Queens. Bronx residents also tended to go to Queens if they moved out of the Bronx. I.N.S. data on intended residence of new entrants also show disproportionately heavy settlement in Queens. So while it does seem clear that the native born and foreign born are not moving to the same boroughs in New York City, there is insufficient evidence of a conscious attempt at "balkanization".

This work represents the difficulties as well as the desirability of looking intensely at a single metropolitan area. Other studies of this topic have focused on inter-state or other long-distance moves and have used the states or a cross-section of metro areas. Yet while such migration is clearly easier to analyze empirically than smaller-scale moves, it constitutes a relatively small share of overall mobility in a big city like New York. Since immigration is so unevenly distributed spatially and since New York is so different from other cities, both in terms of the number of immigrants as well as space constraints and housing costs, the relationship between immigration and native migration in New York is quite likely to be different than that experienced in other metro areas. Studies which generalize by using a panel of metro areas or the fifty states do so at the expense of true understanding of any particular area, particularly an outlier such as New York.

These results neither support or refute those from a prior study which examined immigration and native migration in New York City at the zip code level (Hempstead, 2001). The prior study found a consistently negative relationship between immigration and net migration, but had no information about migrant destinations, whereas the PUMS micro data reveal that nearly half of all moves are local. The present study finds a significant relationship between immigration and one kind of local move, but found that immigration had no effect on non-local moves. Relative to the zip code study the present work uses a far more descriptive measure of internal migration, but has a somewhat weaker measure of immigration. In both cases the results, while suggestive, must be interpreted with extreme trepidation. In the future, perhaps the use of administrative data or data from the American Community Survey will allow a more rigorous examination of the relationship between immigration and native migration in New York City and other localities.

References

Butcher, K.F. and Card, D. 1991. Immigration and wages: Evidence from the 1980s. American Economic

Review. 81: 292-96.

Card, David. 2001. Immigrant inflows, native outflows and the local labor market impact of higher immigration. Journal of Labor Economics, 19:22-64.

Card, David, and DeNardo, John. 2000. Do immigrant inflows lead to native outflows? American Economic Review, 90:360-367.

Card, David. 1990. The impact of the Mariel boatlift on the Miami labor market. Industrial and Labor Relations Review. 43:245-257.

Ellis, Mark and Wright, Richard. 1998. The balkanization metaphor in the analysis of U.S. immigration. 1998. Annals of the Association of American Geographers 88:4(686-698).

Filer, R. 1992. The effect of immigrant arrival on migratory patterns of native workers. In G.

Borjas and R. Freeman (Eds.). Immigration and the work force pp. 245-269. National Bureau

of Economic Research. Chicago: University of Chicago Press.

Frey, William H. 1995a. Immigration and internal migration ‘flight’: A California case study.

Population and the Environment, 16:353-375.

Frey, William H. 1995b. Immigration and internal migration "flight" from U.S. metropolitan

areas: toward a new demographic Balkanization. Urban Studies 32:733-57.

Frey, William H. 1998. The diversity myth. American Demographics. June: 39-43.

Greenwood, M. and J. McDowell. 1999. Legal U.S. immigration: Influences on gender, age and

skill composition. Kalamazoo: Upjohn.

Hempstead, Katherine. 2001. Immigration and net migration in New York City: A small area

analysis. Policy Studies Journal, 30:2.

Immigration and Naturalization Service. 1999. Legal Immigration, Fiscal Year 1998. Office of

Policy and Planning, Statistics Branch.

Kritz, Mary M. and Gurak, Douglas T. 2001. The impact of immigration on the internal migration of natives and immigrants. Demography. 38:133-145.

New York City Department of City Planning. 1999. The Newest New Yorkers, 1995-1996.

Department of City Planning, New York City, DCP #99-08.

NewYork City Department of City Planning, 2001. Population: 2000 Census Summary, http://www.ci.nyc.ny.us/html/dcp/html/pop2000.html#population

White, Michael and Hunter, Lori. 1993. The migratory response of native-born workers to the presence of immigrants in the labor market. Paper presented at the 1993 meetings of the Population Association of America.

White, Michael and Imai, Y. 1994. The impact of U.S. immigration on the internal migration of the native-born population, 1981-1990. Population Research and Policy Review 17:2(141-166).

Wright, Richard A., Ellis, Mark, and Reibel, Michael. 1997. The linkage between immigration

and internal migration in large metropolitan areas in the United States. Economic Geography.

73: 234-254.