v13! .1...
17:51.5...
51:33.12. ..

u I... .. 3.3155 2.
3.11:3..35 ..
1
1.3.5:...- 1
3.1.2,;-

:31
.4 ,

r
r.
Ira.
t 3.. 71....
.. .13....

It:
lilxtev...

x... . . .i
yu............4,.

.. 55......

5.1.5

...r

tasty!!!
a. . u. ..
.s.i.......u...r
,..
I»... -r...§.:........x «
!.......tina.. .a...i...2....2
. {2:3
3.

. at
391‘. tutti?!)
lair... .IJiu...’ I

N:A¢’ql..|n 7 .

7..
It»..tl'lx)r 2 I
n . v ..

...1¢..1\\5:..( .. 1.)"... ‘12:...)

I .31}...xr\n. .x ..
(1....

r:
. fail:

:

Ii. 2;. 1...
r......:l:ra e
Ixnlssltt.

r...

.isma:
311.9. .x 221......
1:3ztril 1::
It: 1]..
. hot—ail-

: . , .1...
Illerr..:.l.
. Eviklll. .1»

.J .aav.

.! 11)....

2:11;...nrlk
32.8....

1. at" .1. . . . .
. » ink 23.111:
$5.25....

I
2...... .
.‘lalylnlt? 3.8;
3;. 81:111....) I 1:.
353i» t!!ly..t... I.
.. L 5.3. lit1:.5:>
s... t 3 9|

.Q;....t_ 4
7:17:31

$1....
.li.\f.4.. 1.?!
till! 1.... 3.
£253.53.

arl'ill

 

7:.»
a.

2...: nio......~..lr. Iv
rtlttzlp .l
‘50‘

R

x
4!!!- Kit!
.1... ~)L|.
.‘II:€'. .rvg.
ul!!k..$lt1:i9)t

it

$5....vlgxrsit
.3..'\!5?s=uf)1sv‘
.121. I'll: . 3-1.}:—

y)..r,..

.3161 6.3)..- .5...» 59.x .\
5.}! IA}: 035...;

9:91.:- .s,

THESIS

}
/
‘4

J

llllllllllllllllllllllllllllllll|l||llllllllllllllllllllll

l
l
3 1293 01563 4748 I

This is to certify that the

thesis entitled

CLASSIFICATION OF CORPS OF ENGINEERS PROJECTS
FOR ECONOMIC IMPACT ASSESSMENT
presented by

Dennis Robert Becker

has been accepted towards fulfillment
of the requirements for

MaSter' 5 degree in

 

Park, Recreation & Tourism Resources

KW

Major professor

Date Ajléé}
/

0-7639 MSU is an Affirmative Action/Equal Opportunity Institution

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES retum on or before date due.

 

DATE DUE DATE DUE DATE DUE
Maven;

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmative Action/Equal Opportunity Institution
cinW.W3—D.1

CLASSIFICATION OF CORPS OF ENGINEERS PROJECTS
FOR ECONOMIC IMPACT ASSESSMENT

By

Dennis Robert Becker

A THESIS
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Park, Recreation & Tourism Resources

1997

ABSTRACT

CLASSIFICATION OF CORPS OF ENGINEERS PROJECTS
FOR ECONOMIC IMPACT ASSESSMENT

By

Dennis Robert Becker

Formal methods for performing economic impact assessments (EIA) are often
time consuming and expensive. This study developed a classification system that
shortcuts formal EIA processes by working as a “look-up” table in assigning the data
required for input-output modeling. That is, the system provides visitor spending profiles
and regional sales multipliers for Corps of Engineers (COE) projects that lack primary
data. Socio-economic variables from previous research were used to construct the
classification system. ANOVA results showed that the classification system successfully
segmented COE projects into homogeneous categories having statistically different class
means for many of the variables. Results of the discriminant analysis showed that fifty-
five percent of the projects were correctly classified, and that regional population, number
of retail establishments, per capita retail sales, percent retail sales, median household
income, and distance to the nearest metropolitan area were valid discriminators.
Recommendations for reducing the error in the classification system and for using it to

perform EIAS are provided.

ACKNOWLEDGMENTS

This thesis is a result of research funded by the US. Army Corps of Engineers
Waterways Experiment Station (Vicksburg, MS). Without their funding this project
would not have been possible.

I would especially like to thank Dr. Dennis B. Propst of the Department of Park,
Recreation and Tourism Resources. As my major professor and advisor, he was
instrumental in all stages of my research and study while at Michigan State University,
and supportive of my educational and professional endeavors. Additionally, Dr. Propst
provided valuable guidance in the development of the methods proposed in this project
and the practical applications of their uses. Valuable contributions were also made by
committee members, Dr. Daniel Stynes of the Department of Park, Recreation and
Tourism Resources, and Dr. Larry Leefers of the Department of Forestry. Dr. Stynes was
instrumental in the conceptual development of the “look-up” table proposed in this study.

Several people have been with me in seeing this project through, and in
supporting me through the difficult times. I would like thank Dr. Joseph Fridgen, Park,
Recreation & Tourism Resources department chair, for his generous contributions.
Above all, I would like to thank my fellow graduate students and peers for their help and
assistance, and most importantly to my friends and family for their unending support and

encouragement to go the extra mile. This thesis would not be possible without them.

iii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. vi
LIST OF FIGURES .......................................................................................................... vii
CHAPTER 1
INTRODUCTION ............................................................................................................... 1
Economic Significance of Corps of Engineers Recreation Areas ............................ 3
Study Purpose .......................................................................................................... 4
CHAPTER 2
REVIEW OF LITERATURE .............................................................................................. 6
Recreation Classification Systems ........................................................................... 7
Classification in Regional Economics ..................................................................... 9
Factors Influencing Recreation Spending .............................................................. ll
Variations in Regional Impacts of Visitor Spending ............................................. 16
Concepts and Tools of Input-Output Modeling in Recreation ............................... 20
Statement of the Problem ....................................................................................... 23
Study Objectives .................................................................................................... 24
CHAPTER 3
METHODS ........................................................................................................................ 26
Operational Definitions .......................................................................................... 26
Sample Population ................................................................................................. 28
Delineation of Economic Impact Regions ............................................................. 30
Secondary Data Sources ......................................................................................... 32
Classification Variables ............................................................................. 34
Twelve Lakes Study ................................................................................... 36
The Classification System ...................................................................................... 38
Tests for Validity of the Classification System ..................................................... 40
CHAPTER 4
RESULTS AND DISCUSSION ........................................................................................ 42
Determinants of Spending and Economic Impacts: Objectives One and Two ...... 43
The Classification System ...................................................................................... 47
Analysis of Variance .................................................................................. 54
Discriminant Analysis ................................................................................ 59
iv

CHAPTER 5

SUMMARY AND CONCLUSIONS ................................................................................ 76

Summary of Findings ............................................................................................. 76

Study Limitations ................................................................................................... 81

Recommendations for Management and Planning ................................................ 83
Procedures for Estimating Total Economic Impacts of Recreation

Spending .................................................................................................... 83

Recommendations for Future Study .......................................................... 88

LITERATURE CITED ...................................................................................................... 90

APPENDICES ................................................................................................................... 95

Appendix A. Counties Located Within 30 Miles of Corps of Engineers

Projects ............................................................................................ 95

Appendix B. Final Data Set of Independent Variables ....................................... 103

Appendix C. Independent Variable Descriptives ................................................ 108
Appendix D. Analysis of Variance Between Classes Using Dunnett’s T3

Multiple Comparison Method ...................................................... 114

Appendix E. Qualitative Descriptions of Initial Classification Categories ........ 119

Table 1.
Table 2.
Table 3.
Table 4.
Table 5.

Table 6.
Table 7.

Table 8.
Table 9.

Table 10.

Table 1 1.

Table 12.
Table 13.

Table 14.

LIST OF TABLES

Independent Variables ...................................................................................... 33
Initial Classification Dimension Breakpoints ................................................... 40
Pearson Product Moment Correlations for Independent Variables .................. 45
Initial Classification of Corps of Engineers Projects ........................................ 49
Average Per Person Trip Spending Within 30 Miles for Day-Users and.
Campers ........................................................................................................... 50
Average Per Person Trip Spending Within 30 Miles for Non-Campers and
Hotel Visitors ................................................................................................... 5 1
Levene Test for Homogeneity of Class Variance ............................................. 55
AN OVA Test for Differences in Class Means .................................................. 57
Results of the Discriminant Functions Obtained From the Independent
Variables in the Second Discriminant Analysis ............................................... 65
Structure Matrix of Correlations Between Variables and Standardized
Discriminant Functions in the Second Discriminant Analysis ........................ 66
Predicted Class Memberships of the Initial and the Second Discriminant
Classification Systems ..................................................................................... 67
Casewise Results of the Discriminant Analyses ............................................... 68
Structure Matrix of Correlations Between Variables and the Standardized
Discriminant Function in the Third Discriminant Analysis ............................. 73
Predicted Row Memberships of the Initial and the Third Discriminant
Classification Systems ..................................................................................... 74

vi

LIST OF FIGURES

Figure 1. Location of Corps of Engineers Projects Used to Develop the Classification
System. .............................................................................................................. 29
Figure 2. Diagram of Initial Classification Scheme .......................................................... 39

vii

Chapter 1

INTRODUCTION

The allocation of scarce resources is a critical issue facing many public resource
management agencies. Public agencies charged with the care and management of these
resources must secure public funding on a competitive basis based on perceived and real
benefits and impacts on public welfare. For socially optimal uses of limited public funds,
quantification of effects of resource allocations are necessary (Leitch and Leistritz, 1985).

Many public entities, including state and local governments, and federal resource
agencies carry out impact studies to quantify the economic effects of proposed actions or
policies on a region’s economy. The methods used to forecast and understand economic
behaviors and impacts created by proposed economic activities are collectively called
economic impact assessments (EIA) (Propst and Gavrilis, 1987). Building on theory and
knowledge in regional economics, socio-economic assessments, and mathematical
economics and econometrics, EIAs typically involve analyses of the intersectoral
relationships which exist within a region. Knowledge of the linkages and flows
(interdependencies) of economic activity within an economy allows for an increased
awareness of possible implications of proposed actions and policies (Propst and Gavrilis,

1987). In the context of public resource management agencies, it is possible with EIAs to

anticipate the economic consequences of a proposed action or policy, such as increasing
recreation user fees. It is also possible with EIAs to make annual projections required for
effective planning, project business output by economic sector, and show the
consequences of alternative development scenarios (Stynes and Propst, 1996).
Increasingly, EIAs are also used as performance indicators in developing agency budgets
and formulating policies (Alward and Lofting, 1985; Archer and Fletch, 1996; Johnson,
Oberrniller and Radtke, 1989). Estimates of economic impacts of visitor spending assist
the Corps of Engineers in justifying recreation budgets in Congressional hearings.
Economic impact assessments are utilized on a regular basis by all seven federal
land management agencies in the United States (U SFS, FWS, NRCS, BLM, NPS, BOR,
and COE) (Clarke and McCool, 1996). Agencies such as the US Forest Service (U SFS)
and the Fish and Wildlife Service (FWS) use EIAs to analyze the economic impacts of
designating public lands as wilderness for the purpose of endangered species protection.
The Natural Resources and Conservation Service (NRCS) uses EIAs to measure the
economic implications of different types of agricultural practices, such as conventional
versus alternative soil tillage techniques. The National Park Service (N PS) often uses
EIAs to measure the economic impacts of recreation activity, and the implications of
those impacts in terms of jobs resulting from visitor spending. Similarly, the Bureau of
Reclamation (BOR) utilizes EIAs to measure the economic implications of recreation on
lands developed primarily for water conservation (Clarke and McCool, 1996). As
illustrated, EIAs have a variety of uses in natural resource planning for federal agencies

as well as for state and local agencies for similar purposes.

Traditional methods of quantifying economic impacts of proposed actions or
policies on a region’s economy require extensive knowledge of EIA concepts,
requirements, and limitations. Additionally, methods which rely primarily on primary
visitor expenditure data often require large investments of time and money for data
collection and aggregation. Often times all that is needed for agency planning purposes is
a crude estimate of economic impacts. The development of simpler tools, aimed at aiding
in the estimation of economic impacts of changes in consumer demand or supply of
resources, would allow resource management agencies to quickly and cost effectively
justify or prioritize budgetary allocations. The research in this study focuses on the
development of a simplified economic assessment tool for the US. Army Corps of

Engineers (COE) as it relates to recreation visitor spending while visiting COE areas.

Economic Significance of Corps of Engineers Recreation Areas

Currently the Corps of Engineers operates and manages 462 water resource
development projects in the United States for purposes of commercial navigation, flood
control, hydroelectric power, and public recreation. A large majority of these projects
exist in close proximity to highly populated areas providing significant opportunities for
people of all ages and backgrounds to participate in recreation activities, particularly
water-based activities. Given that Corps of Engineers visitors account for roughly one
third of all recreation visitor days to federally managed recreation areas, Corps of

Engineers projects can play a significant role in the economic health and vitality of a

region (Jackson, 1997). While engaging in recreation activities at COE projects, visitors
spend money in several sectors of the economy for trip-related goods and services. For
instance, a family camping at a COE project may purchase groceries from a local store,
eat at a restaurant, or buy souvenirs. The degree to which the local economy is affected
by this type of spending behavior is potentially significant in light of the number of visits
taken each year to Corps of Engineers areas by campers and other types of recreation
visitors. With the aid of economic impact assessments, the Corps of Engineers is capable
of measuring the economic implications of visitor spending in terms of jobs, income, and

sales generated (Jackson, Stynes and Propst, 1994).

Study Purpose

The purpose of this study is to develop a classification system to obtain the
necessary information needed to quickly and cost effectively estimate impacts for Corps
of Engineers projects. The proposed classification system is one method that can be used
to reduce the cost and time constraints required of traditional economic impact
assessment techniques.

Critical to the development of the classification system is the identification of
factors which influence Corps of Engineers recreation visitor spending and the regional
economic impacts of that spending. Based on differing levels of these factors, economic
impact assessments can be estimated at Corps of Engineers projects lacking primary

visitor expenditure data. Differing levels of these factors provide the basis for the

classification of a sample of Corps of Engineers projects into relatively homogeneous
classes. Within each class there exists at least one COE project for which visitor
spending profiles and the economic impacts of that spending have been measured with
primary survey-based data. Assuming the dimensions of the classification system are
valid and accurately represent socio-economic conditions around COE projects, the
values from surveyed projects can be assigned to non-surveyed projects.

The information and tools developed in this study will allow the Corps of
Engineers an alternative method of quickly and effectively deriving the required inputs
for an economic impact assessment. With these inputs the Corps of Engineers can readily
estimate regional economic impacts of proposed recreation management strategies to
justify or prioritize budgetary allocations for recreation areas without having to collect

primary visitor expenditure data at each site.

 

Chapter 2

REVIEW OF LITERATURE

Measuring the economic impacts of recreation visitors with traditional economic
impact assessment methods is a time consuming and complex process. Sometimes this
complex process is necessary to predict visitor spending and to obtain accurate
measurements of that spending. However, often times all that is needed for agency
planning purposes is an estimate of economic impacts which can be obtained without
these complex models (Propst et al., 1985). The classification of recreation areas based
on site and regional characteristics is a less expensive approach that bypasses traditional
methods of measuring economic impacts. A review of literature is used to identify
applications of classification systems in recreation and regional economics. It is also
used to identify factors which influence the degree of visitor spending at recreation sites,
and the impacts of that spending on the local economy. Input-output modeling in
regional economics is also examined to provide a brief background of the economic

concepts important to EIAs and their application to the proposed classification system.

Recreation Classification Systems

Classification is an integral part of science, and often the degree of classification
is a measure of various disciplines’ advancement of knowledge. Chemistry reached its
first important stage of development with Mendeleev’s periodic table of the elements.
Botany reached its when Linnaeus developed a taxonomic, or classification system, that
furthered the understanding of plants and their relationships to each other (Gorelik and
Skripkin, 1989). Likewise, recreation and tourism has employed classification systems to
segment different types of visitors for marketing and management purposes (Bevins and
Zwick, 1993; Clark and Stankey, 1978; and Outdoor Recreation Resource Review
Commission, 1962).

Recreation management often uses some type of classification as one step in the
planning process (Bevins and Zwick, 1993). Armed with classification tools, agency
managers and researchers can compare visitor experiences related to opportunities
provided, identify patterns of visitor demand, and make short-term forecasts for
management and planning purposes. For instance, classifying campgrounds as being
rustic or modern can be used to determine the level of maintenance required for each
type. Two of the more-well known classification systems used in recreation planning are
the Outdoor Recreation Resource Review Commission (ORRRC) and the Recreation
Opportunity Spectrum (ROS). These systems are used to illustrate the use of

classification as a recreation planning and management tool.

The ORRRC in 1958 decided that an intensive nationwide study should be made
of outdoor recreation to provide a systematic approach for meeting growing recreation
demand after World War II. The purpose of the study was to identify and forecast
outdoor recreation demand in the United States until the year 2000. Additionally, the
commission was to identify available resources and the policies needed to ensure those
needs were adequately and efficiently met initially and in the future. The commission
recommended an inventory system which classified natural resources by type of resource
and recreational value ranging from “High-Density Recreation Areas,” (areas intensively
developed and managed for mass use) to “Primitive Areas” (undisturbed roadless areas
characterized by natural, wild conditions). With the classification system, planners and
managers could identify the available resources for a given area, and make decisions how
best those resources should be used. The implementation of this system was a major step
forward in a coordinated national recreation effort to provide a consistent and effective
method of planning for all land-managing agencies and promoted logical adjustment of
recreation activities to the available areas (ORRRC, 1962).

The ROS was developed as an outdoor recreation resource supply inventory and
classification system for multiple use natural resource planning. Developed by the US.
Forest Service, it is a framework for integrating recreational opportunities and non-
recreational activities. Unlike the ORRRC inventory system which based classification
solely on the physical setting of the resources, the ROS classes are defined by a spectrum
of recreational opportunities and their associated physical, social, and managerial
settings. Classes of the ROS range from “primitive” (provides an opportunity for

isolation to feel a part of the natural environment), to “modern urbanized” (provides an

opportunity to experience affiliation with individuals and groups, wildland challenges are
unimportant) (Clarke and Stankey, 1978). The central notion of the spectrum is to offer
recreation visitors alternative settings in which they can derive a variety of experiences.
If full ranges of opportunity settings are provided, changes in demand are more easily
accommodated because the kinds of features an activity requires are more likely to be
available.

Classification in recreation and tourism, not unlike other disciplines of science,
offers a means by which to further understanding of relationships between objects
important to planning and management. Building on a priori and statistical approaches to
classification in regional economics, a theoretically grounded system can be developed to
identify variations in regional economic impacts and visitor spending. Break points in
the data can be identified and non-surveyed Corps of Engineers projects can then be

assigned to appropriate classes.

Classification in Regional Economics

The primary building blocks of classification systems, including those used in
regional economics, are the patterns of relationships, phenomena, or processes which
exist between objects (Abler, 1971). Regional economics is concerned with explaining
why businesses and industries locate where they do, understanding consumer’s
purchasing behaviors, understanding and predicting regional development and growth,

and identifying the structure and implications of interregional transactions. Economic

10

impact assessment procedures are analytical tools used by regional economists to predict
patterns of regional and interregional relationships (Richardson, 1969). Likewise,
classification of socio-economic data such as demographic characteristics, geographic
locations, and perceived product benefits is used to segment consumers into
homogeneous groups based on market demand (Gunter and Fumham, 1992). Classifying
recreation visitors according to their patterns of spending and their regional economic
effects, gives researchers or regional economists an analytical tool capable of projecting
social and economic impacts created by proposed economic activities or shifts in
consumer demand (Bevins and Zwick, 1993).

Prior knowledge of interrelationships between variables should be the
fundamental building block for all classification systems, whether statistical in nature, or
not. Each classification system, developed from prior knowledge, allows for a systematic
discrimination of objects into homogeneous groups providing a means by which to
predict performance and behavior (Gorelik and Skripkin, 1989). Prior knowledge of
interrelationships between variables, based on general theory, is often used to segment
different populations into homogeneous groups. Thus, a priori classification of Corps of
Engineers projects can be a useful tool in describing and comparing attributes pertinent to
recreation planning and management (Bevins and Zwick, 1993).

Some studies in regional economics rely heavily on statistical classification to
identify market segments, and to verify hypothesized relationships developed from a
priori investigations of observations (Punj and Stewart, 1983; Bevins and Zwick, 1993).
Unlike a priori classification methods, many statistical methods, such as cluster and

factor analysis, make no prior assumptions about interrelationships between different

1]

populations. Cluster and factor analysis are purely empirical methods of classification
and as such are primarily inductive techniques. Punj and Stewart (1983) suggest that
caution should be taken when using inductive techniques as they lack a conceptual
(deductive) understanding of the relationships that exist, and individually provide little
utility for planning and management. Statistical methods, such as regression and
discriminant analysis, on the other hand rely on conceptual hypotheses. Cluster and
discriminant analyses are used in regional economics and marketing to identify
differences in consumer attitudes and preferences to predict consumer behavior.
Regardless of the method used to classify objects or to verify the classification,
statistical verification of hypothesized interrelationships which exist between consumers
and the areas where they purchase goods and services further enhances an EIA (Punj and
Stewart, 1983). As such, statistical verification of a priori relationships between
recreation visitors and regional economies allows researchers and managers to identify

the strengths and weaknesses of the a priori set of assumptions.

Factors Influencing Recreation Spending

There is a rich history of economic literature pertaining to relationships between
consumer demand for products and services, and industry supply factors relating to price
influences and inputs. Notably, classic studies by Samuelson (1963) and Schultz (1938)
provided economists with ‘operationally meaningful’ theorems and hypotheses on market
equilibriums, and the demand for goods and services based on price. However, relatively

little thought was initially given to spatial aspects within traditional economic theories.

12

Only within the past forty or so years have location theories, gravity—type spatial
interaction models, and regional economics become an important element of mainstream
economics extensively used today in policy and management decisions (Richardson,
1969; Fujita, 1990).

Location theory states that businesses tend to locate where they can maximize
profits relative to production costs. Thus, the spatial position of industries and consumers
is of foremost importance, overshadowing other motivations for location such as personal
considerations and satisfactions (Richardson, 1969). Location theory aids in identifying
regional interdependencies between specific sectors of the economy, such as the building
of a convenience store in relation to a newly opened marina. Inasmuch, location theory
allows for an estimation of the potential of new economic activity to attract new spending
to a region (Propst and Gavrilis, 1987).

Gravity-type spatial interaction models provide a theoretical base for analyzing
the spatial relationships between locations of economic sectors and consumers. It is
assumed that recreation visitors traveling to different sites will attempt to minimize trip
production costs by identifying goods and services that can be purchased for lesser
amounts, while maintaining satisfaction.l Differences in travel time and search costs
imply differences in total acquisition costs regardless of product prices (English and
Bergstrom, 1994). Building on Hotelling’s theory of industry agglomeration
(Richardson, 1969), recreation-related sectors, such as retail establishments, have a

tendency to cluster together. The clustering of economic sectors creates demand

 

lThe notion of minimizing trip cost while maximizing satisfaction is referred to as the Household
Production Theory (English and Bergstrom, 1994).

l3

extemalities, including reduction in search costs, and increases the likelihood that visitors
will find acceptable prices. Gravity—type spatial interaction models have recognized that
the locational distribution of goods and services is a critical determinant in minimizing
trip cost while maximizing satisfaction, and that the clustering of economic sectors is an
attractive force for economic activity. Building on this notion, retail businesses may
concentrate in and around recreation areas to provide goods and services to visitors,
consequently increasing a recreation site’s attractiveness (English and Bergstrom, 1994).
Various studies have focused on an assortment of regional and recreation site
characteristics thought to influence visitor expenditures. Lieber and Allton (1983), and
Lieber, F esenmaier, and Bristow (1989) focused on regional and public recreation site
aspects in Illinois. Lieber and Allton (1983) identified seven site and visitation
characteristics expected to relate to variations in levels of visitor expenditures. These
included: (1) type of recreation area (surveyed in a park or conservation area), (2)
significance of resources (subjective agency rating of attractiveness), (3) degree of
urbanization (whether a large urban area was within the region), (4) length of stay (day-
use or overnight), (5) special events (at time of survey), (6) recreation season, and (7)
distance traveled to recreation site (resident or non-resident). Their study revealed that
neither the type of recreation area nor the ability of the resource base to attract visitors
had much affect on average expenditures per party trip. They found greater non-resident
visitor spending at recreation sites in close proximity to urbanized areas than at recreation
sites in non-urban areas. They also found that visitor expenditures within the study
region increased as the distance traveled by non-residents increased, and increased during

special events and hunting seasons. Similarly, Lieber et al. (1989) concluded that the

l4

locational distribution of recreation sites and visitor origin were important factors
influencing levels of visitor spending. Surveys of public recreation users showed that
total recreation spending per year was higher in areas that possessed a greater number of
alternative recreation sites. Typically, spending per trip was lower within these regions.
However, overall economic impact to the region was greater. They also concluded that
the type of recreation site visited influenced levels of spending. For instance, visitors
were likely to spend more while visiting state parks with concessions and convenience
stores in the area, than while visiting conservation areas with relatively few services.
Results of the Dawson et al. (1989) study suggested that regional characteristics
are an important factor in determining visitor spending. Dawson et al. (1989) suggested
that variations in visitor spending are significantly influenced by the degree of economic
development within a region, and the amount of tourism and recreation development.
Results of visitor expenditure surveys in Great Basin National Park identified retail trade,
restaurants and taverns, hotels and motels, amusement services (e. g. golf courses and
theaters), and transportation services (e. g. gasoline and car rental) as the primary
recreation-related economic sectors where visitors spend money. Analyses of the
regional economic impacts from this spending indicated that increased visitation to the
region resulted in relatively small increases in local economic activity, illustrated by the
small amount of indirect effects. They concluded that in order for the region to increase
recreation and tourism revenues, indirect economic impacts of visitor expenditures must
be increased. For instance, restaurants should purchase their goods and services, used to
meet recreation visitor demand, from the local area to receive an increased impact of

visitor spending. Furthermore, it was hypothesized that increasing the supply and

15

diversity of recreation-related businesses would increase visitor’s length of stay and total
spending within the region by providing more opportunities to spend money.

Additional variables presumed to be important to the classification of recreation
visitors and their expenditure patterns have been investigated in many other studies as
well. Taylor et al. (1993) compared recreation visitors to historical sites with other types
of recreation visitors. They identified specific demographic and trip-related variables that
were thought to influence levels of visitor spending while visiting historic sites. Results
of the study indicate that average daily per-person expenditures increased as household
income increased. Furthermore, regression coefficients indicated that visitor
expenditures decreased as the party size increased, the number of nights within the region
increased, and if the group camped while in the area. Additionally, there was a positive
relationship between regional visitor expenditures and visiting a historical site. However,
historical sites were rarely the destination of visitors.

While researchers have a good idea of the economic sectors influenced by
recreation and tourism (Archer and Fletcher, 1996, Dawson et al., 1993; Douglas and
Harpman, 1995; and Johnson et al., 1989), there are only a limited number of published
studies aimed at predicting recreation visitor spending. Based on the previously reviewed
studies, it appears that travel distance, type of recreation activity, length of stay, type of
lodging, degree of urbanization, and degree of recreation-related sectors within a region
are the most common variables used to predict variations in visitor spending.
Additionally, demographic variables (e.g. age, income, and occupation), season, type of
recreation site, special events, and locational distribution of recreation sites have been

useful in identifying variations of visitor spending. Variables influencing levels of visitor

l6

spending may also contribute to variations in regional impacts of visitor spending. For
example, Taylor et al. (1993) demonstrated that significant levels of visitor spending had

a significant impact on a region’s economy.

Variations in Regional Impacts of Visitor Spending

Economic impacts of visitor spending may be divided into direct, indirect, and
induced effects. The sum of these three is termed the total effects. A change in final
consumer demand (direct effects) is a measurement of the immediate economic impacts
within a defined region resulting from the increase or decrease in consumer activity
within the region. For example, an increase in the number of visitors staying overnight in
hotels would directly yield increased sales in the hotel sector, as well as changes in wages
and salaries, taxes, and supplies and services. To continue the example, indirect effects
are the resulting impacts on sectors that supply goods and services to hotels, such as linen
supply companies. Induced effects of visitor spending are the changes in economic
activity resulting from household spending of income earned by employees and
contractors in the hotel and linen supply industries. The sales, income, and jobs that
result from household spending of added wage, salary, or proprietors’ incomes are the
induced effects. Together, indirect and induced effects are referred to as secondary
effects (Palmer, Siverts, and Sullivan, 1985).

A common tool for assessing the direct and secondary effects of visitor spending

is input-output (I/O) analysis (Leitch and Leistritz, 1985). Input-output analysis traces

17

the intersectoral flows of visitor spending in a region to identify changes in sales, income,
and jobs associated with participation in recreation activities. It may be used to guide
investment decision-making by demonstrating the income and employment changes
resulting fiom alternative policy scenarios. Input-output modeling is used to measure the
amount of indirect and/or induced activity associated with the direct effects. The numeric
representation of this measurement is termed a multiplier. For instance, the numeric
representation of the amount of respending of visitor dollars by a restaurant to purchase
more food and beverages to meet future demand is a multiplier (Palmer et al., 1985).
Generally, economic impact assessments utilize Type I, or Type II or III multipliers to
estimate a given impact on a region from a specific economic activity. Type I multipliers
are simply the ratio of the direct and indirect impacts on a region from a given economic
activity, whereas, Type II and III multipliers are the direct and secondary impacts on a
region, including induced effects (Stynes and Propst, 1996). 2

There are many kinds of multipliers that reflect secondary effects from different
measures of economic activity such as sales, income, or employment. Sales multipliers
in recreation (also termed output multipliers) measure the value of all sales required. to
meet visitor demand. Employee income multipliers measure the wages and salaries
necessary to produce this output, and employment is an estimate of the number of jobs
required to produce this level of sales (Jackson et al., 1994). Sales multipliers are often
used in EIAs because they provide a detailed picture of the sectors most affected by

changes in consumer spending (Palmer et al., 1985). The interdependencies those sectors

 

2Type II multipliers are essentially the same as Type 111, other than Type I I involves a technical difference
in how induced effects are computed.

18

require with other sectors within the region to produce goods and services indicate the
region’s ability to “capture” the secondary impacts of visitor spending (Stynes and
Propst, 1996).

Variations in sales multipliers from region to region are thought to be a fimction
of several factors including population demand, regional ability to meet consumer
demand, and spatial location. Population demand, as defined by Richardson (1969), and
Fujita (1990) is the primary factor in determining which goods and services will be
produced in a region. Richardson (1969) proposed that supply and demand could be
measured in terms of a population potential scale. The scale assumes that population
density, regardless of size, exerts an influence on consumers to purchase goods and
services from producers within a population center. Influence varies directly with size,
and reduces as distance increases. Not unlike central place theory, the population
potential model assumes that consumers will travel from complementary regions to
population centers to purchase goods and services in direct proportion to the size of a
population center. Central place theory is concerned with the hierarchy of
interdependence among trade centers, and the supply of goods and services to
surrounding populations (complementary regions). In theory, the central place hierarchy
results from a direct relationship between the size and functions of central places and the
distances consumers are from trade centers (Richardson, 1969; Fujita, 1990).

Richardson’s (1969) population demand model also assumes that per capita retail
sales increase as population increases, thus supporting the assumption that consumers
travel to population centers to purchase goods and services. At the point at which it is no

longer economically or otherwise feasible to travel to a particular population center,

l9

consumers will opt to purchase goods and services elsewhere. Additionally, areas with
higher populations should exhibit a greater amount of economic diversity within the
region due to a greater diversity of goods and services demanded. As the sales of goods
and services between economic sectors of a region increases, secondary impacts of visitor
spending should increase as well (Dawson et al., 1993), thus increasing regional sales
multipliers.

Regional sales multipliers may be influenced by other factors such as changes in
visitor spending. Depending on the degree of intersectoral dependencies within a region,
visitor spending may have little impact on the regional economy in terms of secondary
impacts. However, a significant increase or decrease in the amount of visitor spending
within a region has the potential to significantly affect the local economy. If support
services for hotels, such as food catering and linen supply services, are available within a
region, hotels businesses are not forced to seek these services outside the region. Hence,
the local area captures a greater percentage of the economic impacts of visitors seeking
overnight accommodations at hotels within the region. Additionally, Type III regional
sales multipliers are affected to a large degree by the induced household spending of
income earned from recreation related businesses. The respending of employee
household income within a local area is dependent on the number of places to purchase
household items and personal goods. If these places do not exist within the region,
induced effects of visitor spending may be lost. Therefore, even with high levels of
visitor spending within a region, economic impacts may be potentially small.

Past studies reviewed in this section have shown that visitor spending increases as

the number of places to purchase recreation-related goods and services increases. Lieber

20

and Allton (1983) found that annual resident and non-resident visitor spending increased
as areas became more urbanized. However, they also found that average per party trip
spending actually decreased for resident visitors in the same urbanized areas, indicating
that residents took trips more frequently and of shorter duration.

The process of tying together factors which influence visitor spending and the
factors which influence regional sales multipliers is a complicated process. A discussion
of the concepts used in 1/0 modeling and economic impact analysis is needed to
understand the importance of the intersectoral relationships that exist between recreation-

related businesses, and how they are impacted by visitor spending.

Concepts and Tools of Input-Output Modeling in Recreation

Economic base theory states that a region’s economy can be divided into basic
and non-basic sectors. Basic sectors are involved in the export of products, and non-basic
in the supply of local goods and services. All non-basic economic activities, especially a
region’s trade and service activities, are induced by the expansion or decline of basic
export industries (Richardson, 1969). Tourism is an export industry, therefore an
assessment of regional impacts produced from a change in economic activity of
recreation and tourism is feasible.

To quantify the economic impacts recreation visitors have on an area, I/O models
estimate the secondary effects of visitor spending within a defined region. These

secondary effects are calculated for the different types of recreation visitors (visitor

21

segments) by tracing the flow of dollars between the economic sectors required to
produce those goods and services visitors purchase (Stynes and Propst, 1996). Since
most of the categories in which visitor spending accrues do not exist as economic sectors,
visitor expenditures for each visitor segment must be margined to their respective sectors.
A bridge table that transforms visitor spending to a final demand vector handles the
process of margining. The final demand vector can then be applied to the regional I/O
model. Total economic impacts are estimated by multiplying the final demand vector by
a set of multipliers (Stynes and Propst, 1992).

Primary visitor expenditure data are often used to determine average visitor
spending by different visitor segments, and to determine where goods and services were
purchased (Bergstrom, Cordell, Ashley and Watson, 1990; Douglas and Harpman, 1995;
Johnson et al., 1989; and Propst et al., 1992). Due to the time and labor expenses of data
collection, alternative methods of [/0 analysis may be desired. Depending on the EIA’s
application, non-survey or partial survey techniques for estimating visitor spending and
regional impacts may be the preferred alternatives. Both techniques utilize secondary
data in the transaction matrix in the I/O model (Hewings, 1985). The transaction matrix
refers to the intersectoral relationships between businesses as to how much they purchase
from each other to provide goods and services to consumers (Palmer et al., 1985). The
non-survey-based technique uses existing input-output tables, generally of national data,
and modifies them for regional uses. The partial survey technique is similar to the non-
survey technique in that it utilizes national tables and modifies them for regional

purposes. However, the partial survey method also uses local data and updates regional

22

tables periodically. Therefore, many researchers regarded it as the best alternative to
costly primary survey-based data (Hewings, 1985).

One of the more common I/O models used in recreation planning and
management is IMPLAN. The US. Forest Service developed IMPLAN in 1979 as a
computer-based system capable of developing non-survey and partial survey models. It
provides a static view of economic activity down to the county level, including detailed
information on final demand and payments for recreation related sectors. The analytical
capabilities of the IMPLAN system can be separated into two broad categories: (1) the
estimation of impacts originating from changes in final demands, and (2) the evaluation
of constraints upon sectoral gross outputs. Economic impacts are expressed by the
changes in regional income and earnings, employment, gross output and various other
parameters (Alward and Lofting, 1985).

A limitation of using IMPLAN for assessing the economic impact of a change in
final recreation demand in a region is that it lacks clearly defined recreation sectors.
Identification of those sectors impacted by visitor spending that are unique to recreation
are identified in many research studies such as those done by Archer and Fletcher (1996),
Dawson et al. (1993), Douglas and Harpman (1995), Johnson et al. (1989), and Stynes
and Propst (1996). Through the use of the Mico-Implan Recreation Economic Impact
Estimation System (MI-REC), IMPLAN can be used for recreation applications by
attributing visitor expenditures for thirty three specific goods and services to recreation
related Standard Industrial Classification (SIC) sectors (Stynes and Propst, 1996). The
MI-REC model identifies these sectors within IMPLAN and automatically margins and

bridges visitor spending for the regional I/O model. Additionally, MI-REC imports the

23

impact reports produced in IMPLAN and generates an assortment of summary reports
important to an economic impact assessment (Stynes and Propst, 1996).

With knowledge of how economic impact analyses are done, methods and tools
can be developed to tailor regional I/O models to particular applications. This study is
concerned with developing a “look-up” classification scheme that short cuts the
traditional I/O approach altogether. This scheme does not require primary visitor
expenditure data nor I/O modeling software to assess the economic impacts of visitor

spending.

Statement of the Problem

Economic impact assessments have a variety of uses in recreation and tourism,
including quantitative evaluations of Corps of Engineers recreation programs and
facilities. Since they estimate economic impacts of changes in recreation demand or
shifts in agency policy and management strategies, the Corps of Engineers use EIAs to
justify or prioritize budgetary allocations. With prior knowledge of what differing types
of recreation visitors spend on different goods and services and what the regional
implications of that spending are, the Corps of Engineers can forecast economic effects of
proposed actions or policies. As previously stated, much of the data needed to make
estimates of visitor spending and its impacts on a region are obtain with survey-driven
expenditure studies. Given the expensive nature of collecting primary data, this study

proposes a classification system that can use existing spending data to estimate economic

24

impacts of Corps of Engineers agency policies and management strategies. Visitor
spending profiles and regional multipliers currently exist for a sample of COE projects
where primary survey-based data were previously collected (Propst et al., 1992). The
classification system can be used to assign these spending profiles and sales multipliers to
COE projects that lack primary expenditure data. Thus, the classification system would
greatly reduce time and costs associated with traditional I/O methods. However, to
assign these values to non-surveyed projects the factors that influence visitor spending
and its regional impacts must be identified. Once these relevant factors have been
identified, Corps of Engineers projects can be classified into categories based on differing
levels of these factors. The classification system developed in this study utilizes readily
available secondary data identified in the literature and examines their effectiveness in

explaining variations in visitor spending and regional impacts of the differing classes.

Study Objectives

The classification system should be capable of assigning non-surveyed Corps of
Engineers projects to classes for the purpose of accurately identifying visitor expenditure
data and the subsequent economic impacts of that spending. To develop a classification

system and utilize it as a “look-up” table, the following objectives must be met.

1. Identify Corps of Engineers project and regional socio-economic variables, from past

research studies, that influence levels of visitor spending.

2.

25

Identify Corps of Engineers project and regional socio-economic variables, from past
research studies, that provide indicators of levels of regional direct and secondary

effects of visitor spending to Corps of Engineers projects.

Develop a classification system based on those project and regional socio-economic

variables identified in the literature as being relevant to the first two objectives.

Determine the effectiveness and validity of the classification system using statistical

procedures.

Chapter 3

METHODS

A classification system capable of assigning spending profiles and regional
multipliers, for recreation areas where primary expenditure data does not exist, must rely
on secondary data sources. With secondary data sources, associated cost requirements
typical of input-output models using primary data can be reduced. Previously identified
factors thought to influence visitor spending and regional economic impacts are
incorporated into an initial classification matrix to segment a sample of Corps of
Engineers projects into homogeneous classes. This section will focus on the methods
used to classify Corps of Engineers projects and the methods used to test the validity of

the proposed system.

Operational Definitions

To standardize results with past research, the following terms as defined in Propst
et al. (1992) study of visitor expenditures at Corps of Engineers projects were used in this

study as operational definitions.

26

27

Region — An area of interest defined to identify what spending and economic activity to
include in an economic impact assessment. Measures of economic impacts
should include only businesses within the defined area. The economic region
used in this study is comprised of those counties that are within 30 road miles of
Corps of Engineers project boundaries.

Visits — A Visit is defined as the entry of one person onto a COE project to engage in one
or more recreation activities.

Non-Resident Visitors — Recreation visitors to Corps of Engineers projects who reside
outside the economic region used in the economic impact assessment.

Resident Visitors — Recreation visitors to Corps of Engineers projects who reside within
the economic region used in the economic impact assessment.

Visitor segments — Homogeneous groups of recreation visitors in terms of similar
spending patterns and preferences for certain goods and services. Each segment,
such as day-use boaters, or non-resident campers are characterized by
homogeneous spending profiles for recreation related goods and services.

Recreation-related businesses — Those retail and service sector businesses and industries
within a defined region directly affected by Corps of Engineers visitor spending.
Those businesses include, but are not limited to hotels and lodging, restaurants,
miscellaneous retail stores, grocery stores, gasoline and automotive services, and
recreation and amusement establishments.

Input-Output Models — An input-output model is a representation of the flows of

economic activity within a region. The model captures what each business or

28

sector must purchase from every other sector in order to produce a dollar’s worth
of goods or services.

Sales Multiplier — Multipliers capture the size of the secondary effects, usually as a ratio
of total effects to direct effects. The sales multipliers used in this study measure
the value of all sales, including indirect and induced (Type 111), required to meet
the demand created by visitors to Corps of Engineers projects.

Spending profiles - Average visitor spending for various goods and services purchased by
members of different recreation visitor segments to Corps of Engineers projects.
Spending profiles for COE visitors are a representation of only those expenditures

on recreation related goods and services within the defined economic region.

Sample Population

The units of analysis for this study are Corps of Engineers projects. A sample of
108 Corps of Engineers projects was drawn from the total population of 462 projects.
Locations of those projects are display in Figure l. The subset of 108 COE projects
consisted primarily of those projects having a high degree of recreation visitation in 1994.
The sample set of projects accounted for more than 65 percent of all COE visitation, but
less than one-third of the projects. The following sources provided the 108 COE projects:
the National Resource Management System (N RMS) developed and maintained by the
Corps of Engineers (US. Army Corps of Engineers, 1994), Propst et al. (1992), and

Ward, Roach, Loomis, Ready and Henderson (1996). The top one hundred COE projects,

29

 

 

 

 

 

Figure 1. Location of Corps of Engineers Projects Used to Develop the Classification
System (N = 108).
ranked by annual visitation in 1994 (U.S.C.O.E., 1994), were included in the study with
the assumption that these projects would account for a greater variety of visitor spending
than projects with a limited number of visitors. Thus, there would be sufficient variation
in spending and sales multipliers to develop a useful classification system into which any
of the remaining COE projects would fit. Other COE projects in the sample set include
those used by Propst et al. (1992) in a previous economic impact assessment study of
twelve projects. Primary visitor segment expenditure data from the 12 lakes study were
used as the basis for variations in spending explored in this study. Lastly, an additional
twenty-two COE projects were added to the classification system due to their previously
defined economic regions of 30 miles (Ward et al., 1996). After excluding overlapping

projects and the exclusion of an additional nine others, due to undefined service areas, the

30

final sample set of COE projects was reduced to 108 (Appendix A). Undefined service
areas, in most instances, were a result of a river project lacking clearly defined property
boundaries. In other cases inconsistent project data in the NRMS made regional
delineation difficult. Inconsistencies included multiple names for projects, and exclusion
of project boundaries. As was the case in Mississippi River Pool systems, data for
several of the areas were lumped together in a non-standardized format. The following is
a list of the excluded projects with their 1994 project visitation rankings in parentheses:
Mississippi River Pools 11-22 (2), Riverlands - Upper (6), Lake Okeechobbee and
Waterway (10), Cape Cod Canal (11), Mississippi River Headwaters Lakes Project (23),
Black Warrior and Tombigbee Lakes (26), John Paul Hammerschmidt Lake (76), Duluth

- Superior Harbor (95), and Mississippi River Pool No. 3 (97) (U .8. Army COE, 1994).

Delineation of Economic Impact Regions

Defining the regions surrounding Corps of Engineers projects was a vital step in
developing the classification system. The decision to include or exclude counties in
regional economic impact assessments may alter the extent of intersectoral economic
relationships, consequently modifying regional multipliers . A change in multiplier size,
in turn, may alter the class to which a given project is assigned. For instance the decision
to include a county with a Metropolitan Statistical Area (MSA) into an I/O model may

reclassify a particular project into a category with differing regional characteristics and

31

spending patterns. Not including a county with a MSA may result in a project being
classified into a category with lower spending patterns and regional impacts.

For the purposes of the classification system, the defined regions surrounding
COE projects were comprised of those counties which had a majority of their land mass
and economic activity within thirty road miles of a project’s shoreline. Therefore, some
areas that fell within thirty miles were not included in the impact region due to a lack of
economic activity and/or their relative geographic position to COE projects.
Additionally, counties were used to delineate economic impact regions because most
socio-economic data relevant to regional economics is aggregated at the county level and
up. For example, IMPLAN utilizes only county level data from US Census Bureau
sources to estimate I/O models (Alward and Lofting, 1985). Counties within each of the
defined economic regions for each of the COE projects are summarized in Appendix A.

Thirty miles was chosen to define economic regions for two primary reasons.
First, thirty miles is roughly the average distance across most counties in the east and
midwest where a majority of the projects are located. Secondly, the researchers in the 12
lakes study felt that thirty miles was a reasonable distance within which visitors could
recall whether or not they purchased goods or services (Propst et al., 1992). Other studies
of COE projects such as Ward et al. (1996) used thirty miles as functional economic
boundaries. However, they included only those counties that fell within thirty miles of a
project’s dam. A significant number of recreation areas exist on the upper ends of
reservoirs, opposite COE dams. Therefore to fully account for visitor expenditures within
thirty miles of recreation sites, this study utilized an operational definition of visitor

spending within thirty road miles of a project’s shoreline.

32

The first step in identifying counties to include in economic impact regions was to
use a computer-based mapping system with a geographic information system called
ArcView (ESRI, 1996). With the aid of ArcView, counties partially or wholly within
thirty miles of COE project shorelines were electronically downloaded into spreadsheet

format. The criteria used to determine inclusion or exclusion of counties were:

Include counties if:

1. They are entirely with the thirty-mile area surrounding the project.

2. They have greater than one half of their landmass within the thirty-mile
area.

3. They have half their landmass within the thirty-mile area and half
outside the area, ONLY IF they are easily accessible', OR an MSA
falls within the service area and can easily access the project.

4. They have less than half their landmass within the thirty-mile area
ONLY when a majority of the population of their MSA(s) is within the
thirty-mile area, AND easy access is possible.

Exclude counties if:
1. They are entirely outside the thirty-mile area surrounding the project.
2. They have less than half of their landmass within the thirty-mile area.
3. Transportation to a project is difficult, regardless of the amount of
landmass within the thirty-mile area.

lProject inaccessibility is defined as the difficulty or inability to travel to the
project site in thirty miles or less. Difficulty or inability to travel may be a result
of a lack of transportation systems (e. g., interstate highways and roads), or
presence of natural barriers (e. g., lakes and mountains).

Secondary Data Sources

The data used in this study are a representation of those factors identified in the
literature as having an influence on visitor spending and the regional impacts of that
spending. Selection criteria for the data were that they come from secondary data sources

that were readily available and that the data be easily obtainable. Secondary data sources

33

included the US. Census Bureau, the NRMS (U.S.C.O.E., 1994), and spending profiles
from Propst et a1. (1992). US. Census Bureau data were electronically downloaded from
government library sources from the following databases: USA Counties 1994 (U .8.
Bureau of the Census, 1996), 1992 Economic Census (US. Bureau of the Census, 1996),
and County Business Patterns 1993 (U .8. Bureau of the Census, 1996).

The variables thought to account for variations in visitor spending and regional
multipliers were summed for counties in each region for each of the 108 Corps of

Engineers projects. Variables and their data sources are listed in Table 1.

Table 1. Independent Variablesl

 

 

Independent Variables Secondary Data Source
Type 111 Sales Multipliers2 IMPLAN

Regional Population USA Counties 19943

Median Household Income (8) USA Counties 1994

Number of Retail Establishments 1992 Economic Census3
Annual Retail Sales ($) 1992 Economic Census
Percent Retail Sales4 County Business Patterns, 19933
Per Capita Retail Sales ($)5

Annual COE Project Visitation 1994 NRMS6

Distance to nearest MSA (miles)7 1994 NRMS

Percentage of Visits Attributed to Boating 1994 NRMS

Percentage of Visits Attributed to Camping 1994 NRMS

 

lComplete database of independent variables and their units of measurement are found in Appendix B

2Regional sales multipliers were calculated with input-output model software, IMPLAN. [/0 models were
estimated for each region in the study sample using 1992 Census Bureau data.

3Goverenment library data sources obtained from the US. Bureau of the Census (1996).

4Ratio of total non-agricultural sales in a region attributed to retail trade

5Computed from annual retail sales per 1000 residents residing in the region

6Corps of Engineers project data derived from the 1994 Natural Resource Management System (NRMS)
maintained by the Corps of Engineers.

7Road miles to nearest Metropolitan Statistical Area (MSA) from Corps of Engineer projects

34

Classification Variables

Variables were identified in the literature as being relevant to variations in visitor
spending and regional sales multipliers. The variables found to influence levels of Visitor
spending include location factors (Richardson, 1969), gravity-type spatial interaction
factors (English and Bergstrom, 1994), and economic industry agglomeration factors
(Fujita, 1990). Following the assumptions of the gravity-type spatial interaction models
presented by English and Bergstrom (1994), and Richardson (1969), a region’s retail
attractiveness increases as retail establishments agglomerate, which in turn causes an
increase in per capita retail sales. Additionally, average total visitor spending per
recreation trip for retail items may increase as a region’s retail attractiveness increases.
Thus, the number of retail establishments and their levels of sales were used as indicators
of “retail attractiveness.”

Variables thought to influence variations in regional sales multipliers were also
derived from the literature based on theories in regional economics as they pertain to
pOpulation demand models, location factors, and regional trade centers (Richardson,

1969; Fujita, 1990). These theories share in common the postulates that as a regional
trade center increases in population, a greater number of consumers are drawn to that area
to purchase goods and services. As the number of consumers increases, the demand for a
greater diversity of goods and services increases. Thus, as population increases the
diversity of economic sectors within a region does as well (Richardson, 1969). An
increase in population and the number of economic sectors within a region to meet

population demand should lead to greater output and sectoral interdependencies. Hence,

35

there should be higher regional sales multipliers in regions with higher populations and
greater economic diversity.

The literature suggests other variables are also factors influencing levels of trip
spending and regional impacts. Included are recreation site characteristics such as annual
visitation, type of activity participated in (e. g. boating and camping), and distance to
nearest economic service areas (distance to nearest MSA). Research such as that done by
Bergstrom et al. (1990), and Dawson et al. (1993) shows that total recreation visitation in
a region is capable of producing a significant enough amount of economic activity to
influence levels of visitor spending and sales multipliers. Other studies suggest that the
type of activity participated in is a significant factor in what visitor spending preferences
are while visiting an area (Douglas and Harpman, 1994; Lieber and Allton, 1983; and
Taylor et al., 1993). Richardson (1969) suggests that distance to economic service
centers is a critical factor influencing visitor spending. A metropolitan statistical area is
an example of an economic service center. An MSA is a core area (defined in counties)
containing a large population nucleus (greater than 50,000 population) with adjacent
communities having a high degree of economic and social integration (National Institute
of Standards and Technology, 1995). Additionally, demographic data, such as household
income and age influence levels of visitor spending. As income increases, expenditures

on recreation related goods and services would increase non-proportionately (Gunter and

F umham, 1992).

36

Twelve Lakes Study

The 12 lakes study conducted by Propst et a1. (1992) was used in the present
research study to identify variations in recreation visitor spending in Corps of Engineers
project regions. The goal of the study was to predict visitor expenditures associated with
the recreational use of a sample of twelve COE projects. The intent was to develop
nationally representative spending profiles for recreation visitors that portrayed spending
behavior in three ways: (1) total amount spent on a recreation trip, (2) distribution of that
spending among economic sectors, and (3) geographic location of spending in relation to
a given COE project.

Site selection criteria of the twelve sample lakes required projects to represent a
full range of spending behavior by COE visitors. Priority was given to projects that:

1. Received recreation use from a broad range of visitor segments.

2. Differed in adjacent county population sizes (assuming that variation in

population size is an index of opportunities for local spending).

3. Were located in different geographic regions of the continental United States.

4. Varied according to lake size and degree of recreation development.

Consistent with the literature review presented in this thesis, the authors argued
that types of activities visitors engage in, distance traveled to a destination, and length of
stay affect the total amount spent and the distribution of spending among economic
sectors. Thus, twelve user segments representing visitor origin, type of overnight
accommodations (if applicable), and activity participation were defined. Spending

profiles were developed for each of the visitor segments at each of the twelve projects

based on reported expenditures for different goods and services by different visitor

37

segments. The following visitor segments, defined by the 12 lakes study, were used to
obtain homogeneous segments of visitors with similar patterns of recreation-related
expenditures.
D/R/B: day user, resident, boater
D/R/NB: day user, resident, nonboater

D/NR/B: day user, nonresident, boater
D/NR/N B: day user, nonresident, nonboater

O/R/C/B: overnight user, resident, camper, boater
OfR/C/NB: overnight user, resident, camper, nonboater
O/NR/C/B: overnight user, nonresident, camper, boater
O/N R/C/N B: overnight user, nonresident, camper, nonboater
O/R/NC/B: overnight user, resident, noncamper, boater
O/R/N C/N B: overnight user, resident, noncamper, boater
O/NR/NC/B: overnight user, resident, noncamper, boater
O/NR/NC/NB: overnight user, resident, noncamper, boater
The survey of the twelve sample lakes yielded 3,144 on-site interviews. Of the
COE visitors interviewed on-site, 2,190 completed mailback questionnaires of recreation
trip expenditures. It was found from the expenditure surveys that average spending for
overnight visitors was more than 7.5 times greater than day-users. Additionally, it was
also found that visitors who engaged in boating activities typically spent more for
recreation related goods and services per party trip than visitor who did not boat. It is
assumed that projects with similar project and regional characteristics will attract similar
visitor segments with similar spending profiles. Therefore, it should be valid to assign

the spending profiles and regional multipliers developed in, the 12 lakes study to other

projects with similar site and regional characteristics.

38

The Classification System

The third study objective was to develop a classification system capable of
assigning non-surveyed Corps of Engineers projects to classes for the purpose of
identifying site specific visitor expenditure data and regional impacts of visitor spending.
The benefit segmentation approach used in marketing research (Gunter and F urnham,
1992) was the basis of the a priori classification system. The benefit segmentation
approach typically uses some type of classification technique to partition objects into
homogeneous classes. The classes are then described in terms of class means, standard
deviations, and standard errors of the variables used in developing the a priori classes.
The classes are then described in terms of additional variables not included in forming the
initial classes. The variables listed in Table 1 were used to describe the classification
classes summarized in Appendices C and D. The classes of variables in Table 1 were
also used to meet study objective four. Objective four was to determine the effectiveness
and validity of the classification system based on the independent variables identified in
the literature.

The sample of COE projects were initially classified into homogeneous classes
based on three of the independent variables: regional sales multipliers, number of retail
establishments within a region, and regional population. These three variables were used
in the classification scheme because of the magnitude of their importance relayed in the
literature. Additionally, both the numbers of retail establishments and regional
population appear to significantly influence visitor spending and regional impacts of that

spending (Richardson, 1969; Fujita, 1990). Likewise, Type III regional sales multipliers

39

were included in the classification scheme to identify levels of regional economic impacts
across different regions. Type III multipliers were used because they represent both the
direct and secondary effects of visitor spending. MI-REC was used with IMPLAN to
estimate Type 111 sales multipliers used in this study. The multipliers are aggregate
estimates of the amount of sales (output) generated in each sector present within each of
the defined regions surrounding the sample of 108 COE projects. Sales multipliers used
in this study were calculated as follows:

Type 111 sales multiplier = Direct Sales + Indirect Sales + Induced Sales

Direct Sales

The classification system for the sample of 108 COE projects consisted of a 3 X
4-classification matrix scheme with twelve categories, depicted in figure 2. Categories
ranged from regions having high numbers of retail establishments, high populations and
high sales multipliers, to regions with a low number of retail establishments, low
population, and low or medium levels of sales multipliers. Three of the twelve classes
were excluded from the classification scheme resulting in nine classes. The three

excluded classes included high, medium, and low sales multipliers classes within the

 

 

 

High Number High Number Low Number Low Number
of Retail Est./ of Retail Est./ of Retail Est./ of Retail Est./
High Population Low Population High Population Low Population
High Multipliers
Medium Multipliers

 

 

m Multipliers

 

 

 

 

 

 

 

Figure 2. Diagram of Initial Classification Scheme

40

“Low Retail Establishments/High Population” category. All regions having high
populations also had a high number of retail establishments. The classification
dimensions were established by identifying natural breaks in the initial classification
variables depicted in Table 2. Natural breaks were identified by personal judgment based
on variable means and distribution among classes (Appendix B). Ranges of retail
establishments and regional populations were large with a significant variation in ranges
above the break points. Due to a limited number of regions with high values of
establishments and populations, break points were established at lower levels to more
closely reflect medians. Corps of Engineers projects classified in this manner created

classes of projects with similar site and regional socio-economic characteristics.

Table 2. Initial Classification Dimension Breakpoints

 

 

High Medium1 Low
Number of Retail Establishments 1,350 + N/A 0 - 1,349
Regional Population 500,000 + N/A 0 — 499,999
Type 111 Sales Multipliers 2.3 + 2.0 - 2.29 0 - 1.99

 

'The number of retail establishments and total regional population was split into categories having high and
low levels with no designation of medium levels in the classification system due to the difficulty in
establishing natural break points.

Tests for Validity of the Classification System

Objective four was to determine the effectiveness and validity of the classification

system using statistical procedures. A Pearson product moment correlation table was

41

computed to determine the magnitude of relationships between the independent variables
chosen to represent those factors identified in the literature review. The correlation table
was used to determined if the identified variables were consistent with past theories in
regional economics, and to aid in the interpretation of the analyses of variances and the
discriminant analyses used to test the validity of the classification system.

Analysis of variance, or ANOVA, was used in this study to determine if class
means for each of the independent variables across the nine classes were statistically
different. The existence of statistical differences provides evidence that the classification
system has some validity in segmenting COE projects into homogeneous classes by
factors thought to influence spending and regional impacts of that spending. Analysis of
variance was also used to select independent variables to include in the discriminant
analysis.

Independent variables found to have statistically significant differences between
class means were included in the discriminant analysis to determine if they were good
discriminating variables between the different classes. If characteristics which
distinguish between the differing levels of visitor spending and regional impacts of that
spending are found for the sample of 108 Corps of Engineers projects, the remainder of
the 462 projects should also appropriately be classified by these characteristics. Thus,

discriminant analysis was used to seek a set of satisfactory discriminating variables.

Chapter 4

RESULTS AND DISCUSSION

The major goal of this study was to classify Corps of Engineers projects into
homogeneous groups based on differing levels of factors thought to influence levels of
visitor spending and the regional impacts of that spending. The intent of the
classification system is to have the capability of assigning spending profiles and sales
multipliers to COE projects without having to collect primary visitor expenditure data.

The results are presented as they relate to the four study objectives. The first
objective was to identify COE project and regional socio-economic variables from past
research that influence levels of visitors spending. The second objective was to identify
COE project and regional socio-economic variables from past research that provide
indicators of regional direct and secondary effects of visitor spending to recreation areas.
The third study objective was to develop a classification system based on differing levels
of the factors identified in objectives one and two. The fourth objective was to determine
the validity of the classification system using statistical procedures. The first and second
study objectives of identifying socio-economic factors that influence levels of visitor
spending and economic impacts were accomplished with the literature review. Therefore,

a majority of the discussion of results will focus on the last two objectives.

42

43

Determinants of Spending and Economic Impacts:
Objectives One and Two

To review, factors that influence levels of visitor spending include gravity—type
spatial interaction factors (English and Bergstrom, 1994), economic industry
agglomeration (F ujita, 1990), location factors (Richardson, 1969), and retail
attractiveness factors (Richardson, 1969). Independent variables chosen to represent
these factors were: regional population, number of retail establishments, per capita retail
sales, type of activity engaged in, distance to MSA, and median household income.

Factors that were identified in the literature that influence regional economic
impacts of visitor spending are based on theories in regional economics. Those factors
include population demand factors, regional trade center factors (Richardson, 1969;
Fujita, 1990), and location factors (Richardson, 1969). Independent variables chosen to
represent these factors were: Type III regional sales multipliers, regional population,
annual COE project visitation, number of retail establishments, per capita retail sales,
distance to the nearest MSA, and percent retail sales. Variables that were readily
obtained from current and reliable secondary data sources were used in this study to
represent the identified factors (Table 1). The independent variables chosen were used in
the classification system to segment Corps of Engineers projects into homogeneous
classes, and to test the validity of the system.

Observed correlations between the independent variables in Table l were used to
determine which relationships were consistent with theories in regional economics, and to

aid in the interpretation of statistical tests of the classification system’s validity. Many of

44

the variables thought to influence visitor spending and regional economic impacts are
significantly correlated (Table 3). For instance, regional population is nearly directly
proportional to the number of retail establishments (r = .997, p < .001). Fujita’s (1990)
population demand models and retail agglomeration theories state that business
establishments increase as population density increases. Consistent with theory, the
number of retail establishments surrounding COE projects increase at a similar rate to
regional population.

Further support of the population demand model is the high correlations between
population and per capita retail sales, and per capita retail sales and distance to the nearest
MSA. As theorized by the population demand model, consumers have an increased
propensity to purchase goods and. services from within population centers in proportion to
the size of the population center and distance traveled. Thus, as regional populations
increased around COE projects, per capita retail sales increased. Additionally, as regional
per capital retail sales increased, travel distance to the nearest MSA decreased.

Regional population was negatively correlated with percent retail sales (p < .001).
This finding is also consistent with the literature. Following Warntz’s population
demand model (Richardson, 1969), large regional populations have a greater diversity of
economic sectors by which to meet increased consumer demands (e. g. wholesale,
transportation, and manufacturing). Thus, the percentage of total sales attributed to retail
trade within highly populated regions will decrease. As further support, Type 111 sales
multipliers increased as the percentage of total sales attributed to retail trade decreased (p

< .01). This finding implies that regions with low percentages of retail sales contained

45

:5. v 0.

 

 

S. v a...
00. v 0_
000.0 $0.0 $0.0. 0m0.0 000.0 000.0 $0.0- 0000.0- 008.0- 000.0 0000000 000002
000.0 ~m0N.0 000.0 Nmmmd 03.0 $20.0- 000.0- $00.0- N030- 00000 00000 0000000
000.0 000.0 000.0 000.0- 0000. Nwom.0- $000- 020- <m2 00. 00080.5
000. 0 20.0 000.0- _00m.0- NS m0- mm 0 0.0- 0m0.0- 000.030 0.00000
83 ES 080. 08.0. 08.0- 00.0- 050000 6800
000.0 80.0 000.0 000.0- 000.0 000.0065 moo 000500
000.0 0300 LON.0 m0wm0 m0_0m 28.0.0 0:000 000
000.0 $00.0 mm 0 .0 30000003005 00000
83 23 8020000 08000
83 000032 33
$0 00.00000 30% @2000 000080 00000000 000000; $0 020m gum 000000000 002000004
2008003 00000 <02 00 0000000 0000000 moo 090% 20000 00000000 003m
000002 0000000 000055 00002 00000 000

 

00000E0> 3000000000000 000 0000000000 0000002 0000000 0000000 .m 050-0

46

more of the economic sectors needed to produce the goods and services purchased by
retail establishments than regions with high percentages of retail sales. However,
regional population was weakly correlated with Type 111 sales multipliers. Although
regions having high degrees of economic diversity also had high sales multipliers, some
sparsely populated regions also had high sales multipliers. Likely this is a function of
isolated areas with low populations meeting consumer demand within the region. Thus,
an adequate number of secondary sectors (e. g. retail, wholesale, transportation, and
manufacturing) which are dependent upon each to produce goods and services were
present in these regions to minimize economic leakages and increase sales multipliers.

Additionally, per capita retail sales and the number of retail establishments were
found to be positively correlated (p < .01). This result supports English and Bergstrom’s
(1994) gravity-type spatial interaction models that assume that consumers are attracted to
areas with clusters of economic sectors (i.e. retail establishments) where they can
minimize trip production costs and maximize satisfaction.

Several variables were not correlated with Type 111 sales multipliers. They
include all of the following: distance to the nearest MSA, regional population, number of
retail establishments, and number of COE visitors. Warntz’s population demand model
(Richardson, 1969) assumes that economic activity will increase as the distance to a
population density (MSA) decreases. Although the relationships between the two
independent variables are in the theorized direction, they were not significant. Other
variables found not to have significant correlations include COE project visitation and

regional population. From Propst et a1. (1992) study of twelve COE projects, it was

47

determined that a majority of visitors reside within the economic regions of the projects.
Hence, highly populated regions would likely have a greater number of visitors.
However, this relationship does not hold true. The data in Appendices B and E indicate
that regions with small reservoirs or limited recreational facilities often have smaller
visitation numbers regardless of the population size within the region. Thus, reservoir
size and degree of recreation facility development appear to be variables that mediate the
relationships between population size and visitation.

The variables chosen to represent factors which are thought to influence levels of
visitor spending and regional economic effects appear to be consistent with past research
and theory. Therefore, it is assumed that the independent variables listed in Table 1 are
appropriate indicators of the identified factors and are capable of segmenting Corps of

Engineers projects into homogeneous classes.

The Classification System

The third objective of the study was to develop a classification system based on
differing levels of factors though to influence visitor spending and regional economic
effects. The classification system should be capable of assigning non—surveyed Corps of
Engineers proj ects to homogeneous classes for the purpose of economic impact
assessment. Once the classification system has been established, predetermined spending
profiles and sales multipliers are available for each class. Thus, the classification system

works as a “look-up” table by assigning spending profiles and sales multipliers to non-

48

surveyed projects assigned to particular classes with similar economic characteristics. If
valid, using the classification system as a “look-up” table significantly reduces the time
and expense involved in many economic impact analyses. A step-by-step process of
assigning profiles and multipliers to non-surveyed COE projects is outlined in Chapter 5.

Following the benefit segmentation approach used in marketing research (Gunter
and Furnham, 1992), regional sales multipliers, number of retail establishments, and
regional population was used to segment COE projects into initial classes. As previously
mentioned, these three variables were used in the initial classification scheme because of
the magnitude of their importance relayed in the literature and because of readily
available, annual databases. Descriptive statistics of the independent variables listed in
Table 1, including the three classification variables, are summarized in Appendix C by
each of the classification categories derived from the scheme.

The sample set of 108 Corps of Engineers projects was classified into the nine
categories summarized in Table 4. In general, COE projects are reasonably distributed
across the classification categories. The high sales multiplier category contains 27
percent (29 of 108) of the projects. The medium multiplier category contains 47 percent
(51) of the projects, and the low multiplier category has 26 percent (28). Additionally,
the “High Retail Establishments/ High Population” category contains 29 percent (31) of
the sample of COE projects across all multiplier levels, while 23 percent (25) of the 108
COE projects are classified into categories with high numbers of retail establishments but
low populations. The remaining 48 percent (52) of the projects are classified into

categories with low retail establishments and low populations.

49

0005 000.0..— m. @003 ..0 00 000000 0. 300.800 $0050.00 00 00.000~

 

 

 

 

 

 

 

 

a. n 0 <0 02800 0. u 20 .0 u 70
0< .0w0000 .0 0000 3 m2 .000 0000000
02 .0000 0.00m <0 0000000
>00 .05. 00.50 0.. 0.0500
<> .0000 000000 Z... .3003. 0.00 A00 000000
0.... .8052 0 .2000 E .800 .523 3:00.02
Om 0.0.0 000 0.300 MO .03000 -.< 0000003 07. .0000 0:00 00.00 B04
00 000005 00.0... 0< 0000002 00.0 07. .0000 00v. .3 02 00000.. 00005.. m.
00 6.00.800 00.0.0 0.). 0000002 d .0000 0000 <0 .0000 00080.2
<3 6.000 00.0.0 _Q~ 000020.:Q 00 Z. .8002 E> 0x0 00.500 00:00.0 E
08 u 70 0.... .050 08.0 0.. u 20 0.... .06.. 003 0: u E
N... .000000 000.03 .00 00.0.0.0 :00 X... .0001. 080x00-
OS. 600000003 v.0 .0000 0.00000 0.00 .0...>0o000m
00 02.0.50 00 .5000 >0 005. 0:02
>3 0.0208800 00 0:05.200 00 .850 2.0 .0 8.3 an." - 0.0
X0. 6.0200000 7...- .0000 000000 <0 00033. v.0 .0000 0000.00 00 .0000 0000 00:00.02
x... 0.05.00 0000 <0 .0000 V.005 02 .03. 0-. 0000 000. .0001. 0:30.300 00 .0020 B000 8.0m
m< 000300). N... .0000 00000 E .05 000000 MO .0000 00000.03. 00 0.05000 000002
OS. .0003... 000.). 0C. 0000 00000 Z0 00000.0 .2 0000 X0. .0000 .000 000 00 00060000
<3 .000000 0030-. .000 0000.0: 0C0 00.300000. 00 0000000..- 800m a 20 0000.). 00.0
>0 00.00 .8000 :00 00.800sz <0 £00000 20.038080 <0 .0000 00300.0 30 .0080 000<
OS. .0082... .0 000.0 0.00 000000030 5 0< 0A000.). 00 5020: > X..- .0000 0006005 X..- .m00.00< >0
8. u 70 a n 70 5 550.3 a. u 20
<0 .0000: 302 0< .0000 0.00,.- 030 00000 .m .3 <0 40.0 00.0
Om .0000m 00.00 02 .0000 080.0000 <0 030007. 00.0.03 <0 .800 00000.0 A+ 0.0
XE .0000 0003 <0 000000.000 >3 00880.0 OS. .0000 0030080 0.4. .0000 .0 0300 0000.002
5 .0000 0000000 0< 0.0000 00m. 0< 002000 Z0 .0000000m E #0050000 0.0.00 00.00
0.). .000... 00 000000 _0< 00.50000 >v. $0.000 <0 .0000 00000000 X0 .0000 000000
m< £000.52 .90 .0000 :— EQ 000533: 0. <0 .0...>00.>0m .30 000.20 .0100 .0 0
88.80 .0250 8.0380 33. 88.80 0050 as. 33. 0. 08.83 800.000 0000
82.. 02.3 088003000 :53. 33 .+ 02.0 058003000 :93. .00: .+ 02.0 058003000 :93. 00.0

 

0000.500 2000.000 00 00.000 00 000.000.0020 .0000. .0 0.00.0

 

50

Recreation spending data were then provided for each of the nine categories. To
do so, average per person trip spending developed in the 12 lakes study (Propst et al.,
1992) were aggregated for day-user, camping, non-camping, and hotel spending
categories (Tables 5 and 6). The hotel spending category included only those visitors
who stayed in hotels or with family or friends while visiting COE projects. Spending
categories of day-users, campers, and hotels represent only visitors in those segments.
However, the non-camping visitor segment is an aggregation of both day-users and hotel
spending segments. Spending categories of residents and non-residents were aggregated
to obtain visitor expenditure profiles based on larger sample sizes, since many of the

original spending segments in Propst et a1. (1992) had relatively few survey respondents.

Table 5. Average Per Person Trip Spending Within 30 Miles for Day-Users and

 

 

Campers]
Average Trip Average Trip
Classification Corps of Spending Per Spending Per Percent
Class Engineers Project Day-User S2 Camper $2 Campers ("/o)3

Class I J. Percy Priest 17.71 (118) 64.84 (15) 20
Class II Willamette 10.93 (201) 66.77 (52) 34
Class [11 Gabe 13.09 (53) 62.15 (38) 13

Ouachita 10.35 (30) 69.13 (65) 18
Class IV None - - -
Class V McNary 14.92 (56) 65.54 (27) 2

Raystown 21.09 (88) 72.30 (150) 9

Shelbyville 10.55 (99) 63.77 (32) 14
Class VI Cumberland 22.45 (26) 77.81 (66) 49

Mendocino 24.44 (25) 102.90 (35) 17

Milford 16.48 (47) 54.50 (178) 13
Class VI] Sidney Lanier 19.06 (65) 54.81 (62) 8
Class VIII None - - -
Class IX Dworshak 10.75 (49) 40.79 (89) 1 1

 

lAverage per person trip spending derived from Propst et a1. (1992) study of 12 Corps of Engineers
projects.

2The number of survey responses are reported in parentheses (N). Average per person trip spending
includes both resident and non-resident visitors.

3Natural Resource Management System (Corps of Engineers, 1994)

51

Table 6. Average Per Person Trip Spending Within 30 Miles for Non-Campersl and
Hotel Visitorsz‘3

 

 

Average Trip Average Trip Percent
Classification Corps of Spending Per Spending Per Non-Campers
Class Engineers Project Non-Camper (5)4 Hotel Visitor (3)4 ("/o)5
Class I J. Percy Priest 24.92 (127) 1 19.56 (9) 80
Class II Willamette 1 1.44 (202) , 114.00 (1) 66
Class III Oahe 49.54 (86) 108.07 (33) 87
Ouachita 81.70 (98) 113.18 (68) 82
Class IV None - - -
Class V McNary 20.76 (61) 86.16 (5) 98
Raystown 45.10 (126) 100.71 (38) 91
Shelbyville 29.85 (125) 103.36 (26) 86
Class VI Cumberland 128.84 (113) 160.64 (87) 51
Mendocino 23.91 (29) 20.62 (4) 83
Milford 20.90 (64) 33.14 (17) 87
Class VII Sidney Lanier 55.21 (110) 107.42 (45) 92
Class VIII None - - -
Class IX Dworshak 16.23 (60) 40.64 (1 l) 89

 

lNon-Campers include both day-users and hotel visitors.

2l-lotel visitors include visitors who stayed in hotels or with family or friends while visiting Corps of
Engineers (COE) projects.

3Average per person trip spending derived from Propst et al. (1992) study of 12 COE projects.

4The number of survey responses are reported in parentheses (N). Average per person trip spending
includes both resident and non-resident visitors.

5Derived from the percentage of campers reported in the Natural Resource Management System (Corps of
Engineers, 1994).

In general, visitor spending across the classification categories are what one
would expect. Average per person trip spending was significantly higher for campers
than day-users (Table 5). Average spending for non-campers generally was lower than
for hotel visitors. Additionally, hotel visitors typically spent the greatest amounts on
recreation related items within 30 miles of COE projects (Table 6). However, campers at
Lakes Mendocino, Milford, and Dworshak reported spending more for related goods and

services than hotel visitors. Caution should be taken when comparing hotel visitors at

these projects, as the number of survey respondents was small (ranging from 4 to 17

52

respondents). Similarly, the number of respondents at many of the COE projects in the
Propst et al. (1992) study was small for hotel visitors.

Regions with high numbers of retail establishments and high populations
generally received higher levels of visitor spending within the local area (30 miles).
However, campers reported higher spending while visiting both Cumberland Lake and
Lake Mendocino which are located in sparsely populated areas with low numbers of retail
establishments (Class VI). Additionally, non-campers (Table 6) at the J. Percy Priest and
Willamette projects (Classes I and II respectively) reported fewer expenditures for
recreation related goods and services than many of the other non—camping segments in
categories with low numbers of retail establishments and low populations.

A good distribution of the COE projects in the twelve lakes study across the nine
classification categories is important for the purpose of assigning spending profiles and
sales multipliers to as many non—surveyed projects as possible. Of the nine categories,
seven are represented by one or more projects from the Propst et al. (1992) study.
Classes 1V and VIII were not represented by a project in the study. Hence, 90 of the 108
(83%) projects in the classification system are represented by primary visitor spending
data from the Propst et al. (1992) study.

Categories with two or more COE projects from the Propst et al. (1992) study
included Classes 111, V, and VI. Generally, visitors within these categories reported
similar spending patterns in each of the visitor segments (day—users, campers, non-
campers, and hotel Visitors). For instance, visitors to Lakes Oahe and Ouachita (Class
111) reported similar trip expenditures for both the day—user and camper visitor segments.

However, this pattern was not consistent as Lakes Oahe and Ouachita (Class III) had

53

differing spending for the non-camping visitor segment. Similarly, non—campers at
Raystown (Class V) and Cumberland Lake (Class VI) reported differing spending
behavior than the other projects in their respective classes. In fact, Cumberland Lake
non-campers and hotel visitors reported significantly higher expenditures than any of the
remaining eleven projects in the study. Further investigation reveals that Cumberland
Lake has an unique visitation pattern of houseboaters (classified by the COE as hotel
visitors) who spent a significantly higher amount on recreation related goods and services
(Appendix E, and Propst et al., 1992). Therefore, Cumberland Lake may be a poor
representation of non-camper and hotel visitor spending profiles for Class VI.

Caution should also be taken when using spending profiles from Lake Sidney
Lanier (Class VII) because of its close proximity to Metropolitan Atlanta. Due to the
selection criteria for choosing counties to include in the I/O model (Chapter 3), Fulton
county (which includes Metropolitan Atlanta) was excluded from the analysis. If Fulton
county was included in the I/O model that produced the Type 111 sales multipliers, the
Lake Sidney Lanier region would likely have been classified into Class 1. Hence, Lake
Sidney Lanier spending profiles may not be representative of visitors to Class VII
projects.

For the purpose of assigning visitor spending profiles and regional sale multipliers
to non-surveyed COE projects, analysis of variance was used to determine whether the
classification categories were statistically different from one another. Additionally,
correlations between independent variables and discriminant analysis were used to test

the classification system’s ability to segment projects into homogeneous classes.

54

Analysis of Variance

Analysis of variance (ANOVA) was used in this study to determine if class means
for each of the independent variables were statistically different across the nine classes.
The existence of statistical differences provides evidence that the classification system
has some validity in segmenting COE projects into homogeneous classes by factors
thought to influence spending and multipliers. Those variables found to have statistical
differences in means between classes were used in the discriminant analysis to measure
the amount of variance explained by the classification variables.

Analysis of variance is a statistical technique used in testing the null hypothesis
that the class means are equal in the population. It compares the sample variance
estimated from the class means to that estimated within the classes. If the observed
differences in class means are not statistically significant, they are assumed to be a
function of natural variability among sample means from the same population (Norusis,
1992). Analysis of variance procedures require the following assumptions: each of the
classes is an independent random sample from a normal population, and the variances of
the classes are equal (Norusis, 1992). To test the null hypothesis that class variances are
equal, a Levene test was used. The Levene test computes the absolute difference between
the independent variable’s value and the class means, and performs a one-way analysis of
variance on those differences (often referred to as Homogeneity-of-variance) (Norusis,
1992). Small significance levels of the Levene statistic (< .05) indicate that the null
hypothesis of equality of class variances should be rejected. Likewise, a large

significance level confirms that the variances are relatively equal. Results of the Levene

55

test for homogeneity of class variances are summarized in Table 7. The AN OVA
indicates that roughly half of the independent variables have equal variances across the
nine classes. However, each of the variables used in the initial classification scheme
(Type 1H sales multipliers, number of retail establishments, and regional population) do
not have equal class variances. The effect of violating the variance assumption is not
clear. Some authors argue that AN OVA is robust and the assmnptions need not be

strictly adhered to (Klecka, 1975; and Lewis-Beck, 1982).

Table 7. Levene Test for Homogeneity of Class Variance

 

 

Levene Between Within
A priori Variable Statistic Classes df Classes df Significance1
Type 111 Sales Multipliers 2.129 8 99 0.040
Regional Population 26.342 8 99 <0.001
Retail Establishments 24.984 8 99 <0.001
Per Capita Retail Sales 1.566 8 99 0.145
Annual COE Visitation 2.089 8 99 0.044
Percent Camping 1.030 8 99 0.419
Percent Boating 0.300 8 99 0.964
Distance to MSA 1.868 8 99 0.073
Percent Retail Sales 1 .139 8 99 0.344
Median Household Income 1.554 8 99 0.149

 

lSignificance less than 0.05 indicates rejection of null hypothesis of equality of class variances.

To test the null hypothesis that the class means for the nine classes are equal, the
between-classes mean square is divided by the within-classes mean square. This ratio of
the variance among class means divided by the number of observations within each class
is the F statistic. Small observed significance levels of the F statistic (<0.05) indicate that

the null hypothesis of equality of class means can be rejected. Hence, observations are

56

from statistically different populations. Results of the F statistic for each of the
independent variables between each of the classification categories are summarized in
Table 8. All the independent variables, with the exception of percentage of the COE
visitors who camped and total COE project visitation, had statistically different means
across the nine classes. Similar results were obtained from a Kruskal-Wallis
nonparametric analysis of variance, with identical variables identified as having
significantly different variances between classes. Therefore, results of the ANOVA are
assumed valid.

The existence of statistical differences provides evidence that the classification
system has some validity in segmenting COE projects into homogeneous classes by
factors thought to influence spending and multipliers. However, a significant F value is
reported if only two class means are unequal. In order to determine exactly which class
means were statistically different from others, a multiple comparison post hoc procedure
was used. The Dunnett T3 post hoc procedure was chosen because it assumes that class
variances are not equal (SPSS, 1997). Results of Dunnett T3 method are summarized in
Appendix D.

The Dunnett T3 multiple comparisons indicate that some of the independent
variables were not statistically different across the nine classification categories. Total
COE project visitation, percent campers, and percent boaters did not have significant
differences in means for any of the classes. However, the percentage of boaters
previously had a statistical significant ANOVA of 0.029 (Table 8). Therefore, these
variables were not good discriminators between the initial classification categories.

Distance to the nearest MSA and percent retail sales had significantly different means

Table 8. ANOVA Test for Differences in Class Means

57

 

 

Sum of Mean

A priori Variable Squares df Mean Square F Significance

Type [D Regional Sales Multipliers
Between Groups 5.059 8 2.13 0.632 54.347 <0.001
Within Groups 1.152 99 0.012

Regional Population
Between Groups 1.1E+14 8 7.6E+05 1.3E+13 6.325 <0.001
Within Groups 2.1E+l4 99 2.1E+12

Number of Retail Establishments
Between Groups 2.9E+09 8 4.2E+03 3.6E+08 7.354 <0.001
Within Groups 4.9E+09 99 4.9E+07

Per Capita Retail Sales
Between Groups l.5E+08 8 6.7E+03 1.9E+07 16.499 <0.001
Within Groups 1.1E+08 99 1.1E+06

Annual COE Project Visitation
Between Groups 5.6E+l3 8 2.3E+06 6.9E+12 1.497 0.168
Within Groups 4.6E+14 99 6.7E+12

Percent Campers
Between Groups 0.082 8 .1 1 0.010 0.864 0.549
Within Groups 1.171 99 0.012

Percent Boaters
Between Groups 0.283 8 .22 0.035 2.258 0.029
Within Groups 1.549 99 0.016

Distance to Nearest MSA
Between Groups 3.5E+04 8 29 4.3E+03 5.194 <0.001
Within Groups 8.3 E+04 99 3.3E+03

Percent Retail Sales
Between Groups 392.954 8 28.1 49.1 19 8.047 <0.001
Within Groups 604.306 99 6.104

Median Household Income
Between Groups 1.8E+09 8 2.4E+04 2.2E+08 16.432 <0.001
Within Groups 1.3E+09 99 1.3 E+07

 

58

between some classes. However, a majority of the class means for these variables were
not different. Thus, these variables too were not the best discriminators between the
initial classes.

Variables that stood out as consistently having significant differences in means
across many of the classes included: Type III sales multipliers, regional population,
number of retail establishments, per capita retail sales, and median household income.
The only class mean for Type 111 sales multipliers that was not consistently different from
other class means at the 0.001 level was Class VIII. Regional population, number of
retail establishments, per capita retail sales, and median household income exhibited
statistical differences for many of the initial classes at the 0.05 significance level.
Therefore, it is assumed that these variables, including Type III sales multipliers, were
good discriminators between the initial classes of COE projects. These variables also
provide evidence that the classification system has some validity in segmenting COE
projects into homogeneous classes by factors thought to influence spending and regional
impacts of that spending. To statistically determine the effectiveness of all the
independent variables in classifying COE projects into homogeneous classes, those
variables identified by ANOVA as having significant between class differences in class
means (Type 111 sales multipliers, regional population, number of retail establishments,
per capita retail sales, median household income, distance to the nearest MSA, percent

retail sales, and percent boaters) were used in discriminant analysis.

59

Discriminant Analysis

Several factors were identified in the literature as influencing visitor spending and
ensuing economic impacts. The independent variables listed in Table 1 represent many
of those factors. Independent variables found to have statistically significant differences
between class means were included in a discriminant analysis to determine if they are
good discriminating variables between the nine classes. “Once a set of variables is found
which provides satisfactory discrimination for cases with known group membership, a set
of classification functions can be derived which will permit the classification of new
cases with known memberships” (Klecka, 1975, p. 436). Thus, if characteristics which
distinguish between differing levels of visitor spending and regional sales multipliers are
found for the sample of 108 Corps of Engineers projects, the remainder of the 462
projects could be classified using these characteristics.

Discriminant analysis attempts to delineate between the classes by forming one or
more linear combinations of the discriminating variables. The mathematical
representation of these linear combinations of discriminating variables are called

discriminant functions (Morrison, 1969). The linear discriminant equation

Zi = [)0 + b1X1j+ b2X2i+ +anm'

is similar to the multiple linear regression equation. The X {3 represent the values of the

independent variable for the ith observation, the b’s are coefficients estimated from the

data, and Z,‘ is the discriminant score for the ith observation (Morrison, 1969). For

60

sample populations to be significantly different, their discriminant scores must be
different. Therefore, independent variables are chosen so that the values of the
discriminant function differ as much as possible between groups, or that the discriminant
scores maximize the ratio of between-classes sum of squares divided by the within-
classes sum of squares. The first function, in the case of two or more functions, has the
largest ratio of between-classes to within—classes sums of squares. All subsequent
functions are uncorrelated and have smaller sums of squares ratios than previous
functions (N orusis, 1986). Thus, the discriminant functions determine which class an
observation has membership in based on maximum discriminant scores. Additionally,
the larger the value of bj (discriminant coefficient for the jth variable), the more
important variable X i is in discriminating between classes (Morrison, 1969).

For the linear discriminant function to be “optimal,” that is, to provide a
classification rule that minimizes the probability of misclassification, certain assumptions
about the data should be met. Those assumptions are that each class should be from a
sample with a multivariate normal population, and the population covariance matrices
between classes should be equal (N orusis, 1986). However, given the robust nature of
discriminant analysis, many researchers feel that these assumptions need not be strongly
adhered to (Klecka, 1975). Although, Lewis-Beck (1982) cited other researchers that felt

that the violations of the assumptions may render the results almost useless and invalid.

61

Examination of the variables listed in Table 1 reveals that only the Type 111 sales
multiplier, per capita retail sales, and percent boating have normal distributions.3 Many
of the variables with non-normal distributions are heavily skewed in the positive
direction. Possibly this is a function of a majority of the sample COE projects being
drawn from the top one hundred projects ranked in visitation. If total regional population
in any way affects visitation, population factors may be skewing the data. Although
results of the correlation matrix (Table 3) show that visitation and regional population are
not highly correlated, many of the projects which had comparatively high visitation were
located in urban regions (Appendices B and E). Therefore, comparing these projects to
the remaining projects with relatively low visitation rates may influence the distribution
of the data. Additionally, since the test of equal covariance is sensitive to the normality
assumption (N orusis, 1986), equal covariance will likely be violated.

Equally critical to the validity of the discriminant function is the absence of
specification errors and measurements errors in the model. Absence of specification error
means that variables have not been improperly included or excluded from the analysis
(Lewis-Beck, 1982). Given the nature of the study, variables that were readily obtained
from current secondary data sources were included in the analysis. Although, the
variables used in this study are thought to represent factors identified in the literature, not
all of the identified factors are represented by an independent variable. Variables not

included in the analyses that are thought to influence levels of visitor spending and

*

3A Lilliefors test of normality was used to determine levels of significance of each of the independent
variables, with a Kurtosis test to examine variable skewness. To adhere to the normality assumption,
variables with nonnormal distributions were transformed by their natural log, 2 scores and squares.
However, each attempt to obtain a normal distribution to adhere to the normality assumption was
unsuccessful.

62

regional impacts include travel distance, length of stay, party size, and type of lodging.
Variables such as length of stay and type of lodging often are obtained from visitor
expenditure surveys. However, these surveys were not available for each project in the
sample set. Likewise, travel distance and visitor origin was not available from secondary
data sources.

Absence of measurement error is a critical assumption of any analysis. If the data
do not reflect what is intended to be measured, any generalizations are likely
inappropriate (Lewis-Beck, 1982). Many of the variables used in the classification
system were obtained from government census sources where measurement error was
assumed minimal. The levels of measurement error for visitation variables obtained from
the NRMS are suspect. The Corps of Engineers estimates total visitation by multiplying
the number of vehicles entering project areas by a predetermined number of passengers
(N RMS, 1994). Therefore, total visitation is only a crude estimate of true visits.
Additionally, length of stay bias is introduced in this measurement, as a visitor staying for
ten days is considered equal to a visitor staying only for a few minutes. However,
estimates of project visitation were not included in the discriminant analysis due to their
statistically insignificant differences between class means.

Separate discriminant analyses were performed to test the initial classification’s
ability to segment COE projects, to obtain a parsimonious set of discriminating variables,
and to predict sales multipliers. Initially a discriminant function was obtained for the
variables included in the classification scheme (Type III regional sales multipliers,
number of retail establishments and regional population). The purpose of the first

discriminant analysis was to determine the adequacy of independent variable break points

63

(Table 2) in segmenting COE projects into homogeneous classes. One might suspect that
the percent of projects correctly classified would be artificially inflated, as the data used
for estimation were also used for validation (Aaker et al., 1995). Therefore, additional
variables not included in the initial classification scheme were included in a second and
third discriminant analyses to obtain a parsimonious set of discriminating variables, and
to predict sales multipliers. Based on the AN OVA, variables found to have significant
differences in means between the nine classes (Type 111 sales multipliers, regional
population, number of retail establishments, per capita retail sales, median household
income, distance to the nearest MSA, percent retail sales, and percent boaters) were
included in the second and third analyses.

The first discriminant analysis (3-variable model) correctly classified 70.4 percent
of the projects (32 of the 108 projects were misclassified). Investigations of those
misclassified projects revealed that nearly half (14) of the 32 projects were misclassified
to the “Medium Multiplier/High Retail Establishments/Low Population” category (Class
V). All fourteen projects in this class were predicted to belong to the “Medium
Multiplier/Low Retail Establishments /Low Population” category (Class VI). Thus, the
break point between high and low retail establishments is likely the reason for
misclassification. Similarly, 6 of the 12 projects in Class I were predicted to belong to
Class II, III, or IV. The initial variables (sales multipliers, number of retail
establishments, and regional population) for each of the misclassified COE projects had
values close to break points listed in Table 2 (Appendix B).

Again, one might suspect that the percent of projects correctly classified would be

artificially inflated, as the data used for estimation were also used for validation (Aaker et

64

al., 1995). The purpose of the first discriminant analysis was to determine the adequacy
of independent variable break points (Table 2) in segmenting COE projects into
homogeneous classes. However, the three variable model is inappropriate for the goal
obtaining a parsimonious list of discriminating variables which are capable of segmenting
COE projects into homogeneous classes.

Therefore, a second discriminant analysis was performed with the following
variables: regional population, the number of retail establishments, per capita retail sales,
distance to the nearest MSA, percent retail sales, percent boaters and median household
income. The Type III regional sales multiplier was excluded from the second analysis
since, along with visitor spending profiles, sales multipliers are to be assigned (predicted)
for non-surveyed projects. Logically it would be inappropriate to include Type 111 sales
multipliers into the discriminant function since the function’s task was to predict the
multipliers. Results of the discriminant function using the above variables are
summarized in Table 9. Seven separate functions were calculated for each linear
combination of independent variables and used in the analysis.4 Functions one and two
accounted for 85.3 percent of the variance, with function one having an eigenvalue
greater than 1.0. Functions three through seven did not significantly contribute to the
analysis (Norusis, 1986). The Wilks’ Lambda for the seven functions was 0.085 at the
<0.001 significance level. The small lambda indicates that class means are statistically

different, and the within-class variability is small compared to the total variability

(Norusis, 1986).

 

4Eight functions were obtained by default due to seven independent variables in the analysis.

65

Table 9. Results of the Discriminant Functions Obtained from the Independent Variables
in the Second Discriminant Analysis.l

 

 

% of Cumulative Canonical
Function2 Eigenvalue Variance % Correlation
1 3.269 73.6 73.6 0.875
2 0.520 11.7 85.3 0.585
3 0.299 6.7 92.0 0.480
4 0.151 3.4 95.4 0.362
5 0.110 2.5 97.9 0.315
6 0.086 1.9 99.8 0.282
7 0.008 0.2 100.0 0.091

 

‘Variables included in the discriminant functions were number of retail establishments, regional
population, per capita retail sales, percent retail sales, distance to nearest Metropolitan Statistical Area,
percent boaters, and median household income.

2Wilks’ Lambda of the seven discriminant functions was 0.085 at the <0.001 significance level.

Independent variables with the highest correlations to the standardized
discriminant functions are thought to contribute more to the overall discriminant function
and are the best discriminating variables in the analysis. Median household income and
per capita retail sales had the largest correlations with the first standardized discriminant
function for the second discriminant analysis (.609 and .588 respectively). In the second
function per capita retail sales (.505) and the distance to the nearest MSA (-.263) had the
largest correlations (Table 10). However, caution should be taken when interpreting
correlations in the structure matrix, as strong intercorrelations between the independent
variables influence their magnitudes and signs. From the results of the correlation matrix
(Table 3) it was shown that the number of retail establishments and regional population
were highly correlated (0.997, with p < 0.001). Thus, their contribution to the
discriminant functions were shared and it was not possible to assess the degree of

importance of the individual correlations to the functions (Norusis, 1986).

66

Table 10. Structure Matrix of Correlations Between Variables and Standardized
Discriminant Functions in the Second Discriminant Analysis.

 

 

 

F unction1

Independent Variable 1 2

Number of Retail Establishments 0.3 522 0.199
Regional Population 0.306 0.226
Per Capita Retail Sales 0.588 0.505
Percent Retail Sales -0.397 -0.256
Distance to Nearest MSA -0.306 -0.263
Percent Boaters -0.13 8 -0. 101
Median Household Income 0.609 0.030

 

lFunctions three through seven have been excluded from the analysis since they do not significantly
contribute to the total variance. Functions one and two accounted for 85.3 percent of the variance.

2Within-classes correlations between discriminating variables and standardized discriminant functions.

Class membership of the sample of COE projects was predicted by statistical
classification using the functions obtained in the second discriminant analysis.
Comparison of the discriminant classification system to the a priori system resulted in
54.6 percent of projects (59) being classified into the correct class. Observation of the 49
misclassified projects reveals that 8 projects were misclassified in Class V, 7 were
misclassified in Classes 1 and VI, 6 in Classes 111 and IX, 5 in Classes II and IV, 4 in
Class VIII, and l in Class VII (Table l 1).

Misclassified COE projects generally fell into adjacent classes (i.e. predicted
Class I projects falling into Classes 11, IV or V, but none of the others). These projects
often had values of Type III sales multipliers, number of retail establishments, and
regional populations similar to the breakpoints (Table 2). However, six COE projects
were predicted by the discriminant function to belong to classes that did not directly

adjoin the initial classification category (Table 12). For instance, Pine Flat project (Class

67

Table 11. Predicted Class Memberships of the Initial and the Second Discriminant
Classification Systems.l

 

 

 

 

Initial Class Predicted Class Membership by the Discriminant Function
Membership I II III IV V VI VII VIII IX Total
I 52 1 l 3 2 - - - - 123
11 - 2 1 - 3 1 - - - 7
III - 1 4 - 1 1 - 1 2 10
IV 1 - - 9 l - 3 - - 14
V - 2 - - 6 3 - - 3 14
VI - - 2 - 3 16 - - 2 23
VII - - - 1 - - 4 - - 5
VIII - - l - 1 2 - - - 4
IX - - 1 - - 5 - - 13 19
Total 6 6 10 13 17 28 7 l 20 108

 

 

 

 

 

l54.6% of Corps of Engineers projects correctly classified.

2Number of Corps of Engineers projects predicted by the discriminant function to have membership in a
class.

3Number of Corps of Engineers projects predicted by the a priori classification system to have membership
in a class.

I), and Lakes Woodruff (Class VIII) and Dannelly (Class IX) was predicted by the
discriminant function to belong to Class 111. However, there were no observable patterns
among the values of the independent variables for these three projects.

Similarly, both projects Norfork and Pomme de Terre were predicted by the
discriminant function to belong Class IX when initially classified into Class III. Closer
observation of these projects reveals that they had similar values to each other as well as
to the mean values of Class IX for per capita retail sales, distance to the nearest MSA,

percent retail sales, and percent boaters (Appendix B and C).

68

Table 12. Casewise Results of the Discriminant Analyses (N=108)l

 

 

Initial Second Initial Third
Corps of Engineers Classification Discriminant Classification Discriminant
Project Category Analysis Category2 Analysis2
Canyon Lake 1 1 1 2*
Cheatham 1 4* l 2*
David D. Terry 1 2* l 2*
Hansen Dam 1 1 l I
J. Percy Priest 1 4* 1 2*
Pine Flat 1 3* 1 1
Saylorville Lake 1 I l 2*
Sepulveda Dam 1 1 l 1
Shenango River 1 5* 1 2*
Smithville Lake 1 5* l 2*
Whittier Narrows l l l 1
William H. Harsha 1 4* l 2*
Barkley 2 6* l 2*
Beaver 2 2 l 2*
Bluestone Lake 2 5* l 2*
Stockton Lake 2 2 l 1
Table Rock 2 5* l 2*
Whitney Lake 2 5* l 2*
Willamette 2 3* l 1
Bull Shoals 3 3 l 1
Englebright 3 3 l 3*
New Hogan 3 8* 1 l
Norfork 3 9* l 2*
Oahe 3 2* 1 l
Ouachita 3 3 1 l
Pomme de Terre 3 9* l 3*
Rathbun Lake 3 6* l 2*
Sharpe 3 3 I 2*
Waco Lake 3 5* 1 1
Addicks Darn 4 4 2 2
Alum Creek Lake 4 4 2 2
Blue Marsh Lake 4 7* 2 2
Bonneville 4 4 2 2
Chatfield Lake 4 4 2 2
Cherry Creek 4 4 2 2
Deer Creek Lake 4 1* 2 2
Grapevine Lake 4 4 2 2
Hartwell Lake 4 4 2 2
J. Strom Thurmon 4 5* 2 2
Joe Pool Lake 4 7* 2 2
Keystone Lake 4 4 2 2
Lewisville Lake 4 7* 2 2

 

'Discriminant analysis was computed using regional population, number of retail establishments, per capita
retail sales, percent boaters, distance to the nearest metropolitan statistical area, percent retail sales, and
median household income as independent variables.

2Type 111 sales multipliers were transformed into dummy variables and used as the dependent variable

*Misclassified Corps of Engineers projects

69

Table 12. Continued

 

 

Initial Second Initial Third

Corps of Engineers Classification Discriminant Classification Discriminant
Project Category Analysis Category Analysis
Oologah Lake 4 4 2 2
Cecil M. Harden 5 2* 2 1*
Center Hill 5 6* 2 2
John H Kerr 5 9* 2 3*
Kaweah 5 9* 2 3*
Lake 0' The Pines 5 5 2 2
McNary 5 5 2 2
Murray 5 2* 2 1*
Nolin River Lake 5 6* 2 2
Raystown 5 6* 2 2
Senecavill Lake 5 5 2 2
Shelbyville 5 5 2 2
Success 5 5 2 2
Texoma Lake 5 5 2 2
West Point Lake 5 9* 2 3*
Barren River Lake 6 6 2 2
Belton Lake 6 5* 2 3*
Black Butte 6 9* 2 2
Cordell Hull 6 6 2 2
Cumberland 6 6 2 2
Dardanell 6 6 2 2
Degray Lake 6 6 2 2
Eufaula Lake 6 6 2 2
Fort Gibson Lake 6 9* 2 3*
Greers Ferry Lake 6 6 2 3*
Harry S. Truman 6 6 2 3*
Laurel River 6 6 2 2
Lower Granite 6 3* 2 2
Mark Twain Lake 6 6 2 3*
Mendocino 6 5* 2 2
Milford 6 6 2 2
Millwood 6 6 2 2
Sam Rayburn 6 6 2 2
Somerville Lake 6 3* 2 1*
Summersville 6 6 2 2
Tenkiller Ferry 6 6 2 2
Wappapello 6 6 2 2
Wright Patman 6 5* 2 2
Allatoona Lake 7 7 3 2*
B Everett Jordan 7 4* 3 2*
Falls Lake 7 7 3 2*
Lavon Lake 7 7 3 2*
Sidney Lanier 7 7 3 2*
Monroe Lake 8 5* 3 2*
Rend Lake 8 6* 3 2*
W. Kerr Scott 8 6* 3 2*
Woodruff 8 3* 3 1*

 

70

Table 12. Continued

 

 

Initial Second Initial Third

Corps of Engineers Classification Discriminant Classification Discriminant
Project Category Analysis Category Analysis
Arkabutla Lake 9 9 3 3
Blue Mountain 9 9 3 2*
Canton Lake 9 9 3 3
Carlyle Lake 9 9 3 2*
Dale Hollow 9 6* 3 2*
Dannelly 9 3 * 3 3
Dworshak 9 9 3 3
Eastman 9 9 3 3
Grenada Lake 9 9 3 3
Hensley 9 9 3 3
Lake Celilo 9 6* 3 2*
Lake Seminole 9 6* 3 3
Lake Umatilla 9 9 3 3
Lewis and Clark 9 6* 3 2*
Nimrod 9 9 3 3
Philpott Lake 9 9 3 2*
Rough River Lake 9 6* 3 2*
Sardis Lake 9 9 3 3
Walter F. George 9 9 3 3

 

The remaining COE project that was predicted by the discriminant function to
belong to a class not directly adjoining the initial class was New Hogan. New Hogan was
initially classified into Class III, but predicted to belong to Class VIII. However, values
of the independent variables for New Hogan used to obtain the discriminant function
were not consistently similar to Class VIII variable means.

Classes that had poor classification rates (less than 40 percent correct
classification) were Classes 11 and VIII (Table 11). The discriminant function assigned
two of the original seven projects to Class 11. Of the five misclassified projects, three
were predicted to belong to Class V which has lower Type III sales multipliers and
similar values of per capita retail sales and percent retail sales. The discriminant function

assigned zero out of the original four projects to Class VIII. Two of the four projects

71

were predicted to belong to Class VI (Rend Lake, and W. Kerr Scott). Both Rend Lake
and W. Kerr Scott had values of retail establishments and Type 111 sales multipliers
similar to the break points (Appendix B).

Classes having greater than 60 percent correct classification rate included Classes
IV, VI, VII and IX, which represent 61 of the 108 COE projects. The discriminant
function assigned 9 or the original 14 projects to Class IV (64%). Three of the 5
misclassified projects in Class IV were predicted by the discriminant function to belong
Class VII which had lower Type 111 sales multipliers. Sales multipliers for both Joe Pool
and Lewisville Lakes were similar to the break points. The discriminant function
assigned 16 of the original 23 projects to Class VI (70%). Three of the 7 misclassified
projects in this category were predicted to belong to Class V. However, none had values
consistently similar to each other. Likewise, 4 of the 5 projects (80%) in Class VII were
correctly classified. The lone misclassified project (B. Everett Jordon) in this class was
predicted to belong to Class IV which was not an adjacent category. Additionally, the
discriminant function assigned 13 of the original 19 projects to Class IX (68%). Five of
the 6 misclassified projects were predicted to belong to class VI. With exception of
Lewis and Clark, each of the misclassified projects had similar values for each of the
independent variables included in the analysis. Lewis and Clark had a smaller sales
multiplier (1.52) than the remaining misclassified projects that had multipliers close to
the break point of 2.0.

To determine if a 54.6 percent “correct” classification rate is satisfactory, the
percentage of correctly classified Corps of Engineers projects was compared to the prior

probability that projects could be classified correctly by chance alone. To compute the

72

prior probability of chance classification, a proportional chance criterion was used
(Aaker, Kuma and Day, 1995). Proportional chance was computed by squaring the
original proportion of COE projects within each class (number of projects within each
class divided by 108). These squared class values were then added together to obtain a

proportional chance classification. The proportional chance rate was:

(12/108)2 + (7/108)2 +(10/108)2 + (14/108)2 + (14/108)2 + (23/108)2 + (5/108)2
+ (4/108)2 + (19/108)2 = 13.9

Thus, there is only a 13.9 percent chance that the COE projects might have been correctly
classified by chance. Compared to the 54.6 percent correct classification figure, the
seven variables included in the discriminant analysis perform significantly greater than
random chance classification for the sample of 108 Corps of Engineers projects. Thus,
the remainder of the 462 projects can be classified by these characteristics, and similar
classification results should be expected.

A third discriminant analysis was run using the same independent variables used
in the second analysis, only using the Type 111 sales multiplier as the dependent variable
as opposed to the class membership used in the first two. Type III sales multipliers were
transformed into dummy variables representing the three levels of multipliers (High:
2.3+, Medium: 2.0-2.29, and Low: under 1.99). The purpose of the analysis was to
determine how well Type 111 sales multipliers could be predicted within each of the given
levels (rows) of COE projects (i.e. Row 1: Classes I-III, Row 2: Classes IV-VI, and Row

3: Classes VII-IX). If the independent variables satisfactory discriminate between

73

different levels of Type 111 sales multipliers, multipliers can be effectively assigned to
non-surveyed projects based on the given regional and project characteristics.

Two discriminant functions were obtained from the three variable model (Rows l-
3) where the first function accounted for 84.6 percent of the total variance. Independent
variables within the first function with the strongest correlations were per capita retail
sales and percent retail sales, followed closely by regional population and the number of
retail establishments (Table 13). Therefore, it is assumed that per capita retail sales and
percent retail sales are the best discriminating variables between levels of sales
multipliers. The degree of discrimination for both regional population and the number of

retail establishments is unknown since they are highly correlated (Table 3).

Table 13. Structure Matrix of Correlations Between Variables and the Standardized
Discriminant Function in the Third Discriminant Analysis.l

 

 

Independent Variable Function 12
Per Capita Retail Sales -0.5893
Percent Retail Sales 0.419
Regional Population -0.338
Number of Retail Establishments -0.330
Distance to Nearest MSA 0.240
Median Household Income -0.210
Percent Boaters 0.179

 

lDiscrimianant function was obtained using Type 111 sales multipliers as the dependent variable.

2Function two was excluded from the analysis since is does not significantly contribute to the total
variance. Function one accounted for 84.6 percent of the variance.

3Within-classes correlations between discriminating variables and standardized discriminant functions.

74

Using Type III sales multipliers as the dependent variable resulted in a correct
classification of 59.3 percent (64) of the 108 COE projects (Table 14). Thus, there were
44 misclassified projects within the three levels of sales multipliers (Table 12). Eighteen
of those misclassified projects were in Row 1 where 16 projects were predicted to belong
in Row 2 by the discriminant function. Likewise, 16 projects were misclassified in Row
3 where all but one was predicted to belong in Row 2. Row 2 had a correct classification
rate of 80.4 percent (41/51). Of the misclassified projects in Row 2, 7 were predicted to

belong to Row 3.

Table 14. Predicted Row Memberships of the Initial and the Third Discriminant
Classification Systems.l

 

 

 

 

Predicted Row Membership by
Initial Row the Discriminant Function
Membership 1 2 3 Total
1 (Classes I-III) 112 16 2 293
2 (Classes IV-VI) 3 41 7 51
3 (Classes VII-IX) 1 15 12 28
Total 15 72 21 108

 

 

 

 

 

159.3% of original grouped Corps of Engineers projects correctly classified using Type 111 sales multipliers
as the dependent variable.

2Number of Corps of Engineers projects predicted by the discriminant function to have membership in a
class.

3Number of Corps of Engineers projects predicted by the a priori classification system to have membership
in a class.

Nearly all the misclassified COE projects had. values of Type 111 sales multipliers
close to the break points of the particular row the discriminant function predicted
membership for. In other words, many of the projects predicted to belong to Row 2

rather than Row 1 had values of Type 111 sales multipliers close to the 2.30 break point.

75

Likewise, many of the misclassified projects in Row 3 had values of sales multipliers
similar to the 1.99 break point. Thus, the initial classification system may perform
significantly better in the third discriminant analysis if ranges of Type 111 sales multipliers
were slightly increased to reflect the misclassification results.

To determine if a 59.3 percent “correct” classification rate is adequate, the
percentage of correctly classified Corps of Engineers projects was compared to the prior
probability that projects could be classified correctly by chance alone. The proportional
chance criteria (Aaker, Kuma and Day, 1995) computed a prior probability of chance
classification of 36.2 percent ((29/108)2 + (51/108)2 + (28/108)2 = 36.2). Thus, the initial
classification system performed moderately well when predicting rows (i.e. sales
multipliers) of COE projects. However, other independent variables need to be explored
to increase the third discriminant analysis’ ability to predict Type 111 sales multipliers.

Otherwise, some sales multipliers may be incorrectly assigned to COE project regions.

Chapter 5

SUMMARY AND CONCLUSIONS

Several factors were identified in the literature that influence levels of visitor
spending and regional economic impacts. Considering differing levels of these factors,
COE projects were classified into homogeneous groups. The resulting classes of COE
projects were statistically tested to determine the validity of the classification system. A
summary of the findings from these methods are presented, along with the limitations of
the design and the data. Additionally, the step-by-step process of assigning visitor
spending profiles and regional sales multipliers to non-surveyed projects is outlined for
the “look-up” table. Conclusions are drawn on the utility of the “look-up” table for

public recreation agencies based on the findings and limitations. Additionally,

recommendations are made for future testing and refinements to the classification system.

Summary of Findings

The first study objective was to identify Corps of Engineers project and regional
socio-economic factors, from past research studies, that influence levels of visitor

spending. Factors identified as influencing visitor spending include: gravity-type spatial

76

77

interaction factors (English and Bergstrom, 1994), economic industry agglomeration
(Fujita, 1990), location factors (Richardson, 1969), and retail attractiveness factors
(Richardson, 1969). Additionally, travel distance, type of recreation activity, length of
stay, type of lodging, degree of urbanization, and number of recreation related sectors
within a region were the most common variables used to predict variations in visitor
spending. Several variables were chosen from readily available secondary data sources to
represent these factors. They included number of retail establishments, regional
population, per capita retail sales, distance to the nearest MSA, median household
income, and percentage of boaters and campers. However, not all factors were included
because important variables such as length of stay, travel distance, visitor origin, and type
of lodging were not available at the secondary data level. Because of the importance
relayed in the literature, regional population and the number of retail establishments were
used to develop the initial, nine category classification system (Table 4).

Recreation spending data were then provided for each of the nine categories. To
do so, average per person trip spending developed in the 12 lakes study (Propst et al.,
1992) were aggregated for day-user, camping, hotel (visitors staying in hotels or with
family or friends), and non-camping (day-users and hotel visitors) spending categories
(Tables 5 and 6). The spending averages from the one or more of the twelve lakes from
Propst et al. (1992) were assigned to all the projects in a given class. Since none of the
twelve projects fell into classes IV and, VIII, additional primary spending data will need
to be collected. In the interim, spending averages from adjacent classes may be used for
Class IV and VIII projects. When two or more of the twelve lakes fell into a given class

(i.e., Classes 111, V and VI), the spending averages were generally similar enough to allow

78

any set of averages to be used. However, there were exceptions. In particular, estimates
of visitor spending for Cumberland Lake and Lake Sidney Lanier may be misleading.
Cumberland Lake (Class VI) experienced a unique visitation pattern of houseboaters who
spent a significant amount on trip-related items. Additionally, the Lake Sidney Lanier
region is located just outside Metropolitan Atlanta giving it regional characteristics much
more like Class 1 than Class VII. Thus, these projects may be poor representations of the
non-surveyed projects in their respective classes. This is not a problem for Class VI as
the spending averages for Lakes Mendocino or Milford may be used. However, Lake
Sidney Lanier is the only primary data project in Class VII. Judgment and the additional
information contained in Appendices B, C and B will be necessary in deciding whether or
not to use Sidney Lanier’s averages or adjacent class averages to represent Class VII.
Objective two was to identify COE project and regional socio-economic factors
that explain variations in regional economic impacts of visitor spending. Factors
identified in the literature that influence economic impacts are based on theories in
regional economics. Those factors included population demand, regional trade center
factors (Richardson, 1969; Fujita, 1990), and location factors (Richardson, 1969).
Several variables were chosen from readily available secondary data sources to represent
these factors. They included Type 111 sales multipliers, regional population, distance to
the nearest MSA, percent retail sales, percent campers and boaters, and COE project
visitation. Along with the number of retail establishments, Type 111 sales multipliers and
regional population were used to develop the classification system. From the results of
the discriminant analysis, it was shown that these three variables, along with distance to

the nearest MSA and percent retail sales were good discriminating variables. However,

79

COE project visitation, percent camping and percent boating did not have statistical
differences in class means, therefore, do not appear to be good discriminators.

Objective three was to develop a classification system based on project and
regional socio-economic factors identified in objectives one and two. The goal was to
use the smallest number of variables possible that would a) indicate variations in visitor
spending and b) explain variations in regional economic impacts. In addition, the data
had to be readily available from reliable data sets. Results of the initial classification
system (using Type 111 sales multipliers, number of retail establishments and regional
population) were encouraging. Corps of Engineers projects were distributed relatively
well across the nine classes. Only Classes VII and VIII were represented by five or fewer
projects, and Class VI by twenty or more. Additionally, seven of the nine classes were
represented by one or more projects in Propst et al. (1992) study of twelve COE projects.
Thus, more than 80 percent of the projects in the sample set of 108 are represented for the
purpose of assigning spending data. For those projects not represented by a COE project
(Classes IV and VIII) from the Propst et al. (1992) study, visitor spending data can be
estimated by the methods outlined in the “look-up” table section to follow in this chapter.

To accomplish objective four (statistical examination of the system) the first step
was to include each of the independent variables listed in Table l in an ANOVA to
determine if their means were statistically different between the nine classification
categories. All the independent variables, except for percent campers and total COE
project visitation, were found to have significant differences in class means. Therefore, it
was assumed that variables with different class means came from statistically different

populations and could be used to segment COE projects. The Dunnett T3 multiple

80

comparison post hoc procedure showed that Type 111 sales multipliers, regional
population, the number of retail establishments, per capita retail sales, and median
household income were consistently different across many of the initial classification
classes. Therefore, they were assumed to be good discriminators between COE projects.
Those variables that were not statistically significant across any of the nine classes
included the percentage of campers and boaters, and annual COE project visitation. The
remaining variables (percent retail sales and distance to the nearest MSA) were
statistically different across some of the classes. However, the percentage of retail sales
and the distance to the nearest MSA were not as good at distinguishing between classes
as those previously. All variables identified by the AN OVA as having significant
differences between class means were included in the discriminant analyses.

Based on results of the discriminant analyses, the classification scheme did an
adequate job of segmenting COE projects into homogeneous classes. Using the seven
independent variables identified as having statistically different class means, the initial
classification scheme and the discriminant function (Table 11) agreed on class
membership 55 percent of time. Classes that had poor classification rates (less than 40
percent correct classification) were Classes 11 and VIII (Table 11). Classes having greater
than 60 percent correct classification rate included Classes IV, VI, VII and IX, which
represent 61 of the 108 COE projects. In most cases, projects “incorrectly” classified
were near the borders of the classes defined by breakpoints (Table 2) derived by visual
examination of the data. For example, results of the last discriminant analysis (Table 14)
suggests that the initial classification system would perform significantly better if ranges

of Type 111 sales multipliers were slightly increased to reflect the misclassification results.

8]

For instance, the classification system might predict Type III sales multipliers better if the
center range was increased from 2.00 - 2.29, to 1.90 - 2.29.

Overall, it is concluded that the classification system is appropriate for assigning
spending profiles and sales multipliers to the majority (80 percent) of the Corps of
Engineers projects in the study sample. The statistical differences between class means
for variables used in the classification system provided evidence of the system’s validity
in segmenting projects. Results of the discriminant analysis supported the choice of the
appropriate discriminating variables upon which the classification was built.
Improvements to the classification system includes a) expansion of the ranges of the Type
III sales multipliers, and b) primary visitor spending data collection for representative

projects in Classes IV, VIII, and possibly VII (See “Study Limitations”).

Study Limitations

Given the dynamic relationships between recreation spending and the impacts of
that spending across regions, it is impossible to have a completely encompassing data set
which lacks specification errors. In meeting the first two study objectives, variables were
chosen from secondary data sources to represent those factors which influence visitor
spending and regional economic impacts. Relying on readily available secondary
sources, data were obtained from the US. Census Bureau, the Corps of Engineers, and
the Propst et a1. (1992) study of twelve COE projects. In this study, specification error is

present since not all of the factors identified in the literature were included in the

82

classification system. Not all factors were included because secondary data were not
available for important variables such as length of stay, group size, travel distance, visitor
origin, and type of lodging. These variables were identified in the literature as being
determinants of visitor expenditures.

Caution should be taken when using secondary data sources, as measurement
error can be a critical problem. Measurement error was suspected in this study, as Corps
of Engineers’ visitation data was questionable. Additionally, distance to the nearest MSA
is not the best indication of regional economic activity. Although, those regions with
MSAs within thirty miles are more economically diverse, other regions appeared to be
just as diverse but did not contain an MSA (Appendix B and E).

Small sample size is another limitation of the study. With small sample sizes it
was difficult to interpret some of the statistical comparisons. Likewise, many of the
classes in the initial classification system had small samples, except for maybe Classes VI
and IX which had 23 and 19 projects respectively. In addition, two of the nine classes
were not represented by Corps of Engineers projects from the 12 lakes study (Propst et
al., 1992). Hence, there were not enough COE projects with previously defined spending
profiles and sales multipliers to represent all 108 sample projects. This is not a serious
problem for multipliers because they can be derived from IMPLAN’s secondary data
sources and matrix computations. However, it is a problem for spending profiles because
primary data are required. Since there is no primary spending data for Classes IV and
VIII and since Lake Sidney Lanier may be more appropriately assigned to Class I than
VII, expenses must either be incurred for data collection or spending profiles must be

chosen from one of the neighboring classes in the “look-up” table.

83

Recommendations for Management and Planning

Given the limitations, the proposed classification system has utility in assigning
visitor spending averages and regional sales multipliers to Corps of Engineers projects
that lack primary visitor expenditure data. The classification system can be used as a
“look-up” table to place COE projects into homogeneous classes (Table 4) and find the
corresponding spending data and sales multipliers that reflect the socio-economic
characteristics of an area. The procedures for assigning the spending averages and sales
multipliers are described in the following section for the purpose of computing total

economic impacts of recreation visitors.

Procedures for Estimating Total Economic Impacts of Recreation Spending

To compute total economic impacts of visitor spending, three basic pieces of
information are required: total visitation by visitor segment, average visitor spending,
and a region’s effective sales multiplier. Average per person visitor spending (Tables 4,5,
and 6), and Type 111 sales multipliers (Table 4, and Appendix B or C) can be obtained
from the classification system. Total visitation by visitor segment must be derived from
total COE project visitation reported in the NRMS (U .S.C.O.E., 1994). This is done by
multiplying total project visitation by percent campers (U.S.C.O.E., 1994) to obtain total
camper visits. Total camper visits can then be subtracted from total project visits to

obtain non-camper visits.

84

To determine which spending profiles and sales multipliers from the classification

system to use in computing total economic impacts, the following steps are provided to

described the process:

1. If the COE project in question is one of the 108 in the sample set (Appendix A

and Table 4), this step may be skipped. For the remaining 354 projects, the next

step is to determine membership in one of the nine classes.

a)

b)

For those projects not in the sample set of 108, a determination must first
be made about which counties to include in the economic region. Users
have the option of defining their own regions based on the requirements of
their particular scenario, or using the criteria outlined in Chapter 3
(“Delineation of Economic Impact Regions”). It should be noted that the
spending profiles derived in the Propst et al. (1992) study and the
classification system are based on 30-mile regions. Thus, it is
recommended that the criteria in Chapter 3 be adhered to as closely as
possible.

Once a region has been defined, county level data can be aggregated from
Census Bureau sources for the remaining 354 projects to determine class
membership. The number of retail establishments can be obtained from
the 1992 Economic Census (US. Bureau of the Census, 1997) and
regional population can be obtained from USA Counties 1994 (U .8.
Bureau of the Census, 1997). These data can generally be downloaded

from libraries or US. Government web pages on the internet.

d)

85

Users can now begin the process of determining which class most
represents their project of interest. This is accomplished by first
determining which column in Table 4 a particular project belongs to (i.e.
“High Retail Establishments/High Population,” “High Retail
Establishments/Low Population,” and “Low Retail Establishments/Low
Population”) based on the data aggregated in the previous step. Once the
user has determined which column the project is most similar to, the
number of possible classes is reduced to three.

To determining which class within a particular column a project belongs,
Appendix B and E are used as a guide in finding a region with similar
economic characteristics. Appendix B provides a qualitative description of
each of the nine classes. The descriptions rely on information about each
of the twelve projects from the Propst et al. (1992) study. By studying
each of the three classes within the chosen column, a decision can be made
about which class most reflects the economic activity within the study

project’s region.

Once the final decision has been made as to which class most reflects the

economic activity of a particular COE project, the spending profiles can be

assigned. For all 462 projects, choose the appropriate spending figures from

Tables 5 and 6, one for campers and one for non-campers. Multiply the number

of campers and non-campers by the appropriate Tables 5 and 6 spending figures.

The results will be total spending (in 1990 dollars) separately for campers and

86

non-campers. It should be noted that the same can be done for day-users and

hotel visitors if the total number of visitors within each visitor segment is known.

a)

b)

If two or more COE projects with differing spending profiles from the
Propst et al. (1992) study represent a class, Appendix B should be used to
determine which project most represents the study project’s region.
Descriptions of each of the surveyed projects should serve as a guide in
determining which project is most similar to the project of interest in terms
of its economic structure.

However, if there are no projects from the Propst et al. (1992) study to
represent a particular class (i.e., Classes IV and VIII), the user has the
option of using Appendix E to choose a surveyed project that most reflects
the project in question. The user of the “look-up” table also has the option
of creating a new spending profile from the range of profiles given in
Tables 5 and 6. This can be accomplished by looking at the profiles given
for classes on each side of the unrepresented class. For instance, day-user
spending profiles can be created for Class VIII by taking the average
between profiles for Class VII and IX. However, it should be noted that
spending patterns between classes are not consistent and newly derived
estimates of spending may not reflect actual visitor expenditures.

Therefore, caution should be taken when using these methods.

Using the information in Appendices B, C and E, choose an appropriate sales

multiplier. To assign a multiplier from the given range of multipliers, users have

87

the option of choosing an average multiplier or one from either end of the range,
based on variable comparisons. This multiplier is multiplied by the appropriate
“capture rate,” which averages 65 percent across the twelve COE project regions
(Propst et al., 1992; Stynes and Propst, 1996). The result is called the effective
sales multiplier (Stynes and Propst, 1996). Since not all of the total camper and
non-camper expenditures are captured by a local region’s economy (i.e. there is
some “leakage”), the aggregate Type III sales multiplier must be reduced by the
capture rate to account for such leakage (See Stynes and Propst, 1996, for more

discussion of the capture rate and effective sales multiplier concepts).

4. To compute total economic impacts (sales), multiply the total expenditure figures
from step 2 by the effective sales multiplier from step 3. The results may then be
summed across campers and non-campers. Since the value will be in 1990
dollars, it may converted to more recent year dollars by applying the relevant
Consumer Price Index (CPI). Total economic impacts will be for sales (output)
only and may be converted to aggregate income (wages) or employment effects

with knowledge of appropriate income/sales or employment/sales ratios.

The “short-cut” method introduced here will allow recreation planners and
managers a means by which to quickly and cost effectively analyze economic impacts of
recreation visitor spending on regional economies. Although, the proposed classification
system has been developed for Corps of Engineers purposes, the “look-up” table method

has utility in estimating economic impacts of recreation visitor spending for many

88

applications. With modification of the data, the proposed “look-up” table method could
be used by different kinds of natural resource recreation agencies interested in quickly
estimating the regional economic impacts of their visitors. The Corps of Engineers uses
these estimates of economic impacts for quantifying the significance of their recreation
programs to Congress for budget purposes. Additionally, a quick and cost effective
system of estimating economic impacts will allow the Corps of Engineers and other
recreation agencies to assess the economic implications of proposed policies and
management plans prior to implementation. Hence, the economic ramifications of
proposed policies can be determined at non-surveyed projects prior to agency

implementation.

Recommendations for Future Study

Future research should focus on further testing and refinements of the
classification system. One recommendation is to collect primary data for a sample of
Corps of Engineers projects not included in the initial classification system. Economic
impacts would be estimated for each of these projects using the IMPLAN/MI-REC input-
output model. Results of the impact assessments would then be compared to results of
economic impact analyses where spending profiles and regional sales multipliers were
assigned to projects based on the “look-up” table methods. Comparison of the two

methods would indicate the degree of error for the “look-up’ table.

89

A method of testing the validity of the statistical classification systems would be
to aggregate data used in this study for an additional sample of Corps of Engineers
projects. This set would constitute a “hold-out” sample to test the discriminant function
(Aaker et al., 1995). Using the discriminant function obtained with the initial sample, the
hold-out sample of COE projects can be classified. If the percentage of correctly
classified projects for the hold-out sample is similar to that obtained with the first sample,
it is assumed that the discriminant function is valid. However, if the discriminant
fimction does a poor job of classifying the additional projects, a new discriminant
function should be obtained with different discriminating variables.

Results of the classification system can also be used to identify variables that are
good discriminators between classes of Corps of Engineers projects. Identified variables
can be used in multiple regression analyses to predict levels of spending, as well as
predict regional economic impacts from visitor spending. Thus, regression analysis can
be used to assign visitor spending profiles and regional sales multipliers to COE projects
for which no primary expenditure data have been collected.

Discriminating variables that explain uniqueness of regional economies such as
population densities, degree of urbanization, regional diversity, and other factors
identified in the literature but not included in the analyses, should be considered for
future investigations. Variables such as these may be useful in further refinement of the
classification system. Additionally, other variables are needed to further determine if the
initial classification scheme is valid or if class memberships are functions of other factors

not identified in this thesis.

LITERATURE CITED

LITERATURE CITED

Aaker, David A., Kumar, V., and George, Day S. (1995). Discriminant Analysis and
Canonical Analysis. Ch. 19. Marketing Research, 5th edition. New York, N.Y.:
John Wiley and Sons, Inc.

Abler, Ronald. (1971). Spatial Organization. Englewood Cliffs, N.J.: Prentice Hall, Inc.

Alward, Gregory S., and Lofting, Everard M. (1985). Opportunities for Analyzing the
Economic Impacts of Recreation an Tourism Expenditures Using IMPLAN. Paper
prepared for presentation at the thirtieth Annual Meeting of the Regional Science
Association.

Archer, Brian, and Fletch, John. (1996). The Economic Impact of Tourism in the
Seychelles. Annals of Tourism Research, 23(1), 32-47.

Bergstrom, J. C., Cordell, H. K., Ashley, G. A., and Watson, A. E. (1990). Economic
Impacts of Recreational Spending on Rural Areas: A Case Study. Economic
Development Quarterly, 4(1), 29-39.

Bevins, Malcolm 1., and Zwick, Rodney R. (1993). Classification of Recreation and
Tourism Communities Using Cluster Analysis and Data Management Techniques.
In 8D. Reiling (Ed), Measuring Tourism Impacts at the Community Level, 11-
19. Maine Agricultural Experiment Station Miscellaneous Report 374.
University of Maine, Orono, ME.

Clark, Roger N., and Stankey, George H. (1978). The Recreation Opportunity Spectrum:
A Framework for Planning, Management, and Research. General Technical
Report PNW-98. Portland, OR: US. Department of Agriculture, Forest Service,
Pacific Northwest Forest and Range Experiment Station.

Clarke, Jeanne N., and McCool, Daniel C. (1996). Staking Out the Terrain: Power and

Performance Among Natural Resource Agencies. Albany, N.Y.: State University
of New York Press.

90

91

Dawson, S. A., Blahna, D. J ., and Keith, J. E. (1993). Expected and Actual Regional
Economic Impacts of Great Basin National Park. Journal of Park and Recreation
Administration, 11(1), 45-57.

Douglas, Aaron J ., and Harpman, David A. (1995). Estimating Recreation Employment
Effects with IMPLAN for the Glen Canyon Darn Region. Journal of
Environmental Management, 44, 233-247.

English, Donald B.K., and Bergstrom, John C. (1994). The Conceptual Links Between
Recreation Site Development and Regional Economic Impacts. Journal of
Regional Science, 34(4), 599-611.

Environmental Systems Research Institute, Inc. (ESRI). (1996). Arc View GIS Version
3.0. Santa Fe, NM.

Fujita, Masahisa. (1990). Spatial Interactions and Agglomeration in Urban Economics.
In Manas Chatterji and Robert E. Kuenne (Ed), New Frontiers in Regional
Science, 184-221. Washington Square, N.Y.: New York University Press.

Gorelik, A.L., and Skripkin, VA. (1989). Methods of Recognition. New York, N.Y.:
Gordon and Breach Science Publishers.

Gunter, B. and Furnham, A. (1992). Consumer Profiles: An Introduction to
Psychographics. New York, NY: Rutledge, Chapman, and Hall, Inc.

Hewings, Geoffrey J .D. (1985). Regional Input-Output Analysis. Beverly Hills, CA:
Sage Publications, Inc.

Jackson, R.S., Stynes, D.J., and Propst, DB. (1994). An assessment of the National
Economic Effects of the US. Army Corps of Engineers Recreation Program,
Miscellaneous Paper R-94-2, US. Army Engineers Waterways Experiment
Station, Vicksburg, MS.

Jackson, Scott. (1997). Personal Interviews. US. Army Corps of Engineers Waterways
Experiment Station, Vicksburg, MS.

Johnson, R. L., Obermiller, F, and Radtke, H. (1989). The Economic Impact of Tourism
Sales. Journal of Leisure Research, 21(2): 140-154.

Klecka, William R. (1975). Discriminant Analysis. In Nie, Norman H., Hull, C. Hadlai,
Jenkins, Jean G., Steinbrenner, Karin, and Bent, Dale H. (Ed.). SPSS Statistical
Package for the Social Sciences: Second Edition. New York, NY: McGraw-Hill
Book Company.

92

Leitch, Jay A., and Leistritz, F. Larry. (1985). Techniques for Assessing the Secondary
Impacts of Recreation and Tourism. In D.B. Propst (Ed), Assessing the
Economic Impacts of Recreation and Tourism, 23-27. U.S.D.A., US. Forest
Service, Southeastern Forest Experiment Station, Asheville, NC.

Lewis-Beck, Michael S. (1982). Applied Regression: An Introduction. Beverly Hills:
Sage Productions.

Lieber, S. R., Allton, D. J. (1983). Visitor Expenditures and the Economic Impact of
Public Recreation Facilities in Illinois. In S. R. Lieber, and D. R. Fesenmaier
(Ed), Recreation Planning and Management, 36-54. State College,
Pennsylvania: Venture Publishing;

Lieber, S. R., Fesenmaier, D. R., and Bristow, R. S. (1989). Recreation Expenditures

and Opportunity Theory: The Case of Illinois. Journal of Leisure Research,
21(2),106-123.

Morrison, Donald G. (1969). On the Interpretation of Discriminant Analysis. Journal of
Marketing Research, 6, 156-163.

National Institute of Standards and Technology. (1995). Metropolitan Areas (Including
MSAs, CMSAs, PMASs, and NECMAs). FIPS PUB 8-6, US. Department of
Commerce. Washington: US. Government Printing Office.

Norusis, Marija J. (1986). SPSS/PC+ for the IBM PC/XT/A T: Advanced Statistics.
Chicago, IL: SPSS Inc.

Norusis, Marija J. (1992). SPSS/PC+ Base System User’s Guide: Version 5.0. Chicago,
IL: SPSS Inc.

Outdoor Recreation Resources Review Commission. (1962). Outdoor Recreation for
America, Report to the President and Congress. Washington, DC: US.
Government Printing Office.

Palmer, Charles, Siverts, Eric, and Sullivan, Jay. (1985) IMPLAN Version 1.1: Analysis
Guide. U.S.D.A., Forest Service, Land Management Planning System Section,
Fort Collins, CO.

Propst, Dennis B, and Gavrilis, Dimitris G. (1987). Role of Economic Impact
Assessment Procedures in Recreational Fisheries Management. Transactions of
the American Fisheries Society, 116, 450-460.

93

Propst, D. B., Gavrilis, D. G., Cordell, H.K., and Hansen, W. (1985). Assessing the
Secondary Economic Impacts of Recreation and Tourism: Work Team
Recommendations. In D.B. Propst (Ed), Assessing the Economic Impacts of
Recreation and Tourism, 52-63. USDA. Forest Service, Southeastern Forest
Experiment Station, Asheville, NC.

Propst, Dennis B., Stynes, Daniel J ., Lee, Ju Hee, and Jackson, R. Scott. (1992).
Development of Spending Profiles for Recreation Visitors to Corps of Engineers
Projects, Technical Report R-92-4, US. Army Engineer Waterways Experiment
Station, Vicksburg, MS.

Punj, Girish, and Stewart, David W. (1983). Cluster Analysis in Marketing Research:
Review and Suggestions for Application. Journal of Marketing Research, 20,
1 34-148.

Richardson, Harry W. (1969). Regional Economics: Location Theory, Urban Structure,
and Regional Change. New York, N.Y.: Praeger Publishers.

Rielly, W. J. (1929). Methods for the Study of Retail Relationships. University of
Texas Bulletin, 2944. In Harry W. Richardson (1969), Regional Economics:
Location Theory, Urban Structure, and Regional Change, 132-137.

Samuelson, Paul A. (1963). Foundations of Economic Analysis. Cambridge, MA:
Harvard University Press.

Schultz, H. (1938). The Theory and Measurement of Demand. In Paul A. Samuelson
(1963), Foundations of Economic Analysis. Cambridge, MA: Harvard University
Press, 104-116.

SPSS, Inc. (1997). SPSS: Advanced Statistics 7.5. Chicago, IL: SPSS Inc.

Stynes, Daniel J ., and Propst, Dennis B. (1996). MI-REC 2.0 User’s Manual.
Department of Park, Recreation and Tourism Resources, Michigan State
University, Michigan State University Agricultural Experiment Station, East
Lansing, MI.

Taylor, D. T., Fletcher, R. R., and Clabaugh, T. (1993). A Comparison of
Characteristics, Regional Characteristics, and Economic Impact of Visitors to
Historical Sites with Other Recreational Visitors. Journal of Travel Research,
32(1), 30-35.

US. Army Corps of Engineers. (1994). Natural Resource Management System (NRMS).
US. Army Corps of Engineers, Department of the Army, Washington, DC.

94

US. Bureau of the Census. (1997). County Business Patterns, 1993. CD-ROM.
Washington DC: The Bureau, [1996].

US. Bureau of the Census. (1997). Economic Census, 1992: Report series I H. CD-
ROM. Washington, DC: The Bureau, [1996].

US. Bureau of the Census. (1997). USA Counties 1994. CD-ROM. Washington, DC:
The Bureau, [1996].

Ward, F. A., Roach, B. A., Loomis, J. B., Ready, R. C., and Henderson, J. E. (1996).
Regional Recreation Demand Models for Large Reservoirs: Database
Development, Model Estimation, and Management Applications. Technical
Report R-96-2, US. Army Corps of Engineers Waterways Experiment Station,
Vicksburg, MS.

Yen, Steven T., Adamowicz, Wiktor L. (1994). Participation, Trip Frequency and Site
Choice: A Multinominal-Poisson Hurdle Model of Recreation Demand.
Canadian Journal of Agricultural Economics, 42, 65-76.

APPENDICES

APPENDIX A

Appendix A. Counties Located Within 30 miles of Corps of Engineers Projects

 

 

Project Name County Project Name County
Canyon Lake, TX Bexar Smithville Lake, MO Leavenworth, KS
Blanco Wyandotte, KS
Comal Buchanan
Guadalupe Clay
Hays Clinton
Kendall Platte
Cheatham, TN Cheatham Whittier Narrows, CA Los Angeles
Davidson Orange
Dickson
Houston William. H Harsha, KY Bracken
Montgomery Campbell
Robertson Kenton
Williamson Pendleton
Brown, OH
David D. Terry, AR Faulkner Clermont, OH
Lonoke Hamilton, OH
Pulaski Barkley, KY Caldwell
Saline Calloway
Christian
Hansen Dam, CA Los Angeles Crittenden
Graves
J. Percy Priest, TNl Cannon Livingston
Cheatham Lyon
Davidson Marshall
Dickson McCracken
Robertson Trigg
Rutherford Dickson, TN
Sumner Houston, TN
Williamson Humphreys, TN
Wilson Montgomery, TN
Stewart, TN
Pine Flat, CA Fresno
Beaver, AR Benton
Saylorville, 1A Boone Carroll
Dallas Madison
Polk Washington
Story Barry, MO
Warren
Whitney, TX Bosque
Sepulveda Dam, CA Los Angeles Hill
Johnson
Shenango, PA Mahoning McLennan
Trumbull Somervell
Lawrence
Mercer Willamette, ORl Lane

 

'Corps of Engineers project included in the 12 lakes study (Propst et al., 1992)

95

96

Appendix A. Continued

 

 

Project Name County Project Name County
Bluestone, WV Bland, VA Bull Shoals, AR Baxter
Giles, VA Boone
Montgomery, VA Marion
Pulaski, VA Ozark, MO
Mercer Taney, MO
Monroe Dewey
Raleigh Hughes
Summers Potter
Radford, VA Stanley
Carroll Sully
Marion Walworth
Barry, MO
Christian, MO Ouachita, AR' Garland
Stone, MO Hot Spring
Taney, MO Montgomery
Englebright, CA Nevada Oahe, SDl Burleigh, ND
Sutter Emmons, ND
Yuba Morton, ND
Campbell
New Hogan, CA Amador Dewey
Calaveras Hughes
Potter
Norfork, AR Baxter Stanley
Fulton Sully
Izard Walworth
Marion ,
Stone Pomme de Terre, MO Cedar
Ozark, MO Dallas
Hickory
Rathbun Lake, IA Appanoose Polk
Davis
Lucas Lake Sharpe, SD Buffalo
Monroe Hughes
Wapello Lyman
Wayne Waco Lake, TX McLennan
Cherry Creek, CO Arapahoe Deer Creek, OH Fayette
Denver Franklin
Douglas Madison
Jefferson Pickaway
Ross
Addicks, TX Fort Bend
Harris

Waller

 

Appendix A. Continued

97

 

 

Project Name County Project Name County
Alum Creek Lake, OH Delaware Blue Marsh, PA Berks
Franklin Lancaster
Knox Lebanon
Licking Schuylkill
Marion
Morrow Bonneville, OR Hood River
Union Multnomah
Skamania, WA
Grapevine Lake, TX Dallas
Denton Chatfield Lake, CO Arapahoe
Tarrant Denver
Douglas
Hartwell Lake, GA Banks Jefferson
Elbert
Franklin Keystone Lake, OK Creek
Habersham Osage
Hart Pawnee
Madison Tulsa
Rabun
Stephens Lewisville Lake, TX Collin
White Cooke
Jackson, NC Dallas
Transylvania, NC Denton
Abbeville, SC Tarrant
Anderson, SC
Greenville, SC Oologah Lake, OK Craig
Oconee, SC Nowata
Pickens, SC Rogers
Tulsa
J Strom Thurmond, SC Columbia, GA Washington
Elbert, GA
Lincoln, GA Cecil M. Harden, IN Clay
McDuffie, GA Montgomery
Richmond, GA Parke
Wilkes, GA Putnam
Abbeville Vermillion
Aiken Vigo
Edgefield
Greenwood Raystown, PAl Bedford
McCormick Blair
Fuhon
Huntingdon

Mifflin

 

Appendix A. Continued

98

 

 

Project Name County Project Name County
Joe Pool Lake, TX Dallas Center Hill, TN Bledsoe
Ellis Cannon
Tarrant De Kalb
Grundy
Lake 0' The Pines, TX Camp Jackson
Cass Putnam
Gregg Smith
Harrison Van Buren
Marion Warren
Morris White
Titus Wilson
Upshur
John H. Kerr, NC Franklin
McNary, ORI Umatilla Granville
Benton, WA Person
Franklin, WA Vance
Warren
Murray, AR Faulkner Brunswick, VA
Pulaski Charlotte, VA
Saline Halifax, VA
Lunenburg, VA
Nolin River, KY Barren Mecklenburg, VA
Butler South Boston, VA
Edmonson
Grayson Kaweah, CA Tulare
Green
Hardin Texoma Lake, TX Bryan, OK
Hart Carter, OK
Larue Johnson, OK
Warren Love, OK
Marshall, OK
Senecaville, OH Belmont Murray, OK
Guernsey Cooke
Monroe Grayson
Muskingum
Noble West Point, AL Chambers
Lee
Shelbyville, ILl Christian Randolph
Coles Carroll, GA
Cumberland Coweta, GA
Douglas Harris, GA
Effingham Heard, GA
Macon Meriwether, GA
Moultrie Troup, GA

Shelby

 

99

Appendix A. Continued

 

 

Project Name County Project Name County
Success, CA Tulare Barren River Lake, KY Allen
Barren
Cumberland, KYl Adair Edmonson
Casey Metcalfe
Clinton Monroe
Cumberland Simpson
Lincoln Warren
McCreary Macon, TN
Metcalfe
Pulaski Belton Lake, TX Bell
Rockcastle Coryell
Russell
Wayne Black Butte, CA Glenn
Pickett, TN Tehama
Dardanelle, AR Conway Cordell Hull, TN Cumberland, KY
Franklin Monroe, KY
Johnson Clay
Logan De Kalb
Perry Jackson
Pope Macon
Yell Overton
Putnam
Eufaula Lake, OK Haskell Smith
Latimer Trousdale
McIntosh Wilson
Muskogee
Okmulgee Degray, AR Clark
Pittsburg Hempstead
Hot Spring
Fort Gibson, OK Osage Montgomery
Washington Pike
Lower Granite, WA Latah, ID
Greers Ferry Lake, AR Clebume Nez Perce, ID
Independence Asotin
Stone Garfield
Van Buren Whitman
White
Harry S. Truman, MO Bates
Mark Twain Lake, MO Audrain Benton
Marion Cedar
Monroe Henry
Ralls Hickory
Randolph Johnson
Shelby Pettis

St. Clair

 

 

100

Appendix A. Continued

 

 

Project Name County Project Name County
Laurel River, KY Knox Carlyle, IL Bond
Laurel Clinton
McCreary Fayette
Pulaski Marion
Whitley Washington
Wappapello, MO Butler Mendocino, CAl Mendocino
Carter
Stoddard Milford, KSI Clay
Wayne Dickinson
Geary
Wright Patrnan, TX Miller, AR Riley
Bowie
Cass Millwood, AR Hempstead
Morris Howard
Little River
Sam Rayburn, TX Angelina Miller
Jasper Sevier
Nacogdoches Bowie, TX
Sabine
San Augustine Somerville, TX Brazos
Shelby Burleson
Lee
Allatoona Lake, GA Bartow Washington
Cherokee
Cobb Summersville, WV Clay
Dawson Fayette
Floyd Nicholas
Forsyth
Fulton Tenkiller, OK Adair
Gordon Cherokee
Paulding Haskell
Pickens Muskogee
Polk Sequoyah
B Everett Jordon, NC Chatham Falls Lake, NC Durham
Durham Franklin
Hamett Granville
Lee Orange
Orange Person
Wake Vance
Wake
Lavon Lake, TX Collin
Dallas
Hunt

Rockwall

 

 

Appendix A. Continued

101

 

 

Project Name County Project Name County
Sidney Lanier, GAl Banks Rend Lake, IL Franklin
Barrow Hamilton
Cherokee Jackson
Dawson Jefferson
De Kalb Marion
Forsyth Perry
Franklin Washington
Gwinnett Williamson
Habersham
Hall W. Kerr Scott, NC Alexander
Jackson Alleghany
Lumpkin Ashe
Pickens Caldwell
Stephens Watauga
Union Wilkes
White
Woodruff, AL Autauga
Monroe, IN Bartholomew Dallas
Brown Lowndes
Greene Montgomery
Jackson
Lawrence Arkabuta, MS De Soto
Martin Tate
Monroe Tunica
Morgan
Owen Blue Mountain, AR Logan
Yell
Dale Hollow, TN Clinton, KY
Cumberland, KY Canton, OK Blaine
Monroe, KY Dewey
Wayne, KY Major
Clay
F entress Lake Umatilla, OR Sherman
Jackson Klickitat, WA
Macon
Overton Lewis and Clark, SD Cedar, NE
Pickett Knox, NE
Putnam Bon Homme
Clay
Dannelly, AL Dallas Yankton
Lowndes
Marengo Nimrod, AR Perry
Wilcox Yell
Dworshak, ID' Clearwater

Lewis

 

 

 

 

 

 

Appendix A. Continued

102

 

 

Project Name County Project Name County
Eastman, CA Madera Philpott Lake, VA Floyd
Mariposa Franklin
Henry
Grenada Lake, MS Calhoun Martinsville
Carroll Patrick
Grenada
Montgomery Rough River Lake, KY Breckinridge
Tallahatchie Grayson
Yalobusha Hancock
Hardin
Hensley, CA Madera Meade
Ohio
Lake Celilo, WA Hood River, OR
Sherman, OR Sardis Lake, MS Lafayette
Wasco, OR Marshall
Klickitat Panola
Tate
Lake Seminole, FL Gadsden Yalobusha
Jackson
Decatur, GA Walter F. George, AL Barbour
Grady, GA Henry
Miller, GA Russell
Seminole, GA Calhoun, GA
Clay, GA
Early, GA
Quitman, GA
Randolph, GA

Stewart, GA

 

 

 

APPENDIX B

 

 

 

 

Am SEED E 56386 was < 3233‘. 89 5&2 0233. 35988 voaumflmmm 2: E onEoE 3:38 we 38.8528
dumb :82 8 3:553 common a 5 83m 33333.5: 38 mo Quad m
38.85 50:95 do 380 Bot $.26 85. Ecumufiw :fizomoboz 58$: 05 8 3:8 98M 4
EPA 5a Eon 0:3 Beam; muoofimcm .8 3.80 39 do 2883 m
3% En 9:8 0:3 28G; muoofimam mo 3.80 38 mo Evoked N
beam 83 We3 3 a3: a a... .283 a some?

 

 

 

 

 

m 3.3 gm 2 .3: 3m 8384 3.3 3.3% $3 832 ohm up .3523

w 4. a 23 m .3: 3m 0%. ES $4.0 RN83 33 $0.08 mom m< 28¢ 29¢

m 3.: 3 om as: .3: 84.82 mm; 838.” 33 3.2:” NE 02 .83 seesaw

a no“ Ex 2 3m {am 08.23 and 23:3 3: 23: SA >3 .ousaam

m m. a 3m o aim 0% 80.50.» :3 83:.” $3 832 com m< 533m

2 EN 3: 8 as $3 898$ 23 Snack 23 832. SN § .323

_ 03 mm o x? 3% 80.33 3% $32k $3 83% N3 20 .3853

5.52..qu— 35 \ 3532333 :5:— awi \ 53.232 33 :3: u: use

a 92 4.3 2 .3: s: oo~.a~._ $3 magmas Es 03.2 3 m2 C. in“: m .353

N we. now o so 3 80.35 23 293mg 2 :m 53%.: m2 <0 .9652 5:23

0 EN 98 m 3% .3 02.83 $3 $3.32 83 5.2% a: o: .33 2:223

4 3A in o .3: 3e 83:; 2% 3.23 33. 83: 32 E .3553

a 9mm cow o so :30 03:3 33 34.320 3.? Samoa «2 <0 .53 25:33

W EN EN o x: am 0:33 23 $0.23. 83 9.3% OZ 5 .2323

_ «.8 is o 3:. .32 83% $3 $583 20.... 23.2 0: <0 a: 25

_ 9mm 98 o .30 .30 08.83 Ea 332$ 3.? 338.0 *3 <6 .53 :35:

4 EN EN 2 .3: .34 08.3; $3 $.33. 22 5.3 EN 3.. .35 d 225

a fix a: 3 so .3: SNNQN a; £32; £3. $.95 03 E .5536

0 SN «.3 o as: so 852:... :2 ages ”of 3.32 NE E .83 83.8

a 2m ”.3 o .3: .33 08.80.” 23 no: 2.. 33 3.3.3 EN :23 Jana 32 a

8.523.... :2: \ 3552353 :2»: $5133.32 83 a9: a new

omoscsoo 20 0:52: mmofim 4 92:5 mzzaom wEaEmo cos—Em; a mufim 3V 33m acofinflzfimm couflsaom 5:9:32 38.3.5 EouEwcm

323.3 223.5: :53. <32 2 ”252$ Nessa .835 moo :33 :23 :93 do 328m 83 do 380
38... 5:52 E85; 8.835 3.52 5.90 Lun— 55E: Z 09?

 

m6388> Eggnog: mo 8m Sam Emmi .m iguana/w

103

 

104

 

 

 

 

 

 

m 23 SN 3. .3: _ .32 £3.82 3.3.x 3.23 ca... 3858 m _ .N :0 .83 3200

m 23 mm“ o .3: 3m 8383 5.” 83.3.3. 23..” 33333 :.~ 0? .83 2:333

3 23 EN : .3: .30 2.3.8} $3 9.32.3 38.3. 38.3% m2 :0 .83.: 25333:

m 3 m SN o .33. .32 832.: an...” $38.3 08.2 :3qu 2: E .83 38 83

: SN 23 mm .32 .33 833.3 5.... 33.833 Sum $3.3m SN 0m .88:ng 585 3

2 33 33 o .33 .33 8038.: :33 3.3383 82.3 3.35 :3. <0 .83 :03:

M 3mm 52 o .32 .3: 852.: 83 833.8 9.3.2 $333.3 2: 0:. .83.: 8:380

m EN :3 2 3m .33 838.: 5.3 338.: 3:3 33.83 83 :0 .8380 .8:

3. 33 :2 o .38 .3: 83.383 :3.” 33.33.: 8% 33.3.; 83 00 .8380 E20

.. Sm :2 o .30 .33 83.20.: :3 33.33.: 85 Emma: 33 O0 .83 22.820

m 92 2m 2 .3: .33 80.33 ma.” 3.33 E... 3.38 a; mo 338.com

.. 3m SN o .32 .35 32.3.; 33. 838.3 338 3.3.2.3 cam <3 .532 SE

3 3.8 SN 0 .33 3m 80.33.: 3.3 3.3.3 33 2333 2 .N :o .83 8:080 :52

m 3m 3m 0 .30 .30 83383 2 2 233.3 25.: SM. _ an 83 0Q .8363

meta—sack

:3: \ 8532333 =5»: :3: < .8332 83 5332 ">— 820

_ 3 new 0 .3: 3 23.82 83 33.83.: 2.: .: 832 03 0? .83 8.3

0 we 9% 8 .32 .32 832.: 23 383:. E m 5.; 92 S .83 .5353

:. _.: 3.3 9. .33 3mm 83%.: 33. 3.33. :3. 8th 32 oz .83 20 8:58

o S: 3% mm .33. 3:. 8983 3.3. 83% 2 m 3.3 NZ 5.. 25052

N 33 EN 2 .3: m .38 $332 om; 3.33 3. 3.3.2.0 mam <0 .535: 302

m N: 3.8 E .32 3m 83.: 3 :3 832 E 03.8 02 am .25.; 83

m 3.8. 23 m .38 .32 83: 22 522.: :2 8?: N mom <0 £332.33

m S: 23 3 .33 .30 83.083 $5 233. 3.3 _ :5: EN m... .285 :3

m 2: EN em .38 .32 838.: cm? 38.2... OR 33.0: NZ .m< .8285

2 EN 38 N .38 .3: 85:5 5.» :38; 33 3%: N3 .3 .83

comma—snea— Fow— \ flsoaaflsaumm :50“ Box— \ hum—Q53: mgaw cum: ”E mun—U

3:550 3V 0885 33m 32:5 wcgom wEnEwU 8:865 3v 83m 6V mofim acacimzfiamm cows—anon 5:332 60.8.5 flooEwam

32qu 22.88: :83. <3 9 8335 8:835 .863 :88. :93 .58: :0 323.8: 83 3o 380
.88. 5:52 E88: 8535 E88; ESE: moo £530 8: 35:52 09:

 

3003:3380 .m £98:me

105

 

 

 

 

 

 

N EN 2a m $0 $2 soamoa $3 83$ 03 82$ ”ON v5 €35 com
o 0.: gm ow $2 $2 08.23 8% 238; E: 5.2: 92 mo .83 «Beam
m 3: 3m mm $2 $3 3.23 8% 3&8 am 53” SN 5. semen
5 I: mom 3 $_ m $m 8&3? 23 8mg; m3 «~33 SN m< "0:055
: m9 2m av $_ _ $§ §.§.m 8% 33.8; 83 mag SN E if .328
N 99 9mm om $3 $8 83: $3 83$ 3m 81: gm <0 .225 fism
N 9% 2m o $2 $v 83ng £3 83:} $2 632 com 5 35 Exam
w 2: 3.». fl $2 $2 83: 2 23 83$; 83 88% Ba 3 “83 55. 55m
v as mom 2 $_~ $2 82% $2 R35 Now 332 2a flmm 65::
fl :8. MSN om $- $2 83% :3 83% 8“ £9; 03 .5 5:688:
2 3; mm on $2 $3 33:; 33 3.33 2 3 832 8a .5. afifigo
zeta—amen
:3 \ afiafizsmm :53. :3 \ Ears: 8% £532 “5 :20
a 58 EN mm $3 $5 Saga 3} we .83 33 Ranch. 22 i £8 33
w ”.2 mom o $¢N $3 818% 83 $38; $2 Rmnmvm v.2 E: "33 «593
_ EN EN o $3 $w 838 83 82me 32 :54 mm mod 5 £88m
m 95 mom 5 $m $o 98.33 $8 $2.53 as} Swim 22 mo u=$8§m
a we mom Om $5 $v 89.83 33 80.82 23 05th SN 5 £58 562
m SN v.2 o $» $N 83$; 33 32:3 23. 833 m2 m< $352
w mom :m o $3 $5 83%; $3. $33; 83 $.va Rd E 3:5 2:10 83
_ 3N SN mm $8 $5 8%: 83 83qu 32 53 mm 8N <0 “€23
: EN SN. 2 $3 $2 3.8% ES 3922 82 $38 2 .N 02 Sum .: 52
: EN EN 8 $w $mm 08.33 5% 83”? $3 $3.2 mom E “a: 385
0 SN 3a a $2 $5 8&2: $2: $352 93 322 SN 7: “SEE .2 =80
w OS <2 mm $8 $3 $323 3 2 9.2 _~.~ as: $.29.” mg as. usieofi
m 3m 3m om $3 $m 08.33 ~50 $.82 83 33% .n fl .N .E “:53me
m EN EN o $2 $~ 33m? $3 8983 $3 $32 SN .mo ,Ezoz
EOE-«Enca— Bcw— \ afloaflmm—Aaumm— =waowm gum—w \ 50:93—52 mow—am 35:32 u> 320
35:30 A8 0885 83m 32:5 wccaom 3:380 8:885 GV 8.5m A8 83m afiEfiESmm cots—30m 8:332 60on floofiwam
3.830% 293283 :83. <22 8 .5536; =o=§£> 803$ Exam =Som Exam mo 3:23”. 83w .«o 880
.38. 5:62 385m 3535 :58?— Eoeom moo SEmU 5m 39:32 up;

 

835300 .m fivaomm/x

106

 

 

 

 

 

 

 

 

.V SN 3a o :9. go conga. $3 892$ 9...". 8.... .m 8.. ,2 H.583

o 3.. mom on $3 $2 85%.. .36 8.3.3 Sm". 33.8 3.. 02 a8... :3. .3

M. EN EN 3 ohm gm. 332.. R3 :38. Sq. gram m... d .83 2.3.

a 0.8 tom 0 $2 $2 83%.. 33 Same? nova 8.8». R... z. .822).

gun—Each Baa \ flauaaflzsmm Exam awi \ 5.32:2 33m Bed "E 820

.. 3m SN 5 :2. x». Smarfi 2% ow~.mom..~ 33”. 833.6. a... .3 £3 :83

5 EN 9% o $2 ox... 8183 33 3mg? $3 9.32. m... oz .83 2...”.

0 3a SN 0 £2 £3 :33. as... $.23 Rod 23% 2... oz .822 5.3,“. m

. . com .8 o x: 8.. 2563... 83 832%.. was Swan. 8.. <0 6.3 £8.22

2 EN EN 2 $2 5 v3.2.3 83. N. 32... 3an 2.3.3.. a... _<0 5.5. 55%

5.3.2.9. an... . 355......22. :53. a»... . 3.3.2:: 8.8 E. ”.5 3.20

v ...N N... o $2 3.. 25.8.... mm... 03.8... to :32 RN 5 ”55.... 2%....

v 3. 33 mm. :8 gm 89$..N mm? 23% 2m 23.. w; 02 6:03.35

m w: 9% mm $2 $8 80.3%. ms... 233 20 $33. Ea v5 5:359

m 3. Odo. mm :2 0.... Sam. 3 Q} 3.33 NS 8me SN >3 2:35.55

v N. . N .8 a. $2 :3 83mm. 23 832.. 3.: M. . 2.: SN 5 u...toeom

o w... a. . m 3 $8 Ax... 8:8... 23.0 2553... 2.2 5.8. a. .N .C 62.3. :am

e com gm mm :8 oi 83$ 3% $38.. 3.... 33,... m. .N ”2 683...:

o EN EN om. x... 82 3mg... 3% 33.3 $0 mm . ,8 RN 0: 9.3 5?; is).

m nmm on 3 :2 § 23.3. $3 323 N... 03mm. 03. <3 6.55 533

m .3. .2». fl 0% x... 83$ 33 833.. 8... 82.... SN 5. .35. .223

w 3.. ...m m $8 0% 823.. SE fivdg 3.... £32 3... o: 552.. .m E...

m 3: .2 R XS go 832w 3w :3: ma mam. m3 m< 08.3 b3. 320

€92.35. 85.2.9. :3 . 3555.33»... :38. E: . 3.32:2 8.5 53.32 .5 3.5

82.500 3. 2:8... 83m $2.8. 9.58m wcfifiuo 8:855 3. 83m 3. 83m flanngnsm. noun—30m 5.32:2 .885 aooswcm

.5833. 29.3.8: :53. $2 9 8:335 8.5.“; 80.8... .53. :33. .53. .o .223. 8.3, .0 3.8
.38: 5622 2.85.. 8.335 .589. 2.8.3 moo 8.9.0 mi .3852 0...:

 

336.80 .m £223....

 

107

 

 

 

a _.E :m o gov £3 mownmmnxy ”um—é wmmflflm owe $.on a; A< ”0986 m 5:85

m 2: v. _ m 3. gem fie ooflsm; 2.3V >368 o: $on 2: m2 “8.5 anew
o N. _ N mam ow $2 £3 oofimwfim mmohm Nvonmvo; 53 $06: $4 >v. «is :32 nmsom

m oém msm um $3 gm 2:32.— Smfi momfii. own «$63 34 <> dung tog—Em
N 3: 3m om xx Ax.” Sagan 89m SS» 2: Eng w: 5. 62.52

m v.2 cam 8 £0: fen 856mm..— wmnw ommdvm Em @3an mm; mm £50 23 £33
m 3%. 0.2 o gmm oi. mogdwonm amflmm Show 2: wwoxfl we; do 5:325 834
0 m9 Ném om o\omm X.” @345.— oomnm ommmmon 2.x mepoE co; 15 6.055% 8:5
v mew Dom om $2 {on Qmfiom; 3&6 mmomoov mmv ommhwm mo; <3 ”2:00 8:5

_ :N mom mm $3. $2 ooafi: $.06 wvmflmov omv mmodm 5: <0 $2803

0 0.2 how. oo °\om~ $0 oomaom 1N ow 1v mmwncmm cow Efimw 0: m2 6&3 25:20
N Mom msm nm :6 $3 comma 3.5” wmmnmvm Rm $3.: 2... <0 “5:35
v ”2 wdm o fomm Rom 2:68; 9R6 cwonmov Rm 30.3 n: 12 5.3ch
: of Wm». o2 $3 fimo coo; E6 cmmw $3.3 So; @3wa 9: E “BB—om 2am
m new 0.8 cm 82 $2 vmndofim omow m5; 2. wow mcmdfl mm; d 6230

m o. K :3 3 $2 $2 335: mvoflm 32.3 #2 Slew mm; MO £9ch

m m? cam ov Ram oi. oofimmm mpg. 33.2 mm: mmnwm mm; N? 58582 0:5

m o. E #3 2 $3 $0 89de am; voodmm cow Mai: 54 m2 433me

N :8 v.3 a: $2 «9: $32 83 31% NS 3&2 m3 2: £555
552:9:— Bo‘. \ flauaaflzaumm 533— Bed \ SEE—.2 83m to; an mus—U

3::on 3V 0885 33m 32:5 wagon 33an 55885 a 83m GV 33m acoEfi__nSmm coax—30m 3:332 80.85 Eoofiwcm
Ecofiom 22.8.8: :83— <mE 8 .8336; 5:865 80.85 :39; 330% 233— mo Ecoaom 83m mo 350

.88. 5:52 885m 8535 2.8ch :58?— mOU «=an 5m .3822 09C.

 

325200 .m iguana?“

 

APPENDIX C

 

 

 

 

 

33m 2: 33 Gm." a; 33 £3 2: .33
so; 2: m8 Sm S RN 8m 2 dam 33 \ mama Ed 2:: 33
81m 83 Sam 23 2: a3 «.8; v dog 33 \ used Em \ 2:2 33
3an $3 $92 :3 SE 83 om; m dam fin: \ fig :3: 2:2 33
8m; 8m 83 $5 mm 3m ”3 mm dog 33 \ 30¢ 39: 2:2 8382
Roam $3 :3 E} c: 9% RE 3 dog 33 \ aaom an: as: :38:
23m $3 ”on: 2% . 35 23 Ed 3 com ”3: \ 3% flat as: 8282
:2 E 08 v2 5 Rm 2: 2 dog 33 \ mag 30,: 2:: $5
83 £3 :3 32 m2 ”3 S3 5 d8 33 \ £3 :3: 2:: E:
2 in 83 $33“ Ea Sum 3&2 own: 2 dam Em \ 3% ES 232 E:
sausaflasfi :59: a. 3:52

NE «3 m3 8.” 8... v2 2.” we .33
mi N? a: 0: mod N; m: 2 .95 33 \ mama 33 \ an: 33
84 8; 2: a: mod :3 o? v d8 33 \ 32% fig: .232 33
a; m: a._ m: :3 So a: m don E: \ Eva awn: 2:2 33
RN 8d mg 23 go 2.0 2 .N 8 dog 33 \ Eva 33 \ 2:2 8382
mg Sm : .m 8N So 2 .o # 3 3 do; 33 \ Exam :3: as: 85.602
8m 2: :a wad 3o mod 2a 2 d8 E: \ E3 if :32 5282
NE 03 wow 03 3o 3 .o 2% 2 do; 33 \ fig 39: as: 3.:
N3 and 3m mom 3o 3o 9%. 5 d8 33 \ Eng :3: as: E:
o: 5m Now V2 3.9 Ed 9% 2 com Em \ msom awn: 232 $5
:82:me 52552 95cm Saab 23an 5304 “ohm :ocmgom :82 Z whims—=2 83m Runaway— E as

amend €22 Sm BEBE .Bm .Em

85880 $3

 

mufiatomoa 05.3wa “newcomonfi .O x65mm<

108

 

 

 

 

109

 

 

32.3 ”8.: :3” $3.2 :2 Sam $23 2: :35
:38 33.2 28$ 8&2 3: .33. 38.8 2 .33 33 \ :93 33 \ 2:2 33
33m ENE 53 83: 9mm; Exm 383mm 2 .33 33 \ :83 22: \ 2:2 33
23.93. 3838 :32 3.3 33 mm}. mm}; m .83 22: \ 2332 22m \ 2:2 33
$.53 832 3%,: $3: MS 33 93.2 mm 2:: 33 \ 293 33 \ 2:2 8:252
828 «8.2 £33 238 m2 822 22.8 2 .33 33 \ 283 2:22 \ 2:2 8:252
332 $.33 82% So“: 83 83: 82.3 2 .38 22: \ 2832 2:2: 2:2 8:252
$32 $92 833 83: 83 93.: 83m 2 .23 33 \ 293 33 \ 2:2 22:
woman $22 :33 $22 an :82 31mm 3 .23 33 \ :33 22:: 2:2 22:
3:2: 333 820mm 22.8 $3 83 £de 2 :3 22m \ 293 22: \ 2:2 22:
3V 2:85 22.09.32 5:32

5.28.: 23.2 812:; $3.22 :32 $3.25 2.3m: 2: .33
$3: :32 83: 82$ 23.: :32 8mm: 2 do: 33 \ 233 33 \ 2:2 33
318m 332m :32 2: 832 32.9. :50 22.2; 2 :3 33 \ :33 22: \ 2:2 33
8332 33% R322 83% 32:8 2 nowm 85:: m :3 22m \ 2832 22: \ 2:2 33
838 8:5 23.32 832 $2: 832 22:22 MN 2:: 33 \ :33 33 \ 2:2 8:232
32.2: :32: 33mm 3828 ”8.2 2223 E33 2 .33 33 \ :33 2:5 \ 2:2 8:252
33.85 83% 22.83 333: 23.2 m 2 as: $333 2 .23 22: \ :93 2:5 \ 2:2 8:252
83:. Show 382 N32: 85: mmmdm SSS 2 :3 33 \ 2:32 33 \ 2:2 :2:
8:22 $3.38 P223 $38 22.8 $2: :32 3 do: 33 \ :93 :2: \ 2:2 2me
£222.: $38 $3.33 :83: 83:2 2 32.: 8:222 Q :3 :2: \ 2532 22m \ 2:2 2:5
825me 82552 980m 5&3 950m 8262 8:5 cocmgoo :82 Z Eta—sac:— .aafiwom

owawm :82 28 BESE .3m .9m
853:8 23

 

3:25:00 .0 59825.4:

110

 

 

 

 

 

«.2: 2.3 :2: m3: 2.: 2.3. 2:: :2 .28.:
3.2: 2:: 3:: 2.2: n: 2.: m2: 2 :o: 33 \ 29:: 33 \ 2:2 33
3:3. :2: 3:: 3.3: n: :2 2:: 2 :o: 33 \ :90: 223 2:2 33
:2: :23 9:: 2.2: :.: 2.: :33. m .::: 2:2: \ :33 222: 2:2 33
:2: mm: 2: :2: n: n: :.:m m: .:o: 33 \ :23: 33 \ 2:2 8:232
:3 2w: 3.:: :3: 2.: 3.: :2: 2: :o: 33 \ :23: 2:2: \ 2:2 8:252
2:: 2:: 3:: 3.3.3. :.: 2.3. :2: 2: .8: 2:2: \ :23: 222: \ 2:2 8:232
2:: 2.3: :23. 3.3: 3.: mm 2.3: :2 :o: 33 \ :93: 33 \ 2:2 22::
:23. :2: 2:3. 3:: :.: 2.: :2: 3 .8: 33 \ :28: 22:: \ 2:2 2:2:
2:: 3:: 3.33 3.3.3. :.: 2.: mm: 2 .8: 22:: \ :23: 22:: 2:2 :2:
83m .3qu Begum

3m: : mm :3 2 mm :3 2:: 3:3
:2: : N: 3: w :3 22 :2 :o: 33 \ :23: 33 \ 2:2 33
:2 : :2 2- :_ :3 3: 2 :o: 33 \ 2%: 22m \ 2:2 33
m2 : mm 2:- : :m :2 m :o: 222:: \ :93: 2mm: \ 2:2 33
mm: : 3: :3 3 2m 3.: m: .::: 33 \ :83: 33 \ 2:2 8:232
:: : :m 3 : .3 a 2_ .::: 33 \ 223: 2:2: \ 2:2 8:232
33 : 3: _ 2 2: : 2: :o: 22: \ :23: 22: \ 2:2 8:252
32 : :3 m 2 :2 :m :_ :3 33 \ :23: 33 \ 2:2 22:
8 : :3 : w a :m 3 .8: 33 \ 22:3: 22:2: \ 2:2 22::
:3 : :_ _- m : m N: :0: 2:2: \ :23: 22:: \ 2:2 2:2:
825me 82552 cgom 8&3 250m $264 :85 53350 :32 Z Ame—=5 <92 39:82 8 3::ng

owfim :32 :8 3335 .9m .9m

8:22:23 :3

8:25:00 .0 5953523

111

 

 

 

 

 

:::.3::.2: ::3: :::.2::.: ::2.:::.: 3::.:::.: :::.::3.:: ::3.3m3.: :3 :33
:2:.:2:.: :::.2: ::3.::: :::n::: 3:3: :::.::: :::.::2 :: :o: 33 \ :53: 33 \ ::3): 33
:::.:::a: :::.:::.: ::3.:3::: ::2h:2:.: :::.::: :::.:: m :::.:::a: 2 .8: 33 \ :33: ::::: ::3: 33
:::n:::,:: :::.:::.3 :: :.:::.:: :33h:2:.2 :::.:::a: :::.:::,: :2:.3::“:: : :0: :::: \ :53: ::E: ::3: 33
:::.::2“: :::.::2 2::.3::a: :::.:23 :2:.3: :3:.::: :::.::: :: :o: 33 \ :93: 33 \ ::3: 82:2):
:32.:: :.2 :::.:3:.: 232:2: ::::3:.: 22:.::: 2::“::3 :::.:::.: 2: :o: 33 \ :33: :3: \ ::3: 8232,:
::::2:3o:: :::::::.: :2:.::2.:: :::a::::: ::3.::3.: :::.::2a:: :::,:::.2: 2: 9.: :::: \ :33: ::::: ::3: 53:02
::3.:::,: :::.::: 32:.2::n: 3::.::2 2::: :: :::n2:2 :::.3:3 :: .8: 33 : :50: 33 \ ::3): ::::
:::.:::.: :3:::::.: :3:.::3.: :33,:3:.: :::.::: ::3,::2 2::,:::.: 3 :o: 33 : :30: :3: \ ::3): E:
:::.3::.2: :::.::3.: :::n:::fl:2 :::.:::a: ::2.::2fi: :22n2::::: ::::32:h:: :: .::: :::: : :53: ::::\ ::3: ::::
:2:: ::3:

3m:.:: :::.: 2::: m::.: :m: :33 ::3. :2 :89:
22:: :::.: :::.: 2::.2 ::: ::: 3::.2 :: .::: 33 \ :83: 33 : ::3): 33
:23.3 :2:: 2::.3 23:.: ::: ::3 :23: 2 .8: 33 \ :93: ::E: ::3): 33
:3:.: :2:.3 :2:: 2::.3 :2: 3:: :::.: : :o: :::: : :93: :::: \ ::3: 33
3::.3 :2:.2 ::2.: 2:“: ::: ::: 8:: :: :o: 33 \ :53: 33 : ::3: 828::
3::.:: 2:3.2 :::“3 :::.: ::: ::2: ::3“: 2: do: 33 \ =33: ::::\ ::3): 52:52
3::.: ::3: 2:: :::.3 ::: 2:: ::2: 2: 3: :::: \ =93: ::::\ ::3: 3:82,:
::3: :::.2 ::3 3:3.: ::2 2:2“: 323.: :: :o: 33 \ ::3: 33 \ ::3: :3:
::3: ::2.: 8:: :::.: ::2 :: :a: :32:3 3 .8: 33 \ :93: ::::\ ::3: :5
::3: 3::a: ::n: :2::3 2:: :8: :::.3 :: .3: :3: \ :33: :::: \ ::3: ::::
8::Ex:2 82552 venom 8&3 950m 8264 :25 53:30:: :82 Z SEaU ham 83m :83—

owqam 5:22 :8 BEBE .3m .9m

8:02.28 ::3

8:53:00 .0 itching:

112

 

 

 

 

 

8.: 8.: 3.: :3... 8.: 2.: 33.: 9: :93
32 8.: :2 2.: 8.: 2.: m3: 2 .8: 33 \ :93 33 \ :32 33
as 2.: :2 2.: 8.: 2.: :3: v .8: 33 \ :93 :3: \ :3: 33
2.: 3:: an: 2 .o 3.: 2.: 33.: m .8: :3: \ :93 :3: \ :22 33
$8 8.: 33.: :8 8.: 3 .o 33.: m3 .8: 33 \ :93 33 \ 3:2 92:02
33.: 8.: 33.: 2.0 8.: 8.: 2 .o E .8: 33 \ :93 :3: \ :32 3:82
M3,: 8.: 33.: 8.: 8.: 2.: 2 .o E .8: :3: \ :93 :3: \ :22 3:82
:3 2.: :8 :3: 3o 2.: :2 S .8: 33 \ :93 33 \ :32 :3:
3.: 8.: 33.: we: 8.: 3.0 :3: 3 .8: 33 \ :93 :3: \ 3:2 :3:
:3 8.: 33.: 8.: 8.0 2 .o m _ .o 3: .8: :3: \ :93 :3: \ :32 :3:
wig—3m “59.3:

833.: 8.3: 33.33 3333.: 33.:3 38323 32.2.33 3: .93
83.85 8%: 8:383 38.38.: 83 : m 38.33.: 33333 2 .8: 33 \ :93 33 \ :32 33
833%.: $3.33: 83.5.: 33.38.: 0:32 93.83 3:33.: 3 .8: 33 \ :93 :3: \ :32 33
333%.: 38.33.: 3.80.: 83338 2:38 23.233 8838.: m .8: :3: \ :93 :3: \ :32 33
36.233 83.9: 83.833 333.93; 383mm 89.25 33383 33 .8: 33 \ :93 33 \ :32 9:82
833%.: 8:3? 33.333 3:332 539. 33125 E9333 3 .8: 33 \ :93 :3: \ 3:2 8:82
8:33.: 83.82 838.: $3: 3833 2333 32.33,: E .8: :3: \ :93 :3: \ :32 9282
83833 8332 £3.33 €0.23. :33; m5: 2 83.32 S .8: 33 \ :93 33 \ :32 :3:
8353 8:83.: 33:3 3833.: 33:33: 32.: m3 933: 3 3 .8: 33 \ :93 :3: \ :32 :3:
83583 8333 33.83 83 8 33.3% 33.233 8383 3: .8: :3: \ :93 :3: \ :32 :3:
525an EEEEE 950m 5ng 959m $304 .535 noggin :82 2 55565 HOD :2:—.33

092$ :82 :8 EZBE .Em .Ew

853:3 .38

325qu .U 36:25::

113

 

 

8.0 00.0 2.0 00.0 8.0 3.0 :0 we :85.
8.0 No.0 2.0 m0.0 m0.0 2.0 9.0 0: do: Box: \ =83: 261: :62 301:
2.0 00.0 0N0 00.0 8.0 m0.0 2.0 :3 do: 304 \ :83: :3: \ 2:2 >51:
w0.0 8.0 00.0 :00 :00 m0.0 m0.0 m .003 :wi \ :83: :3: \ ::2 >51:
03.0 00.0 2.0 w0.0 No.0 2.0 M: .0 mm do: 30‘: \ 630% 30x: \ 3:2 E:602
mm.0 00.0 3.0 m0.0 30.0 M: .0 2.0 :3: do: >64 \ =83“ awn: \ 2:2 E:602
3.0 00.0 :0 No.0 No.0 80.0 50.0 3 do: awn: \ :93: :mi \ :62 8:602
0N0 No.0 0N0 00.0 M00 00.0 M: .0 0: do: 261: \ :93: Bow: \ 2:2 :33
000 No.0 wmd :00- 00.0 3.0 E .0 n do: Box: \ 2833 :wE \ 2:2 :36:
:38 88 3: .o 8.8 8.: 88 8.: 3: 8.: :3: \ :93 :3: : 33>: :3:
8:35:82 52552 650m 5:03 6::om :0304 :85 :033300 :302 Z 33:50 2:00:09
03:53 :302 :0: 630:: Em 6%

88:38: :8

60:88:00 .0 36:03:33

APPENDIX D

Appendix D. Analysis of Variance Between Classes Using Dunnett’s T3 Multiple
Comparison Method.1

 

Independent Classes
Variable Class I II III IV V VI VII VIII IX

Type 111 Regional Sales Multipliers
I

 

 

 

II -
III -
IV Hut: and: *** _
V *** “a: ah“! _
VI *u *** *** _
VII *4“: *** *** “no: “no: *** _
VIII *** *** *** * * =1- _
IX *** *** *u “a: an“ an”: _
Independent Classes
Variable Class I II III IV V VI VII VIII IX
Regional Population
1 _
ll -
lll *** -
lV * u _
V *** an: -
VI H * “a: _
VII -
VIll * -
IX *** ** *** * * _

 

|Dunnett’s T3 Multiple Comparison method is a Post hoc procedure assuming unequal class variances.
* Mean difference is significant at the 0.05 level

** Mean difference is significant at the 0.01 level

*** Mean difference is significant at the 0.00] level

114

Appendix D. Continued

 

Independent
Variable Class

I

11

Number of Retail Establishments

I

II
III
IV
V
VI
VII
VIII

III

Classes
V VI

VII

VIII

IX

 

 

Independent
Variable Class

Per Capita Retail Sales

I

II
III
IV
V
VI
VII
VIII

1

[11

Classes
V VI

VII

VIII

IX

 

116

Appendix D. Continued

 

Independent Classes
Variable Class I II III IV V VI VII VIH IX

 

Annual Corps of Engineers Project Visitation
I

II -
III -
IV _
V _
VI -
VII _
VIII _

 

 

Independent Classes
Variable Class I II III IV V VI VII VIII IX

 

Percent Campers

VIII -

 

Appendix D. Continued

 

 

 

 

Independent Classes
Variable Class I II III V VI VII VIII IX
Percent Boaters
1 _
II -
III -
IV
V _
VI -
VII -
VIII -
IX -
Independent Classes
Variable Class I II III VI VII VIII IX

 

Distance to the Nearest MSA
1 _

II -
III -

VIII
IX **

118

Appendix D. Continued

 

Independent Classes
Variable Class I II III IV V VI VII VIII IX
Percent Retail Sales
1 _

 

 

II -

III -

IV -
V * _

VI * 4: * :0: t _

VII -
VIII * -
IX * * * _
Independent Classes
Variable Class I II III IV V VI VII VIII IX

 

Median Household Income
1 _

11 H _
Ill ** -
IV an: ** _
V * ** _
VI “or “on: _
VII * * -
VIII -
1X *Ihk *** xx _

 

APPENDIX E

Appendix E. Qualitative Descriptions of Initial Classification Categories

Class I: High Multiplier / High Retail / High Population (N = 12)

Each project in this class is within 30 miles of a Metropolitan Statistical Area (MSA), and
has a large number of retail establishments within the region, as well as high retail sales.
Most of the projects in this class have a reportedly low level of camping.

The Corps of Engineers project from the Propst et a1. (1992) study that was classified into
this category was J. Percy Priest. J. Percy Priest is located just outside Nashville, TN in a
relatively rural area. Of the 8.9 million visitors in 1994, approximately 20 percent were
estimated to have camped within the project’s boundaries, while 14 percent of the visitors
engaged in some form of boating while at the project (NRMS, 1994). A ten county
region was aggregated for J. Percy Priest. The region reported a total of $32 billion in
sales to final consumers in 1990 (final demand) (Propst et al., 1992). The percentage of
those sales to COE visitors by individual recreation related sectors were: Hotel (8.3%),
Recreation (3.3%), Eating and Drinking Establishments (2.8%), and Retail and Wholesale
(1.8%) sectors. Nonresident visitors accounted for 20 percent of all visitors to J. Percy
Priest and 38 percent of visitor spending in the local area. Average nonresident visitor
spending was $124.68 within 30 miles, while average resident spending was $51.61.
Average spending for all COE visitors within 30 miles was $66.37, and $88.48 per party
tr1p.

Class 11: High Multiplier / High Retail / Low Population (N = 7)

Corps of Engineers projects in this class generally were further away from MSAs,
although many were still within 30 miles. Apart from Beaver, AR, each of the projects in
this class are located in relatively rural areas in the vicinity of other popular recreation
destination areas. For instance, the Willamette, OR projects are located in the vicinity of
the Pacific Ocean and many other recreation areas. Additionally, Barkley, KY is located
adjacent to the Tennessee Valley Authority’s popular Land Between the Lakes Recreation
Area, and Stockton Lake, MO and Table Rock, AR are located adjacent to the Ozarks.

The Corps of Engineers project from the Propst et al. (1992) study that was classified into
this category was Willamette. Willamette is really an aggregation of three small projects
(Cottage Grove, Fall Creek, and Fern Ridge) surrounding Eugene in Lane County, OR.
Approximately 1.6 million visitors came to these three projects in 1994. Of those
visitors, approximately 34 percent were estimated to have camped within the project’s
boundaries, while 45 percent of the visitors engaged in some form of boating (NRMS,
1994). In Propst et al. (1992) study of the Willamette project, Lane County reported a
total of $7.2 billion in sales to final consumers in 1990 (final demand). The percentage of
those sales to COE visitors by individual recreation related sectors were: Hotel (5.0%),

119

120

Recreation (2.1%), Eating and Drinking Establishments (1.5%), and Retail and Wholesale
(1.3%) sectors. Nonresident visitors accounted for 16 percent of all visitors to Willamette
and 30 percent of visitor spending in the local area. Average nonresident visitor spending
was $101.48 within 30 miles, while average resident spending was $44.53. Average
spending for all COE visitors within 30 miles was $53.64, and $72.54 per party trip.

Class III: High Multiplier / Low Retail / Low Population (N = 10)

Corps of Engineers projects in this class are located in relatively rural areas, except for
Waco, Englebright, and New Hogan. Waco is a primary provider of recreation
opportunities, particularly day-use, for visitors living in nearby Waco, TX. Englebright
and New Hogan are both small reservoirs located outside Yuba City, CA and Medesto,
CA respectively. Neither Englebright nor New Hogan had a comparatively equal amount
of recreation facilities to other projects in this class. As such, neither project is a
significant provider of recreation opportunities within their respective regions.
Additionally, many of the projects within this class are located in popular recreation
areas, such as the Ozark Mountains in Missouri and Arkansas, and Sierra Nevada
Mountains in California.

The Corps of Engineers projects from the Propst et al. (1992) study that were classified
into this category were Lakes Oahe, SD and Ouachita, AR. Lake Oahe is a large river
project spanning 231 miles from Bismarck, the capital of North Dakota, to Pierre, the
capital of South Dakota along the Missouri River. Of the 1.7 million visitors in 1994,
approximately 13 percent were estimated to have camped within the project’s boundaries,
while 30 percent of the visitors engaged in some form of boating while at the project

(N RMS, 1994). A thirteen county region was aggregated for Lake Oahe. The region
reported a total of $3.8 billion in sales to final consumers in 1990 (final demand) (Propst
et al., 1992). The percentage of those sales to COE visitors by individual recreation
related sectors were: Hotel (30.1%), Eating and Drinking Establishments (9.4%),
Recreation (6.5%), and Retail and Wholesale Trade (3.7%) sectors. Nonresident visitors
accounted for 44 percent of all visitors to Lake Oahe and 71 percent of visitor spending in
the local area. Average nonresident visitor spending was $125.54 within 30 miles, while
average resident spending was $40.02. Average spending for all COE visitors within 30
miles was $77.82, and $131.09 per party trip.

Lake Ouachita is located 25 miles from Little Rock, AR on the Ouachita and Black
Rivers in southern Arkansas and northern Louisiana. Of the 1.2 million visitors in 1994,
approximately 18 percent were estimated to have camped within the project’s boundaries,
while 26 percent of the visitors engaged in some form of boating while at the project

(N RMS, 1994). A five county region was aggregated for Lake Ouachita. The region
reported a total of $3.3 billion in sales to final consumers in 1990 (final demand) (Propst
etal., 1992). The percentage of those sales to COE visitors by individual recreation
related sectors were: Hotel (19.2%), Eating and Drinking Establishments (9.9%), Retail

121

and Wholesale Trade (5.3%), and Recreation (4.4%) sectors. Nonresident visitors
accounted for 65 percent of all visitors to Lake Ouachita and 74 percent of visitor
spending in the local area. Average nonresident visitor spending was $129.91 within 30
miles, while average resident spending was $86.48. Average spending for all COE
visitors within 30 miles was $114.71, and $177.14 per party trip.

Class IV: Medium Multiplier / High Retail / High Population (N = 14)

Except for Bonneville, OR and Oologah Lake, OK, Corps of Engineers projects classified
into this category generally were within 30 miles of an MSA. Additionally, 9 of the 14
projects in this class were located immediately adjacent to an MSA. As such there were a
significant number of places to purchase recreation related goods and services.
Additionally, adjacent to or within each of these project’s boundaries were a number of
resort style recreation areas providing golf courses, commercial and private lodging, and
an assortment of urban type recreation facilities. Other than J. Strom Thurmond, project
visitation was predominately day-use. Approximately 25 percent of 1994 visitors
camped while at J. Strom Thurmond.

There were no Corps of Engineers projects from the Propst et al. (1992) study classified
into this category. Thus, primary visitor spending data was not available to develop
visitor spending profiles for this class. However, crude estimates of spending profiles can
still be estimated for COE projects in this class if levels of retail establishments and
regional population are known. For instance, the spending profiles of an adjacent class
may be examined for projects in this class if the levels of retail establishments and
regional population are somewhat similar (i.e. projects in this class having numbers of
retail establishments and regional population close to the break points). However,
caution should be taken when using this method as estimated spending profiles may not
reflect visitors in this class. Additionally, spending profiles derived from adjacent classes
for projects in this class should error on the side of conservative.

Class V: Medium Multiplier / High Retail / Low Population (N = 14)

Corps of Engineers projects classified into this class were in a variety of locations,
ranging from remote to more urbanized areas. A majority of the projects in this class had
relatively few campers during the 1994 season, except for Center Hill, TN. More than 50
percent of COE Visitors were estimated to have camped or stayed in commercial lodges
while at Center Hill.

The Corps of Engineers projects from the Propst et a1. (1992) study that were classified
into this category were McNary Lock and Dam, OR, Raystown, PA and Lake
Shelbyville, KY. McNary Lock and Dam, located in the Columbia River George

122

National Scenic Area adjacent to Richland, OR and Kennewick, OR, backs up 64 miles
of water to create Lake Wallula. Of the 4.9 million visitors in 1994, approximately 2
percent were estimated to have camped within the project’s boundaries, while 12 percent
of the visitors engaged in some form of boating while at the project (NRMS, 1994).
Many of the project’s visitors come to view the large hydroelectric plant in operation. A
four county region was aggregated for McNary. The region reported a total of $8.4
billion in sales to final consumers in 1990 (final demand) (Propst et al., 1992). The
percentage of those sales to COE visitors by individual recreation related sectors were:
Hotel (46.7%), Eating and Drinking Establishments (8.7%), Recreation (6.9%), and
Retail and Wholesale Trade (3.2%) sectors. Nonresident visitors accounted for 38
percent of all visitors to McNary and 64 percent of visitor spending in the local area.
Average nonresident visitor spending was $72.82 within 30 miles, while average resident
spending was $25.31. Average spending for all COE visitors within 30 miles was
$43.51, and $71.38 per party trip.

Raystown is located 30 miles from Altoona, PN. The project has thirteen public use
facilities, a private recreation complex with cruise vessels, and overnight
accommodations. Of the 980,000 visitors in 1994, approximately 9 percent were
estimated to have camped within the project’s boundaries, while 28 percent of the visitors
engaged in some form of boating while at the project (NRMS, 1994). An eight county
region was aggregated for Raystown. The region reported a total of $13 billion in sales to
final consumers in 1990 (final demand) (Propst et al., 1992). The percentage of those
sales to COE visitors by individual recreation related sectors were: Recreation (2.2%),
Hotel (1.4%), and Eating and Drinking Establishments (0.7%) sectors. Nonresident
visitors accounted for 63 percent of all visitors to Raystown and 75 percent of visitor
spending in the local area. Average nonresident visitor spending was $53.01 within 30
miles, while average resident spending was $30.87. Average spending for all COE
visitors within 30 miles was $44.86, and $80.39 per party trip.

Lake Shelbyville is located 22 miles from Decatur, IL. Of the 2 million visitors in 1994,
approximately 14 percent were estimated to have camped within the project’s boundaries,
while 20 percent of the visitors engaged in some form of boating while at the project

(N RMS, 1994). A ten county region was aggregated for Lake Shelbyville. The region
reported a total of $12 billion in sales to final consumers in 1990 (final demand) (Propst
et al., 1992). The percentage of those sales to COE visitors by individual recreation
related sectors were: (10.8%), Eating and Drinking Establishments (2.8%), Recreation
(2.5%), and Retail and Wholesale Trade (1.1%) sectors. Nonresident visitors accounted
for 44 percent of all visitors to Lake Shelbyville and 71 percent of visitor spending in the
local area. Average nonresident visitor spending was $85.86 within 30 miles, while
average resident spending was $27.63. Average spending for all COE visitors within 30
miles was $53.31, and $87.67 per party trip.

123

Class VI: Medium Multiplier / Low Retail / Low Population (N = 23)

Except for Belton Lake, TX and Wright Patman, TX, Corps of Engineers projects
classified into this category generally were in rural areas lacking an MSA within 30 miles
of their boundaries. As such, there were fewer retail establishments and less per capita
retail sales in comparison to the previous classes. Additionally, the percentage of total
sales within the projects’ regions attributed to retail trade was generally higher within this
class. Many of the projects were predominately day-use. However, several projects
received more than a fifth of their visitation as campers.

The Corps of Engineers projects from the Propst et al. (1992) study that were classified
into this category were Lake Cumberland, KY, Mendocino, CA and Milford Lake, KS.
Lake Cumberland is located 50 miles south of Lexington, KY adjacent to the Daniel
Boone National Forest. Of the 7.4 million visitors in 1994, approximately 50 percent
were estimated to have camped within the project’s boundaries, while 15 percent of the
visitors engaged in some form of boating while at the project (NRMS, 1994). The project
is a popular area for many of the more than seven million visitors to stay overnight on
houseboats, and engage in related recreation activities associated with houseboats, such
as fishing and swimming. An eighteen county region was aggregated for Lake
Cumberland. The region reported a total of $6.3 billion in sales to final consumers in
1990 (final demand) (Propst et al., 1992). The percentage of those sales to COE visitors
by individual recreation related sectors were: Hotel (256.3%)1, Recreation (41.7%),
Eating and Drinking Establishments (35.8%), and Retail and Wholesale Trade (15.8%)
sectors. Nonresident visitors accounted for 59 percent of all visitors to Lake Cumberland,
and 78 percent of visitor spending in the local area. Average nonresident visitor spending
was $172.91 within 30 miles, while average resident spending was $70.47. Average
spending for all COE visitors within 30 miles was $131.21, and $197.91 per party trip.

Mendocino is located on the Russian River 30 miles Santa Rosa, CA. The project has
four campgrounds with limited recreational facilities. Of the 500,000 visitors in 1994,
approximately 17 percent were estimated to have camped within the project’s boundaries,
while 22 percent of the visitors engaged in some form of boating while at the project
(NRMS, 1994). In Propst et a1. (1992) study of Mendocino, Mendocino county reported
$2 billion in sales to final consumers in 1990 (final demand). The percentage of those
sales to COE visitors by individual recreation related sectors were: Hotel (7.3%),
Recreation (4.4%), Eating and Drinking Establishments (3.4%), and the Retail and
Wholesale (1.7%) sectors. Nonresident Visitors accounted for 41 percent of all visitors to
Mendocino, and 30 percent of visitor spending in the local area. Average nonresident
visitor spending was $62.86 within 30 miles, while average resident spending was

 

' Estimates of hotel spending exceeded 100% of 1990 sales in the hotel sector. This is likely due to
allocation of houseboat rentals to the hotel sector. In fact, the houseboat rental business is so extensive
that Lake Cumberland can be considered an outlier in terms of visitor spending and economic effects.
Thus, it would be more appropriate to make comparisons to Lakes Milford and Mendocino in Class VI.

124

$101.54. Average spending for all COE visitors within 30 miles was $85.84, and
$126.20 per party trip.

Milford Lake is located 70 miles west of Topeka, KS off Interstate 70, adjacent to Fort
Riley Military Base. The project has four marinas and nine developed recreation areas
providing visitors with several places to purchase recreation related goods and services.
Of the 570,000 visitors in 1994, approximately 13 percent were estimated to have camped
within the project’s boundaries, while 21 percent of the visitors engaged in some form of
boating while at the project (NRMS, 1994). A five county region was aggregated for
Milford. The region reported a total of $2.9 billion in sales to final consumers in 1990
(final demand) (Propst et al., 1992). The percentage of those sales to COE visitors by
individual recreation related sectors were: Hotel (5.9%), Eating and Drinking
Establishments (1.9%), Recreation (1.6%), and the Retail and Wholesale (1.6%) sectors.
Nonresident visitors accounted for 41 percent of all visitors to Milford Lake, and 51
percent of visitor spending in the local area. Average nonresident visitor spending was
$102.05 within 30 miles, while average resident spending was $67.97. Average spending
for all COE visitors within 30 miles was $81.87, and $128.49 per party trip.

Class VII: Low Multiplier / High Retail / High Population (N = 5)

The five Corps of Engineers projects assigned to this class were predominately day-use
facilities. Except for Lake Sidney Lanier, the projects in this class were located adjacent
to MSAs, and had comparatively high per capita retail sales. Given the fact that these
projects are located adjacent to MSA, it is surprising that sales multipliers are low.
Additionally, the projects in this class had relatively high amounts of boating activity in
relation to their numbers of visitors, with the exception of B. Everett Jordon. B. Everett
Jordon was estimated to have had only 12 percent of its visitors participate in boating
activities while the other projects were estimated to have from 23 to 35 percent of their
visitors participate in boating.

The Corps of Engineers project from the Propst et al. (1992) study that was classified into
this class was Lake Sidney Lanier. The project is located 45 miles from Metropolitan
Atlanta, GA, were a majority of the visitors reside. More than sixty public recreation
areas are located on Lake Sidney Lanier, including a golf course, resorts, commercial
lodging and other places for visitors to purchase recreation related goods and services.
The region surrounding Lake Sidney Lanier is an economically large and diverse area,
compared to other COE projects. However, a majority of the population and economic
activity lie just outside 30 miles which partially explains the low multipliers. Of the 6.7
million visitors in 1994, approximately 8 percent were estimated to have camped within
the project’s boundaries, while 32 percent of the visitors engaged in some form of boating
while at the project (N RMS, 1994). A twenty county region was aggregated for Lake
Sidney Lanier. The region reported a total of $41 billion in sales to final consumers in
1990 (final demand) (Propst et al., 1992). The percentage of those sales to COE visitors

125

by individual recreation related sectors were: Hotel (2.6%), Recreation (2.5%), Eating
and Drinking Establishments (1.5%), and Retail and Wholesale (0.7%) sectors.
Nonresident visitors accounted for 23 percent of all visitors to Lake Sidney Lanier, and
29 percent of visitor spending in the local area. Average nonresident visitor spending was
$72.58 within 30 miles, while average resident spending was $51.64. Average spending
for all COE visitors within 30 miles was $56.39, and $71.49 per party trip.

Class VIII: Low Multiplier / High Retail / Low Population (N = 4)

The four Corps of Engineers projects classified into this category were located in sparsely
populated regions compared to many of the other projects. Even Woodruff, AL and
Monroe, IN do not exceed 400,000 people within the region and they were located just
outside Montgomery, AL and Bloomington, IN respectively. Additionally, the number of
retail establishments and per capita retail sales were lower for this class than for Class
VII.

There were no Corps of Engineers projects from the Propst et al. (1992) study classified
into this category. Thus, primary visitor spending data was not available to develop
visitor spending profiles for this class. However, crude estimates of spending profiles can
still be estimated for COE projects in this class if levels of retail establishments and
regional population are known. For instance, the spending profiles of an adjacent class
may be examined for projects in this class if the levels of retail establishments and
regional population are somewhat similar (i.e. projects in this class having numbers of
retail establishments and regional population close to the break points). However,
caution should be taken when using this method as estimated spending profiles may not
reflect visitors in this class. Additionally, spending profiles derived from adjacent classes
for projects in this class should error on the side of conservative.

Class IX: Low Multiplier / Low Retail / Low Population (N = 19)

Most of the projects classified into this category were located in rural areas with
comparatively sparsely populated economic regions. As such, there were a limited
number of places to purchase recreation related goods and services. Dale Hollow, TN
was the only project in this class with more than 1,000 retail establishments (1,001). Per
capita retail sales were lower for the regions in this class than in others. A majority of the
projects in this class received little camping during the 1994 season. However, as high as
63 and 25 percent of the visitors in 1994 were estimated to have camped at Dale Hollow
and Lake Eastman, CA, respectively. All the projects in this class received between 11
and 52 percent of their recreational use in the form of boating, except for Lake Eastman
which had none.

126

The Corps of Engineers project from the Propst et al. (1992) study that was classified into
this category was Dworshak Reservoir. The project is located in a largely rural area
(1994 population of 12,300) 145 miles from Spokane, WA. Limited recreation facilities
were located within the Dworshak region, as well as places to purchase recreation related
goods and services. Of the 109,000 visitors in 1994, approximately 11 percent were
estimated to have camped within the project’s boundaries, while 36 percent of the visitors
engaged in some form of boating while at the project (NRMS, 1994). A four county
region was aggregated for Dworshak. The region reported a total of $2.1 billion in sales
to final consumers in 1990 (final demand) (Propst et al., 1992). The percentage of those
sales to COE visitors by individual recreation related sectors were: (1.2%), Recreation
(0.6%), Eating and Drinking Establishments (0.6%), and Retail and Wholesale (0.3%)
sectors. Nonresident visitors accounted for 53 percent of all visitors to Dworshak, and 43
percent of visitor spending in the local area. Average nonresident visitor spending was
$38.44 within 30 miles, while average resident spending was $55.90. Average spending
for all COE visitors within 30 miles was $46.71, and $81.92 per party trip.

MICHIGAN SRTE UNIV. LIBRARIES
11111111111|1111111111111111111111
31293015634748