Scientific Data Documentation
LMAs And Commuting Zone Codes
DSN: CC37.CODES.LMA.Y92
ACKNOWLEDGEMENTS
This documentation came from the following:
Pp. 69-79 in Singelmann, Joachim and Forrest A. Deseran (eds.),
Inequalities in Labor Market Areas. Boulder, CO: Westview
Press, 1993.
THE LOCAL LABOR MARKET CONCEPT
The local labor market is a key source of inequality. It encompasses a
mix of national, regional, and particularistic labor processes. The local
labor market also consists of an opportunity structure influenced in part by
national and regional macroeconomic forces and also by the local social and
economic organization of production. No matter how obvious it may seem,
it is nonetheless important to observe that two workers with the same human
capital characteristics can face profoundly different employment prospects in
different locales.
Local labor markets vary along many dimensions, including industry mix,
union density, presence of defense installations and/or contractors, and skill
levels of workers. Another critical aspect of area differentiation metropo-
litan-nonmetropolitan dimension, is of primary importance for our approach to
delineating local labor markets. Our objective is to develop a employ
meaningful designations for nonmetropolitan as well as metropolitan labor
market areas.
One major reason for our interest in nonmetropolitan labor market areas
is our appreciation of the wide range of rural economies and labor markets.
Indeed, it can be argued that U.S. rural economies exhibit greater variation
than do urban economies. As the proportion of the U.S. labor force involved
in agriculture has declined over the last half century, rural areas have come
to exhibit a great variety of employment scenarios. While some local rural
economies are healthy and robust, other areas are in serious decline. It is
critical that we understand more about these thriving and declining non-
metropolitan labor market areas.
ALTERNATIVE WAYS TO MAP SOCIAL AND ECONOMIC SPACE
Though a variety of ways to map social and economic space are available,
most of them harbor an urban bias that makes them less than satisfactory for
mapping all areas of the U.S. One of the most common measures is the
Metropolitan Statistical Area (MSA) defined by a committee of researchers
from a variety of federal agencies, including the Bureau of the Census, the
Bureau of Labor Statistics, and the Economic Research Service.(1) MSAs are
primarily defined on the basis of population size and density of counties or
county equivalents. Surrounding counties are included in MSAs if there is
sufficient commuting from the surrounding neighboring counties into the
central MSA county. In this sense, the MSA is a reasonable measure of the
local area in which people live and work. However, the MSA concept has a
critical drawback for examining the entire range of local labor markets:
the approach fails to account for about two-thirds of the land area of the
United States. Counties not included in MSAs are treated as one large
nonmetropolitan residual area. Since there is much greater variation among
rural areas than among metropolitan areas, MSAs are particularly unsatisfac-
tory as labor market proxies for rural areas.
Unlike MSAs, other measures of local areas that encompass the entire U.S.
exist. Like MSAs, however, these exhaustive area definitions also have an
urban bias. The BEA (Bureau of Economic Analysis) county groups, defined
originally by Berry (1968) and revised in the 1970s, are premised on central
place theory, which holds that economic activities tend to originate in large
urban areas and filter down the urban hierarchy from large cities to smaller
urban areas and out to the rural hinterlands (Christaller 1966; Losch 1980).
Following this concept, the BEA county groups begin with a large urban center
and aggregate surrounding (usually non-metropolitan) counties to that center.
In theory and in practice, the possibility of an exclusively nonmetropolitan
labor market area is precluded.
Another exhaustive measure, the Rand/McNally Basic Trade Areas (BTAs), is
based on newspaper circulation, physiography, population distribution,
economic activities, highway facilities, railroad services, suburban
transportation, and field reports of knowledgeable trade analysts. The BTAs
were designed to constitute trade areas, not local labor markets. While
persons might be willing to commute a considerable distance once a month or
even once a week to shop, they may be less willing to do so daily to work.
Like the urban bias in the BEA approach, the Rand/McNally delineation
typically contains an urban center and draws in surrounding counties to that
center of economic gravity.
(1)The Office of Management and Budget and the Bureau of the Census are
sponsoring a "Metropolitan Concepts and Statistics Project" for the year
2000. One aim of this project is to expand the MSA concept to include
standard statistical areas for all parts of the country (Adams 1991; Berry
1991; Frey and Speare 1991; Morrill 1991).
COMMUTING ZONES
Our approach to defining local labor markets has the distinct advantage
of minimizing urban bias. We begin with counties as basic building blocks of
local labor market areas, as do those who construct MSAs, BEA county groups,
and BTAs. Initially, we aggregate counties solely on the basis of commuting
ties; we are not concerned with absolute population size until later in the
process. Moreover, we are not concerned with state boundaries. Counties are
grouped without regard for state lines.
We begin with a county-by-county matrix of place of residence by place of
work, similar to the sample matrix shown in Table 5.1. In this matrix, the
diagonal represents the number of people who live and work in the same county.
The off-diagonal cells represent the number of people who live in one county,
but work in another. For example, a cell in the upper half of the sample
matrix shows that 857 people live in county A and work in county B, and the
corresponding cell in the lower half of the matrix shows 1,827 people living
in county B and working in county A.
Table 5.1 Frequency Matrix for Nine Sample Counties
County of Residence A B C D E
A 244,327 857 1,261 300 168
B 1,827 45,188 125 0 0
C 3,826 121 14,709 0 12
D 1,808 0 175 16,343 216
E 5,132 14 103 0 5,775
F 6,668 22 7 0 99
G 760 0 0 0 11
H 12,214 59 46 0 0
I 836 576 0 0 0
County of Residence F G H I TOTAL
A 752 390 3,216 450 251,721
B 0 0 0 182 47,322
C 0 18 18 0 18,777
D 0 0 0 0 18,542
E 152 28 28 0 11,209
F 5,717 190 181 0 12,884
G 292 19,640 439 0 21,142
H 230 378 11,679 0 24,606
I 0 0 446 2,735 4,593
Unlike the use of commuting data in the MSA approach, we include commuters
in both halves of the matrix in our measure of county commuting ties or two-
way commuting flows. This focus on two-way commuting lows reduces the urban
bias typically found in other measures of local social and economic space,
Conceptually, the two-way flow treats the relationship between counties as one
of reciprocity rather than dependence.
Our strategy for mapping social and economic space is to cluster counties
together on the basis of the strength of these two-way commuting ties. We
measure the strength of these two-way commuting ties with a proportional flow
measure:
(commuters from county i to j) + (commuters from county j to i)
------------------------------------------
(resident labor force of smaller county)
By adjusting the number of commuters by the size of the labor force in the
smaIler county, we ensure that no large county will dominate the analysis
simply because of its absolute size.(2)
This proportional flow measure is computed for all county pairs in the
matrix.(3) The proportional flow measures are then analyzed using a hierar-
chical clustering technique. Cluster analysis is an iterative process that
successively groups counties into larger and larger aggregates based on the
similarity (or dissimilarity) between the counties.
In our case, the proportional flow measure described above provides a
measure of similarity between counties. Counties with the strongest com-
puting ties are grouped together first; with each successive iteration, addi-
tional counties are either clustered together to make a new grouping, or add-
ed to previously formed county groups. Eventually, the cluster analysis
algorithm groups all of the counties together into a single cluster.
Counties that cluster together in early iterations are more strongly
linked than are counties brought into clusters at the final stages of the
process. Cluster analysis, however, does not provide statistical tests to
determine appropriate cut-off points among sets of commuting ties. It does
provide an average normalized distance measure for each stage in the itera-
tion. In our analysis of 1980 commuting flows (Tolbert and Killian 1987), we
used this distance measure to ensure that all clusters had at least the same
minimum commuting linkage. It is important to note that this does not mean
that all clusters have equally strong commuting ties.
Counties that cluster together before a given iteration in the analysis
are referred to as Commuting Zones (or CZs). Counties that fail to cluster
before our cut-off stage are referred to as commuting isolates. In our work
with journey-to-work data from the 1980 census, this procedure resulted in 763
(764 counting Alaska(6)) Commuting Zones.
These CZs are notably smaller than either the BEA county groups or the
Rand/McNally Basic Trade Areas. And with the exception of Alaska, they have
considerable face validity in terms of mapping the areas in which people live
and work. Moreover, over 500 of the CZs do not contain any metropolitan
counties, thus emphasizing the diversity found in nonmetropolitan America.
(2)We experimented with a variety of possible measures of the commuting
flows, especially focusing on measures that controlled for either the size of
the larger county or for the sum of the two counties' resident labor forces.
In cases where one county was significantly larger than the other, these two
alternative measures produced quite similar results. Both of these alterna-
tives tended to draw smaller counties into more distant large metropolitan
areas and away from slightly smaller population centers, reproducing the
central place model of social and economic counties that are more likely to
be independent of major cities.
(3)To aid in computation, several large matrices for different regions of
the U.S. were analyzed. Details of the original data preparations can be
found in Tolbert and Killian (1987).
LABOR MARKET AREAS
One major reason for incorporating space into our conceptual models of
social and economic inequality is the assumption that the structural context
of the local labor market will have a unique, independent impact on individual
social and economic processes. To include these local labor markets into our
empirical statistical models of inequality, we must have access to
individual-level data that include information about each individual's geo-
graphic location. Several individual-level survey data sets do include some
information about where people live, especially if they reside in very large
urban area. However, survey data samples are not typically large enough
samples to permit evaluation of the distinct effects of place, especially in
the case of individuals who live in nonmetropolitan areas and whose place of
residence is usually characterized only as rural. Therefore, in identifying
local labor market areas, we want to construct a geography that can be used
with large data files that permit detail on individuals and their places of
residence and places of work. Our delineation of labor market areas with
the 1980 census (Tolbert and Killian 1987) was carried out with the intent of
acquiring a public use microdata sample (PUMS) from the Bureau of the
Census.(4)
For confidentiality purposes, the Bureau of the Census requires that any
geographic identifiers on the PUMS must contain a minimum of 100,000
persons. To use the labor market geography with the PUMS, we aggregated the
commuting zones and commuting isolates into 382 county groups with at least
100,000 population, referred to as Labor Market Areas or LMAs. Though the
population criterion is clearly arbitrary, Reynolds and Maki (1990) have
argued that establishing a minimum size prevents many small LMAs from domi-
nating cross-regional labor market analyses.
Whenever possible, we based this second-stage aggregation of commuting
zones on the single strongest commuting relationship. In some cases, counties
were truly isolated with essentially no commuting to other counties. We
aggregated these counties based on a visual inspection of physical proximity
and the lack of any obvious physiographic barriers. In other cases,
especially in metropolitan areas, the first-stage commuting zones were large
enough to satisfy the population criterion. Unless surrounding commuting
zones were too small, we treated these larger areas as free-standing LMAs.
These same procedures were subsequently used by Steahr (1990) to develop
detailed labor market areas for New England. Since the 1980 journey-to-work
data were released for New England in minor civil division (MCD) detail,
Steahr was able to work with subcounty commuting data. His results do not
adhere to the 100,000 population criterion and cannot be used with individual-
level census data. Nonetheless, Steahr's work serves as an indication of the
utility of the labor market area methodology.
(4)A consortium of land-grant universities, other universities, regional
development centers, and the U.S. Department of Agriculture contributed to the
development efforts. The resulting PUMS, Public Use Microdata Sample D, is
available from the Bureau or from the Inter-University Consortium for
Political and Social at the University of Michigan.
CZS AND LMAS IN PERSPECTIVE: 1990 AND BEYOND
The commuting zones and labor market areas developed with the 1980 census
data have been employed by researchers in a variety of ways. The geography
has clearly been useful in both academic and applied settings, especially for
researchers interested in documenting the heterogeneity of rural areas in the
United States.
The S-229 U.S.D.A. regional research project plans a replication of the
labor market and commuting zone delineation for the 1990 census journey-
to-work data that are scheduled to be available in 1993. Though we expect
few if any procedural changes in the delineation approach, we will be very
interested in the extent of correspondence between the 1980 and 1990 map-
ping results. Those labor market areas that consist of the same counties in
1980 and 1990 will provide intriguing longitudinal research opportunities.
Those labor market areas that have changed in county composition will
challenge us to understand the macroeconomic and social bases of those
realignments. Changes in the socioeconomic well-being of residents of
changing market areas will also be of interest.
In anticipating the latter sort of question-i.e., the influence of the mar-
ket area on households and individuals-we very much hope to see a 1990
equivalent of the 1980 Public Use Microdata Sample D, which contained the 1980
labor market delineation. Under the auspices of the U.S.D.A. S-229 Technical
Committee, we are developing a proposal to obtain a public use microdata
sample of the 1990 census data. Though such a 1990 data file alone would be an
invaluable resource, the juxtaposition of the 1990 data file and geography
with the 1980 data and geography holds much research promise.
REFERENCES
1. Adams, John S. 1991. "An Examination of Conceptual Issues and Proposals for
New Approaches to Classifying Settled Areas of the United States." Preli-
minary report to the Metropolitan Concepts and Statistics Project, U.S.
Bureau of the Census.
2. Berry, Brian J. 1968. "Metropolitan Area Definition: A Re-Evaluation of
Concept and Statistical Practice." Working Paper Number 28, U.S. Bureau
of the Census, Washington, D.C.
1991. "Capturing Evolving Realities: Statistical Areas for the
American Future." Preliminary report to the Metropolitan Concepts and
Statistics Project, U.S. Bureau of the Census.
3. Christaller, W. 1966. C. W. Baskin(trans.), Central Places in Southern
Germany. Englewood Cliffs, NJ: Prentice-Hall.
4. Frey, William H., and Alden Speare, Jr. 1991. "Metropolitan Areas as
Functional Communities." Preliminary report to the Metropolitan Concepts
and Statistics Project, U.S. Bureau of the Census.
5. Losch, A. 1980. In William H. Woglom (trans.), The Economics of Location. New
Haven: Yale University Press.
6. Morrill, Richard. 1991. "Report on Metropolitan Concepts and Statistics."
Preliminary report to the Metropolitan Concepts and Statistics Project,
U.S. Bureau of the Census.
7. Reynolds, Paul D., and Wilber R. Maki. 1990. "Business Volatility and
Economic Growth." Final project report submitted to the U.S. Small
Business Administration. Minneapolis: Regional Economic Development
Associates, Inc.
8. Steahr, Thomas E. 1990. "Local Labor Market Areas in New England." Storrs,
CT: Department of Agricultural Economics, University of Connecticut.
9. Tolbert, Charles M., and M. S. Killian. 1987. "Labor Market Areas for the
United States." Staff Report No. AGES870721. Agriculture and Rural
Economy Division, Economic Research Service, U.S. Department of Agri-
culture.