ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”. Bangkok, Mav 23-25. 2001 
is defined as the central contiguous urbanized area. This 
leaves out only a small piece of urbanized area in the 
southeast corner of Medina County, which is fairly 
independent based on its commuting pattern. The 
contiguous urbanized area includes 882 TAZs, accounting 
for 89.3% of the population, 88.8% of resident workers and 
92.6% jobs in the five-county region in 1990. 
MEASURING JOB PROXIMITY 
An index of job proximity is proposed to measure the spatial 
separation between a worker’s residence and all suitable 
jobs for him/her. Two points are critical for the index design. 
One is the emphasis of physical distance (in contrast to 
behavioral distance) between workers and jobs without 
consideration of other socioeconomic barriers. This index 
intends to capture the essence of traditional perception of 
locational advantage. The other is the concept of suitable 
jobs. For instance, a job opening of computer specialist is 
meaningless to one without appropriate education 
background. Identifying suitable jobs for workers is a 
challenging task. One approach is the so-called occupation- 
industry (O-l) methodology used by researchers to identify 
entry-level jobs for the welfare-to-work transition (e.g., Sen 
et al., 1999; Leete et al., 2000; Sawicki and Moody, 2000). 
The other approach is based on wage (e.g., Shen, 1998; 
Immergluck, 1998). The O-l method requires extensive 
detailed data such as jobs in each occupational category by 
industry. For instance, Sen et al. (1999) use over 150 
occupations across all industries for identifying entry-level 
jobs. Such information is not available in the CTPP data. 
This research adopts the wage-based approach and 
considers jobs within the same wage group of workers as 
the suitable ones. 
(1) Basic Model 
Among all suitable jobs for a resident worker, a job nearby is 
more likely to be taken than a remote one. Similar to the 
rationale of Huff’s (1963) model, the probability or the portion 
of resident workers at a TAZ i taking jobs at a TAZ j (P M ) is 
determined by the influence of jobs at j among the influences 
of all jobs at various locations: 
J .Id ? 
p _ J U 
k-l 
where J is the number of suitable jobs in a TAZ, p is the 
distance friction 1 , and n the total number of TAZs. Since our 
interest is the physical distance between residents and jobs, 
d is the straight-line aerial distance between two TAZs. 
Job proximity for TAZ i (Dj) is defined as the average of 
distances between this TAZ (as origin) and all job sites (as 
destinations) weighted by Pij: 
7=1 
Adding a superscript g to index jobs of a particular wage 
group (J 9 ) and rearranging the above equation, it can be 
written as 
' Like any gravity-based model, it is important to 
determine the distance friction 0. A regression on the CTPP 
journey-to-work data yields 0=0.7422, based on a simple 
model such as T- tj — aW ; J ■d¡j ^ , where T is the number 
of commuters between residence location i with Wj workers 
and job location j with J, jobs. 
D; =- 
1 
7=1 
jKd~ ß 
J ‘J 
(1) 
Equation (1) computes the proximity to jobs of wage group g 
for workers of the same wage group. The larger the value of 
D, the farther the resident workers are located away from 
their suitable jobs, thus a less advantaged location. 
(2) Illustrative Cases 
A special case is used to help interpret the index. When 0=1, 
and denoting the total number of jobs in the study area as N, 
equation (1) can be rearranged as 
Df 1 =(J ] /N)d- l + U 2 fN)d7 2 l +...-k(J n IN)dr n [ . (2) 
The inverse of distance, a gravity kernel, is used as a scalar 
factor for jobs (i.e., a close job is more influential than a 
remote one). Therefore, the inverse of job proximity at i (i.e., 
aggregated influences of all jobs on i) is the average of 
influences of individual job sites weighted by the number of 
jobs there. 
Why do not we simply define job proximity as the average of 
distances weighted by the number of jobs from resident 
workers? In other words, 
D; = Ui / N)d n + {J 2 I N)d i2 +... + (J„ / N)d in . (3) 
We use a simplified example to explain the difference. In 
Figure 2, two job sites I and II are 102 km apart and each 
has 10 suitable jobs. A resident worker A is 100 km from site 
I and just 2 km from site II. A resident worker B are located 
in the middle with 51 km from each job site. We would 
expect worker A to enjoy a better job proximity than worker B 
since job site II is just 2 km away. However, equation (3) 
would yield the same job proximity for both workers (51). 
Equation (2) confirms our expectation with a far better job 
proximity for worker A (3.92) than that for worker B (51), 
confirming the expectation. A close job is weighted more 
than a remote one in equation (2) whereas equation (3) does 
not distinguish them. 
(3) Overall Job Proximity 
For this research, we segment the workers or jobs into five 
wage groups. See Table 1. The wage groups are designed 
to allow a reasonable wide range for defining suitable jobs 
for workers in each group. 
Table 1. Number of Workers in Five Wage Groups 
(Cleveland, 1990) 
Group 
Name 
Wage 
Range 
No. 
Workers 
Percentage 
1 
Low wage 
less than 
$10,000 
215,542 
23.40 
II 
Lower- 
medium 
wage 
$10,000- 
20,000 
239,468 
26.00 
III 
Medium 
wage 
$20,000- 
30,000 
190,220 
20.65 
IV 
Upper- 
medium 
wage 
$30,000- 
50,000 
196,966 
21.39 
V 
High wage 
more than 
$50,000 
78,799 
8.56 
A TAZ is usually composed of workers of various wage 
groups. The overall (total) job proximity at a TAZ i 
291