THE EVOLUTION OF INDUSTRIAL LAND USE WITHIN
THE KNOXVILLE METROPOLITAN REGION: AN ANALYSIS
USING AERIAL PHOTOGRAPHY AND HISTORICAL DATA
FOR THE PURPOSE OF LAND USE MODELING
ROBERT B. HONEA
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
Dr. (Jim) James C. Wilkinson
Of the many individuals who have contributed to this study,
foremost is Dr. Clark I. Cross, Chairman of my Supervisory Conurttee.
I am appreciative for his professional and personal assistance. His
periodic encouragement has been most opportune. Dr. Gary Shannon has
contributed both to this study and at other stages of my graduate
career. I am grateful for his help. Dr. James R. Anderson has been a
source of encouragement before and during my graduate tenure. To him
I extend my appreciation. To Dr. David L. Niddrie 1 extend my appre-
ciation for many engaging discussions. His intellectual vigor is a
source of professional inspiration. The lectures of Dr. Clayton
Curti. provided many ideas which arc :c uded in tniS study. To him
my sincere appreciation is extended. Gratitude is also extended to
Dr. Sh3nnon McCune, Chairman of 'he De'partment of Geography and other
menibers of the faculty for their assistance.
who contributed to the research effort: Mr. C. T. N. (Ted) Paludan at
Marshall Space Flight Center; Dr. C. W. (Pete) Craven, Mr. Richard C.
DurCee, anid Dr. Michael M. McCarthy at Oak' Ridge National La:boratory;
rand > i~mcT.'ily ,.'ho patiently occupied tlhe remainder -f ::y huro while i.
hibern ated in my office.
TABLE OF CONTENTS
ACKNOWLEGEMENTS . . . . .
LIS" OF TABLES . . . . . .
LIST OF ILLUSTRATIONS. . . . .
ABSTRACT . . . . . .
I. INTRODUCTION . . . . .
- . i-i
. . . . . . . vii.
. . . . . . ix
Problem Background ..
Statement of Problem . .
Model Operation . . .
Motivation for Study . .
Organization of Study. .
II. THEORETICAL AND EMPIRICAL ST'IDIES RELEVANT 10 'THE
DEVELOPMENT OF INDUSTRIAL SITE-SELECTION ALGORITHM . .
Selected Theoretical Works . . .
'er IsolerLe Stat . . . . ..
lUber den Standort der Iridustr -cn .. . .....
The Location of Economiic Rehavior. ..
Imperfect Competition Ir.dI the Divlnonl., neb,) pr ..
'ihe Theories of David M. Smith aiu
Melvin L. Greenhut . . . . . .....
Empirical Studies . . . .
The Studies of Allen Prod and Richard Londale .
Historical Studies of Industrial Location Factors.
Land Use Modeling Studies. . . . . . . . .
4A icrtative List of Variables. ..
III. DESCRIPTION UF THE STUDY AREA AND ANALYSIS PROCEDURES. .
The Study Region . . . . . . .....
Industry within the Region . . . . . ..
Food products industries (SIC 20) . ..
Tobacco products industries (SIC 21) . .
Textile mill product-; industries (SIC 22).
. . . . . .
S. . . . . .
Clothing and related products (SIC 23). .. . 58
Lumber and wood products except furniture
(SIC 24) .. . .. . . . . . 58
Furniture and fixtures industries (SIC 25) . 64
Paper and pulp products industries (SIC 26) . 68
Printing and publishing industries (SIC 27) . 68
Chemical products industries (SIC 28). . . . 6
Petroleum refining and paving and roofing
products (SIC 29). . . . . . . . 74
Rubber and plastic products industries (SIC 30). 74
Leather products industries (SIC 31) ..... 74
Stone, clay, and gland; products industries
(SIC 32) . . . . . . . . .. . 81
Primary metal industries (SIC 33). . . . . 81
Fabricated metal prodLcts industries (SIC 34). . 81
Machinery (except electrical) industries
(SIC 35) . .... . . . . . .. ... . 87
Electrical machinery, equipment, and supplies
industries (SIC 36). ... ... . . . 87
Transportation equipment industries (SIC 37) .. 94
Instrument and related products industries
(SIC 38) .. . ...... . ...... 94
Miscellaneous industries (SIC 39). ... .... 94
Data Collection. . . . . . ... . . . . 101
Statistical Procedures . . . . . . . 106
IV. EMPIRICAL ANALYSIS . . . . . . . . . . ll
Descriptive Statistic' . .. . .. . I.1
Description of Sample. . . . .. ... .. .. 113
Description of Variable Measurements .. ...... . 112
Results of the Factor Analysis .. .. .. ... 148
V. SUMMARY AND CONCLUSIONS . . . . . . .. 158
Summary. . ... . . . . . . . . . 158
Conclusions. .. . . . . . . . . . 159
Conclusions Regarding the Form of the Site-Selection
Algorithm . . . . . . ... . . .. 159
Conclusions Regarding the Proper Form of Model
Operation . . . . . . . . . . 160
Conclusions Regarding the Use of Aerial Photography
in Model Design. . . . . . . . . . 160
Conclusions Regarding Recommendations for Future
Research . . .. . . . . 161
RESEARCH BI LIOGRAPHY. . . . . . . . . 163
Periodicals ........ ................ 163
Books . . . . . . . . . . . . . . 168
Government Doci..,ients and Agency and Institute Research
Reports. . . . . . . . . . ..... 172
Unpublished Sources. . . . . . . . .... .177
A. Sample Survey Forms, . . ... .. . . . .. . 179
B. Map Overlays ........ .. ............. .182
BIOGRAPHICAL SKETCH . . . . . . ..... .. 184
LIST OF TABLES
1. Factors Most Influential in the Location Decisions
of Florida Industries, 1956-1957 . . . . . .
2. List of Possible Factors Influencing Industry Location
as Utilized in the Carrier and Schriver Survey . .
3. Ten Factors Most Frequently Mentioned by Tennessee
Firms as Affecting Location Decisions . . . .
4. Ten Location Factors with Highest Mean Number Points
Assigned by Tennessee Firms Interviewed. . . . .
Location Factors Suggested by Bullington . .
Tentative List of Variables . . ...
Expansion of ETDD Industries Between 1952-1973
Distribution of Sample by SIC Category . .
9. Variable -
i0. Variable -
L1. Variable -
12. Variable -
!3. Variable -
14. Variable -
15. Variable -
16, Variable -
i7. Variable -
[8. Variable -
[9. Variable -
20. Variable -
21. Variable -
Slope of Land . . . . . . . . .
Drainage . . . . . . . . .
Clearing-Cover Condition . . . . .
Distance to Center of Town . . . . .
Distance to Major Throughfare . . . . .
Density of Land Use in Immnediate Vicinity . .
Rating of Price of Land . . . . . .
Proportion of Urban Area Within Two and a
Distance to Major Highway . . . . . .
Distance to Secondary Road . . . . .
Distance co Railway . . . . . . .
Distance to Waterway . . . ......
Distance to Airport . . . . . . .
Tab e Page
22. Variable Distance to Interstate. . . . . . . 131
23. Variable Oveiall Quality of Accessibility-Then . . 133
24. Variable Overall Quality of Accessibility-Now. ... . 134
25. Variable City Water Availability . . . . ... 135
26. Variable City Sewage Availability. . . . . ... 135
27. Variable Gas Availability. . . . . .. . . 136
28. Variable -- Did Community Have Zoning Then? . ... . .136
29. Variable Was Site Zoned for Industry?. . . . .. 137
30. Variable Was Industry Already in Area?. . . . 137
31. Variable Overall Rating of r'oltigous Land
Use Compatability .... . . . . . ... 159
.'. Vari.able Condition of NeighL ohood . ... . . 140
33 Yariable Density of Land Use . ... .. .. .. ... 141
34. Variable Nearby Community Service. . . . . . .143
55. Variable Was the Site in an Industrial Park? ... .. 144
6. c.'" ibje Was was o Qu .ality of the i'a-?. . . . 1J,4
37. Variable Proximity to Knoxville. . . . . . ... 46
33. Variable Amount of Other Industry Located Nearby . 147
39. Correlation Matrix ... .... .. . . . 149
A(j. Eigen Values ..... . . . .. . . .. 152
41, Unrotated Factor Matrix. .. . . .. .. 154
42. Varimax Rotated Factor Matrix. . . . . . . .. 155
LIST OF FIGURES
Food Products (Plant It!:.ti
- Food Products (Employment)
- Textile Products (Plant Loc
- Textile Products (Employmen
Urban Land-Use Location Scheme. .
Smith's Model with Comparisons. .
East Tennessee Development District
Physiography . . . . .
Towns and Communities . . .
Plant Location 1943 . .
Employment 1943 . . . . .
Plant Location 1953 . ....
Employment 1953 . . . ....
Plant Locations 1963 .. .
Employment 1963 . . .
Plant Locations 1973. . ..
Employment 1973 . . . . .
Apparel and TRelared 'roduci. (Plant L.oc:ti
- Apparel and Related Products (Employricent).
- Lumber and Wood Products (Plant Locacions)
-Lumber and Wood Products (Employment). .
- Furniture and Fixtures (Plant Locations) .
- Furniture and Fixtures (Employment). . .
- Paper and Pulp Products (Plant Locations).
. . . . . 11
. . . . 17
. . . 40
. . . . . 42
. . . . . 43
. . . . . 45
. . . 46
S . . . . . 47
. . . . . 48
. . . 50
. . . . . 51
. . . .. 52
on,; . . . 56
. . . . 57
nations) . . . 59
t). . . . . 60
25. SIC 26 Paper and Pulp Products (Employnent) . . .. 70
26. SIC 27 Printing and Publishing (Plant Locations). .. 71
2-. SIC 27 Printing and Publishing (Employment) . . .. 72
28. SIC 28 Chemical Products (Plant Locations). . . ... 73
29. SIC 28 Chemical Products employmentt) . . . ... '75
30. SIC 29 Petroleum Refining and Paving and Roofing
Products (Plant Locations) ...... . 76
31. SIC 29 Petroleum Refining and Paving and Roofing
Products (Employment). . . . . . ... 77
32. SIC 30 Rubber and Plastic Products (Plant Locations) 78
33. SIC 30 R bber and Plastic Products (Fmploymet) 79
34. SIC 31 Leather Products (Plant Locations) . ... 80
35. SIC 31 -Leather Products (Employn.ent) . .. . .... 81
36. SIC 32 Stone, Clay and Glass Products (Plant
Locations) . . . . . . . . . 83
37. SIC 32 Stone, Clay and Glass Products (Employment). 84
38. STC 33 Primary cetal T.c'ustcies (Plant I.ocation) . 85
39. SIC 33 Primary Metal Industries (Employment).. . . 86
40. SIC 34 Fabricated Metal Products (Plant Locations). 88
41. SIC 34 Fabricated Metal Products (Employrment) .. .. 89
47. IC .'S MAchi nery ,Pia.nt Locations) ....... .. 90
43. SIC 35 Machinery (Employ.ent) . . . . 91
44. SIC 36 Electrical Machinery, Equipment and
Supplies (Plant Locations) . ... . ..... 92
45. SIC 36 Electrical Machinery, Equipment and
Supplies (Employment) . . . . . . 93
46. SIC 37 Transportation Equipment (Plant Locations) . 95
47. SIC 37 Transportation Equipment (Employment). ... . 96
48. SIC 38 Instruments and Related Products (Plant
Locations) . ... . . . . . . . 97
49. SIC 38 Instruments and Related Products (Employment). 98
50. SIC 39 Miscellaneous Industries (Plant Locations) . 99
51. SIC 39 Miscellaneous Industries (Employment). ... 100
52. 1958 Stereo Image of Beta-Tok Industrial Site .... .102
53. 1974 Stereo Image of Beta-'ek Industlial Site . .103
Abstract of Dissertation Presented to the Graduate Council
in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy
THE EVOLUTION OF INDUSTRIAL LAND USE WITHIN
THE KNOXVILLE METROPOLITAN REGION: AN ANALYSIS
USING AERIAL PHOTOGRAPHY AND HISTORICAL DATA
FOR T1HE PURPOSE OF LAND USE MODELING
Robert B. Honea
Chairman: Dr. Clark I. Cross
Major Department: Geography
This research effort represents a portion of a much larger
research goal, the identification of the determinants affecting land
use change. This study focuses, however, upon only one aspect of this
problem, the conversion of land to industrial use. The approach
assumes that the determinants of Lhe land use conversion process are
found in the "market place," where land transactions among buyers and
sellers occur. Only one side of the market transaction process is
studied, however, namely that of ite purchaser's desires (in this
case, the industrial developer) in securing an ideal or suitable site.
The problem was to identify the ideal qualities, quantities or attri-
butes desired in an industrial site and to formulate a general algo-
rithmic statement to identify potential industrial sites.
Research procedures involv'-!d ihe developing o7 a .ist of var:iabl es
previously noted in the literature to be related to industrial site
selection and streamlining the list to a set suitable for statistical
testing. A sample of 157 industrLes which have located (or relocated)
in the 16-county Knoxville metropolitan region since 1950 was selected
for analyst s. Ui current a&nd 1'slio-. al a:..i l. ,,hctrn.ra ch data
were collected on the location characteristics of each industrial
site. These data were then subjected to factor analysis to determine
the interrelations of variables, to minimize the list of variables
needed to describe the industrial site-selection process, and to
determine if the preconceived ideas concerning the factors affecting
the process were valid. Seven factors accounting for 72 percent of
the variance found in the original data were identified.
The 30 variables studied did not group as previously conceived
but nevertheless the factors present a logical arrangement of vari-
ables. Four factors are almost singularly identified by accessibility
characteristics. Reexamination of the accessibility factors indicates
they could be combined, thereby reducing the number of indexes to two
or three. The results suggest that a zoned or designated place for
industry to locate, open space suitable for industrial development,
and accessibility to the site acc the three most important considera-
tions in industrial site-selectio~i decisions. Any algorithm intended
to simulate industrial land use coriversion should be constructed with
these parameters as major components.
Most urban and regional planners seek to answer the question,
"What would happen if . .?" The answer is not simply determined.
"Even though equipped with the best information system, no . .
decision maker, however adept, can anticipate or visualize the many
relationships which must be considered in a comprehensive, regional
plan."1 If, however, we could specify the correct development pro-
cesses and represent them symbolically as a set of equations, the
computer can approximate through simulation, years of regional
growth in a matter of minutes. Before we can accurately simulate
regional growth, the development processes must first be defined and
understood, and by observing the limited success of previous land use
modeling efforts we become acutely aware that we do not sufficiently
1Charles R. Meyers, Jr., "New Tools for Regional Planning," AIA
Journal, October, 1971, p. 1.
2Models should not be judged simply on a success-failure basis
but rather in terms of the models' ability to answer questions about
real world. In the past, some models have been more successful than
others. For a review of previous modeling efforts one should consult:
Ira S. Lowry, Seven Models of Urban Development: A Structural Com-
parison (Santa Monica, California: RAND Corporation, 1967); B. W.
Mar and IV. T. Newell, Assessment of Selected RANN Environmental Modeling
Efforts. A report prepared for the Environmental Systems and Resources
We would expect land use change or growth to be stimulated or
brought about by certain "triggering" events or mechanisms. Many
early land use modeling efforts assumed the "triggering" mechanism to
be brought about by the location of new industry. We now recognize
that other stimulating mechanisms such as improvements in transpor-
tation or expanding demand for personal services also may prompt
regional growth.3 This research problem, however, is predicated upon
the assumption that industry (specifically export industry) is the
major stimulus for urban and regional growth and thus deserves initial
consideration in any modeling effort.
Brown et al. recognized "the critical importance of industrial
location. .yet land-use modelers have devoted surprisingly little
effort to analyzing the determinants of industrial location."4 Only
a few research efforts have made serious attempts to model the basic
determinants of industrial location choices and "for the most part
these attempts have been quite limited and crude."5 The focus of
Division, National Science Foundation (Seattle: University of Washing-
ton, June, 1973); and 0. Stradel and B. G. Hutchinson, Notes for a
Short Course on Practical Applications of Regional Development Models
(Waterloo, Ontario: University of Waterloo, The Transport Group,
Department of Civil Engineering, November, 1971).
3Homer Hoyt was among the first urban economists to study the
growth and development of urban land. His primary axiom was that "no
city could grow by taking in its own wash." Homer Hoyt, The Structure
and Growth of Residential Neighborhoods in American Cities (Washington,
D. C.: U. S. Government Printing Office, 1939).
4james II. Brown and others, Empirical Models of Urban Land Use:
Suggestions on Research Objectives and Organizations, Exploratory
Report 6 (New York: Columbia University Press, 1972), p. 82.
this study is on the conversion of land from prior uses to industrial
Statement of Problem
An initial requisite of any model purporting to replicate and
predict industrial land use development is the identification of those
land parcels having the greatest probability to change to industrial
land use. It is necessary, therefore, to construct an algorithm which
calculates the suitability of some land parcel for a specific indus-
trial land use. The primary objective of this research effort is to
develop a site-selection algorithm which identifies land parcels
having the highest probability of attracting industrial land use.
More specifically this research attempts to identify and measure the
importance of the intrinsic site characteristics which appear to
control the conversion of land to industrial use within the 16-county,
Knoxville metropolitan region and to formulate a simulation algorithm
to identify potential industrial sites.
Intrinsic site characteristics refer to the qualities possessed
by some land parcel which identify its suitability for a specific land
use. In the case of industrial land use, examples of some of the
attributes which should be examined are: proximity to rail service;
access to nearby highways; site preparation costs; availability of
city services; or whether the site is within a developed industrial
park. The importance of these attributes may be measured and subse-
quently aggregated to form a combined measure or index.
The following is a list of indexes which are often identified as
affecting industrial land use choices:
Site Preparation Costs--The cost of converting rural or open
land to manufacturing use. Matters which might be taken
into account are foundation conditions, drainage conditions,
slope of the land, and type of vegetation cover. This
factor does not include the purchase price of the land.
Market Price of Land--The per-acre cost of land to the
manufacturer who wishes to build a new plant or expand his
existing plant. This would include property taxes but not
necessarily taxes on improvements to the property.
Proximity to Suitable Labor Force--The number of suitable
workers within easy access of the site.
Transportation Accessibility--The ease with which people or
materials can be moved from the plant site to roads, rail-
roads, airports, or waterways. In this factor, the concern
is with the linkage between the plant and major transporta-
tion facilities within the immediate area.
Utilities--The kinds and quality of utilities available at
the potential plant site. Some utilities which might be
considered are water, sewer, and gas.
Compatibility with Existing Land Uses--The compatibility of
general manufacturing activity with other existing land uses
adjacent to the potential site.
Neighborhood or Community Attractiveness and Amenities--The
condition and density of dwelling units and business estab-
lishments in the immediate area of the plant site, and
proximity to hospitals, schools, parks, and churches.
Industrial Park Space--The availability of suitable buildings
or land in an industrial park. The appearance of the park
and the quality of services produced are some of the con-
siderations to be examined.
This list was compiled from both a review of literature and
preliminary study. Each factor is a composite of several variables
collectively identified as an index. Particular combinations of
variables are not unique but simply represent logical groupings based
upon the literature search. It is convenient for the purpose of this
study to utilize this list as a beginning point to be refined after
Each index posited is considered to reflect an information unit
to be evaluted in the selection of industrial sites at the intra-urban
scale. By combining these indices it is possible to develop an
aggregate measure of the suitability of a site for industrial use.
Before doing so, however, a weight should be attached to each index to
reflect its relative importance in the location decision. A general
mathematical form for the site selection algorithm follows:
LUS. W. I.
i=l i, ,
where LUS = Attractiveness score for land use category 2
(in this case k = industrial land use),
N = Number of Indices,
W. = Index weight for the i index and the th land
use category, and
.th th 6
I. = i index for the Z land use category.
In modeling industrial land use development attractiveness scores
are utilized in the following manner: Projected industry growth
Oak Ridge National Laboratory, Regional Environmental System
Analysis, A Research Proposal Submitted to the National Science
Foundation, Research Applied to National Needs (RANN) Feb. 1, 1973,
Land use models are constantly evolvi;ig and thus one can only
speak of the framework of the model in a hypothetical manner. This
rity or may not become the structure of the nodel and represents only
a tentative view as to the operating manner of the iodel.
provided via a socioeconomic model is distributed initially to various
subregions through a subregional allocation algorithm. Industrial
expansion within the subregion is distributed to available land par-
cels on the basis of attractiveness scores calculated in the above
manner. The availability of various land parcels is determined from
knowledge of the probability that specific land parcels will be
offered for consumption.
The land use model as presently conceptualized is perhaps best
characterized as a stochastic (Monte Carlo) model. Projected growth
in land use will be awarded to various land parcels on the basis of
the land use scores (LUSk). Basically, however, the model is heuris-
tic in that as the land use conversion processes are better under-
stood, analytic-deterministic procedures will be substituted for
Motivation for Study
It has been noted that "urban spatial organization is the outcome
of a process which allocates activities to sites. In our society,
8The assumption that all land is potentially available is not a
valid premise in that not all property owners are willing to sell
property. Cadastral data and ownership characteristics necessary to
develop a site-availability algorithm are not readily available and,
therefore, very few land use models have included this consideration.
The sociopolitical modeling team at Oak Ridge National Laboratory is
currently studying the methods whereby local opinion and knowledge of
owner characteristics may be utilized to identify land parcels which
may be available for various land use activities. See: Osbin L.
Ervin and Charles R. Meyers, Jr., The Utilization of Local Opinion
on Land Use Simulation Modeling: A Delphi Approach (Oak Ridge,
Tennessee: Oak Ridge National Laboratory, ORNL-NSF Environmental
Program, 1973). Olaf Helmar, The Delphi Method for Systemizing
Judgments about the Future (Los Angeles: University of California,
the process is mainly one of transactions between owners of real
estate and those who wish to rent or purchase space for homes and
businesses." The market process of "transactions between willing
buyers and willing sellers determines the spatial organization of
urban activities ." and thus should dictate the methodological
structure of land use models.10
The market place however is not perfect. Individual land specu-
lation, family or corporate property gifts, over-building by contrac-
tors, land use planners over-estimating demand, governments exercising
property rights, the tendency for land uses to remain intact and the
perpetuation of mistakes rather than corrections, all spoil the simple
modeling of market transactions. However, after investigating numerous
approaches toward modeling land use development, the market place
appears to be the most viable way to approach the simulation of urban
activities.11 The site-selection algorithm seeks only to replicate
the buyer's considerations in selecting an adequate industrial site.
Seller considerations and market perturbations are not considered at
The results of this research effort are intended to integrate
with a "holistic" environmental model to be developed later. The
motivation which stimulated this research problem, however, was the
9John P. Crecine, Computer Simulation in Urban Research (Santa
Monica, California: The RAND Corporation, 1967), p. 2.
Ira S. Lowry, Seven Models of Urban Development: A Structural
Comparison (Santa Monica, California: The RAND Corporation, 1967),
desire to understand the land use development process. After some
preliminary analysis and review of previous research it became appar-
ent that understanding and modeling land use change contained problems
to occupy several dissertations. The decision was made, therefore, to
focus upon one aspect of the land use evolutionary process, the develop-
ment of industrial land use, in the hope that by expanding the under-
standing of this process, rigor could be added to the eventual
development of a comprehensive land use model.
Organization of Study
The organization of this study is: the present chapter attempts
to introduce the reader to the research problem and to rationalize the
need for the study. Chapter II discusses the literature from which a
tentative list of site-selection variables was developed. In Chapter
III, the test region is described and the research methodologies and
procedures explained. Chapter IV presents the results of an empirical
study conducted to identify and determine the relative importance of
various location factors in the 16-county Knoxville region. Finally,
Chapter V summarizes the conclusions derived from the empirical study
and presents recommendations for the construction of a site-selection
algorithm and recommendations for further research.
THEORETICAL AND EMPIRICAL STUDIES RELEVANT TO
THE DEVELOPMENT OF INDUSTRIAL SITE-SELECTION ALGORITHM
The purpose of this chapter is to relate to this research effort
ideas expressed within selected theoretical and empirical studies and
to discuss the evolution of the indexes listed in the previous chapter.
Discussion of some works may be abbreviated depending upon their
relevance to industrial site-selection processes. Theoretical works
involving classical location theory are discussed first, followed by
a discussion of empirical studies of industrial location factors.
The final sections discuss approaches utilized in other land use
models and the development of a tentative list of variables.
Selected Theoretical Works
The location of industry in an intrametropolitan area is the
result of a complex interaction of variables that can best be under-
stood only through careful examination of historical events. Classi-
cal location theory, however, abstracts from reality by use of the
principle ceteris paribus where one or two variables are permitted to
vary while all other variables are held constant. Von Thtinen's and
Weber's studies represent early contributions to location theory and,
thus, deserve fir.:t consideration in this study. More recent theo-
retical works follow.
Der Isolierte Staat1
Von ThUnen's work dealt with a hypothetical agricultural land use
system whereby transportation costs and economic rent were used to
explain the location of specific types of agricultural production as
they were concentrically displaced around a market center. In a
similar manner, one might expect industrial, commercial, and residen-
tial land uses to arrange concentrically around an urban center (Figure
1). For a specific land use the utility (accessibility) derived from
any location would decline with increasing distance from the CBD as in
(a). This, of course, assumes that the CBD is the most accessible
point within the urban area. The optimal price one would pay for
utility or accessibility is illustrated in (b). Cost incurred in
obtaining utility (Y) is represented by the area (XYZA) but the profit
to be derived at utility (Y) is (ABCZ). The profit is similar to the
land rent of von Thinen's agricultural model. Thus, rent for various
land uses can be expressed as a function of distance from the CBD as
in (c). Based upon the comparative bid-rent capabilities of each land
use, one would expect (RR1) to represent the rent function for
residential land use, (QQ1) for wholesaling and industry land use,
Johann Heinrich von ThUnen, Der Isolierte Staat in Beziehung
Auf Landwirtschaft Und Nationalikonomies (Berlin: Schumacher-Zarchin,
1875) in K. W. Kapp and L. L. Kapp, Eds., Readings in Economics (New
York: Barnes and Noble, 1949).
2Michael E. Eliot Hurst, A Geography of Economic Behavior: An
Introduction, Belmont, California: Duxbury Press, 1972, p. 231.
and (PP1) for commercial and service activities. The intersections at
Y and Z would form the boundaries between the various land uses.
Commercial activities are the highest bidders for sites and
usually occupy the most accessible places within the city. But commer-
cial land uses are also found in the industrial and residential zones.
Similarly, residential uses may also be found in other zones. For
example, a four-story apartment building may yield several times more
rent as a one-story commercial operation on the same site and thus may
out-bid competitors for the property.4
Bid-rent functions are a very important real world phenomenon
encountered in explaining industrial site selections. Industries
seeking highly accessible sites upon which to build must compete
against other bidders for those same sites. Consequently cost per
unit acre, distance from CBD and proximity to transportation facili-
ties were included as variables to be examined in this study.
Uber den Standort der Industrien5
Alfred Weber was among the first economists to pose a general
theory of industrial plant location. The optimal location for an
industrial plant was seen to be a formation of three factors:
Ronald Reed Boyce, The Bases of Economic Geography: An Essay
on the Spatial Characteristics of Man's Economic Activities (Atlanta:
Holt, Rinehart and Winston, Inc., 1974), p. 264.
5Alfred Weber, Uber den Standort der Industrien (Chicago: Univer-
sity of Chicago Press, 1928).
transportation costs, labor costs, and agglomerative forces. Weber
theorized that the optimal location would be found:
1. where total transportation costs per unit of output were at
a minimum or,
2. where transportation diseconomies were offset by savings
through agglomeration factors or access to labor.
Of particular interest to this study are Weber's ideas concerning-
agglomerative factors and proximity to labor. In general, the ideas
posited by Weber are of greater importance at the regional or sub-
regional levels of industrial location than at the site level. Some
of Weber's agglomerative factors, however, relate to industrial site
characteristics and, thus, are considered in this analysis.
Weber's agglomeration factors are:
(1) The joint development of industries which promotes the
attraction of auxiliary industries and increases the efficiency of
large scale production and utilization of special technical equipment.
(2) The development and growth of specialized labor due to the
greater opportunity for work in the area.
(3) The greater accessibility to raw material suppliers who can
provide material regularly and on short notice.
(4) The reduction in overhead costs, such as gas, water, elec-
tricity, roads, and communications.6
6Robert G. Turner, "General Theories of Plant Location: A
Survey," AIDC Journal, VI (October, 1971), pp. 25-26.
The Location of Economic Behavior7
The A.\nercan economist, Edtgar Iloover initially attempted to
improve i''eber's explanation of industrial location )by including the
consideration of such variables as the size of mIarket area and insti-
tutional forces in his theoretical analysis. Yet Hoover, although
critical of Weber's analysis, ultimately proceeded along the same
lines to present his location theory primarily in terms of costs,
Hoover, however, did expand Weber's agglomerative forces to include
the importance of banks, utilities, fire and police protection,
climate, property tax, and lower interest rates.
Hoover also developed some generalizations in the locational
habits of light industry versus heavy industry. Because of the
necessity of "handling of large quantities of goods either coming in
from elsewhere or being shipped out, heavier types of manufacturing,
warehousing, and wholesaling prefer locations in transshipment zones
along rail or waterways." "Manufactures, wholesalers and warehouses
of less bulky goods need not be located on railroads or waterfronts at
all, since they can be served by truck." They locate more in response
to "the attractions of labor supply, cheap land, and nearness to local
suppliers or customers. As a rule, they are found interspersed with
commercial and inferior residence uses."'
Edgar M. Hoover, The Location of Economic Behavior (New York:
Turner, op. cit., p. 27.
Hoover, op. cit., pp. 128-129.
The site considerations of industries of the 1930's and 40's have
changed in more recent times but several basic locational rules as
expressed by Hoover and Weber remain intact. Accordingly, several
variables such as neighborhood amenities, neighborhood compatibility,
proximity to labor force, availability of utilities, adjacent land
uses, and transportation accessibility have been included in this
Imperfect Competition and the Duopoly Debate
A number of location theorists believed pure competition was not
a suitable theoretical structure for the study of plant locations, and
sought to explain locations in terms of the competition between two
firms attempting to capture the largest share of a market area.
Fetterl0 and Hotelling11 were among the earliest to expound on
duopoly location theory. They were followed by Lerner and Singer,2
Smithies,13 and Chamberlin4 who expanded the original concept.
10Frank A. Fetter, "The Economic Law of Market Areas," Quarterly
Journal of Economics, XXXVIII (May, 1924), pp. 520-529.
11Harold Hotelling, "Stability in Competition," Economic Journal,
XXXIV (March, 1929), pp. 41-57.
12A. P. Lerner and H. W. Singer, "Some Notes on Duopoly and
Spatial Competition," Journal of Political Economy, XLV (April,
1937), pp. 145-186.
13Arthur Smithies, "Optimal Location in Spatial Competition," The
Journal of Political Economy, XLIX (June, 1941), pp. 423-439.
1E. H. Chamberlin, The Theory of Monopolistic Competition
(Cambridge: Harvard University Press, 1936), pp. 194-196.
Devletogloul recently commented on the economic irrationalit) of
the approach and presented arguments against the theoretical base for
such location activities. Imrpo:rtant contributions, however, are still
to be noted. Duopoly theory sugg-sts consideration should be given to
the repelling or attracting properties of industries which depend upon
The Theories of David M. Smithl7 and Melvin L. Greenhut18
Smith's and Greenhut's contribution to the theory of industrial
location incorporated the minimum costs approach of Weber with the
maximum profit solutions of manufacturing location posited by August
Lisch.19 Smith calls this the maximin solution. The concept devel-
oped is illustrated in Figure 2.2 In (a), the costs of production
are permitted to vary over space (distance) and revenue obtained
(demand) is kept constant. This is essentially the Weber solution.
The basic concept of Lbsch's model is shown in (b), where revenue is
1Nicos E. Devletoglou, "A Dissenting View of Duopoly and Spatial
Competition," Economica (May, 1965), pp. 140-160.
1Turner, op. cit., p. 31.
1 DIvid M. Smith, "A Theoretical tFramework for Geographical
Studies oE Industrial Location," Economic Geography XLII (April,
1966), pp. 95-113.
18Melvin L. Greenhut, Plant Location in Theory and in Practice
(Chapel Hill: University of North Carolina Press, 1956).
William H. Weglom and Wolfgang F. Stalper (Translators), The
Economics of Location, by August Lbsch (New York: John Wiley and
20Smith, op. cit., p. 96.
Smith, op. cit., p. 96.
DISTANCE --a b
SMITH'S MODEL WITH COMPARISONS (AFTER SMITH, p. 96)
permitted to vary over space and costs are held constant. Finally, in
(c) the combined solution suggested by Smith is offered with maximum
profit occurring at A, where costs are lowest and profit is highest.
Note that maximum revenue, however, is obtained at B.21
An interesting variation of Smith's model was the introduction of
noneconomic factors, in particular, the concept of psychic income.22
This innovation permitted social, psychological, or other personal
factors to be entered into the model, hence relaxing the assumption of
economic man. Such considerations according to Smith tend to divert
the location of a plant from the ideal site to locations closer to the
owner's home, a golf course, or perhaps a parochial school.
Smith suggests that stochastic procedures may ultimately have to
be used to simulate industrial location decisions as personal factors
cannot be accounted for by rigorous mathematical reasoning.23 This
research assumes that personal considerations may be accounted for by
noting neighborhood amenities near the potential site. The importance
of housing quality, proximity to churches, hospitals, schools, or
parks and personal services availability in the immediate vicinity of
the site are examined in this study.
The following empirical studies were significant resources in the
development of a tentative list of site-selection variables. Two
21Ibid., pp. 96-97.
22Ibid., p. 108.
23David M. Smith, Industrial Location: An Economic Geographical
Analysis (New York: John Wiley and Sons, Inc., 1971), pp. 269-273.
types of studies are presented: those studies directed primarily
toward the industrial developer who seeks new industry for a commu-
nity; and those studies which attempt to anAilyze industrial location
coraiderations by way of a large sample of new industries and to cate-
gorize the locational considerations by industry types. Only one
study is directed toward the development of a site-selection algorithm
for a land use model.
The Studies of Allen Pred24 and Richard Lonsdale2
Allen Pred has compiled a study of the history and present status
of industrial location decisions within a metropolitan region. These
patterns discussed pose interesting hypotheses for empirical analysis
but the interest of this research is directed primarily to the site
characteristics discussed by Pred. It should be noted that Pred's
analysis focused upon a single metropolitan area whereas this study
encompasses a region with a hierarchy of urban places.
In discussing location patterns, Pred identifies seven types:
1. Ubiquitous industries concentrated near the CBD The market
area of these industries is generally coincident with that of the
metropolis or city. Food processing industries, specifically bakery
goods, package foods, and fresh milk products, are Lome of the examples
of these types of industries.
24Allen R. Pred, "The Intrametropolitan Location of American
Manufacturing," Annals of the Association of American Geographers, LIV
(June, 1964), pp. 165-180.
25Richard E. Lonsdale, "Rural Labor as an Attraction for Indus-
try," AIDC Journal, IV (October, 1969), pp. 11-17.
2. Centrally located "communication-economy" industries Job-
printing industries, newspaper printing, and advertising printing
would be representative industries of this category.
3. Local market industries with local raw material sources -
These industries show a high degree of randomness in their locational
pattern but with some tendency toward CBD locations. Samples would be
ice plants, concrete brick and block industries or industries whose
raw materials are by-products of other large-scale industries such as
the pulp and paper products industry.
4. Non-local market industries with high value products Typi-
cally these industries provide a high value per unit weight product
and are insensitive to transport considerations within the local
region. The pattern is "at least superficially irrational." Computer
and related industries and chemical industries are typical examples.
5. Noncentrally located "communication-economy" industries The
subset of industries includes those which are not necessarily pulled
to any functional area of the city but rather tend to cluster together
in any suitable area primarily because of the necessity to "keep
abreast of the latest innovations or forthcoming contracts." Elec-
tronic, military equipment, and space age industries such as found in
Huntsville, Alabama, or Houston, Texas, are examples.
6. Non-local market industries on the waterfront Industries
where primary raw materials are imported by water or whose finished
product is often moved by water comprise this group. Petroleum
refining, coffee roasting, and sugar refining are prominent among
these types of industries.
7. Industries oriented toward nationLral markets These indus-
tries have extensive market areas and arc influenced by high transpor-
tation costs on a bulky fieiishcd product. A large percentages o-fr those
industries have a tendency to locate d:n t h side of the metrc olis
facing the most important market regii.. Pred identifies the Noway :
Lowland refining indust-ry anid th: i)e;. -c_ utomo(i.ie industry a.s
Richard Lonsdale in a study oi the locationa.1 habits of rural
industry notes that those industr~i-;. a .fected by transportation coAi.
tend to locate in urban areas whiJ.; ;.:o:;e highly affected by labor
cost, gravitate to rural areas. (t addition, Lonscda.ie note rural
firms tend to space themselves out; ij.t'. .irde- to assure a labor supply
industries wi-Lh tendencies toward i:rl 0oc.-tions are apiparc
food "rodC..ti's, textit, l J. umber :,d-w' .,*/ y r \, t;s,, p"p.cr 'p ourcG" ,;
chen.cal, and electrical machinery :L,-ecially routine assembly. I.o
profit margins, keen competition, and nigh petrc tageO2 of produce in
workers. are s5ome basicc character i.s;i.-. of thest :i J:Lus ries.
The patterns of in T.ustrial location detailed by Fred and Lon:',d:l
are rimari- ; identified with a spatl.. level slightAy higher tha-: j.11
;sit specific l e,'el which is the foc 'r:; this sty In relation i.
the ... jvelcpmen -. 1'e i. to-select i.o-. '.i rithim, '.a La ions in ::.
rerrn re signif. ic r it is obvi'.ort: sep.r t :tings wil
have -o be developeJd for each i ndi.sty type, but this is not a prjim',ry
6Pred, op, cit., pp. 175-1.78,
Lonsdale, :p. cit.
objective of this study. It was anticipated that after developing a
general site-selection algorithm additional research would permit the
matching of various types of manufacturing to specific types of sites.
Historical Studies of Industrial Location Factors8
Numerous empirical studies exist of the factors associated with
actual industrial location events. Although many of these studies
were undertaken by academicians, the viewpoint is that of the indus-
trial developer seeking to attract new industry or expand existing
industry within the community.
Characteristically, investigations of this type are not concerned
with a specified theoretical framework for approaching the problem
28Discussion within this section is based mostly upon a review of
the following articles: J. S. Bullington, "Utilization of State-Wide
Site Evaluation Committee to Aide in the Location or Relocation of
Plant Facilities," AIDC Journal, IV (October, 1969), pp. 27-42; James
E. Chapman and William H. Wells, "Factors in Industrial Location in
Atlanta, 1946-1955," Atlanta Economic Review, IX (September, 1959),
pp. 3-8; Ronald E. Carrier and William R. Schriver, "Location Theory:
An Empirical Model and Selected Findings," Land Economics, XLIC
(November, 1968), pp. 450-460, and a more complete explanation of the
study: Ronald E. Carrier and William R. Schriver, Plant Location
Analysis: An Investigation of Plant Location in Tennessee (Memphis:
Memphis State University, 1969); Melvin L. Greenhut and Marshall R.
Colberg, Factors in the Location of Florida Industry (Tallahassee:
The Florida State University, 1962); T. E. McMillan, "Why Manufacturers
Change Plant Location versus Determinants of Plant Location," Land
Economics, XLI (August, 1965), pp. 239-243; N. J. Stefaniak, Indus-
trial Location within the Urban Area: A Case Study of Locational
Characteristics of 950 Manufacturing Plants in Milwaukee County
(Milwaukee: Wisconsin Commerce Reports, 1962); Charles M. Tiebout,
"Location Theory, Empirical Evidence and Economic Evolution," Regional
Science Association, Papers, III (1957), pp. 74-86; U. S. Department
of Commerce, Industrial Location Determinants, 1971-1975 (Washington,
D. C.: U. S. Department of Commerce, Economic Development Administra-
tion, February, 1973); and D. C. Williams and Donnie L. Daniel,
"Industrial Sites for Small Communities," AIDC Journal, VI (April,
1971), pp. 33-39.
but rather to learn the reasons for the location decision from persons
acquainted with the location event. None of these studies were con-
cerned with the construction of an algorithm to simulate the process
of industrial land use development.
Greenhut and Colberg29 analyzed factors influencing the decision
of 400 manufacturers locating in the State of Florida between 1956 and
1957. Location considerations were divided into three groups: demand
(market) considerations, costs (assembly) considerations, and personal
(psychic) considerations. Access to markets and potential markets
(Table 1) rated the highest among the location factors with the
remaining factors surprisingly low. The study, however, was slanted
toward measuring regional and subregional factors and thus was of
limited value to this study.
The extensive study undertaken by Carrier and Schriver30 of
plant locations in Tennessee between 1955 and 1965 was conducted
within the framework of existing location theory and, in part, did
focus upon site-location factors. Many of the variables included
in this analysis are based upon the conclusions reached in this study.
Carrier and Schriver identified six classes of location factors
believed capable of affecting plant locations: (1) personal factors,
(2) procurement-cost factors, (3) processing-cost factors, (4) distri-
bution-cost factors, (5) location demand factors (including locational
interdependency considerations), and (6) certainty factors. (This
2Greenhut and Colberg, op. cit.
30Carrier and Schriver, op. cit.
Carrier and Schriver, op. cit., p. 451.
Percentage of 400 Plant
Location Factors Listing as Primary Factor
Access to markets 51.9
Anticipation of market growth 12.8
Good labor relations 1.7
Lower wages 2.6
Ease of attracting out-of-state
personnel, including research 4.7
Low freight cost on obtaining raw
materials and components 7.7
Low cost on freight on shipping final
Climate as it affects operations 1.8
Community facilities (education, police,
medical, etc.) 2.9
All other factors 3.2
FACTORS MOST INFLUENTIAL IN THE LOCATION
DECISIONS OF FLORIDA INDUSTRIES, 1956-1957
class was identified after the interviews.) Certainty factors were
defined as the confidence that the "prevailing and forecasted data used
to identify the site offering maximum profits would persist into the
Persons involved in the selection of sites for 307 manufacturing
plants were interviewed. Each respondent was asked to select six
factors from those listed in Table 2 and to distribute 100 points among
these six in order to indicate the relative importance each factor
contributed to the total plant location decision.
Of the 36 factors listed in Table 2, low cost and availability of
labor was mentioned most frequently as the primary factor affecting
the location decision (Table 3). Personal considerations without
economic advantages received the highest average number of points,
followed by low cost and availability of labor (Table 4).
On the basis of the interviews the authors grouped industries
according to the six factors previously listed:
(1) Personal Factors Miscellaneous manufacturing, furniture and
fixtures, and food and kindred products were highly sensitive to
personal factors with most of those firms being "home-grown."
(2) Procurement-Cost Factors Industries which need large
volumes of low-unit-value or perishable raw materials were character-
istically affected by this group of factors. Food and kindred prod-
ucts, stone, clay and glass products, and lumber and wood products
industries indicated greater sensitivity to these factors.
1. Personal Factors;
Personal with economic
Personal without economic
2. Procurement-Cost Factors:
Better service from seller
of raw materials and
Low cost on raw materials
Availability of low cost
3. Processing-Cost Factors:
Low cost and availability
Low cost of fuel
Low cost of electric power
Low cost of financing project
through Area Redevelopment
Low cost of satisfactory
type of water
Adequate waste disposal
Low cost of building and
Low cost of financing plant
through revenue or
general obligation bonds
Favorable community and state
Available existing plant
Available existing building
of building site
4. Distribution-Cost Factors:
Low freight cost,
5. Location-Demand Factors:
Greater demand in
Greater demand poten-
tial in the area
6. Certainty Factors:
Nearness to metro-
and zoning laws
of the town
Size of city
Data provided by
Chamber of Commerce,
by local manufac-
Recreation, a good
place to live, etc.
Nearness to corporate
neutral in labor-
Progress in racial
Data provided by the
LIST OF POSSIBLE FACTORS INFLUENCING INDUSTRY
LOCATION AS UTILIZED IN THE CARRIER AND SCHRIVER SURVEY33
Ibid., p. 453.
Factor of Firms Rank
Low cost and availability of labor 65.6 1
Low cost of electric power 36.0 2
Favorable labor management relations 35.7 3
Community leaders' cooperation 32.2 4
Low cost of building and land 19.8 5
Low freight cost, finished product 17.9 6
Available existing plant 17.5 7
Favorable community and state tax
structure 17.2 8
Low cost of financing plant through
revenue or general obligation bonds 16.9 9
Available existing building 16.6 10
TEN FACTORS MOST FREQUENTLY MENTIONED BY TENNESSEE
FIRMS AS AFFECTING THE LOCATION DECISION
Factor of Firms Rank
Personal without economic advantages 49.2 1
Low cost and availability of labor 38.0 2
Available existing plant 35.9 3
Personal with economic advantages 32.9 4
Availability of low cost raw materials 31.9 5
Greater demand in area 30.2 6
Greater demand potential in area 29.8 7
Low cost of financing project through
area Redevelopment Administration 29.6 8
Available existing building 27.6 9
Nearness to corporate headquarters 26.0 10
TEN LOCATION FACTORS WITH HIGHEST MEAN NUMBER
POINTS ASSIGNED BY TENNESSEE FIRMS INTERVIEWED
(3) Processing-Cost Factors These factors are associated with
in-plant costs in assembling or processing the finished product,
e.g., labor, energy, external services, capital, land costs, etc.
Electrical machinery, apparel and related products, and textile mill
products industries were affected by these factors.
(4) Distribution-Cost Factors These factors reflect the costs
incurred in shipping the finished products to the buyer. Among the
most sensitive to these factors were food and kindred products,
miscellaneous manufacturing, and paper and allied products industries.
(5) Location-Demand Factors Industries affected by these
factors are highly sensitive to market-demand in terms of proximity.
Included in this category are paper and allied products, printing and
publishing, and primary metal industries.
(6) Certainty Factors The validity of existing and forecasted
data is considered to be highly important by industries affected by
these considerations. In other words, these industries want to know
the future stability of costs in production and the probable con-
tinuance of existing markets. Printing and publishing, leather and
leather products, and transportation industries were highly sensitive
to these factors.
It is obvious that the scope of the Carrier and Schriver study is
much broader than the objectives of this study. Its utility is
limited for this research purpose. The factors considered by Carrier
and Schriver span several spatial levels of locational decisions. The
result is that factors which may be very important at the site-
selection level are weighted low in comparison to the total list of
factors. Also, the disproportionate number of factors offered for
consideration under the six categories tends to skew the weightings.
Finally, the lack of a very large sample in specific SIC categories
tends to decrease the validity of the results of the weightings and,
therefore, the conclusions reached regarding the typical locational
patterns of specific industries.
Bullington34 offers a scheme to locate potential industrial
sites on a state-wide basis by suggesting the scoring of location
factors on an ordinal scale and aggregating them into an index to
determine the site potential for specific industries. Local and
state governments could then match the qualities of the industrial
sites available in the community to specific industries. The factors
which Bullington suggests are listed in Table 5.
The U. S. Department of Commerce recently published the partially
aggregated results of an extensive 5-digit industrial location survey
conducted by mail throughout the U.S. The purpose of the survey
was "to assist the nation's underdeveloped and declining areas in the
development of their economic resources and potentials."36 Only
manufacturing industries demonstrating "reasonable" growth between
1958 and 1967 were selected for inclusion in the survey. Survey
forms were mailed to a total of 2,950 companies in 254 different
34J. S. Bullington, "Utilization of a State-Wide Site Evalua-
tion Committee to Aide in the Location or Relocation of Plant
Facilities," AIDC Journal, IV (October, 1969), pp. 27-42.
35U. S. Department of Commerce, Industrial Location Determi-
nants, 1971-1975 (Washington, D. C.: U. S. Department of Commerce,
Economic Development Administration, February, 1973).
3Ibid, p. 1.
a. Size of Parcel
b. Shape of Parcel
e. Flood Record
f. Condition and
g. Underground Water
h. Soil Bearing Capacity
(This referred to the
potential friction or
good-will prompted by
the location of industry)
b. Secondary Roads
d. City Water
e. City Sewer
f. Limitations of Site
a. Commercial Air Service
b. Water Transport
c. Location in State
d. Mileage Rate
e. Airport Facilities
f. Comprehensive Planning
g. Retail Accommodations
i. Community Appearance
k. Presentation of Facts
1. Sanitary Sewer and
LOCATION FACTORS SUGGESTED BY BULLINGTON
SIC categories. One form, Survey of Industrial Location Determiinants,
was to be completed by all companies to i.Jcntif'y the locational and
operating characteristics of existing plants.
Only the results of the Survey of Industrial Location Det-rmi-
nants have been published to date and its usefulness for this study
is limited because of the highly disaggregated form of the data.
Hopefully, these data have been digitized by now and can be utilized
for future analysis purposes.
The attributes measured or assessed in the survey are only
slightly coincident with those sought in this analysis. Also the
survey spans several spatial levels of locational decisions. Never-
theless, some of the variables utilized in the survey were included
in this analysis.
Land Use Modeling Studies
Site-selection algorithms are certainly not novel to land use
modeling methodologies. Numerous modeling efforts have utilized
various allocation systems to distribute projected change in land
use. Most of these approaches have been different from this approach
in framework and in terms of the spatial scale of allocation.
Previous land-use models usually operate at the census tract or
count) level simply because data (e.g., census materials) to calibrate
the models are more readily available at those levels. The following
studies, however, have approached the allocation problem similar to
this study and consequently are briefly discussed below.
Among the first land-use models to utilize a site-selection
algorithm to distribute projected land use change was one developed
at the University of North Carolina by Donelly, Chapin, and Weiss.37
The primary purpose of the model was to simulate the growth and spa-
tial spread of residential land use. Historical data (1948-1960) were
utilized to calculate an attractiveness probability to simulate the
conversion of rural land to urban residential use. Changes were
recorded by means of 1000-foot cells composed of nine 2.5-acre land
The model operated in the following manner:
1. All land within the city unsuitable for development is
eliminated from consideration at the beginning and the
supply of land remaining is identified as available for
2. For each 1000 foot cell, a measure of relative value is
established, i.e., land value, as a measure of its attrac-
tiveness for residential development.
3. The effect that "priming" (expansion of municipal services,
commercial services, and industrial development) decisions
will have on modifying the value of the property is then
calculated for each cell. These are assumed to be exoge-
nously given but in this case the exact amount is known from
historical data from 1948 to 1960.
4. Land parcels are then "reassessed" to obtain a new attrac-
37Thomas G. Donnelly, F. Stuart Chapin, and Shirley F. Weiss,
A Probabilistic Model for Residential Growth (Chapel ilill: University
of North Carolina, Institute for Research in Social Science, 1964).
5. Density conIstraints (numiibe.rs of units per acre/year) are
6. Finally, the known gro'H,;h of residential households between
1948 and 1960 is allocated on a probability basis.38
The drawbacks to the model are: (1) the need to have historical
data in order to develop the transitiGn probabilities; (2) the disre-
gard of the effect of changes in industrial location and employment;
and (3) the vast amount of programming needed to load and run the
model for a small region such as a single city. The concepts gene-
rated by this study, however, have become basic to many other modeling
efforts and were equally important in this study.
The Pittsburgh industrial location model, INIMP39 (Industrial
Impact Model), is similar to the hypothetical design of this modeling
approach; the major difference is that growth is distributed to
census tracts and consequently the variables are more aggregated than
those being considered in this study.
Four variables (attributes of census tracts) and one constraint
were adopted as being sufficiently discriminatory to determine site
locations. These are: weighted mean unit-assessed value of land;
weighted mean unit-assessed value of buildings; weighted mean struc-
tural density, and amount of industrial clustering. These measures
were determined by census tract. The constraint can either be imposed
3Ibid, p. 11.
3Steven H. Putman, "Intraurban Industrial Location Model Design
and Implementations," Regional Science Association, Papers, IXX
(1966), pp. 199-214.
artificially as in the case of zoning controls or non-existence of
services; or directly imposed by the model operator. On the basis of
the aggregated scores of the above indexes, the :nodel distributes a
portion of projected city-wide employment changes among exist ,ig
facilities and, upon reaching certain critical values of saturation,
switches to a separate routine to distribute new facilities to census
tracts having the highest scores for the remainder of the projected
industrial employment growth. The distribution to various census
tracts is accomplished by way of the maximum score. However, in the
event of a tie, the allocation algorithm switches to a Monte Carlo
routine. Similar but more elegant models of this type are being
developed at Harvard40 and the University of British Columbia.41
A Tentative List of Variables
Each theoretical study reviewed approached the problem of
explaining industrial location in terms of three components: demand,
cost, and personal factors. Carrier and Schriver subdivided the
process into six components: personal factors, procurement-cost
factors, processing-cost factors, distribution-cost factors, location-
demand factors, and certainty factors. Only processing costs, pro-
curement costs, distributing costs, personal costs, and certainty
4Carl Steintz and Peter Rogers. A System Analysis Model of
Urbanization and Change: An Experiment in Interdisciplinary Educa-
tion (Cambridge: M.I.T. Press, 1971).
4M. A. Goldberg, Quantitative Approaches to Land Management
(Vancouver, B. C.: University of British Columbia, The Resource
Science Center, 1970).
factors have any direct relationship to site location considerations.
Certainty factors may be considered a variation of location-demand
costs. At the site level, procurement costs and distribution costs
are sensitive to ,ne variable, accessibility. Considerations of
freight rates, transport modes, proximity to raw materials, supplies,
etc. are more related to locational considerations at the subregional
In an effort to arrive at a tentative set of variables to be
utilized in this analysis, a list of variables mentioned in the
industrial survey literature as being related to the site selection
decision was compiled. This list is found in Table 6.
A number of variables may be eliminated as not related to the
site-selection process while others can be combined with other vari-
ables. After considerable study the following list of variables was
developed to be used in a survey of the site conditions of past
industrial location events. Selection was based upon the relevance of
the variable to industrial site-selection and the ability to obtain
measures of the variable. The variables are listed below:
I. Site Preparation Cost
a. Slope of land
c. Clearing-Cover conditions
II. Market Price of Land
a. Distance to center of town
b. Distance to nearest major thoroughfare
c. Density of urban use in immediate vicinity
d. Overall rating of price of land from 1 to 10
III. Proximity to Work Force
a. Proportion of nearest city within 2-1/2 miles
b. Population of nearest community
1. Industrial Park Space
2. Industrial Park Quality
3. Industry Nearby
4. Zoning and Building
5. Pollution Regulations
6. Land Costs
7. Site Preparation Costs
10. Soil Conditions
11. Place to Dump Effluent
12. Processing Water
14. Municipal Water
16. Natural Gas Service
17. Proximity to Local
18. Proximity to Local
19. Proximity to Sup-
20. Water Transport
21. Railroad Transport
22. Highway Transport
23. Commercial Airport
24. Distance to CBD
25. Community Transportation
26. Community Parking
27. Neighborhood Services
Parks, Gas Stations, etc.)
28. Nearby Housing
29. Nearby Labor
30. Community Cooperation
31. Community Stability
32. Community Wealth
33. Community Taxes
34. Community Progressiveness
35. Community Attractiveness
36. Community Labor Climate
37. Community University
38. Nearby Government and
39. Community Wage Rates
40. Space for Expansion
TENTATIVE LIST OF VARIABLES
IV. Transportation Accessibility
a. Distance to major highxi',-
1). Di stance to secoindaryl roa
c. Distance to rail
d. Distance to airport
e. Waterway service
f. Distance to nearest Interstate interchange
g. Overall quality of accessibility from 1 to 10
a. Water available
b. Gas available
c. Sewerage available
VI. Compatibility with Existing Land Uses
a. Did community have zoning at time of location?
b. Was site zoned for industry?
c. Was zone changed to accept industry?
d. Was industry already in immediate area?
e. Overall rating of contiguous land use compatibility
VII. Neighborhood or Community Attractiveness and Amenities
a. Condition of neighborhood
b. Density of land use in immediate vicinity
c. Nearby community services
VIII. Industrial Park Space
a. Was the site in an industrial park?
b. Overall rating of the quality of park?
Additional Data Collected
a. Proximity of site to Knoxville
b. Amount of other industry located nearby at time
c. Was building already there?
This list may omit variables which should be considered and,
therefore, should not be considered exhaustive. However, it is antic-
ipated that in measuring the importance of each variable some may be
eliminated thus simplifying the site-selection algorithm. The next
chapter explains the methodology and research procedures used to test
the importance of each variable.
CILI fER III
DESCRIPTION OF THE STUDY AREA
AND ANALYSIS PROCEDURES
This chapter describes the research methodologies utilized in
this study. The first part of the chapter describes the eastern
Tennessee study region, the second part the data collection proce-
dures, and the final part the analysis procedures.
The Study Region
The study region encompasses 16 counties surrounding and including
Knoxville, Tennessee, (Figure 3) and represents an administrative
entity called the East Tennessee Development District (ETDD). The
region spans 6,500 square miles and contains a population of approxi-
mately 750,000 people. Its selection for this study was based upon
the availability of data in the Oak Ridge National Laboratory (ORNL)
Data Base.1 ORNL selected the region on the basis of "the diversity
of the region, the availability of data, the presence of cooperative
and interested user groups, close proximity, etc."2
1Oak Ridge National Laboratory is currently developing a "holis-
tic" environmental model for the ETDD region. This research was
supported by this program. ORNL-NSF Environmental Program, Regional
Environmental Systems Analysis (A Research Proposal Submitted to the
National Science Foundation, February, 1972).
Ibid, p. 3.
The ETDD region is centered in the southern portion of the Ridge
and Valley Province (the "CGreat Vally"'), bordered to the northwest by
the Cumberland Plateau and Mountains, and to the southeast by the
Great Smoky Mountain complex (Figure 4). Both the Smokies and nhe
Cumberlands are characterized by steep slopes and forest cover, with
the Cumberlands distinguished by strip mining scars.
The area is drained by the Tennessee River system, the natural
flow of which has been vastly altered by the Tennessee Valley Author-
ity (TVA). By harnessing the power of the river system to produce low
cost electrical power and developing a navigable channel to Knoxville,
TVA has become the major development agency within the study region.
As a result much of the industrial development in the region has been
structured by TVA activities.
The largest urban center in the region is Knoxville which serves
as the major economic and transportation focus for the region. Sur-
rounding Knoxville are Oak Ridge, Maryville-Alcoa, and more distant
Morristown, each with between 20 to 35,000 in population. These
cities perform subregional functions. Remaining urban centers are
small in population and are mostly located within the confines of the
plateau and the Blue Ridge (Figure 5).
Industry within the Region
This analysis has concentrated upon secondary manufacturing (SIC
20 through 39) and excluded extraction industry (SIC 10 through 19).3
The location determinants of extraction (or primary) industry are
Industries are referred to by their Standard Industrial Code
(SIC) number throughout this study.
dictated more by the distribution of raw materials and consequently
the locational criteria of these industries will be different from
secondary industries. For this reason primary industry is not con-
sidered in this st-.dy.
Categorically one could state that most of the industry within
the region is concentrated in the Knoxville area. Of the 1002 indus-
tries in the region, approximately 46 percent are located in the
immediate vicinity of Knoxville. Knoxville also has the greatest
diversity of industry, whereas industry in other communities is
characterized by specific categories. (An example is the concen-
tration of furniture industry in Morristown.)
Figures 6 through 13 provide a visual overview of the dynamics of
industrial development in the region since 1943. Since 1953 a 48
percent turnover in industry has occurred. Of the 1002 industries
within the region in 1973, 488 have developed in the 16-county region
Table 7 provides a statistical summary of the new industry-new
employment expansion (or decline) from 1952 to 1973. The table does
not reflect the absolute employment expansion rate but rather the
amount of employment gained or lost by the birth or death of specific
two-digit SIC industries. A high average expansion rate indicates new
plants have developed in the region and have significantly increased
the total employment in that specific industry.
In the following discussion each industry type is briefly
described. Maps are included to show the regional distribution of
specific industries. Graphs illustrating the historical employment
tr-- 4- -,-- 1-
.. *. .
I. *. *
I \ !---L~;~,-.- t
I I I
r. ~ .
I I i i i i i
*. a -o g g
__^ / .~ m oo -
*6 S-^ e SA a-
^-- ^r--- 1r
S_ S ** I <
\ ^____i i ..
*:. .. ;' 7 :'
S j i- - -i- -- -Y -.
^_ ,_^ p -" lv ^ ; "i
,t~ P *!'-v;
a I 6 a i i
I industry Expansion
SIC 1955 1958 1960 1963 1966 1969 1973 1952-1973
Food +9.4 +10.0 +0.5 +8.7 +6.5 -3.2 +11.8 +3.8
+1.4 +3.2 +7.6
-6.9 +10.1 +.5
+36.6 +12.0 +.8
+8.3 -3.9 +5.2
+.3 -11.0 -4.4 +25.5
-22.8 -4.6 -6.9
+4.7 +36.8 +6.3 +15.3
EXPANSION OF ETDD INDUSTRIES BETWEEN 1952-1973
195' 1958 1960 1963
1966 1969 1973
-4/.7 -1.3 -6.9 4409.8 --21.1 -30.1 +19.1
+26.4 -17.2 +22.5
+37.3 -1.2 -3.9
-15.7 +2.6 -8.4 +
+3.7 +439.0 -13.2 +136.5
+,.5 -1.1 +9.5 -3.3
34.5 +4.4 +11.6 -26.1
*This table is based upon ORNL industrial location data. Expans:
rates are determined on the basis of new employment prompted by
new plants (or death of plants) since the previous time period.
base year is 1952.
TABLL 7 Contiiinued
_(__ ~_ _____~_ ___
-3.4 +1i.2 -8,6
,.3.1 -15., +55.3 +0.5
trends relative to total regional employment and the industry's share
of total regional manufacturing employment accompany each map.4
Food products industries (SIC 20)5
Of the 103 plants existing within the region, 66 are located in
the Knoxville region. Most of the remaining plants are located in the
southern half of the region bordering the Smoky Mountains near the
better crop producing areas along the margins of the Nolichucky,
Pigeon, Little Tennessee, and the French Broad Rivers (Figure 14).
Approximately 75 percent of these industries employ less than 100
persons. Employment growth has been gradual and in accord with total
regional growth. Proportionally, however, employment has been de-
clining (Figure 15).
Tobacco products industries (SIC 21)
No tobacco industries exist within the study region.
Textile mill products industries (SIC 22)
The textile industry is distributed broadly from northeast to
southwest through the center of the region with Knoxville having the
greatest concentration followed by Morristown and Sweetwater (Figure
4The maps and graphs utilized in this chapter were prepared by a
CALCOMP plotter driven by an IBM 360 computer. The data utilized were
originally collected by Osbin L. Ervin. For a complete presentation
of these data, see: C. R. Mcyers, Jr., 0. L. Ervin, D. L. Wilson, and
P. A. Lesslie, Spatial Distributions and Employment Trends of Manufac-
turing Industries in East Tennessee (19413-1973) (Oak Ridge, Tennessee,
Oak Ridge National Laboratory, June, 1974).
5Transparent overlays found in Appendix B may be superimposed
over each industry map for spatial referencing.
/ (Dn 0 0 l
OS Sh Oh S 0
o ^ 4
os ll l o h se oII I S sI
SO1 S 0t S L0t S zOt S -
^ t 'fl 2 z
I '2 M
Sol s oi s cot 1 9O so S
16). Average employment in the industry is approxi-atcly 210 persons
per plant. Rockwood and Harriman have only four plants but share
nearly 50% of the total employment inl the region. The distribution
of plants reflects a tendency to gravitate to re1.Ltively small communi-
ties as suggested by Lonsdale6 and Pred.7 Textile industry has declined
in terms of the percent of total regional employment in recent years
perhaps resulting from the introduction of higher wage industry into
the region (Figure 17).
Clothing and related products (SIC 23)
Knoxville leads the area with the greatest concentration in both
plants and employment (Figure 18). Average employment is approxi-
mately 315 persons; however, the median is closer to 150 employees.
This type of industry though concentrated in Knoxville is scattered
throughout the region with numerous small communities having at least
one (sometimes two or three) small employment industry. In terms of
new industry expansion the clothing industry expanded at an average
rate of 10.9 percent between 1953 and 1973. Historically, however,
growth has been erratic (Figure 19).
Lumber and wood products except furniture (SIC 24)
This industry is widely scattered throughout the region mostly in
small communities (Figure 20). It .s a small employment industry
mainly utilizing local raw materials. In recent years, an overall
Richard E. Lonsdale, cited in Chapter 2.
Allen Pred, cited in Chapter 2.
Io, s, ~ .... ,. .,... I I I... Ji
OS Sh Oh SE OE SZ O SI O S 2
I.......I .....iii 111111 I I
li 1 I 1 II I I I I - - - .. . . .
SO ?U UL e
111111 i I I
sOI s vUI
OS Sh Oh SE OE SZ OZ SI O S
I ~ .. i i I 1 1 1
I i li i
I I I I I I I il lI I I A I I i i I r
cOI s ZoI s
Ii n n
sOl s h01
a a k
a \ Pa a
decline in the industry has occurred partly due to a decline in small-
operated lumbering in general for the region. The average employment
per plant is only 32 persons with the median closer to 15 persons per
The growth prospects for this industry are rated good by the TVA
Regional Planning Office but it is expected to be in terms of large,
permanent sawmills utilizing local oak, hickory, and poplar forest
resources (Figure 21). Those species are being increasingly utilized
by the furniture manufacturers.8
Furniture and fixtures industries (SIC 25)
The furniture and related products industries (e.g. woodworking
industry) is heavily concentrated in Morristown (with 20 of the 46
plants) even though Knoxville has 17 plants (Figure 22). The average
number of employees per plant in Knoxville, however, is 161 persons;
whereas the average in Morristown is approximately 310 persons.
The furniture industry located in Morristown not because of the
local timber sources but because of cheap labor and good rail connec-
tions to large mills in the South and to market regions in the North.
With increasing reliance upon local timber, however, the growth
prospects for this industry are seen to be significant (Figure 23).
Communication with Bill Ogden, TVA Regional Planning Office,
Knoxville, Tennessee, May 3, 1973.
I ... I I I I I I I I I i I I I I I II I ... I I I... It r
OS S O S SE OE S2 0 SI 01 S
A s. I A . I I I . I 1 I 1 I I 1 I I I I I I I
I,,,,,~ ~ ,
I I I I I I I I I
EOl s 201 s
III I I I I I Jl ll ) I I I I l llI II I I l ill lu I I
sOl S hO I S 01 S O S 2s
Paper and pulp products industries (SIC 26)
The paper and pulp industry is concentrated in the Knoxville
area both in terms of employment and number of plants (Figure 24).
Intuitively, one suspects market-Kemand to be a strong locational
factor except for the plants in Morristown which are locationally
related to the by-products produced by the furniture industry.
Employment averages 78 persons with the largest plant employment only
190 persons. Total employment in the paper and pulp products industry
is relatively low compared to other industries, only 1377 employees
in 1973. The average growth rate, however, has been significant
Printing and publishing industries (SIC 27)
Printing and publishing includes local newspaper printing and
for obvious reasons is widely dispersed throughout the 16-county
region (Figure 26). Knoxville has the greatest concentration with
over 75 percent of the employment and 2/3 of the total number of
establishments. Total employment is low for the region (1670 employees
in 1970) with an average of 19 workers per plant. Growth has increased
along with population with no great expansion predicted (Figure 27).
Chemical products industries (SIC 28)
The shunning of large urban places by chemical plants noted in
Chapter II at first appearance is not substantiated by the regional
pattern of chemical industries within ETDD (Figure 28). Industries
are found in the other communities (e.g., Morristown and Oak Ridge)
but Knoxville is the leader in number of plants with all but six of
the 33 plants in the region. However, the Knoxville industries are
os S O S O E se oO S O s 01 s So
,T C II I I I |
sl 1I L
111 1 I I I I 1! i i 1
Il l I i i
I . .
I i l ;
.. ^ _i ._-
+ i --_ --
OS Sh Ot SE O0 SZ 0 S 1O SI 0
I I I I I I I I I
SI I I I I I I
II III I I I
SW z u
I l I I
C- ------ ----
_ _ I _ __ ____ ___
_ _i I
I ft SI I
relatively small (the largest has 69 employees) wh;hle the ENKA plant
located at Lowland (six miles south of Mlorrito-..n) employees over
4500 persons. Large employment figures in Oak Ridge and M'orristown
(Lowland) skew the average to 308 persons per plant; therefore, the
median 10 per plant is more descriptive of the true employment
picture. Employment in the chemical industry has been erratic in the
last two decades (Figure 29) and future expansion is difficult to
Petroleum refining and paving and roofing products (SIC 291
Employment and plant number are relatively insignificant relative
to total industry within the region with all but one plant (seven
employees) located in Knoxville (Figure 30). Little change in employ-
ment has been noted in recent years (Figure 31).
Rubber and plastic products industries (SIC 30)
Ten of the 12 rubber and plastic products industries are found
in Knoxville presenting over 60 percent of the employment (Figure
32). The average number of employees is 68 people with 20 employees
as the median. Between 1952 and 1973, the new industry growth rate
has averaged 16.5 percent but the total employment in 1952 was only
514 persons (Figure 33).
Leather products industries (SIC 31)
The leather product industries are few in number (15), most
occurring in Knoxville, Dandridge (near Jefferson City), and Morris-
town (Figure 34). The average expansion rate has been significant,
24.6 percent even though the base year employment (1952) was only 546
1111 1 1 1 1 1
lilt I IO I I0 I IS
I I I I I I l l I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 I
sOI s h01 s
cOI S 20I s
_________-T,________ /______\_ _____________ _______________ ___________________________
i^ k,-^r -~------ --- s
e- L L_________
os si Oh SE oE SZ OZ SI 0 S
I 1 l i l l i l i 1 I l i I 1
I i I1 i
I I !
11i i i
sOl s 01 s
OI s 1O
I i iI I I Ih SP
OS Sih A SE O S 0 S O[ S
11111l(1 I IiiilIi
liii1 I I I
. . . .. I I I I3
SUL #"I I LL .
' ' "
persons. Employment per plant averages 100 persons as does the
median, indicating a fairly symmetrical distribution. Expansion in
the future may be expected to continue but probably not at such a
high rate (Figure 35).
Stone, clay, and glass products industries (SIC 32)
These represent strong local raw material, local market indus-
tries, and thus their location is determined by the chance occurrence
of resources and the distribution of urban places. The spatial
pattern supports this observation (Figure 36). The greatest concen-
tration, as expected, is in the Knoxville vicinity with 2/3 of the
total number of plants (68). The average employment per plant is 32
persons. The median is 14 employees per plant. Expansion has kept
pace with population growth in recent years and probably will con-
tinue to expand in response to total regional growth (Figure 37).
Primary metal industries (SIC 33)
This industry type is distributed throughout the region simply
because of the large number of foundries which historically developed
in East Tennessee (Figure 38). The largest employment industry is
Alcoa Aluminum in Maryville-Alcoa with 5000 employees (70 percent of
the total employment). Median employment, however, is only 50 persons
per plant. Employment has dropped steadily since 1963 but is expected
to level out in the future (Figure 39).
Fabricated metal products industries (SIC 34)
This classification includes welding and machine tool industries
which usually develop locally and are local-market oriented. This
N E- 0
Sol hol S co Z I
I ---~F -4
pC- __ __
~-: I I I+T
I ~ i I -I
IO S. Oh SI I SZ I OZ SI ... S I
OS Sh Ob SIE OE Sa 02 ST l s F!S
I I I Ii I I I 1 lI II I I I I I
Iii 1) ii i i1 i I
DI s s0i s Oi s s
OS Sh Ot SE OE SZ o. ST 0o s
/11111 I I lii ii r H
1 c 1"V
. -. I=
I . .
explains the wide distribution of the industry Figure 40). Most of
the 73 plants are found in Knoxville (upproxi ,:itely 75 percent) indi-
cating some response to raw material availability which is greater in
the Knoxville area. The average number of persons employed is 45 but
the median is only 14 employees per plant. The industry has grown
significantly between 1952 and 1973 probably in response to national
growth in the transportation and recreational vehicle markets (Figure
Machinery (except electrical) industries (SIC 35)
The machinery industry is heavily concentrated in the Knoxville
region and to a lesser extent in Oak Ridge and Morristown (Figure
42). Many of these industries perform supporting functions such as
the refurbishing of industrial tools (for example, the woodworking
tool industries in Morristown). The largest employer has 639 employees
but the average for the region is 27 persons per plant and the median
is only 7 persons per plant. Total employment has remained fairly
constant in the last decade (Figure 43).
Electrical machinery, equipment, and supplies industries (SIC 36)
The electrical machinery industry has grown from 578 employees
in 1952 to 2726 employees in 1973, averaging a growth rate of 60.9
percent. The largest employer is Magnavox in Morristown with 716
employees in 1970. Average employment is 120 persons; the median is
40 persons per plant. Most of the industry is concentrated in Knox-
ville and dispersed lightly elsewhere (Figure 44). In recent years,
employment has expanded significantly (Figure 45).