McLafferty e Preston - 1996

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

zyxw

zyxwv
zyxwvu
Spatial Mismatch and Employment in a Decade of

zyxw
zyxwvutsrq
Restructuring*

Sara McLafferty
Hunter College
Valerie Preston
York University
Spatial barriers to employment limit women’s job opportunities, but their effects differ among racial/ethnic minority
groups. This study evaluates the degree of spatial mismatch for minority women and men by comparing the
commuting times of African American, Latino, and white workers in the New York metropolitan region. Using
Public Use Microdata for 1980 and 1990, we perform a partial decomposition analysis to assess the role of spatial
mismatch in lengthening commuting times for minority workers. T h e results show that African American men and
women living in the center of the region have poorer spatial access to employment than their white counterparts.
In the suburbs, African American women and Latinas suffer no spatial mismatch; rather, their longer commuting
times reflect greater reliance on mass transit. Comparison with 1980 findings reveals little change in spatial mismatch
over time despite significant economic and social restructuring in the 1980s. Spatial barriers still limit employment
prospects for the majority of minority women living at the core of the region. Key Words: gender, urban labor
markets, spatial mismatch, race.

A lmost 30 years ago, John Kain proposed


the “spatial mismatch hypothesis” to ex-
plain the high rates of poverty and unemploy-
Microdata Sample (PUMS) for both years, we
analyze changes in employment, wages, and
commuting time by gender and race during
ment among African American inner city resi- the 1980s. T h e second goal is to assess changes
dents. Residing in segregated areas distant in the degree of spatial mismatch for employed
from and poorly connected to major centers of African Americans and Latinos in the center
employment growth, African Americans were and suburbs of the New York region. Differ-
said to face strong geographic barriers to find- ences in commuting time among gender and
ing and keeping well-paid jobs (Kain 1968). race groups are analyzed, using a partial de-
Kain’s hypothesis stimulated a large body of composition analysis (Blinder 1973), to evalu-
research on racial differences in unemploy- ate the importance of spatial barriers compared
ment, wages, and commuting in urban areas to the effects of class, labor market segmenta-
(Holzer 1991), providing a foundation for de- tion, household structure, and access to trans-
bates on urban poverty and the urban under- portation. O u r findings show a significant and
class (Wilson 1987). T h e mismatch hypothesis, persistent spatial mismatch for African Ameri-
however, clearly reflects the time in which it can men and women residing in the highly
was written. Cities have changed since the urbanized, central parts of the metropolitan
1960s as a result of economic restructuring, region. In contrast, for African Americans and
changes in family structure, suburbanization, Latinos in the suburbs, there is no evidence of
and the growing polarization between rich and spatial mismatch; rather, differences in com-
poor. These present a continually shifting set muting time reflect differences in income, la-
of opportunities, conditions, and constraints bor market segmentation, and access to trans-
for minority women and men in large cities; portation. This study expands our previous re-
yet the implications for spatial mismatch are search on spatial mismatch (McLafferty and
poorly understood. Preston 1992) by computing an explicit meas-
This research has two main objectives. T h e ure of the degree of spatial mismatch, that
first is to describe the changing context of controls for other known influences on com-
spatial mismatch in the New York CMSA from
1980 to 1990. Using data from the Public Use

zyxwvut
zyxwvutsr
muting time. W e also explore changes in spa-
tial mismatch during the 1980s.

‘This research was supported hy rhr National Science Foundauon under grant #SES 9012916. We also rhankjim Huff, the anonymous re-
viewers, and the editor for their thoughtful comments. A Faculty Development grant From York University provided travel support.

Professional Geographcr, 48(4) 1996, pages 42043 1 0 Copyright 1996 by Association of American Geographers.
Initial submission, October 1995; revised submission, May 1996; final acccptance, May 1996.
Published by Blackwcll Publishers, 238 Main Strcet, Cambridge, MA 02142, and 108 Cowley Road, Oxford, OX4 IJF, UK.
Background
z
zy
zyxwvut
The spatial mismatch hypothesis describes the
combined effects of residential segregation and
economic restructuring on minorities’ spatial
Spatial Mimatch and Employment

Anumonwo et al. 1994). Minority women are


421

overrepresented in service occupations like do-


mestic service, clerical work, and health aid,
and in certain manufacturing occupations such
as textile machine work and assembly. In addi-

zy
access to employment opportunities. The cen- tion to these occupational differences, many
tral argument is that African American, and African American and Latina women are single
possibly Latino, inner-city residents have parents, bearing full responsibilities for child
poorer spatial access to jobs and employment care and economic support of their families
information because of their concentration in (Preston et al. 1993).
segregated residential areas with few nearby What are the implications for spatial m i s -
job opportunities (Kain 1968; Ihlanfeldt 1992). match? Minority men and women may have
Since then, many researchers have tested the different housing opportunities, reflecting the
spatial mismatch hypothesis by assessing racial gendered and racialized nature of residential
differences in unemployment, incomes, and choice. While both women and men face racial
commuting times and distances (Zax and Kain discrimination in housing markets, as sole par-
1991; Ihlanfeldt 1992; Cooke 1993; Rodgers ents, women also confront discrimination
1994). Although some recent studies question against women with children, or they may
the validity of the spatial mismatch hypothesis, need to live near family or day care services for
the weight of empirical evidence supports it, a t help with child care. Along with these differ-
least in part. As Holzer (1991, 1 18) concludes ences in residential location, labor market seg-
in a recent review: “Blacks in central-city areas mentation means that minority women have
have less access to employment than have different job opportunities than minority men.
blacks or whites in the suburbs.” In the context of economic restructuring, mi-
Despite the large literature on the spatial nority women’s concentration in service jobs
mismatch, two important issues have not been and low-wage manufacturing may give them
adequately addressed, and these serve as gen- better spatial access to employment than mi-
eral motivations for the research presented nority men. How have residential choice proc-
here. The first concerns the lack of attention esses and economic restructuring affected spa-
to women’s economic and domestic roles and tial mismatch for African American and Latino
employment problems in the spatial mismatch women and men during the past decade?

zyxwv
literature. With a few recent exceptions (Mad- The second issue concerns the need to ana-
den and Chiu 1990; Ihlanfeldt 1992; McLaf- lyze changes in spatial mismatch over time. As
ferty and Preston 1992), research has focused noted earlier, the mismatch hypothesis reflects
on commuting and employment issues for mi- the time period in which it was developed-a
nority men, neglecting women. Yet African period in which few African Americans or Lat-
American and Latina women’s work is criti- inos lived in the suburbs and deindustrializa-
cally important to the economic well-being of tion had just begun. In the years since then,
households and communities. Their high rates the spatial distributions of jobs and residences
of labor force participation, full-time employ- have changed significantly. When linked with
ment, and single parenthood give women a key broad patterns of labor market segmentation,
economic role (Zinn 1989). Incorporating gen- these changes produce an evolving web of job
der in the analysis of spatial mismatch requires opportunities for African American and Latino
an explicit recognition of differences in labor men and women. More African Americans and
market experiences, choices, and constraints Latinos live in the suburbs, albeit often in a
for minority men and women. relatively small number of communities lo-
Facing racial and gender discrimination, mi- cated near the city boundary (Schneider and
nority women are “doubly disadvantaged” in Phelan 1993). At the same time, employment
urban labor markets (Amott 1992). Their opportunities continue to grow and decline
earnings are less, on average, than those of unevenly within metropolitan regions, as
white women and minority men, and they are manufacturing firms move to the periphery
more likely to work in the low-wage, secon- and service jobs in certain sectors expand se-
dary segment of the labor force (Johnston- lectively in some central and suburban loca-
422 zyxwvu
zyxw
Volume 48, Number 4, November 1996

tions (Mollenkopf and Castells 1991). T h e im-


plications of these changes for spatial mis-
match are poorly understood. Theoretically,
the loss of manufacturing jobs should lead to
an increasing spatial mismatch for minorities
monwo 1992), given the number of possible
factors, a multivariate approach is desirable.
To take into the account the many factors
affecting commuting times, we employ partial
decomposition analysis (Blinder 1973; Oaxaca
in the center, especially minority men. On the 1973). this method is widely used in analyzing
other hand, for those with appropriate skills wage discrimination (Goldin 1990) and has
and education, the growth of producer service also been used to study racial differences in
jobs in the cores of large cities may lead to unemployment (Ihlanfeldt and Sjoquist 1990).
improved spatial access to employment for T h e method involves a comparison of African
those resident in the center. In the suburbs, we Americans’ (or Latinos’) actual average com-
may see the emergence of a spatial mismatch muting time with the average commuting time
as minority residents cluster in communities they could be expected to have given their
near the city boundary, and jobs tnove to more earnings, education, access to transportation,
distant suburban and exurban areas. and other socio-demographic characteristics.

zyxw
To analyze these issues, we present a case T h e difference between actual and expected
study of the spatial mismatch hypothesis for commuting times is a measure of spatial mis-
the New York metropolitan r e g o n from 1980 match.
to 1990. T h e next section describes the meth- For each gendedrace group in the center
ods and data used in evaluating the mismatch and suburbs, we estimate the following model:
hypothesis. In the subsequent section, we ana-
Time = f(D,E,H,T),
lyze general trends in residential and work-
place location in the metropolitan region and where:

zyxwvuts
changes in coinmuting times by gender and
D is a set of demographic variables includ-
race. Finally, we discuss the results of the de-
ing age and education.
composition analysis of changes in spatial mis-
E is a set of economic variables measuring
match for employed persons over time for the
the individual’s earnings, occupation, in-
central and suburban parts of the metropolitan
dustry of employment and household in-
area.
come.
H is a set of variables describing household
Methods and Data characteristics such as marital status and
presence of children.
To compare levels of spatial access to employ-
T is a set of variables describing the mode
ment among race and ethnic groups, many
of transportation used.
studies have analyzed commuting times or dis-
tances for employed workers (Gordon et al. Models are estimated separately for each group
1989; Holzer 1991). Longer work trips for via least squares. To determine expected com-
African Americans or Latinos are thought to muting times, the mean values on all variables
indicate mismatch since workers must travel for one group (i.e., African American women)
longer, on average, to reach their jobs. Such an are substituted in the equation for a corre-
approach, however, neglects the many factors sponding reference group (i.e., white women).
that constrain and motivate workers’ commut- In particular, if b,vf is the vector of parameter
ing decisions. Long commuting trips d o not values for the reference group equation (white
necessarily signal a spatial mismatch. High-in- women) and X,f is the vector of mean values
come workers, and those with children, often on all variables for the comparison group (Af-

zyx
choose to live in attractive residential areas rican American women), the expected com-
distant from work to gain other amenities muting time for African American women
(Brun and Fagnani 1994). In these cases the (td,Vff) is:
long worktrip results from a conscious choice
rather than a spatial constraint. Although some Lf,d = C[bwfX,fI
researchers have analyzed these confounding T h e resulting value is the expected commut-
factors separately in their studies of commut- ing time for an “average” African American
ing times (Gordon et al. 1989; Johnston-Anu- woman if she were able to make the same
commuting choices as an equivalent white
woman. Because we have controlled all factors
except residential location in determining the
Spatial Mimatch and Employment zyz
spatial mismatch debate-are not considered.
The sample size for 1980 is approximately
423

180,000, while for 1990 it is more than double


expected commuting times, the difference be- that-400,000. The PUMS data provide all the
tween actual and expected times provides a information needed to estimate the regression
measure of spatial mismatch. The actual time models, although several variables had to be
includes the effects of residential location, simplified to achieve mathematical tractability.
whereas the expected time does not. Thus, a As in all census information, the identifica-
positive difference between actual and ex- tion of raciallethnic groups is problematic.
pected times indicates a spatial mismatch. To Race typically refers to biological differences
check the sensitivity of the results, one can also of sociocultural significance to a society that
reverse the reference and comparison groups, “racializes” those differences. Ethnicity refers
in which case a negative value would indicate to cultural differences reflecting religion, na-
a spatial mismatch. tional origin, and language. The race variable
The above model represents well-known in the census reports an individual’s self iden-
hypotheses about the determinants of com- tification as a member of a racial group. We
muting times. Economic factors, such as earn- used this variable to identify the African
ings, occupation, and education are expected American and white populations. We defined
to have positive effects on commuting time, as Latino those people who identified their
with workers in high-status, primary occupa- race as “Spanish,” and those persons regardless
tions, with high earnings and education, hav- of race who stated they were of Hispanic ori-
ing longer times (Ihlanfeldt 1992). Industry of gin. While there are contradictions inherent in
employment is expected to affect commuting any effort to define race and ethnicity, these
time independent of wages and occupation, are among the most significant aspects of social
insofar as it reflects the geographic concentra- identity. Race as socially constructed in the
tion of employment in different sectors. In the United States profoundly influences daily ex-
New York region, jobs in producer services are perience. Inequality in American society is still
strongly concentrated in Manhattan, and this rooted in race and ethnicity (Johnson et al.
clustered spatial pattern should result in longer 1994).
commuting times. The economic variables include household
The effects of household characteristics on income, weekly earnings, industry and occupa-
commuting times are expected to differ by race tion of employment. We combined industries
and gender. In juggling the competing de- into three groups: “old core” industries, pro-
mands of domestic work, child care, and paid ducer services, and consumer services. The
employment, women may opt to work near “old core” includes manufacturing, wholesal-
home to free time for domestic responsibilities ing, construction, and transportation (Buck
(Johnston-Anumonwo 1992). In contrast, mi- et al. 1992). The producer service sector in-
nority women’s more constrained residential cludes jobs in finance, insurance and real es-
and job opportunity sets may give them less tate, and business services; the consumer serv-
ability to reduce work trips when faced with ice sector comprises jobs in personal services,
competing domestic and employment de- health, education, social services, retailing, and
mands. Finally, mode of transportation government. T o measure occupational seg-
strongly affects commuting time. Workers mentation, we created a dummy variable that
who rely on mass transit spend more time differentiates high-status primary occupations
commuting than do workers who travel by car, from low-status secondary occupations accord-
regardless of distance (Singell and Lillydahl ing to the classification described by Amott
1986). (1992). Although these two categories greatly
In analyzing these relationships, we utilized simplify occupational differentiation, using
data from the 1980 and 1990 Public Use Mi- more detailed classifications had little impact
crodata Samples (PUMS) for the New York on the results.
metropolitan region. Our samples include only We analyzed household characteristics based
people employed in the paid labor force. Un- on combinations of marital status and the pres-
employed workers-an important group in the ence of children in the household. Using
424 zyxwvu
zyxw
zyxwvu
Volume 48, Number 4, November 1996

dummy variables, four situations were identi-


fied: (1) married with children a t home, (2)

or widowed) with children a t home, (3) mar-


ried with no children at home, and (4) single
of capital that stimulated rapid expansion in
financial services and related producer services
single (i.e., never married, divorced, separated,
while failing to arrest the loss of manufacturing
jobs (Sassen 1991; Mollenkopf and Castells
1991). In the center, between 1980 and 1990,
with no children a t home. Access to transpor- the old core’s share of total employment in-
tation is captured by two variables: the number creased 1.7%, largely because of the revival of
of cars available in the household and a dummy small-scale manufacturing in the inner ring of
variable equal to one if the person uses mass counties surrounding Manhattan (Table 1). In
transit for commuting to work. We singled out the suburbs, the old core’s employment share
fell 1.8%.
mass transit (bus, subway, light rail, and ferry)
because it involved significantly longer travel Service industries in the New York region
times than other modes of transportation. are increasingly differentiated spatially and
The study area consists of the 24 counties functionally. Manhattan remains the preemi-
that make up the New York Consolidated nent location for producer services in the re-
gion, with the highest concentrations of em-
Metropolitan Statistical Area. It is divided into
two parts, the center and the suburbs, that ployment and the most sophisticated financial
have experienced distinct population and em- and business activities. In 1990, 19.9% of jobs
ployment trends. The center includes Manhat- at the center of the New York region were in
tan and the other urbanized counties of New producer services (Table 1). However, employ-
York City and the nearby urban counties in ment in producer services has increased stead-
New Jersey, including the cities of Newark and ily in the suburbs. In addition to back-office
Jersey City. These counties are characterized activities, producer service firms oriented to a

zyxwvut
by high population densities, high reliance on local or regional market have located in the
mass transit, and low median household in- suburbs where by 1990, producer services ac-
come levels. In contrast, the suburbs include counted for 18.7% of jobs (Table 1).
all counties outside the urban core. In this Many argue that these sectoral shifts have
group are older suburban counties like polarized the labor market between well-paid
Westchester, NY,and more distant exurban professionals and managers employed in pri-
counties in New York, New Jersey, and Con- mary occupations and poorly paid service,
necticut. The suburban counties typically have sales, and clerical workers in secondary occu-
moderate population densities, low rates of pations (Sassen 1991). Suburban residents are
mass transit use, and relatively high median still more likely than are central residents to
incomes. However, there is considerable diver- work in primary occupations (Table 1). The

zyxwv
sity among suburban counties, particularly be- high proportion of suburban residents in pri-
tween the older, high-income suburbs near the mary occupations results in large wage differ-

zyxwvu
center where population and employment are entials between central and suburban residents.
stable or declining, and the exurban counties In 1990, suburban workers still earned sub-
that have seen rapid growth in employment stantially more than central workers despite
and population in the last decade. large wage increases throughout the region
(Table 1).
Changes in the New York Region: The New York region is typical of older and
1980-1 990 larger cities of the Northeast and Midwest,
where segregation is most enduring (Massey
Three aspects of New York‘s changing geog- and Denton 1993; Farley and Frey 1994). Per-
raphy are relevant for our analysis: continued sistent segregation is readily apparent from our
deindustrialization and the relocation of manu- analysis of workers living in different parts of
facturing jobs to the periphery of the region, the region. African American and Latino
continuing growth and differentiation of serv- workers are still concentrated a t the center of
ice sector employment, and the persistence of the region where, in 1990, they accounted for
residential segregation. New York is one of 38% of the resident workforce while in the
several cities transformed by the globalization suburbs their share was 10% (Table 1). For
zyxwvuts zyz
zy
zyxwvutsrqpo 1980
Central
1990
Spatial Mismatch and Employment

Table 7 Characteristics of Workers in the New York CMSA, 1980-1990 (percentages?

Changeb
Location

1980
Suburbs
1990
425

Change
Industry
Old Core 38.1 39 8 1.7 40 3 38 5 -1 8
Consumer Service 41.7 40 3 -1.4 43 9 42 8 -1 1
Producer Service 20.3 19 9 -0 4 158 18.7 29

zyxw
Race
African American 192 19 5 03 52 52 00
Latino 14 4 18.5 41 34 49 15
White 62 2 53.7 -8 5 88.6 86 4 -2 2
Sex
Men 54 7 52 6 -2 1 58 2 55 2 -3 0
Women 45 3 47 4 21 41 8 44 8 30
Occupation
Primary 47 1 48 1 10 54 9 58 7 38
Secondary 52 9 51 9 -1 0 45 1 41 3 -3 8
Means of Transportation
Automobile 41 5 45 7 42 81 8 84 5 -2 7
Transit 45 4 41 0 -4 4 11 4 94 -2 0

zyxwvu
Other 13 1 13 3 02 68 61 07
Mean weekly earnings

zyxwvu
(dollars)c $284 88 $623 84 $323 09 $713 21
Mean cornmuting time
Iminutesl 34 0 32 3 28 5 27 6
"All statistics are reported for workers disaggregated by place of residence.
'Change.s in shares do not sum to zero because of roundzng exceptfor race where the list of raciakhnic groups is not exhaustive.
'Mean weekly earnings are reported in 1979 and 1989 dollars, respectively.

white workers living in the center, their share points, replaced by driving and other means of
of the resident workforce declined by more transportation, mainly walking to work. In the
than 8%, largely as a result of a growing Asian suburbs, transit use also declined but only by
population. 2.0 percentage points. Increased reliance on
The changing spatial patterns of employ- the automobile has contributed to slightly
ment and residence in New York mean that shorter commuting times in both the center
women's work experiences have unfolded dif- and the suburbs. Between I980 and 1990,
ferently in each part of the region. Mirroring mean commuting times fell by 1.6 minutes for
national trends, women's participation in the central workers and by 0.4 minutes for subur-
labor market increased steadily during the ban workers.
1980s. By 1990, women accounted for 47.4% The 1980s saw considerable movement of
of workers living at the center of the region women into primary occupations. white
and 44.8% of those in the suburbs (Table 1). women, in particular, made substantial ad-
The biggest increase was in the suburbs where vances. In the center, almost half of white fe-
women's share of the resident labor force in- male workers, 47.7%, were employed in pri-
creased by 3.0 percentage points. At the center mary occupations by 1990, an increase of 11.9
of the region where women have historically percentage points. Progress has been slower
had higher-than-average participation in paid for minority women living a t the center of the
employment, the increase was much smaller, region. The proportions of minority women in
only 1.7 percentage points. primary occupations are lower (28.6% and
Despite substantial investments in public 23.1% for African American and Latina
transit during the 1980s, workers in both parts women, respectively), and the magnitudes of
of the region were less likely to commute by change are smaller (5.4 and 3.5%). Occupa-
transit in 1990 than in 1980. The decline was tional change was faster in the suburbs, where
most precipitous at the center of the region, the small number of minority women have
where transit use decreased by 4.4 percentage made relatively more progress in moving into
zyxwvu
zyxwvu
zyxw
zy
426 Volume 48, Number 4, November 1996
primary occupations. For example, the share of Spatial Mismatch Results
African American women in primary occupa-
~~

tions increased by 7.8% in the suburbs com- The longer average commuting times for Af-
pared with a slightly smaller increase of 5.4% rican Americans and Latinos suggest a persist-
in the center. In both the center and the sub- ence of the spatial mismatch during the 1980s.
urbs, Latinas have made the least progress However, as noted earlier, those comparisons

zyxw
moving into primary occupations. fail to account for class, economic status, and
Commuting times suggest that spatial mis- other factors that affect commuting time. We
match is a continuing problem. In 1980 and incorporated these factors by estimating the
1990, minority workers commuted longer regression equations described earlier for each
times than white workers, although gender and genderhace group in the center and suburbs.
race differences have diminished slightly (Ta- Table 3 shows coefficients of determination
ble 2). Commuting times are still significantly and sample sizes for the regression models for
shorter for minority workers in the suburbs 1980 and 1990, respectively. Given the large
than for those in the central city. African sample sizes, the models fit moderately well,
American women and Latinas living in the with R2 values ranging from .20 to .40. All
suburbs have average commuting times that values were significantly different from zero
are 11.1 minutes and 9.2 minutes less than (p = 0.05).
those of their counterparts in the central city. T h e parameters of the regression models are
The gender gap in commuting time varies not shown here because of the exceptionally
significantlybetween center and suburb. At the large number of values. Instead, we identify
center of the New York region, minority men the most significant and consistent results to
and women commute equally long times, while provide a context for the discussion of spatial
in the suburbs, there is a modest gender dif- mismatch. (For a more detailed discussion of
ference in commuting times for Latinos and gender and racial differences in the determi-
African Americans as well as a large gender nants of commuting time, see McLafferty and
difference for whites. On the one hand, the Preston, forthcoming.) In all models, mode of
existence of gender differences in commuting transportation has the most significant impact
time among minority workers is an encourag- on commuting times. Regardless of gender,
ing indication that, in New York's suburbs, race, or residential location, workers relying on
Latina and African American women may be mass transit spend from 20 to 40 minutes

zyxwvut
wealthy enough that they can accommodate longer commuting than do comparable work-
household and employment responsibilities by ers who use other modes. Economic and class
working close to home. On the other hand, variables are also significant. For all groups in
Latina and African American women in the all contexts, the log of weekly earnings has a
suburbs still commute longer than white sub- strong positive association with time. High

zyxwvut
urban women. earnings both enable and encourage workers

Table 2 Mean Commuting Times by Gender and Race, 1980-1990 (minutes)


1980 1990 Change
Central
African American men 39 1 37 3 -1 8
African American women 40 4 37 8 -2 6
Latino men 34 8 33 3 -1 5
Latina women 35 7 33 8 -1 9
White men 32 5 31 9 -0 8
White women 30 8 28 8 -2 0
Suburbs
African American men 28 7 29 5 08
African American women 25 6 26 7 11
Latino men 31 2 28 3 -2 9
Latina women 25 0 24 6 -0 4
White men 33 1 32 1 -1 0
White women 22 0 23 0 10
Centei
zyxwvu
zyxwv
zyxwvut z
zy
zyxwvuts
Table 3 Goodness of Fjr Statistics
1980
African
American Latino White
Spatial Mismatch and Employment 42 7
commute from two to five minutes longer, on
average, than their unmarried counterparts, ir-
respective of the presence of children. Mar-
riage, then, appears to either encourage or en-
able men to commute longer. This result may
Men R? ia 22 27

Women
N 6799 6386 24920 reflect the unequal gender division of labor in
R’ 23 28 36
N 7859 4555 20552 many married couple households, or the ten-
Suburbs
Men
dency to purchase a suburban residence upon
R2 31 34 38
N 2074 1623 40109 marriage. It may explain, in part, the very large
Women R2 29 36 40 gender disparity in commuting time in subur-
N 221 1 1182 29944
ban areas where married couples predominate.
1990 The spatial mismatch results are shown in
African Figure 1 for 1980 and 1990, respectively. For

zyxwvutsrqpo
American Latino White
each gender and race/ethnic group, we present
Center
Men R’ 20 23 25 the actual and expected average commuting
N 7735 9886 30364 times. The difference (actual-expected) indi-
Women R2 25 30 34
N 10684 7873 26623 cates spatial mismatch. All results use the white
Suburbs population as a reference group; however, we
Men R’ 21 23 24
N 3524 4113 77726 found that reversing the reference group had

z
Women RZ 29 29 32 little impact on the findings. As is evident in
N 4260 3218 63436
Figure 1, in both 1980 and 1990, a significant
spatial mismatch existed for African American
from all race and ethnic groups to lengthen men and women living in the center of the
their commuting trips. Other economic and metropolitan region. In 1980, African Ameri-
class variables have smaller and less consistent can women’s actual commuting times were six
effects on commuting time. Older workers, minutes higher than expected and men’s were
and workers with more years of education, five minutes higher. These differences are
often have longer work trips. In many cases, large in comparison with eight- to ten-minute
workers in producer services have significantly gap in average commuting times between Af-
longer commutes, reflecting the continued rican Americans and whites, indicating that
spatial concentration of employment in this poor spatial access to employment, as reflected
sector in lower Manhattan. in long commuting times, accounts for a siz-
While the impacts of transportation, class, able portion of observed racial disparities. By
and economic variables are similar for men and 1990, spatial mismatch had reduced slightly, to
women, the effects of household variables are five minutes for women and almost four min-
strikingly different. Among women, only white utes for men. Given the overall decline in
women appear to reduce their commuting commuting times for all groups, however, the
times to accommodate domestic responsibili- spatial mismatch is still large and significant.
ties. In both the center and suburbs, married For Latinos in the center, our results pro-
white women with children have shorter com- vide less evidence of spatial mismatch. In 1980,
mutes than other white women, indicating that the difference between actual and expected
they are able or choose to reduce commuting times was 1.98 minutes for Latina women and
time in order to free time for child care and 0.8 minutes for men; by 1990 the differences
other domestic chores. In contrast, African had changed to 2.16 and 0.60 minutes, respec-
American and Latina women have no such tively. The greater mismatch for women than
opportunity, and their commuting times vary for men indicates Latinas’ poorer geographical
little with marital status and the presence of access to the types of jobs in which they nor-
children. For all race/ethnic groups in the cen- mally work. Many Latina women work in pri-
ter and suburbs, the effects of household vari- vate household, health care, and service jobs
ables are substantially larger for men than for that are concentrated in congested Manhattan.
women. Marriage is the key dimension of Although our model controls for occupation,
household structure for men. Married men the very broad categories used here may not
428 zyxwvu
zyxw
zyx
Volume 48, Number 4, November 1996

zyxwvut A
Spatial Mismatch: Center

‘ A
0
i
i A I
I

20 J zyxwvutsrqp
1980
Black Women
1990 1980 1990
Latina Women
1980 1990
Black Men
1980 1990
Latino Men
AActual

-Expected

zyx
Spatial Mismatch: Suburbs

zyxwv
E

zyxwvutsrq
i=

-5
<>
35

zyxwvut
25

zyxwvuts
20 1
P

1980 1990
Black Women
1980 1990
Latina Women
h X

1980 1990
Black Men
E
X

1980 1990
Latino Men

Figure 1: Spatial mismatch by gender, race, and residentla/ location, 1980 and 1990.
I
A Actual
-Expected
-1

capture these detailed, gendered, occupational longer average commuting times in the sub-
effects. urbs are likely the result of mass transit use.
In sharp contrast to the results for workers Our data indicate that even in the suburbs,
living in the center, for employed suburban African Americans and Latinos are much more
workers we find no signs of spatial mismatch. reliant on mass transit than are white men and
After taking into account socioeconomic, geo- women. Minority suburban women, in particu-
graphic, and household factors, African Ameri- lar, are two to three times more likely to com-
cans’ and Latinos’ average commuting times mute by mass transit than are their white coun-
are approximately equal to the expected values terparts.
and thus are similar to those of white suburban These findings confirm past studies that in-
women with the same characteristics. This dicate that spatial mismatch is a more severe
suggests that the longer commuting times for problem for urban than suburban residents
African Americans and Latinos in the suburbs (Zax and Kain 1991). African American work-
are largely the result of differences in mode of ers living in central areas face a spatial mis-
transportation, earnings, education, and house- match that necessitates longer commuting
hold structure, rather than an explicitly spatial times than those of white workers with similar
mismatch. Theoretically, minorities’ lower av- social and economic characteristics and access
erage earnings and education levels should lead to transportation. This spatial mismatch is not
to shorter commuting times; therefore, the less for African American women than for Af-
rican American men; indeed it is greater. Inso-
far as our spatial mismatch measure shows the
effects of residential segregation, these results
suggest that in the center, African American
women’s residential choices are even more
Spatial Mismatch and Employment

lem with strong and obvious geographic di-


mensions.
z
zy429

Analyzing these issues calls for additional


research in several areas. First, the measure-
ment of spatial mismatch needs to be further
geographically constrained than those of Afri- refined and the sensitivity of results explored

zyxwvut
can American men. Also, for both men and to assess the generality and robustness of these
women, spatial mismatch has remained re- findings. The models estimated in this paper
markably stable during the 1980s. Although may have measurement and specification er-
the degree of spatial mismatch decreased rors that can affect the estimates of spatial
slightly over time, it is still large in both abso- mismatch (Hashimoto and Kochin 1980). Spe-
lute and relative terms. cifically, the occupation and transportation
variables were highly simplified and thus may
fail to represent finer-grained effects. Using a
Conclusions single dummy variable to measure transit use,
for example, assumes that mass transit adds a
What are the implications of these findings for fixed amount to individuals’ commute times. In
urban policy making? Our results confirm reality, the effect of mass transit is apt to vary
what many others have argued-that African with travel distance. More detailed measures of
Americans in central cities continue to face economic, household, and transportation fac-
significant spatial barriers in traveling to work. tors should be incorporated to estimate more
These lead both women and men to spend accurately workers’ commuting times. How-
more time commuting, time that is unavailable ever, given the variability of individual and
for other purposes. While we have not looked household circumstances and the complexity
a t the impacts on employment and work per- of residential and workplace location decision
formance, longer commuting times could cer- making, it is unlikely that any aggregate model
tainly pose a significant barrier to finding and can account for more than a fraction of the
holding a job (Rodgers 1994). Reducing these variation in workers’ commuting times (Han-
barriers will require highly focused job infor- son and Pratt 1995).
mation and placement policies that are sensi- Second, we have only analyzed spatial mis-
tive to the employment needs of African match for employed workers and thus are
American men and women in central city omitting one of the most important groups in
neighborhoods. Community-based economic the spatial mismatch debate, the unemployed.
development is also essential for generating job Recent research on the links between spatial
opportunities and local business ownership access to jobs and unemployment gives con-
within a reasonable commuting range. Along flicting results (Ihlanfeldt 1992; Cooke 1993).
with these job-focused strategies, policies to Further work is needed to unravel the impacts
reduce racial segregation and discrimination in of poor geographical access to potential em-
housing markets will also have important ployment opportunities on job information,
benefits. job search, and unemployment for African
The situation in the suburbs differs signifi- American and Latino women and men (Rodg-
cantly, with little evidence of an explicitly spa- ers 1994).
tial mismatch. For the fraction of African Finally, further research on the spatial mis-
Americans and Latinos who now reside in sub- match of minority women is sorely needed.
urban areas, living in the suburbs carries no We have emphasized gender differences in la-
spatial penalty, at least in terms of commuting bor market segmentation as a basis for expect-
time. Despite this, African American women ing gender disparities in spatial mismatch, but
and Latinas in the suburbs continue to spend gender affects minority women’s decisions and
more time commuting, on average, than white circumstances well beyond the economic
women, mainly because of their much greater sphere. Factors such as access to child care and
reliance on mass transit. Although not a purely other services, proximity to family, and dis-
spatial mismatch, in the sense of the term used crimination in housing against women with
in this paper, access to transportation is a prob- children influence women’s and men’s residen-
zyxwvu
zyxw
430 Volume 48, Number 4, November 1996
tial and employment decisions in profoundly
different ways. Qualitative and quantitative
studies are needed to explore how these proc-
design of a multi-city study. Urban Geography
15377-89.
Johnston-Anumonwo, I. 1992. The influence of
household type on gender differences in work trip

zyxwvuts
esses unfold in inner-city and suburban set-
distance. Profissional Geographer 44:161-69.
tings. Understanding women's decisions and
Johnston-Anumonwo, I., S. McLafferty, and V.
responsibilities in geographically, socially, and Preston. 1994. Gender, race and the spatial con-
economically constrained environments is a text of women's employment. In Gender in Urban
significant topic for future spatial mismatch Research, ed. J. Garber and R. Turner, 236-55.
research. W Thousand O a k , CA: Sage Publications.
Kain, J. 1968. Housing segregation, Negro employ-
ment, and metropolitan decentralization. Quar-
Literature Review terly 'journal of Economics 82:175-97.
Madden, J., and L. Chui. 1990. The wage effects of
Amon, T . 1992. Women and the US Economy Today. residential location and unemployment on em-
New York: Basic books. ployed married women. Urban Studies 27:353-69.
Blinder, A. 1973. Wage discrimination: Reduced Massey, D. and N . Denton. 1993. American Apart-

zyxw
zyxwvutsrq
form and structural estimates. Human Resources heid: SeFegation arid the Making of the Underclass.
8:436-55.

zyxwvut
Bnin, J. and J. Fagnani. 1994. Lifestyles and loca-
tional choices-Trade-offs and compromises: A
case study of middle-class couples living in the
Ile-de-France region. Urban Studies 3 1:92 1-34.
Buck, N., M. Drennan, and K. Newton. 1992. Dy-
namics of the metropolitan economy. In Divided
Cities, New York and London in the Contemporary
World, ed. S. Fainstein, I. Gordon, and M. Harloe,
-.
Cambridge, MA: IIarvard University.
McLafferty, S., and V. Preston. 1992. Spatial mis-
match and labor market segmentation for African-
American and Latina women. Economic Gcogruphy
68:406-3 1.
Forthcoming. Gender, race and the deter-
minants of commuting: New York in 1990. Urban
Geop-aphy.
Mollenkopf, J., and M. Castells. 1991. Dual City,
68-104. Cambridge, MA: Blackwell. Resmcturing New York. New York: Kussell Sage
Cooke, T . 1993. Proximity to job opportunities and Foundation.
African-American inale job opportunities: A test Oaxaca, R. 1973. Male-female wage differentials in
of the spatial mismatch hypothesis in Indianapolis. urban labor markets. Ivitcmational Economic Review
Profissional Geographer 4S :407- 15. 143693-709.
Farlcy, R. and W. Frey. 1994. Changes in the segre- Preston, V., S. McLafferty, and E. Hamilton. 1993.

zyxwvutsr
gation of whites from blacks during the 1980s: T h e impact of family status on black, white and
Small steps toward a more integrated society. Hispanic women's commuting. Urban Geography
American Sociological Review 5 9 2 3 4 5 . 14:22 8-50.
Goldin, C. 1990. Understanding the Gender Gap. New Rodgers, C. 1994. Job search and unemployment
York: Oxford University Press. duration: Implications for the spatial mismatch
Gordon, P., A. Kumar, and H. hchardson. 1989. hypothesis. Research Paper No. 9437. Morgan-
'The spatial mismatch hypothesis: Some new evi- town, WV: West Virginia University, Regional
dence. Urban Studies 26:3 15-26. Research Institute.
Hanson, S., and C;. Pratt. 1995. Gender, Work and Sassen, S. 1991. The Global City: New York, London,
Space. New York: Routlcdge. Tokyo. Princeton, NJ: Princeton University Press.
Hashimoto, M., and L. Kochin. 1980. A bias in the Schneider, M., and T . Phelan. 1993. Black subur-
statistical estimation of the effects of discrimina- banization in the 1980s. Demography 30:269-79.

zyxwvuts
tion. Economic Inquiiy 18:478-86. Singell: L., and J. Lillydahl. 1986. An empirical
analysis of the commute to work patterns of males

zyxw
Ilolzer, H . 1991. T h e spatial mismatch hypothesis:
What has the evidence shown? Urban Studies and females in two-earner households. Urban
28:105-22. Studies 23:1119-29.
Ihlanfeldt, K. 1992. Job Accessibility and the Employ- Waldinger, R., and T. Bailey. 1991. The changing
ment and School Enrollment of Teenagers. Kalama- ethnichacia1 division of labor. In Dual City, Rc-
zoo: W. E. Upjohn. structuring New York, ed. J. Mollenkopf and M.
Ihlanfeldt, K., and D. Sjoquist. 1990. Job accessibil- Castells, 43-78. New York: Russell Sage Founda-
ity and racial differences in youth employment tion.
rates. American Economic Review 80:267-76. Wilson, W. J. 1987. The Truly Disadvantaged. Chi-
Johnson, J., Jr., M. Oliver, and L. Bobo. 1994. Un- cago: University of Chicago Press.
derstanding the contours of deepening urban in- Zax, J., and J. Kain. 1991. Commutes, quits and
equality: Theoretical underpinnings and research moves. Journal of Urban Economics 293153-65.
zyxwvutsz
zy
Zinn, M. 1989. Family, race and poverty in the
eighties. Signs: Journal of Women in Culture and
Race, Gender, and Spatial Segmentation

markets, spatial analysis, and the geography of health


and health care.
43 1

zyxwv
Sociery 14:856-74.
VALERIE P R E S T O N is Associate Professor in the
SARA McLAFFERTY is Associate Professor in the Department of Geography, York University, North
Department of Geography, Hunter College, 695 York, Ontario M3J 1P3, Canada. H e r research in-

z
Park Avenue, N e w York, NY 10021. H e r research terest include gender and urban labor markets, im-
interests include gender and race in urban labor migration, and urban social policy.

Race, Gender, and Spatial Segmentation in the Twin Cities+

Elvin K. Wyly
Rutgers University
zyxw
This study analyzes commuting trends in a relatively vibrant setting during the 1980s to determine (a) how labor
market segmentation correlates with differences in the spatial dimensions of local labor markets, and (b) whether
this link represents a direct spatial effect, independent of earnings, travel mode, and part-time work. I use 1980 and
1990 PUMS data to analyze changes in racial and gender divisions in the workforce, and I develop an estimate of
work trip distance to adjust for different travel modes. For all groups except white men, employment in a job
“typical” of one’s gender and racial group is associated with more localized commutes, hut this effect is strongly
mediated by variations in earnings and part-time work. Using a covariance structure model to control for these
effects, I find no independent link between segmentation and longer comniutes among African Americans. Earnings

zyxwv
and commute distances remained unchanged over the decade for African Americans, providing no evidence of a
purely spatial mismatch manifest in lengthening work trips without corresponding wage gains. T h e spatial dimen-
sions of an employment mismatch for inner-city minorities are concealed through the replacement of production
jobs by poorly paid service work in the expanding downtown economy of a vihrant regional center. Key Words:
spatial mismatch, labor market segmentation, commuting.

Introduction tween labor market segmentation and mis-


match, and the resurgence of interest in the

FL esearch on the economic restructuring of


ecent decades highlights a deepening ra-
cia1 polarization of the North American city,
spatial mismatch largely ignores the circum-
stances of minority women-despite a growing
body of evidence on the durability of gender
as evidenced by rising unemployment and pov- divisions in urban labor markets (Hanson and
erty among urban African Americans (Wilson Pratt 1988b, 1995; Tomaskovic-Devey 1993).
1987; Kasarda 1989; Galster 1991; Massey and Consequently, the relations between racial and
Denton 1993). This research has reopened de- gender divisions in the workforce and urban
bate over John Kain’s (1968) celebrated “spa- spatial restructuring remain unclear. This
tial mismatch” hypothesis, which correlates study analyzes changes in the spatial dimen-
worsening inner-city unemployment and pov- sions of labor market segmentation in a rela-
erty with residential segregation and intraur- tively vibrant setting, focusing on the commut-
ban locational shifts in employment growth. ing patterns of male and female African
Yet with few exceptions (McLafferty and Pre- American and white workers in the Minneapo-
ston 1992), analysts overlook the relations be- lis-St. Paul region during the 1980s.

Professional Geographer, 48(4) 1996, p a g e 43 -1zyxwvutsrqp


‘I thank John S . Adams and three anrinymour reviewers for vdluabk criticisms and suggestions on earlier versions of h s manuscript. This pa-
per is a suhstannally revised and expanded version of research presented a t the 1995 annual meeting of thc Asaociaoon of American Geogra-
phers.

0 Copyright 1996 by Association of American Geographers.


Initial submission, August 1995; revised submissions, April 1996, May 1996; final acceptance, May 1996.
Published by Blackwell Publishers, 238 Main Street, Cambridge, M A 02142, and 108 Cowley Road, Oxford, OX4 IJF, UK.

You might also like