Appendix B
The Derivation of Jobs-to-Workers Ratios1
This paper was prepared by Virginia Carlson under a Cooperative Agreement between HUD and the School of Architecture and Urban Planning of the University of Wisconsin-Milwaukee. Essentially, Carlson's work picks up where Bania and Leete left off; it uses their zip code level occupational job estimates and locations as input data to develop ratios of job seekers to jobs within neighborhoods of Cleveland, Columbus, and Toledo. To do this, Carlson determines the number of jobs available to welfare recipients based on where they live, where jobs are located, and how many other people are likely competing for the same jobs. A distinction is made between jobs that are "accessible" within commuting distance and jobs that are "available" taking competition into account. Carlson employs GIS technology in order to generate commuting ring patterns for mandated residents and their competitors. Section IV of this report relies heavily on Carlson's work.
There are several intermediate steps that Carlson develops, in conjunction with HUD, in order to reach a bottom-line. They are: determining the number and location of public housing residents affected by welfare reform; determining the number and location of all welfare recipients within an MSA; determining the number and location of other low-skilled competitors for entry-level jobs; determining the number and type of available jobs; and determining typical commuting patterns within an area.
Of special note is that Carlson's work offers important advancements to the knowledge and methodology of labor market analysis especially through the use of place-specific job and worker variables as overlays onto geographic boundary files and through the accounting for activity in nearby geographies (including the number of competitors contained therein). Of note also is that the use of sophisticated GIS software along with the Census of Transportation Planning Package (CTPP) permitted a degree of specification of commuting patterns not achieved elsewhere in the literature. Moreover, previous work in this area rarely, if ever, accounted for the effects of competition for jobs among individuals in nearby communities or in more suburban locations. Finally, Carlson employs both "distance ring" and CTPP contours to estimate job seeker-to-job ratios. The use of CTPP contours permits calculations of commute zones based on existing transportation infrastructure and tabulations of places to which people within specified areas actually do travel.
This project posed a unique challenge -- to estimate job availability for welfare recipients given several parameters: the location and number of individuals on assistance, low-skilled labor market competitors, the number and type of jobs of available, and typical commuting patterns. Previous work on job availability in the context of welfare reform has been very thorough with regard to careful estimation of appropriately-skilled jobs, and with accounting for labor market competition (Carlson and Theodore 1995, Kleppner and Theodore 1997, Steuernagel 1995). However, less research has been done which makes more geographically precise estimates by considering the match between the location of jobs and typical commuting sheds for workers. This paper first looks at the means by which recipients, competitors, and jobs are estimated, then turns to a discussion of the geographic derivation of workers-to-jobs ratios.
Geography, Commuting, and Job Seekers-to-Jobs Ratios
The goal was to develop seekers-to-jobs ratios for individuals in neighborhoods. How many jobs are available to a typical TANF recipient in a given neighborhood, taking into account jobs available in their commuting reach, and given that there is competition for jobs from individuals in their neighborhood as well as other neighborhoods? Geographic specification comes into play here both in the definition of jobs available within a commute, and in the identification of possible competitors. The derivation of our ratios required five steps: the definition of neighborhoods, the definition of neighborhood commute zones, identification of jobs within commute zones for each neighborhood, the specification of job seekers, and the final calculations of ratios.
It should noted that the methodology discussed here represents a unique application of Geographic Information Systems (GIS) analysis. The complex geographic analysis presented here would not have been possible without the use of a sophisticated spatial analysis software. ARC/INFO 7.11 for NT, a (GIS) package developed by Environmental Research Systems Institute, Inc. (ESRI) of Redlands, CA, was used for geocoding and mapping. The extensive subsetting and overlaying of geographic coverages with attached jobs and worker variables, including the identification of associated and nearby geographies, represents a sophistication in job availability analysis not performed previously.
Specifically, the analysis for neighborhoods counts neighborhood jobs seekers and accessible, but also takes into account the fact that jobs ultimately available to community residents also depends on labor market activity in nearby neighborhoods. In addition, this paper employs a more sophisticated method of estimating commuting patterns than has been used in previous job availability studies. With the use of the Census of Transportation Planning Package, the methodology is better able to capture existing transportation infrastructure.
The Metropolitan Areas and their Neighborhoods
Three metropolitan areas were examined: the Cleveland MSA (Geauga, Lake, Lorain, Median, and Cuyahoga counties); Columbus metropolitan area (Franklin, Licking, Fairfield, Pickaway, Delaware and Madison counties) and Toledo (Lucas, Sandusky, Wood and Fulton counties). In each area, neighborhoods are defined at the zip code level. This resulted in 98 neighborhoods in Cleveland,2 101 neighborhoods in Columbus and 58 neighborhoods in Toledo. Zip codes were chosen for several reasons. Zip code boundary files were readily available for use with our spatial analysis software. In addition, input data were zip-code friendly: the jobs data, as discussed above, were generated for zip codes; and the job seekers data (recipients and competitors) could be easily converted from census tracts to zip codes. Zip codes are a standard method of defining neighborhoods in much social science research.
Commute Zones
A search of the existing literature revealed one previous attempt at estimating worker/jobs ratios at the neighborhood level that used commuting zones (Henle and Kinsella 1996). However, Henle and Kinsella used the standard "distance rings" or "trade areas" method to draw a typical commuting zone around individual communities. Used widely in retail market research, such distance rings consist of a geographic boundary drawn from a central point (in Henle and Kinsella's research, the center of the neighborhood) where the distance from the central point to the boundary is based on consumers' typical driving or commuting distance. The activity located at the central point can expect to draw customers from the area inside this distance ring. Conversely, for this paper, the boundary can be drawn using typical work-travel distances from the central point. The area within the circle thus inscribed is an estimate of all places that individuals located at the central point can be expected to be able to reach for available jobs. This area is referred to as a commute zone. Distance rings analysis was performed for Cleveland, Columbus and Toledo, as mentioned above. Analysis was performed for typical commuting distances for public transportation and for automobile separately.
Although distance rings are an acceptable method of estimating typical commute zones around a central point, such rings do not take into account travel obstacles and opportunities afforded by highways, special bus routes and other "real" transportation infrastructure (Peterson 1997). In order to take the existing travel system into account, this paper uses information on commuting patterns available from the Census of Transportation Planning Package (CTPP), for a special analysis of Cleveland.
Twelve sets of workers-to-jobs ratios were obtained for this project using these two definitions of commute zones (and two definitions of jobs). In Cleveland, four sets of ratios were obtained, using both CTPP-defined commute zones and distance ring-defined commute zones, where each of these rings was drawn twice: once using public transportation distances and once using auto distances. Ratios were calculated using the definition of "female-dominated entry-level jobs" as discussed previously. Four sets were also obtained for Toledo and Columbus. In these two cities, only distance-ring commute zones were drawn, once for public transportation distances and once for autos. However, the calculation of ratios was done twice, once using "all entry-level" jobs and once using "female-dominated" jobs.
Although the specific methodology for distance rings and CTPP commuting contours will be discussed separately below, there are some elements common to both methods which can be explained here. First, after an examination of typical travel times for low-skilled workers in several metropolitan areas and a general discussion among researchers involved in the project, an expected commuting time of 45 minutes was established. This commute time was used both for travel by public transportation and travel by auto.
Second, an explanation of the structure of the CTPP file will help clarify some of the discussion found below. The geographic unit of analysis for the CTPP is called a Traffic Analysis Zone (TAZ). Depending upon population in and trips generated from, TAZs can vary in size from half-square miles to four or five square miles. The information used here is in the form of TAZ pairs: information about travel modes and times are given for pairs of TAZs, consisting of the TAZ of origin and the TAZ of destination.
Finally, the CTPP file does not give information for all public transportation modes as a single item. Public transportation is split between bus, train, trolley, etc. The public transportation analysis here is, therefore, averages for bus and train times in the Cleveland CMSA, and for bus only in Columbus and Toledo.
Distance Rings
The drawing of distance rings required the definition of an appropriate number of miles for the radius, given typical work-travel distances. The CTPP is used for this. Distances between the central points of TAZs were calculated using a State Plane coordinate system based on distances measured in feet. Then, all origin/destination pairs were selected for which median travel time between was 45 minutes (two sets in each city -- one for public transportation and one for autos). These pairs were then used to find the median distance for travel times of 45 minutes.3 A ring was then drawn around every neighborhood. The edge of the zip code was used as the starting point for measuring the ring, so that the ring is the same shape as the zip code. The distance between the zip code edge and the commute zone boundary is the distance given by the median commute distance for 45 minutes.
An argument could be made that the distance ring should have been based on the maximum distance traveled within 45 minutes rather than a median. The median was chosen, however, because a wide range of travel distances was observed and it was decided that commute boundaries should not reflect one unusually accessible situation. In this, the median distance may underestimate the full range of accessibility for some neighborhoods. However, the choice was made to use the edge of the zip code rather than a central point from which to begin drawing the commute zone in order to partially compensate for this possibility.
Census of Transportation
Planning Package (CTPP) Contours
As mentioned above, a CTPP-contours analysis attempts to account for existing transportation infrastructure, which is overlooked by a simple distance-rings analysis. Since the CTPP reports time and distance traveled between TAZs, it is possible to draw commute zones based on places to which persons from particular communities really do travel. In essence, a CTPP-derived commute zone can be thought of as "fingers" emanating from a neighborhood rather than the "area ring" surrounding a neighborhood one obtains from a distance ring procedure. These finger contours arise because not all areas within a theoretical ring are within the same commuting time. Highways, arterial streets, and bus schedules make some areas more accessible than others.
A CTPP-contour commute zone was drawn for every neighborhood, where the contour is defined as all TAZs accessible in 45 minutes or less for a given mode of transportation (public transportation or auto). To do this, zip-code boundary files were overlaid with TAZ boundary files so that TAZs within zip-code neighborhoods could be identified. The neighborhood was thus then defined as a set of origin TAZs. All destination TAZs within the 45 minute limit were then selected by our spatial analysis software to be part of the CTPP-contour commute zone for that neighborhood.
Distance Rings vs. The Census
Of Transportation Planning Package
Although a "fingers vs. circle" analogy can begin to describe the way in which a CTPP-contour commute zone differs from a distance-ring commute zone, an examination of the actual CTPP contours shows that there are many "holes" or empty areas in the "fingers." A destination TAZ does not get counted as part of a neighborhood's commute zone if no one from that neighborhood commutes to the TAZ for work. What is suggested, therefore, is that the CTPP reflects not only the existing transportation infrastructure, but also the outcomes of complex social and labor market processes. In this, "holes" may exist for several reasons. Jobs may not be located in the empty areas. Or, the jobs may be of a nature such that few or none of the neighborhood residents hold those jobs. It may be also that businesses in these areas may not have historically hired residents from neighborhoods containing public housing residents few or none of neighborhood residents hold those jobs.
Thus, these empty areas found within the commute zones delivered by the CTPP method suggests that a CTPP-based analysis takes into account not only existing transportation infrastructure, but also the historical operations of labor markets and the social nature of the employer-employee relationship. Therefore, it may be that the CTPP offers a more rigorous method by which to specify typical; commute zones for neighborhoods. A distance-rings analysis may indicate what jobs are nearby, but cannot account for transportation nodes, micro-level locations of specific occupational niches, or for the geographical scope of residents' historical job search activity and success.4
Identification of Accessible Jobs
But how many jobs were within each of these commute zones? Low-skilled jobs, as defined previously, were attached to the underlying GIS geography in order to permit identification of such jobs within commute zones. Although job totals are given for zip codes, commute zone boundaries split these zip codes inasmuch as residents commuted only halfway or so "into" a zip code. To compensate, zip code totals were converted into densities (total jobs/zip code area). As commute zones were drawn, either with distance rings or by overlaying destination TAZs, the software counted the total number of jobs in a given zone as given by the underlying densities. These densities differed in areas of the commute zone wherever the commute zone crossed zip code boundaries.
Identification of Job Seekers
As discussed above, competitors and TANF recipients make up the definition of job seekers. Each neighborhood (zip code) was assigned a number of job seekers according to the method discussed previously.
Calculation of Seekers/Available Jobs Ratios
As may be apparent, commute zones for neighborhoods overlap. Zip codes are contiguous, so that drawing commute zones for all neighborhoods results in a series of overlapping polygons, creating sub-polygons. These sub-polygons represent areas where two or more neighborhoods each have a claim on jobs. In the CTPP-based commute zones, these sub-polygons are destination TAZs.
That is, not only do job seekers in one community face competition for jobs in their commute zone from fellow neighborhood residents, but jobs in their commute zone may be "claimed" by individuals in other neighborhoods whose commute zone overlaps their own. These other claimants must be taken into consideration when determining how many of the "accessible" jobs are actually "available" to residents of a particular neighborhood. If not, and all jobs in a commute zone are attributed to a neighborhood, jobs in areas where commute zones overlap will be counted more than once. The GIS-basis of our analysis permitted allocation of jobs in a manner not previously attempted. Henle and Kinsella's analysis using commute zones merely normalized the resulting total number of jobs attributed to all communities by giving each neighborhood a proportion of existing jobs based on its proportion of the total of double-counted jobs. Although an improvement over job double-counting, what this method does not do is to account for variation in the number of seekers across neighborhoods. For example, more claimants may be found two neighborhoods away than in the neighborhood next door. This may be either because there are more people in the further neighborhood, or because there are fewer jobs in that neighborhood's commute zone and so residents claim more jobs further away from home.
Instead, an algorithm was developed that takes into account variation in claimants and in job availability across neighborhood commute zones. In essence, each sub-polygon (for distance rings) or TAZ (for the CTPP) in the metropolitan area was assigned a number of jobs (as explained above in "Identification of Accessible Jobs"). Then, all the neighborhoods that had that sub-polygon or TAZ in their commute zone were identified. The jobs in the smaller areas were then allocated as available to a respective neighborhood based on relative concentrations of job seekers and accessible jobs.
Total jobs allocated as available to the ith neighborhood is the sum of jobs allocated from each TAZ or sub-polygon in its commute zone:
Ai = S Ain ,
n=1
where Ain = jobs from the nth TAZ or sub-polygon allocated as available to seekers from the ith neighborhood.
Job allocation from the nth TAZ (or sub-polygon) in a commute zone to seekers from the ith neighborhood is itself based on the seekers from the ith neighborhood allocated to the TAZ or sub-polygon as a proportion of seekers from all neighborhoods allocated to the TAZ or sub-polygon:
Ain = Jn * Bin/Bn ,
where Jn = total jobs in TAZ n,
Bin = seekers from the ith neighborhood allocated to the nth TAZ or sub-polygon in their commute zone, and
Bn = total seekers, from all neighborhoods, allocated to the nth TAZ or sub-polygon.
Seekers from the ith neighborhood allocated to the nth TAZ or sub-polygon is given by (total number of seekers from the ith neighborhood) * (total jobs in the nth TAZ/total jobs in the ith neighborhood's commute zone):
Bin = Ti * Jn/Ki ,
where Ti = total seekers from the ith neighborhood and
Ki = total jobs in all of the TAZ's or sub-polygons in the ith neighborhood's commute zone.
These ratios were calculated for entry-level "female jobs" as defined above. In Toledo and Columbus, ratios based on "all entry-level" were also calculated to serve as a comparison.
Endnotes
1 Virginia Carlson, Department of Urban Planning, The School of Architecture and Urban Planning, University of Wisconsin-Milwaukee.
2 The Cleveland MSA actually contains 147 zip codes. However, an analysis we performed using the Census of Transportation Planning Package (CTPP), discussed below, relied on the geography found in that data set. Numerous zip codes in the Cleveland metropolitan area are not found in the CTPP; for example, information on Ashtabula County, part of the metropolitan area, is not contained in the CTPP files.
3 These distances were as follows: Cleveland auto 8.52 miles, public transportation 4.69 miles; Columbus auto 9.57 miles, bus 4.73 miles; Toledo auto 10.42 miles, bus 5.22 miles.
4 The job seeker-to-jobs ratios generated by The CTPP method and distance rings analysis are autocorrelated at 81 percent (Pearson's r=.9).
References
Carlson, Virginia L. and Nikolas C. Theodore, 1995. Are There Enough Jobs?: Welfare Reform and Labor Market Reality. DeKalb, IL: Office for Social Policy research, Northern Illinois University.
Kleppner, Paul and Nikolas C. Theodore, 1997. Work After Welfare: Is the Midwest's Booming Economy Creating Enough Jobs? DeKalb, IL: Office for Social Policy Research, Northern Illinois University.
Peterson, Keith, 1997. "A Trade Area Primer." Business Geographics (September), Fort Collins, CO: Geoplace.
Steuernagel, Bruce J., 1995. The Job Gap Study. St. Paul, MN: JOBS NOW Coalition.
|
| 
